From 82524746c27fa418c250a56dd7606b9d3fc79826 Mon Sep 17 00:00:00 2001 From: Franck Bui-Huu Date: Mon, 12 May 2008 21:21:05 +0200 Subject: rcu: split list.h and move rcu-protected lists into rculist.h Move rcu-protected lists from list.h into a new header file rculist.h. This is done because list are a very used primitive structure all over the kernel and it's currently impossible to include other header files in this list.h without creating some circular dependencies. For example, list.h implements rcu-protected list and uses rcu_dereference() without including rcupdate.h. It actually compiles because users of rcu_dereference() are macros. Others RCU functions could be used too but aren't probably because of this. Therefore this patch creates rculist.h which includes rcupdates without to many changes/troubles. Signed-off-by: Franck Bui-Huu Acked-by: Paul E. McKenney Acked-by: Josh Triplett Signed-off-by: Andrew Morton Signed-off-by: Ingo Molnar --- drivers/infiniband/hw/ipath/ipath_verbs.c | 1 + drivers/infiniband/hw/ipath/ipath_verbs_mcast.c | 3 +-- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c index e0ec540042b..5d830d87ebc 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs.c @@ -35,6 +35,7 @@ #include #include #include +#include #include "ipath_kernel.h" #include "ipath_verbs.h" diff --git a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c index 9e5abf9c309..d73e3223287 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c @@ -31,8 +31,7 @@ * SOFTWARE. */ -#include -#include +#include #include "ipath_verbs.h" -- cgit v1.2.3 From d21c95c569c462da20d491b75d0a45bd70ddc1bf Mon Sep 17 00:00:00 2001 From: Jonathan Corbet Date: Fri, 16 May 2008 13:40:30 -0600 Subject: Add "no BKL needed" comments to several drivers This documents the fact that somebody looked at the relevant open() functions and concluded that, due to their trivial nature, no locking was needed. Signed-off-by: Jonathan Corbet --- drivers/infiniband/core/ucm.c | 1 + drivers/infiniband/hw/ipath/ipath_file_ops.c | 1 + 2 files changed, 2 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index d7a6881b571..3e6a8ff6d76 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -1155,6 +1155,7 @@ static unsigned int ib_ucm_poll(struct file *filp, return mask; } +/* No BKL needed here: no global resources used */ static int ib_ucm_open(struct inode *inode, struct file *filp) { struct ib_ucm_file *file; diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c index b472b15637f..e80cfbd4f3f 100644 --- a/drivers/infiniband/hw/ipath/ipath_file_ops.c +++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c @@ -1812,6 +1812,7 @@ done: return ret; } +/* No BKL needed here */ static int ipath_open(struct inode *in, struct file *fp) { /* The real work is performed later in ipath_assign_port() */ -- cgit v1.2.3 From 057e7c7ff9f91a36a761588c53826bd6a710aeba Mon Sep 17 00:00:00 2001 From: Jonathan Corbet Date: Fri, 16 May 2008 14:01:12 -0600 Subject: infiniband: more BKL pushdown Be extra-cautious and protect the remaining open() functions. Signed-off-by: Jonathan Corbet --- drivers/infiniband/core/user_mad.c | 7 ++++++- drivers/infiniband/core/uverbs_main.c | 9 +++++++-- 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index 3aa2db54eae..1fc17940a76 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -47,6 +47,7 @@ #include #include #include +#include #include @@ -783,14 +784,17 @@ static int ib_umad_open(struct inode *inode, struct file *filp) struct ib_umad_file *file; int ret = 0; + lock_kernel(); spin_lock(&port_lock); port = umad_port[iminor(inode) - IB_UMAD_MINOR_BASE]; if (port) kref_get(&port->umad_dev->ref); spin_unlock(&port_lock); - if (!port) + if (!port) { + unlock_kernel(); return -ENXIO; + } mutex_lock(&port->file_mutex); @@ -819,6 +823,7 @@ static int ib_umad_open(struct inode *inode, struct file *filp) out: mutex_unlock(&port->file_mutex); + unlock_kernel(); return ret; } diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index cc1afa28c18..8ede1e475ce 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -45,6 +45,7 @@ #include #include #include +#include #include @@ -616,14 +617,17 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp) struct ib_uverbs_file *file; int ret; + lock_kernel(); spin_lock(&map_lock); dev = dev_table[iminor(inode) - IB_UVERBS_BASE_MINOR]; if (dev) kref_get(&dev->ref); spin_unlock(&map_lock); - if (!dev) + if (!dev) { + unlock_kernel(); return -ENXIO; + } if (!try_module_get(dev->ib_dev->owner)) { ret = -ENODEV; @@ -644,6 +648,7 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp) filp->private_data = file; + unlock_kernel(); return 0; err_module: @@ -651,7 +656,7 @@ err_module: err: kref_put(&dev->ref, ib_uverbs_release_dev); - + unlock_kernel(); return ret; } -- cgit v1.2.3 From f2b9857eee17797541b845782ade4d7a9d50f843 Mon Sep 17 00:00:00 2001 From: Jonathan Corbet Date: Sun, 18 May 2008 15:32:43 -0600 Subject: Add a bunch of cycle_kernel_lock() calls All of the open() functions which don't need the BKL on their face may still depend on its acquisition to serialize opens against driver initialization. So make those functions acquire then release the BKL to be on the safe side. Signed-off-by: Jonathan Corbet --- drivers/infiniband/core/ucm.c | 3 ++- drivers/infiniband/hw/ipath/ipath_file_ops.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index 3e6a8ff6d76..b25675faaaf 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -45,6 +45,7 @@ #include #include #include +#include #include @@ -1155,11 +1156,11 @@ static unsigned int ib_ucm_poll(struct file *filp, return mask; } -/* No BKL needed here: no global resources used */ static int ib_ucm_open(struct inode *inode, struct file *filp) { struct ib_ucm_file *file; + cycle_kernel_lock(); file = kmalloc(sizeof(*file), GFP_KERNEL); if (!file) return -ENOMEM; diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c index e80cfbd4f3f..35f301c88b5 100644 --- a/drivers/infiniband/hw/ipath/ipath_file_ops.c +++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c @@ -39,6 +39,7 @@ #include #include #include +#include #include #include "ipath_kernel.h" @@ -1812,10 +1813,10 @@ done: return ret; } -/* No BKL needed here */ static int ipath_open(struct inode *in, struct file *fp) { /* The real work is performed later in ipath_assign_port() */ + cycle_kernel_lock(); fp->private_data = kzalloc(sizeof(struct ipath_filedata), GFP_KERNEL); return fp->private_data ? 0 : -ENOMEM; } -- cgit v1.2.3 From 6b0ee363b294c3724224909dcb0b80f7dac3dfd6 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Tue, 20 May 2008 19:16:02 +0200 Subject: infiniband-ucma: BKL pushdown Signed-off-by: Arnd Bergmann --- drivers/infiniband/core/ucma.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index ca4cf3a511a..195f97302fe 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -38,6 +38,7 @@ #include #include #include +#include #include #include @@ -1156,6 +1157,7 @@ static int ucma_open(struct inode *inode, struct file *filp) if (!file) return -ENOMEM; + lock_kernel(); INIT_LIST_HEAD(&file->event_list); INIT_LIST_HEAD(&file->ctx_list); init_waitqueue_head(&file->poll_wait); @@ -1163,6 +1165,7 @@ static int ucma_open(struct inode *inode, struct file *filp) filp->private_data = file; file->filp = filp; + unlock_kernel(); return 0; } -- cgit v1.2.3 From 5b2d281acb04a29fcdc76ced5ca6099565a0747f Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Fri, 27 Jun 2008 14:43:20 -0700 Subject: IB/uverbs: BKL is not needed for ib_uverbs_open() Remove explicit lock_kernel() calls and document why the code is safe. Signed-off-by: Roland Dreier Signed-off-by: Jonathan Corbet --- drivers/infiniband/core/uverbs_main.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 8ede1e475ce..fdfcf7910d9 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -45,7 +45,6 @@ #include #include #include -#include #include @@ -611,23 +610,32 @@ static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) return file->device->ib_dev->mmap(file->ucontext, vma); } +/* + * ib_uverbs_open() does not need the BKL: + * + * - dev_table[] accesses are protected by map_lock, the + * ib_uverbs_device structures are properly reference counted, and + * everything else is purely local to the file being created, so + * races against other open calls are not a problem; + * - there is no ioctl method to race against; + * - the device is added to dev_table[] as the last part of module + * initialization, the open method will either immediately run + * -ENXIO, or all required initialization will be done. + */ static int ib_uverbs_open(struct inode *inode, struct file *filp) { struct ib_uverbs_device *dev; struct ib_uverbs_file *file; int ret; - lock_kernel(); spin_lock(&map_lock); dev = dev_table[iminor(inode) - IB_UVERBS_BASE_MINOR]; if (dev) kref_get(&dev->ref); spin_unlock(&map_lock); - if (!dev) { - unlock_kernel(); + if (!dev) return -ENXIO; - } if (!try_module_get(dev->ib_dev->owner)) { ret = -ENODEV; @@ -648,7 +656,6 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp) filp->private_data = file; - unlock_kernel(); return 0; err_module: @@ -656,7 +663,6 @@ err_module: err: kref_put(&dev->ref, ib_uverbs_release_dev); - unlock_kernel(); return ret; } -- cgit v1.2.3 From 5e19cf663be534c7c15a35a86fa7ddc9f797e4f4 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Tue, 8 Jul 2008 14:40:05 -0700 Subject: RDMA/cxgb3: Fix regression caused by class_device -> device conversion The change to iwch_provider.c in commit f4e91eb4 ("IB: convert struct class_device to struct device") undid the fix done in commit 7f049f2f ("RDMA/cxgb3: Hold rtnl_lock() around ethtool get_drvinfo call"). It removed the calls to rtnl_lock() that serialized the iw_cxgb3 ethtool ops calls into the cxgb3 driver. This locking is needed to avoid messing up the internal state of the cxgb3 driver. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/iwch_provider.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 8934178a23e..95f82cfb6c5 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1096,7 +1096,9 @@ static ssize_t show_fw_ver(struct device *dev, struct device_attribute *attr, ch struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev; PDBG("%s dev 0x%p\n", __func__, dev); + rtnl_lock(); lldev->ethtool_ops->get_drvinfo(lldev, &info); + rtnl_unlock(); return sprintf(buf, "%s\n", info.fw_version); } @@ -1109,7 +1111,9 @@ static ssize_t show_hca(struct device *dev, struct device_attribute *attr, struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev; PDBG("%s dev 0x%p\n", __func__, dev); + rtnl_lock(); lldev->ethtool_ops->get_drvinfo(lldev, &info); + rtnl_unlock(); return sprintf(buf, "%s\n", info.driver); } -- cgit v1.2.3 From feae1ef116ed381625d3731c5ae4f4ebcb3fa302 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Fri, 11 Jul 2008 13:54:40 -0700 Subject: IB/umad: BKL is not needed for ib_umad_open() Remove explicit lock_kernel() calls and document why the code is safe. Signed-off-by: Roland Dreier Signed-off-by: Jonathan Corbet --- drivers/infiniband/core/user_mad.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index 1fc17940a76..ef6dd825fee 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -47,7 +47,6 @@ #include #include #include -#include #include @@ -778,23 +777,33 @@ static long ib_umad_compat_ioctl(struct file *filp, unsigned int cmd, } #endif +/* + * ib_umad_open() does not need the BKL: + * + * - umad_port[] accesses are protected by port_lock, the + * ib_umad_port structures are properly reference counted, and + * everything else is purely local to the file being created, so + * races against other open calls are not a problem; + * - the ioctl method does not affect any global state outside of the + * file structure being operated on; + * - the port is added to umad_port[] as the last part of module + * initialization so the open method will either immediately run + * -ENXIO, or all required initialization will be done. + */ static int ib_umad_open(struct inode *inode, struct file *filp) { struct ib_umad_port *port; struct ib_umad_file *file; int ret = 0; - lock_kernel(); spin_lock(&port_lock); port = umad_port[iminor(inode) - IB_UMAD_MINOR_BASE]; if (port) kref_get(&port->umad_dev->ref); spin_unlock(&port_lock); - if (!port) { - unlock_kernel(); + if (!port) return -ENXIO; - } mutex_lock(&port->file_mutex); @@ -823,7 +832,6 @@ static int ib_umad_open(struct inode *inode, struct file *filp) out: mutex_unlock(&port->file_mutex); - unlock_kernel(); return ret; } -- cgit v1.2.3 From 40753caa364bfba60ebd5e2a8bdf366ef175d03c Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:53:56 -0500 Subject: [SCSI] iscsi class, iscsi_tcp/iser: add host arg to session creation iscsi offload (bnx2i and qla4xx) allocate a scsi host per hba, so the session creation path needs a shost/host_no argument. Software iscsi/iser will follow the same behabior as before where it allcoates a host per session, but in the future iser will probably look more like bnx2i where the host's parent is the hardware (rnic for iser and for bnx2i it is the nic), because it does not use a socket layer like how iscsi_tcp does. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index aeb58cae9a3..efc121986c5 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -368,6 +368,7 @@ static struct iscsi_transport iscsi_iser_transport; static struct iscsi_cls_session * iscsi_iser_session_create(struct iscsi_transport *iscsit, struct scsi_transport_template *scsit, + struct Scsi_Host *shost, uint16_t cmds_max, uint16_t qdepth, uint32_t initial_cmdsn, uint32_t *hostno) { -- cgit v1.2.3 From d3826721b198001c55353b1c54e10843068aae63 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:53:57 -0500 Subject: [SCSI] iscsi class, iscsi drivers: remove unused iscsi_transport attrs max_cmd_len and max_conn are not really used. max_cmd_len is always 16 and can be set by the LLD. max_conn is always one since we do not support MCS. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 1 - drivers/infiniband/ulp/iser/iscsi_iser.h | 1 - 2 files changed, 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index efc121986c5..32f5d5e79ab 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -592,7 +592,6 @@ static struct iscsi_transport iscsi_iser_transport = { .host_template = &iscsi_iser_sht, .conndata_size = sizeof(struct iscsi_conn), .max_lun = ISCSI_ISER_MAX_LUN, - .max_cmd_len = ISCSI_ISER_MAX_CMD_LEN, /* session management */ .create_session = iscsi_iser_session_create, .destroy_session = iscsi_session_teardown, diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index a8c1b300e34..66a2f30ada0 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -96,7 +96,6 @@ /* support upto 512KB in one RDMA */ #define ISCSI_ISER_SG_TABLESIZE (0x80000 >> SHIFT_4K) #define ISCSI_ISER_MAX_LUN 256 -#define ISCSI_ISER_MAX_CMD_LEN 16 /* QP settings */ /* Maximal bounds on received asynchronous PDUs */ -- cgit v1.2.3 From 756135215ec743be6fdce2bdebe8cdb9f8a231f6 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:53:59 -0500 Subject: [SCSI] iscsi: remove session and host binding in libiscsi bnx2i allocates a host per netdevice but will use libiscsi, so this unbinds the session from the host in that code. This will also be useful for the iser parent device dma settings fixes. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 74 +++++++++++++++++++++++--------- 1 file changed, 54 insertions(+), 20 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 32f5d5e79ab..5a750042e2b 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -74,6 +74,10 @@ #include "iscsi_iser.h" +static struct scsi_host_template iscsi_iser_sht; +static struct iscsi_transport iscsi_iser_transport; +static struct scsi_transport_template *iscsi_iser_scsi_transport; + static unsigned int iscsi_max_lun = 512; module_param_named(max_lun, iscsi_max_lun, uint, S_IRUGO); @@ -363,40 +367,64 @@ iscsi_iser_conn_start(struct iscsi_cls_conn *cls_conn) return iscsi_conn_start(cls_conn); } -static struct iscsi_transport iscsi_iser_transport; +static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session) +{ + struct Scsi_Host *shost = iscsi_session_to_shost(cls_session); + + iscsi_session_teardown(cls_session); + scsi_remove_host(shost); + iscsi_host_teardown(shost); + scsi_host_put(shost); +} static struct iscsi_cls_session * -iscsi_iser_session_create(struct iscsi_transport *iscsit, - struct scsi_transport_template *scsit, - struct Scsi_Host *shost, - uint16_t cmds_max, uint16_t qdepth, - uint32_t initial_cmdsn, uint32_t *hostno) +iscsi_iser_session_create(struct Scsi_Host *shost, + uint16_t cmds_max, uint16_t qdepth, + uint32_t initial_cmdsn, uint32_t *hostno) { struct iscsi_cls_session *cls_session; struct iscsi_session *session; int i; - uint32_t hn; struct iscsi_cmd_task *ctask; struct iscsi_mgmt_task *mtask; struct iscsi_iser_cmd_task *iser_ctask; struct iser_desc *desc; + if (shost) { + printk(KERN_ERR "iscsi_tcp: invalid shost %d.\n", + shost->host_no); + return NULL; + } + + shost = scsi_host_alloc(&iscsi_iser_sht, 0); + if (!shost) + return NULL; + shost->transportt = iscsi_iser_scsi_transport; + shost->max_lun = iscsi_max_lun; + shost->max_id = 0; + shost->max_channel = 0; + shost->max_cmd_len = 16; + + iscsi_host_setup(shost, qdepth); + + if (scsi_add_host(shost, NULL)) + goto free_host; + *hostno = shost->host_no; + /* * we do not support setting can_queue cmd_per_lun from userspace yet * because we preallocate so many resources */ - cls_session = iscsi_session_setup(iscsit, scsit, + cls_session = iscsi_session_setup(&iscsi_iser_transport, shost, ISCSI_DEF_XMIT_CMDS_MAX, - ISCSI_MAX_CMD_PER_LUN, sizeof(struct iscsi_iser_cmd_task), sizeof(struct iser_desc), - initial_cmdsn, &hn); + initial_cmdsn); if (!cls_session) - return NULL; - - *hostno = hn; - session = class_to_transport_session(cls_session); + goto remove_host; + session = cls_session->dd_data; + shost->can_queue = session->cmds_max; /* libiscsi setup itts, data and pool so just set desc fields */ for (i = 0; i < session->cmds_max; i++) { ctask = session->cmds[i]; @@ -413,6 +441,13 @@ iscsi_iser_session_create(struct iscsi_transport *iscsit, } return cls_session; + +remove_host: + scsi_remove_host(shost); +free_host: + iscsi_host_teardown(shost); + scsi_host_put(shost); + return NULL; } static int @@ -589,12 +624,11 @@ static struct iscsi_transport iscsi_iser_transport = { .host_param_mask = ISCSI_HOST_HWADDRESS | ISCSI_HOST_NETDEV_NAME | ISCSI_HOST_INITIATOR_NAME, - .host_template = &iscsi_iser_sht, .conndata_size = sizeof(struct iscsi_conn), - .max_lun = ISCSI_ISER_MAX_LUN, + .sessiondata_size = sizeof(struct iscsi_session), /* session management */ .create_session = iscsi_iser_session_create, - .destroy_session = iscsi_session_teardown, + .destroy_session = iscsi_iser_session_destroy, /* connection management */ .create_conn = iscsi_iser_conn_create, .bind_conn = iscsi_iser_conn_bind, @@ -633,8 +667,6 @@ static int __init iser_init(void) return -EINVAL; } - iscsi_iser_transport.max_lun = iscsi_max_lun; - memset(&ig, 0, sizeof(struct iser_global)); ig.desc_cache = kmem_cache_create("iser_descriptors", @@ -650,7 +682,9 @@ static int __init iser_init(void) mutex_init(&ig.connlist_mutex); INIT_LIST_HEAD(&ig.connlist); - if (!iscsi_register_transport(&iscsi_iser_transport)) { + iscsi_iser_scsi_transport = iscsi_register_transport( + &iscsi_iser_transport); + if (!iscsi_iser_scsi_transport) { iser_err("iscsi_register_transport failed\n"); err = -EINVAL; goto register_transport_failure; -- cgit v1.2.3 From a4804cd6eb19318ae8d08ea967cfeaaf5c5b68a6 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:00 -0500 Subject: [SCSI] iscsi: add iscsi host helpers This finishes the host/session unbinding, by adding some helpers to add and remove hosts and the session they manage. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 17 ++++++----------- 1 file changed, 6 insertions(+), 11 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 5a750042e2b..62e35e503e4 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -371,10 +371,8 @@ static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session) { struct Scsi_Host *shost = iscsi_session_to_shost(cls_session); - iscsi_session_teardown(cls_session); - scsi_remove_host(shost); - iscsi_host_teardown(shost); - scsi_host_put(shost); + iscsi_host_remove(shost); + iscsi_host_free(shost); } static struct iscsi_cls_session * @@ -396,7 +394,7 @@ iscsi_iser_session_create(struct Scsi_Host *shost, return NULL; } - shost = scsi_host_alloc(&iscsi_iser_sht, 0); + shost = iscsi_host_alloc(&iscsi_iser_sht, 0, ISCSI_MAX_CMD_PER_LUN); if (!shost) return NULL; shost->transportt = iscsi_iser_scsi_transport; @@ -405,9 +403,7 @@ iscsi_iser_session_create(struct Scsi_Host *shost, shost->max_channel = 0; shost->max_cmd_len = 16; - iscsi_host_setup(shost, qdepth); - - if (scsi_add_host(shost, NULL)) + if (iscsi_host_add(shost, NULL)) goto free_host; *hostno = shost->host_no; @@ -443,10 +439,9 @@ iscsi_iser_session_create(struct Scsi_Host *shost, return cls_session; remove_host: - scsi_remove_host(shost); + iscsi_host_remove(shost); free_host: - iscsi_host_teardown(shost); - scsi_host_put(shost); + iscsi_host_free(shost); return NULL; } -- cgit v1.2.3 From 5d91e209fb21fb9cc765729d4c6a85a9fb6c9187 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:01 -0500 Subject: [SCSI] iscsi: remove session/conn_data_size from iscsi_transport This removes the session and conn data_size fields from the iscsi_transport. Just pass in the value like with host allocation. This patch also makes it so the LLD iscsi_conn data is allocated with the iscsi_cls_conn. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 16 +++------------- 1 file changed, 3 insertions(+), 13 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 62e35e503e4..9b34946eb00 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -279,7 +279,7 @@ iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx) struct iscsi_cls_conn *cls_conn; struct iscsi_iser_conn *iser_conn; - cls_conn = iscsi_conn_setup(cls_session, conn_idx); + cls_conn = iscsi_conn_setup(cls_session, sizeof(*iser_conn), conn_idx); if (!cls_conn) return NULL; conn = cls_conn->dd_data; @@ -290,10 +290,7 @@ iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx) */ conn->max_recv_dlength = 128; - iser_conn = kzalloc(sizeof(*iser_conn), GFP_KERNEL); - if (!iser_conn) - goto conn_alloc_fail; - + iser_conn = conn->dd_data; /* currently this is the only field which need to be initiated */ rwlock_init(&iser_conn->lock); @@ -301,10 +298,6 @@ iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx) iser_conn->iscsi_conn = conn; return cls_conn; - -conn_alloc_fail: - iscsi_conn_teardown(cls_conn); - return NULL; } static void @@ -313,10 +306,9 @@ iscsi_iser_conn_destroy(struct iscsi_cls_conn *cls_conn) struct iscsi_conn *conn = cls_conn->dd_data; struct iscsi_iser_conn *iser_conn = conn->dd_data; - iscsi_conn_teardown(cls_conn); if (iser_conn->ib_conn) iser_conn->ib_conn->iser_conn = NULL; - kfree(iser_conn); + iscsi_conn_teardown(cls_conn); } static int @@ -619,8 +611,6 @@ static struct iscsi_transport iscsi_iser_transport = { .host_param_mask = ISCSI_HOST_HWADDRESS | ISCSI_HOST_NETDEV_NAME | ISCSI_HOST_INITIATOR_NAME, - .conndata_size = sizeof(struct iscsi_conn), - .sessiondata_size = sizeof(struct iscsi_session), /* session management */ .create_session = iscsi_iser_session_create, .destroy_session = iscsi_iser_session_destroy, -- cgit v1.2.3 From b40977d95fb3a1898ace6a7d97e4ed1a33a440a4 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:03 -0500 Subject: [SCSI] iser: fix handling of scsi cmnds during recovery. After the stop_conn callback has returned the LLD should not touch the scsi cmds. iscsi_tcp and libiscsi use the conn->recv_lock and suspend_rx field to halt recv path processing, but iser does not have any protection. This patch modifies iser so that userspace can just call the ep_disconnect callback, which will halt all recv IO, before calling the stop_conn callback so we do not have to worry about the conn->recv_lock and suspend rx field. iser just needs to stop the send side from accessing the ib conn. Fixup to handle when the ep poll fails and ep disconnect is called from Erez. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 42 +++++++++++++++++++++++++++++--- drivers/infiniband/ulp/iser/iscsi_iser.h | 5 ++++ drivers/infiniband/ulp/iser/iser_verbs.c | 14 ++++++++++- 3 files changed, 57 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 9b34946eb00..8a1bfb7277c 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -305,10 +305,18 @@ iscsi_iser_conn_destroy(struct iscsi_cls_conn *cls_conn) { struct iscsi_conn *conn = cls_conn->dd_data; struct iscsi_iser_conn *iser_conn = conn->dd_data; + struct iser_conn *ib_conn = iser_conn->ib_conn; - if (iser_conn->ib_conn) - iser_conn->ib_conn->iser_conn = NULL; iscsi_conn_teardown(cls_conn); + /* + * Userspace will normally call the stop callback and + * already have freed the ib_conn, but if it goofed up then + * we free it here. + */ + if (ib_conn) { + ib_conn->iser_conn = NULL; + iser_conn_put(ib_conn); + } } static int @@ -340,12 +348,29 @@ iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session, iser_conn = conn->dd_data; ib_conn->iser_conn = iser_conn; iser_conn->ib_conn = ib_conn; + iser_conn_get(ib_conn); conn->recv_lock = &iser_conn->lock; return 0; } +static void +iscsi_iser_conn_stop(struct iscsi_cls_conn *cls_conn, int flag) +{ + struct iscsi_conn *conn = cls_conn->dd_data; + struct iscsi_iser_conn *iser_conn = conn->dd_data; + struct iser_conn *ib_conn = iser_conn->ib_conn; + + iscsi_conn_stop(cls_conn, flag); + /* + * There is no unbind event so the stop callback + * must release the ref from the bind. + */ + iser_conn_put(ib_conn); + iser_conn->ib_conn = NULL; +} + static int iscsi_iser_conn_start(struct iscsi_cls_conn *cls_conn) { @@ -564,6 +589,17 @@ iscsi_iser_ep_disconnect(__u64 ep_handle) if (!ib_conn) return; + if (ib_conn->iser_conn) + /* + * Must suspend xmit path if the ep is bound to the + * iscsi_conn, so we know we are not accessing the ib_conn + * when we free it. + * + * This may not be bound if the ep poll failed. + */ + iscsi_suspend_tx(ib_conn->iser_conn->iscsi_conn); + + iser_err("ib conn %p state %d\n",ib_conn, ib_conn->state); iser_conn_terminate(ib_conn); } @@ -622,7 +658,7 @@ static struct iscsi_transport iscsi_iser_transport = { .get_conn_param = iscsi_conn_get_param, .get_session_param = iscsi_session_get_param, .start_conn = iscsi_iser_conn_start, - .stop_conn = iscsi_conn_stop, + .stop_conn = iscsi_iser_conn_stop, /* iscsi host params */ .get_host_param = iscsi_host_get_param, .set_host_param = iscsi_host_set_param, diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 66a2f30ada0..bd5c1a554ea 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -242,6 +242,7 @@ struct iser_device { struct iser_conn { struct iscsi_iser_conn *iser_conn; /* iser conn for upcalls */ enum iser_ib_conn_state state; /* rdma connection state */ + atomic_t refcount; spinlock_t lock; /* used for state changes */ struct iser_device *device; /* device context */ struct rdma_cm_id *cma_id; /* CMA ID */ @@ -314,6 +315,10 @@ void iscsi_iser_recv(struct iscsi_conn *conn, int iser_conn_init(struct iser_conn **ib_conn); +void iser_conn_get(struct iser_conn *ib_conn); + +void iser_conn_put(struct iser_conn *ib_conn); + void iser_conn_terminate(struct iser_conn *ib_conn); void iser_rcv_completion(struct iser_desc *desc, diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index d19cfe605eb..5daed2bd710 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -328,6 +328,17 @@ static void iser_conn_release(struct iser_conn *ib_conn) kfree(ib_conn); } +void iser_conn_get(struct iser_conn *ib_conn) +{ + atomic_inc(&ib_conn->refcount); +} + +void iser_conn_put(struct iser_conn *ib_conn) +{ + if (atomic_dec_and_test(&ib_conn->refcount)) + iser_conn_release(ib_conn); +} + /** * triggers start of the disconnect procedures and wait for them to be done */ @@ -349,7 +360,7 @@ void iser_conn_terminate(struct iser_conn *ib_conn) wait_event_interruptible(ib_conn->wait, ib_conn->state == ISER_CONN_DOWN); - iser_conn_release(ib_conn); + iser_conn_put(ib_conn); } static void iser_connect_error(struct rdma_cm_id *cma_id) @@ -496,6 +507,7 @@ int iser_conn_init(struct iser_conn **ibconn) init_waitqueue_head(&ib_conn->wait); atomic_set(&ib_conn->post_recv_buf_count, 0); atomic_set(&ib_conn->post_send_buf_count, 0); + atomic_set(&ib_conn->refcount, 1); INIT_LIST_HEAD(&ib_conn->conn_list); spin_lock_init(&ib_conn->lock); -- cgit v1.2.3 From 0af967f5d4f2dd1e00618d34ac988037d37a6c3b Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:04 -0500 Subject: [SCSI] libiscsi, iscsi_tcp, iser: add session cmds array accessor Currently to get a ctask from the session cmd array, you have to know to use the itt modifier. To make this easier on LLDs and so in the future we can easilly kill the session array and use the host shared map instead, this patch adds a nice wrapper to strip the itt into a session->cmds index and return a ctask. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 8 +------- drivers/infiniband/ulp/iser/iser_initiator.c | 23 ++++++++++------------- 2 files changed, 11 insertions(+), 20 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 8a1bfb7277c..7b146886906 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -98,7 +98,6 @@ iscsi_iser_recv(struct iscsi_conn *conn, struct iscsi_hdr *hdr, char *rx_data, int rx_data_len) { int rc = 0; - uint32_t ret_itt; int datalen; int ahslen; @@ -114,12 +113,7 @@ iscsi_iser_recv(struct iscsi_conn *conn, /* read AHS */ ahslen = hdr->hlength * 4; - /* verify itt (itt encoding: age+cid+itt) */ - rc = iscsi_verify_itt(conn, hdr, &ret_itt); - - if (!rc) - rc = iscsi_complete_pdu(conn, hdr, rx_data, rx_data_len); - + rc = iscsi_complete_pdu(conn, hdr, rx_data, rx_data_len); if (rc && rc != ISCSI_ERR_NO_SCSI_CMD) goto error; diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 08dc81c46f4..b82a5f2d4d3 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -537,13 +537,11 @@ void iser_rcv_completion(struct iser_desc *rx_desc, { struct iser_dto *dto = &rx_desc->dto; struct iscsi_iser_conn *conn = dto->ib_conn->iser_conn; - struct iscsi_session *session = conn->iscsi_conn->session; struct iscsi_cmd_task *ctask; struct iscsi_iser_cmd_task *iser_ctask; struct iscsi_hdr *hdr; char *rx_data = NULL; int rx_data_len = 0; - unsigned int itt; unsigned char opcode; hdr = &rx_desc->iscsi_header; @@ -559,19 +557,18 @@ void iser_rcv_completion(struct iser_desc *rx_desc, opcode = hdr->opcode & ISCSI_OPCODE_MASK; if (opcode == ISCSI_OP_SCSI_CMD_RSP) { - itt = get_itt(hdr->itt); /* mask out cid and age bits */ - if (!(itt < session->cmds_max)) + ctask = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt); + if (!ctask) iser_err("itt can't be matched to task!!! " - "conn %p opcode %d cmds_max %d itt %d\n", - conn->iscsi_conn,opcode,session->cmds_max,itt); - /* use the mapping given with the cmds array indexed by itt */ - ctask = (struct iscsi_cmd_task *)session->cmds[itt]; - iser_ctask = ctask->dd_data; - iser_dbg("itt %d ctask %p\n",itt,ctask); - iser_ctask->status = ISER_TASK_STATUS_COMPLETED; - iser_ctask_rdma_finalize(iser_ctask); + "conn %p opcode %d itt %d\n", + conn->iscsi_conn, opcode, hdr->itt); + else { + iser_ctask = ctask->dd_data; + iser_dbg("itt %d ctask %p\n",hdr->itt, ctask); + iser_ctask->status = ISER_TASK_STATUS_COMPLETED; + iser_ctask_rdma_finalize(iser_ctask); + } } - iser_dto_buffs_release(dto); iscsi_iser_recv(conn->iscsi_conn, hdr, rx_data, rx_data_len); -- cgit v1.2.3 From 2747fdb25726caa1a89229f43d99ca50af72576a Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:08 -0500 Subject: [SCSI] iser: convert ib_iser to support merged tasks Convert ib_iser to support merged tasks. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 82 ++++++++++++++-------------- drivers/infiniband/ulp/iser/iscsi_iser.h | 14 ++--- drivers/infiniband/ulp/iser/iser_initiator.c | 42 +++++++------- 3 files changed, 68 insertions(+), 70 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 7b146886906..baecca1ed42 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -124,15 +124,23 @@ error: /** - * iscsi_iser_cmd_init - Initialize iSCSI SCSI_READ or SCSI_WRITE commands + * iscsi_iser_task_init - Initialize ctask + * @ctask: iscsi ctask * - **/ + * Initialize the ctask for the scsi command or mgmt command. + */ static int -iscsi_iser_cmd_init(struct iscsi_cmd_task *ctask) +iscsi_iser_task_init(struct iscsi_cmd_task *ctask) { - struct iscsi_iser_conn *iser_conn = ctask->conn->dd_data; + struct iscsi_iser_conn *iser_conn = ctask->conn->dd_data; struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + /* mgmt ctask */ + if (!ctask->sc) { + iser_ctask->desc.data = ctask->data; + return 0; + } + iser_ctask->command_sent = 0; iser_ctask->iser_conn = iser_conn; iser_ctask_rdma_init(iser_ctask); @@ -140,9 +148,9 @@ iscsi_iser_cmd_init(struct iscsi_cmd_task *ctask) } /** - * iscsi_mtask_xmit - xmit management(immediate) task + * iscsi_iser_mtask_xmit - xmit management(immediate) ctask * @conn: iscsi connection - * @mtask: task management task + * @ctask: ctask management ctask * * Notes: * The function can return -EAGAIN in which case caller must @@ -151,20 +159,19 @@ iscsi_iser_cmd_init(struct iscsi_cmd_task *ctask) * **/ static int -iscsi_iser_mtask_xmit(struct iscsi_conn *conn, - struct iscsi_mgmt_task *mtask) +iscsi_iser_mtask_xmit(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { int error = 0; - debug_scsi("mtask deq [cid %d itt 0x%x]\n", conn->id, mtask->itt); + debug_scsi("ctask deq [cid %d itt 0x%x]\n", conn->id, ctask->itt); - error = iser_send_control(conn, mtask); + error = iser_send_control(conn, ctask); - /* since iser xmits control with zero copy, mtasks can not be recycled + /* since iser xmits control with zero copy, ctasks can not be recycled * right after sending them. * The recycling scheme is based on whether a response is expected - * - if yes, the mtask is recycled at iscsi_complete_pdu - * - if no, the mtask is recycled at iser_snd_completion + * - if yes, the ctask is recycled at iscsi_complete_pdu + * - if no, the ctask is recycled at iser_snd_completion */ if (error && error != -ENOBUFS) iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED); @@ -173,7 +180,7 @@ iscsi_iser_mtask_xmit(struct iscsi_conn *conn, } static int -iscsi_iser_ctask_xmit_unsol_data(struct iscsi_conn *conn, +iscsi_iser_task_xmit_unsol_data(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_data hdr; @@ -190,24 +197,27 @@ iscsi_iser_ctask_xmit_unsol_data(struct iscsi_conn *conn, error = iser_send_data_out(conn, ctask, &hdr); if (error) { ctask->unsol_datasn--; - goto iscsi_iser_ctask_xmit_unsol_data_exit; + goto iscsi_iser_task_xmit_unsol_data_exit; } ctask->unsol_count -= ctask->data_count; debug_scsi("Need to send %d more as data-out PDUs\n", ctask->unsol_count); } -iscsi_iser_ctask_xmit_unsol_data_exit: +iscsi_iser_task_xmit_unsol_data_exit: return error; } static int -iscsi_iser_ctask_xmit(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask) +iscsi_iser_task_xmit(struct iscsi_cmd_task *ctask) { + struct iscsi_conn *conn = ctask->conn; struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; int error = 0; + if (!ctask->sc) + return iscsi_iser_mtask_xmit(conn, ctask); + if (ctask->sc->sc_data_direction == DMA_TO_DEVICE) { BUG_ON(scsi_bufflen(ctask->sc) == 0); @@ -223,25 +233,29 @@ iscsi_iser_ctask_xmit(struct iscsi_conn *conn, if (!iser_ctask->command_sent) { error = iser_send_command(conn, ctask); if (error) - goto iscsi_iser_ctask_xmit_exit; + goto iscsi_iser_task_xmit_exit; iser_ctask->command_sent = 1; } /* Send unsolicited data-out PDU(s) if necessary */ if (ctask->unsol_count) - error = iscsi_iser_ctask_xmit_unsol_data(conn, ctask); + error = iscsi_iser_task_xmit_unsol_data(conn, ctask); - iscsi_iser_ctask_xmit_exit: + iscsi_iser_task_xmit_exit: if (error && error != -ENOBUFS) iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED); return error; } static void -iscsi_iser_cleanup_ctask(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) +iscsi_iser_cleanup_task(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + /* mgmt tasks do not need special cleanup */ + if (!ctask->sc) + return; + if (iser_ctask->status == ISER_TASK_STATUS_STARTED) { iser_ctask->status = ISER_TASK_STATUS_COMPLETED; iser_ctask_rdma_finalize(iser_ctask); @@ -394,10 +408,8 @@ iscsi_iser_session_create(struct Scsi_Host *shost, struct iscsi_cls_session *cls_session; struct iscsi_session *session; int i; - struct iscsi_cmd_task *ctask; - struct iscsi_mgmt_task *mtask; + struct iscsi_cmd_task *ctask; struct iscsi_iser_cmd_task *iser_ctask; - struct iser_desc *desc; if (shost) { printk(KERN_ERR "iscsi_tcp: invalid shost %d.\n", @@ -425,28 +437,19 @@ iscsi_iser_session_create(struct Scsi_Host *shost, cls_session = iscsi_session_setup(&iscsi_iser_transport, shost, ISCSI_DEF_XMIT_CMDS_MAX, sizeof(struct iscsi_iser_cmd_task), - sizeof(struct iser_desc), initial_cmdsn); if (!cls_session) goto remove_host; session = cls_session->dd_data; - shost->can_queue = session->cmds_max; + shost->can_queue = session->scsi_cmds_max; /* libiscsi setup itts, data and pool so just set desc fields */ for (i = 0; i < session->cmds_max; i++) { - ctask = session->cmds[i]; + ctask = session->cmds[i]; iser_ctask = ctask->dd_data; ctask->hdr = (struct iscsi_cmd *)&iser_ctask->desc.iscsi_header; ctask->hdr_max = sizeof(iser_ctask->desc.iscsi_header); } - - for (i = 0; i < session->mgmtpool_max; i++) { - mtask = session->mgmt_cmds[i]; - desc = mtask->dd_data; - mtask->hdr = &desc->iscsi_header; - desc->data = mtask->data; - } - return cls_session; remove_host: @@ -659,10 +662,9 @@ static struct iscsi_transport iscsi_iser_transport = { /* IO */ .send_pdu = iscsi_conn_send_pdu, .get_stats = iscsi_iser_conn_get_stats, - .init_cmd_task = iscsi_iser_cmd_init, - .xmit_cmd_task = iscsi_iser_ctask_xmit, - .xmit_mgmt_task = iscsi_iser_mtask_xmit, - .cleanup_cmd_task = iscsi_iser_cleanup_ctask, + .init_task = iscsi_iser_task_init, + .xmit_task = iscsi_iser_task_xmit, + .cleanup_task = iscsi_iser_cleanup_task, /* recovery */ .session_recovery_timedout = iscsi_session_recovery_timedout, diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index bd5c1a554ea..96a600f127c 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -298,15 +298,15 @@ extern int iser_debug_level; /* allocate connection resources needed for rdma functionality */ int iser_conn_set_full_featured_mode(struct iscsi_conn *conn); -int iser_send_control(struct iscsi_conn *conn, - struct iscsi_mgmt_task *mtask); +int iser_send_control(struct iscsi_conn *conn, + struct iscsi_cmd_task *ctask); -int iser_send_command(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask); +int iser_send_command(struct iscsi_conn *conn, + struct iscsi_cmd_task *ctask); -int iser_send_data_out(struct iscsi_conn *conn, +int iser_send_data_out(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask, - struct iscsi_data *hdr); + struct iscsi_data *hdr); void iscsi_iser_recv(struct iscsi_conn *conn, struct iscsi_hdr *hdr, @@ -326,7 +326,7 @@ void iser_rcv_completion(struct iser_desc *desc, void iser_snd_completion(struct iser_desc *desc); -void iser_ctask_rdma_init(struct iscsi_iser_cmd_task *ctask); +void iser_ctask_rdma_init(struct iscsi_iser_cmd_task *ctask); void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *ctask); diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index b82a5f2d4d3..4ea78fbeee9 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -300,13 +300,13 @@ int iser_conn_set_full_featured_mode(struct iscsi_conn *conn) } static int -iser_check_xmit(struct iscsi_conn *conn, void *task) +iser_check_xmit(struct iscsi_conn *conn, void *ctask) { struct iscsi_iser_conn *iser_conn = conn->dd_data; if (atomic_read(&iser_conn->ib_conn->post_send_buf_count) == ISER_QP_MAX_REQ_DTOS) { - iser_dbg("%ld can't xmit task %p\n",jiffies,task); + iser_dbg("%ld can't xmit ctask %p\n",jiffies,ctask); return -ENOBUFS; } return 0; @@ -316,7 +316,7 @@ iser_check_xmit(struct iscsi_conn *conn, void *task) /** * iser_send_command - send command PDU */ -int iser_send_command(struct iscsi_conn *conn, +int iser_send_command(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_iser_conn *iser_conn = conn->dd_data; @@ -395,7 +395,7 @@ send_command_error: /** * iser_send_data_out - send data out PDU */ -int iser_send_data_out(struct iscsi_conn *conn, +int iser_send_data_out(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask, struct iscsi_data *hdr) { @@ -470,10 +470,11 @@ send_data_out_error: } int iser_send_control(struct iscsi_conn *conn, - struct iscsi_mgmt_task *mtask) + struct iscsi_cmd_task *ctask) { struct iscsi_iser_conn *iser_conn = conn->dd_data; - struct iser_desc *mdesc = mtask->dd_data; + struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iser_desc *mdesc = &iser_ctask->desc; struct iser_dto *send_dto = NULL; unsigned long data_seg_len; int err = 0; @@ -485,7 +486,7 @@ int iser_send_control(struct iscsi_conn *conn, return -EPERM; } - if (iser_check_xmit(conn,mtask)) + if (iser_check_xmit(conn, ctask)) return -ENOBUFS; /* build the tx desc regd header and add it to the tx desc dto */ @@ -498,14 +499,14 @@ int iser_send_control(struct iscsi_conn *conn, iser_reg_single(device, send_dto->regd[0], DMA_TO_DEVICE); - data_seg_len = ntoh24(mtask->hdr->dlength); + data_seg_len = ntoh24(ctask->hdr->dlength); if (data_seg_len > 0) { regd_buf = &mdesc->data_regd_buf; memset(regd_buf, 0, sizeof(struct iser_regd_buf)); regd_buf->device = device; - regd_buf->virt_addr = mtask->data; - regd_buf->data_size = mtask->data_count; + regd_buf->virt_addr = ctask->data; + regd_buf->data_size = ctask->data_count; iser_reg_single(device, regd_buf, DMA_TO_DEVICE); iser_dto_add_regd_buff(send_dto, regd_buf, @@ -535,7 +536,7 @@ send_control_error: void iser_rcv_completion(struct iser_desc *rx_desc, unsigned long dto_xfer_len) { - struct iser_dto *dto = &rx_desc->dto; + struct iser_dto *dto = &rx_desc->dto; struct iscsi_iser_conn *conn = dto->ib_conn->iser_conn; struct iscsi_cmd_task *ctask; struct iscsi_iser_cmd_task *iser_ctask; @@ -559,7 +560,7 @@ void iser_rcv_completion(struct iser_desc *rx_desc, if (opcode == ISCSI_OP_SCSI_CMD_RSP) { ctask = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt); if (!ctask) - iser_err("itt can't be matched to task!!! " + iser_err("itt can't be matched to ctask!!! " "conn %p opcode %d itt %d\n", conn->iscsi_conn, opcode, hdr->itt); else { @@ -577,7 +578,7 @@ void iser_rcv_completion(struct iser_desc *rx_desc, kmem_cache_free(ig.desc_cache, rx_desc); /* decrementing conn->post_recv_buf_count only --after-- freeing the * - * task eliminates the need to worry on tasks which are completed in * + * ctask eliminates the need to worry on ctasks which are completed in * * parallel to the execution of iser_conn_term. So the code that waits * * for the posted rx bufs refcount to become zero handles everything */ atomic_dec(&conn->ib_conn->post_recv_buf_count); @@ -589,7 +590,7 @@ void iser_snd_completion(struct iser_desc *tx_desc) struct iser_conn *ib_conn = dto->ib_conn; struct iscsi_iser_conn *iser_conn = ib_conn->iser_conn; struct iscsi_conn *conn = iser_conn->iscsi_conn; - struct iscsi_mgmt_task *mtask; + struct iscsi_cmd_task *ctask; int resume_tx = 0; iser_dbg("Initiator, Data sent dto=0x%p\n", dto); @@ -612,15 +613,10 @@ void iser_snd_completion(struct iser_desc *tx_desc) if (tx_desc->type == ISCSI_TX_CONTROL) { /* this arithmetic is legal by libiscsi dd_data allocation */ - mtask = (void *) ((long)(void *)tx_desc - - sizeof(struct iscsi_mgmt_task)); - if (mtask->hdr->itt == RESERVED_ITT) { - struct iscsi_session *session = conn->session; - - spin_lock(&conn->session->lock); - iscsi_free_mgmt_task(conn, mtask); - spin_unlock(&session->lock); - } + ctask = (void *) ((long)(void *)tx_desc - + sizeof(struct iscsi_cmd_task)); + if (ctask->hdr->itt == RESERVED_ITT) + iscsi_put_ctask(ctask); } } -- cgit v1.2.3 From 2261ec3d686e35c1a6088ab7f00a1d02b528b994 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:11 -0500 Subject: [SCSI] iser: handle iscsi_cmd_task rename This handles the iscsi_cmd_task rename and renames the iser cmd task to iser task. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 114 ++++++++-------- drivers/infiniband/ulp/iser/iscsi_iser.h | 24 ++-- drivers/infiniband/ulp/iser/iser_initiator.c | 190 +++++++++++++-------------- drivers/infiniband/ulp/iser/iser_memory.c | 77 +++++------ 4 files changed, 203 insertions(+), 202 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index baecca1ed42..86d9c42f0d3 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -124,33 +124,33 @@ error: /** - * iscsi_iser_task_init - Initialize ctask - * @ctask: iscsi ctask + * iscsi_iser_task_init - Initialize task + * @task: iscsi task * - * Initialize the ctask for the scsi command or mgmt command. + * Initialize the task for the scsi command or mgmt command. */ static int -iscsi_iser_task_init(struct iscsi_cmd_task *ctask) +iscsi_iser_task_init(struct iscsi_task *task) { - struct iscsi_iser_conn *iser_conn = ctask->conn->dd_data; - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_conn *iser_conn = task->conn->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; - /* mgmt ctask */ - if (!ctask->sc) { - iser_ctask->desc.data = ctask->data; + /* mgmt task */ + if (!task->sc) { + iser_task->desc.data = task->data; return 0; } - iser_ctask->command_sent = 0; - iser_ctask->iser_conn = iser_conn; - iser_ctask_rdma_init(iser_ctask); + iser_task->command_sent = 0; + iser_task->iser_conn = iser_conn; + iser_task_rdma_init(iser_task); return 0; } /** - * iscsi_iser_mtask_xmit - xmit management(immediate) ctask + * iscsi_iser_mtask_xmit - xmit management(immediate) task * @conn: iscsi connection - * @ctask: ctask management ctask + * @task: task management task * * Notes: * The function can return -EAGAIN in which case caller must @@ -159,19 +159,19 @@ iscsi_iser_task_init(struct iscsi_cmd_task *ctask) * **/ static int -iscsi_iser_mtask_xmit(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) +iscsi_iser_mtask_xmit(struct iscsi_conn *conn, struct iscsi_task *task) { int error = 0; - debug_scsi("ctask deq [cid %d itt 0x%x]\n", conn->id, ctask->itt); + debug_scsi("task deq [cid %d itt 0x%x]\n", conn->id, task->itt); - error = iser_send_control(conn, ctask); + error = iser_send_control(conn, task); - /* since iser xmits control with zero copy, ctasks can not be recycled + /* since iser xmits control with zero copy, tasks can not be recycled * right after sending them. * The recycling scheme is based on whether a response is expected - * - if yes, the ctask is recycled at iscsi_complete_pdu - * - if no, the ctask is recycled at iser_snd_completion + * - if yes, the task is recycled at iscsi_complete_pdu + * - if no, the task is recycled at iser_snd_completion */ if (error && error != -ENOBUFS) iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED); @@ -181,27 +181,27 @@ iscsi_iser_mtask_xmit(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) static int iscsi_iser_task_xmit_unsol_data(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask) + struct iscsi_task *task) { struct iscsi_data hdr; int error = 0; /* Send data-out PDUs while there's still unsolicited data to send */ - while (ctask->unsol_count > 0) { - iscsi_prep_unsolicit_data_pdu(ctask, &hdr); + while (task->unsol_count > 0) { + iscsi_prep_unsolicit_data_pdu(task, &hdr); debug_scsi("Sending data-out: itt 0x%x, data count %d\n", - hdr.itt, ctask->data_count); + hdr.itt, task->data_count); /* the buffer description has been passed with the command */ /* Send the command */ - error = iser_send_data_out(conn, ctask, &hdr); + error = iser_send_data_out(conn, task, &hdr); if (error) { - ctask->unsol_datasn--; + task->unsol_datasn--; goto iscsi_iser_task_xmit_unsol_data_exit; } - ctask->unsol_count -= ctask->data_count; + task->unsol_count -= task->data_count; debug_scsi("Need to send %d more as data-out PDUs\n", - ctask->unsol_count); + task->unsol_count); } iscsi_iser_task_xmit_unsol_data_exit: @@ -209,37 +209,37 @@ iscsi_iser_task_xmit_unsol_data_exit: } static int -iscsi_iser_task_xmit(struct iscsi_cmd_task *ctask) +iscsi_iser_task_xmit(struct iscsi_task *task) { - struct iscsi_conn *conn = ctask->conn; - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_conn *conn = task->conn; + struct iscsi_iser_task *iser_task = task->dd_data; int error = 0; - if (!ctask->sc) - return iscsi_iser_mtask_xmit(conn, ctask); + if (!task->sc) + return iscsi_iser_mtask_xmit(conn, task); - if (ctask->sc->sc_data_direction == DMA_TO_DEVICE) { - BUG_ON(scsi_bufflen(ctask->sc) == 0); + if (task->sc->sc_data_direction == DMA_TO_DEVICE) { + BUG_ON(scsi_bufflen(task->sc) == 0); debug_scsi("cmd [itt %x total %d imm %d unsol_data %d\n", - ctask->itt, scsi_bufflen(ctask->sc), - ctask->imm_count, ctask->unsol_count); + task->itt, scsi_bufflen(task->sc), + task->imm_count, task->unsol_count); } - debug_scsi("ctask deq [cid %d itt 0x%x]\n", - conn->id, ctask->itt); + debug_scsi("task deq [cid %d itt 0x%x]\n", + conn->id, task->itt); /* Send the cmd PDU */ - if (!iser_ctask->command_sent) { - error = iser_send_command(conn, ctask); + if (!iser_task->command_sent) { + error = iser_send_command(conn, task); if (error) goto iscsi_iser_task_xmit_exit; - iser_ctask->command_sent = 1; + iser_task->command_sent = 1; } /* Send unsolicited data-out PDU(s) if necessary */ - if (ctask->unsol_count) - error = iscsi_iser_task_xmit_unsol_data(conn, ctask); + if (task->unsol_count) + error = iscsi_iser_task_xmit_unsol_data(conn, task); iscsi_iser_task_xmit_exit: if (error && error != -ENOBUFS) @@ -248,17 +248,17 @@ iscsi_iser_task_xmit(struct iscsi_cmd_task *ctask) } static void -iscsi_iser_cleanup_task(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) +iscsi_iser_cleanup_task(struct iscsi_conn *conn, struct iscsi_task *task) { - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; /* mgmt tasks do not need special cleanup */ - if (!ctask->sc) + if (!task->sc) return; - if (iser_ctask->status == ISER_TASK_STATUS_STARTED) { - iser_ctask->status = ISER_TASK_STATUS_COMPLETED; - iser_ctask_rdma_finalize(iser_ctask); + if (iser_task->status == ISER_TASK_STATUS_STARTED) { + iser_task->status = ISER_TASK_STATUS_COMPLETED; + iser_task_rdma_finalize(iser_task); } } @@ -408,8 +408,8 @@ iscsi_iser_session_create(struct Scsi_Host *shost, struct iscsi_cls_session *cls_session; struct iscsi_session *session; int i; - struct iscsi_cmd_task *ctask; - struct iscsi_iser_cmd_task *iser_ctask; + struct iscsi_task *task; + struct iscsi_iser_task *iser_task; if (shost) { printk(KERN_ERR "iscsi_tcp: invalid shost %d.\n", @@ -436,7 +436,7 @@ iscsi_iser_session_create(struct Scsi_Host *shost, */ cls_session = iscsi_session_setup(&iscsi_iser_transport, shost, ISCSI_DEF_XMIT_CMDS_MAX, - sizeof(struct iscsi_iser_cmd_task), + sizeof(struct iscsi_iser_task), initial_cmdsn); if (!cls_session) goto remove_host; @@ -445,10 +445,10 @@ iscsi_iser_session_create(struct Scsi_Host *shost, shost->can_queue = session->scsi_cmds_max; /* libiscsi setup itts, data and pool so just set desc fields */ for (i = 0; i < session->cmds_max; i++) { - ctask = session->cmds[i]; - iser_ctask = ctask->dd_data; - ctask->hdr = (struct iscsi_cmd *)&iser_ctask->desc.iscsi_header; - ctask->hdr_max = sizeof(iser_ctask->desc.iscsi_header); + task = session->cmds[i]; + iser_task = task->dd_data; + task->hdr = (struct iscsi_cmd *)&iser_task->desc.iscsi_header; + task->hdr_max = sizeof(iser_task->desc.iscsi_header); } return cls_session; diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 96a600f127c..05431f270fe 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -173,7 +173,7 @@ struct iser_data_buf { /* fwd declarations */ struct iser_device; struct iscsi_iser_conn; -struct iscsi_iser_cmd_task; +struct iscsi_iser_task; struct iser_mem_reg { u32 lkey; @@ -197,7 +197,7 @@ struct iser_regd_buf { #define MAX_REGD_BUF_VECTOR_LEN 2 struct iser_dto { - struct iscsi_iser_cmd_task *ctask; + struct iscsi_iser_task *task; struct iser_conn *ib_conn; int notify_enable; @@ -265,7 +265,7 @@ struct iscsi_iser_conn { rwlock_t lock; }; -struct iscsi_iser_cmd_task { +struct iscsi_iser_task { struct iser_desc desc; struct iscsi_iser_conn *iser_conn; enum iser_task_status status; @@ -299,13 +299,13 @@ extern int iser_debug_level; int iser_conn_set_full_featured_mode(struct iscsi_conn *conn); int iser_send_control(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask); + struct iscsi_task *task); int iser_send_command(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask); + struct iscsi_task *task); int iser_send_data_out(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask, + struct iscsi_task *task, struct iscsi_data *hdr); void iscsi_iser_recv(struct iscsi_conn *conn, @@ -326,9 +326,9 @@ void iser_rcv_completion(struct iser_desc *desc, void iser_snd_completion(struct iser_desc *desc); -void iser_ctask_rdma_init(struct iscsi_iser_cmd_task *ctask); +void iser_task_rdma_init(struct iscsi_iser_task *task); -void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *ctask); +void iser_task_rdma_finalize(struct iscsi_iser_task *task); void iser_dto_buffs_release(struct iser_dto *dto); @@ -338,10 +338,10 @@ void iser_reg_single(struct iser_device *device, struct iser_regd_buf *regd_buf, enum dma_data_direction direction); -void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *ctask, +void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *task, enum iser_data_dir cmd_dir); -int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *ctask, +int iser_reg_rdma_mem(struct iscsi_iser_task *task, enum iser_data_dir cmd_dir); int iser_connect(struct iser_conn *ib_conn, @@ -361,10 +361,10 @@ int iser_post_send(struct iser_desc *tx_desc); int iser_conn_state_comp(struct iser_conn *ib_conn, enum iser_ib_conn_state comp); -int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask, +int iser_dma_map_task_data(struct iscsi_iser_task *iser_task, struct iser_data_buf *data, enum iser_data_dir iser_dir, enum dma_data_direction dma_dir); -void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask); +void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task); #endif diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 4ea78fbeee9..35af60a23c6 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -66,46 +66,46 @@ static void iser_dto_add_regd_buff(struct iser_dto *dto, /* Register user buffer memory and initialize passive rdma * dto descriptor. Total data size is stored in - * iser_ctask->data[ISER_DIR_IN].data_len + * iser_task->data[ISER_DIR_IN].data_len */ -static int iser_prepare_read_cmd(struct iscsi_cmd_task *ctask, +static int iser_prepare_read_cmd(struct iscsi_task *task, unsigned int edtl) { - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; struct iser_regd_buf *regd_buf; int err; - struct iser_hdr *hdr = &iser_ctask->desc.iser_header; - struct iser_data_buf *buf_in = &iser_ctask->data[ISER_DIR_IN]; + struct iser_hdr *hdr = &iser_task->desc.iser_header; + struct iser_data_buf *buf_in = &iser_task->data[ISER_DIR_IN]; - err = iser_dma_map_task_data(iser_ctask, + err = iser_dma_map_task_data(iser_task, buf_in, ISER_DIR_IN, DMA_FROM_DEVICE); if (err) return err; - if (edtl > iser_ctask->data[ISER_DIR_IN].data_len) { + if (edtl > iser_task->data[ISER_DIR_IN].data_len) { iser_err("Total data length: %ld, less than EDTL: " "%d, in READ cmd BHS itt: %d, conn: 0x%p\n", - iser_ctask->data[ISER_DIR_IN].data_len, edtl, - ctask->itt, iser_ctask->iser_conn); + iser_task->data[ISER_DIR_IN].data_len, edtl, + task->itt, iser_task->iser_conn); return -EINVAL; } - err = iser_reg_rdma_mem(iser_ctask,ISER_DIR_IN); + err = iser_reg_rdma_mem(iser_task,ISER_DIR_IN); if (err) { iser_err("Failed to set up Data-IN RDMA\n"); return err; } - regd_buf = &iser_ctask->rdma_regd[ISER_DIR_IN]; + regd_buf = &iser_task->rdma_regd[ISER_DIR_IN]; hdr->flags |= ISER_RSV; hdr->read_stag = cpu_to_be32(regd_buf->reg.rkey); hdr->read_va = cpu_to_be64(regd_buf->reg.va); iser_dbg("Cmd itt:%d READ tags RKEY:%#.4X VA:%#llX\n", - ctask->itt, regd_buf->reg.rkey, + task->itt, regd_buf->reg.rkey, (unsigned long long)regd_buf->reg.va); return 0; @@ -113,43 +113,43 @@ static int iser_prepare_read_cmd(struct iscsi_cmd_task *ctask, /* Register user buffer memory and initialize passive rdma * dto descriptor. Total data size is stored in - * ctask->data[ISER_DIR_OUT].data_len + * task->data[ISER_DIR_OUT].data_len */ static int -iser_prepare_write_cmd(struct iscsi_cmd_task *ctask, +iser_prepare_write_cmd(struct iscsi_task *task, unsigned int imm_sz, unsigned int unsol_sz, unsigned int edtl) { - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; struct iser_regd_buf *regd_buf; int err; - struct iser_dto *send_dto = &iser_ctask->desc.dto; - struct iser_hdr *hdr = &iser_ctask->desc.iser_header; - struct iser_data_buf *buf_out = &iser_ctask->data[ISER_DIR_OUT]; + struct iser_dto *send_dto = &iser_task->desc.dto; + struct iser_hdr *hdr = &iser_task->desc.iser_header; + struct iser_data_buf *buf_out = &iser_task->data[ISER_DIR_OUT]; - err = iser_dma_map_task_data(iser_ctask, + err = iser_dma_map_task_data(iser_task, buf_out, ISER_DIR_OUT, DMA_TO_DEVICE); if (err) return err; - if (edtl > iser_ctask->data[ISER_DIR_OUT].data_len) { + if (edtl > iser_task->data[ISER_DIR_OUT].data_len) { iser_err("Total data length: %ld, less than EDTL: %d, " "in WRITE cmd BHS itt: %d, conn: 0x%p\n", - iser_ctask->data[ISER_DIR_OUT].data_len, - edtl, ctask->itt, ctask->conn); + iser_task->data[ISER_DIR_OUT].data_len, + edtl, task->itt, task->conn); return -EINVAL; } - err = iser_reg_rdma_mem(iser_ctask,ISER_DIR_OUT); + err = iser_reg_rdma_mem(iser_task,ISER_DIR_OUT); if (err != 0) { iser_err("Failed to register write cmd RDMA mem\n"); return err; } - regd_buf = &iser_ctask->rdma_regd[ISER_DIR_OUT]; + regd_buf = &iser_task->rdma_regd[ISER_DIR_OUT]; if (unsol_sz < edtl) { hdr->flags |= ISER_WSV; @@ -158,13 +158,13 @@ iser_prepare_write_cmd(struct iscsi_cmd_task *ctask, iser_dbg("Cmd itt:%d, WRITE tags, RKEY:%#.4X " "VA:%#llX + unsol:%d\n", - ctask->itt, regd_buf->reg.rkey, + task->itt, regd_buf->reg.rkey, (unsigned long long)regd_buf->reg.va, unsol_sz); } if (imm_sz > 0) { iser_dbg("Cmd itt:%d, WRITE, adding imm.data sz: %d\n", - ctask->itt, imm_sz); + task->itt, imm_sz); iser_dto_add_regd_buff(send_dto, regd_buf, 0, @@ -300,13 +300,13 @@ int iser_conn_set_full_featured_mode(struct iscsi_conn *conn) } static int -iser_check_xmit(struct iscsi_conn *conn, void *ctask) +iser_check_xmit(struct iscsi_conn *conn, void *task) { struct iscsi_iser_conn *iser_conn = conn->dd_data; if (atomic_read(&iser_conn->ib_conn->post_send_buf_count) == ISER_QP_MAX_REQ_DTOS) { - iser_dbg("%ld can't xmit ctask %p\n",jiffies,ctask); + iser_dbg("%ld can't xmit task %p\n",jiffies,task); return -ENOBUFS; } return 0; @@ -317,37 +317,37 @@ iser_check_xmit(struct iscsi_conn *conn, void *ctask) * iser_send_command - send command PDU */ int iser_send_command(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask) + struct iscsi_task *task) { struct iscsi_iser_conn *iser_conn = conn->dd_data; - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; struct iser_dto *send_dto = NULL; unsigned long edtl; int err = 0; struct iser_data_buf *data_buf; - struct iscsi_cmd *hdr = ctask->hdr; - struct scsi_cmnd *sc = ctask->sc; + struct iscsi_cmd *hdr = task->hdr; + struct scsi_cmnd *sc = task->sc; if (!iser_conn_state_comp(iser_conn->ib_conn, ISER_CONN_UP)) { iser_err("Failed to send, conn: 0x%p is not up\n", iser_conn->ib_conn); return -EPERM; } - if (iser_check_xmit(conn, ctask)) + if (iser_check_xmit(conn, task)) return -ENOBUFS; edtl = ntohl(hdr->data_length); /* build the tx desc regd header and add it to the tx desc dto */ - iser_ctask->desc.type = ISCSI_TX_SCSI_COMMAND; - send_dto = &iser_ctask->desc.dto; - send_dto->ctask = iser_ctask; - iser_create_send_desc(iser_conn, &iser_ctask->desc); + iser_task->desc.type = ISCSI_TX_SCSI_COMMAND; + send_dto = &iser_task->desc.dto; + send_dto->task = iser_task; + iser_create_send_desc(iser_conn, &iser_task->desc); if (hdr->flags & ISCSI_FLAG_CMD_READ) - data_buf = &iser_ctask->data[ISER_DIR_IN]; + data_buf = &iser_task->data[ISER_DIR_IN]; else - data_buf = &iser_ctask->data[ISER_DIR_OUT]; + data_buf = &iser_task->data[ISER_DIR_OUT]; if (scsi_sg_count(sc)) { /* using a scatter list */ data_buf->buf = scsi_sglist(sc); @@ -357,15 +357,15 @@ int iser_send_command(struct iscsi_conn *conn, data_buf->data_len = scsi_bufflen(sc); if (hdr->flags & ISCSI_FLAG_CMD_READ) { - err = iser_prepare_read_cmd(ctask, edtl); + err = iser_prepare_read_cmd(task, edtl); if (err) goto send_command_error; } if (hdr->flags & ISCSI_FLAG_CMD_WRITE) { - err = iser_prepare_write_cmd(ctask, - ctask->imm_count, - ctask->imm_count + - ctask->unsol_count, + err = iser_prepare_write_cmd(task, + task->imm_count, + task->imm_count + + task->unsol_count, edtl); if (err) goto send_command_error; @@ -380,15 +380,15 @@ int iser_send_command(struct iscsi_conn *conn, goto send_command_error; } - iser_ctask->status = ISER_TASK_STATUS_STARTED; + iser_task->status = ISER_TASK_STATUS_STARTED; - err = iser_post_send(&iser_ctask->desc); + err = iser_post_send(&iser_task->desc); if (!err) return 0; send_command_error: iser_dto_buffs_release(send_dto); - iser_err("conn %p failed ctask->itt %d err %d\n",conn, ctask->itt, err); + iser_err("conn %p failed task->itt %d err %d\n",conn, task->itt, err); return err; } @@ -396,11 +396,11 @@ send_command_error: * iser_send_data_out - send data out PDU */ int iser_send_data_out(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask, + struct iscsi_task *task, struct iscsi_data *hdr) { struct iscsi_iser_conn *iser_conn = conn->dd_data; - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; + struct iscsi_iser_task *iser_task = task->dd_data; struct iser_desc *tx_desc = NULL; struct iser_dto *send_dto = NULL; unsigned long buf_offset; @@ -413,7 +413,7 @@ int iser_send_data_out(struct iscsi_conn *conn, return -EPERM; } - if (iser_check_xmit(conn, ctask)) + if (iser_check_xmit(conn, task)) return -ENOBUFS; itt = (__force uint32_t)hdr->itt; @@ -434,7 +434,7 @@ int iser_send_data_out(struct iscsi_conn *conn, /* build the tx desc regd header and add it to the tx desc dto */ send_dto = &tx_desc->dto; - send_dto->ctask = iser_ctask; + send_dto->task = iser_task; iser_create_send_desc(iser_conn, tx_desc); iser_reg_single(iser_conn->ib_conn->device, @@ -442,15 +442,15 @@ int iser_send_data_out(struct iscsi_conn *conn, /* all data was registered for RDMA, we can use the lkey */ iser_dto_add_regd_buff(send_dto, - &iser_ctask->rdma_regd[ISER_DIR_OUT], + &iser_task->rdma_regd[ISER_DIR_OUT], buf_offset, data_seg_len); - if (buf_offset + data_seg_len > iser_ctask->data[ISER_DIR_OUT].data_len) { + if (buf_offset + data_seg_len > iser_task->data[ISER_DIR_OUT].data_len) { iser_err("Offset:%ld & DSL:%ld in Data-Out " "inconsistent with total len:%ld, itt:%d\n", buf_offset, data_seg_len, - iser_ctask->data[ISER_DIR_OUT].data_len, itt); + iser_task->data[ISER_DIR_OUT].data_len, itt); err = -EINVAL; goto send_data_out_error; } @@ -470,11 +470,11 @@ send_data_out_error: } int iser_send_control(struct iscsi_conn *conn, - struct iscsi_cmd_task *ctask) + struct iscsi_task *task) { struct iscsi_iser_conn *iser_conn = conn->dd_data; - struct iscsi_iser_cmd_task *iser_ctask = ctask->dd_data; - struct iser_desc *mdesc = &iser_ctask->desc; + struct iscsi_iser_task *iser_task = task->dd_data; + struct iser_desc *mdesc = &iser_task->desc; struct iser_dto *send_dto = NULL; unsigned long data_seg_len; int err = 0; @@ -486,27 +486,27 @@ int iser_send_control(struct iscsi_conn *conn, return -EPERM; } - if (iser_check_xmit(conn, ctask)) + if (iser_check_xmit(conn, task)) return -ENOBUFS; /* build the tx desc regd header and add it to the tx desc dto */ mdesc->type = ISCSI_TX_CONTROL; send_dto = &mdesc->dto; - send_dto->ctask = NULL; + send_dto->task = NULL; iser_create_send_desc(iser_conn, mdesc); device = iser_conn->ib_conn->device; iser_reg_single(device, send_dto->regd[0], DMA_TO_DEVICE); - data_seg_len = ntoh24(ctask->hdr->dlength); + data_seg_len = ntoh24(task->hdr->dlength); if (data_seg_len > 0) { regd_buf = &mdesc->data_regd_buf; memset(regd_buf, 0, sizeof(struct iser_regd_buf)); regd_buf->device = device; - regd_buf->virt_addr = ctask->data; - regd_buf->data_size = ctask->data_count; + regd_buf->virt_addr = task->data; + regd_buf->data_size = task->data_count; iser_reg_single(device, regd_buf, DMA_TO_DEVICE); iser_dto_add_regd_buff(send_dto, regd_buf, @@ -538,8 +538,8 @@ void iser_rcv_completion(struct iser_desc *rx_desc, { struct iser_dto *dto = &rx_desc->dto; struct iscsi_iser_conn *conn = dto->ib_conn->iser_conn; - struct iscsi_cmd_task *ctask; - struct iscsi_iser_cmd_task *iser_ctask; + struct iscsi_task *task; + struct iscsi_iser_task *iser_task; struct iscsi_hdr *hdr; char *rx_data = NULL; int rx_data_len = 0; @@ -558,16 +558,16 @@ void iser_rcv_completion(struct iser_desc *rx_desc, opcode = hdr->opcode & ISCSI_OPCODE_MASK; if (opcode == ISCSI_OP_SCSI_CMD_RSP) { - ctask = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt); - if (!ctask) - iser_err("itt can't be matched to ctask!!! " + task = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt); + if (!task) + iser_err("itt can't be matched to task!!! " "conn %p opcode %d itt %d\n", conn->iscsi_conn, opcode, hdr->itt); else { - iser_ctask = ctask->dd_data; - iser_dbg("itt %d ctask %p\n",hdr->itt, ctask); - iser_ctask->status = ISER_TASK_STATUS_COMPLETED; - iser_ctask_rdma_finalize(iser_ctask); + iser_task = task->dd_data; + iser_dbg("itt %d task %p\n",hdr->itt, task); + iser_task->status = ISER_TASK_STATUS_COMPLETED; + iser_task_rdma_finalize(iser_task); } } iser_dto_buffs_release(dto); @@ -578,7 +578,7 @@ void iser_rcv_completion(struct iser_desc *rx_desc, kmem_cache_free(ig.desc_cache, rx_desc); /* decrementing conn->post_recv_buf_count only --after-- freeing the * - * ctask eliminates the need to worry on ctasks which are completed in * + * task eliminates the need to worry on tasks which are completed in * * parallel to the execution of iser_conn_term. So the code that waits * * for the posted rx bufs refcount to become zero handles everything */ atomic_dec(&conn->ib_conn->post_recv_buf_count); @@ -590,7 +590,7 @@ void iser_snd_completion(struct iser_desc *tx_desc) struct iser_conn *ib_conn = dto->ib_conn; struct iscsi_iser_conn *iser_conn = ib_conn->iser_conn; struct iscsi_conn *conn = iser_conn->iscsi_conn; - struct iscsi_cmd_task *ctask; + struct iscsi_task *task; int resume_tx = 0; iser_dbg("Initiator, Data sent dto=0x%p\n", dto); @@ -613,31 +613,31 @@ void iser_snd_completion(struct iser_desc *tx_desc) if (tx_desc->type == ISCSI_TX_CONTROL) { /* this arithmetic is legal by libiscsi dd_data allocation */ - ctask = (void *) ((long)(void *)tx_desc - - sizeof(struct iscsi_cmd_task)); - if (ctask->hdr->itt == RESERVED_ITT) - iscsi_put_ctask(ctask); + task = (void *) ((long)(void *)tx_desc - + sizeof(struct iscsi_task)); + if (task->hdr->itt == RESERVED_ITT) + iscsi_put_task(task); } } -void iser_ctask_rdma_init(struct iscsi_iser_cmd_task *iser_ctask) +void iser_task_rdma_init(struct iscsi_iser_task *iser_task) { - iser_ctask->status = ISER_TASK_STATUS_INIT; + iser_task->status = ISER_TASK_STATUS_INIT; - iser_ctask->dir[ISER_DIR_IN] = 0; - iser_ctask->dir[ISER_DIR_OUT] = 0; + iser_task->dir[ISER_DIR_IN] = 0; + iser_task->dir[ISER_DIR_OUT] = 0; - iser_ctask->data[ISER_DIR_IN].data_len = 0; - iser_ctask->data[ISER_DIR_OUT].data_len = 0; + iser_task->data[ISER_DIR_IN].data_len = 0; + iser_task->data[ISER_DIR_OUT].data_len = 0; - memset(&iser_ctask->rdma_regd[ISER_DIR_IN], 0, + memset(&iser_task->rdma_regd[ISER_DIR_IN], 0, sizeof(struct iser_regd_buf)); - memset(&iser_ctask->rdma_regd[ISER_DIR_OUT], 0, + memset(&iser_task->rdma_regd[ISER_DIR_OUT], 0, sizeof(struct iser_regd_buf)); } -void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask) +void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task) { int deferred; int is_rdma_aligned = 1; @@ -646,17 +646,17 @@ void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask) /* if we were reading, copy back to unaligned sglist, * anyway dma_unmap and free the copy */ - if (iser_ctask->data_copy[ISER_DIR_IN].copy_buf != NULL) { + if (iser_task->data_copy[ISER_DIR_IN].copy_buf != NULL) { is_rdma_aligned = 0; - iser_finalize_rdma_unaligned_sg(iser_ctask, ISER_DIR_IN); + iser_finalize_rdma_unaligned_sg(iser_task, ISER_DIR_IN); } - if (iser_ctask->data_copy[ISER_DIR_OUT].copy_buf != NULL) { + if (iser_task->data_copy[ISER_DIR_OUT].copy_buf != NULL) { is_rdma_aligned = 0; - iser_finalize_rdma_unaligned_sg(iser_ctask, ISER_DIR_OUT); + iser_finalize_rdma_unaligned_sg(iser_task, ISER_DIR_OUT); } - if (iser_ctask->dir[ISER_DIR_IN]) { - regd = &iser_ctask->rdma_regd[ISER_DIR_IN]; + if (iser_task->dir[ISER_DIR_IN]) { + regd = &iser_task->rdma_regd[ISER_DIR_IN]; deferred = iser_regd_buff_release(regd); if (deferred) { iser_err("%d references remain for BUF-IN rdma reg\n", @@ -664,8 +664,8 @@ void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask) } } - if (iser_ctask->dir[ISER_DIR_OUT]) { - regd = &iser_ctask->rdma_regd[ISER_DIR_OUT]; + if (iser_task->dir[ISER_DIR_OUT]) { + regd = &iser_task->rdma_regd[ISER_DIR_OUT]; deferred = iser_regd_buff_release(regd); if (deferred) { iser_err("%d references remain for BUF-OUT rdma reg\n", @@ -675,7 +675,7 @@ void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask) /* if the data was unaligned, it was already unmapped and then copied */ if (is_rdma_aligned) - iser_dma_unmap_task_data(iser_ctask); + iser_dma_unmap_task_data(iser_task); } void iser_dto_buffs_release(struct iser_dto *dto) diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index cac50c4dc15..48f2a601fc2 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -101,13 +101,13 @@ void iser_reg_single(struct iser_device *device, /** * iser_start_rdma_unaligned_sg */ -static int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask, +static int iser_start_rdma_unaligned_sg(struct iscsi_iser_task *iser_task, enum iser_data_dir cmd_dir) { int dma_nents; struct ib_device *dev; char *mem = NULL; - struct iser_data_buf *data = &iser_ctask->data[cmd_dir]; + struct iser_data_buf *data = &iser_task->data[cmd_dir]; unsigned long cmd_data_len = data->data_len; if (cmd_data_len > ISER_KMALLOC_THRESHOLD) @@ -140,37 +140,37 @@ static int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask, } } - sg_init_one(&iser_ctask->data_copy[cmd_dir].sg_single, mem, cmd_data_len); - iser_ctask->data_copy[cmd_dir].buf = - &iser_ctask->data_copy[cmd_dir].sg_single; - iser_ctask->data_copy[cmd_dir].size = 1; + sg_init_one(&iser_task->data_copy[cmd_dir].sg_single, mem, cmd_data_len); + iser_task->data_copy[cmd_dir].buf = + &iser_task->data_copy[cmd_dir].sg_single; + iser_task->data_copy[cmd_dir].size = 1; - iser_ctask->data_copy[cmd_dir].copy_buf = mem; + iser_task->data_copy[cmd_dir].copy_buf = mem; - dev = iser_ctask->iser_conn->ib_conn->device->ib_device; + dev = iser_task->iser_conn->ib_conn->device->ib_device; dma_nents = ib_dma_map_sg(dev, - &iser_ctask->data_copy[cmd_dir].sg_single, + &iser_task->data_copy[cmd_dir].sg_single, 1, (cmd_dir == ISER_DIR_OUT) ? DMA_TO_DEVICE : DMA_FROM_DEVICE); BUG_ON(dma_nents == 0); - iser_ctask->data_copy[cmd_dir].dma_nents = dma_nents; + iser_task->data_copy[cmd_dir].dma_nents = dma_nents; return 0; } /** * iser_finalize_rdma_unaligned_sg */ -void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask, +void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *iser_task, enum iser_data_dir cmd_dir) { struct ib_device *dev; struct iser_data_buf *mem_copy; unsigned long cmd_data_len; - dev = iser_ctask->iser_conn->ib_conn->device->ib_device; - mem_copy = &iser_ctask->data_copy[cmd_dir]; + dev = iser_task->iser_conn->ib_conn->device->ib_device; + mem_copy = &iser_task->data_copy[cmd_dir]; ib_dma_unmap_sg(dev, &mem_copy->sg_single, 1, (cmd_dir == ISER_DIR_OUT) ? @@ -186,8 +186,8 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask, /* copy back read RDMA to unaligned sg */ mem = mem_copy->copy_buf; - sgl = (struct scatterlist *)iser_ctask->data[ISER_DIR_IN].buf; - sg_size = iser_ctask->data[ISER_DIR_IN].size; + sgl = (struct scatterlist *)iser_task->data[ISER_DIR_IN].buf; + sg_size = iser_task->data[ISER_DIR_IN].size; p = mem; for_each_sg(sgl, sg, sg_size, i) { @@ -200,7 +200,7 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask, } } - cmd_data_len = iser_ctask->data[cmd_dir].data_len; + cmd_data_len = iser_task->data[cmd_dir].data_len; if (cmd_data_len > ISER_KMALLOC_THRESHOLD) free_pages((unsigned long)mem_copy->copy_buf, @@ -378,15 +378,15 @@ static void iser_page_vec_build(struct iser_data_buf *data, } } -int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask, - struct iser_data_buf *data, - enum iser_data_dir iser_dir, - enum dma_data_direction dma_dir) +int iser_dma_map_task_data(struct iscsi_iser_task *iser_task, + struct iser_data_buf *data, + enum iser_data_dir iser_dir, + enum dma_data_direction dma_dir) { struct ib_device *dev; - iser_ctask->dir[iser_dir] = 1; - dev = iser_ctask->iser_conn->ib_conn->device->ib_device; + iser_task->dir[iser_dir] = 1; + dev = iser_task->iser_conn->ib_conn->device->ib_device; data->dma_nents = ib_dma_map_sg(dev, data->buf, data->size, dma_dir); if (data->dma_nents == 0) { @@ -396,20 +396,20 @@ int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask, return 0; } -void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask) +void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task) { struct ib_device *dev; struct iser_data_buf *data; - dev = iser_ctask->iser_conn->ib_conn->device->ib_device; + dev = iser_task->iser_conn->ib_conn->device->ib_device; - if (iser_ctask->dir[ISER_DIR_IN]) { - data = &iser_ctask->data[ISER_DIR_IN]; + if (iser_task->dir[ISER_DIR_IN]) { + data = &iser_task->data[ISER_DIR_IN]; ib_dma_unmap_sg(dev, data->buf, data->size, DMA_FROM_DEVICE); } - if (iser_ctask->dir[ISER_DIR_OUT]) { - data = &iser_ctask->data[ISER_DIR_OUT]; + if (iser_task->dir[ISER_DIR_OUT]) { + data = &iser_task->data[ISER_DIR_OUT]; ib_dma_unmap_sg(dev, data->buf, data->size, DMA_TO_DEVICE); } } @@ -420,21 +420,21 @@ void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask) * * returns 0 on success, errno code on failure */ -int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask, +int iser_reg_rdma_mem(struct iscsi_iser_task *iser_task, enum iser_data_dir cmd_dir) { - struct iscsi_conn *iscsi_conn = iser_ctask->iser_conn->iscsi_conn; - struct iser_conn *ib_conn = iser_ctask->iser_conn->ib_conn; + struct iscsi_conn *iscsi_conn = iser_task->iser_conn->iscsi_conn; + struct iser_conn *ib_conn = iser_task->iser_conn->ib_conn; struct iser_device *device = ib_conn->device; struct ib_device *ibdev = device->ib_device; - struct iser_data_buf *mem = &iser_ctask->data[cmd_dir]; + struct iser_data_buf *mem = &iser_task->data[cmd_dir]; struct iser_regd_buf *regd_buf; int aligned_len; int err; int i; struct scatterlist *sg; - regd_buf = &iser_ctask->rdma_regd[cmd_dir]; + regd_buf = &iser_task->rdma_regd[cmd_dir]; aligned_len = iser_data_buf_aligned_len(mem, ibdev); if (aligned_len != mem->dma_nents) { @@ -444,13 +444,13 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask, iser_data_buf_dump(mem, ibdev); /* unmap the command data before accessing it */ - iser_dma_unmap_task_data(iser_ctask); + iser_dma_unmap_task_data(iser_task); /* allocate copy buf, if we are writing, copy the */ /* unaligned scatterlist, dma map the copy */ - if (iser_start_rdma_unaligned_sg(iser_ctask, cmd_dir) != 0) + if (iser_start_rdma_unaligned_sg(iser_task, cmd_dir) != 0) return -ENOMEM; - mem = &iser_ctask->data_copy[cmd_dir]; + mem = &iser_task->data_copy[cmd_dir]; } /* if there a single dma entry, FMR is not needed */ @@ -474,8 +474,9 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask, err = iser_reg_page_vec(ib_conn, ib_conn->page_vec, ®d_buf->reg); if (err) { iser_data_buf_dump(mem, ibdev); - iser_err("mem->dma_nents = %d (dlength = 0x%x)\n", mem->dma_nents, - ntoh24(iser_ctask->desc.iscsi_header.dlength)); + iser_err("mem->dma_nents = %d (dlength = 0x%x)\n", + mem->dma_nents, + ntoh24(iser_task->desc.iscsi_header.dlength)); iser_err("page_vec: data_size = 0x%x, length = %d, offset = 0x%x\n", ib_conn->page_vec->data_size, ib_conn->page_vec->length, ib_conn->page_vec->offset); -- cgit v1.2.3 From 7970634b81a6e3561954517bca42615542c4535b Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:12 -0500 Subject: [SCSI] iscsi class: user device_for_each_child instead of duplicating session list Currently we duplicate the list of sessions, because we were using the test for if a session was on the host list to indicate if the session was bound or unbound. We can instead use the target_id and fix up the class so that drivers like bnx2i do not have to manage the target id space. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 86d9c42f0d3..3a89039e9a9 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -437,7 +437,7 @@ iscsi_iser_session_create(struct Scsi_Host *shost, cls_session = iscsi_session_setup(&iscsi_iser_transport, shost, ISCSI_DEF_XMIT_CMDS_MAX, sizeof(struct iscsi_iser_task), - initial_cmdsn); + initial_cmdsn, 0); if (!cls_session) goto remove_host; session = cls_session->dd_data; -- cgit v1.2.3 From 412eeafa0a51a8d86545d0be637bf84e4374fccf Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:14 -0500 Subject: [SCSI] iser: Modify iser to take a iscsi_endpoint struct in ep callouts and session setup This hooks iser into the iscsi endpoint code. Previously it handled the lookup and allocation. This has been made generic so bnx2i and iser can share it. It also allows us to pass iser the leading conn's ep, so we know the ib_deivce being used and can set it as the scsi_host's parent. And that allows scsi-ml to set the dma_mask based on those values. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 87 ++++++++++++++------------------ drivers/infiniband/ulp/iser/iscsi_iser.h | 4 +- drivers/infiniband/ulp/iser/iser_verbs.c | 14 +---- 3 files changed, 43 insertions(+), 62 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 3a89039e9a9..42e95b83309 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -262,24 +262,6 @@ iscsi_iser_cleanup_task(struct iscsi_conn *conn, struct iscsi_task *task) } } -static struct iser_conn * -iscsi_iser_ib_conn_lookup(__u64 ep_handle) -{ - struct iser_conn *ib_conn; - struct iser_conn *uib_conn = (struct iser_conn *)(unsigned long)ep_handle; - - mutex_lock(&ig.connlist_mutex); - list_for_each_entry(ib_conn, &ig.connlist, conn_list) { - if (ib_conn == uib_conn) { - mutex_unlock(&ig.connlist_mutex); - return ib_conn; - } - } - mutex_unlock(&ig.connlist_mutex); - iser_err("no conn exists for eph %llx\n",(unsigned long long)ep_handle); - return NULL; -} - static struct iscsi_cls_conn * iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx) { @@ -335,6 +317,7 @@ iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session, struct iscsi_conn *conn = cls_conn->dd_data; struct iscsi_iser_conn *iser_conn; struct iser_conn *ib_conn; + struct iscsi_endpoint *ep; int error; error = iscsi_conn_bind(cls_session, cls_conn, is_leading); @@ -343,12 +326,14 @@ iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session, /* the transport ep handle comes from user space so it must be * verified against the global ib connections list */ - ib_conn = iscsi_iser_ib_conn_lookup(transport_eph); - if (!ib_conn) { + ep = iscsi_lookup_endpoint(transport_eph); + if (!ep) { iser_err("can't bind eph %llx\n", (unsigned long long)transport_eph); return -EINVAL; } + ib_conn = ep->dd_data; + /* binds the iSER connection retrieved from the previously * connected ep_handle to the iSCSI layer connection. exchanges * connection pointers */ @@ -401,21 +386,17 @@ static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session) } static struct iscsi_cls_session * -iscsi_iser_session_create(struct Scsi_Host *shost, +iscsi_iser_session_create(struct iscsi_endpoint *ep, uint16_t cmds_max, uint16_t qdepth, uint32_t initial_cmdsn, uint32_t *hostno) { struct iscsi_cls_session *cls_session; struct iscsi_session *session; + struct Scsi_Host *shost; int i; struct iscsi_task *task; struct iscsi_iser_task *iser_task; - - if (shost) { - printk(KERN_ERR "iscsi_tcp: invalid shost %d.\n", - shost->host_no); - return NULL; - } + struct iser_conn *ib_conn; shost = iscsi_host_alloc(&iscsi_iser_sht, 0, ISCSI_MAX_CMD_PER_LUN); if (!shost) @@ -426,7 +407,15 @@ iscsi_iser_session_create(struct Scsi_Host *shost, shost->max_channel = 0; shost->max_cmd_len = 16; - if (iscsi_host_add(shost, NULL)) + /* + * older userspace tools (before 2.0-870) did not pass us + * the leading conn's ep so this will be NULL; + */ + if (ep) + ib_conn = ep->dd_data; + + if (iscsi_host_add(shost, + ep ? ib_conn->device->ib_device->dma_device : NULL)) goto free_host; *hostno = shost->host_no; @@ -529,34 +518,37 @@ iscsi_iser_conn_get_stats(struct iscsi_cls_conn *cls_conn, struct iscsi_stats *s stats->custom[3].value = conn->fmr_unalign_cnt; } -static int -iscsi_iser_ep_connect(struct sockaddr *dst_addr, int non_blocking, - __u64 *ep_handle) +static struct iscsi_endpoint * +iscsi_iser_ep_connect(struct sockaddr *dst_addr, int non_blocking) { int err; struct iser_conn *ib_conn; + struct iscsi_endpoint *ep; - err = iser_conn_init(&ib_conn); - if (err) - goto out; + ep = iscsi_create_endpoint(sizeof(*ib_conn)); + if (!ep) + return ERR_PTR(-ENOMEM); - err = iser_connect(ib_conn, NULL, (struct sockaddr_in *)dst_addr, non_blocking); - if (!err) - *ep_handle = (__u64)(unsigned long)ib_conn; + ib_conn = ep->dd_data; + ib_conn->ep = ep; + iser_conn_init(ib_conn); -out: - return err; + err = iser_connect(ib_conn, NULL, (struct sockaddr_in *)dst_addr, + non_blocking); + if (err) { + iscsi_destroy_endpoint(ep); + return ERR_PTR(err); + } + return ep; } static int -iscsi_iser_ep_poll(__u64 ep_handle, int timeout_ms) +iscsi_iser_ep_poll(struct iscsi_endpoint *ep, int timeout_ms) { - struct iser_conn *ib_conn = iscsi_iser_ib_conn_lookup(ep_handle); + struct iser_conn *ib_conn; int rc; - if (!ib_conn) - return -EINVAL; - + ib_conn = ep->dd_data; rc = wait_event_interruptible_timeout(ib_conn->wait, ib_conn->state == ISER_CONN_UP, msecs_to_jiffies(timeout_ms)); @@ -578,14 +570,11 @@ iscsi_iser_ep_poll(__u64 ep_handle, int timeout_ms) } static void -iscsi_iser_ep_disconnect(__u64 ep_handle) +iscsi_iser_ep_disconnect(struct iscsi_endpoint *ep) { struct iser_conn *ib_conn; - ib_conn = iscsi_iser_ib_conn_lookup(ep_handle); - if (!ib_conn) - return; - + ib_conn = ep->dd_data; if (ib_conn->iser_conn) /* * Must suspend xmit path if the ep is bound to the diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 05431f270fe..cdf48763b08 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -174,6 +174,7 @@ struct iser_data_buf { struct iser_device; struct iscsi_iser_conn; struct iscsi_iser_task; +struct iscsi_endpoint; struct iser_mem_reg { u32 lkey; @@ -241,6 +242,7 @@ struct iser_device { struct iser_conn { struct iscsi_iser_conn *iser_conn; /* iser conn for upcalls */ + struct iscsi_endpoint *ep; enum iser_ib_conn_state state; /* rdma connection state */ atomic_t refcount; spinlock_t lock; /* used for state changes */ @@ -313,7 +315,7 @@ void iscsi_iser_recv(struct iscsi_conn *conn, char *rx_data, int rx_data_len); -int iser_conn_init(struct iser_conn **ib_conn); +void iser_conn_init(struct iser_conn *ib_conn); void iser_conn_get(struct iser_conn *ib_conn); diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 5daed2bd710..81b45d4d9aa 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -325,7 +325,7 @@ static void iser_conn_release(struct iser_conn *ib_conn) iser_device_try_release(device); if (ib_conn->iser_conn) ib_conn->iser_conn->ib_conn = NULL; - kfree(ib_conn); + iscsi_destroy_endpoint(ib_conn->ep); } void iser_conn_get(struct iser_conn *ib_conn) @@ -494,15 +494,8 @@ static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *eve return ret; } -int iser_conn_init(struct iser_conn **ibconn) +void iser_conn_init(struct iser_conn *ib_conn) { - struct iser_conn *ib_conn; - - ib_conn = kzalloc(sizeof *ib_conn, GFP_KERNEL); - if (!ib_conn) { - iser_err("can't alloc memory for struct iser_conn\n"); - return -ENOMEM; - } ib_conn->state = ISER_CONN_INIT; init_waitqueue_head(&ib_conn->wait); atomic_set(&ib_conn->post_recv_buf_count, 0); @@ -510,9 +503,6 @@ int iser_conn_init(struct iser_conn **ibconn) atomic_set(&ib_conn->refcount, 1); INIT_LIST_HEAD(&ib_conn->conn_list); spin_lock_init(&ib_conn->lock); - - *ibconn = ib_conn; - return 0; } /** -- cgit v1.2.3 From 88dfd340b9dece8fcaa1a2d4c782338926c017f7 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:16 -0500 Subject: [SCSI] iscsi class: Add session initiatorname and ifacename sysfs attrs. This adds two new attrs used for creating initiator ports and binding sessions to hardware. The session level initiatorname: Since bnx2i does a scsi_host per host device, we need to add the iface initiator port settings on the session, so we can create multiple initiator ports (each with different inames) per device/scsi_host. The current iname reflects that qla4xxx can have one iname per hba, and we are allocating a host per session for software. The iname on the host will remain so we can export and set the hba level qla4xxx setting. The ifacename attr: To bind a session to a some peice of hardware in userspace we maintain some mappings, but during boot or iscsid restart (iscsid contains the user space part of the driver) we need to be able to figure out which of those host mappings abstractions maps to certain sessions. This patch adds a ifacename attr, which userspace can set to id the host side of the endpoint across pivot_roots and iscsid restarts. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 42e95b83309..08edbaf8922 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -629,7 +629,8 @@ static struct iscsi_transport iscsi_iser_transport = { ISCSI_USERNAME | ISCSI_PASSWORD | ISCSI_USERNAME_IN | ISCSI_PASSWORD_IN | ISCSI_FAST_ABORT | ISCSI_ABORT_TMO | - ISCSI_PING_TMO | ISCSI_RECV_TMO, + ISCSI_PING_TMO | ISCSI_RECV_TMO | + ISCSI_IFACE_NAME | ISCSI_INITIATOR_NAME, .host_param_mask = ISCSI_HOST_HWADDRESS | ISCSI_HOST_NETDEV_NAME | ISCSI_HOST_INITIATOR_NAME, -- cgit v1.2.3 From 913e5bf435617aa529919a4f7567f849f9f35f9f Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 21 May 2008 15:54:18 -0500 Subject: [SCSI] libiscsi, iser, tcp: remove recv_lock The recv lock was defined so the iscsi layer could block the recv path from processing IO during recovery. It turns out iser just set a lock to that pointer which was pointless. We now disconnect the transport connection before doing recovery so we do not need the recv lock. For iscsi_tcp we still stop the recv path incase older tools are being used. This patch also has iscsi_itt_to_ctask user grab the session lock and has the caller access the task with the lock or get a ref to it in case the target is broken and sends a tmf success response then sends data or a response for the command that was supposed to be affected bty the tmf. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 20 ++++++++++---------- drivers/infiniband/ulp/iser/iscsi_iser.h | 2 -- drivers/infiniband/ulp/iser/iser_initiator.c | 6 ++++++ 3 files changed, 16 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 08edbaf8922..c02eabd383a 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -281,9 +281,6 @@ iscsi_iser_conn_create(struct iscsi_cls_session *cls_session, uint32_t conn_idx) conn->max_recv_dlength = 128; iser_conn = conn->dd_data; - /* currently this is the only field which need to be initiated */ - rwlock_init(&iser_conn->lock); - conn->dd_data = iser_conn; iser_conn->iscsi_conn = conn; @@ -342,9 +339,6 @@ iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session, ib_conn->iser_conn = iser_conn; iser_conn->ib_conn = ib_conn; iser_conn_get(ib_conn); - - conn->recv_lock = &iser_conn->lock; - return 0; } @@ -355,12 +349,18 @@ iscsi_iser_conn_stop(struct iscsi_cls_conn *cls_conn, int flag) struct iscsi_iser_conn *iser_conn = conn->dd_data; struct iser_conn *ib_conn = iser_conn->ib_conn; - iscsi_conn_stop(cls_conn, flag); /* - * There is no unbind event so the stop callback - * must release the ref from the bind. + * Userspace may have goofed up and not bound the connection or + * might have only partially setup the connection. */ - iser_conn_put(ib_conn); + if (ib_conn) { + iscsi_conn_stop(cls_conn, flag); + /* + * There is no unbind event so the stop callback + * must release the ref from the bind. + */ + iser_conn_put(ib_conn); + } iser_conn->ib_conn = NULL; } diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index cdf48763b08..a547edeea96 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -263,8 +263,6 @@ struct iser_conn { struct iscsi_iser_conn { struct iscsi_conn *iscsi_conn;/* ptr to iscsi conn */ struct iser_conn *ib_conn; /* iSER IB conn */ - - rwlock_t lock; }; struct iscsi_iser_task { diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 35af60a23c6..c3608392213 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -558,7 +558,12 @@ void iser_rcv_completion(struct iser_desc *rx_desc, opcode = hdr->opcode & ISCSI_OPCODE_MASK; if (opcode == ISCSI_OP_SCSI_CMD_RSP) { + spin_lock(&conn->iscsi_conn->session->lock); task = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt); + if (task) + __iscsi_get_task(task); + spin_unlock(&conn->iscsi_conn->session->lock); + if (!task) iser_err("itt can't be matched to task!!! " "conn %p opcode %d itt %d\n", @@ -568,6 +573,7 @@ void iser_rcv_completion(struct iser_desc *rx_desc, iser_dbg("itt %d task %p\n",hdr->itt, task); iser_task->status = ISER_TASK_STATUS_COMPLETED; iser_task_rdma_finalize(iser_task); + iscsi_put_task(task); } } iser_dto_buffs_release(dto); -- cgit v1.2.3 From 8e9a20cee4511be4560f9c858d9994eb6913731e Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Mon, 16 Jun 2008 10:11:33 -0500 Subject: [SCSI] libiscsi, iscsi_tcp, ib_iser: fix setting of can_queue with old tools. This patch fixes two bugs that are related. 1. Old tools did not set can_queue/cmds_max. This patch modifies libiscsi so that when we add the host we catch this and set it to the default. 2. iscsi_tcp thought that the scsi command that was passed to the eh functions needed a iscsi_cmd_task allocated for it. It only needed a mgmt task, and now it does not matter since it all comes from the same pool and libiscsi handles this for the drivers. ib_iser had copied iscsi_tcp's code and set can_queue to its max - 1 to handle this. So this patch removes the max -1, and just sets it to the max. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index c02eabd383a..a56931e0397 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -595,7 +595,6 @@ static struct scsi_host_template iscsi_iser_sht = { .name = "iSCSI Initiator over iSER, v." DRV_VER, .queuecommand = iscsi_queuecommand, .change_queue_depth = iscsi_change_queue_depth, - .can_queue = ISCSI_DEF_XMIT_CMDS_MAX - 1, .sg_tablesize = ISCSI_ISER_SG_TABLESIZE, .max_sectors = 1024, .cmd_per_lun = ISCSI_MAX_CMD_PER_LUN, -- cgit v1.2.3 From 969a60f9db3f879f95bd37026a3c3bf02cc2568f Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:43 -0700 Subject: IB/srp: Remove use of cached P_Key/GID queries The SRP initiator is currently using ib_find_cached_pkey() and ib_get_cached_gid() in situations where the uncached ib_find_pkey() and ib_query_gid() functions serve just as well: sleeping is allowed and performance is not an issue. Since we want to eliminate the cached operations in the long term, convert SRP to use the uncached variants. Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/srp/ib_srp.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 435145709dd..81cc59ca559 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -49,8 +49,6 @@ #include #include -#include - #include "ib_srp.h" #define DRV_NAME "ib_srp" @@ -183,10 +181,10 @@ static int srp_init_qp(struct srp_target_port *target, if (!attr) return -ENOMEM; - ret = ib_find_cached_pkey(target->srp_host->srp_dev->dev, - target->srp_host->port, - be16_to_cpu(target->path.pkey), - &attr->pkey_index); + ret = ib_find_pkey(target->srp_host->srp_dev->dev, + target->srp_host->port, + be16_to_cpu(target->path.pkey), + &attr->pkey_index); if (ret) goto out; @@ -1883,8 +1881,7 @@ static ssize_t srp_create_target(struct device *dev, if (ret) goto err; - ib_get_cached_gid(host->srp_dev->dev, host->port, 0, - &target->path.sgid); + ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid); shost_printk(KERN_DEBUG, target->scsi_host, PFX "new target: id_ext %016llx ioc_guid %016llx pkey %04x " -- cgit v1.2.3 From 929555a2baed9b0b050d03532655bfd721a43c44 Mon Sep 17 00:00:00 2001 From: Christophe Jaillet Date: Mon, 14 Jul 2008 23:48:43 -0700 Subject: RDMA/nes: Remove unnecessary memset() Remove an explicit memset(..., 0, ...) of a 'listener' structure allocated with kzalloc(). Signed-off-by: Christophe Jaillet Acked-by: Faisal Latif Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_cm.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index 9a4b40fae40..6aa531d5276 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -1603,7 +1603,6 @@ static struct nes_cm_listener *mini_cm_listen(struct nes_cm_core *cm_core, return NULL; } - memset(listener, 0, sizeof(struct nes_cm_listener)); listener->loc_addr = htonl(cm_info->loc_addr); listener->loc_port = htons(cm_info->loc_port); listener->reused_node = 0; -- cgit v1.2.3 From a9474917099e007c0f51d5474394b5890111614f Mon Sep 17 00:00:00 2001 From: Sean Hefty Date: Mon, 14 Jul 2008 23:48:43 -0700 Subject: RDMA: Fix license text The license text for several files references a third software license that was inadvertently copied in. Update the license to what was intended. This update was based on a request from HP. Signed-off-by: Sean Hefty Signed-off-by: Roland Dreier --- drivers/infiniband/core/addr.c | 41 +++++++++++++++++++++++------------------ drivers/infiniband/core/cma.c | 42 +++++++++++++++++++++++------------------- 2 files changed, 46 insertions(+), 37 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c index 781ea595037..e4eb8be3bb0 100644 --- a/drivers/infiniband/core/addr.c +++ b/drivers/infiniband/core/addr.c @@ -4,28 +4,33 @@ * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2005 Intel Corporation. All rights reserved. * - * This Software is licensed under one of the following licenses: + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: * - * 1) under the terms of the "Common Public License 1.0" a copy of which is - * available from the Open Source Initiative, see - * http://www.opensource.org/licenses/cpl.php. + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: * - * 2) under the terms of the "The BSD License" a copy of which is - * available from the Open Source Initiative, see - * http://www.opensource.org/licenses/bsd-license.php. + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. * - * 3) under the terms of the "GNU General Public License (GPL) Version 2" a - * copy of which is available from the Open Source Initiative, see - * http://www.opensource.org/licenses/gpl-license.php. + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. * - * Licensee has the right to choose one of the above licenses. - * - * Redistributions of source code must retain the above copyright - * notice and one of the license notices. - * - * Redistributions in binary form must reproduce both the above copyright - * notice, one of the license notices in the documentation - * and/or other materials provided with the distribution. + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. */ #include diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 671f1373805..e5bd6172a1f 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -4,29 +4,33 @@ * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. * Copyright (c) 2005-2006 Intel Corporation. All rights reserved. * - * This Software is licensed under one of the following licenses: + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: * - * 1) under the terms of the "Common Public License 1.0" a copy of which is - * available from the Open Source Initiative, see - * http://www.opensource.org/licenses/cpl.php. + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: * - * 2) under the terms of the "The BSD License" a copy of which is - * available from the Open Source Initiative, see - * http://www.opensource.org/licenses/bsd-license.php. + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. * - * 3) under the terms of the "GNU General Public License (GPL) Version 2" a - * copy of which is available from the Open Source Initiative, see - * http://www.opensource.org/licenses/gpl-license.php. - * - * Licensee has the right to choose one of the above licenses. - * - * Redistributions of source code must retain the above copyright - * notice and one of the license notices. - * - * Redistributions in binary form must reproduce both the above copyright - * notice, one of the license notices in the documentation - * and/or other materials provided with the distribution. + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. */ #include -- cgit v1.2.3 From 164ba0893c27a216557396320b6063fdac040392 Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Mon, 14 Jul 2008 23:48:43 -0700 Subject: IB/sa: Fail requests made while creating new SM AH This patch solves a race that occurs after an event occurs that causes the SA query module to flush its SM address handle (AH). When SM AH becomes invalid and needs an update it is handled by the global workqueue. On the other hand this event is also handled in the IPoIB driver by queuing work in the ipoib_workqueue that does multicast joins. Although queuing is in the right order, it is done to 2 different workqueues and so there is no guarantee that the first to be queued is the first to be executed. This causes a problem because IPoIB may end up sending an request to the old SM, which will take a long time to time out (since the old SM is gone); this leads to a much longer than necessary interruption in multicast traffer. The patch sets the SA query module's SM AH to NULL when the event occurs, and until update_sm_ah() is done, any request that needs sm_ah fails with -EAGAIN return status. For consumers, the patch doesn't make things worse. Before the patch, MADs are sent to the wrong SM so the request gets lost. Consumers can be improved if they examine the return code and respond to EAGAIN properly but even without an improvement the situation is not getting worse. Signed-off-by: Moni Levy Signed-off-by: Moni Shoua Signed-off-by: Roland Dreier --- drivers/infiniband/core/sa_query.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index cf474ec2707..78ea8157d62 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -361,7 +361,7 @@ static void update_sm_ah(struct work_struct *work) { struct ib_sa_port *port = container_of(work, struct ib_sa_port, update_task); - struct ib_sa_sm_ah *new_ah, *old_ah; + struct ib_sa_sm_ah *new_ah; struct ib_port_attr port_attr; struct ib_ah_attr ah_attr; @@ -397,12 +397,9 @@ static void update_sm_ah(struct work_struct *work) } spin_lock_irq(&port->ah_lock); - old_ah = port->sm_ah; port->sm_ah = new_ah; spin_unlock_irq(&port->ah_lock); - if (old_ah) - kref_put(&old_ah->ref, free_sm_ah); } static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event) @@ -413,8 +410,17 @@ static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event event->event == IB_EVENT_PKEY_CHANGE || event->event == IB_EVENT_SM_CHANGE || event->event == IB_EVENT_CLIENT_REREGISTER) { - struct ib_sa_device *sa_dev; - sa_dev = container_of(handler, typeof(*sa_dev), event_handler); + unsigned long flags; + struct ib_sa_device *sa_dev = + container_of(handler, typeof(*sa_dev), event_handler); + struct ib_sa_port *port = + &sa_dev->port[event->element.port_num - sa_dev->start_port]; + + spin_lock_irqsave(&port->ah_lock, flags); + if (port->sm_ah) + kref_put(&port->sm_ah->ref, free_sm_ah); + port->sm_ah = NULL; + spin_unlock_irqrestore(&port->ah_lock, flags); schedule_work(&sa_dev->port[event->element.port_num - sa_dev->start_port].update_task); @@ -519,6 +525,10 @@ static int alloc_mad(struct ib_sa_query *query, gfp_t gfp_mask) unsigned long flags; spin_lock_irqsave(&query->port->ah_lock, flags); + if (!query->port->sm_ah) { + spin_unlock_irqrestore(&query->port->ah_lock, flags); + return -EAGAIN; + } kref_get(&query->port->sm_ah->ref); query->sm_ah = query->port->sm_ah; spin_unlock_irqrestore(&query->port->ah_lock, flags); -- cgit v1.2.3 From 9670e553915e67fb68f13258644342c68dc26b84 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:44 -0700 Subject: IB/mlx4: Optimize QP stamping The idea is that for QPs with fixed size work requests (eg selective signaling QPs), before stamping the WQE, we read the value of the DS field, which gives the effective size of the descriptor as used in the previous post. Then we stamp only that area, since the rest of the descriptor is already stamped. When initializing the send queue buffer, make sure the DS field is initialized to the max descriptor size so that the subsequent stamping will be done on the entire descriptor area. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/qp.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index a80df22deae..4b0ac5d68c4 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -129,9 +129,10 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size) int ind; void *buf; __be32 stamp; + struct mlx4_wqe_ctrl_seg *ctrl; - s = roundup(size, 1U << qp->sq.wqe_shift); if (qp->sq_max_wqes_per_wr > 1) { + s = roundup(size, 1U << qp->sq.wqe_shift); for (i = 0; i < s; i += 64) { ind = (i >> qp->sq.wqe_shift) + n; stamp = ind & qp->sq.wqe_cnt ? cpu_to_be32(0x7fffffff) : @@ -141,7 +142,8 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size) *wqe = stamp; } } else { - buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1)); + ctrl = buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1)); + s = (ctrl->fence_size & 0x3f) << 4; for (i = 64; i < s; i += 64) { wqe = buf + i; *wqe = cpu_to_be32(0xffffffff); @@ -1063,6 +1065,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, for (i = 0; i < qp->sq.wqe_cnt; ++i) { ctrl = get_send_wqe(qp, i); ctrl->owner_opcode = cpu_to_be32(1 << 31); + if (qp->sq_max_wqes_per_wr == 1) + ctrl->fence_size = 1 << (qp->sq.wqe_shift - 4); stamp_send_wqe(qp, i, 1 << qp->sq.wqe_shift); } -- cgit v1.2.3 From fd91b1bf1bb6fb443cb8c7600c7314f093b31f40 Mon Sep 17 00:00:00 2001 From: "Robert P. J. Day" Date: Mon, 14 Jul 2008 23:48:44 -0700 Subject: IB/ipath: Simplify code using ARRAY_SIZE() macro Signed-off-by: Robert P. J. Day Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ipath/ipath_iba7220.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ipath/ipath_iba7220.c b/drivers/infiniband/hw/ipath/ipath_iba7220.c index 8eee7830f04..fb70712ac85 100644 --- a/drivers/infiniband/hw/ipath/ipath_iba7220.c +++ b/drivers/infiniband/hw/ipath/ipath_iba7220.c @@ -2228,8 +2228,8 @@ static void ipath_autoneg_send(struct ipath_devdata *dd, int which) 0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x40000001, 0x1388, 0x15e, /* rest 0's */ }; - dcnt = sizeof(madpayload_start)/sizeof(madpayload_start[0]); - hcnt = sizeof(hdr)/sizeof(hdr[0]); + dcnt = ARRAY_SIZE(madpayload_start); + hcnt = ARRAY_SIZE(hdr); if (!swapped) { /* for maintainability, do it at runtime */ for (i = 0; i < hcnt; i++) { -- cgit v1.2.3 From f3781d2e89f12dd5afa046dc56032af6e39bd116 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:44 -0700 Subject: RDMA: Remove subversion $Id tags They don't get updated by git and so they're worse than useless. Signed-off-by: Roland Dreier --- drivers/infiniband/core/agent.h | 2 -- drivers/infiniband/core/cache.c | 2 -- drivers/infiniband/core/cm.c | 2 -- drivers/infiniband/core/core_priv.h | 2 -- drivers/infiniband/core/device.c | 2 -- drivers/infiniband/core/fmr_pool.c | 2 -- drivers/infiniband/core/mad_priv.h | 2 -- drivers/infiniband/core/mad_rmpp.c | 2 -- drivers/infiniband/core/mad_rmpp.h | 2 -- drivers/infiniband/core/packer.c | 2 -- drivers/infiniband/core/sa_query.c | 2 -- drivers/infiniband/core/sysfs.c | 2 -- drivers/infiniband/core/ucm.c | 2 -- drivers/infiniband/core/ud_header.c | 2 -- drivers/infiniband/core/umem.c | 2 -- drivers/infiniband/core/user_mad.c | 2 -- drivers/infiniband/core/uverbs.h | 2 -- drivers/infiniband/core/uverbs_cmd.c | 2 -- drivers/infiniband/core/uverbs_main.c | 2 -- drivers/infiniband/core/verbs.c | 2 -- drivers/infiniband/hw/mthca/mthca_allocator.c | 2 -- drivers/infiniband/hw/mthca/mthca_av.c | 2 -- drivers/infiniband/hw/mthca/mthca_catas.c | 2 -- drivers/infiniband/hw/mthca/mthca_cmd.c | 2 -- drivers/infiniband/hw/mthca/mthca_cmd.h | 2 -- drivers/infiniband/hw/mthca/mthca_config_reg.h | 2 -- drivers/infiniband/hw/mthca/mthca_cq.c | 2 -- drivers/infiniband/hw/mthca/mthca_dev.h | 2 -- drivers/infiniband/hw/mthca/mthca_doorbell.h | 2 -- drivers/infiniband/hw/mthca/mthca_eq.c | 2 -- drivers/infiniband/hw/mthca/mthca_mad.c | 2 -- drivers/infiniband/hw/mthca/mthca_main.c | 2 -- drivers/infiniband/hw/mthca/mthca_mcg.c | 2 -- drivers/infiniband/hw/mthca/mthca_memfree.c | 2 -- drivers/infiniband/hw/mthca/mthca_memfree.h | 2 -- drivers/infiniband/hw/mthca/mthca_mr.c | 2 -- drivers/infiniband/hw/mthca/mthca_pd.c | 2 -- drivers/infiniband/hw/mthca/mthca_profile.c | 2 -- drivers/infiniband/hw/mthca/mthca_profile.h | 2 -- drivers/infiniband/hw/mthca/mthca_provider.c | 2 -- drivers/infiniband/hw/mthca/mthca_provider.h | 2 -- drivers/infiniband/hw/mthca/mthca_qp.c | 2 -- drivers/infiniband/hw/mthca/mthca_reset.c | 2 -- drivers/infiniband/hw/mthca/mthca_srq.c | 2 -- drivers/infiniband/hw/mthca/mthca_uar.c | 2 -- drivers/infiniband/hw/mthca/mthca_user.h | 1 - drivers/infiniband/hw/mthca/mthca_wqe.h | 2 -- drivers/infiniband/ulp/ipoib/ipoib.h | 2 -- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_fs.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_ib.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_main.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 2 -- drivers/infiniband/ulp/ipoib/ipoib_vlan.c | 2 -- drivers/infiniband/ulp/iser/iscsi_iser.c | 3 --- drivers/infiniband/ulp/iser/iscsi_iser.h | 2 -- drivers/infiniband/ulp/iser/iser_initiator.c | 2 -- drivers/infiniband/ulp/iser/iser_memory.c | 2 -- drivers/infiniband/ulp/iser/iser_verbs.c | 2 -- drivers/infiniband/ulp/srp/ib_srp.c | 2 -- drivers/infiniband/ulp/srp/ib_srp.h | 2 -- 62 files changed, 124 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/agent.h b/drivers/infiniband/core/agent.h index fb9ed1489f9..6669287009c 100644 --- a/drivers/infiniband/core/agent.h +++ b/drivers/infiniband/core/agent.h @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: agent.h 1389 2004-12-27 22:56:47Z roland $ */ #ifndef __AGENT_H_ diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c index e85f7013de5..68883565b72 100644 --- a/drivers/infiniband/core/cache.c +++ b/drivers/infiniband/core/cache.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: cache.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index a47fe64e5c3..55738eead3b 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: cm.c 4311 2005-12-05 18:42:01Z sean.hefty $ */ #include diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h index 7ad47a4b166..05ac36e6acd 100644 --- a/drivers/infiniband/core/core_priv.h +++ b/drivers/infiniband/core/core_priv.h @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: core_priv.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef _CORE_PRIV_H diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 5ac5ffee05c..7913b804311 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: device.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/core/fmr_pool.c b/drivers/infiniband/core/fmr_pool.c index 1286dc1b98b..4507043d24c 100644 --- a/drivers/infiniband/core/fmr_pool.c +++ b/drivers/infiniband/core/fmr_pool.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: fmr_pool.c 2730 2005-06-28 16:43:03Z sean.hefty $ */ #include diff --git a/drivers/infiniband/core/mad_priv.h b/drivers/infiniband/core/mad_priv.h index 8b75010016e..05ce331733b 100644 --- a/drivers/infiniband/core/mad_priv.h +++ b/drivers/infiniband/core/mad_priv.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mad_priv.h 5596 2006-03-03 01:00:07Z sean.hefty $ */ #ifndef __IB_MAD_PRIV_H__ diff --git a/drivers/infiniband/core/mad_rmpp.c b/drivers/infiniband/core/mad_rmpp.c index a5e2a310f31..d0ef7d61c03 100644 --- a/drivers/infiniband/core/mad_rmpp.c +++ b/drivers/infiniband/core/mad_rmpp.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mad_rmpp.c 1921 2005-03-02 22:58:44Z sean.hefty $ */ #include "mad_priv.h" diff --git a/drivers/infiniband/core/mad_rmpp.h b/drivers/infiniband/core/mad_rmpp.h index f0616fd2249..3d336bff114 100644 --- a/drivers/infiniband/core/mad_rmpp.h +++ b/drivers/infiniband/core/mad_rmpp.h @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mad_rmpp.h 1921 2005-02-25 22:58:44Z sean.hefty $ */ #ifndef __MAD_RMPP_H__ diff --git a/drivers/infiniband/core/packer.c b/drivers/infiniband/core/packer.c index c972d723576..019bd4b0863 100644 --- a/drivers/infiniband/core/packer.c +++ b/drivers/infiniband/core/packer.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: packer.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index 78ea8157d62..1341de793e5 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: sa_query.c 2811 2005-07-06 18:11:43Z halr $ */ #include diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index 95756551cf7..36a0ef97c6a 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: sysfs.c 1349 2004-12-16 21:09:43Z roland $ */ #include "core_priv.h" diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index b25675faaaf..9494005d1c9 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ucm.c 4311 2005-12-05 18:42:01Z sean.hefty $ */ #include diff --git a/drivers/infiniband/core/ud_header.c b/drivers/infiniband/core/ud_header.c index 997c07db6d8..8ec7876bedc 100644 --- a/drivers/infiniband/core/ud_header.c +++ b/drivers/infiniband/core/ud_header.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ud_header.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c index a1768dbb072..6f7c096abf1 100644 --- a/drivers/infiniband/core/umem.c +++ b/drivers/infiniband/core/umem.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: uverbs_mem.c 2743 2005-06-28 22:27:59Z roland $ */ #include diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index 208c7f34323..268a2d23b7c 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: user_mad.c 5596 2006-03-03 01:00:07Z sean.hefty $ */ #include diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h index 376a57ce1b4..b3ea9587dc8 100644 --- a/drivers/infiniband/core/uverbs.h +++ b/drivers/infiniband/core/uverbs.h @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: uverbs.h 2559 2005-06-06 19:43:16Z roland $ */ #ifndef UVERBS_H diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 2c3bff5fe86..112b37cd689 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: uverbs_cmd.c 2708 2005-06-24 17:27:21Z roland $ */ #include diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 0f34858e31e..aeee856c406 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: uverbs_main.c 2733 2005-06-28 19:14:34Z roland $ */ #include diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 05042089de6..9f399d3a42b 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -34,8 +34,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: verbs.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_allocator.c b/drivers/infiniband/hw/mthca/mthca_allocator.c index a7630670961..c5ccc2daab6 100644 --- a/drivers/infiniband/hw/mthca/mthca_allocator.c +++ b/drivers/infiniband/hw/mthca/mthca_allocator.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_allocator.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_av.c b/drivers/infiniband/hw/mthca/mthca_av.c index 4b111a852ff..32f6c631545 100644 --- a/drivers/infiniband/hw/mthca/mthca_av.c +++ b/drivers/infiniband/hw/mthca/mthca_av.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_av.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c index e948158a28d..40573e48439 100644 --- a/drivers/infiniband/hw/mthca/mthca_catas.c +++ b/drivers/infiniband/hw/mthca/mthca_catas.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id$ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c index 54d230ee7d6..c33e1c53c79 100644 --- a/drivers/infiniband/hw/mthca/mthca_cmd.c +++ b/drivers/infiniband/hw/mthca/mthca_cmd.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_cmd.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.h b/drivers/infiniband/hw/mthca/mthca_cmd.h index 8928ca4a932..6efd3265f24 100644 --- a/drivers/infiniband/hw/mthca/mthca_cmd.h +++ b/drivers/infiniband/hw/mthca/mthca_cmd.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_cmd.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef MTHCA_CMD_H diff --git a/drivers/infiniband/hw/mthca/mthca_config_reg.h b/drivers/infiniband/hw/mthca/mthca_config_reg.h index afa56bfaab2..75671f75cac 100644 --- a/drivers/infiniband/hw/mthca/mthca_config_reg.h +++ b/drivers/infiniband/hw/mthca/mthca_config_reg.h @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_config_reg.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef MTHCA_CONFIG_REG_H diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c index 20401d2ba6b..f788fce71ac 100644 --- a/drivers/infiniband/hw/mthca/mthca_cq.c +++ b/drivers/infiniband/hw/mthca/mthca_cq.c @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_cq.c 1369 2004-12-20 16:17:07Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h index 7bc32f8e377..2997d8d564e 100644 --- a/drivers/infiniband/hw/mthca/mthca_dev.h +++ b/drivers/infiniband/hw/mthca/mthca_dev.h @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_dev.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef MTHCA_DEV_H diff --git a/drivers/infiniband/hw/mthca/mthca_doorbell.h b/drivers/infiniband/hw/mthca/mthca_doorbell.h index b374dc395be..14f51ef97d7 100644 --- a/drivers/infiniband/hw/mthca/mthca_doorbell.h +++ b/drivers/infiniband/hw/mthca/mthca_doorbell.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_doorbell.h 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c index 8bde7f98e58..4e36aa7cb3d 100644 --- a/drivers/infiniband/hw/mthca/mthca_eq.c +++ b/drivers/infiniband/hw/mthca/mthca_eq.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_eq.c 1382 2004-12-24 02:21:02Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index 8b7e83e6e88..640449582ab 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_mad.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c index 200cf13fc9b..fb9f91b60f3 100644 --- a/drivers/infiniband/hw/mthca/mthca_main.c +++ b/drivers/infiniband/hw/mthca/mthca_main.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_main.c 1396 2004-12-28 04:10:27Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_mcg.c b/drivers/infiniband/hw/mthca/mthca_mcg.c index a8ad072be07..3f5f9487920 100644 --- a/drivers/infiniband/hw/mthca/mthca_mcg.c +++ b/drivers/infiniband/hw/mthca/mthca_mcg.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_mcg.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_memfree.c b/drivers/infiniband/hw/mthca/mthca_memfree.c index d5862e5d99a..1f7d1a29d2a 100644 --- a/drivers/infiniband/hw/mthca/mthca_memfree.c +++ b/drivers/infiniband/hw/mthca/mthca_memfree.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id$ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_memfree.h b/drivers/infiniband/hw/mthca/mthca_memfree.h index a1ab06847b7..da9b8f9b884 100644 --- a/drivers/infiniband/hw/mthca/mthca_memfree.h +++ b/drivers/infiniband/hw/mthca/mthca_memfree.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id$ */ #ifndef MTHCA_MEMFREE_H diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c index 820205dec56..8489b1e81c0 100644 --- a/drivers/infiniband/hw/mthca/mthca_mr.c +++ b/drivers/infiniband/hw/mthca/mthca_mr.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_mr.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_pd.c b/drivers/infiniband/hw/mthca/mthca_pd.c index c1e950764bd..266f14e4740 100644 --- a/drivers/infiniband/hw/mthca/mthca_pd.c +++ b/drivers/infiniband/hw/mthca/mthca_pd.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_pd.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_profile.c b/drivers/infiniband/hw/mthca/mthca_profile.c index 605a8d57fac..d168c254061 100644 --- a/drivers/infiniband/hw/mthca/mthca_profile.c +++ b/drivers/infiniband/hw/mthca/mthca_profile.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_profile.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_profile.h b/drivers/infiniband/hw/mthca/mthca_profile.h index e76cb62d8e3..62b009cc873 100644 --- a/drivers/infiniband/hw/mthca/mthca_profile.h +++ b/drivers/infiniband/hw/mthca/mthca_profile.h @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_profile.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef MTHCA_PROFILE_H diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index be34f99ca62..87ad889e367 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -32,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_provider.c 4859 2006-01-09 21:55:10Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_provider.h b/drivers/infiniband/hw/mthca/mthca_provider.h index 934bf954403..c621f8794b8 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.h +++ b/drivers/infiniband/hw/mthca/mthca_provider.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_provider.h 1349 2004-12-16 21:09:43Z roland $ */ #ifndef MTHCA_PROVIDER_H diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index 09dc3614cf2..3b1c5baf13b 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_qp.c 1355 2004-12-17 15:23:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_reset.c b/drivers/infiniband/hw/mthca/mthca_reset.c index 91934f2d9db..acb6817f606 100644 --- a/drivers/infiniband/hw/mthca/mthca_reset.c +++ b/drivers/infiniband/hw/mthca/mthca_reset.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_reset.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_srq.c b/drivers/infiniband/hw/mthca/mthca_srq.c index a5ffff6e102..4fabe62aab8 100644 --- a/drivers/infiniband/hw/mthca/mthca_srq.c +++ b/drivers/infiniband/hw/mthca/mthca_srq.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_srq.c 3047 2005-08-10 03:59:35Z roland $ */ #include diff --git a/drivers/infiniband/hw/mthca/mthca_uar.c b/drivers/infiniband/hw/mthca/mthca_uar.c index 8b728486410..ca5900c96fc 100644 --- a/drivers/infiniband/hw/mthca/mthca_uar.c +++ b/drivers/infiniband/hw/mthca/mthca_uar.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id$ */ #include /* PAGE_SHIFT */ diff --git a/drivers/infiniband/hw/mthca/mthca_user.h b/drivers/infiniband/hw/mthca/mthca_user.h index e1262c942db..5fe56e81073 100644 --- a/drivers/infiniband/hw/mthca/mthca_user.h +++ b/drivers/infiniband/hw/mthca/mthca_user.h @@ -29,7 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * */ #ifndef MTHCA_USER_H diff --git a/drivers/infiniband/hw/mthca/mthca_wqe.h b/drivers/infiniband/hw/mthca/mthca_wqe.h index b3551a8dea1..341a5ae881c 100644 --- a/drivers/infiniband/hw/mthca/mthca_wqe.h +++ b/drivers/infiniband/hw/mthca/mthca_wqe.h @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: mthca_wqe.h 3047 2005-08-10 03:59:35Z roland $ */ #ifndef MTHCA_WQE_H diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index ca126fc2b85..0dcbab3203c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib.h 1358 2004-12-17 22:00:11Z roland $ */ #ifndef _IPOIB_H diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 97e67d36378..91c95929991 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id$ */ #include diff --git a/drivers/infiniband/ulp/ipoib/ipoib_fs.c b/drivers/infiniband/ulp/ipoib/ipoib_fs.c index 8b882bbd1d0..961c585da21 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_fs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_fs.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_fs.c 1389 2004-12-27 22:56:47Z roland $ */ #include diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index f429bce24c2..eca8518d79a 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -31,8 +31,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_ib.c 1386 2004-12-27 16:23:17Z roland $ */ #include diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 2442090ac8d..f217b1edd0a 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_main.c 1377 2004-12-23 19:57:12Z roland $ */ #include "ipoib.h" diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 3f663fb852c..4a6538b9301 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -30,8 +30,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_multicast.c 1362 2004-12-18 15:56:29Z roland $ */ #include diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index 8766d29ce3b..810790ae753 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_verbs.c 1349 2004-12-16 21:09:43Z roland $ */ #include "ipoib.h" diff --git a/drivers/infiniband/ulp/ipoib/ipoib_vlan.c b/drivers/infiniband/ulp/ipoib/ipoib_vlan.c index 1cdb5cfb0ff..b08eb56196d 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_vlan.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_vlan.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ipoib_vlan.c 1349 2004-12-16 21:09:43Z roland $ */ #include diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index aeb58cae9a3..356fac6d105 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -42,9 +42,6 @@ * Zhenyu Wang * Modified by: * Erez Zilber - * - * - * $Id: iscsi_iser.c 6965 2006-05-07 11:36:20Z ogerlitz $ */ #include diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index a8c1b300e34..0e10703cf59 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -36,8 +36,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: iscsi_iser.h 7051 2006-05-10 12:29:11Z ogerlitz $ */ #ifndef __ISCSI_ISER_H__ #define __ISCSI_ISER_H__ diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 08dc81c46f4..31ad498bdc5 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: iser_initiator.c 6964 2006-05-07 11:11:43Z ogerlitz $ */ #include #include diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index cac50c4dc15..81e49cb10ed 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: iser_memory.c 6964 2006-05-07 11:11:43Z ogerlitz $ */ #include #include diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index d19cfe605eb..77cabee7cc0 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -29,8 +29,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: iser_verbs.c 7051 2006-05-10 12:29:11Z ogerlitz $ */ #include #include diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 81cc59ca559..ed7c5f72cb8 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ib_srp.c 3932 2005-11-01 17:19:29Z roland $ */ #include diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h index 63d2ae72406..e185b907fc1 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.h +++ b/drivers/infiniband/ulp/srp/ib_srp.h @@ -28,8 +28,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: ib_srp.h 3932 2005-11-01 17:19:29Z roland $ */ #ifndef IB_SRP_H -- cgit v1.2.3 From f89271da32bc1a636cf4eb078e615930886cd013 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:44 -0700 Subject: IPoIB: Copy small received SKBs in connected mode The connected mode implementation in the IPoIB driver has a large overhead in the way SKBs are handled in the receive flow. It usually allocates an SKB with as big as was used in the currently received SKB and moves unused fragments from the old SKB to the new one. This involves a loop on all the remaining fragments and incurs overhead on the CPU. This patch, for small SKBs, allocates an SKB just large enough to contain the received data and copies to it the data from the received SKB. The newly allocated SKB is passed to the stack and the old SKB is reposted. When running netperf, UDP small messages, without this pach I get: UDP UNIDIRECTIONAL SEND TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 14.4.3.178 (14.4.3.178) port 0 AF_INET Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 114688 128 10.00 5142034 0 526.31 114688 10.00 1130489 115.71 With this patch I get both send and receive at ~315 mbps. The reason that send performance actually slows down is as follows: When using this patch, the overhead of the CPU for handling RX packets is dramatically reduced. As a result, we do not experience RNR NAK messages from the receiver which cause the connection to be closed and reopened again; when the patch is not used, the receiver cannot handle the packets fast enough so there is less time to post new buffers and hence the mentioned RNR NACKs. So what happens is that the application *thinks* it posted a certain number of packets for transmission but these packets are flushed and do not really get transmitted. Since the connection gets opened and closed many times, each time netperf gets the CPU time that otherwise would have been given to IPoIB to actually transmit the packets. This can be verified when looking at the port counters -- the output of ifconfig and the oputput of netperf (this is for the case without the patch): tx packets ========== port counter: 1,543,996 ifconfig: 1,581,426 netperf: 5,142,034 rx packets ========== netperf 1,1304,089 Signed-off-by: Eli Cohen --- drivers/infiniband/ulp/ipoib/ipoib.h | 1 + drivers/infiniband/ulp/ipoib/ipoib_cm.c | 19 +++++++++++++++++++ drivers/infiniband/ulp/ipoib/ipoib_main.c | 6 ++++++ 3 files changed, 26 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 0dcbab3203c..8754b364f22 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -95,6 +95,7 @@ enum { IPOIB_MCAST_FLAG_ATTACHED = 3, MAX_SEND_CQE = 16, + IPOIB_CM_COPYBREAK = 256, }; #define IPOIB_OP_RECV (1ul << 31) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 91c95929991..6223fc39af7 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -523,6 +523,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) u64 mapping[IPOIB_CM_RX_SG]; int frags; int has_srq; + struct sk_buff *small_skb; ipoib_dbg_data(priv, "cm recv completion: id %d, status: %d\n", wr_id, wc->status); @@ -577,6 +578,23 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) } } + if (wc->byte_len < IPOIB_CM_COPYBREAK) { + int dlen = wc->byte_len; + + small_skb = dev_alloc_skb(dlen + 12); + if (small_skb) { + skb_reserve(small_skb, 12); + ib_dma_sync_single_for_cpu(priv->ca, rx_ring[wr_id].mapping[0], + dlen, DMA_FROM_DEVICE); + skb_copy_from_linear_data(skb, small_skb->data, dlen); + ib_dma_sync_single_for_device(priv->ca, rx_ring[wr_id].mapping[0], + dlen, DMA_FROM_DEVICE); + skb_put(small_skb, dlen); + skb = small_skb; + goto copied; + } + } + frags = PAGE_ALIGN(wc->byte_len - min(wc->byte_len, (unsigned)IPOIB_CM_HEAD_SIZE)) / PAGE_SIZE; @@ -599,6 +617,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) skb_put_frags(skb, IPOIB_CM_HEAD_SIZE, wc->byte_len, newskb); +copied: skb->protocol = ((struct ipoib_header *) skb->data)->proto; skb_reset_mac_header(skb); skb_pull(skb, IPOIB_ENCAP_LEN); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index f217b1edd0a..bfe1dbf9920 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -1302,6 +1302,12 @@ static int __init ipoib_init_module(void) ipoib_max_conn_qp = min(ipoib_max_conn_qp, IPOIB_CM_MAX_CONN_QP); #endif + /* + * When copying small received packets, we only copy from the + * linear data part of the SKB, so we rely on this condition. + */ + BUILD_BUG_ON(IPOIB_CM_COPYBREAK > IPOIB_CM_HEAD_SIZE); + ret = ipoib_register_debugfs(); if (ret) return ret; -- cgit v1.2.3 From 00f7ec36c9324928e4cd23f02e6d8550f30c32ca Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:45 -0700 Subject: RDMA/core: Add memory management extensions support This patch adds support for the IB "base memory management extension" (BMME) and the equivalent iWARP operations (which the iWARP verbs mandates all devices must implement). The new operations are: - Allocate an ib_mr for use in fast register work requests. - Allocate/free a physical buffer lists for use in fast register work requests. This allows device drivers to allocate this memory as needed for use in posting send requests (eg via dma_alloc_coherent). - New send queue work requests: * send with remote invalidate * fast register memory region * local invalidate memory region * RDMA read with invalidate local memory region (iWARP only) Consumer interface details: - A new device capability flag IB_DEVICE_MEM_MGT_EXTENSIONS is added to indicate device support for these features. - New send work request opcodes IB_WR_FAST_REG_MR, IB_WR_LOCAL_INV, IB_WR_RDMA_READ_WITH_INV are added. - A new consumer API function, ib_alloc_mr() is added to allocate fast register memory regions. - New consumer API functions, ib_alloc_fast_reg_page_list() and ib_free_fast_reg_page_list() are added to allocate and free device-specific memory for fast registration page lists. - A new consumer API function, ib_update_fast_reg_key(), is added to allow the key portion of the R_Key and L_Key of a fast registration MR to be updated. Consumers call this if desired before posting a IB_WR_FAST_REG_MR work request. Consumers can use this as follows: - MR is allocated with ib_alloc_mr(). - Page list memory is allocated with ib_alloc_fast_reg_page_list(). - MR R_Key/L_Key "key" field is updated with ib_update_fast_reg_key(). - MR made VALID and bound to a specific page list via ib_post_send(IB_WR_FAST_REG_MR) - MR made INVALID via ib_post_send(IB_WR_LOCAL_INV), ib_post_send(IB_WR_RDMA_READ_WITH_INV) or an incoming send with invalidate operation. - MR is deallocated with ib_dereg_mr() - page lists dealloced via ib_free_fast_reg_page_list(). Applications can allocate a fast register MR once, and then can repeatedly bind the MR to different physical block lists (PBLs) via posting work requests to a send queue (SQ). For each outstanding MR-to-PBL binding in the SQ pipe, a fast_reg_page_list needs to be allocated (the fast_reg_page_list is owned by the low-level driver from the consumer posting a work request until the request completes). Thus pipelining can be achieved while still allowing device-specific page_list processing. The 32-bit fast register memory key/STag is composed of a 24-bit index and an 8-bit key. The application can change the key each time it fast registers thus allowing more control over the peer's use of the key/STag (ie it can effectively be changed each time the rkey is rebound to a page list). Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/core/uverbs_cmd.c | 2 +- drivers/infiniband/core/verbs.c | 46 +++++++++++++++++++++++++++++++++ drivers/infiniband/hw/ehca/ehca_reqs.c | 2 +- drivers/infiniband/hw/ipath/ipath_cq.c | 2 +- drivers/infiniband/hw/ipath/ipath_rc.c | 4 +-- drivers/infiniband/hw/ipath/ipath_ruc.c | 4 +-- drivers/infiniband/hw/ipath/ipath_uc.c | 8 +++--- drivers/infiniband/hw/ipath/ipath_ud.c | 8 +++--- drivers/infiniband/hw/mlx4/cq.c | 12 ++++----- drivers/infiniband/hw/mthca/mthca_cq.c | 4 +-- 10 files changed, 69 insertions(+), 23 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 112b37cd689..56feab6c251 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -917,7 +917,7 @@ ssize_t ib_uverbs_poll_cq(struct ib_uverbs_file *file, resp->wc[i].opcode = wc[i].opcode; resp->wc[i].vendor_err = wc[i].vendor_err; resp->wc[i].byte_len = wc[i].byte_len; - resp->wc[i].imm_data = (__u32 __force) wc[i].imm_data; + resp->wc[i].ex.imm_data = (__u32 __force) wc[i].ex.imm_data; resp->wc[i].qp_num = wc[i].qp->qp_num; resp->wc[i].src_qp = wc[i].src_qp; resp->wc[i].wc_flags = wc[i].wc_flags; diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 9f399d3a42b..e0fbe597586 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -753,6 +753,52 @@ int ib_dereg_mr(struct ib_mr *mr) } EXPORT_SYMBOL(ib_dereg_mr); +struct ib_mr *ib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len) +{ + struct ib_mr *mr; + + if (!pd->device->alloc_fast_reg_mr) + return ERR_PTR(-ENOSYS); + + mr = pd->device->alloc_fast_reg_mr(pd, max_page_list_len); + + if (!IS_ERR(mr)) { + mr->device = pd->device; + mr->pd = pd; + mr->uobject = NULL; + atomic_inc(&pd->usecnt); + atomic_set(&mr->usecnt, 0); + } + + return mr; +} +EXPORT_SYMBOL(ib_alloc_fast_reg_mr); + +struct ib_fast_reg_page_list *ib_alloc_fast_reg_page_list(struct ib_device *device, + int max_page_list_len) +{ + struct ib_fast_reg_page_list *page_list; + + if (!device->alloc_fast_reg_page_list) + return ERR_PTR(-ENOSYS); + + page_list = device->alloc_fast_reg_page_list(device, max_page_list_len); + + if (!IS_ERR(page_list)) { + page_list->device = device; + page_list->max_page_list_len = max_page_list_len; + } + + return page_list; +} +EXPORT_SYMBOL(ib_alloc_fast_reg_page_list); + +void ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list) +{ + page_list->device->free_fast_reg_page_list(page_list); +} +EXPORT_SYMBOL(ib_free_fast_reg_page_list); + /* Memory windows */ struct ib_mw *ib_alloc_mw(struct ib_pd *pd) diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c index f093b0033da..b799b271021 100644 --- a/drivers/infiniband/hw/ehca/ehca_reqs.c +++ b/drivers/infiniband/hw/ehca/ehca_reqs.c @@ -681,7 +681,7 @@ poll_cq_one_read_cqe: wc->dlid_path_bits = cqe->dlid; wc->src_qp = cqe->remote_qp_number; wc->wc_flags = cqe->w_completion_flags; - wc->imm_data = cpu_to_be32(cqe->immediate_data); + wc->ex.imm_data = cpu_to_be32(cqe->immediate_data); wc->sl = cqe->service_level; poll_cq_one_exit0: diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c index a03bd28d9b4..d385e4168c9 100644 --- a/drivers/infiniband/hw/ipath/ipath_cq.c +++ b/drivers/infiniband/hw/ipath/ipath_cq.c @@ -82,7 +82,7 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited) wc->uqueue[head].opcode = entry->opcode; wc->uqueue[head].vendor_err = entry->vendor_err; wc->uqueue[head].byte_len = entry->byte_len; - wc->uqueue[head].imm_data = (__u32 __force)entry->imm_data; + wc->uqueue[head].ex.imm_data = (__u32 __force) entry->ex.imm_data; wc->uqueue[head].qp_num = entry->qp->qp_num; wc->uqueue[head].src_qp = entry->src_qp; wc->uqueue[head].wc_flags = entry->wc_flags; diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c index 108df667d2e..97710522624 100644 --- a/drivers/infiniband/hw/ipath/ipath_rc.c +++ b/drivers/infiniband/hw/ipath/ipath_rc.c @@ -1703,11 +1703,11 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, case OP(SEND_LAST_WITH_IMMEDIATE): send_last_imm: if (header_in_data) { - wc.imm_data = *(__be32 *) data; + wc.ex.imm_data = *(__be32 *) data; data += sizeof(__be32); } else { /* Immediate data comes after BTH */ - wc.imm_data = ohdr->u.imm_data; + wc.ex.imm_data = ohdr->u.imm_data; } hdrsize += 4; wc.wc_flags = IB_WC_WITH_IMM; diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c index a4b5521567f..af051f75766 100644 --- a/drivers/infiniband/hw/ipath/ipath_ruc.c +++ b/drivers/infiniband/hw/ipath/ipath_ruc.c @@ -331,7 +331,7 @@ again: switch (wqe->wr.opcode) { case IB_WR_SEND_WITH_IMM: wc.wc_flags = IB_WC_WITH_IMM; - wc.imm_data = wqe->wr.ex.imm_data; + wc.ex.imm_data = wqe->wr.ex.imm_data; /* FALLTHROUGH */ case IB_WR_SEND: if (!ipath_get_rwqe(qp, 0)) @@ -342,7 +342,7 @@ again: if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE))) goto inv_err; wc.wc_flags = IB_WC_WITH_IMM; - wc.imm_data = wqe->wr.ex.imm_data; + wc.ex.imm_data = wqe->wr.ex.imm_data; if (!ipath_get_rwqe(qp, 1)) goto rnr_nak; /* FALLTHROUGH */ diff --git a/drivers/infiniband/hw/ipath/ipath_uc.c b/drivers/infiniband/hw/ipath/ipath_uc.c index 0596ec16fcb..82cc588b8bf 100644 --- a/drivers/infiniband/hw/ipath/ipath_uc.c +++ b/drivers/infiniband/hw/ipath/ipath_uc.c @@ -379,11 +379,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, case OP(SEND_LAST_WITH_IMMEDIATE): send_last_imm: if (header_in_data) { - wc.imm_data = *(__be32 *) data; + wc.ex.imm_data = *(__be32 *) data; data += sizeof(__be32); } else { /* Immediate data comes after BTH */ - wc.imm_data = ohdr->u.imm_data; + wc.ex.imm_data = ohdr->u.imm_data; } hdrsize += 4; wc.wc_flags = IB_WC_WITH_IMM; @@ -483,11 +483,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE): rdma_last_imm: if (header_in_data) { - wc.imm_data = *(__be32 *) data; + wc.ex.imm_data = *(__be32 *) data; data += sizeof(__be32); } else { /* Immediate data comes after BTH */ - wc.imm_data = ohdr->u.imm_data; + wc.ex.imm_data = ohdr->u.imm_data; } hdrsize += 4; wc.wc_flags = IB_WC_WITH_IMM; diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c index 77ca8ca74e7..36aa242c487 100644 --- a/drivers/infiniband/hw/ipath/ipath_ud.c +++ b/drivers/infiniband/hw/ipath/ipath_ud.c @@ -96,7 +96,7 @@ static void ipath_ud_loopback(struct ipath_qp *sqp, struct ipath_swqe *swqe) if (swqe->wr.opcode == IB_WR_SEND_WITH_IMM) { wc.wc_flags = IB_WC_WITH_IMM; - wc.imm_data = swqe->wr.ex.imm_data; + wc.ex.imm_data = swqe->wr.ex.imm_data; } /* @@ -492,14 +492,14 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, if (qp->ibqp.qp_num > 1 && opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) { if (header_in_data) { - wc.imm_data = *(__be32 *) data; + wc.ex.imm_data = *(__be32 *) data; data += sizeof(__be32); } else - wc.imm_data = ohdr->u.ud.imm_data; + wc.ex.imm_data = ohdr->u.ud.imm_data; wc.wc_flags = IB_WC_WITH_IMM; hdrsize += sizeof(u32); } else if (opcode == IB_OPCODE_UD_SEND_ONLY) { - wc.imm_data = 0; + wc.ex.imm_data = 0; wc.wc_flags = 0; } else { dev->n_pkt_drops++; diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c index 4521319b140..299f20832ab 100644 --- a/drivers/infiniband/hw/mlx4/cq.c +++ b/drivers/infiniband/hw/mlx4/cq.c @@ -663,18 +663,18 @@ repoll: switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) { case MLX4_RECV_OPCODE_RDMA_WRITE_IMM: - wc->opcode = IB_WC_RECV_RDMA_WITH_IMM; - wc->wc_flags = IB_WC_WITH_IMM; - wc->imm_data = cqe->immed_rss_invalid; + wc->opcode = IB_WC_RECV_RDMA_WITH_IMM; + wc->wc_flags = IB_WC_WITH_IMM; + wc->ex.imm_data = cqe->immed_rss_invalid; break; case MLX4_RECV_OPCODE_SEND: wc->opcode = IB_WC_RECV; wc->wc_flags = 0; break; case MLX4_RECV_OPCODE_SEND_IMM: - wc->opcode = IB_WC_RECV; - wc->wc_flags = IB_WC_WITH_IMM; - wc->imm_data = cqe->immed_rss_invalid; + wc->opcode = IB_WC_RECV; + wc->wc_flags = IB_WC_WITH_IMM; + wc->ex.imm_data = cqe->immed_rss_invalid; break; } diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c index f788fce71ac..d9f4735c2b3 100644 --- a/drivers/infiniband/hw/mthca/mthca_cq.c +++ b/drivers/infiniband/hw/mthca/mthca_cq.c @@ -620,13 +620,13 @@ static inline int mthca_poll_one(struct mthca_dev *dev, case IB_OPCODE_SEND_LAST_WITH_IMMEDIATE: case IB_OPCODE_SEND_ONLY_WITH_IMMEDIATE: entry->wc_flags = IB_WC_WITH_IMM; - entry->imm_data = cqe->imm_etype_pkey_eec; + entry->ex.imm_data = cqe->imm_etype_pkey_eec; entry->opcode = IB_WC_RECV; break; case IB_OPCODE_RDMA_WRITE_LAST_WITH_IMMEDIATE: case IB_OPCODE_RDMA_WRITE_ONLY_WITH_IMMEDIATE: entry->wc_flags = IB_WC_WITH_IMM; - entry->imm_data = cqe->imm_etype_pkey_eec; + entry->ex.imm_data = cqe->imm_etype_pkey_eec; entry->opcode = IB_WC_RECV_RDMA_WITH_IMM; break; default: -- cgit v1.2.3 From e7e55829999deaab3f43e201a087731c02c54cf9 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:45 -0700 Subject: RDMA/cxgb3: MEM_MGT_EXTENSIONS support - set IB_DEVICE_MEM_MGT_EXTENSIONS capability bit if fw supports it. - set max_fast_reg_page_list_len device attribute. - add iwch_alloc_fast_reg_mr function. - add iwch_alloc_fastreg_pbl - add iwch_free_fastreg_pbl - adjust the WQ depth for kernel mode work queues to account for fastreg possibly taking 2 WR slots. - add fastreg_mr work request support. - add local_inv work request support. - add send_with_inv and send_with_se_inv work request support. - removed useless duplicate enums/defines for TPT/MW/MR stuff. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/cxio_hal.c | 15 ++- drivers/infiniband/hw/cxgb3/cxio_hal.h | 1 + drivers/infiniband/hw/cxgb3/cxio_wr.h | 90 ++++++++++++++---- drivers/infiniband/hw/cxgb3/iwch_cq.c | 15 ++- drivers/infiniband/hw/cxgb3/iwch_provider.c | 104 +++++++++++++++++++- drivers/infiniband/hw/cxgb3/iwch_provider.h | 8 -- drivers/infiniband/hw/cxgb3/iwch_qp.c | 142 +++++++++++++++++++--------- 7 files changed, 293 insertions(+), 82 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c index 3f441fc57c1..340e4181c76 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.c +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c @@ -145,7 +145,9 @@ static int cxio_hal_clear_qp_ctx(struct cxio_rdev *rdev_p, u32 qpid) } wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe)); memset(wqe, 0, sizeof(*wqe)); - build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 3, 0, qpid, 7); + build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, + T3_COMPLETION_FLAG | T3_NOTIFY_FLAG, 0, qpid, 7, + T3_SOPEOP); wqe->flags = cpu_to_be32(MODQP_WRITE_EC); sge_cmd = qpid << 8 | 3; wqe->sge_cmd = cpu_to_be64(sge_cmd); @@ -558,7 +560,7 @@ static int cxio_hal_init_ctrl_qp(struct cxio_rdev *rdev_p) wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe)); memset(wqe, 0, sizeof(*wqe)); build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 0, 0, - T3_CTL_QP_TID, 7); + T3_CTL_QP_TID, 7, T3_SOPEOP); wqe->flags = cpu_to_be32(MODQP_WRITE_EC); sge_cmd = (3ULL << 56) | FW_RI_SGEEC_START << 8 | 3; wqe->sge_cmd = cpu_to_be64(sge_cmd); @@ -674,7 +676,7 @@ static int cxio_hal_ctrl_qp_write_mem(struct cxio_rdev *rdev_p, u32 addr, build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_BP, flag, Q_GENBIT(rdev_p->ctrl_qp.wptr, T3_CTRL_QP_SIZE_LOG2), T3_CTRL_QP_ID, - wr_len); + wr_len, T3_SOPEOP); if (flag == T3_COMPLETION_FLAG) ring_doorbell(rdev_p->ctrl_qp.doorbell, T3_CTRL_QP_ID); len -= 96; @@ -816,6 +818,13 @@ int cxio_deallocate_window(struct cxio_rdev *rdev_p, u32 stag) 0, 0); } +int cxio_allocate_stag(struct cxio_rdev *rdev_p, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr) +{ + *stag = T3_STAG_UNSET; + return __cxio_tpt_op(rdev_p, 0, stag, 0, pdid, TPT_NON_SHARED_MR, + 0, 0, 0ULL, 0, 0, pbl_size, pbl_addr); +} + int cxio_rdma_init(struct cxio_rdev *rdev_p, struct t3_rdma_init_attr *attr) { struct t3_rdma_init_wr *wqe; diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h index 6e128f6bab0..25a880664e6 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.h +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h @@ -165,6 +165,7 @@ int cxio_reregister_phys_mem(struct cxio_rdev *rdev, u32 * stag, u32 pdid, int cxio_dereg_mem(struct cxio_rdev *rdev, u32 stag, u32 pbl_size, u32 pbl_addr); int cxio_allocate_window(struct cxio_rdev *rdev, u32 * stag, u32 pdid); +int cxio_allocate_stag(struct cxio_rdev *rdev, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr); int cxio_deallocate_window(struct cxio_rdev *rdev, u32 stag); int cxio_rdma_init(struct cxio_rdev *rdev, struct t3_rdma_init_attr *attr); void cxio_register_ev_cb(cxio_hal_ev_callback_func_t ev_cb); diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h index f1a25a821a4..de760e9f1cc 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_wr.h +++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h @@ -72,7 +72,8 @@ enum t3_wr_opcode { T3_WR_BIND = FW_WROPCODE_RI_BIND_MW, T3_WR_RCV = FW_WROPCODE_RI_RECEIVE, T3_WR_INIT = FW_WROPCODE_RI_RDMA_INIT, - T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP + T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP, + T3_WR_FASTREG = FW_WROPCODE_RI_FASTREGISTER_MR } __attribute__ ((packed)); enum t3_rdma_opcode { @@ -89,7 +90,8 @@ enum t3_rdma_opcode { T3_FAST_REGISTER, T3_LOCAL_INV, T3_QP_MOD, - T3_BYPASS + T3_BYPASS, + T3_RDMA_READ_REQ_WITH_INV, } __attribute__ ((packed)); static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop) @@ -103,6 +105,7 @@ static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop) case T3_WR_BIND: return T3_BIND_MW; case T3_WR_INIT: return T3_RDMA_INIT; case T3_WR_QP_MOD: return T3_QP_MOD; + case T3_WR_FASTREG: return T3_FAST_REGISTER; default: break; } return -1; @@ -170,11 +173,54 @@ struct t3_send_wr { struct t3_sge sgl[T3_MAX_SGE]; /* 4+ */ }; +#define T3_MAX_FASTREG_DEPTH 24 +#define T3_MAX_FASTREG_FRAG 10 + +struct t3_fastreg_wr { + struct fw_riwrh wrh; /* 0 */ + union t3_wrid wrid; /* 1 */ + __be32 stag; /* 2 */ + __be32 len; + __be32 va_base_hi; /* 3 */ + __be32 va_base_lo_fbo; + __be32 page_type_perms; /* 4 */ + __be32 reserved1; + __be64 pbl_addrs[0]; /* 5+ */ +}; + +/* + * If a fastreg wr spans multiple wqes, then the 2nd fragment look like this. + */ +struct t3_pbl_frag { + struct fw_riwrh wrh; /* 0 */ + __be64 pbl_addrs[14]; /* 1..14 */ +}; + +#define S_FR_PAGE_COUNT 24 +#define M_FR_PAGE_COUNT 0xff +#define V_FR_PAGE_COUNT(x) ((x) << S_FR_PAGE_COUNT) +#define G_FR_PAGE_COUNT(x) ((((x) >> S_FR_PAGE_COUNT)) & M_FR_PAGE_COUNT) + +#define S_FR_PAGE_SIZE 16 +#define M_FR_PAGE_SIZE 0x1f +#define V_FR_PAGE_SIZE(x) ((x) << S_FR_PAGE_SIZE) +#define G_FR_PAGE_SIZE(x) ((((x) >> S_FR_PAGE_SIZE)) & M_FR_PAGE_SIZE) + +#define S_FR_TYPE 8 +#define M_FR_TYPE 0x1 +#define V_FR_TYPE(x) ((x) << S_FR_TYPE) +#define G_FR_TYPE(x) ((((x) >> S_FR_TYPE)) & M_FR_TYPE) + +#define S_FR_PERMS 0 +#define M_FR_PERMS 0xff +#define V_FR_PERMS(x) ((x) << S_FR_PERMS) +#define G_FR_PERMS(x) ((((x) >> S_FR_PERMS)) & M_FR_PERMS) + struct t3_local_inv_wr { struct fw_riwrh wrh; /* 0 */ union t3_wrid wrid; /* 1 */ __be32 stag; /* 2 */ - __be32 reserved3; + __be32 reserved; }; struct t3_rdma_write_wr { @@ -193,7 +239,8 @@ struct t3_rdma_read_wr { struct fw_riwrh wrh; /* 0 */ union t3_wrid wrid; /* 1 */ u8 rdmaop; /* 2 */ - u8 reserved[3]; + u8 local_inv; + u8 reserved[2]; __be32 rem_stag; __be64 rem_to; /* 3 */ __be32 local_stag; /* 4 */ @@ -201,18 +248,6 @@ struct t3_rdma_read_wr { __be64 local_to; /* 5 */ }; -enum t3_addr_type { - T3_VA_BASED_TO = 0x0, - T3_ZERO_BASED_TO = 0x1 -} __attribute__ ((packed)); - -enum t3_mem_perms { - T3_MEM_ACCESS_LOCAL_READ = 0x1, - T3_MEM_ACCESS_LOCAL_WRITE = 0x2, - T3_MEM_ACCESS_REM_READ = 0x4, - T3_MEM_ACCESS_REM_WRITE = 0x8 -} __attribute__ ((packed)); - struct t3_bind_mw_wr { struct fw_riwrh wrh; /* 0 */ union t3_wrid wrid; /* 1 */ @@ -336,6 +371,11 @@ struct t3_genbit { __be64 genbit; }; +struct t3_wq_in_err { + u64 flit[13]; + u64 err; +}; + enum rdma_init_wr_flags { MPA_INITIATOR = (1<<0), PRIV_QP = (1<<1), @@ -346,13 +386,16 @@ union t3_wr { struct t3_rdma_write_wr write; struct t3_rdma_read_wr read; struct t3_receive_wr recv; + struct t3_fastreg_wr fastreg; + struct t3_pbl_frag pbl_frag; struct t3_local_inv_wr local_inv; struct t3_bind_mw_wr bind; struct t3_bypass_wr bypass; struct t3_rdma_init_wr init; struct t3_modify_qp_wr qp_mod; struct t3_genbit genbit; - u64 flit[16]; + struct t3_wq_in_err wq_in_err; + __be64 flit[16]; }; #define T3_SQ_CQE_FLIT 13 @@ -366,12 +409,18 @@ static inline enum t3_wr_opcode fw_riwrh_opcode(struct fw_riwrh *wqe) return G_FW_RIWR_OP(be32_to_cpu(wqe->op_seop_flags)); } +enum t3_wr_hdr_bits { + T3_EOP = 1, + T3_SOP = 2, + T3_SOPEOP = T3_EOP|T3_SOP, +}; + static inline void build_fw_riwrh(struct fw_riwrh *wqe, enum t3_wr_opcode op, enum t3_wr_flags flags, u8 genbit, u32 tid, - u8 len) + u8 len, u8 sopeop) { wqe->op_seop_flags = cpu_to_be32(V_FW_RIWR_OP(op) | - V_FW_RIWR_SOPEOP(M_FW_RIWR_SOPEOP) | + V_FW_RIWR_SOPEOP(sopeop) | V_FW_RIWR_FLAGS(flags)); wmb(); wqe->gen_tid_len = cpu_to_be32(V_FW_RIWR_GEN(genbit) | @@ -404,6 +453,7 @@ enum tpt_addr_type { }; enum tpt_mem_perm { + TPT_MW_BIND = 0x10, TPT_LOCAL_READ = 0x8, TPT_LOCAL_WRITE = 0x4, TPT_REMOTE_READ = 0x2, @@ -659,7 +709,7 @@ struct t3_cq { static inline void cxio_set_wq_in_error(struct t3_wq *wq) { - wq->queue->flit[13] = 1; + wq->queue->wq_in_err.err = 1; } static inline struct t3_cqe *cxio_next_hw_cqe(struct t3_cq *cq) diff --git a/drivers/infiniband/hw/cxgb3/iwch_cq.c b/drivers/infiniband/hw/cxgb3/iwch_cq.c index 4ee8ccd0a9e..cf5474ae68f 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_cq.c +++ b/drivers/infiniband/hw/cxgb3/iwch_cq.c @@ -81,6 +81,7 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp, wc->wr_id = cookie; wc->qp = &qhp->ibqp; wc->vendor_err = CQE_STATUS(cqe); + wc->wc_flags = 0; PDBG("%s qpid 0x%x type %d opcode %d status 0x%x wrid hi 0x%x " "lo 0x%x cookie 0x%llx\n", __func__, @@ -94,6 +95,11 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp, else wc->byte_len = 0; wc->opcode = IB_WC_RECV; + if (CQE_OPCODE(cqe) == T3_SEND_WITH_INV || + CQE_OPCODE(cqe) == T3_SEND_WITH_SE_INV) { + wc->ex.invalidate_rkey = CQE_WRID_STAG(cqe); + wc->wc_flags |= IB_WC_WITH_INVALIDATE; + } } else { switch (CQE_OPCODE(cqe)) { case T3_RDMA_WRITE: @@ -105,17 +111,20 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp, break; case T3_SEND: case T3_SEND_WITH_SE: + case T3_SEND_WITH_INV: + case T3_SEND_WITH_SE_INV: wc->opcode = IB_WC_SEND; break; case T3_BIND_MW: wc->opcode = IB_WC_BIND_MW; break; - /* these aren't supported yet */ - case T3_SEND_WITH_INV: - case T3_SEND_WITH_SE_INV: case T3_LOCAL_INV: + wc->opcode = IB_WC_LOCAL_INV; + break; case T3_FAST_REGISTER: + wc->opcode = IB_WC_FAST_REG_MR; + break; default: printk(KERN_ERR MOD "Unexpected opcode %d " "in the CQE received for QPID=0x%0x\n", diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 95f82cfb6c5..5d504f3ed68 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -768,6 +768,68 @@ static int iwch_dealloc_mw(struct ib_mw *mw) return 0; } +static struct ib_mr *iwch_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth) +{ + struct iwch_dev *rhp; + struct iwch_pd *php; + struct iwch_mr *mhp; + u32 mmid; + u32 stag = 0; + int ret; + + php = to_iwch_pd(pd); + rhp = php->rhp; + mhp = kzalloc(sizeof(*mhp), GFP_KERNEL); + if (!mhp) + return ERR_PTR(-ENOMEM); + + mhp->rhp = rhp; + ret = iwch_alloc_pbl(mhp, pbl_depth); + if (ret) { + kfree(mhp); + return ERR_PTR(ret); + } + mhp->attr.pbl_size = pbl_depth; + ret = cxio_allocate_stag(&rhp->rdev, &stag, php->pdid, + mhp->attr.pbl_size, mhp->attr.pbl_addr); + if (ret) { + iwch_free_pbl(mhp); + kfree(mhp); + return ERR_PTR(ret); + } + mhp->attr.pdid = php->pdid; + mhp->attr.type = TPT_NON_SHARED_MR; + mhp->attr.stag = stag; + mhp->attr.state = 1; + mmid = (stag) >> 8; + mhp->ibmr.rkey = mhp->ibmr.lkey = stag; + insert_handle(rhp, &rhp->mmidr, mhp, mmid); + PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag); + return &(mhp->ibmr); +} + +static struct ib_fast_reg_page_list *iwch_alloc_fastreg_pbl( + struct ib_device *device, + int page_list_len) +{ + struct ib_fast_reg_page_list *page_list; + + page_list = kmalloc(sizeof *page_list + page_list_len * sizeof(u64), + GFP_KERNEL); + if (!page_list) + return ERR_PTR(-ENOMEM); + + page_list->page_list = (u64 *)(page_list + 1); + page_list->max_page_list_len = page_list_len; + + return page_list; +} + +static void iwch_free_fastreg_pbl(struct ib_fast_reg_page_list *page_list) +{ + kfree(page_list); +} + static int iwch_destroy_qp(struct ib_qp *ib_qp) { struct iwch_dev *rhp; @@ -843,6 +905,15 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd, */ sqsize = roundup_pow_of_two(attrs->cap.max_send_wr); wqsize = roundup_pow_of_two(rqsize + sqsize); + + /* + * Kernel users need more wq space for fastreg WRs which can take + * 2 WR fragments. + */ + ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL; + if (!ucontext && wqsize < (rqsize + (2 * sqsize))) + wqsize = roundup_pow_of_two(rqsize + + roundup_pow_of_two(attrs->cap.max_send_wr * 2)); PDBG("%s wqsize %d sqsize %d rqsize %d\n", __func__, wqsize, sqsize, rqsize); qhp = kzalloc(sizeof(*qhp), GFP_KERNEL); @@ -851,7 +922,6 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd, qhp->wq.size_log2 = ilog2(wqsize); qhp->wq.rq_size_log2 = ilog2(rqsize); qhp->wq.sq_size_log2 = ilog2(sqsize); - ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL; if (cxio_create_qp(&rhp->rdev, !udata, &qhp->wq, ucontext ? &ucontext->uctx : &rhp->rdev.uctx)) { kfree(qhp); @@ -1048,6 +1118,7 @@ static int iwch_query_device(struct ib_device *ibdev, props->max_mr = dev->attr.max_mem_regs; props->max_pd = dev->attr.max_pds; props->local_ca_ack_delay = 0; + props->max_fast_reg_page_list_len = T3_MAX_FASTREG_DEPTH; return 0; } @@ -1088,6 +1159,28 @@ static ssize_t show_rev(struct device *dev, struct device_attribute *attr, return sprintf(buf, "%d\n", iwch_dev->rdev.t3cdev_p->type); } +static int fw_supports_fastreg(struct iwch_dev *iwch_dev) +{ + struct ethtool_drvinfo info; + struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev; + char *cp, *next; + unsigned fw_maj, fw_min; + + rtnl_lock(); + lldev->ethtool_ops->get_drvinfo(lldev, &info); + rtnl_unlock(); + + next = info.fw_version+1; + cp = strsep(&next, "."); + sscanf(cp, "%i", &fw_maj); + cp = strsep(&next, "."); + sscanf(cp, "%i", &fw_min); + + PDBG("%s maj %u min %u\n", __func__, fw_maj, fw_min); + + return fw_maj > 6 || (fw_maj == 6 && fw_min > 0); +} + static ssize_t show_fw_ver(struct device *dev, struct device_attribute *attr, char *buf) { struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev, @@ -1149,8 +1242,10 @@ int iwch_register_device(struct iwch_dev *dev) memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid)); memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6); dev->ibdev.owner = THIS_MODULE; - dev->device_cap_flags = - (IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW); + dev->device_cap_flags = IB_DEVICE_ZERO_STAG | + IB_DEVICE_MEM_WINDOW; + if (fw_supports_fastreg(dev)) + dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS; dev->ibdev.uverbs_cmd_mask = (1ull << IB_USER_VERBS_CMD_GET_CONTEXT) | @@ -1202,6 +1297,9 @@ int iwch_register_device(struct iwch_dev *dev) dev->ibdev.alloc_mw = iwch_alloc_mw; dev->ibdev.bind_mw = iwch_bind_mw; dev->ibdev.dealloc_mw = iwch_dealloc_mw; + dev->ibdev.alloc_fast_reg_mr = iwch_alloc_fast_reg_mr; + dev->ibdev.alloc_fast_reg_page_list = iwch_alloc_fastreg_pbl; + dev->ibdev.free_fast_reg_page_list = iwch_free_fastreg_pbl; dev->ibdev.attach_mcast = iwch_multicast_attach; dev->ibdev.detach_mcast = iwch_multicast_detach; diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.h b/drivers/infiniband/hw/cxgb3/iwch_provider.h index 836163fc542..f5ceca05c43 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.h +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.h @@ -296,14 +296,6 @@ static inline u32 iwch_ib_to_tpt_access(int acc) TPT_LOCAL_READ; } -static inline u32 iwch_ib_to_mwbind_access(int acc) -{ - return (acc & IB_ACCESS_REMOTE_WRITE ? T3_MEM_ACCESS_REM_WRITE : 0) | - (acc & IB_ACCESS_REMOTE_READ ? T3_MEM_ACCESS_REM_READ : 0) | - (acc & IB_ACCESS_LOCAL_WRITE ? T3_MEM_ACCESS_LOCAL_WRITE : 0) | - T3_MEM_ACCESS_LOCAL_READ; -} - enum iwch_mmid_state { IWCH_STAG_STATE_VALID, IWCH_STAG_STATE_INVALID diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c index 99261379922..3b44300a303 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_qp.c +++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c @@ -44,54 +44,39 @@ static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr, switch (wr->opcode) { case IB_WR_SEND: - case IB_WR_SEND_WITH_IMM: if (wr->send_flags & IB_SEND_SOLICITED) wqe->send.rdmaop = T3_SEND_WITH_SE; else wqe->send.rdmaop = T3_SEND; wqe->send.rem_stag = 0; break; -#if 0 /* Not currently supported */ - case TYPE_SEND_INVALIDATE: - case TYPE_SEND_INVALIDATE_IMMEDIATE: - wqe->send.rdmaop = T3_SEND_WITH_INV; - wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey); - break; - case TYPE_SEND_SE_INVALIDATE: - wqe->send.rdmaop = T3_SEND_WITH_SE_INV; - wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey); + case IB_WR_SEND_WITH_INV: + if (wr->send_flags & IB_SEND_SOLICITED) + wqe->send.rdmaop = T3_SEND_WITH_SE_INV; + else + wqe->send.rdmaop = T3_SEND_WITH_INV; + wqe->send.rem_stag = cpu_to_be32(wr->ex.invalidate_rkey); break; -#endif default: - break; + return -EINVAL; } if (wr->num_sge > T3_MAX_SGE) return -EINVAL; wqe->send.reserved[0] = 0; wqe->send.reserved[1] = 0; wqe->send.reserved[2] = 0; - if (wr->opcode == IB_WR_SEND_WITH_IMM) { - plen = 4; - wqe->send.sgl[0].stag = wr->ex.imm_data; - wqe->send.sgl[0].len = __constant_cpu_to_be32(0); - wqe->send.num_sgle = __constant_cpu_to_be32(0); - *flit_cnt = 5; - } else { - plen = 0; - for (i = 0; i < wr->num_sge; i++) { - if ((plen + wr->sg_list[i].length) < plen) { - return -EMSGSIZE; - } - plen += wr->sg_list[i].length; - wqe->send.sgl[i].stag = - cpu_to_be32(wr->sg_list[i].lkey); - wqe->send.sgl[i].len = - cpu_to_be32(wr->sg_list[i].length); - wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr); - } - wqe->send.num_sgle = cpu_to_be32(wr->num_sge); - *flit_cnt = 4 + ((wr->num_sge) << 1); + plen = 0; + for (i = 0; i < wr->num_sge; i++) { + if ((plen + wr->sg_list[i].length) < plen) + return -EMSGSIZE; + + plen += wr->sg_list[i].length; + wqe->send.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey); + wqe->send.sgl[i].len = cpu_to_be32(wr->sg_list[i].length); + wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr); } + wqe->send.num_sgle = cpu_to_be32(wr->num_sge); + *flit_cnt = 4 + ((wr->num_sge) << 1); wqe->send.plen = cpu_to_be32(plen); return 0; } @@ -143,9 +128,12 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr, if (wr->num_sge > 1) return -EINVAL; wqe->read.rdmaop = T3_READ_REQ; + if (wr->opcode == IB_WR_RDMA_READ_WITH_INV) + wqe->read.local_inv = 1; + else + wqe->read.local_inv = 0; wqe->read.reserved[0] = 0; wqe->read.reserved[1] = 0; - wqe->read.reserved[2] = 0; wqe->read.rem_stag = cpu_to_be32(wr->wr.rdma.rkey); wqe->read.rem_to = cpu_to_be64(wr->wr.rdma.remote_addr); wqe->read.local_stag = cpu_to_be32(wr->sg_list[0].lkey); @@ -155,6 +143,57 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr, return 0; } +static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr, + u8 *flit_cnt, int *wr_cnt, struct t3_wq *wq) +{ + int i; + __be64 *p; + + if (wr->wr.fast_reg.page_list_len > T3_MAX_FASTREG_DEPTH) + return -EINVAL; + *wr_cnt = 1; + wqe->fastreg.stag = cpu_to_be32(wr->wr.fast_reg.rkey); + wqe->fastreg.len = cpu_to_be32(wr->wr.fast_reg.length); + wqe->fastreg.va_base_hi = cpu_to_be32(wr->wr.fast_reg.iova_start >> 32); + wqe->fastreg.va_base_lo_fbo = + cpu_to_be32(wr->wr.fast_reg.iova_start & 0xffffffff); + wqe->fastreg.page_type_perms = cpu_to_be32( + V_FR_PAGE_COUNT(wr->wr.fast_reg.page_list_len) | + V_FR_PAGE_SIZE(wr->wr.fast_reg.page_shift-12) | + V_FR_TYPE(TPT_VATO) | + V_FR_PERMS(iwch_ib_to_tpt_access(wr->wr.fast_reg.access_flags))); + p = &wqe->fastreg.pbl_addrs[0]; + for (i = 0; i < wr->wr.fast_reg.page_list_len; i++, p++) { + + /* If we need a 2nd WR, then set it up */ + if (i == T3_MAX_FASTREG_FRAG) { + *wr_cnt = 2; + wqe = (union t3_wr *)(wq->queue + + Q_PTR2IDX((wq->wptr+1), wq->size_log2)); + build_fw_riwrh((void *)wqe, T3_WR_FASTREG, 0, + Q_GENBIT(wq->wptr + 1, wq->size_log2), + 0, 1 + wr->wr.fast_reg.page_list_len - T3_MAX_FASTREG_FRAG, + T3_EOP); + + p = &wqe->pbl_frag.pbl_addrs[0]; + } + *p = cpu_to_be64((u64)wr->wr.fast_reg.page_list->page_list[i]); + } + *flit_cnt = 5 + wr->wr.fast_reg.page_list_len; + if (*flit_cnt > 15) + *flit_cnt = 15; + return 0; +} + +static int iwch_build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr, + u8 *flit_cnt) +{ + wqe->local_inv.stag = cpu_to_be32(wr->ex.invalidate_rkey); + wqe->local_inv.reserved = 0; + *flit_cnt = sizeof(struct t3_local_inv_wr) >> 3; + return 0; +} + /* * TBD: this is going to be moved to firmware. Missing pdid/qpid check for now. */ @@ -238,6 +277,7 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, u32 num_wrs; unsigned long flag; struct t3_swsq *sqp; + int wr_cnt = 1; qhp = to_iwch_qp(ibqp); spin_lock_irqsave(&qhp->lock, flag); @@ -262,15 +302,15 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, t3_wr_flags = 0; if (wr->send_flags & IB_SEND_SOLICITED) t3_wr_flags |= T3_SOLICITED_EVENT_FLAG; - if (wr->send_flags & IB_SEND_FENCE) - t3_wr_flags |= T3_READ_FENCE_FLAG; if (wr->send_flags & IB_SEND_SIGNALED) t3_wr_flags |= T3_COMPLETION_FLAG; sqp = qhp->wq.sq + Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2); switch (wr->opcode) { case IB_WR_SEND: - case IB_WR_SEND_WITH_IMM: + case IB_WR_SEND_WITH_INV: + if (wr->send_flags & IB_SEND_FENCE) + t3_wr_flags |= T3_READ_FENCE_FLAG; t3_wr_opcode = T3_WR_SEND; err = iwch_build_rdma_send(wqe, wr, &t3_wr_flit_cnt); break; @@ -280,6 +320,7 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, err = iwch_build_rdma_write(wqe, wr, &t3_wr_flit_cnt); break; case IB_WR_RDMA_READ: + case IB_WR_RDMA_READ_WITH_INV: t3_wr_opcode = T3_WR_READ; t3_wr_flags = 0; /* T3 reads are always signaled */ err = iwch_build_rdma_read(wqe, wr, &t3_wr_flit_cnt); @@ -289,6 +330,17 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, if (!qhp->wq.oldest_read) qhp->wq.oldest_read = sqp; break; + case IB_WR_FAST_REG_MR: + t3_wr_opcode = T3_WR_FASTREG; + err = iwch_build_fastreg(wqe, wr, &t3_wr_flit_cnt, + &wr_cnt, &qhp->wq); + break; + case IB_WR_LOCAL_INV: + if (wr->send_flags & IB_SEND_FENCE) + t3_wr_flags |= T3_LOCAL_FENCE_FLAG; + t3_wr_opcode = T3_WR_INV_STAG; + err = iwch_build_inv_stag(wqe, wr, &t3_wr_flit_cnt); + break; default: PDBG("%s post of type=%d TBD!\n", __func__, wr->opcode); @@ -307,14 +359,15 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, build_fw_riwrh((void *) wqe, t3_wr_opcode, t3_wr_flags, Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), - 0, t3_wr_flit_cnt); + 0, t3_wr_flit_cnt, + (wr_cnt == 1) ? T3_SOPEOP : T3_SOP); PDBG("%s cookie 0x%llx wq idx 0x%x swsq idx %ld opcode %d\n", __func__, (unsigned long long) wr->wr_id, idx, Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2), sqp->opcode); wr = wr->next; num_wrs--; - ++(qhp->wq.wptr); + qhp->wq.wptr += wr_cnt; ++(qhp->wq.sq_wptr); } spin_unlock_irqrestore(&qhp->lock, flag); @@ -359,7 +412,7 @@ int iwch_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr, wr->wr_id; build_fw_riwrh((void *) wqe, T3_WR_RCV, T3_COMPLETION_FLAG, Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), - 0, sizeof(struct t3_receive_wr) >> 3); + 0, sizeof(struct t3_receive_wr) >> 3, T3_SOPEOP); PDBG("%s cookie 0x%llx idx 0x%x rq_wptr 0x%x rw_rptr 0x%x " "wqe %p \n", __func__, (unsigned long long) wr->wr_id, idx, qhp->wq.rq_wptr, qhp->wq.rq_rptr, wqe); @@ -419,10 +472,10 @@ int iwch_bind_mw(struct ib_qp *qp, sgl.lkey = mw_bind->mr->lkey; sgl.length = mw_bind->length; wqe->bind.reserved = 0; - wqe->bind.type = T3_VA_BASED_TO; + wqe->bind.type = TPT_VATO; /* TBD: check perms */ - wqe->bind.perms = iwch_ib_to_mwbind_access(mw_bind->mw_access_flags); + wqe->bind.perms = iwch_ib_to_tpt_access(mw_bind->mw_access_flags); wqe->bind.mr_stag = cpu_to_be32(mw_bind->mr->lkey); wqe->bind.mw_stag = cpu_to_be32(mw->rkey); wqe->bind.mw_len = cpu_to_be32(mw_bind->length); @@ -430,7 +483,7 @@ int iwch_bind_mw(struct ib_qp *qp, err = iwch_sgl2pbl_map(rhp, &sgl, 1, &pbl_addr, &page_size); if (err) { spin_unlock_irqrestore(&qhp->lock, flag); - return err; + return err; } wqe->send.wrid.id0.hi = qhp->wq.sq_wptr; sqp = qhp->wq.sq + Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2); @@ -441,10 +494,9 @@ int iwch_bind_mw(struct ib_qp *qp, sqp->signaled = (mw_bind->send_flags & IB_SEND_SIGNALED); wqe->bind.mr_pbl_addr = cpu_to_be32(pbl_addr); wqe->bind.mr_pagesz = page_size; - wqe->flit[T3_SQ_COOKIE_FLIT] = mw_bind->wr_id; build_fw_riwrh((void *)wqe, T3_WR_BIND, t3_wr_flags, Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), 0, - sizeof(struct t3_bind_mw_wr) >> 3); + sizeof(struct t3_bind_mw_wr) >> 3, T3_SOPEOP); ++(qhp->wq.wptr); ++(qhp->wq.sq_wptr); spin_unlock_irqrestore(&qhp->lock, flag); -- cgit v1.2.3 From d1f2cd895f8733faa9d79d09d825a2ed80002ac7 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:45 -0700 Subject: IB/mlx4: Configure QPs' max message size based on real device capability ConnectX returns the max message size it supports through the QUERY_DEV_CAP firmware command. When modifying a QP to RTR, the max message size for the QP must be specified. This value must not exceed the value declared through QUERY_DEV_CAP. The current code ignores the max allowed size and unconditionally sets the value to 2^31. This patch sets all QPs to the max value allowed as returned from firmware. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/qp.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 4b0ac5d68c4..76054a72f71 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -908,7 +908,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, attr->path_mtu); goto out; } - context->mtu_msgmax = (attr->path_mtu << 5) | 31; + context->mtu_msgmax = (attr->path_mtu << 5) | + ilog2(dev->dev->caps.max_msg_sz); } if (qp->rq.wqe_cnt) -- cgit v1.2.3 From 6578cf33989a594bab37af988d45d87812b946b8 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:45 -0700 Subject: IB/mlx4: Pass congestion management class MADs to the HCA ConnectX HCAs support the IB_MGMT_CLASS_CONG_MGMT management class, so process MADs of this class through the MAD_IFC firmware command. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/mad.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 4c1e72fc8f5..cdca3a511e1 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -255,7 +255,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, return IB_MAD_RESULT_SUCCESS; } else if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT || in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS1 || - in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2) { + in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2 || + in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_CONG_MGMT) { if (in_mad->mad_hdr.method != IB_MGMT_METHOD_GET && in_mad->mad_hdr.method != IB_MGMT_METHOD_SET) return IB_MAD_RESULT_SUCCESS; -- cgit v1.2.3 From e5a5e7d59af5944a674b9cea420a1fedc60496f2 Mon Sep 17 00:00:00 2001 From: Ralph Campbell Date: Mon, 14 Jul 2008 23:48:46 -0700 Subject: IB/core: Reset to error QP state transition is not allowed I was reviewing the QP state transition diagram in the IB 1.2.1 spec and the code for qp_state_table[], and noticed that the code allows a QP to be modified from IB_QPS_RESET to IB_QPS_ERR whereas the notes for figure 124 (pg 457) specifically says that this transition isn't allowed. This is a clarification from earlier versions of the IB spec, which were ambiguous in this area and suggested that the RESET to ERR transition was allowed. Fix up the qp_state_table[] to make RESET->ERR not allowed. Signed-off-by: Ralph Campbell Signed-off-by: Roland Dreier --- drivers/infiniband/core/verbs.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index e0fbe597586..a7da9be43e6 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -315,7 +315,6 @@ static const struct { } qp_state_table[IB_QPS_ERR + 1][IB_QPS_ERR + 1] = { [IB_QPS_RESET] = { [IB_QPS_RESET] = { .valid = 1 }, - [IB_QPS_ERR] = { .valid = 1 }, [IB_QPS_INIT] = { .valid = 1, .req_param = { -- cgit v1.2.3 From d3809ad0972297fbc7ef0585049ef465d9d8d79d Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:46 -0700 Subject: IB/mthca: Remove extra code for RESET->ERR QP state transition Commit b18aad71 ("IB/mthca: Fix RESET to ERROR transition") added some extra code to handle a QP state transition from RESET to ERROR. However, the latest 1.2.1 version of the IB spec has clarified that this transition is actually not allowed, so we can remove this extra code again. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_qp.c | 26 -------------------------- 1 file changed, 26 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index 3b1c5baf13b..77b52b3adcc 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -848,23 +848,6 @@ out: return err; } -static const struct ib_qp_attr dummy_init_attr = { .port_num = 1 }; -static const int dummy_init_attr_mask[] = { - [IB_QPT_UD] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_QKEY), - [IB_QPT_UC] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_ACCESS_FLAGS), - [IB_QPT_RC] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_ACCESS_FLAGS), - [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | - IB_QP_QKEY), - [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | - IB_QP_QKEY), -}; - int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask, struct ib_udata *udata) { @@ -926,15 +909,6 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask, goto out; } - if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) { - err = __mthca_modify_qp(ibqp, &dummy_init_attr, - dummy_init_attr_mask[ibqp->qp_type], - IB_QPS_RESET, IB_QPS_INIT); - if (err) - goto out; - cur_state = IB_QPS_INIT; - } - err = __mthca_modify_qp(ibqp, attr, attr_mask, cur_state, new_state); out: -- cgit v1.2.3 From 7c27f358209a8ce7c57b584346d7b611e823f1b1 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:46 -0700 Subject: IB/mlx4: Remove extra code for RESET->ERR QP state transition Commit 65adfa91 ("IB/mlx4: Fix RESET to RESET and RESET to ERROR transitions") added some extra code to handle a QP state transition from RESET to ERROR. However, the latest 1.2.1 version of the IB spec has clarified that this transition is actually not allowed, so we can remove this extra code again. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/qp.c | 26 -------------------------- 1 file changed, 26 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 76054a72f71..44bbd6c2e31 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -1132,23 +1132,6 @@ out: return err; } -static const struct ib_qp_attr mlx4_ib_qp_attr = { .port_num = 1 }; -static const int mlx4_ib_qp_attr_mask_table[IB_QPT_UD + 1] = { - [IB_QPT_UD] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_QKEY), - [IB_QPT_UC] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_ACCESS_FLAGS), - [IB_QPT_RC] = (IB_QP_PKEY_INDEX | - IB_QP_PORT | - IB_QP_ACCESS_FLAGS), - [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | - IB_QP_QKEY), - [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | - IB_QP_QKEY), -}; - int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask, struct ib_udata *udata) { @@ -1191,15 +1174,6 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, goto out; } - if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) { - err = __mlx4_ib_modify_qp(ibqp, &mlx4_ib_qp_attr, - mlx4_ib_qp_attr_mask_table[ibqp->qp_type], - IB_QPS_RESET, IB_QPS_INIT); - if (err) - goto out; - cur_state = IB_QPS_INIT; - } - err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state); out: -- cgit v1.2.3 From 3e255eac561672cbc92844b9f16cae9304c2a783 Mon Sep 17 00:00:00 2001 From: Joachim Fenkes Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: IB/ehca: Reject receive work requests if QP is in RESET state Signed-off-by: Joachim Fenkes Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ehca/ehca_reqs.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c index b799b271021..dd9bc68f1c7 100644 --- a/drivers/infiniband/hw/ehca/ehca_reqs.c +++ b/drivers/infiniband/hw/ehca/ehca_reqs.c @@ -544,8 +544,16 @@ int ehca_post_recv(struct ib_qp *qp, struct ib_recv_wr *recv_wr, struct ib_recv_wr **bad_recv_wr) { - return internal_post_recv(container_of(qp, struct ehca_qp, ib_qp), - qp->device, recv_wr, bad_recv_wr); + struct ehca_qp *my_qp = container_of(qp, struct ehca_qp, ib_qp); + + /* Reject WR if QP is in RESET state */ + if (unlikely(my_qp->state == IB_QPS_RESET)) { + ehca_err(qp->device, "Invalid QP state qp_state=%d qpn=%x", + my_qp->state, qp->qp_num); + return -EINVAL; + } + + return internal_post_recv(my_qp, qp->device, recv_wr, bad_recv_wr); } int ehca_post_srq_recv(struct ib_srq *srq, -- cgit v1.2.3 From 6f7bc01a7382641c61ec036d68ff3a9140b48a1c Mon Sep 17 00:00:00 2001 From: Stefan Roscher Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: IB/ehca: In case of lost interrupts, trigger EOI to reenable interrupts During corner case testing, we noticed that some versions of ehca do not properly transition to interrupt done in special load situations. This can be resolved by periodically triggering EOI through H_EOI, if EQEs are pending. Signed-off-by: Stefan Roscher Acked-by: Benjamin Herrenschmidt Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ehca/ehca_irq.c | 9 +++++++-- drivers/infiniband/hw/ehca/hcp_if.c | 10 ++++++++++ drivers/infiniband/hw/ehca/hcp_if.h | 1 + 3 files changed, 18 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c index ce1ab0571be..0792d930c48 100644 --- a/drivers/infiniband/hw/ehca/ehca_irq.c +++ b/drivers/infiniband/hw/ehca/ehca_irq.c @@ -531,7 +531,7 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq) { struct ehca_eq *eq = &shca->eq; struct ehca_eqe_cache_entry *eqe_cache = eq->eqe_cache; - u64 eqe_value; + u64 eqe_value, ret; unsigned long flags; int eqe_cnt, i; int eq_empty = 0; @@ -583,8 +583,13 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq) ehca_dbg(&shca->ib_device, "No eqe found for irq event"); goto unlock_irq_spinlock; - } else if (!is_irq) + } else if (!is_irq) { + ret = hipz_h_eoi(eq->ist); + if (ret != H_SUCCESS) + ehca_err(&shca->ib_device, + "bad return code EOI -rc = %ld\n", ret); ehca_dbg(&shca->ib_device, "deadman found %x eqe", eqe_cnt); + } if (unlikely(eqe_cnt == EHCA_EQE_CACHE_SIZE)) ehca_dbg(&shca->ib_device, "too many eqes for one irq event"); /* enable irq for new packets */ diff --git a/drivers/infiniband/hw/ehca/hcp_if.c b/drivers/infiniband/hw/ehca/hcp_if.c index 5245e13c3a3..415d3a465de 100644 --- a/drivers/infiniband/hw/ehca/hcp_if.c +++ b/drivers/infiniband/hw/ehca/hcp_if.c @@ -933,3 +933,13 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle, r_cb, 0, 0, 0, 0); } + +u64 hipz_h_eoi(int irq) +{ + unsigned long xirr; + + iosync(); + xirr = (0xffULL << 24) | irq; + + return plpar_hcall_norets(H_EOI, xirr); +} diff --git a/drivers/infiniband/hw/ehca/hcp_if.h b/drivers/infiniband/hw/ehca/hcp_if.h index 60ce02b7066..2c3c6e0ea5c 100644 --- a/drivers/infiniband/hw/ehca/hcp_if.h +++ b/drivers/infiniband/hw/ehca/hcp_if.h @@ -260,5 +260,6 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle, const u64 ressource_handle, void *rblock, unsigned long *byte_count); +u64 hipz_h_eoi(int irq); #endif /* __HCP_IF_H__ */ -- cgit v1.2.3 From 97d1cc8055c7b3fbd35bf693775d61102e65d174 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: RDMA/cxgb3: Fix up some ib_device_attr fields - set fw_ver - set hw_ver - set max_qp_wr to something reasonable - set max_cqe to something reasonable Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/cxio_hal.h | 3 ++- drivers/infiniband/hw/cxgb3/iwch.c | 4 ++-- drivers/infiniband/hw/cxgb3/iwch_provider.c | 25 +++++++++++++++++++++++++ 3 files changed, 29 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h index 25a880664e6..a9ff32c3621 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.h +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h @@ -45,13 +45,14 @@ #define T3_CTRL_QP_SIZE_LOG2 8 #define T3_CTRL_CQ_ID 0 -/* TBD */ #define T3_MAX_NUM_RI (1<<15) #define T3_MAX_NUM_QP (1<<15) #define T3_MAX_NUM_CQ (1<<15) #define T3_MAX_NUM_PD (1<<15) #define T3_MAX_PBL_SIZE 256 #define T3_MAX_RQ_SIZE 1024 +#define T3_MAX_QP_DEPTH (T3_MAX_RQ_SIZE-1) +#define T3_MAX_CQ_DEPTH 8192 #define T3_MAX_NUM_STAG (1<<15) #define T3_MAX_MR_SIZE 0x100000000ULL diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c index 71554eacb13..e09cc1a8199 100644 --- a/drivers/infiniband/hw/cxgb3/iwch.c +++ b/drivers/infiniband/hw/cxgb3/iwch.c @@ -74,11 +74,11 @@ static void rnic_init(struct iwch_dev *rnicp) rnicp->attr.vendor_id = 0x168; rnicp->attr.vendor_part_id = 7; rnicp->attr.max_qps = T3_MAX_NUM_QP - 32; - rnicp->attr.max_wrs = (1UL << 24) - 1; + rnicp->attr.max_wrs = T3_MAX_QP_DEPTH; rnicp->attr.max_sge_per_wr = T3_MAX_SGE; rnicp->attr.max_sge_per_rdma_write_wr = T3_MAX_SGE; rnicp->attr.max_cqs = T3_MAX_NUM_CQ - 1; - rnicp->attr.max_cqes_per_cq = (1UL << 24) - 1; + rnicp->attr.max_cqes_per_cq = T3_MAX_CQ_DEPTH; rnicp->attr.max_mem_regs = cxio_num_stags(&rnicp->rdev); rnicp->attr.max_phys_buf_entries = T3_MAX_PBL_SIZE; rnicp->attr.max_pds = T3_MAX_NUM_PD - 1; diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 5d504f3ed68..5f438065739 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1093,6 +1093,29 @@ static int iwch_query_gid(struct ib_device *ibdev, u8 port, return 0; } +static u64 fw_vers_string_to_u64(struct iwch_dev *iwch_dev) +{ + struct ethtool_drvinfo info; + struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev; + char *cp, *next; + unsigned fw_maj, fw_min, fw_mic; + + rtnl_lock(); + lldev->ethtool_ops->get_drvinfo(lldev, &info); + rtnl_unlock(); + + next = info.fw_version + 1; + cp = strsep(&next, "."); + sscanf(cp, "%i", &fw_maj); + cp = strsep(&next, "."); + sscanf(cp, "%i", &fw_min); + cp = strsep(&next, "."); + sscanf(cp, "%i", &fw_mic); + + return (((u64)fw_maj & 0xffff) << 32) | ((fw_min & 0xffff) << 16) | + (fw_mic & 0xffff); +} + static int iwch_query_device(struct ib_device *ibdev, struct ib_device_attr *props) { @@ -1103,6 +1126,8 @@ static int iwch_query_device(struct ib_device *ibdev, dev = to_iwch_dev(ibdev); memset(props, 0, sizeof *props); memcpy(&props->sys_image_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6); + props->hw_ver = dev->rdev.t3cdev_p->type; + props->fw_ver = fw_vers_string_to_u64(dev); props->device_cap_flags = dev->device_cap_flags; props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor; props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device; -- cgit v1.2.3 From eec8845d29504a12fbd434e192d61aed3d9d74fa Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: RDMA/cxgb3: Remove write-only iwch_rnic_attributes fields The members struct iwch_rnic_attributes.vendor_id and .vendor_part_id are write-only, so we might as well get rid of them. Signed-off-by: Roland Dreier Acked-by: Steve Wise --- drivers/infiniband/hw/cxgb3/iwch.c | 2 -- drivers/infiniband/hw/cxgb3/iwch.h | 2 -- 2 files changed, 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c index e09cc1a8199..06086256ca1 100644 --- a/drivers/infiniband/hw/cxgb3/iwch.c +++ b/drivers/infiniband/hw/cxgb3/iwch.c @@ -71,8 +71,6 @@ static void rnic_init(struct iwch_dev *rnicp) idr_init(&rnicp->mmidr); spin_lock_init(&rnicp->lock); - rnicp->attr.vendor_id = 0x168; - rnicp->attr.vendor_part_id = 7; rnicp->attr.max_qps = T3_MAX_NUM_QP - 32; rnicp->attr.max_wrs = T3_MAX_QP_DEPTH; rnicp->attr.max_sge_per_wr = T3_MAX_SGE; diff --git a/drivers/infiniband/hw/cxgb3/iwch.h b/drivers/infiniband/hw/cxgb3/iwch.h index d2409a505e8..3773453b2cf 100644 --- a/drivers/infiniband/hw/cxgb3/iwch.h +++ b/drivers/infiniband/hw/cxgb3/iwch.h @@ -48,8 +48,6 @@ struct iwch_qp; struct iwch_mr; struct iwch_rnic_attributes { - u32 vendor_id; - u32 vendor_part_id; u32 max_qps; u32 max_wrs; /* Max for any SQ/RQ */ u32 max_sge_per_wr; -- cgit v1.2.3 From 468f2239bcc71ae0f345c3fe58c797cf4627daf4 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: RDMA/cma: Add missing newlines to printk()s Signed-off-by: Roland Dreier Acked-by: Sean Hefty --- drivers/infiniband/core/cma.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index e5bd6172a1f..44d190f6781 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -974,7 +974,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE; break; default: - printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d", + printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n", ib_event->event); goto out; } @@ -1450,7 +1450,7 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv, ret = rdma_listen(id, id_priv->backlog); if (ret) printk(KERN_WARNING "RDMA CMA: cma_listen_on_dev, error %d, " - "listening on device %s", ret, cma_dev->device->name); + "listening on device %s\n", ret, cma_dev->device->name); } static void cma_listen_on_all(struct rdma_id_private *id_priv) @@ -2155,7 +2155,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, event.status = 0; break; default: - printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d", + printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n", ib_event->event); goto out; } -- cgit v1.2.3 From a7d834c4bc6be73e8f83eaa5072fac3c5549f7f2 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:47 -0700 Subject: IPoIB/cm: Fix racy use of receive WR/SGL in ipoib_cm_post_receive_nonsrq() For devices that don't support SRQs, ipoib_cm_post_receive_nonsrq() is called from both ipoib_cm_handle_rx_wc() and ipoib_cm_nonsrq_init_rx(), and these two callers are not synchronized against each other. However, ipoib_cm_post_receive_nonsrq() always reuses the same receive work request and scatter list structures, so multiple callers can end up stepping on each other, which leads to posting garbled work requests. Fix this by having the caller pass in the ib_recv_wr and ib_sge structures to use, and allocating new local structures in ipoib_cm_nonsrq_init_rx(). Based on a patch by Pradeep Satyanarayana and David Wilder , with debugging help from Hoang-Nam Nguyen . Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 63 ++++++++++++++++++++++++--------- 1 file changed, 47 insertions(+), 16 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 6223fc39af7..37bf67b2a26 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -111,18 +111,20 @@ static int ipoib_cm_post_receive_srq(struct net_device *dev, int id) } static int ipoib_cm_post_receive_nonsrq(struct net_device *dev, - struct ipoib_cm_rx *rx, int id) + struct ipoib_cm_rx *rx, + struct ib_recv_wr *wr, + struct ib_sge *sge, int id) { struct ipoib_dev_priv *priv = netdev_priv(dev); struct ib_recv_wr *bad_wr; int i, ret; - priv->cm.rx_wr.wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV; + wr->wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV; for (i = 0; i < IPOIB_CM_RX_SG; ++i) - priv->cm.rx_sge[i].addr = rx->rx_ring[id].mapping[i]; + sge[i].addr = rx->rx_ring[id].mapping[i]; - ret = ib_post_recv(rx->qp, &priv->cm.rx_wr, &bad_wr); + ret = ib_post_recv(rx->qp, wr, &bad_wr); if (unlikely(ret)) { ipoib_warn(priv, "post recv failed for buf %d (%d)\n", id, ret); ipoib_cm_dma_unmap_rx(priv, IPOIB_CM_RX_SG - 1, @@ -320,10 +322,33 @@ static int ipoib_cm_modify_rx_qp(struct net_device *dev, return 0; } +static void ipoib_cm_init_rx_wr(struct net_device *dev, + struct ib_recv_wr *wr, + struct ib_sge *sge) +{ + struct ipoib_dev_priv *priv = netdev_priv(dev); + int i; + + for (i = 0; i < priv->cm.num_frags; ++i) + sge[i].lkey = priv->mr->lkey; + + sge[0].length = IPOIB_CM_HEAD_SIZE; + for (i = 1; i < priv->cm.num_frags; ++i) + sge[i].length = PAGE_SIZE; + + wr->next = NULL; + wr->sg_list = priv->cm.rx_sge; + wr->num_sge = priv->cm.num_frags; +} + static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_id, struct ipoib_cm_rx *rx) { struct ipoib_dev_priv *priv = netdev_priv(dev); + struct { + struct ib_recv_wr wr; + struct ib_sge sge[IPOIB_CM_RX_SG]; + } *t; int ret; int i; @@ -331,6 +356,14 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i if (!rx->rx_ring) return -ENOMEM; + t = kmalloc(sizeof *t, GFP_KERNEL); + if (!t) { + ret = -ENOMEM; + goto err_free; + } + + ipoib_cm_init_rx_wr(dev, &t->wr, t->sge); + spin_lock_irq(&priv->lock); if (priv->cm.nonsrq_conn_qp >= ipoib_max_conn_qp) { @@ -349,8 +382,8 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i ipoib_warn(priv, "failed to allocate receive buffer %d\n", i); ret = -ENOMEM; goto err_count; - } - ret = ipoib_cm_post_receive_nonsrq(dev, rx, i); + } + ret = ipoib_cm_post_receive_nonsrq(dev, rx, &t->wr, t->sge, i); if (ret) { ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq " "failed for buf %d\n", i); @@ -361,6 +394,8 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i rx->recv_count = ipoib_recvq_size; + kfree(t); + return 0; err_count: @@ -369,6 +404,7 @@ err_count: spin_unlock_irq(&priv->lock); err_free: + kfree(t); ipoib_cm_free_rx_ring(dev, rx->rx_ring); return ret; @@ -637,7 +673,10 @@ repost: ipoib_warn(priv, "ipoib_cm_post_receive_srq failed " "for buf %d\n", wr_id); } else { - if (unlikely(ipoib_cm_post_receive_nonsrq(dev, p, wr_id))) { + if (unlikely(ipoib_cm_post_receive_nonsrq(dev, p, + &priv->cm.rx_wr, + priv->cm.rx_sge, + wr_id))) { --p->recv_count; ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq failed " "for buf %d\n", wr_id); @@ -1502,15 +1541,7 @@ int ipoib_cm_dev_init(struct net_device *dev) priv->cm.num_frags = IPOIB_CM_RX_SG; } - for (i = 0; i < priv->cm.num_frags; ++i) - priv->cm.rx_sge[i].lkey = priv->mr->lkey; - - priv->cm.rx_sge[0].length = IPOIB_CM_HEAD_SIZE; - for (i = 1; i < priv->cm.num_frags; ++i) - priv->cm.rx_sge[i].length = PAGE_SIZE; - priv->cm.rx_wr.next = NULL; - priv->cm.rx_wr.sg_list = priv->cm.rx_sge; - priv->cm.rx_wr.num_sge = priv->cm.num_frags; + ipoib_cm_init_rx_wr(dev, &priv->cm.rx_wr, priv->cm.rx_sge); if (ipoib_cm_has_srq(dev)) { for (i = 0; i < ipoib_recvq_size; ++i) { -- cgit v1.2.3 From 7f624d023b5fb150831e02c1e4c0f2619ade72c2 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:48 -0700 Subject: RDMA/core: Add iWARP protocol statistics attributes in sysfs This patch adds a sysfs attribute group called "proto_stats" under /sys/class/infiniband/$device/ and populates this group with protocol statistics if they exist for a given device. Currently, only iWARP stats are defined, but the code is designed to allow InfiniBand protocol stats if they become available. These stats are per-device and more importantly -not- per port. Details: - Add union rdma_protocol_stats in ib_verbs.h. This union allows defining transport-specific stats. Currently only iwarp stats are defined. - Add struct iw_protocol_stats to define the current set of iwarp protocol stats. - Add new ib_device method called get_proto_stats() to return protocol statistics. - Add logic in core/sysfs.c to create iwarp protocol stats attributes if the device is an RNIC and has a get_proto_stats() method. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/core/sysfs.c | 120 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 120 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index 36a0ef97c6a..4d104211559 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -663,6 +663,120 @@ static struct class ib_class = { .dev_uevent = ib_device_uevent, }; +/* Show a given an attribute in the statistics group */ +static ssize_t show_protocol_stat(const struct device *device, + struct device_attribute *attr, char *buf, + unsigned offset) +{ + struct ib_device *dev = container_of(device, struct ib_device, dev); + union rdma_protocol_stats stats; + ssize_t ret; + + ret = dev->get_protocol_stats(dev, &stats); + if (ret) + return ret; + + return sprintf(buf, "%llu\n", + (unsigned long long) ((u64 *) &stats)[offset]); +} + +/* generate a read-only iwarp statistics attribute */ +#define IW_STATS_ENTRY(name) \ +static ssize_t show_##name(struct device *device, \ + struct device_attribute *attr, char *buf) \ +{ \ + return show_protocol_stat(device, attr, buf, \ + offsetof(struct iw_protocol_stats, name) / \ + sizeof (u64)); \ +} \ +static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL) + +IW_STATS_ENTRY(ipInReceives); +IW_STATS_ENTRY(ipInHdrErrors); +IW_STATS_ENTRY(ipInTooBigErrors); +IW_STATS_ENTRY(ipInNoRoutes); +IW_STATS_ENTRY(ipInAddrErrors); +IW_STATS_ENTRY(ipInUnknownProtos); +IW_STATS_ENTRY(ipInTruncatedPkts); +IW_STATS_ENTRY(ipInDiscards); +IW_STATS_ENTRY(ipInDelivers); +IW_STATS_ENTRY(ipOutForwDatagrams); +IW_STATS_ENTRY(ipOutRequests); +IW_STATS_ENTRY(ipOutDiscards); +IW_STATS_ENTRY(ipOutNoRoutes); +IW_STATS_ENTRY(ipReasmTimeout); +IW_STATS_ENTRY(ipReasmReqds); +IW_STATS_ENTRY(ipReasmOKs); +IW_STATS_ENTRY(ipReasmFails); +IW_STATS_ENTRY(ipFragOKs); +IW_STATS_ENTRY(ipFragFails); +IW_STATS_ENTRY(ipFragCreates); +IW_STATS_ENTRY(ipInMcastPkts); +IW_STATS_ENTRY(ipOutMcastPkts); +IW_STATS_ENTRY(ipInBcastPkts); +IW_STATS_ENTRY(ipOutBcastPkts); +IW_STATS_ENTRY(tcpRtoAlgorithm); +IW_STATS_ENTRY(tcpRtoMin); +IW_STATS_ENTRY(tcpRtoMax); +IW_STATS_ENTRY(tcpMaxConn); +IW_STATS_ENTRY(tcpActiveOpens); +IW_STATS_ENTRY(tcpPassiveOpens); +IW_STATS_ENTRY(tcpAttemptFails); +IW_STATS_ENTRY(tcpEstabResets); +IW_STATS_ENTRY(tcpCurrEstab); +IW_STATS_ENTRY(tcpInSegs); +IW_STATS_ENTRY(tcpOutSegs); +IW_STATS_ENTRY(tcpRetransSegs); +IW_STATS_ENTRY(tcpInErrs); +IW_STATS_ENTRY(tcpOutRsts); + +static struct attribute *iw_proto_stats_attrs[] = { + &dev_attr_ipInReceives.attr, + &dev_attr_ipInHdrErrors.attr, + &dev_attr_ipInTooBigErrors.attr, + &dev_attr_ipInNoRoutes.attr, + &dev_attr_ipInAddrErrors.attr, + &dev_attr_ipInUnknownProtos.attr, + &dev_attr_ipInTruncatedPkts.attr, + &dev_attr_ipInDiscards.attr, + &dev_attr_ipInDelivers.attr, + &dev_attr_ipOutForwDatagrams.attr, + &dev_attr_ipOutRequests.attr, + &dev_attr_ipOutDiscards.attr, + &dev_attr_ipOutNoRoutes.attr, + &dev_attr_ipReasmTimeout.attr, + &dev_attr_ipReasmReqds.attr, + &dev_attr_ipReasmOKs.attr, + &dev_attr_ipReasmFails.attr, + &dev_attr_ipFragOKs.attr, + &dev_attr_ipFragFails.attr, + &dev_attr_ipFragCreates.attr, + &dev_attr_ipInMcastPkts.attr, + &dev_attr_ipOutMcastPkts.attr, + &dev_attr_ipInBcastPkts.attr, + &dev_attr_ipOutBcastPkts.attr, + &dev_attr_tcpRtoAlgorithm.attr, + &dev_attr_tcpRtoMin.attr, + &dev_attr_tcpRtoMax.attr, + &dev_attr_tcpMaxConn.attr, + &dev_attr_tcpActiveOpens.attr, + &dev_attr_tcpPassiveOpens.attr, + &dev_attr_tcpAttemptFails.attr, + &dev_attr_tcpEstabResets.attr, + &dev_attr_tcpCurrEstab.attr, + &dev_attr_tcpInSegs.attr, + &dev_attr_tcpOutSegs.attr, + &dev_attr_tcpRetransSegs.attr, + &dev_attr_tcpInErrs.attr, + &dev_attr_tcpOutRsts.attr, + NULL +}; + +static struct attribute_group iw_stats_group = { + .name = "proto_stats", + .attrs = iw_proto_stats_attrs, +}; + int ib_device_register_sysfs(struct ib_device *device) { struct device *class_dev = &device->dev; @@ -705,6 +819,12 @@ int ib_device_register_sysfs(struct ib_device *device) } } + if (device->node_type == RDMA_NODE_RNIC && device->get_protocol_stats) { + ret = sysfs_create_group(&class_dev->kobj, &iw_stats_group); + if (ret) + goto err_put; + } + return 0; err_put: -- cgit v1.2.3 From 14cc180f7b032f8484c1a3d0533b1129ffe307fd Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:48 -0700 Subject: RDMA/cxgb3: Add support for protocol statistics - Add a new rdma ctl command called RDMA_GET_MIB to the cxgb3 low level driver to obtain the protocol mib from the rnic hardware. - Add new iw_cxgb3 provider method to get the MIB from the low level driver. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/iwch_provider.c | 62 +++++++++++++++++++++++++++-- 1 file changed, 58 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 5f438065739..18a6609f5e0 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -56,6 +56,7 @@ #include "iwch_provider.h" #include "iwch_cm.h" #include "iwch_user.h" +#include "common.h" static int iwch_modify_port(struct ib_device *ibdev, u8 port, int port_modify_mask, @@ -1245,6 +1246,61 @@ static ssize_t show_board(struct device *dev, struct device_attribute *attr, iwch_dev->rdev.rnic_info.pdev->device); } +static int iwch_get_mib(struct ib_device *ibdev, + union rdma_protocol_stats *stats) +{ + struct iwch_dev *dev; + struct tp_mib_stats m; + int ret; + + PDBG("%s ibdev %p\n", __func__, ibdev); + dev = to_iwch_dev(ibdev); + ret = dev->rdev.t3cdev_p->ctl(dev->rdev.t3cdev_p, RDMA_GET_MIB, &m); + if (ret) + return -ENOSYS; + + memset(stats, 0, sizeof *stats); + stats->iw.ipInReceives = ((u64) m.ipInReceive_hi << 32) + + m.ipInReceive_lo; + stats->iw.ipInHdrErrors = ((u64) m.ipInHdrErrors_hi << 32) + + m.ipInHdrErrors_lo; + stats->iw.ipInAddrErrors = ((u64) m.ipInAddrErrors_hi << 32) + + m.ipInAddrErrors_lo; + stats->iw.ipInUnknownProtos = ((u64) m.ipInUnknownProtos_hi << 32) + + m.ipInUnknownProtos_lo; + stats->iw.ipInDiscards = ((u64) m.ipInDiscards_hi << 32) + + m.ipInDiscards_lo; + stats->iw.ipInDelivers = ((u64) m.ipInDelivers_hi << 32) + + m.ipInDelivers_lo; + stats->iw.ipOutRequests = ((u64) m.ipOutRequests_hi << 32) + + m.ipOutRequests_lo; + stats->iw.ipOutDiscards = ((u64) m.ipOutDiscards_hi << 32) + + m.ipOutDiscards_lo; + stats->iw.ipOutNoRoutes = ((u64) m.ipOutNoRoutes_hi << 32) + + m.ipOutNoRoutes_lo; + stats->iw.ipReasmTimeout = (u64) m.ipReasmTimeout; + stats->iw.ipReasmReqds = (u64) m.ipReasmReqds; + stats->iw.ipReasmOKs = (u64) m.ipReasmOKs; + stats->iw.ipReasmFails = (u64) m.ipReasmFails; + stats->iw.tcpActiveOpens = (u64) m.tcpActiveOpens; + stats->iw.tcpPassiveOpens = (u64) m.tcpPassiveOpens; + stats->iw.tcpAttemptFails = (u64) m.tcpAttemptFails; + stats->iw.tcpEstabResets = (u64) m.tcpEstabResets; + stats->iw.tcpOutRsts = (u64) m.tcpOutRsts; + stats->iw.tcpCurrEstab = (u64) m.tcpCurrEstab; + stats->iw.tcpInSegs = ((u64) m.tcpInSegs_hi << 32) + + m.tcpInSegs_lo; + stats->iw.tcpOutSegs = ((u64) m.tcpOutSegs_hi << 32) + + m.tcpOutSegs_lo; + stats->iw.tcpRetransSegs = ((u64) m.tcpRetransSeg_hi << 32) + + m.tcpRetransSeg_lo; + stats->iw.tcpInErrs = ((u64) m.tcpInErrs_hi << 32) + + m.tcpInErrs_lo; + stats->iw.tcpRtoMin = (u64) m.tcpRtoMin; + stats->iw.tcpRtoMax = (u64) m.tcpRtoMax; + return 0; +} + static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL); static DEVICE_ATTR(fw_ver, S_IRUGO, show_fw_ver, NULL); static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL); @@ -1254,7 +1310,7 @@ static struct device_attribute *iwch_class_attributes[] = { &dev_attr_hw_rev, &dev_attr_fw_ver, &dev_attr_hca_type, - &dev_attr_board_id + &dev_attr_board_id, }; int iwch_register_device(struct iwch_dev *dev) @@ -1325,15 +1381,13 @@ int iwch_register_device(struct iwch_dev *dev) dev->ibdev.alloc_fast_reg_mr = iwch_alloc_fast_reg_mr; dev->ibdev.alloc_fast_reg_page_list = iwch_alloc_fastreg_pbl; dev->ibdev.free_fast_reg_page_list = iwch_free_fastreg_pbl; - dev->ibdev.attach_mcast = iwch_multicast_attach; dev->ibdev.detach_mcast = iwch_multicast_detach; dev->ibdev.process_mad = iwch_process_mad; - dev->ibdev.req_notify_cq = iwch_arm_cq; dev->ibdev.post_send = iwch_post_send; dev->ibdev.post_recv = iwch_post_receive; - + dev->ibdev.get_protocol_stats = iwch_get_mib; dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL); if (!dev->ibdev.iwcm) -- cgit v1.2.3 From 521e575b9a7324a0bca762622139f69582a042bf Mon Sep 17 00:00:00 2001 From: Ron Livne Date: Mon, 14 Jul 2008 23:48:48 -0700 Subject: IB/mlx4: Add support for blocking multicast loopback packets Add support for handling the IB_QP_CREATE_MULTICAST_BLOCK_LOOPBACK flag by using the per-multicast group loopback blocking feature of mlx4 hardware. Signed-off-by: Ron Livne Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/main.c | 7 +++++-- drivers/infiniband/hw/mlx4/mlx4_ib.h | 3 ++- drivers/infiniband/hw/mlx4/qp.c | 21 ++++++++++++++++++--- 3 files changed, 25 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 4d61e32866c..bcf50648fa1 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -90,7 +90,8 @@ static int mlx4_ib_query_device(struct ib_device *ibdev, props->device_cap_flags = IB_DEVICE_CHANGE_PHY_PORT | IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SYS_IMAGE_GUID | - IB_DEVICE_RC_RNR_NAK_GEN; + IB_DEVICE_RC_RNR_NAK_GEN | + IB_DEVICE_BLOCK_MULTICAST_LOOPBACK; if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_PKEY_CNTR) props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR; if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_QKEY_CNTR) @@ -437,7 +438,9 @@ static int mlx4_ib_dealloc_pd(struct ib_pd *pd) static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) { return mlx4_multicast_attach(to_mdev(ibqp->device)->dev, - &to_mqp(ibqp)->mqp, gid->raw); + &to_mqp(ibqp)->mqp, gid->raw, + !!(to_mqp(ibqp)->flags & + MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK)); } static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 5cf994794d2..c4cf5b69eef 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -101,7 +101,8 @@ struct mlx4_ib_wq { }; enum mlx4_ib_qp_flags { - MLX4_IB_QP_LSO = 1 << 0 + MLX4_IB_QP_LSO = 1 << 0, + MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK = 1 << 1, }; struct mlx4_ib_qp { diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 44bbd6c2e31..91590e7fba0 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -511,6 +511,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, } else { qp->sq_no_prefetch = 0; + if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) + qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK; + if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO) qp->flags |= MLX4_IB_QP_LSO; @@ -684,10 +687,15 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd, struct mlx4_ib_qp *qp; int err; - /* We only support LSO, and only for kernel UD QPs. */ - if (init_attr->create_flags & ~IB_QP_CREATE_IPOIB_UD_LSO) + /* + * We only support LSO and multicast loopback blocking, and + * only for kernel UD QPs. + */ + if (init_attr->create_flags & ~(IB_QP_CREATE_IPOIB_UD_LSO | + IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)) return ERR_PTR(-EINVAL); - if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO && + + if (init_attr->create_flags && (pd->uobject || init_attr->qp_type != IB_QPT_UD)) return ERR_PTR(-EINVAL); @@ -1844,6 +1852,13 @@ done: qp_init_attr->cap = qp_attr->cap; + qp_init_attr->create_flags = 0; + if (qp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK) + qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK; + + if (qp->flags & MLX4_IB_QP_LSO) + qp_init_attr->create_flags |= IB_QP_CREATE_IPOIB_UD_LSO; + out: mutex_unlock(&qp->mutex); return err; -- cgit v1.2.3 From 12406734051a26e9fe4c8568e931dfddbb72d431 Mon Sep 17 00:00:00 2001 From: Ron Livne Date: Mon, 14 Jul 2008 23:48:48 -0700 Subject: IPoIB: Use multicast loopback blocking if available Set IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK for IPoIB's UD QPs if supported by the underlying device. This creates an improvement of up to 39% in bandwidth when sending multicast packets with IPoIB, and an improvment of 12% in cpu usage. Signed-off-by: Ron Livne Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index 810790ae753..7b8fa36f509 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -199,7 +199,10 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) init_attr.recv_cq = priv->recv_cq; if (priv->hca_caps & IB_DEVICE_UD_TSO) - init_attr.create_flags = IB_QP_CREATE_IPOIB_UD_LSO; + init_attr.create_flags |= IB_QP_CREATE_IPOIB_UD_LSO; + + if (priv->hca_caps & IB_DEVICE_BLOCK_MULTICAST_LOOPBACK) + init_attr.create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK; if (dev->features & NETIF_F_SG) init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1; -- cgit v1.2.3 From af40da894e96d5c826d38be3ea53ee00d9de0367 Mon Sep 17 00:00:00 2001 From: Vladimir Sokolovsky Date: Mon, 14 Jul 2008 23:48:48 -0700 Subject: IPoIB: add LRO support Add "ipoib_use_lro" module parameter to enable LRO and an "ipoib_lro_max_aggr" module parameter to set the max number of packets to be aggregated. Make LRO controllable and LRO statistics accessible through ethtool. Signed-off-by: Vladimir Sokolovsky Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/Kconfig | 1 + drivers/infiniband/ulp/ipoib/ipoib.h | 11 +++++ drivers/infiniband/ulp/ipoib/ipoib_ethtool.c | 46 +++++++++++++++++++++ drivers/infiniband/ulp/ipoib/ipoib_ib.c | 8 +++- drivers/infiniband/ulp/ipoib/ipoib_main.c | 62 ++++++++++++++++++++++++++++ 5 files changed, 127 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/Kconfig b/drivers/infiniband/ulp/ipoib/Kconfig index 1f76bad020f..691525cf394 100644 --- a/drivers/infiniband/ulp/ipoib/Kconfig +++ b/drivers/infiniband/ulp/ipoib/Kconfig @@ -1,6 +1,7 @@ config INFINIBAND_IPOIB tristate "IP-over-InfiniBand" depends on NETDEVICES && INET && (IPV6 || IPV6=n) + select INET_LRO ---help--- Support for the IP-over-InfiniBand protocol (IPoIB). This transports IP packets over InfiniBand so you can use your IB diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 8754b364f22..2c522572e3c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -50,6 +50,7 @@ #include #include #include +#include /* constants */ @@ -94,6 +95,9 @@ enum { IPOIB_MCAST_FLAG_BUSY = 2, /* joining or already joined */ IPOIB_MCAST_FLAG_ATTACHED = 3, + IPOIB_MAX_LRO_DESCRIPTORS = 8, + IPOIB_LRO_MAX_AGGR = 64, + MAX_SEND_CQE = 16, IPOIB_CM_COPYBREAK = 256, }; @@ -248,6 +252,11 @@ struct ipoib_ethtool_st { u16 max_coalesced_frames; }; +struct ipoib_lro { + struct net_lro_mgr lro_mgr; + struct net_lro_desc lro_desc[IPOIB_MAX_LRO_DESCRIPTORS]; +}; + /* * Device private locking: tx_lock protects members used in TX fast * path (and we use LLTX so upper layers don't do extra locking). @@ -334,6 +343,8 @@ struct ipoib_dev_priv { int hca_caps; struct ipoib_ethtool_st ethtool; struct timer_list poll_timer; + + struct ipoib_lro lro; }; struct ipoib_ah { diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c index 10279b79c44..66af5c1a76e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c @@ -86,11 +86,57 @@ static int ipoib_set_coalesce(struct net_device *dev, return 0; } +static const char ipoib_stats_keys[][ETH_GSTRING_LEN] = { + "LRO aggregated", "LRO flushed", + "LRO avg aggr", "LRO no desc" +}; + +static void ipoib_get_strings(struct net_device *netdev, u32 stringset, u8 *data) +{ + switch (stringset) { + case ETH_SS_STATS: + memcpy(data, *ipoib_stats_keys, sizeof(ipoib_stats_keys)); + break; + } +} + +static int ipoib_get_sset_count(struct net_device *dev, int sset) +{ + switch (sset) { + case ETH_SS_STATS: + return ARRAY_SIZE(ipoib_stats_keys); + default: + return -EOPNOTSUPP; + } +} + +static void ipoib_get_ethtool_stats(struct net_device *dev, + struct ethtool_stats *stats, uint64_t *data) +{ + struct ipoib_dev_priv *priv = netdev_priv(dev); + int index = 0; + + /* Get LRO statistics */ + data[index++] = priv->lro.lro_mgr.stats.aggregated; + data[index++] = priv->lro.lro_mgr.stats.flushed; + if (priv->lro.lro_mgr.stats.flushed) + data[index++] = priv->lro.lro_mgr.stats.aggregated / + priv->lro.lro_mgr.stats.flushed; + else + data[index++] = 0; + data[index++] = priv->lro.lro_mgr.stats.no_desc; +} + static const struct ethtool_ops ipoib_ethtool_ops = { .get_drvinfo = ipoib_get_drvinfo, .get_tso = ethtool_op_get_tso, .get_coalesce = ipoib_get_coalesce, .set_coalesce = ipoib_set_coalesce, + .get_flags = ethtool_op_get_flags, + .set_flags = ethtool_op_set_flags, + .get_strings = ipoib_get_strings, + .get_sset_count = ipoib_get_sset_count, + .get_ethtool_stats = ipoib_get_ethtool_stats, }; void ipoib_set_ethtool_ops(struct net_device *dev) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index eca8518d79a..5d50e5261ee 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -288,7 +288,10 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) if (test_bit(IPOIB_FLAG_CSUM, &priv->flags) && likely(wc->csum_ok)) skb->ip_summed = CHECKSUM_UNNECESSARY; - netif_receive_skb(skb); + if (dev->features & NETIF_F_LRO) + lro_receive_skb(&priv->lro.lro_mgr, skb, NULL); + else + netif_receive_skb(skb); repost: if (unlikely(ipoib_ib_post_receive(dev, wr_id))) @@ -440,6 +443,9 @@ poll_more: } if (done < budget) { + if (dev->features & NETIF_F_LRO) + lro_flush_all(&priv->lro.lro_mgr); + netif_rx_complete(dev, napi); if (unlikely(ib_req_notify_cq(priv->recv_cq, IB_CQ_NEXT_COMP | diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index bfe1dbf9920..fead88f7fb1 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -60,6 +60,15 @@ MODULE_PARM_DESC(send_queue_size, "Number of descriptors in send queue"); module_param_named(recv_queue_size, ipoib_recvq_size, int, 0444); MODULE_PARM_DESC(recv_queue_size, "Number of descriptors in receive queue"); +static int lro; +module_param(lro, bool, 0444); +MODULE_PARM_DESC(lro, "Enable LRO (Large Receive Offload)"); + +static int lro_max_aggr = IPOIB_LRO_MAX_AGGR; +module_param(lro_max_aggr, int, 0644); +MODULE_PARM_DESC(lro_max_aggr, "LRO: Max packets to be aggregated " + "(default = 64)"); + #ifdef CONFIG_INFINIBAND_IPOIB_DEBUG int ipoib_debug_level; @@ -936,6 +945,54 @@ static const struct header_ops ipoib_header_ops = { .create = ipoib_hard_header, }; +static int get_skb_hdr(struct sk_buff *skb, void **iphdr, + void **tcph, u64 *hdr_flags, void *priv) +{ + unsigned int ip_len; + struct iphdr *iph; + + if (unlikely(skb->protocol != htons(ETH_P_IP))) + return -1; + + /* + * In the future we may add an else clause that verifies the + * checksum and allows devices which do not calculate checksum + * to use LRO. + */ + if (unlikely(skb->ip_summed != CHECKSUM_UNNECESSARY)) + return -1; + + /* Check for non-TCP packet */ + skb_reset_network_header(skb); + iph = ip_hdr(skb); + if (iph->protocol != IPPROTO_TCP) + return -1; + + ip_len = ip_hdrlen(skb); + skb_set_transport_header(skb, ip_len); + *tcph = tcp_hdr(skb); + + /* check if IP header and TCP header are complete */ + if (ntohs(iph->tot_len) < ip_len + tcp_hdrlen(skb)) + return -1; + + *hdr_flags = LRO_IPV4 | LRO_TCP; + *iphdr = iph; + + return 0; +} + +static void ipoib_lro_setup(struct ipoib_dev_priv *priv) +{ + priv->lro.lro_mgr.max_aggr = lro_max_aggr; + priv->lro.lro_mgr.max_desc = IPOIB_MAX_LRO_DESCRIPTORS; + priv->lro.lro_mgr.lro_arr = priv->lro.lro_desc; + priv->lro.lro_mgr.get_skb_header = get_skb_hdr; + priv->lro.lro_mgr.features = LRO_F_NAPI; + priv->lro.lro_mgr.dev = priv->dev; + priv->lro.lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY; +} + static void ipoib_setup(struct net_device *dev) { struct ipoib_dev_priv *priv = netdev_priv(dev); @@ -975,6 +1032,8 @@ static void ipoib_setup(struct net_device *dev) priv->dev = dev; + ipoib_lro_setup(priv); + spin_lock_init(&priv->lock); spin_lock_init(&priv->tx_lock); @@ -1152,6 +1211,9 @@ static struct net_device *ipoib_add_port(const char *format, priv->dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM; } + if (lro) + priv->dev->features |= NETIF_F_LRO; + /* * Set the full membership bit, so that we join the right * broadcast group, etc. -- cgit v1.2.3 From 038919f29682b00ea95506e959210fc72d1aaf64 Mon Sep 17 00:00:00 2001 From: Joachim Fenkes Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: IB/ehca: Make device table externally visible This gives ehca an autogenerated modalias and therefore enables automatic loading. Signed-off-by: Joachim Fenkes Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ehca/ehca_main.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c index 482103eb6ea..598844d2edc 100644 --- a/drivers/infiniband/hw/ehca/ehca_main.c +++ b/drivers/infiniband/hw/ehca/ehca_main.c @@ -923,6 +923,7 @@ static struct of_device_id ehca_device_table[] = }, {}, }; +MODULE_DEVICE_TABLE(of, ehca_device_table); static struct of_platform_driver ehca_driver = { .name = "ehca", -- cgit v1.2.3 From ee1e2c82c245a5fb2864e9dbcdaab3390fde3fcc Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy Signed-off-by: Moni Shoua Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 17 ++++++++++-- drivers/infiniband/ulp/ipoib/ipoib_ib.c | 42 ++++++++++++++++++---------- drivers/infiniband/ulp/ipoib/ipoib_main.c | 44 +++++++++++++++++++++++++++--- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 18 ++++++------ 4 files changed, 91 insertions(+), 30 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 2c522572e3c..bb19587c5ea 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -54,6 +54,12 @@ /* constants */ +enum ipoib_flush_level { + IPOIB_FLUSH_LIGHT, + IPOIB_FLUSH_NORMAL, + IPOIB_FLUSH_HEAVY +}; + enum { IPOIB_ENCAP_LEN = 4, @@ -284,10 +290,11 @@ struct ipoib_dev_priv { struct delayed_work pkey_poll_task; struct delayed_work mcast_task; - struct work_struct flush_task; + struct work_struct flush_light; + struct work_struct flush_normal; + struct work_struct flush_heavy; struct work_struct restart_task; struct delayed_work ah_reap_task; - struct work_struct pkey_event_task; struct ib_device *ca; u8 port; @@ -369,6 +376,7 @@ struct ipoib_path { struct rb_node rb_node; struct list_head list; + int valid; }; struct ipoib_neigh { @@ -433,11 +441,14 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_ah *address, u32 qpn); void ipoib_reap_ah(struct work_struct *work); +void ipoib_mark_paths_invalid(struct net_device *dev); void ipoib_flush_paths(struct net_device *dev); struct ipoib_dev_priv *ipoib_intf_alloc(const char *format); int ipoib_ib_dev_init(struct net_device *dev, struct ib_device *ca, int port); -void ipoib_ib_dev_flush(struct work_struct *work); +void ipoib_ib_dev_flush_light(struct work_struct *work); +void ipoib_ib_dev_flush_normal(struct work_struct *work); +void ipoib_ib_dev_flush_heavy(struct work_struct *work); void ipoib_pkey_event(struct work_struct *work); void ipoib_ib_dev_cleanup(struct net_device *dev); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index 5d50e5261ee..66cafa20c24 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -902,7 +902,8 @@ int ipoib_ib_dev_init(struct net_device *dev, struct ib_device *ca, int port) return 0; } -static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event) +static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, + enum ipoib_flush_level level) { struct ipoib_dev_priv *cpriv; struct net_device *dev = priv->dev; @@ -915,7 +916,7 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event) * the parent is down. */ list_for_each_entry(cpriv, &priv->child_intfs, list) - __ipoib_ib_dev_flush(cpriv, pkey_event); + __ipoib_ib_dev_flush(cpriv, level); mutex_unlock(&priv->vlan_mutex); @@ -929,7 +930,7 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event) return; } - if (pkey_event) { + if (level == IPOIB_FLUSH_HEAVY) { if (ib_find_pkey(priv->ca, priv->port, priv->pkey, &new_index)) { clear_bit(IPOIB_PKEY_ASSIGNED, &priv->flags); ipoib_ib_dev_down(dev, 0); @@ -947,11 +948,15 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event) priv->pkey_index = new_index; } - ipoib_dbg(priv, "flushing\n"); + if (level == IPOIB_FLUSH_LIGHT) { + ipoib_mark_paths_invalid(dev); + ipoib_mcast_dev_flush(dev); + } - ipoib_ib_dev_down(dev, 0); + if (level >= IPOIB_FLUSH_NORMAL) + ipoib_ib_dev_down(dev, 0); - if (pkey_event) { + if (level == IPOIB_FLUSH_HEAVY) { ipoib_ib_dev_stop(dev, 0); ipoib_ib_dev_open(dev); } @@ -961,27 +966,34 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event) * we get here, don't bring it back up if it's not configured up */ if (test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) { - ipoib_ib_dev_up(dev); + if (level >= IPOIB_FLUSH_NORMAL) + ipoib_ib_dev_up(dev); ipoib_mcast_restart_task(&priv->restart_task); } } -void ipoib_ib_dev_flush(struct work_struct *work) +void ipoib_ib_dev_flush_light(struct work_struct *work) +{ + struct ipoib_dev_priv *priv = + container_of(work, struct ipoib_dev_priv, flush_light); + + __ipoib_ib_dev_flush(priv, IPOIB_FLUSH_LIGHT); +} + +void ipoib_ib_dev_flush_normal(struct work_struct *work) { struct ipoib_dev_priv *priv = - container_of(work, struct ipoib_dev_priv, flush_task); + container_of(work, struct ipoib_dev_priv, flush_normal); - ipoib_dbg(priv, "Flushing %s\n", priv->dev->name); - __ipoib_ib_dev_flush(priv, 0); + __ipoib_ib_dev_flush(priv, IPOIB_FLUSH_NORMAL); } -void ipoib_pkey_event(struct work_struct *work) +void ipoib_ib_dev_flush_heavy(struct work_struct *work) { struct ipoib_dev_priv *priv = - container_of(work, struct ipoib_dev_priv, pkey_event_task); + container_of(work, struct ipoib_dev_priv, flush_heavy); - ipoib_dbg(priv, "Flushing %s and restarting its QP\n", priv->dev->name); - __ipoib_ib_dev_flush(priv, 1); + __ipoib_ib_dev_flush(priv, IPOIB_FLUSH_HEAVY); } void ipoib_ib_dev_cleanup(struct net_device *dev) diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index fead88f7fb1..b3fd7e8333c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -357,6 +357,23 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter, #endif /* CONFIG_INFINIBAND_IPOIB_DEBUG */ +void ipoib_mark_paths_invalid(struct net_device *dev) +{ + struct ipoib_dev_priv *priv = netdev_priv(dev); + struct ipoib_path *path, *tp; + + spin_lock_irq(&priv->lock); + + list_for_each_entry_safe(path, tp, &priv->path_list, list) { + ipoib_dbg(priv, "mark path LID 0x%04x GID " IPOIB_GID_FMT " invalid\n", + be16_to_cpu(path->pathrec.dlid), + IPOIB_GID_ARG(path->pathrec.dgid)); + path->valid = 0; + } + + spin_unlock_irq(&priv->lock); +} + void ipoib_flush_paths(struct net_device *dev) { struct ipoib_dev_priv *priv = netdev_priv(dev); @@ -393,6 +410,7 @@ static void path_rec_completion(int status, struct net_device *dev = path->dev; struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_ah *ah = NULL; + struct ipoib_ah *old_ah; struct ipoib_neigh *neigh, *tn; struct sk_buff_head skqueue; struct sk_buff *skb; @@ -416,6 +434,7 @@ static void path_rec_completion(int status, spin_lock_irqsave(&priv->lock, flags); + old_ah = path->ah; path->ah = ah; if (ah) { @@ -428,6 +447,17 @@ static void path_rec_completion(int status, __skb_queue_tail(&skqueue, skb); list_for_each_entry_safe(neigh, tn, &path->neigh_list, list) { + if (neigh->ah) { + WARN_ON(neigh->ah != old_ah); + /* + * Dropping the ah reference inside + * priv->lock is safe here, because we + * will hold one more reference from + * the original value of path->ah (ie + * old_ah). + */ + ipoib_put_ah(neigh->ah); + } kref_get(&path->ah->ref); neigh->ah = path->ah; memcpy(&neigh->dgid.raw, &path->pathrec.dgid.raw, @@ -450,6 +480,7 @@ static void path_rec_completion(int status, while ((skb = __skb_dequeue(&neigh->queue))) __skb_queue_tail(&skqueue, skb); } + path->valid = 1; } path->query = NULL; @@ -457,6 +488,9 @@ static void path_rec_completion(int status, spin_unlock_irqrestore(&priv->lock, flags); + if (old_ah) + ipoib_put_ah(old_ah); + while ((skb = __skb_dequeue(&skqueue))) { skb->dev = dev; if (dev_queue_xmit(skb)) @@ -630,8 +664,9 @@ static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev, spin_lock(&priv->lock); path = __path_find(dev, phdr->hwaddr + 4); - if (!path) { - path = path_rec_create(dev, phdr->hwaddr + 4); + if (!path || !path->valid) { + if (!path) + path = path_rec_create(dev, phdr->hwaddr + 4); if (path) { /* put pseudoheader back on for next time */ skb_push(skb, sizeof *phdr); @@ -1046,9 +1081,10 @@ static void ipoib_setup(struct net_device *dev) INIT_LIST_HEAD(&priv->multicast_list); INIT_DELAYED_WORK(&priv->pkey_poll_task, ipoib_pkey_poll); - INIT_WORK(&priv->pkey_event_task, ipoib_pkey_event); INIT_DELAYED_WORK(&priv->mcast_task, ipoib_mcast_join_task); - INIT_WORK(&priv->flush_task, ipoib_ib_dev_flush); + INIT_WORK(&priv->flush_light, ipoib_ib_dev_flush_light); + INIT_WORK(&priv->flush_normal, ipoib_ib_dev_flush_normal); + INIT_WORK(&priv->flush_heavy, ipoib_ib_dev_flush_heavy); INIT_WORK(&priv->restart_task, ipoib_mcast_restart_task); INIT_DELAYED_WORK(&priv->ah_reap_task, ipoib_reap_ah); } diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index 7b8fa36f509..96f9aa79cbb 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -290,15 +290,17 @@ void ipoib_event(struct ib_event_handler *handler, if (record->element.port_num != priv->port) return; - if (record->event == IB_EVENT_PORT_ERR || - record->event == IB_EVENT_PORT_ACTIVE || - record->event == IB_EVENT_LID_CHANGE || - record->event == IB_EVENT_SM_CHANGE || + ipoib_dbg(priv, "Event %d on device %s port %d\n", record->event, + record->device->name, record->element.port_num); + + if (record->event == IB_EVENT_SM_CHANGE || record->event == IB_EVENT_CLIENT_REREGISTER) { - ipoib_dbg(priv, "Port state change event\n"); - queue_work(ipoib_workqueue, &priv->flush_task); + queue_work(ipoib_workqueue, &priv->flush_light); + } else if (record->event == IB_EVENT_PORT_ERR || + record->event == IB_EVENT_PORT_ACTIVE || + record->event == IB_EVENT_LID_CHANGE) { + queue_work(ipoib_workqueue, &priv->flush_normal); } else if (record->event == IB_EVENT_PKEY_CHANGE) { - ipoib_dbg(priv, "P_Key change event on port:%d\n", priv->port); - queue_work(ipoib_workqueue, &priv->pkey_event_task); + queue_work(ipoib_workqueue, &priv->flush_heavy); } } -- cgit v1.2.3 From 1ff66e8c1faee7c2711b84b9c89e1c5fcd767839 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: RDMA/nes: Encapsulate logic nes_put_cqp_request() The iw_nes driver repeats the logic if (atomic_dec_and_test(&cqp_request->refcount)) { if (cqp_request->dynamic) { kfree(cqp_request); } else { spin_lock_irqsave(&nesdev->cqp.lock, flags); list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); spin_unlock_irqrestore(&nesdev->cqp.lock, flags); } } over and over. Wrap this up in functions nes_free_cqp_request() and nes_put_cqp_request() to simplify such code. In addition to making the source smaller and more readable, this shrinks the compiled code quite a bit: add/remove: 2/0 grow/shrink: 0/13 up/down: 164/-1692 (-1528) function old new delta nes_free_cqp_request - 147 +147 nes_put_cqp_request - 17 +17 nes_modify_qp 2316 2293 -23 nes_hw_modify_qp 737 657 -80 nes_dereg_mr 945 860 -85 flush_wqes 501 416 -85 nes_manage_apbvt 648 560 -88 nes_reg_mr 1117 1026 -91 nes_cqp_ce_handler 927 769 -158 nes_alloc_mw 1052 884 -168 nes_create_qp 5314 5141 -173 nes_alloc_fmr 2212 2035 -177 nes_destroy_cq 1097 918 -179 nes_create_cq 2787 2598 -189 nes_dealloc_mw 762 566 -196 Signed-off-by: Roland Dreier Acked-by: Faisal Latif --- drivers/infiniband/hw/nes/nes.h | 4 + drivers/infiniband/hw/nes/nes_hw.c | 60 ++--------- drivers/infiniband/hw/nes/nes_utils.c | 24 +++++ drivers/infiniband/hw/nes/nes_verbs.c | 189 ++++++---------------------------- 4 files changed, 66 insertions(+), 211 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h index 61b46e9c7d2..fe88bec4894 100644 --- a/drivers/infiniband/hw/nes/nes.h +++ b/drivers/infiniband/hw/nes/nes.h @@ -538,6 +538,10 @@ void nes_read_1G_phy_reg(struct nes_device *, u8, u8, u16 *); void nes_write_10G_phy_reg(struct nes_device *, u16, u8, u16, u16); void nes_read_10G_phy_reg(struct nes_device *, u8, u8, u16); struct nes_cqp_request *nes_get_cqp_request(struct nes_device *); +void nes_free_cqp_request(struct nes_device *nesdev, + struct nes_cqp_request *cqp_request); +void nes_put_cqp_request(struct nes_device *nesdev, + struct nes_cqp_request *cqp_request); void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *, int); int nes_arp_table(struct nes_device *, u32, u8 *, u32); void nes_mh_fix(unsigned long); diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index d3278f111ca..80e486653ec 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -2710,39 +2710,11 @@ static void nes_cqp_ce_handler(struct nes_device *nesdev, struct nes_hw_cq *cq) barrier(); cqp_request->request_done = 1; wake_up(&cqp_request->waitq); - if (atomic_dec_and_test(&cqp_request->refcount)) { - nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n", - cqp_request, - le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX])&0x3f); - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } - } else if (cqp_request->callback) { - /* Envoke the callback routine */ - cqp_request->cqp_callback(nesdev, cqp_request); - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } + nes_put_cqp_request(nesdev, cqp_request); } else { - nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n", - cqp_request, - le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f); - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } + if (cqp_request->callback) + cqp_request->cqp_callback(nesdev, cqp_request); + nes_free_cqp_request(nesdev, cqp_request); } } else { wake_up(&nesdev->cqp.waitq); @@ -3149,7 +3121,6 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port, { struct nes_device *nesdev = nesvnic->nesdev; struct nes_hw_cqp_wqe *cqp_wqe; - unsigned long flags; struct nes_cqp_request *cqp_request; int ret = 0; u16 major_code; @@ -3184,15 +3155,9 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port, nes_debug(NES_DBG_QP, "Completed, ret=%u, CQP Major:Minor codes = 0x%04X:0x%04X\n", ret, cqp_request->major_code, cqp_request->minor_code); major_code = cqp_request->major_code; - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + + nes_put_cqp_request(nesdev, cqp_request); + if (!ret) return -ETIME; else if (major_code) @@ -3262,7 +3227,6 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr, void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp, u32 which_wq, u32 wait_completion) { - unsigned long flags; struct nes_cqp_request *cqp_request; struct nes_hw_cqp_wqe *cqp_wqe; int ret; @@ -3294,14 +3258,6 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp, nes_debug(NES_DBG_QP, "Flush SQ QP WQEs completed, ret=%u," " CQP Major:Minor codes = 0x%04X:0x%04X\n", ret, cqp_request->major_code, cqp_request->minor_code); - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); } } diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c index fe83d1b2b17..8f519c86087 100644 --- a/drivers/infiniband/hw/nes/nes_utils.c +++ b/drivers/infiniband/hw/nes/nes_utils.c @@ -567,6 +567,30 @@ struct nes_cqp_request *nes_get_cqp_request(struct nes_device *nesdev) return cqp_request; } +void nes_free_cqp_request(struct nes_device *nesdev, + struct nes_cqp_request *cqp_request) +{ + unsigned long flags; + + nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n", + cqp_request, + le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f); + + if (cqp_request->dynamic) { + kfree(cqp_request); + } else { + spin_lock_irqsave(&nesdev->cqp.lock, flags); + list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); + spin_unlock_irqrestore(&nesdev->cqp.lock, flags); + } +} + +void nes_put_cqp_request(struct nes_device *nesdev, + struct nes_cqp_request *cqp_request) +{ + if (atomic_dec_and_test(&cqp_request->refcount)) + nes_free_cqp_request(nesdev, cqp_request); +} /** * nes_post_cqp_request diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index d617da9bd35..464a98a6e11 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -55,7 +55,6 @@ static void nes_unregister_ofa_device(struct nes_ib_device *nesibdev); * nes_alloc_mw */ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) { - unsigned long flags; struct nes_pd *nespd = to_nespd(ibpd); struct nes_vnic *nesvnic = to_nesvnic(ibpd->device); struct nes_device *nesdev = nesvnic->nesdev; @@ -128,15 +127,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) { " CQP Major:Minor codes = 0x%04X:0x%04X.\n", stag, ret, cqp_request->major_code, cqp_request->minor_code); if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); kfree(nesmr); nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index); if (!ret) { @@ -144,17 +135,8 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) { } else { return ERR_PTR(-ENOMEM); } - } else { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } } + nes_put_cqp_request(nesdev, cqp_request); nesmr->ibmw.rkey = stag; nesmr->mode = IWNES_MEMREG_TYPE_MW; @@ -178,7 +160,6 @@ static int nes_dealloc_mw(struct ib_mw *ibmw) struct nes_hw_cqp_wqe *cqp_wqe; struct nes_cqp_request *cqp_request; int err = 0; - unsigned long flags; int ret; /* Deallocate the window with the adapter */ @@ -204,32 +185,12 @@ static int nes_dealloc_mw(struct ib_mw *ibmw) nes_debug(NES_DBG_MR, "Deallocate STag completed, wait_event_timeout ret = %u," " CQP Major:Minor codes = 0x%04X:0x%04X.\n", ret, cqp_request->major_code, cqp_request->minor_code); - if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } - if (!ret) { - err = -ETIME; - } else { - err = -EIO; - } - } else { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } - } + if (!ret) + err = -ETIME; + else if (cqp_request->major_code) + err = -EIO; + + nes_put_cqp_request(nesdev, cqp_request); nes_free_resource(nesadapter, nesadapter->allocated_mrs, (ibmw->rkey & 0x0fffff00) >> 8); @@ -526,29 +487,11 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd, stag, ret, cqp_request->major_code, cqp_request->minor_code); if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); ret = (!ret) ? -ETIME : -EIO; goto failed_leaf_vpbl_pages_alloc; - } else { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } } - + nes_put_cqp_request(nesdev, cqp_request); nesfmr->nesmr.ibfmr.lkey = stag; nesfmr->nesmr.ibfmr.rkey = stag; nesfmr->attr = *ibfmr_attr; @@ -1487,15 +1430,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, nesqp->hwqp.qp_id, ret, nesdev->cqp.sq_head, nesdev->cqp.sq_tail, cqp_request->major_code, cqp_request->minor_code); if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); nes_free_resource(nesadapter, nesadapter->allocated_qps, qp_num); nes_free_qp_mem(nesdev, nesqp,virt_wqs); kfree(nesqp->allocated_buffer); @@ -1504,18 +1439,10 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, } else { return ERR_PTR(-EIO); } - } else { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } } + nes_put_cqp_request(nesdev, cqp_request); + if (ibpd->uobject) { uresp.mmap_sq_db_index = nesqp->mmap_sq_db_index; uresp.actual_sq_size = sq_size; @@ -1827,32 +1754,15 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries, nes_debug(NES_DBG_CQ, "Create iWARP CQ%u completed, wait_event_timeout ret = %d.\n", nescq->hw_cq.cq_number, ret); if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); if (!context) pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size, mem, nescq->hw_cq.cq_pbase); nes_free_resource(nesadapter, nesadapter->allocated_cqs, cq_num); kfree(nescq); return ERR_PTR(-EIO); - } else { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } } + nes_put_cqp_request(nesdev, cqp_request); if (context) { /* free the nespbl */ @@ -1942,37 +1852,18 @@ static int nes_destroy_cq(struct ib_cq *ib_cq) " CQP Major:Minor codes = 0x%04X:0x%04X.\n", nescq->hw_cq.cq_number, ret, cqp_request->major_code, cqp_request->minor_code); - if ((!ret) || (cqp_request->major_code)) { - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } - if (!ret) { - nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n", + if (!ret) { + nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n", nescq->hw_cq.cq_number); - ret = -ETIME; - } else { - nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n", + ret = -ETIME; + } else if (cqp_request->major_code) { + nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n", nescq->hw_cq.cq_number); - ret = -EIO; - } + ret = -EIO; } else { ret = 0; - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } } + nes_put_cqp_request(nesdev, cqp_request); if (nescq->cq_mem_size) pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size, @@ -2105,15 +1996,8 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, " CQP Major:Minor codes = 0x%04X:0x%04X.\n", stag, ret, cqp_request->major_code, cqp_request->minor_code); major_code = cqp_request->major_code; - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + nes_put_cqp_request(nesdev, cqp_request); + if (!ret) return -ETIME; else if (major_code) @@ -2771,15 +2655,9 @@ static int nes_dereg_mr(struct ib_mr *ib_mr) major_code = cqp_request->major_code; minor_code = cqp_request->minor_code; - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + + nes_put_cqp_request(nesdev, cqp_request); + if (!ret) { nes_debug(NES_DBG_MR, "Timeout waiting to destroy STag," " ib_mr=%p, rkey = 0x%08X\n", @@ -2904,7 +2782,6 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp, /* struct iw_cm_id *cm_id = nesqp->cm_id; */ /* struct iw_cm_event cm_event; */ struct nes_cqp_request *cqp_request; - unsigned long flags; int ret; u16 major_code; @@ -2950,15 +2827,9 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp, nesqp->hwqp.qp_id, cqp_request->major_code, cqp_request->minor_code, next_iwarp_state); } - if (atomic_dec_and_test(&cqp_request->refcount)) { - if (cqp_request->dynamic) { - kfree(cqp_request); - } else { - spin_lock_irqsave(&nesdev->cqp.lock, flags); - list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs); - spin_unlock_irqrestore(&nesdev->cqp.lock, flags); - } - } + + nes_put_cqp_request(nesdev, cqp_request); + if (!ret) return -ETIME; else if (major_code) -- cgit v1.2.3 From 52c8084b740c42af27d5bfa62cec7079d12fbc2b Mon Sep 17 00:00:00 2001 From: Jon Mason Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: RDMA/cxgb3: Propagate HW page size capabilities cxgb3 does not currently report the page size capabilities, and incorrectly reports them internally. This version changes the bit-shifting to a static value (per Steve's request). Signed-off-by: Jon Mason Acked-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/cxio_hal.h | 1 + drivers/infiniband/hw/cxgb3/iwch.c | 2 +- drivers/infiniband/hw/cxgb3/iwch_provider.c | 1 + 3 files changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h index a9ff32c3621..656fe47bc84 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.h +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h @@ -55,6 +55,7 @@ #define T3_MAX_CQ_DEPTH 8192 #define T3_MAX_NUM_STAG (1<<15) #define T3_MAX_MR_SIZE 0x100000000ULL +#define T3_PAGESIZE_MASK 0xffff000 /* 4KB-128MB */ #define T3_STAG_UNSET 0xffffffff diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c index 06086256ca1..4489c89d671 100644 --- a/drivers/infiniband/hw/cxgb3/iwch.c +++ b/drivers/infiniband/hw/cxgb3/iwch.c @@ -80,7 +80,7 @@ static void rnic_init(struct iwch_dev *rnicp) rnicp->attr.max_mem_regs = cxio_num_stags(&rnicp->rdev); rnicp->attr.max_phys_buf_entries = T3_MAX_PBL_SIZE; rnicp->attr.max_pds = T3_MAX_NUM_PD - 1; - rnicp->attr.mem_pgsizes_bitmask = 0x7FFF; /* 4KB-128MB */ + rnicp->attr.mem_pgsizes_bitmask = T3_PAGESIZE_MASK; rnicp->attr.max_mr_size = T3_MAX_MR_SIZE; rnicp->attr.can_resize_wq = 0; rnicp->attr.max_rdma_reads_per_qp = 8; diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 18a6609f5e0..249d99f4a3c 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1130,6 +1130,7 @@ static int iwch_query_device(struct ib_device *ibdev, props->hw_ver = dev->rdev.t3cdev_p->type; props->fw_ver = fw_vers_string_to_u64(dev); props->device_cap_flags = dev->device_cap_flags; + props->page_size_cap = dev->attr.mem_pgsizes_bitmask; props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor; props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device; props->max_mr_size = dev->attr.max_mr_size; -- cgit v1.2.3 From 8294f29767c53e97664a27db9974adea8e2ea95b Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: RDMA/nes: Get rid of ring_doorbell parameter of nes_post_cqp_request() Every caller of nes_post_cqp_request() passed it NES_CQP_REQUEST_RING_DOORBELL, so just remove that parameter and always ring the doorbell. Signed-off-by: Roland Dreier Acked-by: Faisal Latif --- drivers/infiniband/hw/nes/nes.c | 2 +- drivers/infiniband/hw/nes/nes.h | 5 +---- drivers/infiniband/hw/nes/nes_hw.c | 6 +++--- drivers/infiniband/hw/nes/nes_hw.h | 2 +- drivers/infiniband/hw/nes/nes_utils.c | 9 ++++----- drivers/infiniband/hw/nes/nes_verbs.c | 18 +++++++++--------- 6 files changed, 19 insertions(+), 23 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c index a4e9269a29b..d2884e77809 100644 --- a/drivers/infiniband/hw/nes/nes.c +++ b/drivers/infiniband/hw/nes/nes.c @@ -328,7 +328,7 @@ void nes_rem_ref(struct ib_qp *ibqp) set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_ID_IDX, nesqp->hwqp.qp_id); u64temp = (u64)nesqp->nesqp_context_pbase; set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); } } diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h index fe88bec4894..39bd897b40c 100644 --- a/drivers/infiniband/hw/nes/nes.h +++ b/drivers/infiniband/hw/nes/nes.h @@ -94,9 +94,6 @@ #define MAX_DPC_ITERATIONS 128 -#define NES_CQP_REQUEST_NO_DOORBELL_RING 0 -#define NES_CQP_REQUEST_RING_DOORBELL 1 - #define NES_DRV_OPT_ENABLE_MPA_VER_0 0x00000001 #define NES_DRV_OPT_DISABLE_MPA_CRC 0x00000002 #define NES_DRV_OPT_DISABLE_FIRST_WRITE 0x00000004 @@ -542,7 +539,7 @@ void nes_free_cqp_request(struct nes_device *nesdev, struct nes_cqp_request *cqp_request); void nes_put_cqp_request(struct nes_device *nesdev, struct nes_cqp_request *cqp_request); -void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *, int); +void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *); int nes_arp_table(struct nes_device *, u32, u8 *, u32); void nes_mh_fix(unsigned long); void nes_clc(unsigned long); diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 80e486653ec..902b1375a5d 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -3147,7 +3147,7 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port, nes_debug(NES_DBG_QP, "Waiting for CQP completion for APBVT.\n"); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); if (add_port == NES_MANAGE_APBVT_ADD) ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0), @@ -3217,7 +3217,7 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr, nesdev->cqp.sq_head, nesdev->cqp.sq_tail); atomic_set(&cqp_request->refcount, 1); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); } @@ -3249,7 +3249,7 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp, cpu_to_le32(NES_CQP_FLUSH_WQES | which_wq); cqp_wqe->wqe_words[NES_CQP_WQE_ID_IDX] = cpu_to_le32(nesqp->hwqp.qp_id); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); if (wait_completion) { /* Wait for CQP */ diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index 745bf94f3f0..7b81e0ae007 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -1172,7 +1172,7 @@ struct nes_vnic { u32 mcrq_qp_id; struct nes_ucontext *mcrq_ucontext; struct nes_cqp_request* (*get_cqp_request)(struct nes_device *nesdev); - void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *, int); + void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *); int (*mcrq_mcast_filter)( struct nes_vnic* nesvnic, __u8* dmi_addr ); struct net_device_stats netstats; /* used to put the netdev on the adapters logical port list */ diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c index 8f519c86087..fb8cbd71a2e 100644 --- a/drivers/infiniband/hw/nes/nes_utils.c +++ b/drivers/infiniband/hw/nes/nes_utils.c @@ -596,7 +596,7 @@ void nes_put_cqp_request(struct nes_device *nesdev, * nes_post_cqp_request */ void nes_post_cqp_request(struct nes_device *nesdev, - struct nes_cqp_request *cqp_request, int ring_doorbell) + struct nes_cqp_request *cqp_request) { struct nes_hw_cqp_wqe *cqp_wqe; unsigned long flags; @@ -624,10 +624,9 @@ void nes_post_cqp_request(struct nes_device *nesdev, nesdev->cqp.sq_head, nesdev->cqp.sq_tail, nesdev->cqp.sq_size, cqp_request->waiting, atomic_read(&cqp_request->refcount)); barrier(); - if (ring_doorbell) { - /* Ring doorbell (1 WQEs) */ - nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id); - } + + /* Ring doorbell (1 WQEs) */ + nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id); barrier(); } else { diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index 464a98a6e11..e3939d13484 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -118,7 +118,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) { set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, stag); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0), @@ -175,7 +175,7 @@ static int nes_dealloc_mw(struct ib_mw *ibmw) set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ibmw->rkey); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X to complete.\n", @@ -477,7 +477,7 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd, (nesfmr->nesmr.pbls_used-1) : nesfmr->nesmr.pbls_used); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0), @@ -1417,7 +1417,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd, set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ nes_debug(NES_DBG_QP, "Waiting for create iWARP QP%u to complete.\n", @@ -1744,7 +1744,7 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries, cpu_to_le32(((u32)((u64temp) >> 33)) & 0x7FFFFFFF); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ nes_debug(NES_DBG_CQ, "Waiting for create iWARP CQ%u to complete.\n", @@ -1841,7 +1841,7 @@ static int nes_destroy_cq(struct ib_cq *ib_cq) (nescq->hw_cq.cq_number | ((u32)PCI_FUNC(nesdev->pcidev->devfn) << 16))); nes_free_resource(nesadapter, nesadapter->allocated_cqs, nescq->hw_cq.cq_number); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ nes_debug(NES_DBG_CQ, "Waiting for destroy iWARP CQ%u to complete.\n", @@ -1987,7 +1987,7 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, barrier(); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ ret = wait_event_timeout(cqp_request->waitq, (0 != cqp_request->request_done), @@ -2638,7 +2638,7 @@ static int nes_dereg_mr(struct ib_mr *ib_mr) set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ib_mr->rkey); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X completed\n", ib_mr->rkey); @@ -2809,7 +2809,7 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp, set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, (u64)nesqp->nesqp_context_pbase); atomic_set(&cqp_request->refcount, 2); - nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL); + nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ if (wait_completion) { -- cgit v1.2.3 From 70fe1796a5ebc5f955be39bba5c42eee9eb89e1f Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:49 -0700 Subject: RDMA/cxgb3: Set rkey field for new memory windows in iwch_alloc_mw() Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/iwch_provider.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 249d99f4a3c..c9a3893b38e 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -748,6 +748,7 @@ static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd) mhp->attr.type = TPT_MW; mhp->attr.stag = stag; mmid = (stag) >> 8; + mhp->ibmw.rkey = stag; insert_handle(rhp, &rhp->mmidr, mhp, mmid); PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag); return &(mhp->ibmw); -- cgit v1.2.3 From c03d4731b5b6de45b95a10bf1d510dde423d6757 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:50 -0700 Subject: IPoIB: Remove unused IPOIB_MCAST_STARTED code The IPOIB_MCAST_STARTED flag is not used at all since commit b3e2749b ("IPoIB: Don't drop multicast sends when they can be queued"), so remove it. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 1 - drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 8 -------- 2 files changed, 9 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index bb19587c5ea..66a897567ea 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -89,7 +89,6 @@ enum { IPOIB_FLAG_SUBINTERFACE = 5, IPOIB_MCAST_RUN = 6, IPOIB_STOP_REAPER = 7, - IPOIB_MCAST_STARTED = 8, IPOIB_FLAG_ADMIN_CM = 9, IPOIB_FLAG_UMCAST = 10, IPOIB_FLAG_CSUM = 11, diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 4a6538b9301..0b7d129161e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -592,10 +592,6 @@ int ipoib_mcast_start_thread(struct net_device *dev) queue_delayed_work(ipoib_workqueue, &priv->mcast_task, 0); mutex_unlock(&mcast_mutex); - spin_lock_irq(&priv->lock); - set_bit(IPOIB_MCAST_STARTED, &priv->flags); - spin_unlock_irq(&priv->lock); - return 0; } @@ -605,10 +601,6 @@ int ipoib_mcast_stop_thread(struct net_device *dev, int flush) ipoib_dbg_mcast(priv, "stopping multicast thread\n"); - spin_lock_irq(&priv->lock); - clear_bit(IPOIB_MCAST_STARTED, &priv->flags); - spin_unlock_irq(&priv->lock); - mutex_lock(&mcast_mutex); clear_bit(IPOIB_MCAST_RUN, &priv->flags); cancel_delayed_work(&priv->mcast_task); -- cgit v1.2.3 From 5892eff91ad60ba365ae7f75050ce464036c5396 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:50 -0700 Subject: IPoIB: Remove priv->mcast_mutex No need for a mutex around calls to ib_attach_mcast/ib_detach_mcast since these operations are synchronized at the HW driver layer. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 1 - drivers/infiniband/ulp/ipoib/ipoib_main.c | 1 - drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 4 ---- 3 files changed, 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 66a897567ea..b8753222c87 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -277,7 +277,6 @@ struct ipoib_dev_priv { unsigned long flags; - struct mutex mcast_mutex; struct mutex vlan_mutex; struct rb_root path_tree; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index b3fd7e8333c..8be9ea0436e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -1072,7 +1072,6 @@ static void ipoib_setup(struct net_device *dev) spin_lock_init(&priv->lock); spin_lock_init(&priv->tx_lock); - mutex_init(&priv->mcast_mutex); mutex_init(&priv->vlan_mutex); INIT_LIST_HEAD(&priv->path_list); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index 96f9aa79cbb..f50ebe0643e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -61,9 +61,7 @@ int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid) } /* attach QP to multicast group */ - mutex_lock(&priv->mcast_mutex); ret = ib_attach_mcast(priv->qp, mgid, mlid); - mutex_unlock(&priv->mcast_mutex); if (ret) ipoib_warn(priv, "failed to attach to multicast group, ret = %d\n", ret); @@ -77,9 +75,7 @@ int ipoib_mcast_detach(struct net_device *dev, u16 mlid, union ib_gid *mgid) struct ipoib_dev_priv *priv = netdev_priv(dev); int ret; - mutex_lock(&priv->mcast_mutex); ret = ib_detach_mcast(priv->qp, mgid, mlid); - mutex_unlock(&priv->mcast_mutex); if (ret) ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret); -- cgit v1.2.3 From d0de13622d5ac658efe7c51521dbdbe0752aa3dd Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:50 -0700 Subject: IPoIB: Only set Q_Key once: after joining broadcast group The current code will set the Q_Key for any join of a non-sendonly multicast group. The operation involves a modify QP operation, which is fairly heavyweight, and is only really required after the join of the broadcast group. Fix this by adding a parameter to ipoib_mcast_attach() to control when the Q_Key is set. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 2 +- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 4 +++- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 28 ++++++++++++++------------ 3 files changed, 19 insertions(+), 15 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index b8753222c87..7b46e2d7b3c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -485,7 +485,7 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter, #endif int ipoib_mcast_attach(struct net_device *dev, u16 mlid, - union ib_gid *mgid); + union ib_gid *mgid, int set_qkey); int ipoib_mcast_detach(struct net_device *dev, u16 mlid, union ib_gid *mgid); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 0b7d129161e..55ebd950bf2 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -186,6 +186,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_ah *ah; int ret; + int set_qkey = 0; mcast->mcmember = *mcmember; @@ -200,6 +201,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, priv->qkey = be32_to_cpu(priv->broadcast->mcmember.qkey); spin_unlock_irq(&priv->lock); priv->tx_wr.wr.ud.remote_qkey = priv->qkey; + set_qkey = 1; } if (!test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { @@ -212,7 +214,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, } ret = ipoib_mcast_attach(dev, be16_to_cpu(mcast->mcmember.mlid), - &mcast->mcmember.mgid); + &mcast->mcmember.mgid, set_qkey); if (ret < 0) { ipoib_warn(priv, "couldn't attach QP to multicast group " IPOIB_GID_FMT "\n", diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index f50ebe0643e..ba7c8868e6f 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -33,18 +33,13 @@ #include "ipoib.h" -int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid) +int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid, int set_qkey) { struct ipoib_dev_priv *priv = netdev_priv(dev); - struct ib_qp_attr *qp_attr; + struct ib_qp_attr *qp_attr = NULL; int ret; u16 pkey_index; - ret = -ENOMEM; - qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL); - if (!qp_attr) - goto out; - if (ib_find_pkey(priv->ca, priv->port, priv->pkey, &pkey_index)) { clear_bit(IPOIB_PKEY_ASSIGNED, &priv->flags); ret = -ENXIO; @@ -52,12 +47,19 @@ int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid) } set_bit(IPOIB_PKEY_ASSIGNED, &priv->flags); - /* set correct QKey for QP */ - qp_attr->qkey = priv->qkey; - ret = ib_modify_qp(priv->qp, qp_attr, IB_QP_QKEY); - if (ret) { - ipoib_warn(priv, "failed to modify QP, ret = %d\n", ret); - goto out; + if (set_qkey) { + ret = -ENOMEM; + qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL); + if (!qp_attr) + goto out; + + /* set correct QKey for QP */ + qp_attr->qkey = priv->qkey; + ret = ib_modify_qp(priv->qp, qp_attr, IB_QP_QKEY); + if (ret) { + ipoib_warn(priv, "failed to modify QP, ret = %d\n", ret); + goto out; + } } /* attach QP to multicast group */ -- cgit v1.2.3 From 9eae554c171e086c89ab83da2a2d3c8bf958fcb5 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:50 -0700 Subject: IPoIB: Get rid of ipoib_mcast_detach() wrapper ipoib_mcast_detach() does nothing except call ib_detach_mcast(), so just use the core API in the one place that does a multicast group detach. add/remove: 0/1 grow/shrink: 0/1 up/down: 0/-105 (-105) function old new delta ipoib_mcast_leave 357 319 -38 ipoib_mcast_detach 67 - -67 Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 2 -- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 6 +++--- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 12 ------------ 3 files changed, 3 insertions(+), 17 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 7b46e2d7b3c..a89b9fbe1ef 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -486,8 +486,6 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter, int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid, int set_qkey); -int ipoib_mcast_detach(struct net_device *dev, u16 mlid, - union ib_gid *mgid); int ipoib_init_qp(struct net_device *dev); int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 55ebd950bf2..71add7a8d53 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -627,10 +627,10 @@ static int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) IPOIB_GID_ARG(mcast->mcmember.mgid)); /* Remove ourselves from the multicast group */ - ret = ipoib_mcast_detach(dev, be16_to_cpu(mcast->mcmember.mlid), - &mcast->mcmember.mgid); + ret = ib_detach_mcast(priv->qp, &mcast->mcmember.mgid, + be16_to_cpu(mcast->mcmember.mlid)); if (ret) - ipoib_warn(priv, "ipoib_mcast_detach failed (result = %d)\n", ret); + ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret); } return 0; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index ba7c8868e6f..68325119f74 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -72,18 +72,6 @@ out: return ret; } -int ipoib_mcast_detach(struct net_device *dev, u16 mlid, union ib_gid *mgid) -{ - struct ipoib_dev_priv *priv = netdev_priv(dev); - int ret; - - ret = ib_detach_mcast(priv->qp, mgid, mlid); - if (ret) - ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret); - - return ret; -} - int ipoib_init_qp(struct net_device *dev) { struct ipoib_dev_priv *priv = netdev_priv(dev); -- cgit v1.2.3 From c8c2afe360b7366f586f6bece1109a72ea334876 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:51 -0700 Subject: IPoIB: Use rtnl lock/unlock when changing device flags Use of this lock is required to synchronize changes to the netdvice's data structs. Also move the call to ipoib_flush_paths() after the modification of the netdevice flags in set_mode(). Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 8 ++++++-- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 5 ++++- 2 files changed, 10 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 37bf67b2a26..b4269139135 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -1440,7 +1440,9 @@ static ssize_t set_mode(struct device *d, struct device_attribute *attr, ipoib_warn(priv, "enabling connected mode " "will cause multicast packet drops\n"); + rtnl_lock(); dev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_SG | NETIF_F_TSO); + rtnl_unlock(); priv->tx_wr.send_flags &= ~IB_SEND_IP_CSUM; ipoib_flush_paths(dev); @@ -1449,14 +1451,16 @@ static ssize_t set_mode(struct device *d, struct device_attribute *attr, if (!strcmp(buf, "datagram\n")) { clear_bit(IPOIB_FLAG_ADMIN_CM, &priv->flags); - dev->mtu = min(priv->mcast_mtu, dev->mtu); - ipoib_flush_paths(dev); + rtnl_lock(); if (test_bit(IPOIB_FLAG_CSUM, &priv->flags)) { dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG; if (priv->hca_caps & IB_DEVICE_UD_TSO) dev->features |= NETIF_F_TSO; } + dev->mtu = min(priv->mcast_mtu, dev->mtu); + rtnl_unlock(); + ipoib_flush_paths(dev); return count; } diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 71add7a8d53..be1ed38cdcf 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -575,8 +575,11 @@ void ipoib_mcast_join_task(struct work_struct *work) priv->mcast_mtu = IPOIB_UD_MTU(ib_mtu_enum_to_int(priv->broadcast->mcmember.mtu)); - if (!ipoib_cm_admin_enabled(dev)) + if (!ipoib_cm_admin_enabled(dev)) { + rtnl_lock(); dev->mtu = min(priv->mcast_mtu, priv->admin_mtu); + rtnl_unlock(); + } ipoib_dbg_mcast(priv, "successfully joined all multicast groups\n"); -- cgit v1.2.3 From bd3606715effbf37df986548c43bbed0842b49d5 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:51 -0700 Subject: IPoIB: Use dev_set_mtu() to change mtu When the driver sets the MTU of the net device outside of its change_mtu method, it should make use of dev_set_mtu() instead of directly setting the mtu field of struct netdevice. Otherwise functions registered to be called upon MTU change will not get called (this is done through call_netdevice_notifiers() in dev_set_mtu()). Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 2 +- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index b4269139135..87f9f3ef3b2 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -1458,7 +1458,7 @@ static ssize_t set_mode(struct device *d, struct device_attribute *attr, if (priv->hca_caps & IB_DEVICE_UD_TSO) dev->features |= NETIF_F_TSO; } - dev->mtu = min(priv->mcast_mtu, dev->mtu); + dev_set_mtu(dev, min(priv->mcast_mtu, dev->mtu)); rtnl_unlock(); ipoib_flush_paths(dev); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index be1ed38cdcf..1fcc9a898d8 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -577,7 +577,7 @@ void ipoib_mcast_join_task(struct work_struct *work) if (!ipoib_cm_admin_enabled(dev)) { rtnl_lock(); - dev->mtu = min(priv->mcast_mtu, priv->admin_mtu); + dev_set_mtu(dev, min(priv->mcast_mtu, priv->admin_mtu)); rtnl_unlock(); } -- cgit v1.2.3 From df8666198dd058b9498ebdbc52c61957206d30a5 Mon Sep 17 00:00:00 2001 From: Ralph Campbell Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IB/ipath: Use IEEE OUI for vendor_id reported by ibv_query_device() The IB spe. for SubnGet(NodeInfo) and query HCA says that the vendor ID field should be the IEEE OUI assigned to the vendor. The ipath driver was returning the PCI vendor ID instead. This will affect applications which call ibv_query_device(). The old value was 0x001fc1 or 0x001077, the new value is 0x001175. The vendor ID doesn't appear to be exported via /sys so that should reduce possible compatibility issues. I'm only aware of Open MPI as a major application which depends on this change, and they have made necessary adjustments. Signed-off-by: Ralph Campbell Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ipath/ipath_mad.c | 6 +++--- drivers/infiniband/hw/ipath/ipath_verbs.c | 3 ++- 2 files changed, 5 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c index 5f9315d77a4..be4fc9ada8e 100644 --- a/drivers/infiniband/hw/ipath/ipath_mad.c +++ b/drivers/infiniband/hw/ipath/ipath_mad.c @@ -111,9 +111,9 @@ static int recv_subn_get_nodeinfo(struct ib_smp *smp, nip->revision = cpu_to_be32((majrev << 16) | minrev); nip->local_port_num = port; vendor = dd->ipath_vendorid; - nip->vendor_id[0] = 0; - nip->vendor_id[1] = vendor >> 8; - nip->vendor_id[2] = vendor; + nip->vendor_id[0] = IPATH_SRC_OUI_1; + nip->vendor_id[1] = IPATH_SRC_OUI_2; + nip->vendor_id[2] = IPATH_SRC_OUI_3; return reply(smp); } diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c index 7779165b2c2..9e23ab0b51a 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs.c @@ -1497,7 +1497,8 @@ static int ipath_query_device(struct ib_device *ibdev, IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN | IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE; props->page_size_cap = PAGE_SIZE; - props->vendor_id = dev->dd->ipath_vendorid; + props->vendor_id = + IPATH_SRC_OUI_1 << 16 | IPATH_SRC_OUI_2 << 8 | IPATH_SRC_OUI_3; props->vendor_part_id = dev->dd->ipath_deviceid; props->hw_ver = dev->dd->ipath_pcirev; -- cgit v1.2.3 From e112373fd6aa280bd2cbc0d5cc3809115325a1be Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IPoIB/cm: Reduce connected mode TX object size Since IPoIB connected mode does not NETIF_F_SG, we only have one DMA mapping per send, so we don't need a mapping[] array. Define a new struct with a single u64 mapping member and use it for the CM tx_ring. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 7 ++++++- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 12 ++++++------ 2 files changed, 12 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index a89b9fbe1ef..0281c8fecc9 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -157,6 +157,11 @@ struct ipoib_tx_buf { u64 mapping[MAX_SKB_FRAGS + 1]; }; +struct ipoib_cm_tx_buf { + struct sk_buff *skb; + u64 mapping; +}; + struct ib_cm_id; struct ipoib_cm_data { @@ -215,7 +220,7 @@ struct ipoib_cm_tx { struct net_device *dev; struct ipoib_neigh *neigh; struct ipoib_path *path; - struct ipoib_tx_buf *tx_ring; + struct ipoib_cm_tx_buf *tx_ring; unsigned tx_head; unsigned tx_tail; unsigned long flags; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 87f9f3ef3b2..0f2d3045061 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -703,7 +703,7 @@ static inline int post_send(struct ipoib_dev_priv *priv, void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_tx *tx) { struct ipoib_dev_priv *priv = netdev_priv(dev); - struct ipoib_tx_buf *tx_req; + struct ipoib_cm_tx_buf *tx_req; u64 addr; if (unlikely(skb->len > tx->mtu)) { @@ -734,7 +734,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ return; } - tx_req->mapping[0] = addr; + tx_req->mapping = addr; if (unlikely(post_send(priv, tx, tx->tx_head & (ipoib_sendq_size - 1), addr, skb->len))) { @@ -759,7 +759,7 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_cm_tx *tx = wc->qp->qp_context; unsigned int wr_id = wc->wr_id & ~IPOIB_OP_CM; - struct ipoib_tx_buf *tx_req; + struct ipoib_cm_tx_buf *tx_req; unsigned long flags; ipoib_dbg_data(priv, "cm send completion: id %d, status: %d\n", @@ -773,7 +773,7 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc) tx_req = &tx->tx_ring[wr_id]; - ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len, DMA_TO_DEVICE); + ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len, DMA_TO_DEVICE); /* FIXME: is this right? Shouldn't we only increment on success? */ ++dev->stats.tx_packets; @@ -1143,7 +1143,7 @@ err_tx: static void ipoib_cm_tx_destroy(struct ipoib_cm_tx *p) { struct ipoib_dev_priv *priv = netdev_priv(p->dev); - struct ipoib_tx_buf *tx_req; + struct ipoib_cm_tx_buf *tx_req; unsigned long flags; unsigned long begin; @@ -1171,7 +1171,7 @@ timeout: while ((int) p->tx_tail - (int) p->tx_head < 0) { tx_req = &p->tx_ring[p->tx_tail & (ipoib_sendq_size - 1)]; - ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len, + ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len, DMA_TO_DEVICE); dev_kfree_skb_any(tx_req->skb); ++p->tx_tail; -- cgit v1.2.3 From bc3a290b51aaefc6a6af2d6e6d52ed32387c416c Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IPoIB: Double default RX/TX ring sizes Increase IPoIB ring sizes to twice their original sizes (RX: 128->256, TX: 64->128) to act as a shock absorber for high traffic peaks. With the current settings, we have seen cases that there are many calls to netif_stop_queue(), which causes degradation in throughput. Also, larger receive buffer sizes help IPoIB in CM mode to avoid experiencing RNR NAK conditions due to insufficient receive buffers at the SRQ. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 0281c8fecc9..b0ffc9abe8c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -70,8 +70,8 @@ enum { IPOIB_CM_BUF_SIZE = IPOIB_CM_MTU + IPOIB_ENCAP_LEN, IPOIB_CM_HEAD_SIZE = IPOIB_CM_BUF_SIZE % PAGE_SIZE, IPOIB_CM_RX_SG = ALIGN(IPOIB_CM_BUF_SIZE, PAGE_SIZE) / PAGE_SIZE, - IPOIB_RX_RING_SIZE = 128, - IPOIB_TX_RING_SIZE = 64, + IPOIB_RX_RING_SIZE = 256, + IPOIB_TX_RING_SIZE = 128, IPOIB_MAX_QUEUE_SIZE = 8192, IPOIB_MIN_QUEUE_SIZE = 2, IPOIB_CM_MAX_CONN_QP = 4096, -- cgit v1.2.3 From 4522e08ced48baaf28990e2674e940aae9940310 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IB/mthca: Remove "stop" flag for catastrophic error polling timer Since we use del_timer_sync() anyway, there's no need for an additional flag to tell the timer not to rearm. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_catas.c | 15 ++------------- drivers/infiniband/hw/mthca/mthca_dev.h | 1 - 2 files changed, 2 insertions(+), 14 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c index 40573e48439..50e2792e60c 100644 --- a/drivers/infiniband/hw/mthca/mthca_catas.c +++ b/drivers/infiniband/hw/mthca/mthca_catas.c @@ -126,7 +126,6 @@ static void handle_catas(struct mthca_dev *dev) static void poll_catas(unsigned long dev_ptr) { struct mthca_dev *dev = (struct mthca_dev *) dev_ptr; - unsigned long flags; int i; for (i = 0; i < dev->catas_err.size; ++i) @@ -135,13 +134,8 @@ static void poll_catas(unsigned long dev_ptr) return; } - spin_lock_irqsave(&catas_lock, flags); - if (!dev->catas_err.stop) - mod_timer(&dev->catas_err.timer, - jiffies + MTHCA_CATAS_POLL_INTERVAL); - spin_unlock_irqrestore(&catas_lock, flags); - - return; + mod_timer(&dev->catas_err.timer, + jiffies + MTHCA_CATAS_POLL_INTERVAL); } void mthca_start_catas_poll(struct mthca_dev *dev) @@ -149,7 +143,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev) unsigned long addr; init_timer(&dev->catas_err.timer); - dev->catas_err.stop = 0; dev->catas_err.map = NULL; addr = pci_resource_start(dev->pdev, 0) + @@ -180,10 +173,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev) void mthca_stop_catas_poll(struct mthca_dev *dev) { - spin_lock_irq(&catas_lock); - dev->catas_err.stop = 1; - spin_unlock_irq(&catas_lock); - del_timer_sync(&dev->catas_err.timer); if (dev->catas_err.map) { diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h index 2997d8d564e..ee4d073c889 100644 --- a/drivers/infiniband/hw/mthca/mthca_dev.h +++ b/drivers/infiniband/hw/mthca/mthca_dev.h @@ -277,7 +277,6 @@ struct mthca_mcg_table { struct mthca_catas_err { u64 addr; u32 __iomem *map; - unsigned long stop; u32 size; struct timer_list timer; struct list_head list; -- cgit v1.2.3 From c036925ac0e940a4e8525b08e89d2c64fe282c51 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IB/mthca: Use round_jiffies() for catastrophic error polling timer Exactly when the catastrophic error polling timer function runs is not important, so use round_jiffies() to save unnecessary wakeups. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_catas.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c index 50e2792e60c..cc440f90000 100644 --- a/drivers/infiniband/hw/mthca/mthca_catas.c +++ b/drivers/infiniband/hw/mthca/mthca_catas.c @@ -135,7 +135,7 @@ static void poll_catas(unsigned long dev_ptr) } mod_timer(&dev->catas_err.timer, - jiffies + MTHCA_CATAS_POLL_INTERVAL); + round_jiffies(jiffies + MTHCA_CATAS_POLL_INTERVAL)); } void mthca_start_catas_poll(struct mthca_dev *dev) -- cgit v1.2.3 From aed012279d35e88e29fd55737d8821604433f50a Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 14 Jul 2008 23:48:52 -0700 Subject: IB/mthca: Fix check of max_send_sge for special QPs The MLX transport requires two extra gather entries for sends (one for the header and one for the checksum at the end, as the comment says). However the code checked that max_recv_sge was not too big, instead of checking max_send_sge as it should have. Fix the code to check the correct condition. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_qp.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index 77b52b3adcc..f5081bfde6d 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -1249,10 +1249,10 @@ static int mthca_set_qp_size(struct mthca_dev *dev, struct ib_qp_cap *cap, return -EINVAL; /* - * For MLX transport we need 2 extra S/G entries: + * For MLX transport we need 2 extra send gather entries: * one for the header and one for the checksum at the end */ - if (qp->transport == MLX && cap->max_recv_sge + 2 > dev->limits.max_sg) + if (qp->transport == MLX && cap->max_send_sge + 2 > dev->limits.max_sg) return -EINVAL; if (mthca_is_memfree(dev)) { -- cgit v1.2.3 From 96f15c03532282366364ecfd20f04e49b5d96f3a Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:53 -0700 Subject: RDMA/core: Add local DMA L_Key support - Change the IB_DEVICE_ZERO_STAG flag to the transport-neutral name IB_DEVICE_LOCAL_DMA_LKEY, which is used by iWARP RNICs to indicate 0 STag support and IB HCAs to indicate reserved L_Key support. - Add a u32 local_dma_lkey member to struct ib_device. Drivers fill this in with the appropriate local DMA L_Key (if they support it). - Fix up the drivers using this flag. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/amso1100/c2_rnic.c | 2 +- drivers/infiniband/hw/cxgb3/iwch_provider.c | 6 ++++-- drivers/infiniband/hw/nes/nes_hw.c | 2 +- 3 files changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/amso1100/c2_rnic.c b/drivers/infiniband/hw/amso1100/c2_rnic.c index b1441aeb60c..dd05c483564 100644 --- a/drivers/infiniband/hw/amso1100/c2_rnic.c +++ b/drivers/infiniband/hw/amso1100/c2_rnic.c @@ -454,7 +454,7 @@ int __devinit c2_rnic_init(struct c2_dev *c2dev) (IB_DEVICE_RESIZE_MAX_WR | IB_DEVICE_CURR_QP_STATE_MOD | IB_DEVICE_SYS_IMAGE_GUID | - IB_DEVICE_ZERO_STAG | + IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW); /* Allocate the qptr_array */ diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index c9a3893b38e..7ecfd4d638c 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1325,8 +1325,10 @@ int iwch_register_device(struct iwch_dev *dev) memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid)); memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6); dev->ibdev.owner = THIS_MODULE; - dev->device_cap_flags = IB_DEVICE_ZERO_STAG | - IB_DEVICE_MEM_WINDOW; + dev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW; + + /* cxgb3 supports STag 0. */ + dev->ibdev.local_dma_lkey = 0; if (fw_supports_fastreg(dev)) dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS; diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 902b1375a5d..85f26d19a32 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -398,7 +398,7 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) { nesadapter->base_pd = 1; nesadapter->device_cap_flags = - IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW; + IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW; nesadapter->allocated_qps = (unsigned long *)&(((unsigned char *)nesadapter) [(sizeof(struct nes_adapter)+(sizeof(unsigned long)-1))&(~(sizeof(unsigned long)-1))]); -- cgit v1.2.3 From 4ab928f69208d240d3681336f34589e4b151824f Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Mon, 14 Jul 2008 23:48:53 -0700 Subject: RDMA/cxgb3: Fixes for zero STag Handling the zero STag in receive work request requires some extra logic in the driver: - Only set the QP_PRIV bit for kernel mode QPs. - Add a zero STag build function for recv wrs. The uP needs a PBL allocated and passed down in the recv WR so it can construct a HW PBL for the zero STag S/G entries. Note: we need to place a few restrictions on zero STag usage because of this: 1) all SGEs in a recv WR must either be zero STag or not. No mixing. 2) an individual SGE length cannot exceed 128MB for a zero-stag SGE. This should be OK since it's not really practical to allocate such a large chunk of pinned contiguous DMA mapped memory. - Add an optimized non-zero-STag recv wr format for kernel users. This is needed to optimize both zero and non-zero STag cracking in the recv path for kernel users. - Remove the iwch_ prefix from the static build functions. - Bump required FW version. Signed-off-by: Steve Wise --- drivers/infiniband/hw/cxgb3/cxio_hal.c | 12 ++- drivers/infiniband/hw/cxgb3/cxio_wr.h | 13 ++- drivers/infiniband/hw/cxgb3/iwch_provider.c | 4 +- drivers/infiniband/hw/cxgb3/iwch_qp.c | 127 ++++++++++++++++++++++++---- 4 files changed, 130 insertions(+), 26 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c index 340e4181c76..f6d5747153a 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_hal.c +++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c @@ -278,7 +278,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain, if (!wq->qpid) return -ENOMEM; - wq->rq = kzalloc(depth * sizeof(u64), GFP_KERNEL); + wq->rq = kzalloc(depth * sizeof(struct t3_swrq), GFP_KERNEL); if (!wq->rq) goto err1; @@ -302,6 +302,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain, if (!kernel_domain) wq->udb = (u64)rdev_p->rnic_info.udbell_physbase + (wq->qpid << rdev_p->qpshift); + wq->rdev = rdev_p; PDBG("%s qpid 0x%x doorbell 0x%p udb 0x%llx\n", __func__, wq->qpid, wq->doorbell, (unsigned long long) wq->udb); return 0; @@ -1266,13 +1267,16 @@ proc_cqe: wq->sq_rptr = CQE_WRID_SQ_WPTR(*hw_cqe); PDBG("%s completing sq idx %ld\n", __func__, Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2)); - *cookie = (wq->sq + - Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2))->wr_id; + *cookie = wq->sq[Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2)].wr_id; wq->sq_rptr++; } else { PDBG("%s completing rq idx %ld\n", __func__, Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)); - *cookie = *(wq->rq + Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)); + *cookie = wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].wr_id; + if (wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].pbl_addr) + cxio_hal_pblpool_free(wq->rdev, + wq->rq[Q_PTR2IDX(wq->rq_rptr, + wq->rq_size_log2)].pbl_addr, T3_STAG0_PBL_SIZE); wq->rq_rptr++; } diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h index de760e9f1cc..04618f7bfbb 100644 --- a/drivers/infiniband/hw/cxgb3/cxio_wr.h +++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h @@ -39,6 +39,9 @@ #define T3_MAX_SGE 4 #define T3_MAX_INLINE 64 +#define T3_STAG0_PBL_SIZE (2 * T3_MAX_SGE << 3) +#define T3_STAG0_MAX_PBE_LEN (128 * 1024 * 1024) +#define T3_STAG0_PAGE_SHIFT 15 #define Q_EMPTY(rptr,wptr) ((rptr)==(wptr)) #define Q_FULL(rptr,wptr,size_log2) ( (((wptr)-(rptr))>>(size_log2)) && \ @@ -665,6 +668,11 @@ struct t3_swsq { int signaled; }; +struct t3_swrq { + __u64 wr_id; + __u32 pbl_addr; +}; + /* * A T3 WQ implements both the SQ and RQ. */ @@ -681,14 +689,15 @@ struct t3_wq { u32 sq_wptr; /* sq_wptr - sq_rptr == count of */ u32 sq_rptr; /* pending wrs */ u32 sq_size_log2; /* sq size */ - u64 *rq; /* SW RQ (holds consumer wr_ids */ + struct t3_swrq *rq; /* SW RQ (holds consumer wr_ids */ u32 rq_wptr; /* rq_wptr - rq_rptr == count of */ u32 rq_rptr; /* pending wrs */ - u64 *rq_oldest_wr; /* oldest wr on the SW RQ */ + struct t3_swrq *rq_oldest_wr; /* oldest wr on the SW RQ */ u32 rq_size_log2; /* rq size */ u32 rq_addr; /* rq adapter address */ void __iomem *doorbell; /* kernel db */ u64 udb; /* user db if any */ + struct cxio_rdev *rdev; }; struct t3_cq { diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c index 7ecfd4d638c..b89640aa6e1 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.c +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -1007,10 +1007,10 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd, qhp->ibqp.qp_num = qhp->wq.qpid; init_timer(&(qhp->timer)); PDBG("%s sq_num_entries %d, rq_num_entries %d " - "qpid 0x%0x qhp %p dma_addr 0x%llx size %d\n", + "qpid 0x%0x qhp %p dma_addr 0x%llx size %d rq_addr 0x%x\n", __func__, qhp->attr.sq_num_entries, qhp->attr.rq_num_entries, qhp->wq.qpid, qhp, (unsigned long long) qhp->wq.dma_addr, - 1 << qhp->wq.size_log2); + 1 << qhp->wq.size_log2, qhp->wq.rq_addr); return &qhp->ibqp; } diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c index 3b44300a303..9a3be3a9d5d 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_qp.c +++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c @@ -33,10 +33,11 @@ #include "iwch.h" #include "iwch_cm.h" #include "cxio_hal.h" +#include "cxio_resource.h" #define NO_SUPPORT -1 -static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr, +static int build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr, u8 * flit_cnt) { int i; @@ -81,7 +82,7 @@ static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr, return 0; } -static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr, +static int build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr, u8 *flit_cnt) { int i; @@ -122,7 +123,7 @@ static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr, return 0; } -static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr, +static int build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr, u8 *flit_cnt) { if (wr->num_sge > 1) @@ -143,7 +144,7 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr, return 0; } -static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr, +static int build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr, u8 *flit_cnt, int *wr_cnt, struct t3_wq *wq) { int i; @@ -185,7 +186,7 @@ static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr, return 0; } -static int iwch_build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr, +static int build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr, u8 *flit_cnt) { wqe->local_inv.stag = cpu_to_be32(wr->ex.invalidate_rkey); @@ -244,23 +245,106 @@ static int iwch_sgl2pbl_map(struct iwch_dev *rhp, struct ib_sge *sg_list, return 0; } -static int iwch_build_rdma_recv(struct iwch_dev *rhp, union t3_wr *wqe, +static int build_rdma_recv(struct iwch_qp *qhp, union t3_wr *wqe, struct ib_recv_wr *wr) { - int i; - if (wr->num_sge > T3_MAX_SGE) - return -EINVAL; + int i, err = 0; + u32 pbl_addr[T3_MAX_SGE]; + u8 page_size[T3_MAX_SGE]; + + err = iwch_sgl2pbl_map(qhp->rhp, wr->sg_list, wr->num_sge, pbl_addr, + page_size); + if (err) + return err; + wqe->recv.pagesz[0] = page_size[0]; + wqe->recv.pagesz[1] = page_size[1]; + wqe->recv.pagesz[2] = page_size[2]; + wqe->recv.pagesz[3] = page_size[3]; wqe->recv.num_sgle = cpu_to_be32(wr->num_sge); for (i = 0; i < wr->num_sge; i++) { wqe->recv.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey); wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length); + + /* to in the WQE == the offset into the page */ + wqe->recv.sgl[i].to = cpu_to_be64(((u32) wr->sg_list[i].addr) % + (1UL << (12 + page_size[i]))); + + /* pbl_addr is the adapters address in the PBL */ + wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_addr[i]); + } + for (; i < T3_MAX_SGE; i++) { + wqe->recv.sgl[i].stag = 0; + wqe->recv.sgl[i].len = 0; + wqe->recv.sgl[i].to = 0; + wqe->recv.pbl_addr[i] = 0; + } + qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, + qhp->wq.rq_size_log2)].wr_id = wr->wr_id; + qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, + qhp->wq.rq_size_log2)].pbl_addr = 0; + return 0; +} + +static int build_zero_stag_recv(struct iwch_qp *qhp, union t3_wr *wqe, + struct ib_recv_wr *wr) +{ + int i; + u32 pbl_addr; + u32 pbl_offset; + + + /* + * The T3 HW requires the PBL in the HW recv descriptor to reference + * a PBL entry. So we allocate the max needed PBL memory here and pass + * it to the uP in the recv WR. The uP will build the PBL and setup + * the HW recv descriptor. + */ + pbl_addr = cxio_hal_pblpool_alloc(&qhp->rhp->rdev, T3_STAG0_PBL_SIZE); + if (!pbl_addr) + return -ENOMEM; + + /* + * Compute the 8B aligned offset. + */ + pbl_offset = (pbl_addr - qhp->rhp->rdev.rnic_info.pbl_base) >> 3; + + wqe->recv.num_sgle = cpu_to_be32(wr->num_sge); + + for (i = 0; i < wr->num_sge; i++) { + + /* + * Use a 128MB page size. This and an imposed 128MB + * sge length limit allows us to require only a 2-entry HW + * PBL for each SGE. This restriction is acceptable since + * since it is not possible to allocate 128MB of contiguous + * DMA coherent memory! + */ + if (wr->sg_list[i].length > T3_STAG0_MAX_PBE_LEN) + return -EINVAL; + wqe->recv.pagesz[i] = T3_STAG0_PAGE_SHIFT; + + /* + * T3 restricts a recv to all zero-stag or all non-zero-stag. + */ + if (wr->sg_list[i].lkey != 0) + return -EINVAL; + wqe->recv.sgl[i].stag = 0; + wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length); wqe->recv.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr); + wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_offset); + pbl_offset += 2; } for (; i < T3_MAX_SGE; i++) { + wqe->recv.pagesz[i] = 0; wqe->recv.sgl[i].stag = 0; wqe->recv.sgl[i].len = 0; wqe->recv.sgl[i].to = 0; + wqe->recv.pbl_addr[i] = 0; } + qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, + qhp->wq.rq_size_log2)].wr_id = wr->wr_id; + qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, + qhp->wq.rq_size_log2)].pbl_addr = pbl_addr; return 0; } @@ -312,18 +396,18 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, if (wr->send_flags & IB_SEND_FENCE) t3_wr_flags |= T3_READ_FENCE_FLAG; t3_wr_opcode = T3_WR_SEND; - err = iwch_build_rdma_send(wqe, wr, &t3_wr_flit_cnt); + err = build_rdma_send(wqe, wr, &t3_wr_flit_cnt); break; case IB_WR_RDMA_WRITE: case IB_WR_RDMA_WRITE_WITH_IMM: t3_wr_opcode = T3_WR_WRITE; - err = iwch_build_rdma_write(wqe, wr, &t3_wr_flit_cnt); + err = build_rdma_write(wqe, wr, &t3_wr_flit_cnt); break; case IB_WR_RDMA_READ: case IB_WR_RDMA_READ_WITH_INV: t3_wr_opcode = T3_WR_READ; t3_wr_flags = 0; /* T3 reads are always signaled */ - err = iwch_build_rdma_read(wqe, wr, &t3_wr_flit_cnt); + err = build_rdma_read(wqe, wr, &t3_wr_flit_cnt); if (err) break; sqp->read_len = wqe->read.local_len; @@ -332,14 +416,14 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, break; case IB_WR_FAST_REG_MR: t3_wr_opcode = T3_WR_FASTREG; - err = iwch_build_fastreg(wqe, wr, &t3_wr_flit_cnt, + err = build_fastreg(wqe, wr, &t3_wr_flit_cnt, &wr_cnt, &qhp->wq); break; case IB_WR_LOCAL_INV: if (wr->send_flags & IB_SEND_FENCE) t3_wr_flags |= T3_LOCAL_FENCE_FLAG; t3_wr_opcode = T3_WR_INV_STAG; - err = iwch_build_inv_stag(wqe, wr, &t3_wr_flit_cnt); + err = build_inv_stag(wqe, wr, &t3_wr_flit_cnt); break; default: PDBG("%s post of type=%d TBD!\n", __func__, @@ -398,18 +482,24 @@ int iwch_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr, return -EINVAL; } while (wr) { + if (wr->num_sge > T3_MAX_SGE) { + err = -EINVAL; + *bad_wr = wr; + break; + } idx = Q_PTR2IDX(qhp->wq.wptr, qhp->wq.size_log2); wqe = (union t3_wr *) (qhp->wq.queue + idx); if (num_wrs) - err = iwch_build_rdma_recv(qhp->rhp, wqe, wr); + if (wr->sg_list[0].lkey) + err = build_rdma_recv(qhp, wqe, wr); + else + err = build_zero_stag_recv(qhp, wqe, wr); else err = -ENOMEM; if (err) { *bad_wr = wr; break; } - qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, qhp->wq.rq_size_log2)] = - wr->wr_id; build_fw_riwrh((void *) wqe, T3_WR_RCV, T3_COMPLETION_FLAG, Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), 0, sizeof(struct t3_receive_wr) >> 3, T3_SOPEOP); @@ -810,7 +900,8 @@ static int rdma_init(struct iwch_dev *rhp, struct iwch_qp *qhp, init_attr.qp_dma_size = (1UL << qhp->wq.size_log2); init_attr.rqe_count = iwch_rqes_posted(qhp); init_attr.flags = qhp->attr.mpa_attr.initiator ? MPA_INITIATOR : 0; - init_attr.flags |= capable(CAP_NET_BIND_SERVICE) ? PRIV_QP : 0; + if (!qhp->ibqp.uobject) + init_attr.flags |= PRIV_QP; if (peer2peer) { init_attr.rtr_type = RTR_READ; if (init_attr.ord == 0 && qhp->attr.mpa_attr.initiator) -- cgit v1.2.3 From 64c5e613b9dd34ef1281ed6d22478609667ae36a Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Mon, 14 Jul 2008 23:48:53 -0700 Subject: RDMA/addr: Keep pointer to netdevice in struct rdma_dev_addr Keep a pointer to the local (src) netdevice in struct rdma_dev_addr, and copy it in as part of rdma_copy_addr(). Use rdma_translate_ip() in cma_new_conn_id() to reduce some code duplication and also make sure the src_dev member gets set. In a high-availability configuration the netdevice pointer can be used by the RDMA CM to align RDMA sessions to use the same links as the IP stack does under fail-over and route change cases. Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- drivers/infiniband/core/addr.c | 1 + drivers/infiniband/core/cma.c | 8 +++++--- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c index e4eb8be3bb0..09a2bec7fd3 100644 --- a/drivers/infiniband/core/addr.c +++ b/drivers/infiniband/core/addr.c @@ -105,6 +105,7 @@ int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev, memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN); if (dst_dev_addr) memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN); + dev_addr->src_dev = dev; return 0; } EXPORT_SYMBOL(rdma_copy_addr); diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 44d190f6781..5fb506a4177 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -1002,6 +1002,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id, union cma_ip_addr *src, *dst; __be16 port; u8 ip_ver; + int ret; if (cma_get_net_info(ib_event->private_data, listen_id->ps, &ip_ver, &port, &src, &dst)) @@ -1026,10 +1027,11 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id, if (rt->num_paths == 2) rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path; - ib_addr_set_sgid(&rt->addr.dev_addr, &rt->path_rec[0].sgid); ib_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid); - ib_addr_set_pkey(&rt->addr.dev_addr, be16_to_cpu(rt->path_rec[0].pkey)); - rt->addr.dev_addr.dev_type = RDMA_NODE_IB_CA; + ret = rdma_translate_ip(&id->route.addr.src_addr, + &id->route.addr.dev_addr); + if (ret) + goto destroy_id; id_priv = container_of(id, struct rdma_id_private, id); id_priv->state = CMA_CONNECT; -- cgit v1.2.3 From de910bd92137005b5e1ecaf2ce68053d7d7d5350 Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Mon, 14 Jul 2008 23:48:53 -0700 Subject: RDMA/cma: Simplify locking needed for serialization of callbacks The RDMA CM has some logic in place to make sure that callbacks on a given CM ID are delivered to the consumer in a serialized manner. Specifically it has code to protect against a device removal racing with a running callback function. This patch simplifies this logic by using a mutex per ID instead of a wait queue and atomic variable. This means that cma_disable_remove() now is more properly named to cma_disable_callback(), and cma_enable_remove() can now be removed because it just would become a trivial wrapper around mutex_unlock(). Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- drivers/infiniband/core/cma.c | 106 ++++++++++++++++++++---------------------- 1 file changed, 50 insertions(+), 56 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 5fb506a4177..ae11d5cc74d 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -130,8 +130,7 @@ struct rdma_id_private { struct completion comp; atomic_t refcount; - wait_queue_head_t wait_remove; - atomic_t dev_remove; + struct mutex handler_mutex; int backlog; int timeout_ms; @@ -355,26 +354,15 @@ static void cma_deref_id(struct rdma_id_private *id_priv) complete(&id_priv->comp); } -static int cma_disable_remove(struct rdma_id_private *id_priv, +static int cma_disable_callback(struct rdma_id_private *id_priv, enum cma_state state) { - unsigned long flags; - int ret; - - spin_lock_irqsave(&id_priv->lock, flags); - if (id_priv->state == state) { - atomic_inc(&id_priv->dev_remove); - ret = 0; - } else - ret = -EINVAL; - spin_unlock_irqrestore(&id_priv->lock, flags); - return ret; -} - -static void cma_enable_remove(struct rdma_id_private *id_priv) -{ - if (atomic_dec_and_test(&id_priv->dev_remove)) - wake_up(&id_priv->wait_remove); + mutex_lock(&id_priv->handler_mutex); + if (id_priv->state != state) { + mutex_unlock(&id_priv->handler_mutex); + return -EINVAL; + } + return 0; } static int cma_has_cm_dev(struct rdma_id_private *id_priv) @@ -399,8 +387,7 @@ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, mutex_init(&id_priv->qp_mutex); init_completion(&id_priv->comp); atomic_set(&id_priv->refcount, 1); - init_waitqueue_head(&id_priv->wait_remove); - atomic_set(&id_priv->dev_remove, 0); + mutex_init(&id_priv->handler_mutex); INIT_LIST_HEAD(&id_priv->listen_list); INIT_LIST_HEAD(&id_priv->mc_list); get_random_bytes(&id_priv->seq_num, sizeof id_priv->seq_num); @@ -927,7 +914,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) struct rdma_cm_event event; int ret = 0; - if (cma_disable_remove(id_priv, CMA_CONNECT)) + if (cma_disable_callback(id_priv, CMA_CONNECT)) return 0; memset(&event, 0, sizeof event); @@ -984,12 +971,12 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.ib = NULL; cma_exch(id_priv, CMA_DESTROYING); - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; } out: - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); return ret; } @@ -1101,7 +1088,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) int offset, ret; listen_id = cm_id->context; - if (cma_disable_remove(listen_id, CMA_LISTEN)) + if (cma_disable_callback(listen_id, CMA_LISTEN)) return -ECONNABORTED; memset(&event, 0, sizeof event); @@ -1122,7 +1109,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) goto out; } - atomic_inc(&conn_id->dev_remove); + mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING); mutex_lock(&lock); ret = cma_acquire_dev(conn_id); mutex_unlock(&lock); @@ -1144,7 +1131,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) !cma_is_ud_ps(conn_id->id.ps)) ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0); mutex_unlock(&lock); - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); goto out; } @@ -1153,11 +1140,11 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) release_conn_id: cma_exch(conn_id, CMA_DESTROYING); - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(&conn_id->id); out: - cma_enable_remove(listen_id); + mutex_unlock(&listen_id->handler_mutex); return ret; } @@ -1223,7 +1210,7 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event) struct sockaddr_in *sin; int ret = 0; - if (cma_disable_remove(id_priv, CMA_CONNECT)) + if (cma_disable_callback(id_priv, CMA_CONNECT)) return 0; memset(&event, 0, sizeof event); @@ -1267,12 +1254,12 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event) /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.iw = NULL; cma_exch(id_priv, CMA_DESTROYING); - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; } - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); return ret; } @@ -1288,7 +1275,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, struct ib_device_attr attr; listen_id = cm_id->context; - if (cma_disable_remove(listen_id, CMA_LISTEN)) + if (cma_disable_callback(listen_id, CMA_LISTEN)) return -ECONNABORTED; /* Create a new RDMA id for the new IW CM ID */ @@ -1300,19 +1287,19 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, goto out; } conn_id = container_of(new_cm_id, struct rdma_id_private, id); - atomic_inc(&conn_id->dev_remove); + mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING); conn_id->state = CMA_CONNECT; dev = ip_dev_find(&init_net, iw_event->local_addr.sin_addr.s_addr); if (!dev) { ret = -EADDRNOTAVAIL; - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(new_cm_id); goto out; } ret = rdma_copy_addr(&conn_id->id.route.addr.dev_addr, dev, NULL); if (ret) { - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(new_cm_id); goto out; } @@ -1321,7 +1308,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, ret = cma_acquire_dev(conn_id); mutex_unlock(&lock); if (ret) { - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(new_cm_id); goto out; } @@ -1337,7 +1324,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, ret = ib_query_device(conn_id->id.device, &attr); if (ret) { - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(new_cm_id); goto out; } @@ -1353,14 +1340,17 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, /* User wants to destroy the CM ID */ conn_id->cm_id.iw = NULL; cma_exch(conn_id, CMA_DESTROYING); - cma_enable_remove(conn_id); + mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(&conn_id->id); + goto out; } + mutex_unlock(&conn_id->handler_mutex); + out: if (dev) dev_put(dev); - cma_enable_remove(listen_id); + mutex_unlock(&listen_id->handler_mutex); return ret; } @@ -1592,7 +1582,7 @@ static void cma_work_handler(struct work_struct *_work) struct rdma_id_private *id_priv = work->id; int destroy = 0; - atomic_inc(&id_priv->dev_remove); + mutex_lock(&id_priv->handler_mutex); if (!cma_comp_exch(id_priv, work->old_state, work->new_state)) goto out; @@ -1601,7 +1591,7 @@ static void cma_work_handler(struct work_struct *_work) destroy = 1; } out: - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); cma_deref_id(id_priv); if (destroy) rdma_destroy_id(&id_priv->id); @@ -1764,7 +1754,7 @@ static void addr_handler(int status, struct sockaddr *src_addr, struct rdma_cm_event event; memset(&event, 0, sizeof event); - atomic_inc(&id_priv->dev_remove); + mutex_lock(&id_priv->handler_mutex); /* * Grab mutex to block rdma_destroy_id() from removing the device while @@ -1793,13 +1783,13 @@ static void addr_handler(int status, struct sockaddr *src_addr, if (id_priv->id.event_handler(&id_priv->id, &event)) { cma_exch(id_priv, CMA_DESTROYING); - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); cma_deref_id(id_priv); rdma_destroy_id(&id_priv->id); return; } out: - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); cma_deref_id(id_priv); } @@ -2126,7 +2116,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, struct ib_cm_sidr_rep_event_param *rep = &ib_event->param.sidr_rep_rcvd; int ret = 0; - if (cma_disable_remove(id_priv, CMA_CONNECT)) + if (cma_disable_callback(id_priv, CMA_CONNECT)) return 0; memset(&event, 0, sizeof event); @@ -2167,12 +2157,12 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.ib = NULL; cma_exch(id_priv, CMA_DESTROYING); - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; } out: - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); return ret; } @@ -2570,8 +2560,8 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast) int ret; id_priv = mc->id_priv; - if (cma_disable_remove(id_priv, CMA_ADDR_BOUND) && - cma_disable_remove(id_priv, CMA_ADDR_RESOLVED)) + if (cma_disable_callback(id_priv, CMA_ADDR_BOUND) && + cma_disable_callback(id_priv, CMA_ADDR_RESOLVED)) return 0; mutex_lock(&id_priv->qp_mutex); @@ -2596,12 +2586,12 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast) ret = id_priv->id.event_handler(&id_priv->id, &event); if (ret) { cma_exch(id_priv, CMA_DESTROYING); - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return 0; } - cma_enable_remove(id_priv); + mutex_unlock(&id_priv->handler_mutex); return 0; } @@ -2760,6 +2750,7 @@ static int cma_remove_id_dev(struct rdma_id_private *id_priv) { struct rdma_cm_event event; enum cma_state state; + int ret = 0; /* Record that we want to remove the device */ state = cma_exch(id_priv, CMA_DEVICE_REMOVAL); @@ -2767,15 +2758,18 @@ static int cma_remove_id_dev(struct rdma_id_private *id_priv) return 0; cma_cancel_operation(id_priv, state); - wait_event(id_priv->wait_remove, !atomic_read(&id_priv->dev_remove)); + mutex_lock(&id_priv->handler_mutex); /* Check for destruction from another callback. */ if (!cma_comp(id_priv, CMA_DEVICE_REMOVAL)) - return 0; + goto out; memset(&event, 0, sizeof event); event.event = RDMA_CM_EVENT_DEVICE_REMOVAL; - return id_priv->id.event_handler(&id_priv->id, &event); + ret = id_priv->id.event_handler(&id_priv->id, &event); +out: + mutex_unlock(&id_priv->handler_mutex); + return ret; } static void cma_process_remove(struct cma_device *cma_dev) -- cgit v1.2.3 From f507d28bff0601f1a8a96b7939fa3855c50d25b6 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Mon, 14 Jul 2008 23:48:53 -0700 Subject: IB/mlx4: Use kzalloc() for new QPs so flags are initialized to 0 Current code uses kmalloc() and then just does a bitwise OR operation on qp->flags in create_qp_common(), which means that qp->flags may potentially have some unintended bits set. This patch uses kzalloc() and avoids further explicit clearing of structure members, which also shrinks the code: add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-65 (-65) function old new delta create_qp_common 2024 1959 -65 Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/qp.c | 15 ++------------- 1 file changed, 2 insertions(+), 13 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 91590e7fba0..89eb6cbe592 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -454,19 +454,8 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, spin_lock_init(&qp->rq.lock); qp->state = IB_QPS_RESET; - qp->atomic_rd_en = 0; - qp->resp_depth = 0; - - qp->rq.head = 0; - qp->rq.tail = 0; - qp->sq.head = 0; - qp->sq.tail = 0; - qp->sq_next_wqe = 0; - if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE); - else - qp->sq_signal_bits = 0; err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, !!init_attr->srq, qp); if (err) @@ -704,7 +693,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd, case IB_QPT_UC: case IB_QPT_UD: { - qp = kmalloc(sizeof *qp, GFP_KERNEL); + qp = kzalloc(sizeof *qp, GFP_KERNEL); if (!qp) return ERR_PTR(-ENOMEM); @@ -725,7 +714,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd, if (pd->uobject) return ERR_PTR(-EINVAL); - sqp = kmalloc(sizeof *sqp, GFP_KERNEL); + sqp = kzalloc(sizeof *sqp, GFP_KERNEL); if (!sqp) return ERR_PTR(-ENOMEM); -- cgit v1.2.3