From 20509f1bc553ed7fafa88fa8d01c6212d1876d9f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Aug 2005 16:25:34 -0700 Subject: NFS: Drop inode after rename When doing a rename on top of an existing file that is not in use, the inode of the overwritten file will remain in the icache. The fix is to decrement i_nlink of the overwritten inode, like we do for unlink, rmdir etc already. Problem diagnosed by Olaf Kirch. This patch is a slight variation on his fix. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 2df639f143e..94a7fcee062 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1539,7 +1539,8 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, #endif goto out; } - } + } else + new_inode->i_nlink--; go_ahead: /* -- cgit v1.2.3 From 449231d6ddf50ca46b7fb2f76ecf790135222913 Mon Sep 17 00:00:00 2001 From: Olaf Kirch Date: Thu, 25 Aug 2005 16:25:35 -0700 Subject: From: Olaf Kirch [PATCH] Fix miscompare in __posix_lock_file If an application requests the same lock twice, the kernel should just leave the existing lock in place. Currently, it will install a second lock of the same type. Signed-off-by: Olaf Kirch Signed-off-by: Trond Myklebust --- fs/locks.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/locks.c b/fs/locks.c index f7daa5f4894..7eb1d77b920 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -829,12 +829,16 @@ static int __posix_lock_file(struct inode *inode, struct file_lock *request) /* Detect adjacent or overlapping regions (if same lock type) */ if (request->fl_type == fl->fl_type) { + /* In all comparisons of start vs end, use + * "start - 1" rather than "end + 1". If end + * is OFFSET_MAX, end + 1 will become negative. + */ if (fl->fl_end < request->fl_start - 1) goto next_lock; /* If the next lock in the list has entirely bigger * addresses than the new one, insert the lock here. */ - if (fl->fl_start > request->fl_end + 1) + if (fl->fl_start - 1 > request->fl_end) break; /* If we come here, the new and old lock are of the -- cgit v1.2.3 From 9aa48b7e270d13c8781414dce081a42cae20a80d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Aug 2005 16:25:35 -0700 Subject: NFS: Don't expose internal READDIR errors to userspace Fixes a condition whereby the kernel is returning the non-POSIX error EBADCOOKIE to userspace. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 94a7fcee062..c70eabd6d17 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -565,8 +565,6 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) } } unlock_kernel(); - if (desc->error < 0) - return desc->error; if (res < 0) return res; return 0; -- cgit v1.2.3 From eab5c084b858fd95a873fc2b97de9a9ad937b4ed Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 11 Aug 2005 16:25:14 -0400 Subject: [PATCH] NFS: use a constant value for TCP retransmit timeouts Implement a best practice: don't use exponential backoff when computing retransmit timeout values on TCP connections, but simply retransmit at regular intervals. This also fixes a bug introduced when xprt_reset_majortimeo() was added. Test-plan: Enable RPC debugging and watch timeout behavior on a NFS/TCP mount. Version: Thu, 11 Aug 2005 16:02:19 -0400 Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 73 ++++++++++++++++++++++++++++------------------------------ 1 file changed, 35 insertions(+), 38 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 6922469d6fc..b6a1ca508e6 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -358,6 +358,35 @@ out_no_root: return no_root_error; } +static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans) +{ + to->to_initval = timeo * HZ / 10; + to->to_retries = retrans; + if (!to->to_retries) + to->to_retries = 2; + + switch (proto) { + case IPPROTO_TCP: + if (!to->to_initval) + to->to_initval = 60 * HZ; + if (to->to_initval > RPC_MAX_TCP_TIMEOUT) + to->to_initval = RPC_MAX_TCP_TIMEOUT; + to->to_increment = to->to_initval; + to->to_maxval = to->to_initval + (to->to_increment * to->to_retries); + to->to_exponential = 0; + break; + case IPPROTO_UDP: + default: + if (!to->to_initval) + to->to_initval = 11 * HZ / 10; + if (to->to_initval > RPC_MAX_UDP_TIMEOUT) + to->to_initval = RPC_MAX_UDP_TIMEOUT; + to->to_maxval = RPC_MAX_UDP_TIMEOUT; + to->to_exponential = 1; + break; + } +} + /* * Create an RPC client handle. */ @@ -367,22 +396,12 @@ nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data) struct rpc_timeout timeparms; struct rpc_xprt *xprt = NULL; struct rpc_clnt *clnt = NULL; - int tcp = (data->flags & NFS_MOUNT_TCP); - - /* Initialize timeout values */ - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT; - timeparms.to_exponential = 1; + int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP; - if (!timeparms.to_initval) - timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10; - if (!timeparms.to_retries) - timeparms.to_retries = 5; + nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans); /* create transport and client */ - xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP, - &server->addr, &timeparms); + xprt = xprt_create_proto(proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { dprintk("%s: cannot create RPC transport. Error = %ld\n", __FUNCTION__, PTR_ERR(xprt)); @@ -1674,7 +1693,7 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, struct rpc_clnt *clnt = NULL; struct rpc_timeout timeparms; rpc_authflavor_t authflavour; - int proto, err = -EIO; + int err = -EIO; sb->s_blocksize_bits = 0; sb->s_blocksize = 0; @@ -1692,30 +1711,8 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, server->acdirmax = data->acdirmax*HZ; server->rpc_ops = &nfs_v4_clientops; - /* Initialize timeout values */ - - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_exponential = 1; - if (!timeparms.to_retries) - timeparms.to_retries = 5; - proto = data->proto; - /* Which IP protocol do we use? */ - switch (proto) { - case IPPROTO_TCP: - timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 600 * HZ / 10; - break; - case IPPROTO_UDP: - timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 11 * HZ / 10; - break; - default: - return -EINVAL; - } + nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans); clp = nfs4_get_client(&server->addr.sin_addr); if (!clp) { @@ -1740,7 +1737,7 @@ static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, down_write(&clp->cl_sem); if (IS_ERR(clp->cl_rpcclient)) { - xprt = xprt_create_proto(proto, &server->addr, &timeparms); + xprt = xprt_create_proto(data->proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { up_write(&clp->cl_sem); err = PTR_ERR(xprt); -- cgit v1.2.3 From 43118c29dea2b23798bd42a147015cceee7fa885 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 25 Aug 2005 16:25:49 -0700 Subject: [PATCH] RPC: get rid of xprt->stream Now we can fix up the last few places that use the "xprt->stream" variable, and get rid of it from the rpc_xprt structure. Test-plan: Destructive testing (unplugging the network temporarily). Connectathon with UDP and TCP. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/lockd/host.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/lockd/host.c b/fs/lockd/host.c index 82c77df81c5..7901f5b8092 100644 --- a/fs/lockd/host.c +++ b/fs/lockd/host.c @@ -173,11 +173,10 @@ nlm_bind_host(struct nlm_host *host) /* If we've already created an RPC client, check whether * RPC rebind is required - * Note: why keep rebinding if we're on a tcp connection? */ if ((clnt = host->h_rpcclnt) != NULL) { xprt = clnt->cl_xprt; - if (!xprt->stream && time_after_eq(jiffies, host->h_nextrebind)) { + if (time_after_eq(jiffies, host->h_nextrebind)) { clnt->cl_port = 0; host->h_nextrebind = jiffies + NLM_HOST_REBIND; dprintk("lockd: next rebind in %ld jiffies\n", -- cgit v1.2.3 From ed63c003701a314c4893c11eceb9d68f8f46c662 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 25 Aug 2005 16:25:53 -0700 Subject: [PATCH] RPC: remove xprt->nocong Get rid of the "xprt->nocong" variable. Test-plan: Use WAN simulation to cause sporadic bursty packet loss with UDP mounts. Look for significant regression in performance or client stability. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/lockd/host.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs') diff --git a/fs/lockd/host.c b/fs/lockd/host.c index 7901f5b8092..c4c8601096e 100644 --- a/fs/lockd/host.c +++ b/fs/lockd/host.c @@ -188,7 +188,6 @@ nlm_bind_host(struct nlm_host *host) goto forgetit; xprt_set_timeout(&xprt->timeout, 5, nlmsvc_timeout); - xprt->nocong = 1; /* No congestion control for NLM */ xprt->resvport = 1; /* NLM requires a reserved port */ /* Existing NLM servers accept AUTH_UNIX only */ -- cgit v1.2.3 From 03bf4b707eee06706c9db343dd5c905b7ee47ed2 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Thu, 25 Aug 2005 16:25:55 -0700 Subject: [PATCH] RPC: parametrize various transport connect timeouts Each transport implementation can now set unique bind, connect, reestablishment, and idle timeout values. These are variables, allowing the values to be modified dynamically. This permits exponential backoff of any of these values, for instance. As an example, we implement exponential backoff for the connection reestablishment timeout. Test-plan: Destructive testing (unplugging the network temporarily). Connectathon with UDP and TCP. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index b6a1ca508e6..062911e7ceb 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -369,8 +369,8 @@ static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned case IPPROTO_TCP: if (!to->to_initval) to->to_initval = 60 * HZ; - if (to->to_initval > RPC_MAX_TCP_TIMEOUT) - to->to_initval = RPC_MAX_TCP_TIMEOUT; + if (to->to_initval > NFS_MAX_TCP_TIMEOUT) + to->to_initval = NFS_MAX_TCP_TIMEOUT; to->to_increment = to->to_initval; to->to_maxval = to->to_initval + (to->to_increment * to->to_retries); to->to_exponential = 0; @@ -379,9 +379,9 @@ static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned default: if (!to->to_initval) to->to_initval = 11 * HZ / 10; - if (to->to_initval > RPC_MAX_UDP_TIMEOUT) - to->to_initval = RPC_MAX_UDP_TIMEOUT; - to->to_maxval = RPC_MAX_UDP_TIMEOUT; + if (to->to_initval > NFS_MAX_UDP_TIMEOUT) + to->to_initval = NFS_MAX_UDP_TIMEOUT; + to->to_maxval = NFS_MAX_UDP_TIMEOUT; to->to_exponential = 1; break; } -- cgit v1.2.3 From 278c995c8a153bb2a9bc427e931cfb9c8034c9d7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sun, 24 Jul 2005 23:53:01 +0100 Subject: [PATCH] RPC,NFS: new rpc_pipefs patch Currently rpc_mkdir/rpc_rmdir and rpc_mkpipe/mk_unlink have an API that's a little unfortunate. They take a path relative to the rpc_pipefs root and thus need to perform a full lookup. If you look at debugfs or usbfs they always store the dentry for directories they created and thus can pass in a dentry + single pathname component pair into their equivalents of the above functions. And in fact rpc_pipefs actually stores a dentry for all but one component so this change not only simplifies the core rpc_pipe code but also the callers. Unfortuntately this code path is only used by the NFS4 idmapper and AUTH_GSSAPI for which I don't have a test enviroment. Could someone give it a spin? It's the last bit needed before we can rework the lookup_hash API Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/idmap.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c index ffb8df91dc3..1d0a5bf0d26 100644 --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -66,7 +66,6 @@ struct idmap_hashtable { }; struct idmap { - char idmap_path[48]; struct dentry *idmap_dentry; wait_queue_head_t idmap_wq; struct idmap_msg idmap_im; @@ -102,11 +101,8 @@ nfs_idmap_new(struct nfs4_client *clp) memset(idmap, 0, sizeof(*idmap)); - snprintf(idmap->idmap_path, sizeof(idmap->idmap_path), - "%s/idmap", clp->cl_rpcclient->cl_pathname); - - idmap->idmap_dentry = rpc_mkpipe(idmap->idmap_path, - idmap, &idmap_upcall_ops, 0); + idmap->idmap_dentry = rpc_mkpipe(clp->cl_rpcclient->cl_dentry, + "idmap", idmap, &idmap_upcall_ops, 0); if (IS_ERR(idmap->idmap_dentry)) { kfree(idmap); return; @@ -128,7 +124,7 @@ nfs_idmap_delete(struct nfs4_client *clp) if (!idmap) return; - rpc_unlink(idmap->idmap_path); + rpc_unlink(idmap->idmap_dentry); clp->cl_idmap = NULL; kfree(idmap); } -- cgit v1.2.3 From 3063d8a16643190f9e12e9c7e9f1ca56f7e7934e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 25 Aug 2005 16:25:57 -0700 Subject: NFS: Make /proc/mounts display the protocol used by NFSv4 Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 062911e7ceb..358d8ef8c08 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -595,7 +595,6 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt) { NFS_MOUNT_SOFT, ",soft", ",hard" }, { NFS_MOUNT_INTR, ",intr", "" }, { NFS_MOUNT_POSIX, ",posix", "" }, - { NFS_MOUNT_TCP, ",tcp", ",udp" }, { NFS_MOUNT_NOCTO, ",nocto", "" }, { NFS_MOUNT_NOAC, ",noac", "" }, { NFS_MOUNT_NONLM, ",nolock", ",lock" }, @@ -604,6 +603,8 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt) }; struct proc_nfs_info *nfs_infop; struct nfs_server *nfss = NFS_SB(mnt->mnt_sb); + char buf[12]; + char *proto; seq_printf(m, ",v%d", nfss->rpc_ops->version); seq_printf(m, ",rsize=%d", nfss->rsize); @@ -622,6 +623,18 @@ static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt) else seq_puts(m, nfs_infop->nostr); } + switch (nfss->client->cl_xprt->prot) { + case IPPROTO_TCP: + proto = "tcp"; + break; + case IPPROTO_UDP: + proto = "udp"; + break; + default: + snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot); + proto = buf; + } + seq_printf(m, ",proto=%s", proto); seq_puts(m, ",addr="); seq_escape(m, nfss->hostname, " \t\n\\"); return 0; -- cgit v1.2.3 From f134585a7343d71f9be7f0cf97e2145f21dd10c6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 23 Sep 2005 11:08:25 -0400 Subject: Revert "[PATCH] RPC,NFS: new rpc_pipefs patch" This reverts 17f4e6febca160a9f9dd4bdece9784577a2f4524 commit. --- fs/nfs/idmap.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c index 1d0a5bf0d26..ffb8df91dc3 100644 --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -66,6 +66,7 @@ struct idmap_hashtable { }; struct idmap { + char idmap_path[48]; struct dentry *idmap_dentry; wait_queue_head_t idmap_wq; struct idmap_msg idmap_im; @@ -101,8 +102,11 @@ nfs_idmap_new(struct nfs4_client *clp) memset(idmap, 0, sizeof(*idmap)); - idmap->idmap_dentry = rpc_mkpipe(clp->cl_rpcclient->cl_dentry, - "idmap", idmap, &idmap_upcall_ops, 0); + snprintf(idmap->idmap_path, sizeof(idmap->idmap_path), + "%s/idmap", clp->cl_rpcclient->cl_pathname); + + idmap->idmap_dentry = rpc_mkpipe(idmap->idmap_path, + idmap, &idmap_upcall_ops, 0); if (IS_ERR(idmap->idmap_dentry)) { kfree(idmap); return; @@ -124,7 +128,7 @@ nfs_idmap_delete(struct nfs4_client *clp) if (!idmap) return; - rpc_unlink(idmap->idmap_dentry); + rpc_unlink(idmap->idmap_path); clp->cl_idmap = NULL; kfree(idmap); } -- cgit v1.2.3 From ac17b8b57013a3e38d1958f66a218f15659e5752 Mon Sep 17 00:00:00 2001 From: Dave Kleikamp Date: Mon, 3 Oct 2005 15:32:11 -0500 Subject: JFS: make special inodes play nicely with page balancing This patch fixes up a few problems with jfs's reserved inodes. 1. There is no need for the jfs code setting the I_DIRTY bits in i_state. I am ashamed that the code ever did this, and surprised it hasn't been noticed until now. 2. Make sure special inodes are on an inode hash list. If the inodes are unhashed, __mark_inode_dirty will fail to put the inode on the superblock's dirty list, and the data will not be flushed under memory pressure. 3. Force writing journal data to disk when metapage_writepage is unable to write a metadata page due to pending journal I/O. Signed-off-by: Dave Kleikamp --- fs/jfs/jfs_dmap.c | 1 - fs/jfs/jfs_imap.c | 10 ++++++++-- fs/jfs/jfs_metapage.c | 6 ++++++ fs/jfs/jfs_txnmgr.c | 2 -- fs/jfs/super.c | 1 + 5 files changed, 15 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/jfs/jfs_dmap.c b/fs/jfs/jfs_dmap.c index eadf319bee2..51c02bf0787 100644 --- a/fs/jfs/jfs_dmap.c +++ b/fs/jfs/jfs_dmap.c @@ -305,7 +305,6 @@ int dbSync(struct inode *ipbmap) filemap_fdatawrite(ipbmap->i_mapping); filemap_fdatawait(ipbmap->i_mapping); - ipbmap->i_state |= I_DIRTY; diWriteSpecial(ipbmap, 0); return (0); diff --git a/fs/jfs/jfs_imap.c b/fs/jfs/jfs_imap.c index 4021d46da7e..28201b194f5 100644 --- a/fs/jfs/jfs_imap.c +++ b/fs/jfs/jfs_imap.c @@ -56,6 +56,12 @@ #include "jfs_superblock.h" #include "jfs_debug.h" +/* + * __mark_inode_dirty expects inodes to be hashed. Since we don't want + * special inodes in the fileset inode space, we hash them to a dummy head + */ +static HLIST_HEAD(aggregate_hash); + /* * imap locks */ @@ -491,6 +497,8 @@ struct inode *diReadSpecial(struct super_block *sb, ino_t inum, int secondary) /* release the page */ release_metapage(mp); + hlist_add_head(&ip->i_hash, &aggregate_hash); + return (ip); } @@ -514,8 +522,6 @@ void diWriteSpecial(struct inode *ip, int secondary) ino_t inum = ip->i_ino; struct metapage *mp; - ip->i_state &= ~I_DIRTY; - if (secondary) address = addressPXD(&sbi->ait2) >> sbi->l2nbperpage; else diff --git a/fs/jfs/jfs_metapage.c b/fs/jfs/jfs_metapage.c index 13d7e3f1feb..c81c6438fce 100644 --- a/fs/jfs/jfs_metapage.c +++ b/fs/jfs/jfs_metapage.c @@ -395,6 +395,12 @@ static int metapage_writepage(struct page *page, struct writeback_control *wbc) if (mp->nohomeok && !test_bit(META_forcewrite, &mp->flag)) { redirty = 1; + /* + * Make sure this page isn't blocked indefinitely. + * If the journal isn't undergoing I/O, push it + */ + if (mp->log && !(mp->log->cflag & logGC_PAGEOUT)) + jfs_flush_journal(mp->log, 0); continue; } diff --git a/fs/jfs/jfs_txnmgr.c b/fs/jfs/jfs_txnmgr.c index 9b71ed2674f..b660c93c92d 100644 --- a/fs/jfs/jfs_txnmgr.c +++ b/fs/jfs/jfs_txnmgr.c @@ -2396,7 +2396,6 @@ static void txUpdateMap(struct tblock * tblk) */ if (tblk->xflag & COMMIT_CREATE) { diUpdatePMap(ipimap, tblk->ino, FALSE, tblk); - ipimap->i_state |= I_DIRTY; /* update persistent block allocation map * for the allocation of inode extent; */ @@ -2407,7 +2406,6 @@ static void txUpdateMap(struct tblock * tblk) } else if (tblk->xflag & COMMIT_DELETE) { ip = tblk->u.ip; diUpdatePMap(ipimap, ip->i_ino, TRUE, tblk); - ipimap->i_state |= I_DIRTY; iput(ip); } } diff --git a/fs/jfs/super.c b/fs/jfs/super.c index 71bc34b96b2..4226af3ea91 100644 --- a/fs/jfs/super.c +++ b/fs/jfs/super.c @@ -442,6 +442,7 @@ static int jfs_fill_super(struct super_block *sb, void *data, int silent) inode->i_nlink = 1; inode->i_size = sb->s_bdev->bd_inode->i_size; inode->i_mapping->a_ops = &jfs_metapage_aops; + insert_inode_hash(inode); mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); sbi->direct_inode = inode; -- cgit v1.2.3 From b6a47fd8ff08a9d5cd279cdb8d97a619983575fa Mon Sep 17 00:00:00 2001 From: Dave Kleikamp Date: Tue, 11 Oct 2005 09:06:59 -0500 Subject: JFS: Corrupted block map should not cause trap Replace assert statements with better error handling. Signed-off-by: Dave Kleikamp --- fs/jfs/jfs_dmap.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/jfs/jfs_dmap.c b/fs/jfs/jfs_dmap.c index 51c02bf0787..68000a50ceb 100644 --- a/fs/jfs/jfs_dmap.c +++ b/fs/jfs/jfs_dmap.c @@ -74,7 +74,7 @@ static void dbAllocBits(struct bmap * bmp, struct dmap * dp, s64 blkno, int nblocks); static void dbSplit(dmtree_t * tp, int leafno, int splitsz, int newval); -static void dbBackSplit(dmtree_t * tp, int leafno); +static int dbBackSplit(dmtree_t * tp, int leafno); static int dbJoin(dmtree_t * tp, int leafno, int newval); static void dbAdjTree(dmtree_t * tp, int leafno, int newval); static int dbAdjCtl(struct bmap * bmp, s64 blkno, int newval, int alloc, @@ -2466,7 +2466,9 @@ dbAdjCtl(struct bmap * bmp, s64 blkno, int newval, int alloc, int level) * that it is at the front of a binary buddy system. */ if (oldval == NOFREE) { - dbBackSplit((dmtree_t *) dcp, leafno); + rc = dbBackSplit((dmtree_t *) dcp, leafno); + if (rc) + return rc; oldval = dcp->stree[ti]; } dbSplit((dmtree_t *) dcp, leafno, dcp->budmin, newval); @@ -2626,7 +2628,7 @@ static void dbSplit(dmtree_t * tp, int leafno, int splitsz, int newval) * * serialization: IREAD_LOCK(ipbmap) or IWRITE_LOCK(ipbmap) held on entry/exit; */ -static void dbBackSplit(dmtree_t * tp, int leafno) +static int dbBackSplit(dmtree_t * tp, int leafno) { int budsz, bud, w, bsz, size; int cursz; @@ -2661,7 +2663,10 @@ static void dbBackSplit(dmtree_t * tp, int leafno) */ for (w = leafno, bsz = budsz;; bsz <<= 1, w = (w < bud) ? w : bud) { - assert(bsz < le32_to_cpu(tp->dmt_nleafs)); + if (bsz >= le32_to_cpu(tp->dmt_nleafs)) { + jfs_err("JFS: block map error in dbBackSplit"); + return -EIO; + } /* determine the buddy. */ @@ -2680,7 +2685,11 @@ static void dbBackSplit(dmtree_t * tp, int leafno) } } - assert(leaf[leafno] == size); + if (leaf[leafno] != size) { + jfs_err("JFS: wrong leaf value in dbBackSplit"); + return -EIO; + } + return 0; } -- cgit v1.2.3 From cee54fc944422c44e476736c045a9e8053cb0644 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:12 -0700 Subject: NFSv4: Add functions to order RPC calls NFSv4 file state-changing functions such as OPEN, CLOSE, LOCK,... are all labelled with "sequence identifiers" in order to prevent the server from reordering RPC requests, as this could cause its file state to become out of sync with the client. Currently the NFS client code enforces this ordering locally using semaphores to restrict access to structures until the RPC call is done. This, of course, only works with synchronous RPC calls, since the user process must first grab the semaphore. By dropping semaphores, and instead teaching the RPC engine to hold the RPC calls until they are ready to be sent, we can extend this process to work nicely with asynchronous RPC calls too. This patch adds a new list called "rpc_sequence" that defines the order of the RPC calls to be sent. We add one such list for each state_owner. When an RPC call is ready to be sent, it checks if it is top of the rpc_sequence list. If so, it proceeds. If not, it goes back to sleep, and loops until it hits top of the list. Once the RPC call has completed, it can then bump the sequence id counter, and remove itself from the rpc_sequence list, and then wake up the next sleeper. Note that the state_owner sequence ids and lock_owner sequence ids are all indexed to the same rpc_sequence list, so OPEN, LOCK,... requests are all ordered w.r.t. each other. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 42 ++++++++++++++-- fs/nfs/nfs4proc.c | 111 ++++++++++++++++++++++++++-------------- fs/nfs/nfs4state.c | 145 +++++++++++++++++++++++++++++++++++++++++++---------- fs/nfs/nfs4xdr.c | 43 +++++++++++++--- 4 files changed, 265 insertions(+), 76 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index ec1a22d7b87..6ac6708484f 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -92,6 +92,35 @@ struct nfs4_client { unsigned char cl_id_uniquifier; }; +/* + * struct rpc_sequence ensures that RPC calls are sent in the exact + * order that they appear on the list. + */ +struct rpc_sequence { + struct rpc_wait_queue wait; /* RPC call delay queue */ + spinlock_t lock; /* Protects the list */ + struct list_head list; /* Defines sequence of RPC calls */ +}; + +#define NFS_SEQID_CONFIRMED 1 +struct nfs_seqid_counter { + struct rpc_sequence *sequence; + int flags; + u32 counter; +}; + +struct nfs_seqid { + struct list_head list; + struct nfs_seqid_counter *sequence; + struct rpc_task *task; +}; + +static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status) +{ + if (seqid_mutating_err(-status)) + seqid->flags |= NFS_SEQID_CONFIRMED; +} + /* * NFS4 state_owners and lock_owners are simply labels for ordered * sequences of RPC calls. Their sole purpose is to provide once-only @@ -106,12 +135,13 @@ struct nfs4_state_owner { struct nfs4_client *so_client; u32 so_id; /* 32-bit identifier, unique */ struct semaphore so_sema; - u32 so_seqid; /* protected by so_sema */ atomic_t so_count; struct rpc_cred *so_cred; /* Associated cred */ struct list_head so_states; struct list_head so_delegations; + struct nfs_seqid_counter so_seqid; + struct rpc_sequence so_sequence; }; /* @@ -132,7 +162,7 @@ struct nfs4_lock_state { fl_owner_t ls_owner; /* POSIX lock owner */ #define NFS_LOCK_INITIALIZED 1 int ls_flags; - u32 ls_seqid; + struct nfs_seqid_counter ls_seqid; u32 ls_id; nfs4_stateid ls_stateid; atomic_t ls_count; @@ -224,12 +254,16 @@ extern struct nfs4_state * nfs4_get_open_state(struct inode *, struct nfs4_state extern void nfs4_put_open_state(struct nfs4_state *); extern void nfs4_close_state(struct nfs4_state *, mode_t); extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode); -extern void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp); extern void nfs4_schedule_state_recovery(struct nfs4_client *); extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); -extern void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *ls); extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t); +extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter); +extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task); +extern void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid); +extern void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid); +extern void nfs_free_seqid(struct nfs_seqid *seqid); + extern const nfs4_stateid zero_stateid; /* nfs4xdr.c */ diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9701ca8c942..9ba89e7cdd2 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -218,7 +218,6 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st struct nfs_delegation *delegation = NFS_I(inode)->delegation; struct nfs_openargs o_arg = { .fh = NFS_FH(inode), - .seqid = sp->so_seqid, .id = sp->so_id, .open_flags = state->state, .clientid = server->nfs4_state->cl_clientid, @@ -245,8 +244,13 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st } o_arg.u.delegation_type = delegation->type; } + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + if (o_arg.seqid == NULL) + return -ENOMEM; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + /* Confirm the sequence as being established */ + nfs_confirm_seqid(&sp->so_seqid, status); + nfs_increment_open_seqid(status, o_arg.seqid); if (status == 0) { memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid)); if (o_res.delegation_type != 0) { @@ -256,6 +260,7 @@ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *st nfs_async_inode_return_delegation(inode, &o_res.stateid); } } + nfs_free_seqid(o_arg.seqid); clear_bit(NFS_DELEGATED_STATE, &state->flags); /* Ensure we update the inode attributes */ NFS_CACHEINV(inode); @@ -307,16 +312,20 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state goto out; if (state->state == 0) goto out; - arg.seqid = sp->so_seqid; + arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + status = -ENOMEM; + if (arg.seqid == NULL) + goto out; arg.open_flags = state->state; memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data)); status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + nfs_increment_open_seqid(status, arg.seqid); if (status >= 0) { memcpy(state->stateid.data, res.stateid.data, sizeof(state->stateid.data)); clear_bit(NFS_DELEGATED_STATE, &state->flags); } + nfs_free_seqid(arg.seqid); out: up(&sp->so_sema); dput(parent); @@ -345,11 +354,11 @@ int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) return err; } -static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid) +static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid) { struct nfs_open_confirmargs arg = { .fh = fh, - .seqid = sp->so_seqid, + .seqid = seqid, .stateid = *stateid, }; struct nfs_open_confirmres res; @@ -362,7 +371,9 @@ static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nf int status; status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + /* Confirm the sequence as being established */ + nfs_confirm_seqid(&sp->so_seqid, status); + nfs_increment_open_seqid(status, seqid); if (status >= 0) memcpy(stateid, &res.stateid, sizeof(*stateid)); return status; @@ -380,21 +391,21 @@ static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, stru int status; /* Update sequence id. The caller must serialize! */ - o_arg->seqid = sp->so_seqid; o_arg->id = sp->so_id; o_arg->clientid = sp->so_client->cl_clientid; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + nfs_increment_open_seqid(status, o_arg->seqid); if (status != 0) goto out; update_changeattr(dir, &o_res->cinfo); if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { status = _nfs4_proc_open_confirm(server->client, &o_res->fh, - sp, &o_res->stateid); + sp, &o_res->stateid, o_arg->seqid); if (status != 0) goto out; } + nfs_confirm_seqid(&sp->so_seqid, 0); if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); out: @@ -465,6 +476,10 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st set_bit(NFS_DELEGATED_STATE, &state->flags); goto out; } + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + status = -ENOMEM; + if (o_arg.seqid == NULL) + goto out; status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_nodeleg; @@ -490,6 +505,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res); } out_nodeleg: + nfs_free_seqid(o_arg.seqid); clear_bit(NFS_DELEGATED_STATE, &state->flags); out: dput(parent); @@ -667,6 +683,9 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st /* Serialization for the sequence id */ down(&sp->so_sema); + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + if (o_arg.seqid == NULL) + return -ENOMEM; status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_err; @@ -681,6 +700,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st update_open_stateid(state, &o_res.stateid, flags); if (o_res.delegation_type != 0) nfs_inode_set_delegation(inode, cred, &o_res); + nfs_free_seqid(o_arg.seqid); up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&clp->cl_sem); @@ -690,6 +710,7 @@ out_err: if (sp != NULL) { if (state != NULL) nfs4_put_open_state(state); + nfs_free_seqid(o_arg.seqid); up(&sp->so_sema); nfs4_put_state_owner(sp); } @@ -718,7 +739,7 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, * It is actually a sign of a bug on the client or on the server. * * If we receive a BAD_SEQID error in the particular case of - * doing an OPEN, we assume that nfs4_increment_seqid() will + * doing an OPEN, we assume that nfs_increment_open_seqid() will * have unhashed the old state_owner for us, and that we can * therefore safely retry using a new one. We should still warn * the user though... @@ -799,7 +820,7 @@ static void nfs4_close_done(struct rpc_task *task) /* hmm. we are done with the inode, and in the process of freeing * the state_owner. we keep this around to process errors */ - nfs4_increment_seqid(task->tk_status, sp); + nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid); switch (task->tk_status) { case 0: memcpy(&state->stateid, &calldata->res.stateid, @@ -818,6 +839,7 @@ static void nfs4_close_done(struct rpc_task *task) } state->state = calldata->arg.open_flags; nfs4_put_open_state(state); + nfs_free_seqid(calldata->arg.seqid); up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&server->nfs4_state->cl_sem); @@ -865,7 +887,11 @@ int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) calldata->state = state; calldata->arg.fh = NFS_FH(inode); /* Serialization for the sequence id */ - calldata->arg.seqid = state->owner->so_seqid; + calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid); + if (calldata->arg.seqid == NULL) { + kfree(calldata); + return -ENOMEM; + } calldata->arg.open_flags = mode; memcpy(&calldata->arg.stateid, &state->stateid, sizeof(calldata->arg.stateid)); @@ -2729,15 +2755,19 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock /* We might have lost the locks! */ if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) goto out; - luargs.seqid = lsp->ls_seqid; - memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid)); + luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); + status = -ENOMEM; + if (luargs.seqid == NULL) + goto out; + memcpy(luargs.stateid.data, lsp->ls_stateid.data, sizeof(luargs.stateid.data)); arg.u.locku = &luargs; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_lock_seqid(status, lsp); + nfs_increment_lock_seqid(status, luargs.seqid); if (status == 0) - memcpy(&lsp->ls_stateid, &res.u.stateid, - sizeof(lsp->ls_stateid)); + memcpy(lsp->ls_stateid.data, res.u.stateid.data, + sizeof(lsp->ls_stateid.data)); + nfs_free_seqid(luargs.seqid); out: up(&state->lock_sema); if (status == 0) @@ -2783,9 +2813,13 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r .reclaim = reclaim, .new_lock_owner = 0, }; - int status; + struct nfs_seqid *lock_seqid; + int status = -ENOMEM; - if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) { + lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (lock_seqid == NULL) + return -ENOMEM; + if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) { struct nfs4_state_owner *owner = state->owner; struct nfs_open_to_lock otl = { .lock_owner = { @@ -2793,39 +2827,40 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r }, }; - otl.lock_seqid = lsp->ls_seqid; + otl.lock_seqid = lock_seqid; otl.lock_owner.id = lsp->ls_id; memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid)); largs.u.open_lock = &otl; largs.new_lock_owner = 1; arg.u.lock = &largs; down(&owner->so_sema); - otl.open_seqid = owner->so_seqid; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment open_owner seqid on success, and - * seqid mutating errors */ - nfs4_increment_seqid(status, owner); - up(&owner->so_sema); - if (status == 0) { - lsp->ls_flags |= NFS_LOCK_INITIALIZED; - lsp->ls_seqid++; + otl.open_seqid = nfs_alloc_seqid(&owner->so_seqid); + if (otl.open_seqid != NULL) { + status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); + /* increment seqid on success, and seqid mutating errors */ + nfs_increment_open_seqid(status, otl.open_seqid); + nfs_free_seqid(otl.open_seqid); } + up(&owner->so_sema); + if (status == 0) + nfs_confirm_seqid(&lsp->ls_seqid, 0); } else { - struct nfs_exist_lock el = { - .seqid = lsp->ls_seqid, - }; + struct nfs_exist_lock el; memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid)); largs.u.exist_lock = ⪙ arg.u.lock = &largs; + el.seqid = lock_seqid; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment seqid on success, and * seqid mutating errors*/ - nfs4_increment_lock_seqid(status, lsp); } + /* increment seqid on success, and seqid mutating errors*/ + nfs_increment_lock_seqid(status, lock_seqid); /* save the returned stateid. */ - if (status == 0) - memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid)); - else if (status == -NFS4ERR_DENIED) + if (status == 0) { + memcpy(lsp->ls_stateid.data, res.u.stateid.data, sizeof(lsp->ls_stateid.data)); + lsp->ls_flags |= NFS_LOCK_INITIALIZED; + } else if (status == -NFS4ERR_DENIED) status = -EAGAIN; + nfs_free_seqid(lock_seqid); return status; } diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index afe587d82f1..f535c219cf3 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -264,13 +264,16 @@ nfs4_alloc_state_owner(void) { struct nfs4_state_owner *sp; - sp = kmalloc(sizeof(*sp),GFP_KERNEL); + sp = kzalloc(sizeof(*sp),GFP_KERNEL); if (!sp) return NULL; init_MUTEX(&sp->so_sema); - sp->so_seqid = 0; /* arbitrary */ INIT_LIST_HEAD(&sp->so_states); INIT_LIST_HEAD(&sp->so_delegations); + rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); + sp->so_seqid.sequence = &sp->so_sequence; + spin_lock_init(&sp->so_sequence.lock); + INIT_LIST_HEAD(&sp->so_sequence.list); atomic_set(&sp->so_count, 1); return sp; } @@ -553,12 +556,10 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f struct nfs4_lock_state *lsp; struct nfs4_client *clp = state->owner->so_client; - lsp = kmalloc(sizeof(*lsp), GFP_KERNEL); + lsp = kzalloc(sizeof(*lsp), GFP_KERNEL); if (lsp == NULL) return NULL; - lsp->ls_flags = 0; - lsp->ls_seqid = 0; /* arbitrary */ - memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data)); + lsp->ls_seqid.sequence = &state->owner->so_sequence; atomic_set(&lsp->ls_count, 1); lsp->ls_owner = fl_owner; spin_lock(&clp->cl_lock); @@ -673,29 +674,102 @@ void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t f nfs4_put_lock_state(lsp); } -/* -* Called with state->lock_sema and clp->cl_sem held. -*/ -void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp) +struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) +{ + struct rpc_sequence *sequence = counter->sequence; + struct nfs_seqid *new; + + new = kmalloc(sizeof(*new), GFP_KERNEL); + if (new != NULL) { + new->sequence = counter; + new->task = NULL; + spin_lock(&sequence->lock); + list_add_tail(&new->list, &sequence->list); + spin_unlock(&sequence->lock); + } + return new; +} + +void nfs_free_seqid(struct nfs_seqid *seqid) { - if (status == NFS_OK || seqid_mutating_err(-status)) - lsp->ls_seqid++; + struct rpc_sequence *sequence = seqid->sequence->sequence; + struct rpc_task *next = NULL; + + spin_lock(&sequence->lock); + list_del(&seqid->list); + if (!list_empty(&sequence->list)) { + next = list_entry(sequence->list.next, struct nfs_seqid, list)->task; + if (next) + rpc_wake_up_task(next); + } + spin_unlock(&sequence->lock); + kfree(seqid); } /* -* Called with sp->so_sema and clp->cl_sem held. -* -* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or -* failed with a seqid incrementing error - -* see comments nfs_fs.h:seqid_mutating_error() -*/ -void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp) -{ - if (status == NFS_OK || seqid_mutating_err(-status)) - sp->so_seqid++; - /* If the server returns BAD_SEQID, unhash state_owner here */ - if (status == -NFS4ERR_BAD_SEQID) + * Called with sp->so_sema and clp->cl_sem held. + * + * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or + * failed with a seqid incrementing error - + * see comments nfs_fs.h:seqid_mutating_error() + */ +static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid) +{ + switch (status) { + case 0: + break; + case -NFS4ERR_BAD_SEQID: + case -NFS4ERR_STALE_CLIENTID: + case -NFS4ERR_STALE_STATEID: + case -NFS4ERR_BAD_STATEID: + case -NFS4ERR_BADXDR: + case -NFS4ERR_RESOURCE: + case -NFS4ERR_NOFILEHANDLE: + /* Non-seqid mutating errors */ + return; + }; + /* + * Note: no locking needed as we are guaranteed to be first + * on the sequence list + */ + seqid->sequence->counter++; +} + +void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) +{ + if (status == -NFS4ERR_BAD_SEQID) { + struct nfs4_state_owner *sp = container_of(seqid->sequence, + struct nfs4_state_owner, so_seqid); nfs4_drop_state_owner(sp); + } + return nfs_increment_seqid(status, seqid); +} + +/* + * Called with ls->lock_sema and clp->cl_sem held. + * + * Increment the seqid if the LOCK/LOCKU succeeded, or + * failed with a seqid incrementing error - + * see comments nfs_fs.h:seqid_mutating_error() + */ +void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid) +{ + return nfs_increment_seqid(status, seqid); +} + +int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) +{ + struct rpc_sequence *sequence = seqid->sequence->sequence; + int status = 0; + + spin_lock(&sequence->lock); + if (sequence->list.next != &seqid->list) { + seqid->task = task; + rpc_sleep_on(&sequence->wait, task, NULL, NULL); + status = -EAGAIN; + } + spin_unlock(&sequence->lock); + return status; } static int reclaimer(void *); @@ -791,8 +865,6 @@ static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct n if (state->state == 0) continue; status = ops->recover_open(sp, state); - list_for_each_entry(lock, &state->lock_states, ls_locks) - lock->ls_flags &= ~NFS_LOCK_INITIALIZED; if (status >= 0) { status = nfs4_reclaim_locks(ops, state); if (status < 0) @@ -831,6 +903,26 @@ out_err: return status; } +static void nfs4_state_mark_reclaim(struct nfs4_client *clp) +{ + struct nfs4_state_owner *sp; + struct nfs4_state *state; + struct nfs4_lock_state *lock; + + /* Reset all sequence ids to zero */ + list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + sp->so_seqid.counter = 0; + sp->so_seqid.flags = 0; + list_for_each_entry(state, &sp->so_states, open_states) { + list_for_each_entry(lock, &state->lock_states, ls_locks) { + lock->ls_seqid.counter = 0; + lock->ls_seqid.flags = 0; + lock->ls_flags &= ~NFS_LOCK_INITIALIZED; + } + } + } +} + static int reclaimer(void *ptr) { struct reclaimer_args *args = (struct reclaimer_args *)ptr; @@ -864,6 +956,7 @@ restart_loop: default: ops = &nfs4_network_partition_recovery_ops; }; + nfs4_state_mark_reclaim(clp); status = __nfs4_init_client(clp); if (status) goto out_error; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 6c564ef9489..fcd28a29a2f 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -604,7 +604,7 @@ static int encode_close(struct xdr_stream *xdr, const struct nfs_closeargs *arg) RESERVE_SPACE(8+sizeof(arg->stateid.data)); WRITE32(OP_CLOSE); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); return 0; @@ -732,9 +732,9 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lockargs *arg) struct nfs_open_to_lock *ol = opargs->u.open_lock; RESERVE_SPACE(40); - WRITE32(ol->open_seqid); + WRITE32(ol->open_seqid->sequence->counter); WRITEMEM(&ol->open_stateid, sizeof(ol->open_stateid)); - WRITE32(ol->lock_seqid); + WRITE32(ol->lock_seqid->sequence->counter); WRITE64(ol->lock_owner.clientid); WRITE32(4); WRITE32(ol->lock_owner.id); @@ -744,7 +744,7 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lockargs *arg) RESERVE_SPACE(20); WRITEMEM(&el->stateid, sizeof(el->stateid)); - WRITE32(el->seqid); + WRITE32(el->seqid->sequence->counter); } return 0; @@ -775,7 +775,7 @@ static int encode_locku(struct xdr_stream *xdr, const struct nfs_lockargs *arg) RESERVE_SPACE(44); WRITE32(OP_LOCKU); WRITE32(arg->type); - WRITE32(opargs->seqid); + WRITE32(opargs->seqid->sequence->counter); WRITEMEM(&opargs->stateid, sizeof(opargs->stateid)); WRITE64(arg->offset); WRITE64(arg->length); @@ -826,7 +826,7 @@ static inline void encode_openhdr(struct xdr_stream *xdr, const struct nfs_opena */ RESERVE_SPACE(8); WRITE32(OP_OPEN); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); RESERVE_SPACE(16); WRITE64(arg->clientid); @@ -941,7 +941,7 @@ static int encode_open_confirm(struct xdr_stream *xdr, const struct nfs_open_con RESERVE_SPACE(8+sizeof(arg->stateid.data)); WRITE32(OP_OPEN_CONFIRM); WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); return 0; } @@ -953,7 +953,7 @@ static int encode_open_downgrade(struct xdr_stream *xdr, const struct nfs_closea RESERVE_SPACE(8+sizeof(arg->stateid.data)); WRITE32(OP_OPEN_DOWNGRADE); WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); return 0; } @@ -1416,6 +1416,9 @@ static int nfs4_xdr_enc_close(struct rpc_rqst *req, uint32_t *p, struct nfs_clos }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1437,6 +1440,9 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1464,6 +1470,9 @@ static int nfs4_xdr_enc_open_confirm(struct rpc_rqst *req, uint32_t *p, struct n }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1485,6 +1494,9 @@ static int nfs4_xdr_enc_open_noattr(struct rpc_rqst *req, uint32_t *p, struct nf }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1506,6 +1518,9 @@ static int nfs4_xdr_enc_open_downgrade(struct rpc_rqst *req, uint32_t *p, struct }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1525,8 +1540,17 @@ static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_locka struct compound_hdr hdr = { .nops = 2, }; + struct nfs_lock_opargs *opargs = args->u.lock; + struct nfs_seqid *seqid; int status; + if (opargs->new_lock_owner) + seqid = opargs->u.open_lock->lock_seqid; + else + seqid = opargs->u.exist_lock->seqid; + status = nfs_wait_on_sequence(seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1569,6 +1593,9 @@ static int nfs4_xdr_enc_locku(struct rpc_rqst *req, uint32_t *p, struct nfs_lock }; int status; + status = nfs_wait_on_sequence(args->u.locku->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); -- cgit v1.2.3 From 9512135df14f8293b9bc5e8fb22d4279dee5ff66 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:12 -0700 Subject: NFSv4: Fix a potential CLOSE race Once the state_owner and lock_owner semaphores get removed, it will be possible for other OPEN requests to reopen the same file if they have lower sequence ids than our CLOSE call. This patch ensures that we recheck the file state once nfs_wait_on_sequence() has completed waiting. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 110 +++++++++++++++++++++++++++++++++++++---------------- fs/nfs/nfs4state.c | 6 ++- fs/nfs/nfs4xdr.c | 14 ++----- 3 files changed, 86 insertions(+), 44 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9ba89e7cdd2..5154ddf6d9a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -189,6 +189,21 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf nfsi->change_attr = cinfo->after; } +/* Helper for asynchronous RPC calls */ +static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin, + rpc_action tk_exit, void *calldata) +{ + struct rpc_task *task; + + if (!(task = rpc_new_task(clnt, tk_exit, RPC_TASK_ASYNC))) + return -ENOMEM; + + task->tk_calldata = calldata; + task->tk_action = tk_begin; + rpc_execute(task); + return 0; +} + static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) { struct inode *inode = state->inode; @@ -810,11 +825,24 @@ struct nfs4_closedata { struct nfs_closeres res; }; +static void nfs4_free_closedata(struct nfs4_closedata *calldata) +{ + struct nfs4_state *state = calldata->state; + struct nfs4_state_owner *sp = state->owner; + struct nfs_server *server = NFS_SERVER(calldata->inode); + + nfs4_put_open_state(calldata->state); + nfs_free_seqid(calldata->arg.seqid); + up(&sp->so_sema); + nfs4_put_state_owner(sp); + up_read(&server->nfs4_state->cl_sem); + kfree(calldata); +} + static void nfs4_close_done(struct rpc_task *task) { struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; struct nfs4_state *state = calldata->state; - struct nfs4_state_owner *sp = state->owner; struct nfs_server *server = NFS_SERVER(calldata->inode); /* hmm. we are done with the inode, and in the process of freeing @@ -838,25 +866,46 @@ static void nfs4_close_done(struct rpc_task *task) } } state->state = calldata->arg.open_flags; - nfs4_put_open_state(state); - nfs_free_seqid(calldata->arg.seqid); - up(&sp->so_sema); - nfs4_put_state_owner(sp); - up_read(&server->nfs4_state->cl_sem); - kfree(calldata); + nfs4_free_closedata(calldata); } -static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata) +static void nfs4_close_begin(struct rpc_task *task) { + struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; + struct nfs4_state *state = calldata->state; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE], .rpc_argp = &calldata->arg, .rpc_resp = &calldata->res, - .rpc_cred = calldata->state->owner->so_cred, + .rpc_cred = state->owner->so_cred, }; - if (calldata->arg.open_flags != 0) + int mode = 0; + int status; + + status = nfs_wait_on_sequence(calldata->arg.seqid, task); + if (status != 0) + return; + /* Don't reorder reads */ + smp_rmb(); + /* Recalculate the new open mode in case someone reopened the file + * while we were waiting in line to be scheduled. + */ + if (state->nreaders != 0) + mode |= FMODE_READ; + if (state->nwriters != 0) + mode |= FMODE_WRITE; + if (test_bit(NFS_DELEGATED_STATE, &state->flags)) + state->state = mode; + if (mode == state->state) { + nfs4_free_closedata(calldata); + task->tk_exit = NULL; + rpc_exit(task, 0); + return; + } + if (mode != 0) msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE]; - return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata); + calldata->arg.open_flags = mode; + rpc_call_setup(task, &msg, 0); } /* @@ -873,35 +922,30 @@ static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata * int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) { struct nfs4_closedata *calldata; - int status; + int status = -ENOMEM; - /* Tell caller we're done */ - if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { - state->state = mode; - return 0; - } - calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL); + calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); if (calldata == NULL) - return -ENOMEM; + goto out; calldata->inode = inode; calldata->state = state; calldata->arg.fh = NFS_FH(inode); + calldata->arg.stateid = &state->stateid; /* Serialization for the sequence id */ calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid); - if (calldata->arg.seqid == NULL) { - kfree(calldata); - return -ENOMEM; - } - calldata->arg.open_flags = mode; - memcpy(&calldata->arg.stateid, &state->stateid, - sizeof(calldata->arg.stateid)); - status = nfs4_close_call(NFS_SERVER(inode)->client, calldata); - /* - * Return -EINPROGRESS on success in order to indicate to the - * caller that an asynchronous RPC call has been launched, and - * that it will release the semaphores on completion. - */ - return (status == 0) ? -EINPROGRESS : status; + if (calldata->arg.seqid == NULL) + goto out_free_calldata; + + status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_close_begin, + nfs4_close_done, calldata); + if (status == 0) + goto out; + + nfs_free_seqid(calldata->arg.seqid); +out_free_calldata: + kfree(calldata); +out: + return status; } struct inode * diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index f535c219cf3..59c93f37e1b 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -518,7 +518,11 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) newstate |= FMODE_WRITE; if (state->state == newstate) goto out; - if (nfs4_do_close(inode, state, newstate) == -EINPROGRESS) + if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { + state->state = newstate; + goto out; + } + if (nfs4_do_close(inode, state, newstate) == 0) return; } out: diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index fcd28a29a2f..934ec50ea6b 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -602,10 +602,10 @@ static int encode_close(struct xdr_stream *xdr, const struct nfs_closeargs *arg) { uint32_t *p; - RESERVE_SPACE(8+sizeof(arg->stateid.data)); + RESERVE_SPACE(8+sizeof(arg->stateid->data)); WRITE32(OP_CLOSE); WRITE32(arg->seqid->sequence->counter); - WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); + WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data)); return 0; } @@ -950,9 +950,9 @@ static int encode_open_downgrade(struct xdr_stream *xdr, const struct nfs_closea { uint32_t *p; - RESERVE_SPACE(8+sizeof(arg->stateid.data)); + RESERVE_SPACE(8+sizeof(arg->stateid->data)); WRITE32(OP_OPEN_DOWNGRADE); - WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); + WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data)); WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); return 0; @@ -1416,9 +1416,6 @@ static int nfs4_xdr_enc_close(struct rpc_rqst *req, uint32_t *p, struct nfs_clos }; int status; - status = nfs_wait_on_sequence(args->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1518,9 +1515,6 @@ static int nfs4_xdr_enc_open_downgrade(struct rpc_rqst *req, uint32_t *p, struct }; int status; - status = nfs_wait_on_sequence(args->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); -- cgit v1.2.3 From e6dfa553cffcb9740f932311dff42f81d6ac63bb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:13 -0700 Subject: NFSv4: Remove obsolete state_owner and lock_owner semaphores OPEN, CLOSE, etc no longer need these semaphores to ensure ordering of requests. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 6 ------ fs/nfs/nfs4proc.c | 18 ------------------ fs/nfs/nfs4state.c | 15 ++++----------- 3 files changed, 4 insertions(+), 35 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 6ac6708484f..d4fcb5d0ce6 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -125,16 +125,11 @@ static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status * NFS4 state_owners and lock_owners are simply labels for ordered * sequences of RPC calls. Their sole purpose is to provide once-only * semantics by allowing the server to identify replayed requests. - * - * The ->so_sema is held during all state_owner seqid-mutating operations: - * OPEN, OPEN_DOWNGRADE, and CLOSE. Its purpose is to properly serialize - * so_seqid. */ struct nfs4_state_owner { struct list_head so_list; /* per-clientid list of state_owners */ struct nfs4_client *so_client; u32 so_id; /* 32-bit identifier, unique */ - struct semaphore so_sema; atomic_t so_count; struct rpc_cred *so_cred; /* Associated cred */ @@ -183,7 +178,6 @@ struct nfs4_state { struct inode *inode; /* Pointer to the inode */ unsigned long flags; /* Do we hold any locks? */ - struct semaphore lock_sema; /* Serializes file locking operations */ spinlock_t state_lock; /* Protects the lock_states list */ nfs4_stateid stateid; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 5154ddf6d9a..25bab6032df 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -224,7 +224,6 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, /* * OPEN_RECLAIM: * reclaim state on the server after a reboot. - * Assumes caller is holding the sp->so_sem */ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) { @@ -322,7 +321,6 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state }; int status = 0; - down(&sp->so_sema); if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) goto out; if (state->state == 0) @@ -342,7 +340,6 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state } nfs_free_seqid(arg.seqid); out: - up(&sp->so_sema); dput(parent); return status; } @@ -595,7 +592,6 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__); goto out_err; } - down(&sp->so_sema); state = nfs4_get_open_state(inode, sp); if (state == NULL) goto out_err; @@ -620,7 +616,6 @@ static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred set_bit(NFS_DELEGATED_STATE, &state->flags); update_open_stateid(state, &delegation->stateid, open_flags); out_ok: - up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&nfsi->rwsem); up_read(&clp->cl_sem); @@ -631,7 +626,6 @@ out_err: if (sp != NULL) { if (state != NULL) nfs4_put_open_state(state); - up(&sp->so_sema); nfs4_put_state_owner(sp); } up_read(&nfsi->rwsem); @@ -696,7 +690,6 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st } else o_arg.u.attrs = sattr; /* Serialization for the sequence id */ - down(&sp->so_sema); o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); if (o_arg.seqid == NULL) @@ -716,7 +709,6 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st if (o_res.delegation_type != 0) nfs_inode_set_delegation(inode, cred, &o_res); nfs_free_seqid(o_arg.seqid); - up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&clp->cl_sem); *res = state; @@ -726,7 +718,6 @@ out_err: if (state != NULL) nfs4_put_open_state(state); nfs_free_seqid(o_arg.seqid); - up(&sp->so_sema); nfs4_put_state_owner(sp); } /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */ @@ -833,7 +824,6 @@ static void nfs4_free_closedata(struct nfs4_closedata *calldata) nfs4_put_open_state(calldata->state); nfs_free_seqid(calldata->arg.seqid); - up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&server->nfs4_state->cl_sem); kfree(calldata); @@ -2702,7 +2692,6 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock down_read(&clp->cl_sem); nlo.clientid = clp->cl_clientid; - down(&state->lock_sema); status = nfs4_set_lock_state(state, request); if (status != 0) goto out; @@ -2729,7 +2718,6 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock status = 0; } out: - up(&state->lock_sema); up_read(&clp->cl_sem); return status; } @@ -2791,7 +2779,6 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock int status; down_read(&clp->cl_sem); - down(&state->lock_sema); status = nfs4_set_lock_state(state, request); if (status != 0) goto out; @@ -2813,7 +2800,6 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock sizeof(lsp->ls_stateid.data)); nfs_free_seqid(luargs.seqid); out: - up(&state->lock_sema); if (status == 0) do_vfs_lock(request->fl_file, request); up_read(&clp->cl_sem); @@ -2877,7 +2863,6 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r largs.u.open_lock = &otl; largs.new_lock_owner = 1; arg.u.lock = &largs; - down(&owner->so_sema); otl.open_seqid = nfs_alloc_seqid(&owner->so_seqid); if (otl.open_seqid != NULL) { status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); @@ -2885,7 +2870,6 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r nfs_increment_open_seqid(status, otl.open_seqid); nfs_free_seqid(otl.open_seqid); } - up(&owner->so_sema); if (status == 0) nfs_confirm_seqid(&lsp->ls_seqid, 0); } else { @@ -2944,11 +2928,9 @@ static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock int status; down_read(&clp->cl_sem); - down(&state->lock_sema); status = nfs4_set_lock_state(state, request); if (status == 0) status = _nfs4_do_setlk(state, cmd, request, 0); - up(&state->lock_sema); if (status == 0) { /* Note: we always want to sleep here! */ request->fl_flags |= FL_SLEEP; diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 59c93f37e1b..86c08c165ce 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -267,7 +267,6 @@ nfs4_alloc_state_owner(void) sp = kzalloc(sizeof(*sp),GFP_KERNEL); if (!sp) return NULL; - init_MUTEX(&sp->so_sema); INIT_LIST_HEAD(&sp->so_states); INIT_LIST_HEAD(&sp->so_delegations); rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); @@ -362,7 +361,6 @@ nfs4_alloc_open_state(void) memset(state->stateid.data, 0, sizeof(state->stateid.data)); atomic_set(&state->count, 1); INIT_LIST_HEAD(&state->lock_states); - init_MUTEX(&state->lock_sema); spin_lock_init(&state->state_lock); return state; } @@ -444,7 +442,6 @@ nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner) state = __nfs4_find_state_byowner(inode, owner); if (state == NULL && new != NULL) { state = new; - /* Caller *must* be holding owner->so_sem */ /* Note: The reclaim code dictates that we add stateless * and read-only stateids to the end of the list */ list_add_tail(&state->open_states, &owner->so_states); @@ -464,7 +461,7 @@ out: /* * Beware! Caller must be holding exactly one - * reference to clp->cl_sem and owner->so_sema! + * reference to clp->cl_sem! */ void nfs4_put_open_state(struct nfs4_state *state) { @@ -485,7 +482,6 @@ void nfs4_put_open_state(struct nfs4_state *state) /* * Beware! Caller must be holding no references to clp->cl_sem! - * of owner->so_sema! */ void nfs4_close_state(struct nfs4_state *state, mode_t mode) { @@ -496,7 +492,6 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) atomic_inc(&owner->so_count); down_read(&clp->cl_sem); - down(&owner->so_sema); /* Protect against nfs4_find_state() */ spin_lock(&inode->i_lock); if (mode & FMODE_READ) @@ -527,7 +522,6 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) } out: nfs4_put_open_state(state); - up(&owner->so_sema); nfs4_put_state_owner(owner); up_read(&clp->cl_sem); } @@ -553,7 +547,6 @@ __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) * Return a compatible lock_state. If no initialized lock_state structure * exists, return an uninitialized one. * - * The caller must be holding state->lock_sema */ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) { @@ -577,7 +570,7 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f * Return a compatible lock_state. If no initialized lock_state structure * exists, return an uninitialized one. * - * The caller must be holding state->lock_sema and clp->cl_sem + * The caller must be holding clp->cl_sem */ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) { @@ -711,7 +704,7 @@ void nfs_free_seqid(struct nfs_seqid *seqid) } /* - * Called with sp->so_sema and clp->cl_sem held. + * Called with clp->cl_sem held. * * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or * failed with a seqid incrementing error - @@ -750,7 +743,7 @@ void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) } /* - * Called with ls->lock_sema and clp->cl_sem held. + * Called with clp->cl_sem held. * * Increment the seqid if the LOCK/LOCKU succeeded, or * failed with a seqid incrementing error - -- cgit v1.2.3 From 83c9d41e456033000ccfadf535f3098d8739488d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:13 -0700 Subject: NFSv4: Remove nfs4_client->cl_sem from close() path We no longer need to worry about collisions between close() and the state recovery code, since the new close will automatically recheck the file state once it is done waiting on its sequence slot. Ditto for the nfs4_proc_locku() procedure. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 5 ----- fs/nfs/nfs4state.c | 9 +-------- 2 files changed, 1 insertion(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 25bab6032df..611052bacaf 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -820,12 +820,10 @@ static void nfs4_free_closedata(struct nfs4_closedata *calldata) { struct nfs4_state *state = calldata->state; struct nfs4_state_owner *sp = state->owner; - struct nfs_server *server = NFS_SERVER(calldata->inode); nfs4_put_open_state(calldata->state); nfs_free_seqid(calldata->arg.seqid); nfs4_put_state_owner(sp); - up_read(&server->nfs4_state->cl_sem); kfree(calldata); } @@ -2758,7 +2756,6 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock { struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); - struct nfs4_client *clp = server->nfs4_state; struct nfs_lockargs arg = { .fh = NFS_FH(inode), .type = nfs4_lck_type(cmd, request), @@ -2778,7 +2775,6 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock struct nfs_locku_opargs luargs; int status; - down_read(&clp->cl_sem); status = nfs4_set_lock_state(state, request); if (status != 0) goto out; @@ -2802,7 +2798,6 @@ static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock out: if (status == 0) do_vfs_lock(request->fl_file, request); - up_read(&clp->cl_sem); return status; } diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 86c08c165ce..bb357436195 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -481,17 +481,15 @@ void nfs4_put_open_state(struct nfs4_state *state) } /* - * Beware! Caller must be holding no references to clp->cl_sem! + * Close the current file. */ void nfs4_close_state(struct nfs4_state *state, mode_t mode) { struct inode *inode = state->inode; struct nfs4_state_owner *owner = state->owner; - struct nfs4_client *clp = owner->so_client; int newstate; atomic_inc(&owner->so_count); - down_read(&clp->cl_sem); /* Protect against nfs4_find_state() */ spin_lock(&inode->i_lock); if (mode & FMODE_READ) @@ -523,7 +521,6 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) out: nfs4_put_open_state(state); nfs4_put_state_owner(owner); - up_read(&clp->cl_sem); } /* @@ -704,8 +701,6 @@ void nfs_free_seqid(struct nfs_seqid *seqid) } /* - * Called with clp->cl_sem held. - * * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or * failed with a seqid incrementing error - * see comments nfs_fs.h:seqid_mutating_error() @@ -743,8 +738,6 @@ void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) } /* - * Called with clp->cl_sem held. - * * Increment the seqid if the LOCK/LOCKU succeeded, or * failed with a seqid incrementing error - * see comments nfs_fs.h:seqid_mutating_error() -- cgit v1.2.3 From 0a8838f972883112f0a7b259141b24db17583c2d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:14 -0700 Subject: NFSv4: Add missing handling of OPEN_CONFIRM requests on CLAIM_DELEGATE_CUR. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 611052bacaf..f57dba81509 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -56,6 +56,7 @@ #define NFS4_POLL_RETRY_MIN (1*HZ) #define NFS4_POLL_RETRY_MAX (15*HZ) +static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid); static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); @@ -333,11 +334,21 @@ static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data)); status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); nfs_increment_open_seqid(status, arg.seqid); + if (status != 0) + goto out_free; + if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) { + status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode), + sp, &res.stateid, arg.seqid); + if (status != 0) + goto out_free; + } + nfs_confirm_seqid(&sp->so_seqid, 0); if (status >= 0) { memcpy(state->stateid.data, res.stateid.data, sizeof(state->stateid.data)); clear_bit(NFS_DELEGATED_STATE, &state->flags); } +out_free: nfs_free_seqid(arg.seqid); out: dput(parent); @@ -366,7 +377,7 @@ int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) return err; } -static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid) +static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid) { struct nfs_open_confirmargs arg = { .fh = fh, -- cgit v1.2.3 From faf5f49c2d9c0af2847837c232a432cc146e203b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:15 -0700 Subject: NFSv4: Make NFS clean up byte range locks asynchronously Currently we fail to do so if the process was signalled. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 1 + fs/nfs/nfs4proc.c | 161 ++++++++++++++++++++++++++++++++++++----------------- fs/nfs/nfs4state.c | 2 +- fs/nfs/nfs4xdr.c | 9 +-- 4 files changed, 115 insertions(+), 58 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index d4fcb5d0ce6..2215cdee43a 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -249,6 +249,7 @@ extern void nfs4_put_open_state(struct nfs4_state *); extern void nfs4_close_state(struct nfs4_state *, mode_t); extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode); extern void nfs4_schedule_state_recovery(struct nfs4_client *); +extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp); extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f57dba81509..612a9a14aed 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -58,9 +58,9 @@ static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid); static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); -static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *); +static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); -static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception); +static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception); extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus); extern struct rpc_procinfo nfs4_procedures[]; @@ -2422,7 +2422,7 @@ static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen } static int -nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server) +nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server) { struct nfs4_client *clp = server->nfs4_state; @@ -2500,7 +2500,7 @@ static int nfs4_delay(struct rpc_clnt *clnt, long *timeout) /* This is the error handling routine for processes that are allowed * to sleep. */ -int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception) +int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception) { struct nfs4_client *clp = server->nfs4_state; int ret = errorcode; @@ -2763,68 +2763,127 @@ static int do_vfs_lock(struct file *file, struct file_lock *fl) return res; } -static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) +struct nfs4_unlockdata { + struct nfs_lockargs arg; + struct nfs_locku_opargs luargs; + struct nfs_lockres res; + struct nfs4_lock_state *lsp; + struct nfs_open_context *ctx; + atomic_t refcount; + struct completion completion; +}; + +static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata) { - struct inode *inode = state->inode; - struct nfs_server *server = NFS_SERVER(inode); - struct nfs_lockargs arg = { - .fh = NFS_FH(inode), - .type = nfs4_lck_type(cmd, request), - .offset = request->fl_start, - .length = nfs4_lck_length(request), - }; - struct nfs_lockres res = { - .server = server, - }; + if (atomic_dec_and_test(&calldata->refcount)) { + nfs_free_seqid(calldata->luargs.seqid); + nfs4_put_lock_state(calldata->lsp); + put_nfs_open_context(calldata->ctx); + kfree(calldata); + } +} + +static void nfs4_locku_complete(struct nfs4_unlockdata *calldata) +{ + complete(&calldata->completion); + nfs4_locku_release_calldata(calldata); +} + +static void nfs4_locku_done(struct rpc_task *task) +{ + struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; + + nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid); + switch (task->tk_status) { + case 0: + memcpy(calldata->lsp->ls_stateid.data, + calldata->res.u.stateid.data, + sizeof(calldata->lsp->ls_stateid.data)); + break; + case -NFS4ERR_STALE_STATEID: + case -NFS4ERR_EXPIRED: + nfs4_schedule_state_recovery(calldata->res.server->nfs4_state); + break; + default: + if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) { + rpc_restart_call(task); + return; + } + } + nfs4_locku_complete(calldata); +} + +static void nfs4_locku_begin(struct rpc_task *task) +{ + struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU], - .rpc_argp = &arg, - .rpc_resp = &res, - .rpc_cred = state->owner->so_cred, + .rpc_argp = &calldata->arg, + .rpc_resp = &calldata->res, + .rpc_cred = calldata->lsp->ls_state->owner->so_cred, }; + int status; + + status = nfs_wait_on_sequence(calldata->luargs.seqid, task); + if (status != 0) + return; + if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) { + nfs4_locku_complete(calldata); + task->tk_exit = NULL; + rpc_exit(task, 0); + return; + } + rpc_call_setup(task, &msg, 0); +} + +static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) +{ + struct nfs4_unlockdata *calldata; + struct inode *inode = state->inode; + struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp; - struct nfs_locku_opargs luargs; int status; - + status = nfs4_set_lock_state(state, request); if (status != 0) - goto out; + return status; lsp = request->fl_u.nfs4_fl.owner; /* We might have lost the locks! */ if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) - goto out; - luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); - status = -ENOMEM; - if (luargs.seqid == NULL) - goto out; - memcpy(luargs.stateid.data, lsp->ls_stateid.data, sizeof(luargs.stateid.data)); - arg.u.locku = &luargs; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs_increment_lock_seqid(status, luargs.seqid); - - if (status == 0) - memcpy(lsp->ls_stateid.data, res.u.stateid.data, - sizeof(lsp->ls_stateid.data)); - nfs_free_seqid(luargs.seqid); -out: + return 0; + calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); + if (calldata == NULL) + return -ENOMEM; + calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (calldata->luargs.seqid == NULL) { + kfree(calldata); + return -ENOMEM; + } + calldata->luargs.stateid = &lsp->ls_stateid; + calldata->arg.fh = NFS_FH(inode); + calldata->arg.type = nfs4_lck_type(cmd, request); + calldata->arg.offset = request->fl_start; + calldata->arg.length = nfs4_lck_length(request); + calldata->arg.u.locku = &calldata->luargs; + calldata->res.server = server; + calldata->lsp = lsp; + atomic_inc(&lsp->ls_count); + + /* Ensure we don't close file until we're done freeing locks! */ + calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data); + + atomic_set(&calldata->refcount, 2); + init_completion(&calldata->completion); + + status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin, + nfs4_locku_done, calldata); if (status == 0) - do_vfs_lock(request->fl_file, request); + wait_for_completion_interruptible(&calldata->completion); + do_vfs_lock(request->fl_file, request); + nfs4_locku_release_calldata(calldata); return status; } -static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) -{ - struct nfs4_exception exception = { }; - int err; - - do { - err = nfs4_handle_exception(NFS_SERVER(state->inode), - _nfs4_proc_unlck(state, cmd, request), - &exception); - } while (exception.retry); - return err; -} - static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim) { struct inode *inode = state->inode; diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index bb357436195..23834c8fb74 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -600,7 +600,7 @@ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_ * Release reference to lock_state, and free it if we see that * it is no longer in use */ -static void nfs4_put_lock_state(struct nfs4_lock_state *lsp) +void nfs4_put_lock_state(struct nfs4_lock_state *lsp) { struct nfs4_state *state; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 934ec50ea6b..4706192cfb0 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -776,7 +776,7 @@ static int encode_locku(struct xdr_stream *xdr, const struct nfs_lockargs *arg) WRITE32(OP_LOCKU); WRITE32(arg->type); WRITE32(opargs->seqid->sequence->counter); - WRITEMEM(&opargs->stateid, sizeof(opargs->stateid)); + WRITEMEM(opargs->stateid->data, sizeof(opargs->stateid->data)); WRITE64(arg->offset); WRITE64(arg->length); @@ -1587,9 +1587,6 @@ static int nfs4_xdr_enc_locku(struct rpc_rqst *req, uint32_t *p, struct nfs_lock }; int status; - status = nfs_wait_on_sequence(args->u.locku->seqid, req->rq_task); - if (status != 0) - goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -2934,8 +2931,8 @@ static int decode_locku(struct xdr_stream *xdr, struct nfs_lockres *res) status = decode_op_hdr(xdr, OP_LOCKU); if (status == 0) { - READ_BUF(sizeof(nfs4_stateid)); - COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); + READ_BUF(sizeof(res->u.stateid.data)); + COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); } return status; } -- cgit v1.2.3 From 06735b3454824bd561decbde46111f144e905923 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:15 -0700 Subject: NFSv4: Fix up handling of open_to_lock sequence ids Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 69 ++++++++++++++++++++++++++----------------------------- fs/nfs/nfs4xdr.c | 32 +++++++++++--------------- 2 files changed, 45 insertions(+), 56 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 612a9a14aed..35da15342e0 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2889,11 +2889,23 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; + struct nfs_lock_opargs largs = { + .lock_stateid = &lsp->ls_stateid, + .open_stateid = &state->stateid, + .lock_owner = { + .clientid = server->nfs4_state->cl_clientid, + .id = lsp->ls_id, + }, + .reclaim = reclaim, + }; struct nfs_lockargs arg = { .fh = NFS_FH(inode), .type = nfs4_lck_type(cmd, request), .offset = request->fl_start, .length = nfs4_lck_length(request), + .u = { + .lock = &largs, + }, }; struct nfs_lockres res = { .server = server, @@ -2904,56 +2916,39 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *r .rpc_resp = &res, .rpc_cred = state->owner->so_cred, }; - struct nfs_lock_opargs largs = { - .reclaim = reclaim, - .new_lock_owner = 0, - }; - struct nfs_seqid *lock_seqid; int status = -ENOMEM; - lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); - if (lock_seqid == NULL) + largs.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (largs.lock_seqid == NULL) return -ENOMEM; if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) { struct nfs4_state_owner *owner = state->owner; - struct nfs_open_to_lock otl = { - .lock_owner = { - .clientid = server->nfs4_state->cl_clientid, - }, - }; - - otl.lock_seqid = lock_seqid; - otl.lock_owner.id = lsp->ls_id; - memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid)); - largs.u.open_lock = &otl; + + largs.open_seqid = nfs_alloc_seqid(&owner->so_seqid); + if (largs.open_seqid == NULL) + goto out; largs.new_lock_owner = 1; - arg.u.lock = &largs; - otl.open_seqid = nfs_alloc_seqid(&owner->so_seqid); - if (otl.open_seqid != NULL) { - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment seqid on success, and seqid mutating errors */ - nfs_increment_open_seqid(status, otl.open_seqid); - nfs_free_seqid(otl.open_seqid); + status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); + /* increment open seqid on success, and seqid mutating errors */ + if (largs.new_lock_owner != 0) { + nfs_increment_open_seqid(status, largs.open_seqid); + if (status == 0) + nfs_confirm_seqid(&lsp->ls_seqid, 0); } - if (status == 0) - nfs_confirm_seqid(&lsp->ls_seqid, 0); - } else { - struct nfs_exist_lock el; - memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid)); - largs.u.exist_lock = ⪙ - arg.u.lock = &largs; - el.seqid = lock_seqid; + nfs_free_seqid(largs.open_seqid); + } else status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - } - /* increment seqid on success, and seqid mutating errors*/ - nfs_increment_lock_seqid(status, lock_seqid); + /* increment lock seqid on success, and seqid mutating errors*/ + nfs_increment_lock_seqid(status, largs.lock_seqid); /* save the returned stateid. */ if (status == 0) { - memcpy(lsp->ls_stateid.data, res.u.stateid.data, sizeof(lsp->ls_stateid.data)); + memcpy(lsp->ls_stateid.data, res.u.stateid.data, + sizeof(lsp->ls_stateid.data)); lsp->ls_flags |= NFS_LOCK_INITIALIZED; } else if (status == -NFS4ERR_DENIED) status = -EAGAIN; - nfs_free_seqid(lock_seqid); +out: + nfs_free_seqid(largs.lock_seqid); return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 4706192cfb0..c5c75235c5b 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -729,22 +729,18 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lockargs *arg) WRITE64(arg->length); WRITE32(opargs->new_lock_owner); if (opargs->new_lock_owner){ - struct nfs_open_to_lock *ol = opargs->u.open_lock; - RESERVE_SPACE(40); - WRITE32(ol->open_seqid->sequence->counter); - WRITEMEM(&ol->open_stateid, sizeof(ol->open_stateid)); - WRITE32(ol->lock_seqid->sequence->counter); - WRITE64(ol->lock_owner.clientid); + WRITE32(opargs->open_seqid->sequence->counter); + WRITEMEM(opargs->open_stateid->data, sizeof(opargs->open_stateid->data)); + WRITE32(opargs->lock_seqid->sequence->counter); + WRITE64(opargs->lock_owner.clientid); WRITE32(4); - WRITE32(ol->lock_owner.id); + WRITE32(opargs->lock_owner.id); } else { - struct nfs_exist_lock *el = opargs->u.exist_lock; - RESERVE_SPACE(20); - WRITEMEM(&el->stateid, sizeof(el->stateid)); - WRITE32(el->seqid->sequence->counter); + WRITEMEM(opargs->lock_stateid->data, sizeof(opargs->lock_stateid->data)); + WRITE32(opargs->lock_seqid->sequence->counter); } return 0; @@ -1535,16 +1531,14 @@ static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_locka .nops = 2, }; struct nfs_lock_opargs *opargs = args->u.lock; - struct nfs_seqid *seqid; int status; - if (opargs->new_lock_owner) - seqid = opargs->u.open_lock->lock_seqid; - else - seqid = opargs->u.exist_lock->seqid; - status = nfs_wait_on_sequence(seqid, req->rq_task); + status = nfs_wait_on_sequence(opargs->lock_seqid, req->rq_task); if (status != 0) goto out; + /* Do we need to do an open_to_lock_owner? */ + if (opargs->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED) + opargs->new_lock_owner = 0; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -2908,8 +2902,8 @@ static int decode_lock(struct xdr_stream *xdr, struct nfs_lockres *res) status = decode_op_hdr(xdr, OP_LOCK); if (status == 0) { - READ_BUF(sizeof(nfs4_stateid)); - COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); + READ_BUF(sizeof(res->u.stateid.data)); + COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); } else if (status == -NFS4ERR_DENIED) return decode_lock_denied(xdr, &res->u.denied); return status; -- cgit v1.2.3 From 039c4d7a82d8268ec71f59679460b41d0dd9b225 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:16 -0700 Subject: NFS: Fix up a race in the NFS implementation of GETLK ...and fix a memory corruption bug due to improper use of memcpy() on a struct file_lock. Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 27 ++++++++++++++++++--------- 1 file changed, 18 insertions(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 6bdcfa95de9..572d8593486 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -376,22 +376,31 @@ out_swapfile: static int do_getlk(struct file *filp, int cmd, struct file_lock *fl) { + struct file_lock *cfl; struct inode *inode = filp->f_mapping->host; int status = 0; lock_kernel(); - /* Use local locking if mounted with "-onolock" */ - if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) - status = NFS_PROTO(inode)->lock(filp, cmd, fl); - else { - struct file_lock *cfl = posix_test_lock(filp, fl); - - fl->fl_type = F_UNLCK; - if (cfl != NULL) - memcpy(fl, cfl, sizeof(*fl)); + /* Try local locking first */ + cfl = posix_test_lock(filp, fl); + if (cfl != NULL) { + locks_copy_lock(fl, cfl); + goto out; } + + if (nfs_have_delegation(inode, FMODE_READ)) + goto out_noconflict; + + if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM) + goto out_noconflict; + + status = NFS_PROTO(inode)->lock(filp, cmd, fl); +out: unlock_kernel(); return status; +out_noconflict: + fl->fl_type = F_UNLCK; + goto out; } static int do_vfs_lock(struct file *file, struct file_lock *fl) -- cgit v1.2.3 From 834f2a4a1554dc5b2598038b3fe8703defcbe467 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:16 -0700 Subject: VFS: Allow the filesystem to return a full file pointer on open intent This is needed by NFSv4 for atomicity reasons: our open command is in fact a lookup+open, so we need to be able to propagate open context information from lookup() into the resulting struct file's private_data field. Signed-off-by: Trond Myklebust --- fs/exec.c | 12 ++++---- fs/namei.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++----- fs/open.c | 79 ++++++++++++++++++++++++++++++++++++++++++---------- 3 files changed, 157 insertions(+), 27 deletions(-) (limited to 'fs') diff --git a/fs/exec.c b/fs/exec.c index a04a575ad43..d2208f7c87d 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -126,8 +126,7 @@ asmlinkage long sys_uselib(const char __user * library) struct nameidata nd; int error; - nd.intent.open.flags = FMODE_READ; - error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); + error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ); if (error) goto out; @@ -139,7 +138,7 @@ asmlinkage long sys_uselib(const char __user * library) if (error) goto exit; - file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); + file = nameidata_to_filp(&nd, O_RDONLY); error = PTR_ERR(file); if (IS_ERR(file)) goto out; @@ -167,6 +166,7 @@ asmlinkage long sys_uselib(const char __user * library) out: return error; exit: + release_open_intent(&nd); path_release(&nd); goto out; } @@ -490,8 +490,7 @@ struct file *open_exec(const char *name) int err; struct file *file; - nd.intent.open.flags = FMODE_READ; - err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); + err = path_lookup_open(name, LOOKUP_FOLLOW, &nd, FMODE_READ); file = ERR_PTR(err); if (!err) { @@ -504,7 +503,7 @@ struct file *open_exec(const char *name) err = -EACCES; file = ERR_PTR(err); if (!err) { - file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); + file = nameidata_to_filp(&nd, O_RDONLY); if (!IS_ERR(file)) { err = deny_write_access(file); if (err) { @@ -516,6 +515,7 @@ out: return file; } } + release_open_intent(&nd); path_release(&nd); } goto out; diff --git a/fs/namei.c b/fs/namei.c index aa62dbda93a..0d1dff7d3d9 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -28,6 +28,7 @@ #include #include #include +#include #include #include @@ -317,6 +318,18 @@ void path_release_on_umount(struct nameidata *nd) mntput_no_expire(nd->mnt); } +/** + * release_open_intent - free up open intent resources + * @nd: pointer to nameidata + */ +void release_open_intent(struct nameidata *nd) +{ + if (nd->intent.open.file->f_dentry == NULL) + put_filp(nd->intent.open.file); + else + fput(nd->intent.open.file); +} + /* * Internal lookup() using the new generic dcache. * SMP-safe @@ -1052,6 +1065,70 @@ out: return retval; } +static int __path_lookup_intent_open(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags, int create_mode) +{ + struct file *filp = get_empty_filp(); + int err; + + if (filp == NULL) + return -ENFILE; + nd->intent.open.file = filp; + nd->intent.open.flags = open_flags; + nd->intent.open.create_mode = create_mode; + err = path_lookup(name, lookup_flags|LOOKUP_OPEN, nd); + if (IS_ERR(nd->intent.open.file)) { + if (err == 0) { + err = PTR_ERR(nd->intent.open.file); + path_release(nd); + } + } else if (err != 0) + release_open_intent(nd); + return err; +} + +/** + * path_lookup_open - lookup a file path with open intent + * @name: pointer to file name + * @lookup_flags: lookup intent flags + * @nd: pointer to nameidata + * @open_flags: open intent flags + */ +int path_lookup_open(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags) +{ + return __path_lookup_intent_open(name, lookup_flags, nd, + open_flags, 0); +} + +/** + * path_lookup_create - lookup a file path with open + create intent + * @name: pointer to file name + * @lookup_flags: lookup intent flags + * @nd: pointer to nameidata + * @open_flags: open intent flags + * @create_mode: create intent flags + */ +int path_lookup_create(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags, int create_mode) +{ + return __path_lookup_intent_open(name, lookup_flags|LOOKUP_CREATE, nd, + open_flags, create_mode); +} + +int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags) +{ + char *tmp = getname(name); + int err = PTR_ERR(tmp); + + if (!IS_ERR(tmp)) { + err = __path_lookup_intent_open(tmp, lookup_flags, nd, open_flags, 0); + putname(tmp); + } + return err; +} + /* * Restricted form of lookup. Doesn't follow links, single-component only, * needs parent already locked. Doesn't follow mounts. @@ -1416,27 +1493,27 @@ int may_open(struct nameidata *nd, int acc_mode, int flag) */ int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) { - int acc_mode, error = 0; + int acc_mode, error; struct path path; struct dentry *dir; int count = 0; acc_mode = ACC_MODE(flag); + /* O_TRUNC implies we need access checks for write permissions */ + if (flag & O_TRUNC) + acc_mode |= MAY_WRITE; + /* Allow the LSM permission hook to distinguish append access from general write access. */ if (flag & O_APPEND) acc_mode |= MAY_APPEND; - /* Fill in the open() intent data */ - nd->intent.open.flags = flag; - nd->intent.open.create_mode = mode; - /* * The simplest case - just a plain lookup. */ if (!(flag & O_CREAT)) { - error = path_lookup(pathname, lookup_flags(flag)|LOOKUP_OPEN, nd); + error = path_lookup_open(pathname, lookup_flags(flag), nd, flag); if (error) return error; goto ok; @@ -1445,7 +1522,7 @@ int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) /* * Create - we need to know the parent. */ - error = path_lookup(pathname, LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE, nd); + error = path_lookup_create(pathname, LOOKUP_PARENT, nd, flag, mode); if (error) return error; @@ -1520,6 +1597,8 @@ ok: exit_dput: dput_path(&path, nd); exit: + if (!IS_ERR(nd->intent.open.file)) + release_open_intent(nd); path_release(nd); return error; diff --git a/fs/open.c b/fs/open.c index f0d90cf0495..8d06ec911fd 100644 --- a/fs/open.c +++ b/fs/open.c @@ -739,7 +739,8 @@ asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group) } static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, - int flags, struct file *f) + int flags, struct file *f, + int (*open)(struct inode *, struct file *)) { struct inode *inode; int error; @@ -761,11 +762,14 @@ static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, f->f_op = fops_get(inode->i_fop); file_move(f, &inode->i_sb->s_files); - if (f->f_op && f->f_op->open) { - error = f->f_op->open(inode,f); + if (!open && f->f_op) + open = f->f_op->open; + if (open) { + error = open(inode, f); if (error) goto cleanup_all; } + f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); @@ -814,28 +818,75 @@ struct file *filp_open(const char * filename, int flags, int mode) { int namei_flags, error; struct nameidata nd; - struct file *f; namei_flags = flags; if ((namei_flags+1) & O_ACCMODE) namei_flags++; - if (namei_flags & O_TRUNC) - namei_flags |= 2; - - error = -ENFILE; - f = get_empty_filp(); - if (f == NULL) - return ERR_PTR(error); error = open_namei(filename, namei_flags, mode, &nd); if (!error) - return __dentry_open(nd.dentry, nd.mnt, flags, f); + return nameidata_to_filp(&nd, flags); - put_filp(f); return ERR_PTR(error); } EXPORT_SYMBOL(filp_open); +/** + * lookup_instantiate_filp - instantiates the open intent filp + * @nd: pointer to nameidata + * @dentry: pointer to dentry + * @open: open callback + * + * Helper for filesystems that want to use lookup open intents and pass back + * a fully instantiated struct file to the caller. + * This function is meant to be called from within a filesystem's + * lookup method. + * Note that in case of error, nd->intent.open.file is destroyed, but the + * path information remains valid. + * If the open callback is set to NULL, then the standard f_op->open() + * filesystem callback is substituted. + */ +struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry, + int (*open)(struct inode *, struct file *)) +{ + if (IS_ERR(nd->intent.open.file)) + goto out; + if (IS_ERR(dentry)) + goto out_err; + nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt), + nd->intent.open.flags - 1, + nd->intent.open.file, + open); +out: + return nd->intent.open.file; +out_err: + release_open_intent(nd); + nd->intent.open.file = (struct file *)dentry; + goto out; +} +EXPORT_SYMBOL_GPL(lookup_instantiate_filp); + +/** + * nameidata_to_filp - convert a nameidata to an open filp. + * @nd: pointer to nameidata + * @flags: open flags + * + * Note that this function destroys the original nameidata + */ +struct file *nameidata_to_filp(struct nameidata *nd, int flags) +{ + struct file *filp; + + /* Pick up the filp from the open intent */ + filp = nd->intent.open.file; + /* Has the filesystem initialised the file for us? */ + if (filp->f_dentry == NULL) + filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL); + else + path_release(nd); + return filp; +} + struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) { int error; @@ -846,7 +897,7 @@ struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) if (f == NULL) return ERR_PTR(error); - return __dentry_open(dentry, mnt, flags, f); + return __dentry_open(dentry, mnt, flags, f, NULL); } EXPORT_SYMBOL(dentry_open); -- cgit v1.2.3 From 02a913a73b52071e93f4b76db3e86138d19efffd Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:17 -0700 Subject: NFSv4: Eliminate nfsv4 open race... Make NFSv4 return the fully initialized file pointer with the stateid that it created in the lookup w/intent. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 29 ++++++------ fs/nfs/nfs3proc.c | 2 +- fs/nfs/nfs4_fs.h | 4 +- fs/nfs/nfs4proc.c | 137 ++++++++++++++++++++++-------------------------------- fs/nfs/proc.c | 2 +- 5 files changed, 73 insertions(+), 101 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index c70eabd6d17..a6e251f21fd 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -914,7 +914,6 @@ static int is_atomic_open(struct inode *dir, struct nameidata *nd) static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { struct dentry *res = NULL; - struct inode *inode = NULL; int error; /* Check that we are indeed trying to open this file */ @@ -928,8 +927,10 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry dentry->d_op = NFS_PROTO(dir)->dentry_ops; /* Let vfs_create() deal with O_EXCL */ - if (nd->intent.open.flags & O_EXCL) - goto no_entry; + if (nd->intent.open.flags & O_EXCL) { + d_add(dentry, NULL); + goto out; + } /* Open the file on the server */ lock_kernel(); @@ -943,18 +944,18 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry if (nd->intent.open.flags & O_CREAT) { nfs_begin_data_update(dir); - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); nfs_end_data_update(dir); } else - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); unlock_kernel(); - if (IS_ERR(inode)) { - error = PTR_ERR(inode); + if (IS_ERR(res)) { + error = PTR_ERR(res); switch (error) { /* Make a negative dentry */ case -ENOENT: - inode = NULL; - break; + res = NULL; + goto out; /* This turned out not to be a regular file */ case -ELOOP: if (!(nd->intent.open.flags & O_NOFOLLOW)) @@ -962,13 +963,9 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry /* case -EISDIR: */ /* case -EINVAL: */ default: - res = ERR_PTR(error); goto out; } - } -no_entry: - res = d_add_unique(dentry, inode); - if (res != NULL) + } else if (res != NULL) dentry = res; nfs_renew_times(dentry); nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); @@ -1012,7 +1009,7 @@ static int nfs_open_revalidate(struct dentry *dentry, struct nameidata *nd) */ lock_kernel(); verifier = nfs_save_change_attribute(dir); - ret = nfs4_open_revalidate(dir, dentry, openflags); + ret = nfs4_open_revalidate(dir, dentry, openflags, nd); if (!ret) nfs_set_verifier(dentry, verifier); unlock_kernel(); @@ -1135,7 +1132,7 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode, lock_kernel(); nfs_begin_data_update(dir); - error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags); + error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd); nfs_end_data_update(dir); if (error != 0) goto out_err; diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index edc95514046..df80477c5af 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -299,7 +299,7 @@ static int nfs3_proc_commit(struct nfs_write_data *cdata) */ static int nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs_fh fhandle; struct nfs_fattr fattr; diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 2215cdee43a..8a378819905 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -215,8 +215,8 @@ extern int nfs4_proc_setclientid_confirm(struct nfs4_client *); extern int nfs4_proc_async_renew(struct nfs4_client *); extern int nfs4_proc_renew(struct nfs4_client *); extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode); -extern struct inode *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); -extern int nfs4_open_revalidate(struct inode *, struct dentry *, int); +extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); +extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *); extern struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops; extern struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 35da15342e0..c9ecb811963 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -47,6 +47,7 @@ #include #include #include +#include #include "nfs4_fs.h" #include "delegation.h" @@ -947,12 +948,26 @@ out: return status; } -struct inode * +static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state) +{ + struct file *filp; + + filp = lookup_instantiate_filp(nd, dentry, NULL); + if (!IS_ERR(filp)) { + struct nfs_open_context *ctx; + ctx = (struct nfs_open_context *)filp->private_data; + ctx->state = state; + } else + nfs4_close_state(state, nd->intent.open.flags); +} + +struct dentry * nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { struct iattr attr; struct rpc_cred *cred; struct nfs4_state *state; + struct dentry *res; if (nd->flags & LOOKUP_CREATE) { attr.ia_mode = nd->intent.open.create_mode; @@ -966,16 +981,23 @@ nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0); if (IS_ERR(cred)) - return (struct inode *)cred; + return (struct dentry *)cred; state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred); put_rpccred(cred); - if (IS_ERR(state)) - return (struct inode *)state; - return state->inode; + if (IS_ERR(state)) { + if (PTR_ERR(state) == -ENOENT) + d_add(dentry, NULL); + return (struct dentry *)state; + } + res = d_add_unique(dentry, state->inode); + if (res != NULL) + dentry = res; + nfs4_intent_set_file(nd, dentry, state); + return res; } int -nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags) +nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd) { struct rpc_cred *cred; struct nfs4_state *state; @@ -988,18 +1010,30 @@ nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags) if (IS_ERR(state)) state = nfs4_do_open(dir, dentry, openflags, NULL, cred); put_rpccred(cred); - if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0) - return 1; - if (IS_ERR(state)) - return 0; + if (IS_ERR(state)) { + switch (PTR_ERR(state)) { + case -EPERM: + case -EACCES: + case -EDQUOT: + case -ENOSPC: + case -EROFS: + lookup_instantiate_filp(nd, (struct dentry *)state, NULL); + return 1; + case -ENOENT: + if (dentry->d_inode == NULL) + return 1; + } + goto out_drop; + } inode = state->inode; + iput(inode); if (inode == dentry->d_inode) { - iput(inode); + nfs4_intent_set_file(nd, dentry, state); return 1; } - d_drop(dentry); nfs4_close_state(state, openflags); - iput(inode); +out_drop: + d_drop(dentry); return 0; } @@ -1500,7 +1534,7 @@ static int nfs4_proc_commit(struct nfs_write_data *cdata) static int nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs4_state *state; struct rpc_cred *cred; @@ -1522,13 +1556,13 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, struct nfs_fattr fattr; status = nfs4_do_setattr(NFS_SERVER(dir), &fattr, NFS_FH(state->inode), sattr, state); - if (status == 0) { + if (status == 0) nfs_setattr_update_inode(state->inode, sattr); - goto out; - } - } else if (flags != 0) - goto out; - nfs4_close_state(state, flags); + } + if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN)) + nfs4_intent_set_file(nd, dentry, state); + else + nfs4_close_state(state, flags); out: return status; } @@ -2175,65 +2209,6 @@ nfs4_proc_renew(struct nfs4_client *clp) return 0; } -/* - * We will need to arrange for the VFS layer to provide an atomic open. - * Until then, this open method is prone to inefficiency and race conditions - * due to the lookup, potential create, and open VFS calls from sys_open() - * placed on the wire. - */ -static int -nfs4_proc_file_open(struct inode *inode, struct file *filp) -{ - struct dentry *dentry = filp->f_dentry; - struct nfs_open_context *ctx; - struct nfs4_state *state = NULL; - struct rpc_cred *cred; - int status = -ENOMEM; - - dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n", - (int)dentry->d_parent->d_name.len, - dentry->d_parent->d_name.name, - (int)dentry->d_name.len, dentry->d_name.name); - - - /* Find our open stateid */ - cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0); - if (IS_ERR(cred)) - return PTR_ERR(cred); - ctx = alloc_nfs_open_context(dentry, cred); - put_rpccred(cred); - if (unlikely(ctx == NULL)) - return -ENOMEM; - status = -EIO; /* ERACE actually */ - state = nfs4_find_state(inode, cred, filp->f_mode); - if (unlikely(state == NULL)) - goto no_state; - ctx->state = state; - nfs4_close_state(state, filp->f_mode); - ctx->mode = filp->f_mode; - nfs_file_set_open_context(filp, ctx); - put_nfs_open_context(ctx); - if (filp->f_mode & FMODE_WRITE) - nfs_begin_data_update(inode); - return 0; -no_state: - printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__); - put_nfs_open_context(ctx); - return status; -} - -/* - * Release our state - */ -static int -nfs4_proc_file_release(struct inode *inode, struct file *filp) -{ - if (filp->f_mode & FMODE_WRITE) - nfs_end_data_update(inode); - nfs_file_clear_open_context(filp); - return 0; -} - static inline int nfs4_server_supports_acls(struct nfs_server *server) { return (server->caps & NFS_CAP_ACLS) @@ -3145,8 +3120,8 @@ struct nfs_rpc_ops nfs_v4_clientops = { .read_setup = nfs4_proc_read_setup, .write_setup = nfs4_proc_write_setup, .commit_setup = nfs4_proc_commit_setup, - .file_open = nfs4_proc_file_open, - .file_release = nfs4_proc_file_release, + .file_open = nfs_open, + .file_release = nfs_release, .lock = nfs4_proc_lock, .clear_acl_cache = nfs4_zap_acl_attr, }; diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index be23c3fb926..8fef86523d7 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -216,7 +216,7 @@ static int nfs_proc_write(struct nfs_write_data *wdata) static int nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs_fh fhandle; struct nfs_fattr fattr; -- cgit v1.2.3 From 6f926b5ba75c568296ec227e7d782db4ddbdca5c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:18 -0700 Subject: [NFS]: Check that the server returns a valid regular file to our OPEN request Since it appears that some servers don't... Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 4 +++- fs/nfs/nfs4proc.c | 16 ++++++++++++++++ 2 files changed, 19 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index a6e251f21fd..ac331d2d4c4 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -957,10 +957,12 @@ static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry res = NULL; goto out; /* This turned out not to be a regular file */ + case -EISDIR: + case -ENOTDIR: + goto no_open; case -ELOOP: if (!(nd->intent.open.flags & O_NOFOLLOW)) goto no_open; - /* case -EISDIR: */ /* case -EINVAL: */ default: goto out; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index c9ecb811963..3db1c9f0b09 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -419,6 +419,22 @@ static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, stru o_arg->clientid = sp->so_client->cl_clientid; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); + if (status == 0) { + /* OPEN on anything except a regular file is disallowed in NFSv4 */ + switch (o_res->f_attr->mode & S_IFMT) { + case S_IFREG: + break; + case S_IFLNK: + status = -ELOOP; + break; + case S_IFDIR: + status = -EISDIR; + break; + default: + status = -ENOTDIR; + } + } + nfs_increment_open_seqid(status, o_arg->seqid); if (status != 0) goto out; -- cgit v1.2.3 From cdce5d6b94b6182f6d8a5b7b52923933e98cbc92 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:18 -0700 Subject: VFS: Make link_path_walk set LOOKUP_CONTINUE before calling permission(). This will allow nfs_permission() to perform additional optimizations when walking the path, by folding the ACCESS(MAY_EXEC) call on the directory into the lookup revalidation. Signed-off-by: Trond Myklebust --- fs/namei.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/namei.c b/fs/namei.c index 0d1dff7d3d9..aaaa8103623 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -763,6 +763,7 @@ static fastcall int __link_path_walk(const char * name, struct nameidata *nd) struct qstr this; unsigned int c; + nd->flags |= LOOKUP_CONTINUE; err = exec_permission_lite(inode, nd); if (err == -EAGAIN) { err = permission(inode, MAY_EXEC, nd); @@ -815,7 +816,6 @@ static fastcall int __link_path_walk(const char * name, struct nameidata *nd) if (err < 0) break; } - nd->flags |= LOOKUP_CONTINUE; /* This does the actual lookups.. */ err = do_lookup(nd, &this, &next); if (err) -- cgit v1.2.3 From cae7a073a4c5484cc5713eab606bf54b46724ab3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:19 -0700 Subject: NFSv4: Return delegation upon rename or removal of file. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 2 +- fs/nfs/delegation.h | 16 +++++++++++++++- fs/nfs/dir.c | 3 +++ fs/nfs/inode.c | 3 +-- 4 files changed, 20 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 4a36839f0bb..44135af9894 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -142,7 +142,7 @@ static void nfs_msync_inode(struct inode *inode) /* * Basic procedure for returning a delegation to the server */ -int nfs_inode_return_delegation(struct inode *inode) +int __nfs_inode_return_delegation(struct inode *inode) { struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state; struct nfs_inode *nfsi = NFS_I(inode); diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 3f6c45a29d6..8017846b561 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -25,7 +25,7 @@ struct nfs_delegation { int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); -int nfs_inode_return_delegation(struct inode *inode); +int __nfs_inode_return_delegation(struct inode *inode); int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid); struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle); @@ -47,11 +47,25 @@ static inline int nfs_have_delegation(struct inode *inode, int flags) return 1; return 0; } + +static inline int nfs_inode_return_delegation(struct inode *inode) +{ + int err = 0; + + if (NFS_I(inode)->delegation != NULL) + err = __nfs_inode_return_delegation(inode); + return err; +} #else static inline int nfs_have_delegation(struct inode *inode, int flags) { return 0; } + +static inline int nfs_inode_return_delegation(struct inode *inode) +{ + return 0; +} #endif #endif diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index ac331d2d4c4..72f50c0117b 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -801,6 +801,7 @@ static int nfs_dentry_delete(struct dentry *dentry) */ static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode) { + nfs_inode_return_delegation(inode); if (dentry->d_flags & DCACHE_NFSFS_RENAMED) { lock_kernel(); inode->i_nlink--; @@ -1329,6 +1330,7 @@ static int nfs_safe_remove(struct dentry *dentry) nfs_begin_data_update(dir); if (inode != NULL) { + nfs_inode_return_delegation(inode); nfs_begin_data_update(inode); error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); /* The VFS may want to delete this inode */ @@ -1547,6 +1549,7 @@ go_ahead: nfs_wb_all(old_inode); shrink_dcache_parent(old_dentry); } + nfs_inode_return_delegation(old_inode); if (new_inode) d_delete(new_dentry); diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index d7abaa00473..6b0f7fe6bd1 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1671,8 +1671,7 @@ static void nfs4_clear_inode(struct inode *inode) struct nfs_inode *nfsi = NFS_I(inode); /* If we are holding a delegation, return it! */ - if (nfsi->delegation != NULL) - nfs_inode_return_delegation(inode); + nfs_inode_return_delegation(inode); /* First call standard NFS clear_inode() code */ nfs_clear_inode(inode); /* Now clear out any remaining state */ -- cgit v1.2.3 From 642ac54923e0291ae2c8e42edde18d8c9c0a3c84 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:19 -0700 Subject: NFSv4: Return delegations in case we're changing ACLs Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 5 +++++ fs/nfs/nfs4proc.c | 1 + 2 files changed, 6 insertions(+) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 6b0f7fe6bd1..65d5ab45ddc 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -853,6 +853,11 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) filemap_fdatawait(inode->i_mapping); nfs_wb_all(inode); } + /* + * Return any delegations if we're going to change ACLs + */ + if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) + nfs_inode_return_delegation(inode); error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr); if (error == 0) nfs_refresh_inode(inode, &fattr); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3db1c9f0b09..c10dcd12af5 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2405,6 +2405,7 @@ static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen if (!nfs4_server_supports_acls(server)) return -EOPNOTSUPP; + nfs_inode_return_delegation(inode); buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase); ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0); if (ret == 0) -- cgit v1.2.3 From b8e5c4c2978fa666287525a9de2fa648a7581262 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:20 -0700 Subject: NFSv4: If a delegated open fails, ensure that we return the delegation Unless of course the open fails due to permission issues. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index c10dcd12af5..a8be9af610a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -658,6 +658,8 @@ out_err: } up_read(&nfsi->rwsem); up_read(&clp->cl_sem); + if (err != -EACCES) + nfs_inode_return_delegation(inode); return err; } -- cgit v1.2.3 From 550f57470c6506f5ef3c708335dea6bd48bf3dc4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:21 -0700 Subject: NFSv4: Ensure that we recover from the OPEN + OPEN_CONFIRM BAD_STATEID race If the server is in the unconfirmed OPEN state for a given open owner and receives a second OPEN for the same open owner, it will cancel the state of the first request and set up an OPEN_CONFIRM for the second. This can cause a race that is discussed in rfc3530 on page 181. The following patch allows the client to recover by retrying the original open request. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a8be9af610a..ff378126ccd 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -785,6 +785,16 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, exception.retry = 1; continue; } + /* + * BAD_STATEID on OPEN means that the server cancelled our + * state before it received the OPEN_CONFIRM. + * Recover by retrying the request as per the discussion + * on Page 181 of RFC3530. + */ + if (status == -NFS4ERR_BAD_STATEID) { + exception.retry = 1; + continue; + } res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir), status, &exception)); } while (exception.retry); -- cgit v1.2.3 From 4c780a4688b421baa896b59778c05d7e068e479f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:21 -0700 Subject: Fix Connectathon locking test failure We currently fail Connectathon test 6.10 in the case of 32-bit locks due to incorrect error checking. Also add support for l->l_len < 0 to 64-bit locks. Signed-off-by: Trond Myklebust --- fs/locks.c | 42 +++++++++++++++++++++++++----------------- 1 file changed, 25 insertions(+), 17 deletions(-) (limited to 'fs') diff --git a/fs/locks.c b/fs/locks.c index 7eb1d77b920..a1e8b224801 100644 --- a/fs/locks.c +++ b/fs/locks.c @@ -316,21 +316,22 @@ static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, /* POSIX-1996 leaves the case l->l_len < 0 undefined; POSIX-2001 defines it. */ start += l->l_start; - end = start + l->l_len - 1; - if (l->l_len < 0) { + if (start < 0) + return -EINVAL; + fl->fl_end = OFFSET_MAX; + if (l->l_len > 0) { + end = start + l->l_len - 1; + fl->fl_end = end; + } else if (l->l_len < 0) { end = start - 1; + fl->fl_end = end; start += l->l_len; + if (start < 0) + return -EINVAL; } - - if (start < 0) - return -EINVAL; - if (l->l_len > 0 && end < 0) - return -EOVERFLOW; - fl->fl_start = start; /* we record the absolute position */ - fl->fl_end = end; - if (l->l_len == 0) - fl->fl_end = OFFSET_MAX; + if (fl->fl_end < fl->fl_start) + return -EOVERFLOW; fl->fl_owner = current->files; fl->fl_pid = current->tgid; @@ -362,14 +363,21 @@ static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, return -EINVAL; } - if (((start += l->l_start) < 0) || (l->l_len < 0)) + start += l->l_start; + if (start < 0) return -EINVAL; - fl->fl_end = start + l->l_len - 1; - if (l->l_len > 0 && fl->fl_end < 0) - return -EOVERFLOW; + fl->fl_end = OFFSET_MAX; + if (l->l_len > 0) { + fl->fl_end = start + l->l_len - 1; + } else if (l->l_len < 0) { + fl->fl_end = start - 1; + start += l->l_len; + if (start < 0) + return -EINVAL; + } fl->fl_start = start; /* we record the absolute position */ - if (l->l_len == 0) - fl->fl_end = OFFSET_MAX; + if (fl->fl_end < fl->fl_start) + return -EOVERFLOW; fl->fl_owner = current->files; fl->fl_pid = current->tgid; -- cgit v1.2.3 From 6fe43f9e3701f7a9f2be151a5e6cfe94b87e92f9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 14:20:22 -0700 Subject: NFS: Fix rename of directory onto empty directory If someone tries to rename a directory onto an empty directory, we currently fail and return EBUSY. This patch ensures that we try the rename if both source and target are directories, and that we fail with a correct error of EISDIR if the source is not a directory. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 72f50c0117b..eb50c19fc25 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1511,9 +1511,11 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, */ if (!new_inode) goto go_ahead; - if (S_ISDIR(new_inode->i_mode)) - goto out; - else if (atomic_read(&new_dentry->d_count) > 2) { + if (S_ISDIR(new_inode->i_mode)) { + error = -EISDIR; + if (!S_ISDIR(old_inode->i_mode)) + goto out; + } else if (atomic_read(&new_dentry->d_count) > 2) { int err; /* copy the target dentry's name */ dentry = d_alloc(new_dentry->d_parent, -- cgit v1.2.3 From 7f709a48fa798cfa0f2f777c8752e12995054f78 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Oct 2005 23:19:39 -0700 Subject: NFSv4: Fix an oopsable condition in nfs_free_seqid Storing a pointer to the struct rpc_task in the nfs_seqid is broken since the nfs_seqid may be freed well after the task has been destroyed. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 1 - fs/nfs/nfs4state.c | 9 +-------- 2 files changed, 1 insertion(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 8a378819905..45bff1d1a51 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -112,7 +112,6 @@ struct nfs_seqid_counter { struct nfs_seqid { struct list_head list; struct nfs_seqid_counter *sequence; - struct rpc_task *task; }; static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status) diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 23834c8fb74..da0861db57f 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -676,7 +676,6 @@ struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) new = kmalloc(sizeof(*new), GFP_KERNEL); if (new != NULL) { new->sequence = counter; - new->task = NULL; spin_lock(&sequence->lock); list_add_tail(&new->list, &sequence->list); spin_unlock(&sequence->lock); @@ -687,15 +686,10 @@ struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) void nfs_free_seqid(struct nfs_seqid *seqid) { struct rpc_sequence *sequence = seqid->sequence->sequence; - struct rpc_task *next = NULL; spin_lock(&sequence->lock); list_del(&seqid->list); - if (!list_empty(&sequence->list)) { - next = list_entry(sequence->list.next, struct nfs_seqid, list)->task; - if (next) - rpc_wake_up_task(next); - } + rpc_wake_up(&sequence->wait); spin_unlock(&sequence->lock); kfree(seqid); } @@ -754,7 +748,6 @@ int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) spin_lock(&sequence->lock); if (sequence->list.next != &seqid->list) { - seqid->task = task; rpc_sleep_on(&sequence->wait, task, NULL, NULL); status = -EAGAIN; } -- cgit v1.2.3 From 8c233cf9c2ad6f49df753bdce84fddbf00bf6a75 Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Thu, 13 Oct 2005 16:54:27 -0400 Subject: NFSv4: handle no acl attr Stop handing garbage to userspace in the case where a weird server clears the acl bit in the getattr return (despite the fact that they've already claimed acl support.) Signed-off-by: J. Bruce Fields Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index c5c75235c5b..cd762648fa9 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -3255,7 +3255,8 @@ static int decode_getacl(struct xdr_stream *xdr, struct rpc_rqst *req, if (attrlen <= *acl_len) xdr_read_pages(xdr, attrlen); *acl_len = attrlen; - } + } else + status = -EOPNOTSUPP; out: return status; -- cgit v1.2.3 From 1d95db8e1688ed54e143a597c5570631a42fa594 Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Thu, 13 Oct 2005 16:54:32 -0400 Subject: NFSv4: Fix acl buffer size resp_len is passed in as buffer size to decode routine; make sure it's set right in case where userspace provides less than a page's worth of buffer. Signed-off-by: J. Bruce Fields Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index ff378126ccd..2d9357cf3fc 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2357,7 +2357,7 @@ static inline ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size return -ENOMEM; args.acl_pages[0] = localpage; args.acl_pgbase = 0; - args.acl_len = PAGE_SIZE; + resp_len = args.acl_len = PAGE_SIZE; } else { resp_buf = buf; buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase); -- cgit v1.2.3 From 4e51336a00bdcb42960acca52c23e90e9f4e6959 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Oct 2005 14:22:41 -0700 Subject: NFSv4: Final tweak to sequence id Sacrifice queueing fairness for performance. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 2 +- fs/nfs/nfs4state.c | 23 +++++++++++++---------- 2 files changed, 14 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 45bff1d1a51..5c0dd26d098 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -110,8 +110,8 @@ struct nfs_seqid_counter { }; struct nfs_seqid { - struct list_head list; struct nfs_seqid_counter *sequence; + struct list_head list; }; static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status) diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index da0861db57f..c59ef90e956 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -670,15 +670,12 @@ void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t f struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) { - struct rpc_sequence *sequence = counter->sequence; struct nfs_seqid *new; new = kmalloc(sizeof(*new), GFP_KERNEL); if (new != NULL) { new->sequence = counter; - spin_lock(&sequence->lock); - list_add_tail(&new->list, &sequence->list); - spin_unlock(&sequence->lock); + INIT_LIST_HEAD(&new->list); } return new; } @@ -687,10 +684,12 @@ void nfs_free_seqid(struct nfs_seqid *seqid) { struct rpc_sequence *sequence = seqid->sequence->sequence; - spin_lock(&sequence->lock); - list_del(&seqid->list); - rpc_wake_up(&sequence->wait); - spin_unlock(&sequence->lock); + if (!list_empty(&seqid->list)) { + spin_lock(&sequence->lock); + list_del(&seqid->list); + spin_unlock(&sequence->lock); + } + rpc_wake_up_next(&sequence->wait); kfree(seqid); } @@ -746,12 +745,16 @@ int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) struct rpc_sequence *sequence = seqid->sequence->sequence; int status = 0; + if (sequence->list.next == &seqid->list) + goto out; spin_lock(&sequence->lock); - if (sequence->list.next != &seqid->list) { + if (!list_empty(&sequence->list)) { rpc_sleep_on(&sequence->wait, task, NULL, NULL); status = -EAGAIN; - } + } else + list_add(&seqid->list, &sequence->list); spin_unlock(&sequence->lock); +out: return status; } -- cgit v1.2.3 From ec073428281b401f1142cb84b277a5b00c7994c9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Oct 2005 14:22:47 -0700 Subject: NFSv4: Fix up locking for nfs4_state_owner Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 1 + fs/nfs/nfs4proc.c | 2 ++ fs/nfs/nfs4state.c | 20 +++++++++++++++----- 3 files changed, 18 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 5c0dd26d098..78a53f5a9f1 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -126,6 +126,7 @@ static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status * semantics by allowing the server to identify replayed requests. */ struct nfs4_state_owner { + spinlock_t so_lock; struct list_head so_list; /* per-clientid list of state_owners */ struct nfs4_client *so_client; u32 so_id; /* 32-bit identifier, unique */ diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2d9357cf3fc..9c1da34036a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -212,6 +212,7 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, open_flags &= (FMODE_READ|FMODE_WRITE); /* Protect against nfs4_find_state() */ + spin_lock(&state->owner->so_lock); spin_lock(&inode->i_lock); state->state |= open_flags; /* NB! List reordering - see the reclaim code for why. */ @@ -221,6 +222,7 @@ static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, state->nreaders++; memcpy(&state->stateid, stateid, sizeof(state->stateid)); spin_unlock(&inode->i_lock); + spin_unlock(&state->owner->so_lock); } /* diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index c59ef90e956..2d5a6a2b9de 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -267,6 +267,7 @@ nfs4_alloc_state_owner(void) sp = kzalloc(sizeof(*sp),GFP_KERNEL); if (!sp) return NULL; + spin_lock_init(&sp->so_lock); INIT_LIST_HEAD(&sp->so_states); INIT_LIST_HEAD(&sp->so_delegations); rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); @@ -438,20 +439,23 @@ nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner) if (state) goto out; new = nfs4_alloc_open_state(); + spin_lock(&owner->so_lock); spin_lock(&inode->i_lock); state = __nfs4_find_state_byowner(inode, owner); if (state == NULL && new != NULL) { state = new; - /* Note: The reclaim code dictates that we add stateless - * and read-only stateids to the end of the list */ - list_add_tail(&state->open_states, &owner->so_states); state->owner = owner; atomic_inc(&owner->so_count); list_add(&state->inode_states, &nfsi->open_states); state->inode = igrab(inode); spin_unlock(&inode->i_lock); + /* Note: The reclaim code dictates that we add stateless + * and read-only stateids to the end of the list */ + list_add_tail(&state->open_states, &owner->so_states); + spin_unlock(&owner->so_lock); } else { spin_unlock(&inode->i_lock); + spin_unlock(&owner->so_lock); if (new) nfs4_free_open_state(new); } @@ -468,12 +472,14 @@ void nfs4_put_open_state(struct nfs4_state *state) struct inode *inode = state->inode; struct nfs4_state_owner *owner = state->owner; - if (!atomic_dec_and_lock(&state->count, &inode->i_lock)) + if (!atomic_dec_and_lock(&state->count, &owner->so_lock)) return; + spin_lock(&inode->i_lock); if (!list_empty(&state->inode_states)) list_del(&state->inode_states); - spin_unlock(&inode->i_lock); list_del(&state->open_states); + spin_unlock(&inode->i_lock); + spin_unlock(&owner->so_lock); iput(inode); BUG_ON (state->state != 0); nfs4_free_open_state(state); @@ -491,6 +497,7 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) atomic_inc(&owner->so_count); /* Protect against nfs4_find_state() */ + spin_lock(&owner->so_lock); spin_lock(&inode->i_lock); if (mode & FMODE_READ) state->nreaders--; @@ -503,6 +510,7 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode) list_move_tail(&state->open_states, &owner->so_states); } spin_unlock(&inode->i_lock); + spin_unlock(&owner->so_lock); newstate = 0; if (state->state != 0) { if (state->nreaders) @@ -899,6 +907,7 @@ static void nfs4_state_mark_reclaim(struct nfs4_client *clp) list_for_each_entry(sp, &clp->cl_state_owners, so_list) { sp->so_seqid.counter = 0; sp->so_seqid.flags = 0; + spin_lock(&sp->so_lock); list_for_each_entry(state, &sp->so_states, open_states) { list_for_each_entry(lock, &state->lock_states, ls_locks) { lock->ls_seqid.counter = 0; @@ -906,6 +915,7 @@ static void nfs4_state_mark_reclaim(struct nfs4_client *clp) lock->ls_flags &= ~NFS_LOCK_INITIALIZED; } } + spin_unlock(&sp->so_lock); } } -- cgit v1.2.3 From 94c1d3184523efa7109472eb393cee6e954c5d75 Mon Sep 17 00:00:00 2001 From: Peter Wainwright Date: Wed, 26 Oct 2005 01:59:02 -0700 Subject: [PATCH] Fix HFS+ to free up the space when a file is deleted. fsck_hfs reveals lots of temporary files accumulating in the hidden directory "\000\000\000HFS+ Private Data". According to the HFS+ documentation these are files which are unlinked while in use. However, there may be a bug in the Linux hfsplus implementation which causes this to happen even when the files are not in use. It looks like the "opencnt" field is never initialized as (I think) it should be in hfsplus_read_inode. This means that a file can appear to be still in use when in fact it has been closed. This patch seems to fix it for me. Signed-off-by: Anton Altaparmakov Cc: Roman Zippel Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hfsplus/super.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs') diff --git a/fs/hfsplus/super.c b/fs/hfsplus/super.c index fd0f0f050e1..452fc1fdbd3 100644 --- a/fs/hfsplus/super.c +++ b/fs/hfsplus/super.c @@ -50,6 +50,7 @@ static void hfsplus_read_inode(struct inode *inode) init_MUTEX(&HFSPLUS_I(inode).extents_lock); HFSPLUS_I(inode).flags = 0; HFSPLUS_I(inode).rsrc_inode = NULL; + atomic_set(&HFSPLUS_I(inode).opencnt, 0); if (inode->i_ino >= HFSPLUS_FIRSTUSER_CNID) { read_inode: -- cgit v1.2.3 From 34123da66e613602de5a886b05c875b6a91b8ed2 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 19:10:09 -0400 Subject: NFS: Fix a bad cast in nfs3_read_done Signed-off-by: Trond Myklebust --- fs/nfs/nfs3proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index df80477c5af..e4a1cd48195 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -735,7 +735,7 @@ extern u32 *nfs3_decode_dirent(u32 *, struct nfs_entry *, int); static void nfs3_read_done(struct rpc_task *task) { - struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; + struct nfs_read_data *data = (struct nfs_read_data *) task->tk_calldata; if (nfs3_async_handle_jukebox(task)) return; -- cgit v1.2.3 From 0e574af1be5f569a5d7f2800333b0bfb358a5e34 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:38 -0400 Subject: NFS: Cleanup initialisation of struct nfs_fattr Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 1 + fs/nfs/nfs3proc.c | 58 +++++++++++++++++++++++++++---------------------------- fs/nfs/nfs4proc.c | 43 +++++++++++++++++++---------------------- fs/nfs/proc.c | 26 ++++++++++++------------- fs/nfs/read.c | 1 + fs/nfs/write.c | 2 ++ 6 files changed, 66 insertions(+), 65 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index eb50c19fc25..b8a73045e9a 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -532,6 +532,7 @@ static int nfs_readdir(struct file *filp, void *dirent, filldir_t filldir) my_entry.eof = 0; my_entry.fh = &fh; my_entry.fattr = &fattr; + nfs_fattr_init(&fattr); desc->entry = &my_entry; while(!desc->entry->eof) { diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index e4a1cd48195..4b1b48b139f 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -78,7 +78,7 @@ nfs3_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("%s: call fsinfo\n", __FUNCTION__); - info->fattr->valid = 0; + nfs_fattr_init(info->fattr); status = rpc_call(server->client_sys, NFS3PROC_FSINFO, fhandle, info, 0); dprintk("%s: reply fsinfo: %d\n", __FUNCTION__, status); if (!(info->fattr->valid & NFS_ATTR_FATTR)) { @@ -98,7 +98,7 @@ nfs3_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call getattr\n"); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(server->client, NFS3PROC_GETATTR, fhandle, fattr, 0); dprintk("NFS reply getattr: %d\n", status); @@ -117,7 +117,7 @@ nfs3_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr, int status; dprintk("NFS call setattr\n"); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(inode), NFS3PROC_SETATTR, &arg, fattr, 0); if (status == 0) nfs_setattr_update_inode(inode, sattr); @@ -143,8 +143,8 @@ nfs3_proc_lookup(struct inode *dir, struct qstr *name, int status; dprintk("NFS call lookup %s\n", name->name); - dir_attr.valid = 0; - fattr->valid = 0; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_LOOKUP, &arg, &res, 0); if (status >= 0 && !(fattr->valid & NFS_ATTR_FATTR)) status = rpc_call(NFS_CLIENT(dir), NFS3PROC_GETATTR, @@ -174,7 +174,6 @@ static int nfs3_proc_access(struct inode *inode, struct nfs_access_entry *entry) int status; dprintk("NFS call access\n"); - fattr.valid = 0; if (mode & MAY_READ) arg.access |= NFS3_ACCESS_READ; @@ -189,6 +188,7 @@ static int nfs3_proc_access(struct inode *inode, struct nfs_access_entry *entry) if (mode & MAY_EXEC) arg.access |= NFS3_ACCESS_EXECUTE; } + nfs_fattr_init(&fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0); nfs_refresh_inode(inode, &fattr); if (status == 0) { @@ -217,7 +217,7 @@ static int nfs3_proc_readlink(struct inode *inode, struct page *page, int status; dprintk("NFS call readlink\n"); - fattr.valid = 0; + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(inode), NFS3PROC_READLINK, &args, &fattr, 0); nfs_refresh_inode(inode, &fattr); @@ -240,7 +240,7 @@ static int nfs3_proc_read(struct nfs_read_data *rdata) dprintk("NFS call read %d @ %Ld\n", rdata->args.count, (long long) rdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, flags); if (status >= 0) nfs_refresh_inode(inode, fattr); @@ -263,7 +263,7 @@ static int nfs3_proc_write(struct nfs_write_data *wdata) dprintk("NFS call write %d @ %Ld\n", wdata->args.count, (long long) wdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, rpcflags); if (status >= 0) nfs_refresh_inode(inode, fattr); @@ -285,7 +285,7 @@ static int nfs3_proc_commit(struct nfs_write_data *cdata) dprintk("NFS call commit %d @ %Ld\n", cdata->args.count, (long long) cdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0); if (status >= 0) nfs_refresh_inode(inode, fattr); @@ -329,8 +329,8 @@ nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, sattr->ia_mode &= ~current->fs->umask; again: - dir_attr.valid = 0; - fattr.valid = 0; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_CREATE, &arg, &res, 0); nfs_refresh_inode(dir, &dir_attr); @@ -401,7 +401,7 @@ nfs3_proc_remove(struct inode *dir, struct qstr *name) int status; dprintk("NFS call remove %s\n", name->name); - dir_attr.valid = 0; + nfs_fattr_init(&dir_attr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); nfs_refresh_inode(dir, &dir_attr); dprintk("NFS reply remove: %d\n", status); @@ -422,7 +422,7 @@ nfs3_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir, struct qstr ptr->arg.fh = NFS_FH(dir->d_inode); ptr->arg.name = name->name; ptr->arg.len = name->len; - ptr->res.valid = 0; + nfs_fattr_init(&ptr->res); msg->rpc_proc = &nfs3_procedures[NFS3PROC_REMOVE]; msg->rpc_argp = &ptr->arg; msg->rpc_resp = &ptr->res; @@ -465,8 +465,8 @@ nfs3_proc_rename(struct inode *old_dir, struct qstr *old_name, int status; dprintk("NFS call rename %s -> %s\n", old_name->name, new_name->name); - old_dir_attr.valid = 0; - new_dir_attr.valid = 0; + nfs_fattr_init(&old_dir_attr); + nfs_fattr_init(&new_dir_attr); status = rpc_call(NFS_CLIENT(old_dir), NFS3PROC_RENAME, &arg, &res, 0); nfs_refresh_inode(old_dir, &old_dir_attr); nfs_refresh_inode(new_dir, &new_dir_attr); @@ -491,8 +491,8 @@ nfs3_proc_link(struct inode *inode, struct inode *dir, struct qstr *name) int status; dprintk("NFS call link %s\n", name->name); - dir_attr.valid = 0; - fattr.valid = 0; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(inode), NFS3PROC_LINK, &arg, &res, 0); nfs_refresh_inode(dir, &dir_attr); nfs_refresh_inode(inode, &fattr); @@ -524,8 +524,8 @@ nfs3_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path, if (path->len > NFS3_MAXPATHLEN) return -ENAMETOOLONG; dprintk("NFS call symlink %s -> %s\n", name->name, path->name); - dir_attr.valid = 0; - fattr->valid = 0; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_SYMLINK, &arg, &res, 0); nfs_refresh_inode(dir, &dir_attr); dprintk("NFS reply symlink: %d\n", status); @@ -552,11 +552,11 @@ nfs3_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr) int status; dprintk("NFS call mkdir %s\n", dentry->d_name.name); - dir_attr.valid = 0; - fattr.valid = 0; sattr->ia_mode &= ~current->fs->umask; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKDIR, &arg, &res, 0); nfs_refresh_inode(dir, &dir_attr); if (status != 0) @@ -582,7 +582,7 @@ nfs3_proc_rmdir(struct inode *dir, struct qstr *name) int status; dprintk("NFS call rmdir %s\n", name->name); - dir_attr.valid = 0; + nfs_fattr_init(&dir_attr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_RMDIR, &arg, &dir_attr, 0); nfs_refresh_inode(dir, &dir_attr); dprintk("NFS reply rmdir: %d\n", status); @@ -634,7 +634,7 @@ nfs3_proc_readdir(struct dentry *dentry, struct rpc_cred *cred, dprintk("NFS call readdir%s %d\n", plus? "plus" : "", (unsigned int) cookie); - dir_attr.valid = 0; + nfs_fattr_init(&dir_attr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); nfs_refresh_inode(dir, &dir_attr); dprintk("NFS reply readdir: %d\n", status); @@ -676,8 +676,8 @@ nfs3_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr, sattr->ia_mode &= ~current->fs->umask; - dir_attr.valid = 0; - fattr.valid = 0; + nfs_fattr_init(&dir_attr); + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKNOD, &arg, &res, 0); nfs_refresh_inode(dir, &dir_attr); if (status != 0) @@ -698,7 +698,7 @@ nfs3_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call fsstat\n"); - stat->fattr->valid = 0; + nfs_fattr_init(stat->fattr); status = rpc_call(server->client, NFS3PROC_FSSTAT, fhandle, stat, 0); dprintk("NFS reply statfs: %d\n", status); return status; @@ -711,7 +711,7 @@ nfs3_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call fsinfo\n"); - info->fattr->valid = 0; + nfs_fattr_init(info->fattr); status = rpc_call(server->client_sys, NFS3PROC_FSINFO, fhandle, info, 0); dprintk("NFS reply fsinfo: %d\n", status); return status; @@ -724,7 +724,7 @@ nfs3_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call pathconf\n"); - info->fattr->valid = 0; + nfs_fattr_init(info->fattr); status = rpc_call(server->client, NFS3PROC_PATHCONF, fhandle, info, 0); dprintk("NFS reply pathconf: %d\n", status); return status; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9c1da34036a..2a759e8e387 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -494,9 +494,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(dir); struct nfs_delegation *delegation = NFS_I(inode)->delegation; - struct nfs_fattr f_attr = { - .valid = 0, - }; + struct nfs_fattr f_attr; struct nfs_openargs o_arg = { .fh = NFS_FH(dir), .open_flags = state->state, @@ -522,6 +520,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st status = -ENOMEM; if (o_arg.seqid == NULL) goto out; + nfs_fattr_init(&f_attr); status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_nodeleg; @@ -692,9 +691,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st struct nfs4_client *clp = server->nfs4_state; struct inode *inode = NULL; int status; - struct nfs_fattr f_attr = { - .valid = 0, - }; + struct nfs_fattr f_attr; struct nfs_openargs o_arg = { .fh = NFS_FH(dir), .open_flags = flags, @@ -726,6 +723,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); if (o_arg.seqid == NULL) return -ENOMEM; + nfs_fattr_init(&f_attr); status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_err; @@ -824,7 +822,7 @@ static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr, }; int status; - fattr->valid = 0; + nfs_fattr_init(fattr); if (state != NULL) { msg.rpc_cred = state->owner->so_cred; @@ -1107,13 +1105,12 @@ static int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fh static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *info) { - struct nfs_fattr * fattr = info->fattr; struct nfs4_lookup_root_arg args = { .bitmask = nfs4_fattr_bitmap, }; struct nfs4_lookup_res res = { .server = server, - .fattr = fattr, + .fattr = info->fattr, .fh = fhandle, }; struct rpc_message msg = { @@ -1121,7 +1118,7 @@ static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle, .rpc_argp = &args, .rpc_resp = &res, }; - fattr->valid = 0; + nfs_fattr_init(info->fattr); return rpc_call_sync(server->client, &msg, 0); } @@ -1184,7 +1181,7 @@ static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, q.len = p - q.name; do { - fattr->valid = 0; + nfs_fattr_init(fattr); status = nfs4_handle_exception(server, rpc_call_sync(server->client, &msg, 0), &exception); @@ -1221,7 +1218,7 @@ static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, .rpc_resp = &res, }; - fattr->valid = 0; + nfs_fattr_init(fattr); return rpc_call_sync(server->client, &msg, 0); } @@ -1263,7 +1260,7 @@ nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr, struct nfs4_state *state; int status; - fattr->valid = 0; + nfs_fattr_init(fattr); cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0); if (IS_ERR(cred)) @@ -1309,7 +1306,7 @@ static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name, .rpc_resp = &res, }; - fattr->valid = 0; + nfs_fattr_init(fattr); dprintk("NFS call lookup %s\n", name->name); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); @@ -1458,7 +1455,7 @@ static int _nfs4_proc_read(struct nfs_read_data *rdata) dprintk("NFS call read %d @ %Ld\n", rdata->args.count, (long long) rdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, flags); if (!status) renew_lease(server, timestamp); @@ -1495,7 +1492,7 @@ static int _nfs4_proc_write(struct nfs_write_data *wdata) dprintk("NFS call write %d @ %Ld\n", wdata->args.count, (long long) wdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, rpcflags); dprintk("NFS reply write: %d\n", status); return status; @@ -1529,7 +1526,7 @@ static int _nfs4_proc_commit(struct nfs_write_data *cdata) dprintk("NFS call commit %d @ %Ld\n", cdata->args.count, (long long) cdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(server->client, &msg, 0); dprintk("NFS reply commit: %d\n", status); return status; @@ -1769,7 +1766,7 @@ static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name, if (path->len > NFS4_MAXPATHLEN) return -ENAMETOOLONG; arg.u.symlink = path; - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); if (!status) @@ -1818,7 +1815,7 @@ static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry, }; int status; - fattr.valid = 0; + nfs_fattr_init(&fattr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); if (!status) { @@ -1916,7 +1913,7 @@ static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry, int status; int mode = sattr->ia_mode; - fattr.valid = 0; + nfs_fattr_init(&fattr); BUG_ON(!(sattr->ia_valid & ATTR_MODE)); BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode)); @@ -1969,7 +1966,7 @@ static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, .rpc_resp = fsstat, }; - fsstat->fattr->valid = 0; + nfs_fattr_init(fsstat->fattr); return rpc_call_sync(server->client, &msg, 0); } @@ -2016,7 +2013,7 @@ static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, str static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo) { - fsinfo->fattr->valid = 0; + nfs_fattr_init(fsinfo->fattr); return nfs4_do_fsinfo(server, fhandle, fsinfo); } @@ -2039,7 +2036,7 @@ static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle return 0; } - pathconf->fattr->valid = 0; + nfs_fattr_init(pathconf->fattr); return rpc_call_sync(server->client, &msg, 0); } diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index 8fef86523d7..5ef28f08f42 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -61,7 +61,7 @@ nfs_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("%s: call getattr\n", __FUNCTION__); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(server->client_sys, NFSPROC_GETATTR, fhandle, fattr, 0); dprintk("%s: reply getattr: %d\n", __FUNCTION__, status); if (status) @@ -93,7 +93,7 @@ nfs_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call getattr\n"); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(server->client, NFSPROC_GETATTR, fhandle, fattr, 0); dprintk("NFS reply getattr: %d\n", status); @@ -112,7 +112,7 @@ nfs_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr, int status; dprintk("NFS call setattr\n"); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(inode), NFSPROC_SETATTR, &arg, fattr, 0); if (status == 0) nfs_setattr_update_inode(inode, sattr); @@ -136,7 +136,7 @@ nfs_proc_lookup(struct inode *dir, struct qstr *name, int status; dprintk("NFS call lookup %s\n", name->name); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_LOOKUP, &arg, &res, 0); dprintk("NFS reply lookup: %d\n", status); return status; @@ -174,7 +174,7 @@ static int nfs_proc_read(struct nfs_read_data *rdata) dprintk("NFS call read %d @ %Ld\n", rdata->args.count, (long long) rdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, flags); if (status >= 0) { nfs_refresh_inode(inode, fattr); @@ -203,7 +203,7 @@ static int nfs_proc_write(struct nfs_write_data *wdata) dprintk("NFS call write %d @ %Ld\n", wdata->args.count, (long long) wdata->args.offset); - fattr->valid = 0; + nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, flags); if (status >= 0) { nfs_refresh_inode(inode, fattr); @@ -232,7 +232,7 @@ nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, }; int status; - fattr.valid = 0; + nfs_fattr_init(&fattr); dprintk("NFS call create %s\n", dentry->d_name.name); status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0); if (status == 0) @@ -273,12 +273,12 @@ nfs_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr, sattr->ia_size = new_encode_dev(rdev);/* get out your barf bag */ } - fattr.valid = 0; + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0); if (status == -EINVAL && S_ISFIFO(mode)) { sattr->ia_mode = mode; - fattr.valid = 0; + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0); } if (status == 0) @@ -391,7 +391,7 @@ nfs_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path, if (path->len > NFS2_MAXPATHLEN) return -ENAMETOOLONG; dprintk("NFS call symlink %s -> %s\n", name->name, path->name); - fattr->valid = 0; + nfs_fattr_init(fattr); fhandle->size = 0; status = rpc_call(NFS_CLIENT(dir), NFSPROC_SYMLINK, &arg, NULL, 0); dprintk("NFS reply symlink: %d\n", status); @@ -416,7 +416,7 @@ nfs_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr) int status; dprintk("NFS call mkdir %s\n", dentry->d_name.name); - fattr.valid = 0; + nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_MKDIR, &arg, &res, 0); if (status == 0) status = nfs_instantiate(dentry, &fhandle, &fattr); @@ -484,7 +484,7 @@ nfs_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call statfs\n"); - stat->fattr->valid = 0; + nfs_fattr_init(stat->fattr); status = rpc_call(server->client, NFSPROC_STATFS, fhandle, &fsinfo, 0); dprintk("NFS reply statfs: %d\n", status); if (status) @@ -507,7 +507,7 @@ nfs_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, int status; dprintk("NFS call fsinfo\n"); - info->fattr->valid = 0; + nfs_fattr_init(info->fattr); status = rpc_call(server->client, NFSPROC_STATFS, fhandle, &fsinfo, 0); dprintk("NFS reply fsinfo: %d\n", status); if (status) diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 9758ebd4990..43b03b19731 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -215,6 +215,7 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, data->res.fattr = &data->fattr; data->res.count = count; data->res.eof = 0; + nfs_fattr_init(&data->fattr); NFS_PROTO(inode)->read_setup(data); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 5130eda231d..819a65f5071 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -870,6 +870,7 @@ static void nfs_write_rpcsetup(struct nfs_page *req, data->res.fattr = &data->fattr; data->res.count = count; data->res.verf = &data->verf; + nfs_fattr_init(&data->fattr); NFS_PROTO(inode)->write_setup(data, how); @@ -1237,6 +1238,7 @@ static void nfs_commit_rpcsetup(struct list_head *head, data->res.count = 0; data->res.fattr = &data->fattr; data->res.verf = &data->verf; + nfs_fattr_init(&data->fattr); NFS_PROTO(inode)->commit_setup(data, how); -- cgit v1.2.3 From 913a70fc170530f7e1ff0693595155457cc6d0ca Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:38 -0400 Subject: NFS: Convert cache_change_attribute into a jiffy-based value Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 65d5ab45ddc..449df8c8aa3 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1135,7 +1135,7 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) * We may need to keep the attributes marked as invalid if * we raced with nfs_end_attr_update(). */ - if (verifier == nfsi->cache_change_attribute) + if (time_after_eq(verifier, nfsi->cache_change_attribute)) nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); spin_unlock(&inode->i_lock); @@ -1202,7 +1202,7 @@ void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping) if (S_ISDIR(inode->i_mode)) { memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf)); /* This ensures we revalidate child dentries */ - nfsi->cache_change_attribute++; + nfsi->cache_change_attribute = jiffies; } spin_unlock(&inode->i_lock); @@ -1242,7 +1242,7 @@ void nfs_end_data_update(struct inode *inode) nfsi->cache_validity |= NFS_INO_INVALID_DATA; spin_unlock(&inode->i_lock); } - nfsi->cache_change_attribute ++; + nfsi->cache_change_attribute = jiffies; atomic_dec(&nfsi->data_updates); } @@ -1391,7 +1391,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign /* Do we perhaps have any outstanding writes? */ if (nfsi->npages == 0) { /* No, but did we race with nfs_end_data_update()? */ - if (verifier == nfsi->cache_change_attribute) { + if (time_after_eq(verifier, nfsi->cache_change_attribute)) { inode->i_size = new_isize; invalid |= NFS_INO_INVALID_DATA; } -- cgit v1.2.3 From 33801147a8fda6b04d7e9afe1d42f1c01d3d6837 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:39 -0400 Subject: NFS: Optimise inode attribute cache updates Allow nfs_refresh_inode() also to update attributes on the inode if the RPC call was sent after the last call to nfs_update_inode(). Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 54 ++++++++++++++++++++++++++++++++++++++++-------------- fs/nfs/nfs2xdr.c | 1 - fs/nfs/nfs3xdr.c | 1 - fs/nfs/nfs4xdr.c | 4 +--- 4 files changed, 41 insertions(+), 19 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 449df8c8aa3..b7d4f8f13ac 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -785,7 +785,8 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr) else init_special_inode(inode, inode->i_mode, fattr->rdev); - nfsi->read_cache_jiffies = fattr->timestamp; + nfsi->read_cache_jiffies = fattr->time_start; + nfsi->last_updated = jiffies; inode->i_atime = fattr->atime; inode->i_mtime = fattr->mtime; inode->i_ctime = fattr->ctime; @@ -1120,14 +1121,15 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode) goto out; } + spin_lock(&inode->i_lock); status = nfs_update_inode(inode, &fattr, verifier); if (status) { + spin_unlock(&inode->i_lock); dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n", inode->i_sb->s_id, (long long)NFS_FILEID(inode), status); goto out; } - spin_lock(&inode->i_lock); cache_validity = nfsi->cache_validity; nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE; @@ -1247,7 +1249,7 @@ void nfs_end_data_update(struct inode *inode) } /** - * nfs_refresh_inode - verify consistency of the inode attribute cache + * nfs_check_inode_attributes - verify consistency of the inode attribute cache * @inode - pointer to inode * @fattr - updated attributes * @@ -1255,13 +1257,12 @@ void nfs_end_data_update(struct inode *inode) * so that fattr carries weak cache consistency data, then it may * also update the ctime/mtime/change_attribute. */ -int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr) +static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr) { struct nfs_inode *nfsi = NFS_I(inode); loff_t cur_size, new_isize; int data_unstable; - spin_lock(&inode->i_lock); /* Are we in the process of updating data on the server? */ data_unstable = nfs_caches_unstable(inode); @@ -1325,11 +1326,40 @@ int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr) if (!timespec_equal(&inode->i_atime, &fattr->atime)) nfsi->cache_validity |= NFS_INO_INVALID_ATIME; - nfsi->read_cache_jiffies = fattr->timestamp; - spin_unlock(&inode->i_lock); + nfsi->read_cache_jiffies = fattr->time_start; return 0; } +/** + * nfs_refresh_inode - try to update the inode attribute cache + * @inode - pointer to inode + * @fattr - updated attributes + * + * Check that an RPC call that returned attributes has not overlapped with + * other recent updates of the inode metadata, then decide whether it is + * safe to do a full update of the inode attributes, or whether just to + * call nfs_check_inode_attributes. + */ +int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr) +{ + struct nfs_inode *nfsi = NFS_I(inode); + int status; + + if ((fattr->valid & NFS_ATTR_FATTR) == 0) + return 0; + spin_lock(&inode->i_lock); + nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE; + if (nfs_verify_change_attribute(inode, fattr->time_start)) + nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME); + if (time_after(fattr->time_start, nfsi->last_updated)) + status = nfs_update_inode(inode, fattr, fattr->time_start); + else + status = nfs_check_inode_attributes(inode, fattr); + + spin_unlock(&inode->i_lock); + return status; +} + /* * Many nfs protocol calls return the new file attributes after * an operation. Here we update the inode to reflect the state @@ -1365,20 +1395,17 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign goto out_err; } - spin_lock(&inode->i_lock); - /* * Make sure the inode's type hasn't changed. */ - if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) { - spin_unlock(&inode->i_lock); + if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) goto out_changed; - } /* * Update the read time so we don't revalidate too often. */ - nfsi->read_cache_jiffies = fattr->timestamp; + nfsi->read_cache_jiffies = fattr->time_start; + nfsi->last_updated = jiffies; /* Are we racing with known updates of the metadata on the server? */ data_unstable = ! (nfs_verify_change_attribute(inode, verifier) || @@ -1467,7 +1494,6 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsign if (!nfs_have_delegation(inode, FMODE_READ)) nfsi->cache_validity |= invalid; - spin_unlock(&inode->i_lock); return 0; out_changed: /* diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c index d91b69044a4..59049e864ca 100644 --- a/fs/nfs/nfs2xdr.c +++ b/fs/nfs/nfs2xdr.c @@ -143,7 +143,6 @@ xdr_decode_fattr(u32 *p, struct nfs_fattr *fattr) fattr->mode = (fattr->mode & ~S_IFMT) | S_IFIFO; fattr->rdev = 0; } - fattr->timestamp = jiffies; return p; } diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index db4a904810a..0498bd36602 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -174,7 +174,6 @@ xdr_decode_fattr(u32 *p, struct nfs_fattr *fattr) /* Update the mode bits */ fattr->valid |= (NFS_ATTR_FATTR | NFS_ATTR_FATTR_V3); - fattr->timestamp = jiffies; return p; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index cd762648fa9..8b21de8a06f 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -2799,10 +2799,8 @@ static int decode_getfattr(struct xdr_stream *xdr, struct nfs_fattr *fattr, cons goto xdr_error; if ((status = decode_attr_time_modify(xdr, bitmap, &fattr->mtime)) != 0) goto xdr_error; - if ((status = verify_attr_len(xdr, savep, attrlen)) == 0) { + if ((status = verify_attr_len(xdr, savep, attrlen)) == 0) fattr->valid = NFS_ATTR_FATTR | NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4; - fattr->timestamp = jiffies; - } xdr_error: if (status != 0) printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); -- cgit v1.2.3 From decf491f3076190262d4c649bed877650623903a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:39 -0400 Subject: NFS: Don't let nfs_end_data_update() clobber attribute update information Since we almost always call nfs_end_data_update() after we called nfs_refresh_inode(), we now end up marking the inode metadata as needing revalidation immediately after having updated it. This patch rearranges things so that we mark the inode as needing revalidation _before_ we call nfs_refresh_inode() on those operations that need it. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 38 ++++++++++++++++++++++++++++++++------ fs/nfs/nfs3proc.c | 30 +++++++++++++++--------------- fs/nfs/nfs4proc.c | 3 +++ fs/nfs/proc.c | 16 +++++++++++++--- 4 files changed, 63 insertions(+), 24 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index b7d4f8f13ac..6b3156e1535 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1236,13 +1236,12 @@ void nfs_end_data_update(struct inode *inode) struct nfs_inode *nfsi = NFS_I(inode); if (!nfs_have_delegation(inode, FMODE_READ)) { - /* Mark the attribute cache for revalidation */ - spin_lock(&inode->i_lock); - nfsi->cache_validity |= NFS_INO_INVALID_ATTR; - /* Directories and symlinks: invalidate page cache too */ - if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) + /* Directories and symlinks: invalidate page cache */ + if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) { + spin_lock(&inode->i_lock); nfsi->cache_validity |= NFS_INO_INVALID_DATA; - spin_unlock(&inode->i_lock); + spin_unlock(&inode->i_lock); + } } nfsi->cache_change_attribute = jiffies; atomic_dec(&nfsi->data_updates); @@ -1360,6 +1359,33 @@ int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr) return status; } +/** + * nfs_post_op_update_inode - try to update the inode attribute cache + * @inode - pointer to inode + * @fattr - updated attributes + * + * After an operation that has changed the inode metadata, mark the + * attribute cache as being invalid, then try to update it. + */ +int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr) +{ + struct nfs_inode *nfsi = NFS_I(inode); + int status = 0; + + spin_lock(&inode->i_lock); + if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) { + nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS; + goto out; + } + status = nfs_update_inode(inode, fattr, fattr->time_start); + if (time_after_eq(fattr->time_start, nfsi->cache_change_attribute)) + nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE); + nfsi->cache_change_attribute = jiffies; +out: + spin_unlock(&inode->i_lock); + return status; +} + /* * Many nfs protocol calls return the new file attributes after * an operation. Here we update the inode to reflect the state diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c index 4b1b48b139f..92c870d19cc 100644 --- a/fs/nfs/nfs3proc.c +++ b/fs/nfs/nfs3proc.c @@ -266,7 +266,7 @@ static int nfs3_proc_write(struct nfs_write_data *wdata) nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, rpcflags); if (status >= 0) - nfs_refresh_inode(inode, fattr); + nfs_post_op_update_inode(inode, fattr); dprintk("NFS reply write: %d\n", status); return status < 0? status : wdata->res.count; } @@ -288,7 +288,7 @@ static int nfs3_proc_commit(struct nfs_write_data *cdata) nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0); if (status >= 0) - nfs_refresh_inode(inode, fattr); + nfs_post_op_update_inode(inode, fattr); dprintk("NFS reply commit: %d\n", status); return status; } @@ -332,7 +332,7 @@ again: nfs_fattr_init(&dir_attr); nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_CREATE, &arg, &res, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); /* If the server doesn't support the exclusive creation semantics, * try again with simple 'guarded' mode. */ @@ -403,7 +403,7 @@ nfs3_proc_remove(struct inode *dir, struct qstr *name) dprintk("NFS call remove %s\n", name->name); nfs_fattr_init(&dir_attr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); dprintk("NFS reply remove: %d\n", status); return status; } @@ -439,7 +439,7 @@ nfs3_proc_unlink_done(struct dentry *dir, struct rpc_task *task) return 1; if (msg->rpc_argp) { dir_attr = (struct nfs_fattr*)msg->rpc_resp; - nfs_refresh_inode(dir->d_inode, dir_attr); + nfs_post_op_update_inode(dir->d_inode, dir_attr); kfree(msg->rpc_argp); } return 0; @@ -468,8 +468,8 @@ nfs3_proc_rename(struct inode *old_dir, struct qstr *old_name, nfs_fattr_init(&old_dir_attr); nfs_fattr_init(&new_dir_attr); status = rpc_call(NFS_CLIENT(old_dir), NFS3PROC_RENAME, &arg, &res, 0); - nfs_refresh_inode(old_dir, &old_dir_attr); - nfs_refresh_inode(new_dir, &new_dir_attr); + nfs_post_op_update_inode(old_dir, &old_dir_attr); + nfs_post_op_update_inode(new_dir, &new_dir_attr); dprintk("NFS reply rename: %d\n", status); return status; } @@ -494,8 +494,8 @@ nfs3_proc_link(struct inode *inode, struct inode *dir, struct qstr *name) nfs_fattr_init(&dir_attr); nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(inode), NFS3PROC_LINK, &arg, &res, 0); - nfs_refresh_inode(dir, &dir_attr); - nfs_refresh_inode(inode, &fattr); + nfs_post_op_update_inode(dir, &dir_attr); + nfs_post_op_update_inode(inode, &fattr); dprintk("NFS reply link: %d\n", status); return status; } @@ -527,7 +527,7 @@ nfs3_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path, nfs_fattr_init(&dir_attr); nfs_fattr_init(fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_SYMLINK, &arg, &res, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); dprintk("NFS reply symlink: %d\n", status); return status; } @@ -558,7 +558,7 @@ nfs3_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr) nfs_fattr_init(&dir_attr); nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKDIR, &arg, &res, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); if (status != 0) goto out; status = nfs_instantiate(dentry, &fhandle, &fattr); @@ -584,7 +584,7 @@ nfs3_proc_rmdir(struct inode *dir, struct qstr *name) dprintk("NFS call rmdir %s\n", name->name); nfs_fattr_init(&dir_attr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_RMDIR, &arg, &dir_attr, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); dprintk("NFS reply rmdir: %d\n", status); return status; } @@ -679,7 +679,7 @@ nfs3_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr, nfs_fattr_init(&dir_attr); nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFS3PROC_MKNOD, &arg, &res, 0); - nfs_refresh_inode(dir, &dir_attr); + nfs_post_op_update_inode(dir, &dir_attr); if (status != 0) goto out; status = nfs_instantiate(dentry, &fh, &fattr); @@ -775,7 +775,7 @@ nfs3_write_done(struct rpc_task *task) return; data = (struct nfs_write_data *)task->tk_calldata; if (task->tk_status >= 0) - nfs_refresh_inode(data->inode, data->res.fattr); + nfs_post_op_update_inode(data->inode, data->res.fattr); nfs_writeback_done(task); } @@ -819,7 +819,7 @@ nfs3_commit_done(struct rpc_task *task) return; data = (struct nfs_write_data *)task->tk_calldata; if (task->tk_status >= 0) - nfs_refresh_inode(data->inode, data->res.fattr); + nfs_post_op_update_inode(data->inode, data->res.fattr); nfs_commit_done(task); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2a759e8e387..3274f2d354f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -187,8 +187,11 @@ static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinf { struct nfs_inode *nfsi = NFS_I(inode); + spin_lock(&inode->i_lock); + nfsi->cache_validity |= NFS_INO_INVALID_ATTR; if (cinfo->before == nfsi->change_attr && cinfo->atomic) nfsi->change_attr = cinfo->after; + spin_unlock(&inode->i_lock); } /* Helper for asynchronous RPC calls */ diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index 5ef28f08f42..a48a003242c 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -206,7 +206,7 @@ static int nfs_proc_write(struct nfs_write_data *wdata) nfs_fattr_init(fattr); status = rpc_call_sync(NFS_CLIENT(inode), &msg, flags); if (status >= 0) { - nfs_refresh_inode(inode, fattr); + nfs_post_op_update_inode(inode, fattr); wdata->res.count = wdata->args.count; wdata->verf.committed = NFS_FILE_SYNC; } @@ -275,6 +275,7 @@ nfs_proc_mknod(struct inode *dir, struct dentry *dentry, struct iattr *sattr, nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_CREATE, &arg, &res, 0); + nfs_mark_for_revalidate(dir); if (status == -EINVAL && S_ISFIFO(mode)) { sattr->ia_mode = mode; @@ -305,6 +306,7 @@ nfs_proc_remove(struct inode *dir, struct qstr *name) dprintk("NFS call remove %s\n", name->name); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); + nfs_mark_for_revalidate(dir); dprintk("NFS reply remove: %d\n", status); return status; @@ -331,8 +333,10 @@ nfs_proc_unlink_done(struct dentry *dir, struct rpc_task *task) { struct rpc_message *msg = &task->tk_msg; - if (msg->rpc_argp) + if (msg->rpc_argp) { + nfs_mark_for_revalidate(dir->d_inode); kfree(msg->rpc_argp); + } return 0; } @@ -352,6 +356,8 @@ nfs_proc_rename(struct inode *old_dir, struct qstr *old_name, dprintk("NFS call rename %s -> %s\n", old_name->name, new_name->name); status = rpc_call(NFS_CLIENT(old_dir), NFSPROC_RENAME, &arg, NULL, 0); + nfs_mark_for_revalidate(old_dir); + nfs_mark_for_revalidate(new_dir); dprintk("NFS reply rename: %d\n", status); return status; } @@ -369,6 +375,7 @@ nfs_proc_link(struct inode *inode, struct inode *dir, struct qstr *name) dprintk("NFS call link %s\n", name->name); status = rpc_call(NFS_CLIENT(inode), NFSPROC_LINK, &arg, NULL, 0); + nfs_mark_for_revalidate(dir); dprintk("NFS reply link: %d\n", status); return status; } @@ -394,6 +401,7 @@ nfs_proc_symlink(struct inode *dir, struct qstr *name, struct qstr *path, nfs_fattr_init(fattr); fhandle->size = 0; status = rpc_call(NFS_CLIENT(dir), NFSPROC_SYMLINK, &arg, NULL, 0); + nfs_mark_for_revalidate(dir); dprintk("NFS reply symlink: %d\n", status); return status; } @@ -418,6 +426,7 @@ nfs_proc_mkdir(struct inode *dir, struct dentry *dentry, struct iattr *sattr) dprintk("NFS call mkdir %s\n", dentry->d_name.name); nfs_fattr_init(&fattr); status = rpc_call(NFS_CLIENT(dir), NFSPROC_MKDIR, &arg, &res, 0); + nfs_mark_for_revalidate(dir); if (status == 0) status = nfs_instantiate(dentry, &fhandle, &fattr); dprintk("NFS reply mkdir: %d\n", status); @@ -436,6 +445,7 @@ nfs_proc_rmdir(struct inode *dir, struct qstr *name) dprintk("NFS call rmdir %s\n", name->name); status = rpc_call(NFS_CLIENT(dir), NFSPROC_RMDIR, &arg, NULL, 0); + nfs_mark_for_revalidate(dir); dprintk("NFS reply rmdir: %d\n", status); return status; } @@ -579,7 +589,7 @@ nfs_write_done(struct rpc_task *task) struct nfs_write_data *data = (struct nfs_write_data *) task->tk_calldata; if (task->tk_status >= 0) - nfs_refresh_inode(data->inode, data->res.fattr); + nfs_post_op_update_inode(data->inode, data->res.fattr); nfs_writeback_done(task); } -- cgit v1.2.3 From 0c70b50150cfb0b43ff500a8a394a52b4d5f1350 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Tue, 25 Oct 2005 11:48:36 -0400 Subject: NFS: nfs_lookup doesn't need to revalidate the parent directory's inode nfs_lookup() used to consult a lookup cache before trying an actual wire lookup operation. The lookup cache would be invalid, of course, if the parent directory's mtime had changed, so nfs_lookup performed an inode revalidation on the parent. Since nfs_lookup() doesn't use a cache anymore, the revalidation is no longer necessary. There are cases where it will generate a lot of unnecessary GETATTR traffic. See http://bugzilla.linux-nfs.org/show_bug.cgi?id=9 Test-plan: Use lndir and "rm -rf" and watch for excess GETATTR traffic or application level errors. Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index b8a73045e9a..ce8f77dadff 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -853,12 +853,6 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru dentry->d_op = NFS_PROTO(dir)->dentry_ops; lock_kernel(); - /* Revalidate parent directory attribute cache */ - error = nfs_revalidate_inode(NFS_SERVER(dir), dir); - if (error < 0) { - res = ERR_PTR(error); - goto out_unlock; - } /* If we're doing an exclusive create, optimize away the lookup */ if (nfs_is_exclusive_create(dir, nd)) -- cgit v1.2.3 From 56ae19f38f10aad4f27f7e12138a29b295dff07a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:40 -0400 Subject: NFSv4: Add directory post-op attributes to the CREATE operations. Since the directory attributes change every time we CREATE a file, we might as well pick up the new directory attributes in the same compound. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 28 ++++++++++++++++++---- fs/nfs/nfs4xdr.c | 69 +++++++++++++++++++++++++++++++++++++++++++++++-------- 2 files changed, 82 insertions(+), 15 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3274f2d354f..f363fd6c7f4 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -443,7 +443,11 @@ static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner *sp, stru nfs_increment_open_seqid(status, o_arg->seqid); if (status != 0) goto out; - update_changeattr(dir, &o_res->cinfo); + if (o_arg->open_flags & O_CREAT) { + update_changeattr(dir, &o_res->cinfo); + nfs_post_op_update_inode(dir, o_res->dir_attr); + } else + nfs_refresh_inode(dir, o_res->dir_attr); if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { status = _nfs4_proc_open_confirm(server->client, &o_res->fh, sp, &o_res->stateid, o_arg->seqid); @@ -497,7 +501,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(dir); struct nfs_delegation *delegation = NFS_I(inode)->delegation; - struct nfs_fattr f_attr; + struct nfs_fattr f_attr, dir_attr; struct nfs_openargs o_arg = { .fh = NFS_FH(dir), .open_flags = state->state, @@ -507,6 +511,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st }; struct nfs_openres o_res = { .f_attr = &f_attr, + .dir_attr = &dir_attr, .server = server, }; int status = 0; @@ -524,6 +529,7 @@ static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st if (o_arg.seqid == NULL) goto out; nfs_fattr_init(&f_attr); + nfs_fattr_init(&dir_attr); status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_nodeleg; @@ -694,7 +700,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st struct nfs4_client *clp = server->nfs4_state; struct inode *inode = NULL; int status; - struct nfs_fattr f_attr; + struct nfs_fattr f_attr, dir_attr; struct nfs_openargs o_arg = { .fh = NFS_FH(dir), .open_flags = flags, @@ -705,6 +711,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st }; struct nfs_openres o_res = { .f_attr = &f_attr, + .dir_attr = &dir_attr, .server = server, }; @@ -727,6 +734,7 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st if (o_arg.seqid == NULL) return -ENOMEM; nfs_fattr_init(&f_attr); + nfs_fattr_init(&dir_attr); status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_err; @@ -1746,6 +1754,7 @@ static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name, struct nfs_fattr *fattr) { struct nfs_server *server = NFS_SERVER(dir); + struct nfs_fattr dir_fattr; struct nfs4_create_arg arg = { .dir_fh = NFS_FH(dir), .server = server, @@ -1758,6 +1767,7 @@ static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name, .server = server, .fh = fhandle, .fattr = fattr, + .dir_fattr = &dir_fattr, }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK], @@ -1770,10 +1780,12 @@ static int _nfs4_proc_symlink(struct inode *dir, struct qstr *name, return -ENAMETOOLONG; arg.u.symlink = path; nfs_fattr_init(fattr); + nfs_fattr_init(&dir_fattr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); if (!status) update_changeattr(dir, &res.dir_cinfo); + nfs_post_op_update_inode(dir, res.dir_fattr); return status; } @@ -1797,7 +1809,7 @@ static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry, { struct nfs_server *server = NFS_SERVER(dir); struct nfs_fh fhandle; - struct nfs_fattr fattr; + struct nfs_fattr fattr, dir_fattr; struct nfs4_create_arg arg = { .dir_fh = NFS_FH(dir), .server = server, @@ -1810,6 +1822,7 @@ static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry, .server = server, .fh = &fhandle, .fattr = &fattr, + .dir_fattr = &dir_fattr, }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE], @@ -1819,10 +1832,12 @@ static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry, int status; nfs_fattr_init(&fattr); + nfs_fattr_init(&dir_fattr); status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); if (!status) { update_changeattr(dir, &res.dir_cinfo); + nfs_post_op_update_inode(dir, res.dir_fattr); status = nfs_instantiate(dentry, &fhandle, &fattr); } return status; @@ -1895,7 +1910,7 @@ static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry, { struct nfs_server *server = NFS_SERVER(dir); struct nfs_fh fh; - struct nfs_fattr fattr; + struct nfs_fattr fattr, dir_fattr; struct nfs4_create_arg arg = { .dir_fh = NFS_FH(dir), .server = server, @@ -1907,6 +1922,7 @@ static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry, .server = server, .fh = &fh, .fattr = &fattr, + .dir_fattr = &dir_fattr, }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE], @@ -1917,6 +1933,7 @@ static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry, int mode = sattr->ia_mode; nfs_fattr_init(&fattr); + nfs_fattr_init(&dir_fattr); BUG_ON(!(sattr->ia_valid & ATTR_MODE)); BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode)); @@ -1938,6 +1955,7 @@ static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry, status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); if (status == 0) { update_changeattr(dir, &res.dir_cinfo); + nfs_post_op_update_inode(dir, res.dir_fattr); status = nfs_instantiate(dentry, &fh, &fattr); } return status; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 8b21de8a06f..7f91d613d31 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -95,6 +95,8 @@ static int nfs_stat_to_errno(int); #define decode_getattr_maxsz (op_decode_hdr_maxsz + nfs4_fattr_maxsz) #define encode_savefh_maxsz (op_encode_hdr_maxsz) #define decode_savefh_maxsz (op_decode_hdr_maxsz) +#define encode_restorefh_maxsz (op_encode_hdr_maxsz) +#define decode_restorefh_maxsz (op_decode_hdr_maxsz) #define encode_fsinfo_maxsz (op_encode_hdr_maxsz + 2) #define decode_fsinfo_maxsz (op_decode_hdr_maxsz + 11) #define encode_renew_maxsz (op_encode_hdr_maxsz + 3) @@ -336,14 +338,20 @@ static int nfs_stat_to_errno(int); decode_getfh_maxsz) #define NFS4_enc_create_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ + encode_savefh_maxsz + \ encode_create_maxsz + \ + encode_getfh_maxsz + \ encode_getattr_maxsz + \ - encode_getfh_maxsz) + encode_restorefh_maxsz + \ + encode_getattr_maxsz) #define NFS4_dec_create_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ + decode_savefh_maxsz + \ decode_create_maxsz + \ + decode_getfh_maxsz + \ decode_getattr_maxsz + \ - decode_getfh_maxsz) + decode_restorefh_maxsz + \ + decode_getattr_maxsz) #define NFS4_enc_pathconf_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_getattr_maxsz) @@ -1112,6 +1120,17 @@ static int encode_renew(struct xdr_stream *xdr, const struct nfs4_client *client return 0; } +static int +encode_restorefh(struct xdr_stream *xdr) +{ + uint32_t *p; + + RESERVE_SPACE(4); + WRITE32(OP_RESTOREFH); + + return 0; +} + static int encode_setacl(struct xdr_stream *xdr, struct nfs_setaclargs *arg) { @@ -1358,7 +1377,7 @@ static int nfs4_xdr_enc_create(struct rpc_rqst *req, uint32_t *p, const struct n { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 4, + .nops = 7, }; int status; @@ -1366,10 +1385,16 @@ static int nfs4_xdr_enc_create(struct rpc_rqst *req, uint32_t *p, const struct n encode_compound_hdr(&xdr, &hdr); if ((status = encode_putfh(&xdr, args->dir_fh)) != 0) goto out; + if ((status = encode_savefh(&xdr)) != 0) + goto out; if ((status = encode_create(&xdr, args)) != 0) goto out; if ((status = encode_getfh(&xdr)) != 0) goto out; + if ((status = encode_getfattr(&xdr, args->bitmask)) != 0) + goto out; + if ((status = encode_restorefh(&xdr)) != 0) + goto out; status = encode_getfattr(&xdr, args->bitmask); out: return status; @@ -1429,7 +1454,7 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 4, + .nops = 7, }; int status; @@ -1439,6 +1464,9 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); + if (status) + goto out; + status = encode_savefh(&xdr); if (status) goto out; status = encode_open(&xdr, args); @@ -1448,6 +1476,12 @@ static int nfs4_xdr_enc_open(struct rpc_rqst *req, uint32_t *p, struct nfs_opena if (status) goto out; status = encode_getfattr(&xdr, args->bitmask); + if (status) + goto out; + status = encode_restorefh(&xdr); + if (status) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -3218,6 +3252,12 @@ static int decode_renew(struct xdr_stream *xdr) return decode_op_hdr(xdr, OP_RENEW); } +static int +decode_restorefh(struct xdr_stream *xdr) +{ + return decode_op_hdr(xdr, OP_RESTOREFH); +} + static int decode_getacl(struct xdr_stream *xdr, struct rpc_rqst *req, size_t *acl_len) { @@ -3510,13 +3550,17 @@ static int nfs4_xdr_dec_create(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_ goto out; if ((status = decode_putfh(&xdr)) != 0) goto out; + if ((status = decode_savefh(&xdr)) != 0) + goto out; if ((status = decode_create(&xdr,&res->dir_cinfo)) != 0) goto out; if ((status = decode_getfh(&xdr, res->fh)) != 0) goto out; - status = decode_getfattr(&xdr, res->fattr, res->server); - if (status == NFS4ERR_DELAY) - status = 0; + if (decode_getfattr(&xdr, res->fattr, res->server) != 0) + goto out; + if ((status = decode_restorefh(&xdr)) != 0) + goto out; + decode_getfattr(&xdr, res->dir_fattr, res->server); out: return status; } @@ -3654,15 +3698,20 @@ static int nfs4_xdr_dec_open(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_ope status = decode_putfh(&xdr); if (status) goto out; + status = decode_savefh(&xdr); + if (status) + goto out; status = decode_open(&xdr, res); if (status) goto out; status = decode_getfh(&xdr, &res->fh); if (status) goto out; - status = decode_getfattr(&xdr, res->f_attr, res->server); - if (status == NFS4ERR_DELAY) - status = 0; + if (decode_getfattr(&xdr, res->f_attr, res->server) != 0) + goto out; + if ((status = decode_restorefh(&xdr)) != 0) + goto out; + decode_getfattr(&xdr, res->dir_attr, res->server); out: return status; } -- cgit v1.2.3 From 3338c143b4fde2d256016b63043ec8e2c93eba19 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:41 -0400 Subject: NFS: Optimise attribute revalidation on close(). Only force a getattr in nfs_file_flush() if the attribute cache is stale. Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 572d8593486..57d3e77d97e 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -205,8 +205,8 @@ nfs_file_flush(struct file *file) if (!status) { status = ctx->error; ctx->error = 0; - if (!status && !nfs_have_delegation(inode, FMODE_READ)) - __nfs_revalidate_inode(NFS_SERVER(inode), inode); + if (!status) + nfs_revalidate_inode(NFS_SERVER(inode), inode); } unlock_kernel(); return status; -- cgit v1.2.3 From 516a6af641bb50c608329a5bd751acd0d65cc4ab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:41 -0400 Subject: NFS: Add optional post-op getattr instruction to the NFSv4 file close. "Optional" means that the close call will not fail if the getattr at the end of the compound fails. If it does succeed, try to refresh inode attributes. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 9 ++++++++- fs/nfs/nfs4xdr.c | 34 ++++++++++++++++++++++++++++------ 2 files changed, 36 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f363fd6c7f4..7be3d2d15d6 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -865,6 +865,7 @@ struct nfs4_closedata { struct nfs4_state *state; struct nfs_closeargs arg; struct nfs_closeres res; + struct nfs_fattr fattr; }; static void nfs4_free_closedata(struct nfs4_closedata *calldata) @@ -904,6 +905,7 @@ static void nfs4_close_done(struct rpc_task *task) return; } } + nfs_refresh_inode(calldata->inode, calldata->res.fattr); state->state = calldata->arg.open_flags; nfs4_free_closedata(calldata); } @@ -941,6 +943,7 @@ static void nfs4_close_begin(struct rpc_task *task) rpc_exit(task, 0); return; } + nfs_fattr_init(calldata->res.fattr); if (mode != 0) msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE]; calldata->arg.open_flags = mode; @@ -960,6 +963,7 @@ static void nfs4_close_begin(struct rpc_task *task) */ int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) { + struct nfs_server *server = NFS_SERVER(inode); struct nfs4_closedata *calldata; int status = -ENOMEM; @@ -974,8 +978,11 @@ int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid); if (calldata->arg.seqid == NULL) goto out_free_calldata; + calldata->arg.bitmask = server->attr_bitmask; + calldata->res.fattr = &calldata->fattr; + calldata->res.server = server; - status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_close_begin, + status = nfs4_call_async(server->client, nfs4_close_begin, nfs4_close_done, calldata); if (status == 0) goto out; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 7f91d613d31..cd9e26cfa86 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -198,17 +198,21 @@ static int nfs_stat_to_errno(int); #define NFS4_enc_open_downgrade_sz \ (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - op_encode_hdr_maxsz + 7) + op_encode_hdr_maxsz + 7 + \ + encode_getattr_maxsz) #define NFS4_dec_open_downgrade_sz \ (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ - op_decode_hdr_maxsz + 4) + op_decode_hdr_maxsz + 4 + \ + decode_getattr_maxsz) #define NFS4_enc_close_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - op_encode_hdr_maxsz + 5) + op_encode_hdr_maxsz + 5 + \ + encode_getattr_maxsz) #define NFS4_dec_close_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ - op_decode_hdr_maxsz + 4) + op_decode_hdr_maxsz + 4 + \ + decode_getattr_maxsz) #define NFS4_enc_setattr_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ op_encode_hdr_maxsz + 4 + \ @@ -1433,7 +1437,7 @@ static int nfs4_xdr_enc_close(struct rpc_rqst *req, uint32_t *p, struct nfs_clos { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; @@ -1443,6 +1447,9 @@ static int nfs4_xdr_enc_close(struct rpc_rqst *req, uint32_t *p, struct nfs_clos if(status) goto out; status = encode_close(&xdr, args); + if (status != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -1541,7 +1548,7 @@ static int nfs4_xdr_enc_open_downgrade(struct rpc_rqst *req, uint32_t *p, struct { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; @@ -1551,6 +1558,9 @@ static int nfs4_xdr_enc_open_downgrade(struct rpc_rqst *req, uint32_t *p, struct if (status) goto out; status = encode_open_downgrade(&xdr, args); + if (status != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -3403,6 +3413,9 @@ static int nfs4_xdr_dec_open_downgrade(struct rpc_rqst *rqstp, uint32_t *p, stru if (status) goto out; status = decode_open_downgrade(&xdr, res); + if (status != 0) + goto out; + decode_getfattr(&xdr, res->fattr, res->server); out: return status; } @@ -3678,6 +3691,15 @@ static int nfs4_xdr_dec_close(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_cl if (status) goto out; status = decode_close(&xdr, res); + if (status != 0) + goto out; + /* + * Note: Server may do delete on close for this file + * in which case the getattr call will fail with + * an ESTALE error. Shouldn't be a problem, + * though, since fattr->valid will remain unset. + */ + decode_getfattr(&xdr, res->fattr, res->server); out: return status; } -- cgit v1.2.3 From cf809556149f076b7a020c10e066b2b96e79b6a1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:42 -0400 Subject: NFS: Ensure that nfs_link() instantiates the dentry correctly Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index ce8f77dadff..8272ed3fc70 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1432,17 +1432,14 @@ nfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) old_dentry->d_parent->d_name.name, old_dentry->d_name.name, dentry->d_parent->d_name.name, dentry->d_name.name); - /* - * Drop the dentry in advance to force a new lookup. - * Since nfs_proc_link doesn't return a file handle, - * we can't use the existing dentry. - */ lock_kernel(); - d_drop(dentry); - nfs_begin_data_update(dir); nfs_begin_data_update(inode); error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name); + if (error == 0) { + atomic_inc(&inode->i_count); + d_instantiate(dentry, inode); + } nfs_end_data_update(inode); nfs_end_data_update(dir); unlock_kernel(); -- cgit v1.2.3 From 91ba2eeec5e8e86e054937eb3bf5aec5b22b1830 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:42 -0400 Subject: NFSv4: Add post-op attributes to nfs4_proc_link() Optimise attribute revalidation when hardlinking. Add post-op attributes for the directory and the original inode. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 22 +++++++++++++++++----- fs/nfs/nfs4xdr.c | 34 ++++++++++++++++++++++++++++------ 2 files changed, 45 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 7be3d2d15d6..04995e39e86 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1724,22 +1724,34 @@ static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name, static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name) { + struct nfs_server *server = NFS_SERVER(inode); struct nfs4_link_arg arg = { .fh = NFS_FH(inode), .dir_fh = NFS_FH(dir), .name = name, + .bitmask = server->attr_bitmask, + }; + struct nfs_fattr fattr, dir_attr; + struct nfs4_link_res res = { + .server = server, + .fattr = &fattr, + .dir_attr = &dir_attr, }; - struct nfs4_change_info cinfo = { }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK], .rpc_argp = &arg, - .rpc_resp = &cinfo, + .rpc_resp = &res, }; int status; - status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0); - if (!status) - update_changeattr(dir, &cinfo); + nfs_fattr_init(res.fattr); + nfs_fattr_init(res.dir_attr); + status = rpc_call_sync(server->client, &msg, 0); + if (!status) { + update_changeattr(dir, &res.cinfo); + nfs_post_op_update_inode(dir, res.dir_attr); + nfs_refresh_inode(inode, res.fattr); + } return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index cd9e26cfa86..f624b693ce2 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -324,12 +324,18 @@ static int nfs_stat_to_errno(int); encode_putfh_maxsz + \ encode_savefh_maxsz + \ encode_putfh_maxsz + \ - encode_link_maxsz) + encode_link_maxsz + \ + decode_getattr_maxsz + \ + encode_restorefh_maxsz + \ + decode_getattr_maxsz) #define NFS4_dec_link_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ decode_savefh_maxsz + \ decode_putfh_maxsz + \ - decode_link_maxsz) + decode_link_maxsz + \ + decode_getattr_maxsz + \ + decode_restorefh_maxsz + \ + decode_getattr_maxsz) #define NFS4_enc_symlink_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_symlink_maxsz + \ @@ -1357,7 +1363,7 @@ static int nfs4_xdr_enc_link(struct rpc_rqst *req, uint32_t *p, const struct nfs { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 4, + .nops = 7, }; int status; @@ -1369,7 +1375,13 @@ static int nfs4_xdr_enc_link(struct rpc_rqst *req, uint32_t *p, const struct nfs goto out; if ((status = encode_putfh(&xdr, args->dir_fh)) != 0) goto out; - status = encode_link(&xdr, args->name); + if ((status = encode_link(&xdr, args->name)) != 0) + goto out; + if ((status = encode_getfattr(&xdr, args->bitmask)) != 0) + goto out; + if ((status = encode_restorefh(&xdr)) != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -3529,7 +3541,7 @@ out: /* * Decode LINK response */ -static int nfs4_xdr_dec_link(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_change_info *cinfo) +static int nfs4_xdr_dec_link(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_link_res *res) { struct xdr_stream xdr; struct compound_hdr hdr; @@ -3544,7 +3556,17 @@ static int nfs4_xdr_dec_link(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_ch goto out; if ((status = decode_putfh(&xdr)) != 0) goto out; - status = decode_link(&xdr, cinfo); + if ((status = decode_link(&xdr, &res->cinfo)) != 0) + goto out; + /* + * Note order: OP_LINK leaves the directory as the current + * filehandle. + */ + if (decode_getfattr(&xdr, res->dir_attr, res->server) != 0) + goto out; + if ((status = decode_restorefh(&xdr)) != 0) + goto out; + decode_getfattr(&xdr, res->fattr, res->server); out: return status; } -- cgit v1.2.3 From 6caf2c8276d371679a798058e8fdf49f5ff831a3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:43 -0400 Subject: NFSv4: Add post-op attributes to nfs4_proc_rename() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 15 +++++++++++++-- fs/nfs/nfs4xdr.c | 29 ++++++++++++++++++++++++----- 2 files changed, 37 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 04995e39e86..f96bc12c0fa 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1685,13 +1685,20 @@ static int nfs4_proc_unlink_done(struct dentry *dir, struct rpc_task *task) static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name, struct inode *new_dir, struct qstr *new_name) { + struct nfs_server *server = NFS_SERVER(old_dir); struct nfs4_rename_arg arg = { .old_dir = NFS_FH(old_dir), .new_dir = NFS_FH(new_dir), .old_name = old_name, .new_name = new_name, + .bitmask = server->attr_bitmask, + }; + struct nfs_fattr old_fattr, new_fattr; + struct nfs4_rename_res res = { + .server = server, + .old_fattr = &old_fattr, + .new_fattr = &new_fattr, }; - struct nfs4_rename_res res = { }; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME], .rpc_argp = &arg, @@ -1699,11 +1706,15 @@ static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name, }; int status; - status = rpc_call_sync(NFS_CLIENT(old_dir), &msg, 0); + nfs_fattr_init(res.old_fattr); + nfs_fattr_init(res.new_fattr); + status = rpc_call_sync(server->client, &msg, 0); if (!status) { update_changeattr(old_dir, &res.old_cinfo); + nfs_post_op_update_inode(old_dir, res.old_fattr); update_changeattr(new_dir, &res.new_cinfo); + nfs_post_op_update_inode(new_dir, res.new_fattr); } return status; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index f624b693ce2..2a07755bd34 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -314,12 +314,18 @@ static int nfs_stat_to_errno(int); encode_putfh_maxsz + \ encode_savefh_maxsz + \ encode_putfh_maxsz + \ - encode_rename_maxsz) + encode_rename_maxsz + \ + encode_getattr_maxsz + \ + encode_restorefh_maxsz + \ + encode_getattr_maxsz) #define NFS4_dec_rename_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ decode_savefh_maxsz + \ decode_putfh_maxsz + \ - decode_rename_maxsz) + decode_rename_maxsz + \ + decode_getattr_maxsz + \ + decode_restorefh_maxsz + \ + decode_getattr_maxsz) #define NFS4_enc_link_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_savefh_maxsz + \ @@ -1339,7 +1345,7 @@ static int nfs4_xdr_enc_rename(struct rpc_rqst *req, uint32_t *p, const struct n { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 4, + .nops = 7, }; int status; @@ -1351,7 +1357,13 @@ static int nfs4_xdr_enc_rename(struct rpc_rqst *req, uint32_t *p, const struct n goto out; if ((status = encode_putfh(&xdr, args->new_dir)) != 0) goto out; - status = encode_rename(&xdr, args->old_name, args->new_name); + if ((status = encode_rename(&xdr, args->old_name, args->new_name)) != 0) + goto out; + if ((status = encode_getfattr(&xdr, args->bitmask)) != 0) + goto out; + if ((status = encode_restorefh(&xdr)) != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -3533,7 +3545,14 @@ static int nfs4_xdr_dec_rename(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_ goto out; if ((status = decode_putfh(&xdr)) != 0) goto out; - status = decode_rename(&xdr, &res->old_cinfo, &res->new_cinfo); + if ((status = decode_rename(&xdr, &res->old_cinfo, &res->new_cinfo)) != 0) + goto out; + /* Current FH is target directory */ + if (decode_getfattr(&xdr, res->new_fattr, res->server) != 0) + goto out; + if ((status = decode_restorefh(&xdr)) != 0) + goto out; + decode_getfattr(&xdr, res->old_fattr, res->server); out: return status; } -- cgit v1.2.3 From 16e429596dec4d28e16812b3a9be27f18412c567 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:44 -0400 Subject: NFSv4: Add post-op attributes to nfs4_proc_remove() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 27 +++++++++++++++++++++------ fs/nfs/nfs4xdr.c | 25 +++++++++++++++++-------- 2 files changed, 38 insertions(+), 14 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f96bc12c0fa..bab47c4cb41 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1614,11 +1614,17 @@ out: static int _nfs4_proc_remove(struct inode *dir, struct qstr *name) { + struct nfs_server *server = NFS_SERVER(dir); struct nfs4_remove_arg args = { .fh = NFS_FH(dir), .name = name, + .bitmask = server->attr_bitmask, + }; + struct nfs_fattr dir_attr; + struct nfs4_remove_res res = { + .server = server, + .dir_attr = &dir_attr, }; - struct nfs4_change_info res; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE], .rpc_argp = &args, @@ -1626,9 +1632,12 @@ static int _nfs4_proc_remove(struct inode *dir, struct qstr *name) }; int status; - status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); - if (status == 0) - update_changeattr(dir, &res); + nfs_fattr_init(res.dir_attr); + status = rpc_call_sync(server->client, &msg, 0); + if (status == 0) { + update_changeattr(dir, &res.cinfo); + nfs_post_op_update_inode(dir, res.dir_attr); + } return status; } @@ -1646,12 +1655,14 @@ static int nfs4_proc_remove(struct inode *dir, struct qstr *name) struct unlink_desc { struct nfs4_remove_arg args; - struct nfs4_change_info res; + struct nfs4_remove_res res; + struct nfs_fattr dir_attr; }; static int nfs4_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir, struct qstr *name) { + struct nfs_server *server = NFS_SERVER(dir->d_inode); struct unlink_desc *up; up = (struct unlink_desc *) kmalloc(sizeof(*up), GFP_KERNEL); @@ -1660,6 +1671,9 @@ static int nfs4_proc_unlink_setup(struct rpc_message *msg, struct dentry *dir, up->args.fh = NFS_FH(dir->d_inode); up->args.name = name; + up->args.bitmask = server->attr_bitmask; + up->res.server = server; + up->res.dir_attr = &up->dir_attr; msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE]; msg->rpc_argp = &up->args; @@ -1674,7 +1688,8 @@ static int nfs4_proc_unlink_done(struct dentry *dir, struct rpc_task *task) if (msg->rpc_resp != NULL) { up = container_of(msg->rpc_resp, struct unlink_desc, res); - update_changeattr(dir->d_inode, &up->res); + update_changeattr(dir->d_inode, &up->res.cinfo); + nfs_post_op_update_inode(dir->d_inode, up->res.dir_attr); kfree(up); msg->rpc_resp = NULL; msg->rpc_argp = NULL; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 2a07755bd34..3ee3a1669d2 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -306,10 +306,12 @@ static int nfs_stat_to_errno(int); decode_getfh_maxsz) #define NFS4_enc_remove_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - encode_remove_maxsz) + encode_remove_maxsz + \ + encode_getattr_maxsz) #define NFS4_dec_remove_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ - op_decode_hdr_maxsz + 5) + op_decode_hdr_maxsz + 5 + \ + decode_getattr_maxsz) #define NFS4_enc_rename_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_savefh_maxsz + \ @@ -1327,14 +1329,18 @@ static int nfs4_xdr_enc_remove(struct rpc_rqst *req, uint32_t *p, const struct n { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); - if ((status = encode_putfh(&xdr, args->fh)) == 0) - status = encode_remove(&xdr, args->name); + if ((status = encode_putfh(&xdr, args->fh)) != 0) + goto out; + if ((status = encode_remove(&xdr, args->name)) != 0) + goto out; + status = encode_getfattr(&xdr, args->bitmask); +out: return status; } @@ -3512,7 +3518,7 @@ out: /* * Decode REMOVE response */ -static int nfs4_xdr_dec_remove(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_change_info *cinfo) +static int nfs4_xdr_dec_remove(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_remove_res *res) { struct xdr_stream xdr; struct compound_hdr hdr; @@ -3521,8 +3527,11 @@ static int nfs4_xdr_dec_remove(struct rpc_rqst *rqstp, uint32_t *p, struct nfs4_ xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p); if ((status = decode_compound_hdr(&xdr, &hdr)) != 0) goto out; - if ((status = decode_putfh(&xdr)) == 0) - status = decode_remove(&xdr, cinfo); + if ((status = decode_putfh(&xdr)) != 0) + goto out; + if ((status = decode_remove(&xdr, &res->cinfo)) != 0) + goto out; + decode_getfattr(&xdr, res->dir_attr, res->server); out: return status; } -- cgit v1.2.3 From 4f9838c7ecd14f31f701f64fa65ded132fc0db8a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:44 -0400 Subject: NFSv4: Add post-op attributes to NFSv4 write and commit callbacks. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 13 ++++++++++++- fs/nfs/nfs4xdr.c | 28 ++++++++++++++++++++++------ 2 files changed, 34 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bab47c4cb41..933e13b383f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2169,8 +2169,10 @@ nfs4_write_done(struct rpc_task *task) rpc_restart_call(task); return; } - if (task->tk_status >= 0) + if (task->tk_status >= 0) { renew_lease(NFS_SERVER(inode), data->timestamp); + nfs_post_op_update_inode(inode, data->res.fattr); + } /* Call back common NFS writeback processing */ nfs_writeback_done(task); } @@ -2186,6 +2188,7 @@ nfs4_proc_write_setup(struct nfs_write_data *data, int how) .rpc_cred = data->cred, }; struct inode *inode = data->inode; + struct nfs_server *server = NFS_SERVER(inode); int stable; int flags; @@ -2197,6 +2200,8 @@ nfs4_proc_write_setup(struct nfs_write_data *data, int how) } else stable = NFS_UNSTABLE; data->args.stable = stable; + data->args.bitmask = server->attr_bitmask; + data->res.server = server; data->timestamp = jiffies; @@ -2218,6 +2223,8 @@ nfs4_commit_done(struct rpc_task *task) rpc_restart_call(task); return; } + if (task->tk_status >= 0) + nfs_post_op_update_inode(inode, data->res.fattr); /* Call back common NFS writeback processing */ nfs_commit_done(task); } @@ -2233,8 +2240,12 @@ nfs4_proc_commit_setup(struct nfs_write_data *data, int how) .rpc_cred = data->cred, }; struct inode *inode = data->inode; + struct nfs_server *server = NFS_SERVER(inode); int flags; + data->args.bitmask = server->attr_bitmask; + data->res.server = server; + /* Set the initial flags for the task. */ flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 3ee3a1669d2..6f1bf182e0e 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -159,16 +159,20 @@ static int nfs_stat_to_errno(int); op_decode_hdr_maxsz + 2) #define NFS4_enc_write_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - op_encode_hdr_maxsz + 8) + op_encode_hdr_maxsz + 8 + \ + encode_getattr_maxsz) #define NFS4_dec_write_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ - op_decode_hdr_maxsz + 4) + op_decode_hdr_maxsz + 4 + \ + decode_getattr_maxsz) #define NFS4_enc_commit_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ - op_encode_hdr_maxsz + 3) + op_encode_hdr_maxsz + 3 + \ + encode_getattr_maxsz) #define NFS4_dec_commit_sz (compound_decode_hdr_maxsz + \ decode_putfh_maxsz + \ - op_decode_hdr_maxsz + 2) + op_decode_hdr_maxsz + 2 + \ + decode_getattr_maxsz) #define NFS4_enc_open_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ op_encode_hdr_maxsz + \ @@ -1799,7 +1803,7 @@ static int nfs4_xdr_enc_write(struct rpc_rqst *req, uint32_t *p, struct nfs_writ { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; @@ -1809,6 +1813,9 @@ static int nfs4_xdr_enc_write(struct rpc_rqst *req, uint32_t *p, struct nfs_writ if (status) goto out; status = encode_write(&xdr, args); + if (status) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -1820,7 +1827,7 @@ static int nfs4_xdr_enc_commit(struct rpc_rqst *req, uint32_t *p, struct nfs_wri { struct xdr_stream xdr; struct compound_hdr hdr = { - .nops = 2, + .nops = 3, }; int status; @@ -1830,6 +1837,9 @@ static int nfs4_xdr_enc_commit(struct rpc_rqst *req, uint32_t *p, struct nfs_wri if (status) goto out; status = encode_commit(&xdr, args); + if (status) + goto out; + status = encode_getfattr(&xdr, args->bitmask); out: return status; } @@ -4001,6 +4011,9 @@ static int nfs4_xdr_dec_write(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_wr if (status) goto out; status = decode_write(&xdr, res); + if (status) + goto out; + decode_getfattr(&xdr, res->fattr, res->server); if (!status) status = res->count; out: @@ -4024,6 +4037,9 @@ static int nfs4_xdr_dec_commit(struct rpc_rqst *rqstp, uint32_t *p, struct nfs_w if (status) goto out; status = decode_commit(&xdr, res); + if (status) + goto out; + decode_getfattr(&xdr, res->fattr, res->server); out: return status; } -- cgit v1.2.3 From 16c32b71bc53d6f7143702ebb264b4ef20d8f1e5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:45 -0400 Subject: NFSv4: Convert unnecessary XDR warning messages into dprintk() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 17 ++++++----------- 1 file changed, 6 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 6f1bf182e0e..fbbace8a30c 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -2766,8 +2766,7 @@ static int decode_server_caps(struct xdr_stream *xdr, struct nfs4_server_caps_re goto xdr_error; status = verify_attr_len(xdr, savep, attrlen); xdr_error: - if (status != 0) - printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); + dprintk("%s: xdr returned %d!\n", __FUNCTION__, -status); return status; } @@ -2800,8 +2799,7 @@ static int decode_statfs(struct xdr_stream *xdr, struct nfs_fsstat *fsstat) status = verify_attr_len(xdr, savep, attrlen); xdr_error: - if (status != 0) - printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); + dprintk("%s: xdr returned %d!\n", __FUNCTION__, -status); return status; } @@ -2826,8 +2824,7 @@ static int decode_pathconf(struct xdr_stream *xdr, struct nfs_pathconf *pathconf status = verify_attr_len(xdr, savep, attrlen); xdr_error: - if (status != 0) - printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); + dprintk("%s: xdr returned %d!\n", __FUNCTION__, -status); return status; } @@ -2886,8 +2883,7 @@ static int decode_getfattr(struct xdr_stream *xdr, struct nfs_fattr *fattr, cons if ((status = verify_attr_len(xdr, savep, attrlen)) == 0) fattr->valid = NFS_ATTR_FATTR | NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4; xdr_error: - if (status != 0) - printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); + dprintk("%s: xdr returned %d\n", __FUNCTION__, -status); return status; } @@ -2920,8 +2916,7 @@ static int decode_fsinfo(struct xdr_stream *xdr, struct nfs_fsinfo *fsinfo) status = verify_attr_len(xdr, savep, attrlen); xdr_error: - if (status != 0) - printk(KERN_NOTICE "%s: xdr error %d!\n", __FUNCTION__, -status); + dprintk("%s: xdr returned %d!\n", __FUNCTION__, -status); return status; } @@ -3088,7 +3083,7 @@ static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res) p += bmlen; return decode_delegation(xdr, res); xdr_error: - printk(KERN_NOTICE "%s: xdr error!\n", __FUNCTION__); + dprintk("%s: Bitmap too large! Length = %u\n", __FUNCTION__, bmlen); return -EIO; } -- cgit v1.2.3 From bec273b491bd16351a9cdb8e9a51d30afa7fe9f4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Oct 2005 22:12:45 -0400 Subject: NFS: Allow files that are open for write to invalidate caches Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'fs') diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 6b3156e1535..f2781ca4276 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1057,15 +1057,11 @@ int nfs_open(struct inode *inode, struct file *filp) ctx->mode = filp->f_mode; nfs_file_set_open_context(filp, ctx); put_nfs_open_context(ctx); - if ((filp->f_mode & FMODE_WRITE) != 0) - nfs_begin_data_update(inode); return 0; } int nfs_release(struct inode *inode, struct file *filp) { - if ((filp->f_mode & FMODE_WRITE) != 0) - nfs_end_data_update(inode); nfs_file_clear_open_context(filp); return 0; } -- cgit v1.2.3 From 20e5c81fcff89535dced2ed71cf24c6c648ff40e Mon Sep 17 00:00:00 2001 From: "Chen, Kenneth W" Date: Thu, 13 Oct 2005 21:48:42 +0200 Subject: [patch] remove gendisk->stamp_idle field struct gendisk has these two fields: stamp, stamp_idle. Update to stamp_idle is always in sync with stamp and they are always the same. Therefore, it does not add any value in having two fields tracking same timestamp. Suggest to remove it. Also, we should only update gendisk stats with non-zero value. Advantage is that we don't have to needlessly calculate memory address, and then add zero to the content. Signed-off-by: Ken Chen Signed-off-by: Jens Axboe --- fs/partitions/check.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/partitions/check.c b/fs/partitions/check.c index 77e178f1316..1e848648a32 100644 --- a/fs/partitions/check.c +++ b/fs/partitions/check.c @@ -430,7 +430,7 @@ void del_gendisk(struct gendisk *disk) disk->flags &= ~GENHD_FL_UP; unlink_gendisk(disk); disk_stat_set_all(disk, 0); - disk->stamp = disk->stamp_idle = 0; + disk->stamp = 0; devfs_remove_disk(disk); -- cgit v1.2.3 From af4ca457eaf2d6682059c18463eb106e2ce58198 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Fri, 21 Oct 2005 02:55:38 -0400 Subject: [PATCH] gfp_t: infrastructure Beginning of gfp_t annotations: - -Wbitwise added to CHECKFLAGS - old __bitwise renamed to __bitwise__ - __bitwise defined to either __bitwise__ or nothing, depending on __CHECK_ENDIAN__ being defined - gfp_t switched from __nocast to __bitwise__ - force cast to gfp_t added to __GFP_... constants - new helper - gfp_zone(); extracts zone bits out of gfp_t value and casts the result to int Signed-off-by: Al Viro Signed-off-by: Linus Torvalds --- fs/buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/buffer.c b/fs/buffer.c index 1216c0d3c8c..9657696fd6d 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -502,7 +502,7 @@ static void free_more_memory(void) yield(); for_each_pgdat(pgdat) { - zones = pgdat->node_zonelists[GFP_NOFS&GFP_ZONEMASK].zones; + zones = pgdat->node_zonelists[gfp_zone(GFP_NOFS)].zones; if (*zones) try_to_free_pages(zones, GFP_NOFS); } -- cgit v1.2.3 From 27496a8c67bef4d789d8e3c8317ca35813a507ae Mon Sep 17 00:00:00 2001 From: Al Viro Date: Fri, 21 Oct 2005 03:20:48 -0400 Subject: [PATCH] gfp_t: fs/* - ->releasepage() annotated (s/int/gfp_t), instances updated - missing gfp_t in fs/* added - fixed misannotation from the original sweep caught by bitwise checks: XFS used __nocast both for gfp_t and for flags used by XFS allocator. The latter left with unsigned int __nocast; we might want to add a different type for those but for now let's leave them alone. That, BTW, is a case when __nocast use had been actively confusing - it had been used in the same code for two different and similar types, with no way to catch misuses. Switch of gfp_t to bitwise had caught that immediately... One tricky bit is left alone to be dealt with later - mapping->flags is a mix of gfp_t and error indications. Left alone for now. Signed-off-by: Al Viro Signed-off-by: Linus Torvalds --- fs/afs/file.c | 4 ++-- fs/bio.c | 4 ++-- fs/buffer.c | 2 +- fs/dcache.c | 2 +- fs/dquot.c | 2 +- fs/ext3/inode.c | 2 +- fs/hfs/inode.c | 2 +- fs/hfsplus/inode.c | 2 +- fs/inode.c | 2 +- fs/jbd/journal.c | 2 +- fs/jbd/transaction.c | 2 +- fs/jfs/jfs_metapage.c | 4 ++-- fs/mbcache.c | 6 +++--- fs/reiserfs/fix_node.c | 2 +- fs/reiserfs/inode.c | 2 +- fs/xfs/linux-2.6/kmem.c | 22 +++++++++++----------- fs/xfs/linux-2.6/kmem.h | 18 +++++++++--------- fs/xfs/linux-2.6/xfs_aops.c | 2 +- fs/xfs/linux-2.6/xfs_buf.c | 8 ++++---- 19 files changed, 45 insertions(+), 45 deletions(-) (limited to 'fs') diff --git a/fs/afs/file.c b/fs/afs/file.c index 23c12512802..0d576987ec6 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -29,7 +29,7 @@ static int afs_file_release(struct inode *inode, struct file *file); static int afs_file_readpage(struct file *file, struct page *page); static int afs_file_invalidatepage(struct page *page, unsigned long offset); -static int afs_file_releasepage(struct page *page, int gfp_flags); +static int afs_file_releasepage(struct page *page, gfp_t gfp_flags); static ssize_t afs_file_write(struct file *file, const char __user *buf, size_t size, loff_t *off); @@ -279,7 +279,7 @@ static int afs_file_invalidatepage(struct page *page, unsigned long offset) /* * release a page and cleanup its private data */ -static int afs_file_releasepage(struct page *page, int gfp_flags) +static int afs_file_releasepage(struct page *page, gfp_t gfp_flags) { struct cachefs_page *pageio; diff --git a/fs/bio.c b/fs/bio.c index 7d81a93afd4..460554b07ff 100644 --- a/fs/bio.c +++ b/fs/bio.c @@ -778,7 +778,7 @@ static int bio_map_kern_endio(struct bio *bio, unsigned int bytes_done, int err) static struct bio *__bio_map_kern(request_queue_t *q, void *data, - unsigned int len, unsigned int gfp_mask) + unsigned int len, gfp_t gfp_mask) { unsigned long kaddr = (unsigned long)data; unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; @@ -825,7 +825,7 @@ static struct bio *__bio_map_kern(request_queue_t *q, void *data, * device. Returns an error pointer in case of error. */ struct bio *bio_map_kern(request_queue_t *q, void *data, unsigned int len, - unsigned int gfp_mask) + gfp_t gfp_mask) { struct bio *bio; diff --git a/fs/buffer.c b/fs/buffer.c index 9657696fd6d..b1667986442 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -1571,7 +1571,7 @@ static inline void discard_buffer(struct buffer_head * bh) * * NOTE: @gfp_mask may go away, and this function may become non-blocking. */ -int try_to_release_page(struct page *page, int gfp_mask) +int try_to_release_page(struct page *page, gfp_t gfp_mask) { struct address_space * const mapping = page->mapping; diff --git a/fs/dcache.c b/fs/dcache.c index fb10386c59b..e90512ed35a 100644 --- a/fs/dcache.c +++ b/fs/dcache.c @@ -689,7 +689,7 @@ void shrink_dcache_anon(struct hlist_head *head) * * In this case we return -1 to tell the caller that we baled. */ -static int shrink_dcache_memory(int nr, unsigned int gfp_mask) +static int shrink_dcache_memory(int nr, gfp_t gfp_mask) { if (nr) { if (!(gfp_mask & __GFP_FS)) diff --git a/fs/dquot.c b/fs/dquot.c index b9732335bcd..05f3327d64a 100644 --- a/fs/dquot.c +++ b/fs/dquot.c @@ -500,7 +500,7 @@ static void prune_dqcache(int count) * more memory */ -static int shrink_dqcache_memory(int nr, unsigned int gfp_mask) +static int shrink_dqcache_memory(int nr, gfp_t gfp_mask) { if (nr) { spin_lock(&dq_list_lock); diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c index b5177c90d6f..8b38f223279 100644 --- a/fs/ext3/inode.c +++ b/fs/ext3/inode.c @@ -1434,7 +1434,7 @@ static int ext3_invalidatepage(struct page *page, unsigned long offset) return journal_invalidatepage(journal, page, offset); } -static int ext3_releasepage(struct page *page, int wait) +static int ext3_releasepage(struct page *page, gfp_t wait) { journal_t *journal = EXT3_JOURNAL(page->mapping->host); diff --git a/fs/hfs/inode.c b/fs/hfs/inode.c index f1570b9f9de..3f680c5675b 100644 --- a/fs/hfs/inode.c +++ b/fs/hfs/inode.c @@ -46,7 +46,7 @@ static sector_t hfs_bmap(struct address_space *mapping, sector_t block) return generic_block_bmap(mapping, block, hfs_get_block); } -static int hfs_releasepage(struct page *page, int mask) +static int hfs_releasepage(struct page *page, gfp_t mask) { struct inode *inode = page->mapping->host; struct super_block *sb = inode->i_sb; diff --git a/fs/hfsplus/inode.c b/fs/hfsplus/inode.c index d5642705f63..f205773ddfb 100644 --- a/fs/hfsplus/inode.c +++ b/fs/hfsplus/inode.c @@ -40,7 +40,7 @@ static sector_t hfsplus_bmap(struct address_space *mapping, sector_t block) return generic_block_bmap(mapping, block, hfsplus_get_block); } -static int hfsplus_releasepage(struct page *page, int mask) +static int hfsplus_releasepage(struct page *page, gfp_t mask) { struct inode *inode = page->mapping->host; struct super_block *sb = inode->i_sb; diff --git a/fs/inode.c b/fs/inode.c index f80a79ff156..7d331652776 100644 --- a/fs/inode.c +++ b/fs/inode.c @@ -475,7 +475,7 @@ static void prune_icache(int nr_to_scan) * This function is passed the number of inodes to scan, and it returns the * total number of remaining possibly-reclaimable inodes. */ -static int shrink_icache_memory(int nr, unsigned int gfp_mask) +static int shrink_icache_memory(int nr, gfp_t gfp_mask) { if (nr) { /* diff --git a/fs/jbd/journal.c b/fs/jbd/journal.c index 7ae2c4fe506..e4b516ac498 100644 --- a/fs/jbd/journal.c +++ b/fs/jbd/journal.c @@ -1606,7 +1606,7 @@ int journal_blocks_per_page(struct inode *inode) * Simple support for retrying memory allocations. Introduced to help to * debug different VM deadlock avoidance strategies. */ -void * __jbd_kmalloc (const char *where, size_t size, int flags, int retry) +void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry) { return kmalloc(size, flags | (retry ? __GFP_NOFAIL : 0)); } diff --git a/fs/jbd/transaction.c b/fs/jbd/transaction.c index 49bbc2be3d7..13cb05bf604 100644 --- a/fs/jbd/transaction.c +++ b/fs/jbd/transaction.c @@ -1621,7 +1621,7 @@ out: * while the data is part of a transaction. Yes? */ int journal_try_to_free_buffers(journal_t *journal, - struct page *page, int unused_gfp_mask) + struct page *page, gfp_t unused_gfp_mask) { struct buffer_head *head; struct buffer_head *bh; diff --git a/fs/jfs/jfs_metapage.c b/fs/jfs/jfs_metapage.c index 13d7e3f1feb..eeb37d70e65 100644 --- a/fs/jfs/jfs_metapage.c +++ b/fs/jfs/jfs_metapage.c @@ -198,7 +198,7 @@ static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags) } } -static inline struct metapage *alloc_metapage(unsigned int gfp_mask) +static inline struct metapage *alloc_metapage(gfp_t gfp_mask) { return mempool_alloc(metapage_mempool, gfp_mask); } @@ -534,7 +534,7 @@ add_failed: return -EIO; } -static int metapage_releasepage(struct page *page, int gfp_mask) +static int metapage_releasepage(struct page *page, gfp_t gfp_mask) { struct metapage *mp; int busy = 0; diff --git a/fs/mbcache.c b/fs/mbcache.c index b002a088857..298997f1747 100644 --- a/fs/mbcache.c +++ b/fs/mbcache.c @@ -116,7 +116,7 @@ mb_cache_indexes(struct mb_cache *cache) * What the mbcache registers as to get shrunk dynamically. */ -static int mb_cache_shrink_fn(int nr_to_scan, unsigned int gfp_mask); +static int mb_cache_shrink_fn(int nr_to_scan, gfp_t gfp_mask); static inline int @@ -140,7 +140,7 @@ __mb_cache_entry_unhash(struct mb_cache_entry *ce) static inline void -__mb_cache_entry_forget(struct mb_cache_entry *ce, int gfp_mask) +__mb_cache_entry_forget(struct mb_cache_entry *ce, gfp_t gfp_mask) { struct mb_cache *cache = ce->e_cache; @@ -193,7 +193,7 @@ forget: * Returns the number of objects which are present in the cache. */ static int -mb_cache_shrink_fn(int nr_to_scan, unsigned int gfp_mask) +mb_cache_shrink_fn(int nr_to_scan, gfp_t gfp_mask) { LIST_HEAD(free_list); struct list_head *l, *ltmp; diff --git a/fs/reiserfs/fix_node.c b/fs/reiserfs/fix_node.c index 2706e2adffa..45829889dcd 100644 --- a/fs/reiserfs/fix_node.c +++ b/fs/reiserfs/fix_node.c @@ -2022,7 +2022,7 @@ static int get_neighbors(struct tree_balance *p_s_tb, int n_h) } #ifdef CONFIG_REISERFS_CHECK -void *reiserfs_kmalloc(size_t size, int flags, struct super_block *s) +void *reiserfs_kmalloc(size_t size, gfp_t flags, struct super_block *s) { void *vp; static size_t malloced; diff --git a/fs/reiserfs/inode.c b/fs/reiserfs/inode.c index d76ee6c4f9b..5f82352b97e 100644 --- a/fs/reiserfs/inode.c +++ b/fs/reiserfs/inode.c @@ -2842,7 +2842,7 @@ static int reiserfs_set_page_dirty(struct page *page) * even in -o notail mode, we can't be sure an old mount without -o notail * didn't create files with tails. */ -static int reiserfs_releasepage(struct page *page, int unused_gfp_flags) +static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags) { struct inode *inode = page->mapping->host; struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb); diff --git a/fs/xfs/linux-2.6/kmem.c b/fs/xfs/linux-2.6/kmem.c index d2653b589b1..3c92162dc72 100644 --- a/fs/xfs/linux-2.6/kmem.c +++ b/fs/xfs/linux-2.6/kmem.c @@ -45,11 +45,11 @@ void * -kmem_alloc(size_t size, gfp_t flags) +kmem_alloc(size_t size, unsigned int __nocast flags) { - int retries = 0; - unsigned int lflags = kmem_flags_convert(flags); - void *ptr; + int retries = 0; + gfp_t lflags = kmem_flags_convert(flags); + void *ptr; do { if (size < MAX_SLAB_SIZE || retries > MAX_VMALLOCS) @@ -67,7 +67,7 @@ kmem_alloc(size_t size, gfp_t flags) } void * -kmem_zalloc(size_t size, gfp_t flags) +kmem_zalloc(size_t size, unsigned int __nocast flags) { void *ptr; @@ -90,7 +90,7 @@ kmem_free(void *ptr, size_t size) void * kmem_realloc(void *ptr, size_t newsize, size_t oldsize, - gfp_t flags) + unsigned int __nocast flags) { void *new; @@ -105,11 +105,11 @@ kmem_realloc(void *ptr, size_t newsize, size_t oldsize, } void * -kmem_zone_alloc(kmem_zone_t *zone, gfp_t flags) +kmem_zone_alloc(kmem_zone_t *zone, unsigned int __nocast flags) { - int retries = 0; - unsigned int lflags = kmem_flags_convert(flags); - void *ptr; + int retries = 0; + gfp_t lflags = kmem_flags_convert(flags); + void *ptr; do { ptr = kmem_cache_alloc(zone, lflags); @@ -124,7 +124,7 @@ kmem_zone_alloc(kmem_zone_t *zone, gfp_t flags) } void * -kmem_zone_zalloc(kmem_zone_t *zone, gfp_t flags) +kmem_zone_zalloc(kmem_zone_t *zone, unsigned int __nocast flags) { void *ptr; diff --git a/fs/xfs/linux-2.6/kmem.h b/fs/xfs/linux-2.6/kmem.h index ee7010f085b..f4bb78c268c 100644 --- a/fs/xfs/linux-2.6/kmem.h +++ b/fs/xfs/linux-2.6/kmem.h @@ -81,9 +81,9 @@ typedef unsigned long xfs_pflags_t; *(NSTATEP) = *(OSTATEP); \ } while (0) -static __inline unsigned int kmem_flags_convert(gfp_t flags) +static __inline gfp_t kmem_flags_convert(unsigned int __nocast flags) { - unsigned int lflags = __GFP_NOWARN; /* we'll report problems, if need be */ + gfp_t lflags = __GFP_NOWARN; /* we'll report problems, if need be */ #ifdef DEBUG if (unlikely(flags & ~(KM_SLEEP|KM_NOSLEEP|KM_NOFS|KM_MAYFAIL))) { @@ -125,16 +125,16 @@ kmem_zone_destroy(kmem_zone_t *zone) BUG(); } -extern void *kmem_zone_zalloc(kmem_zone_t *, gfp_t); -extern void *kmem_zone_alloc(kmem_zone_t *, gfp_t); +extern void *kmem_zone_zalloc(kmem_zone_t *, unsigned int __nocast); +extern void *kmem_zone_alloc(kmem_zone_t *, unsigned int __nocast); -extern void *kmem_alloc(size_t, gfp_t); -extern void *kmem_realloc(void *, size_t, size_t, gfp_t); -extern void *kmem_zalloc(size_t, gfp_t); +extern void *kmem_alloc(size_t, unsigned int __nocast); +extern void *kmem_realloc(void *, size_t, size_t, unsigned int __nocast); +extern void *kmem_zalloc(size_t, unsigned int __nocast); extern void kmem_free(void *, size_t); typedef struct shrinker *kmem_shaker_t; -typedef int (*kmem_shake_func_t)(int, unsigned int); +typedef int (*kmem_shake_func_t)(int, gfp_t); static __inline kmem_shaker_t kmem_shake_register(kmem_shake_func_t sfunc) @@ -149,7 +149,7 @@ kmem_shake_deregister(kmem_shaker_t shrinker) } static __inline int -kmem_shake_allow(unsigned int gfp_mask) +kmem_shake_allow(gfp_t gfp_mask) { return (gfp_mask & __GFP_WAIT); } diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c index c6c077978fe..7aa39872470 100644 --- a/fs/xfs/linux-2.6/xfs_aops.c +++ b/fs/xfs/linux-2.6/xfs_aops.c @@ -1296,7 +1296,7 @@ linvfs_invalidate_page( STATIC int linvfs_release_page( struct page *page, - int gfp_mask) + gfp_t gfp_mask) { struct inode *inode = page->mapping->host; int dirty, delalloc, unmapped, unwritten; diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index e82cf72ac59..ba4767c04ad 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -64,7 +64,7 @@ STATIC kmem_cache_t *pagebuf_zone; STATIC kmem_shaker_t pagebuf_shake; -STATIC int xfsbufd_wakeup(int, unsigned int); +STATIC int xfsbufd_wakeup(int, gfp_t); STATIC void pagebuf_delwri_queue(xfs_buf_t *, int); STATIC struct workqueue_struct *xfslogd_workqueue; @@ -383,7 +383,7 @@ _pagebuf_lookup_pages( size_t blocksize = bp->pb_target->pbr_bsize; size_t size = bp->pb_count_desired; size_t nbytes, offset; - int gfp_mask = pb_to_gfp(flags); + gfp_t gfp_mask = pb_to_gfp(flags); unsigned short page_count, i; pgoff_t first; loff_t end; @@ -1749,8 +1749,8 @@ STATIC int xfsbufd_force_sleep; STATIC int xfsbufd_wakeup( - int priority, - unsigned int mask) + int priority, + gfp_t mask) { if (xfsbufd_force_sleep) return 0; -- cgit v1.2.3 From c4cdd038318863e912e9b992489f61497f98b442 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Fri, 21 Oct 2005 03:22:39 -0400 Subject: [PATCH] gfp_t: reiserfs mapping_set_gfp_mask() use Signed-off-by: Al Viro Signed-off-by: Linus Torvalds --- fs/reiserfs/xattr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/reiserfs/xattr.c b/fs/reiserfs/xattr.c index 87ac9dc8b38..72e12079867 100644 --- a/fs/reiserfs/xattr.c +++ b/fs/reiserfs/xattr.c @@ -453,7 +453,7 @@ static struct page *reiserfs_get_page(struct inode *dir, unsigned long n) struct page *page; /* We can deadlock if we try to free dentries, and an unlink/rmdir has just occured - GFP_NOFS avoids this */ - mapping->flags = (mapping->flags & ~__GFP_BITS_MASK) | GFP_NOFS; + mapping_set_gfp_mask(mapping, GFP_NOFS); page = read_cache_page(mapping, n, (filler_t *) mapping->a_ops->readpage, NULL); if (!IS_ERR(page)) { -- cgit v1.2.3 From a7fd67062efc5b0fc9a61368c607fa92d1d57f9e Mon Sep 17 00:00:00 2001 From: Kay Sievers Date: Sat, 1 Oct 2005 14:49:43 +0200 Subject: [PATCH] add sysfs attr to re-emit device hotplug event A "coldplug + udevstart" can be simple like this: for i in /sys/block/*/*/uevent; do echo 1 > $i; done for i in /sys/class/*/*/uevent; do echo 1 > $i; done for i in /sys/bus/*/devices/*/uevent; do echo 1 > $i; done Signed-off-by: Kay Sievers Signed-off-by: Greg Kroah-Hartman --- fs/partitions/check.c | 27 ++++++++++++++++++++++++++- 1 file changed, 26 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/partitions/check.c b/fs/partitions/check.c index 77e178f1316..d95a110293f 100644 --- a/fs/partitions/check.c +++ b/fs/partitions/check.c @@ -192,6 +192,7 @@ check_partition(struct gendisk *hd, struct block_device *bdev) struct part_attribute { struct attribute attr; ssize_t (*show)(struct hd_struct *,char *); + ssize_t (*store)(struct hd_struct *,const char *, size_t); }; static ssize_t @@ -201,14 +202,33 @@ part_attr_show(struct kobject * kobj, struct attribute * attr, char * page) struct part_attribute * part_attr = container_of(attr,struct part_attribute,attr); ssize_t ret = 0; if (part_attr->show) - ret = part_attr->show(p,page); + ret = part_attr->show(p, page); + return ret; +} +static ssize_t +part_attr_store(struct kobject * kobj, struct attribute * attr, + const char *page, size_t count) +{ + struct hd_struct * p = container_of(kobj,struct hd_struct,kobj); + struct part_attribute * part_attr = container_of(attr,struct part_attribute,attr); + ssize_t ret = 0; + + if (part_attr->store) + ret = part_attr->store(p, page, count); return ret; } static struct sysfs_ops part_sysfs_ops = { .show = part_attr_show, + .store = part_attr_store, }; +static ssize_t part_uevent_store(struct hd_struct * p, + const char *page, size_t count) +{ + kobject_hotplug(&p->kobj, KOBJ_ADD); + return count; +} static ssize_t part_dev_read(struct hd_struct * p, char *page) { struct gendisk *disk = container_of(p->kobj.parent,struct gendisk,kobj); @@ -229,6 +249,10 @@ static ssize_t part_stat_read(struct hd_struct * p, char *page) p->reads, (unsigned long long)p->read_sectors, p->writes, (unsigned long long)p->write_sectors); } +static struct part_attribute part_attr_uevent = { + .attr = {.name = "uevent", .mode = S_IWUSR }, + .store = part_uevent_store +}; static struct part_attribute part_attr_dev = { .attr = {.name = "dev", .mode = S_IRUGO }, .show = part_dev_read @@ -247,6 +271,7 @@ static struct part_attribute part_attr_stat = { }; static struct attribute * default_attrs[] = { + &part_attr_uevent.attr, &part_attr_dev.attr, &part_attr_start.attr, &part_attr_size.attr, -- cgit v1.2.3 From 53f4654272df7c51064825024340554b39c9efba Mon Sep 17 00:00:00 2001 From: Greg Kroah-Hartman Date: Thu, 27 Oct 2005 22:25:43 -0700 Subject: [PATCH] Driver Core: fix up all callers of class_device_create() The previous patch adding the ability to nest struct class_device changed the paramaters to the call class_device_create(). This patch fixes up all in-kernel users of the function. Signed-off-by: Greg Kroah-Hartman --- fs/coda/psdev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/coda/psdev.c b/fs/coda/psdev.c index 3d1cce3653b..6a3df88accf 100644 --- a/fs/coda/psdev.c +++ b/fs/coda/psdev.c @@ -370,8 +370,8 @@ static int init_coda_psdev(void) } devfs_mk_dir ("coda"); for (i = 0; i < MAX_CODADEVS; i++) { - class_device_create(coda_psdev_class, MKDEV(CODA_PSDEV_MAJOR,i), - NULL, "cfs%d", i); + class_device_create(coda_psdev_class, NULL, + MKDEV(CODA_PSDEV_MAJOR,i), NULL, "cfs%d", i); err = devfs_mk_cdev(MKDEV(CODA_PSDEV_MAJOR, i), S_IFCHR|S_IRUSR|S_IWUSR, "coda/%d", i); if (err) -- cgit v1.2.3 From 7038f1cbac899654cf0515e60dbe3e44d58271de Mon Sep 17 00:00:00 2001 From: Dave Kleikamp Date: Fri, 28 Oct 2005 13:27:40 -0500 Subject: JFS: make sure right-most xtree pages have header.next set to zero The xtTruncate code was only doing this for leaf pages. When a file is horribly fragmented, we may truncate a file leaving an internal page with an invalid head.next field, which may cause a stale page to be referenced. Signed-off-by: Dave Kleikamp --- fs/jfs/jfs_xtree.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/jfs/jfs_xtree.c b/fs/jfs/jfs_xtree.c index a7fe2f2b969..e72f4ebb6e9 100644 --- a/fs/jfs/jfs_xtree.c +++ b/fs/jfs/jfs_xtree.c @@ -3516,16 +3516,10 @@ s64 xtTruncate(tid_t tid, struct inode *ip, s64 newsize, int flag) /* process entries backward from last index */ index = le16_to_cpu(p->header.nextindex) - 1; - if (p->header.flag & BT_INTERNAL) - goto getChild; - - /* - * leaf page - */ - /* Since this is the rightmost leaf, and we may have already freed - * a page that was formerly to the right, let's make sure that the - * next pointer is zero. + /* Since this is the rightmost page at this level, and we may have + * already freed a page that was formerly to the right, let's make + * sure that the next pointer is zero. */ if (p->header.next) { if (log) @@ -3539,6 +3533,12 @@ s64 xtTruncate(tid_t tid, struct inode *ip, s64 newsize, int flag) p->header.next = 0; } + if (p->header.flag & BT_INTERNAL) + goto getChild; + + /* + * leaf page + */ freed = 0; /* does region covered by leaf page precede Teof ? */ -- cgit v1.2.3 From 6cd37cda7ed117d5a13d9b69aeded57b4fd6de14 Mon Sep 17 00:00:00 2001 From: Peter Osterlund Date: Fri, 28 Oct 2005 20:23:39 +0200 Subject: [PATCH] Fix ext3 warning for unused var Fix compile warning in ext3 quota code. Signed-off-by: Peter Osterlund Signed-off-by: Linus Torvalds --- fs/ext3/super.c | 26 ++++++++++++++++---------- 1 file changed, 16 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/ext3/super.c b/fs/ext3/super.c index 9e24ceb019f..097383c1115 100644 --- a/fs/ext3/super.c +++ b/fs/ext3/super.c @@ -510,19 +510,11 @@ static void ext3_clear_inode(struct inode *inode) kfree(rsv); } -static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs) +static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb) { - struct super_block *sb = vfs->mnt_sb; +#if defined(CONFIG_QUOTA) struct ext3_sb_info *sbi = EXT3_SB(sb); - if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA) - seq_puts(seq, ",data=journal"); - else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA) - seq_puts(seq, ",data=ordered"); - else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA) - seq_puts(seq, ",data=writeback"); - -#if defined(CONFIG_QUOTA) if (sbi->s_jquota_fmt) seq_printf(seq, ",jqfmt=%s", (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0"); @@ -539,6 +531,20 @@ static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs) if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) seq_puts(seq, ",grpquota"); #endif +} + +static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs) +{ + struct super_block *sb = vfs->mnt_sb; + + if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA) + seq_puts(seq, ",data=journal"); + else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA) + seq_puts(seq, ",data=ordered"); + else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA) + seq_puts(seq, ",data=writeback"); + + ext3_show_quota_options(seq, sb); return 0; } -- cgit v1.2.3 From c36fc889b5a4fd66cfd9ba80d9e038745d349567 Mon Sep 17 00:00:00 2001 From: Pete Zaitcev Date: Mon, 17 Oct 2005 18:15:54 -0700 Subject: [PATCH] usb: Patch for USBDEVFS_IOCTL from 32-bit programs Dell supplied me with the following test: #include #include #include #include #include main(int argc,char*argv[]) { struct usbdevfs_hub_portinfo hubPortInfo = {0}; struct usbdevfs_ioctl command = {0}; command.ifno = 0; command.ioctl_code = USBDEVFS_HUB_PORTINFO; command.data = (void*)&hubPortInfo; int fd, ret; if(argc != 2) { fprintf(stderr,"Usage: %s /proc/bus/usb//\n",argv[0]); fprintf(stderr,"Example: %s /proc/bus/usb/001/001\n",argv[0]); exit(1); } errno = 0; fd = open(argv[1],O_RDWR); if(fd < 0) { perror("open failed:"); exit(errno); } errno = 0; ret = ioctl(fd,USBDEVFS_IOCTL,&command); printf("IOCTL return status:%d\n",ret); if(ret<0) { perror("IOCTL failed:"); close(fd); exit(3); } else { printf("IOCTL passed:Num of ports %d\n",hubPortInfo.nports); close(fd); exit(0); } return 0; } I have verified that it breaks if built in 32 bit mode on x86_64 and that the patch below fixes it. Signed-off-by: Pete Zaitcev Signed-off-by: Greg Kroah-Hartman --- fs/compat_ioctl.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs') diff --git a/fs/compat_ioctl.c b/fs/compat_ioctl.c index e28a74203f3..a327e03753a 100644 --- a/fs/compat_ioctl.c +++ b/fs/compat_ioctl.c @@ -3050,6 +3050,7 @@ HANDLE_IOCTL(TIOCSSERIAL, serial_struct_ioctl) HANDLE_IOCTL(USBDEVFS_CONTROL32, do_usbdevfs_control) HANDLE_IOCTL(USBDEVFS_BULK32, do_usbdevfs_bulk) HANDLE_IOCTL(USBDEVFS_DISCSIGNAL32, do_usbdevfs_discsignal) +COMPATIBLE_IOCTL(USBDEVFS_IOCTL32) /* i2c */ HANDLE_IOCTL(I2C_FUNCS, w_long) HANDLE_IOCTL(I2C_RDWR, do_i2c_rdwr_ioctl) -- cgit v1.2.3 From dfcd3c0dc426bb75770c34b40e14f2da8845ea62 Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Sat, 29 Oct 2005 18:15:48 -0700 Subject: [PATCH] Convert mempolicies to nodemask_t The NUMA policy code predated nodemask_t so it used open coded bitmaps. Convert everything to nodemask_t. Big patch, but shouldn't have any actual behaviour changes (except I removed one unnecessary check against node_online_map and one unnecessary BUG_ON) Signed-off-by: "Andi Kleen" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/proc/task_mmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index c7ef3e48e35..994612bc72d 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -469,7 +469,7 @@ static int show_numa_map(struct seq_file *m, void *v) seq_printf(m, " interleave={"); first = 1; for_each_node(n) { - if (test_bit(n, pol->v.nodes)) { + if (node_isset(n, pol->v.nodes)) { if (!first) seq_putc(m,','); else -- cgit v1.2.3 From 404351e67a9facb475abf1492245374a28d13e90 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:04 -0700 Subject: [PATCH] mm: mm_init set_mm_counters How is anon_rss initialized? In dup_mmap, and by mm_alloc's memset; but that's not so good if an mm_counter_t is a special type. And how is rss initialized? By set_mm_counter, all over the place. Come on, we just need to initialize them both at once by set_mm_counter in mm_init (which follows the memcpy when forking). Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_aout.c | 1 - fs/binfmt_elf.c | 1 - fs/binfmt_elf_fdpic.c | 7 ------- fs/binfmt_flat.c | 1 - fs/binfmt_som.c | 1 - 5 files changed, 11 deletions(-) (limited to 'fs') diff --git a/fs/binfmt_aout.c b/fs/binfmt_aout.c index dd9baabaf01..72011826f0c 100644 --- a/fs/binfmt_aout.c +++ b/fs/binfmt_aout.c @@ -318,7 +318,6 @@ static int load_aout_binary(struct linux_binprm * bprm, struct pt_regs * regs) current->mm->free_area_cache = current->mm->mmap_base; current->mm->cached_hole_size = 0; - set_mm_counter(current->mm, rss, 0); current->mm->mmap = NULL; compute_creds(bprm); current->flags &= ~PF_FORKNOEXEC; diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index d4b15576e58..918ccc267e4 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -773,7 +773,6 @@ static int load_elf_binary(struct linux_binprm * bprm, struct pt_regs * regs) /* Do this so that we can load the interpreter, if need be. We will change some of these later */ - set_mm_counter(current->mm, rss, 0); current->mm->free_area_cache = current->mm->mmap_base; current->mm->cached_hole_size = 0; retval = setup_arg_pages(bprm, randomize_stack_top(STACK_TOP), diff --git a/fs/binfmt_elf_fdpic.c b/fs/binfmt_elf_fdpic.c index 134c9c0d1f5..dda87c4c82a 100644 --- a/fs/binfmt_elf_fdpic.c +++ b/fs/binfmt_elf_fdpic.c @@ -294,14 +294,7 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm, struct pt_regs *regs &interp_params, ¤t->mm->start_stack, ¤t->mm->start_brk); -#endif - - /* do this so that we can load the interpreter, if need be - * - we will change some of these later - */ - set_mm_counter(current->mm, rss, 0); -#ifdef CONFIG_MMU retval = setup_arg_pages(bprm, current->mm->start_stack, executable_stack); if (retval < 0) { send_sig(SIGKILL, current, 0); diff --git a/fs/binfmt_flat.c b/fs/binfmt_flat.c index 7974efa107b..9d6625829b9 100644 --- a/fs/binfmt_flat.c +++ b/fs/binfmt_flat.c @@ -650,7 +650,6 @@ static int load_flat_file(struct linux_binprm * bprm, current->mm->start_brk = datapos + data_len + bss_len; current->mm->brk = (current->mm->start_brk + 3) & ~3; current->mm->context.end_brk = memp + ksize((void *) memp) - stack_len; - set_mm_counter(current->mm, rss, 0); } if (flags & FLAT_FLAG_KTRACE) diff --git a/fs/binfmt_som.c b/fs/binfmt_som.c index 227a2682d2b..00a91dc25d1 100644 --- a/fs/binfmt_som.c +++ b/fs/binfmt_som.c @@ -259,7 +259,6 @@ load_som_binary(struct linux_binprm * bprm, struct pt_regs * regs) create_som_tables(bprm); current->mm->start_stack = bprm->p; - set_mm_counter(current->mm, rss, 0); #if 0 printk("(start_brk) %08lx\n" , (unsigned long) current->mm->start_brk); -- cgit v1.2.3 From 4294621f41a85497019fae64341aa5351a1921b7 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:05 -0700 Subject: [PATCH] mm: rss = file_rss + anon_rss I was lazy when we added anon_rss, and chose to change as few places as possible. So currently each anonymous page has to be counted twice, in rss and in anon_rss. Which won't be so good if those are atomic counts in some configurations. Change that around: keep file_rss and anon_rss separately, and add them together (with get_mm_rss macro) when the total is needed - reading two atomics is much cheaper than updating two atomics. And update anon_rss upfront, typically in memory.c, not tucked away in page_add_anon_rmap. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/exec.c | 2 +- fs/proc/array.c | 2 +- fs/proc/task_mmu.c | 8 +++----- 3 files changed, 5 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/exec.c b/fs/exec.c index d2208f7c87d..cefadf5ab83 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -330,7 +330,7 @@ void install_arg_page(struct vm_area_struct *vma, pte_unmap(pte); goto out; } - inc_mm_counter(mm, rss); + inc_mm_counter(mm, anon_rss); lru_cache_add_active(page); set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte( page, vma->vm_page_prot)))); diff --git a/fs/proc/array.c b/fs/proc/array.c index d84eecacbea..3e1239e4b30 100644 --- a/fs/proc/array.c +++ b/fs/proc/array.c @@ -438,7 +438,7 @@ static int do_task_stat(struct task_struct *task, char * buffer, int whole) jiffies_to_clock_t(it_real_value), start_time, vsize, - mm ? get_mm_counter(mm, rss) : 0, /* you might want to shift this left 3 */ + mm ? get_mm_rss(mm) : 0, rsslim, mm ? mm->start_code : 0, mm ? mm->end_code : 0, diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index 994612bc72d..bccee7cf9cc 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -29,7 +29,7 @@ char *task_mem(struct mm_struct *mm, char *buffer) "VmPTE:\t%8lu kB\n", (mm->total_vm - mm->reserved_vm) << (PAGE_SHIFT-10), mm->locked_vm << (PAGE_SHIFT-10), - get_mm_counter(mm, rss) << (PAGE_SHIFT-10), + get_mm_rss(mm) << (PAGE_SHIFT-10), data << (PAGE_SHIFT-10), mm->stack_vm << (PAGE_SHIFT-10), text, lib, (PTRS_PER_PTE*sizeof(pte_t)*mm->nr_ptes) >> 10); @@ -44,13 +44,11 @@ unsigned long task_vsize(struct mm_struct *mm) int task_statm(struct mm_struct *mm, int *shared, int *text, int *data, int *resident) { - int rss = get_mm_counter(mm, rss); - - *shared = rss - get_mm_counter(mm, anon_rss); + *shared = get_mm_counter(mm, file_rss); *text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> PAGE_SHIFT; *data = mm->total_vm - mm->shared_vm; - *resident = rss; + *resident = *shared + get_mm_counter(mm, anon_rss); return mm->total_vm; } -- cgit v1.2.3 From b5810039a54e5babf428e9a1e89fc1940fabff11 Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Sat, 29 Oct 2005 18:16:12 -0700 Subject: [PATCH] core remove PageReserved Remove PageReserved() calls from core code by tightening VM_RESERVED handling in mm/ to cover PageReserved functionality. PageReserved special casing is removed from get_page and put_page. All setting and clearing of PageReserved is retained, and it is now flagged in the page_alloc checks to help ensure we don't introduce any refcount based freeing of Reserved pages. MAP_PRIVATE, PROT_WRITE of VM_RESERVED regions is tentatively being deprecated. We never completely handled it correctly anyway, and is be reintroduced in future if required (Hugh has a proof of concept). Once PageReserved() calls are removed from kernel/power/swsusp.c, and all arch/ and driver code, the Set and Clear calls, and the PG_reserved bit can be trivially removed. Last real user of PageReserved is swsusp, which uses PageReserved to determine whether a struct page points to valid memory or not. This still needs to be addressed (a generic page_is_ram() should work). A last caveat: the ZERO_PAGE is now refcounted and managed with rmap (and thus mapcounted and count towards shared rss). These writes to the struct page could cause excessive cacheline bouncing on big systems. There are a number of ways this could be addressed if it is an issue. Signed-off-by: Nick Piggin Refcount bug fix for filemap_xip.c Signed-off-by: Carsten Otte Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/direct-io.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/direct-io.c b/fs/direct-io.c index 0d06097bc99..3931e7f1e6b 100644 --- a/fs/direct-io.c +++ b/fs/direct-io.c @@ -162,6 +162,7 @@ static int dio_refill_pages(struct dio *dio) up_read(¤t->mm->mmap_sem); if (ret < 0 && dio->blocks_available && (dio->rw == WRITE)) { + struct page *page = ZERO_PAGE(dio->curr_user_address); /* * A memory fault, but the filesystem has some outstanding * mapped blocks. We need to use those blocks up to avoid @@ -169,7 +170,8 @@ static int dio_refill_pages(struct dio *dio) */ if (dio->page_errors == 0) dio->page_errors = ret; - dio->pages[0] = ZERO_PAGE(dio->curr_user_address); + page_cache_get(page); + dio->pages[0] = page; dio->head = 0; dio->tail = 1; ret = 0; -- cgit v1.2.3 From 365e9c87a982c03d0af3886e29d877f581b59611 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:18 -0700 Subject: [PATCH] mm: update_hiwaters just in time update_mem_hiwater has attracted various criticisms, in particular from those concerned with mm scalability. Originally it was called whenever rss or total_vm got raised. Then many of those callsites were replaced by a timer tick call from account_system_time. Now Frank van Maarseveen reports that to be found inadequate. How about this? Works for Frank. Replace update_mem_hiwater, a poor combination of two unrelated ops, by macros update_hiwater_rss and update_hiwater_vm. Don't attempt to keep mm->hiwater_rss up to date at timer tick, nor every time we raise rss (usually by 1): those are hot paths. Do the opposite, update only when about to lower rss (usually by many), or just before final accounting in do_exit. Handle mm->hiwater_vm in the same way, though it's much less of an issue. Demand that whoever collects these hiwater statistics do the work of taking the maximum with rss or total_vm. And there has been no collector of these hiwater statistics in the tree. The new convention needs an example, so match Frank's usage by adding a VmPeak line above VmSize to /proc//status, and also a VmHWM line above VmRSS (High-Water-Mark or High-Water-Memory). There was a particular anomaly during mremap move, that hiwater_vm might be captured too high. A fleeting such anomaly remains, but it's quickly corrected now, whereas before it would stick. What locking? None: if the app is racy then these statistics will be racy, it's not worth any overhead to make them exact. But whenever it suits, hiwater_vm is updated under exclusive mmap_sem, and hiwater_rss under page_table_lock (for now) or with preemption disabled (later on): without going to any trouble, minimize the time between reading current values and updating, to minimize those occasions when a racing thread bumps a count up and back down in between. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/compat.c | 1 - fs/exec.c | 1 - fs/proc/task_mmu.c | 23 +++++++++++++++++++++-- 3 files changed, 21 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/compat.c b/fs/compat.c index a719e158e00..8e71cdbecc7 100644 --- a/fs/compat.c +++ b/fs/compat.c @@ -1490,7 +1490,6 @@ int compat_do_execve(char * filename, /* execve success */ security_bprm_free(bprm); acct_update_integrals(current); - update_mem_hiwater(current); kfree(bprm); return retval; } diff --git a/fs/exec.c b/fs/exec.c index cefadf5ab83..9bb55c8cf22 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -1207,7 +1207,6 @@ int do_execve(char * filename, /* execve success */ security_bprm_free(bprm); acct_update_integrals(current); - update_mem_hiwater(current); kfree(bprm); return retval; } diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index bccee7cf9cc..7c89b454904 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -14,22 +14,41 @@ char *task_mem(struct mm_struct *mm, char *buffer) { unsigned long data, text, lib; + unsigned long hiwater_vm, total_vm, hiwater_rss, total_rss; + + /* + * Note: to minimize their overhead, mm maintains hiwater_vm and + * hiwater_rss only when about to *lower* total_vm or rss. Any + * collector of these hiwater stats must therefore get total_vm + * and rss too, which will usually be the higher. Barriers? not + * worth the effort, such snapshots can always be inconsistent. + */ + hiwater_vm = total_vm = mm->total_vm; + if (hiwater_vm < mm->hiwater_vm) + hiwater_vm = mm->hiwater_vm; + hiwater_rss = total_rss = get_mm_rss(mm); + if (hiwater_rss < mm->hiwater_rss) + hiwater_rss = mm->hiwater_rss; data = mm->total_vm - mm->shared_vm - mm->stack_vm; text = (PAGE_ALIGN(mm->end_code) - (mm->start_code & PAGE_MASK)) >> 10; lib = (mm->exec_vm << (PAGE_SHIFT-10)) - text; buffer += sprintf(buffer, + "VmPeak:\t%8lu kB\n" "VmSize:\t%8lu kB\n" "VmLck:\t%8lu kB\n" + "VmHWM:\t%8lu kB\n" "VmRSS:\t%8lu kB\n" "VmData:\t%8lu kB\n" "VmStk:\t%8lu kB\n" "VmExe:\t%8lu kB\n" "VmLib:\t%8lu kB\n" "VmPTE:\t%8lu kB\n", - (mm->total_vm - mm->reserved_vm) << (PAGE_SHIFT-10), + hiwater_vm << (PAGE_SHIFT-10), + (total_vm - mm->reserved_vm) << (PAGE_SHIFT-10), mm->locked_vm << (PAGE_SHIFT-10), - get_mm_rss(mm) << (PAGE_SHIFT-10), + hiwater_rss << (PAGE_SHIFT-10), + total_rss << (PAGE_SHIFT-10), data << (PAGE_SHIFT-10), mm->stack_vm << (PAGE_SHIFT-10), text, lib, (PTRS_PER_PTE*sizeof(pte_t)*mm->nr_ptes) >> 10); -- cgit v1.2.3 From c74df32c724a1652ad8399b4891bb02c9d43743a Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:23 -0700 Subject: [PATCH] mm: ptd_alloc take ptlock Second step in pushing down the page_table_lock. Remove the temporary bridging hack from __pud_alloc, __pmd_alloc, __pte_alloc: expect callers not to hold page_table_lock, whether it's on init_mm or a user mm; take page_table_lock internally to check if a racing task already allocated. Convert their callers from common code. But avoid coming back to change them again later: instead of moving the spin_lock(&mm->page_table_lock) down, switch over to new macros pte_alloc_map_lock and pte_unmap_unlock, which encapsulate the mapping+locking and unlocking+unmapping together, and in the end may use alternatives to the mm page_table_lock itself. These callers all hold mmap_sem (some exclusively, some not), so at no level can a page table be whipped away from beneath them; and pte_alloc uses the "atomic" pmd_present to test whether it needs to allocate. It appears that on all arches we can safely descend without page_table_lock. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/exec.c | 14 +++++--------- 1 file changed, 5 insertions(+), 9 deletions(-) (limited to 'fs') diff --git a/fs/exec.c b/fs/exec.c index 9bb55c8cf22..ba73797eb4c 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -309,25 +309,24 @@ void install_arg_page(struct vm_area_struct *vma, pud_t * pud; pmd_t * pmd; pte_t * pte; + spinlock_t *ptl; if (unlikely(anon_vma_prepare(vma))) - goto out_sig; + goto out; flush_dcache_page(page); pgd = pgd_offset(mm, address); - - spin_lock(&mm->page_table_lock); pud = pud_alloc(mm, pgd, address); if (!pud) goto out; pmd = pmd_alloc(mm, pud, address); if (!pmd) goto out; - pte = pte_alloc_map(mm, pmd, address); + pte = pte_alloc_map_lock(mm, pmd, address, &ptl); if (!pte) goto out; if (!pte_none(*pte)) { - pte_unmap(pte); + pte_unmap_unlock(pte, ptl); goto out; } inc_mm_counter(mm, anon_rss); @@ -335,14 +334,11 @@ void install_arg_page(struct vm_area_struct *vma, set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte( page, vma->vm_page_prot)))); page_add_anon_rmap(page, vma, address); - pte_unmap(pte); - spin_unlock(&mm->page_table_lock); + pte_unmap_unlock(pte, ptl); /* no need for flush_tlb */ return; out: - spin_unlock(&mm->page_table_lock); -out_sig: __free_page(page); force_sig(SIGKILL, current); } -- cgit v1.2.3 From 705e87c0c3c38424f7f30556c85bc20e808d2f59 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:27 -0700 Subject: [PATCH] mm: pte_offset_map_lock loops Convert those common loops using page_table_lock on the outside and pte_offset_map within to use just pte_offset_map_lock within instead. These all hold mmap_sem (some exclusively, some not), so at no level can a page table be whipped away from beneath them. But whereas pte_alloc loops tested with the "atomic" pmd_present, these loops are testing with pmd_none, which on i386 PAE tests both lower and upper halves. That's now unsafe, so add a cast into pmd_none to test only the vital lower half: we lose a little sensitivity to a corrupt middle directory, but not enough to worry about. It appears that i386 and UML were the only architectures vulnerable in this way, and pgd and pud no problem. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/proc/task_mmu.c | 17 ++++++----------- 1 file changed, 6 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index 7c89b454904..7e5e7ec2e36 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -203,13 +203,14 @@ static void smaps_pte_range(struct vm_area_struct *vma, pmd_t *pmd, struct mem_size_stats *mss) { pte_t *pte, ptent; + spinlock_t *ptl; unsigned long pfn; struct page *page; - pte = pte_offset_map(pmd, addr); + pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); do { ptent = *pte; - if (pte_none(ptent) || !pte_present(ptent)) + if (!pte_present(ptent)) continue; mss->resident += PAGE_SIZE; @@ -230,8 +231,8 @@ static void smaps_pte_range(struct vm_area_struct *vma, pmd_t *pmd, mss->private_clean += PAGE_SIZE; } } while (pte++, addr += PAGE_SIZE, addr != end); - pte_unmap(pte - 1); - cond_resched_lock(&vma->vm_mm->page_table_lock); + pte_unmap_unlock(pte - 1, ptl); + cond_resched(); } static inline void smaps_pmd_range(struct vm_area_struct *vma, pud_t *pud, @@ -285,17 +286,11 @@ static inline void smaps_pgd_range(struct vm_area_struct *vma, static int show_smap(struct seq_file *m, void *v) { struct vm_area_struct *vma = v; - struct mm_struct *mm = vma->vm_mm; struct mem_size_stats mss; memset(&mss, 0, sizeof mss); - - if (mm) { - spin_lock(&mm->page_table_lock); + if (vma->vm_mm) smaps_pgd_range(vma, vma->vm_start, vma->vm_end, &mss); - spin_unlock(&mm->page_table_lock); - } - return show_map_internal(m, v, &mss); } -- cgit v1.2.3 From 508034a32b819a2d40aa7ac0dbc8cd2e044c2de6 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:30 -0700 Subject: [PATCH] mm: unmap_vmas with inner ptlock Remove the page_table_lock from around the calls to unmap_vmas, and replace the pte_offset_map in zap_pte_range by pte_offset_map_lock: all callers are now safe to descend without page_table_lock. Don't attempt fancy locking for hugepages, just take page_table_lock in unmap_hugepage_range. Which makes zap_hugepage_range, and the hugetlb test in zap_page_range, redundant: unmap_vmas calls unmap_hugepage_range anyway. Nor does unmap_vmas have much use for its mm arg now. The tlb_start_vma and tlb_end_vma in unmap_page_range are now called without page_table_lock: if they're implemented at all, they typically come down to flush_cache_range (usually done outside page_table_lock) and flush_tlb_range (which we already audited for the mprotect case). Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 3a9b6d179cb..a826a8add5e 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -92,7 +92,7 @@ out: } /* - * Called under down_write(mmap_sem), page_table_lock is not held + * Called under down_write(mmap_sem). */ #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA @@ -308,7 +308,6 @@ hugetlb_vmtruncate_list(struct prio_tree_root *root, unsigned long h_pgoff) vma_prio_tree_foreach(vma, &iter, root, h_pgoff, ULONG_MAX) { unsigned long h_vm_pgoff; - unsigned long v_length; unsigned long v_offset; h_vm_pgoff = vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT); @@ -319,11 +318,8 @@ hugetlb_vmtruncate_list(struct prio_tree_root *root, unsigned long h_pgoff) if (h_vm_pgoff >= h_pgoff) v_offset = 0; - v_length = vma->vm_end - vma->vm_start; - - zap_hugepage_range(vma, - vma->vm_start + v_offset, - v_length - v_offset); + unmap_hugepage_range(vma, + vma->vm_start + v_offset, vma->vm_end); } } -- cgit v1.2.3 From deceb6cd17e6dfafe4c4f81b1b4153bc41b2cb70 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:33 -0700 Subject: [PATCH] mm: follow_page with inner ptlock Final step in pushing down common core's page_table_lock. follow_page no longer wants caller to hold page_table_lock, uses pte_offset_map_lock itself; and so no page_table_lock is taken in get_user_pages itself. But get_user_pages (and get_futex_key) do then need follow_page to pin the page for them: take Daniel's suggestion of bitflags to follow_page. Need one for WRITE, another for TOUCH (it was the accessed flag before: vanished along with check_user_page_readable, but surely get_numa_maps is wrong to mark every page it finds as accessed), another for GET. And another, ANON to dispose of untouched_anonymous_page: it seems silly for that to descend a second time, let follow_page observe if there was no page table and return ZERO_PAGE if so. Fix minor bug in that: check VM_LOCKED - make_pages_present ought to make readonly anonymous present. Give get_numa_maps a cond_resched while we're there. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/proc/task_mmu.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index 7e5e7ec2e36..d2fa42006d8 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -419,7 +419,6 @@ static struct numa_maps *get_numa_maps(const struct vm_area_struct *vma) for_each_node(i) md->node[i] =0; - spin_lock(&mm->page_table_lock); for (vaddr = vma->vm_start; vaddr < vma->vm_end; vaddr += PAGE_SIZE) { page = follow_page(mm, vaddr, 0); if (page) { @@ -434,8 +433,8 @@ static struct numa_maps *get_numa_maps(const struct vm_area_struct *vma) md->anon++; md->node[page_to_nid(page)]++; } + cond_resched(); } - spin_unlock(&mm->page_table_lock); return md; } -- cgit v1.2.3 From 4c21e2f2441dc5fbb957b030333f5a3f2d02dea7 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Sat, 29 Oct 2005 18:16:40 -0700 Subject: [PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/afs/file.c | 4 ++-- fs/buffer.c | 2 +- fs/jfs/jfs_metapage.c | 12 ++++++------ fs/xfs/linux-2.6/xfs_buf.c | 7 ++++--- 4 files changed, 13 insertions(+), 12 deletions(-) (limited to 'fs') diff --git a/fs/afs/file.c b/fs/afs/file.c index 0d576987ec6..4975c9c193d 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -291,8 +291,8 @@ static int afs_file_releasepage(struct page *page, gfp_t gfp_flags) cachefs_uncache_page(vnode->cache, page); #endif - pageio = (struct cachefs_page *) page->private; - page->private = 0; + pageio = (struct cachefs_page *) page_private(page); + set_page_private(page, 0); ClearPagePrivate(page); if (pageio) diff --git a/fs/buffer.c b/fs/buffer.c index b1667986442..2066e4cb700 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -96,7 +96,7 @@ static void __clear_page_buffers(struct page *page) { ClearPagePrivate(page); - page->private = 0; + set_page_private(page, 0); page_cache_release(page); } diff --git a/fs/jfs/jfs_metapage.c b/fs/jfs/jfs_metapage.c index 26091a5f88d..8a53981f9f2 100644 --- a/fs/jfs/jfs_metapage.c +++ b/fs/jfs/jfs_metapage.c @@ -86,7 +86,7 @@ struct meta_anchor { atomic_t io_count; struct metapage *mp[MPS_PER_PAGE]; }; -#define mp_anchor(page) ((struct meta_anchor *)page->private) +#define mp_anchor(page) ((struct meta_anchor *)page_private(page)) static inline struct metapage *page_to_mp(struct page *page, uint offset) { @@ -108,7 +108,7 @@ static inline int insert_metapage(struct page *page, struct metapage *mp) if (!a) return -ENOMEM; memset(a, 0, sizeof(struct meta_anchor)); - page->private = (unsigned long)a; + set_page_private(page, (unsigned long)a); SetPagePrivate(page); kmap(page); } @@ -136,7 +136,7 @@ static inline void remove_metapage(struct page *page, struct metapage *mp) a->mp[index] = NULL; if (--a->mp_count == 0) { kfree(a); - page->private = 0; + set_page_private(page, 0); ClearPagePrivate(page); kunmap(page); } @@ -156,13 +156,13 @@ static inline void dec_io(struct page *page, void (*handler) (struct page *)) #else static inline struct metapage *page_to_mp(struct page *page, uint offset) { - return PagePrivate(page) ? (struct metapage *)page->private : NULL; + return PagePrivate(page) ? (struct metapage *)page_private(page) : NULL; } static inline int insert_metapage(struct page *page, struct metapage *mp) { if (mp) { - page->private = (unsigned long)mp; + set_page_private(page, (unsigned long)mp); SetPagePrivate(page); kmap(page); } @@ -171,7 +171,7 @@ static inline int insert_metapage(struct page *page, struct metapage *mp) static inline void remove_metapage(struct page *page, struct metapage *mp) { - page->private = 0; + set_page_private(page, 0); ClearPagePrivate(page); kunmap(page); } diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index ba4767c04ad..4cd46abe843 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -181,8 +181,9 @@ set_page_region( size_t offset, size_t length) { - page->private |= page_region_mask(offset, length); - if (page->private == ~0UL) + set_page_private(page, + page_private(page) | page_region_mask(offset, length)); + if (page_private(page) == ~0UL) SetPageUptodate(page); } @@ -194,7 +195,7 @@ test_page_region( { unsigned long mask = page_region_mask(offset, length); - return (mask && (page->private & mask) == mask); + return (mask && (page_private(page) & mask) == mask); } /* -- cgit v1.2.3 From 96527980d4cb8f65fe49efdbc4ab92c0837d42f6 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 29 Oct 2005 18:16:42 -0700 Subject: [PATCH] hugetlbfs: move free_inodes accounting Move hugetlbfs accounting into ->alloc_inode / ->destroy_inode. This keeps the code simpler, fixes a loeak where a failing inode allocation wouldn't decrement the counter and moves hugetlbfs_delete_inode and hugetlbfs_forget_inode closer to their generic counterparts. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 78 ++++++++++++++++++++++++++++------------------------ 1 file changed, 42 insertions(+), 36 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index a826a8add5e..8e9d4363336 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -224,8 +224,6 @@ static void truncate_hugepages(struct address_space *mapping, loff_t lstart) static void hugetlbfs_delete_inode(struct inode *inode) { - struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(inode->i_sb); - hlist_del_init(&inode->i_hash); list_del_init(&inode->i_list); list_del_init(&inode->i_sb_list); @@ -238,12 +236,6 @@ static void hugetlbfs_delete_inode(struct inode *inode) security_inode_delete(inode); - if (sbinfo->free_inodes >= 0) { - spin_lock(&sbinfo->stat_lock); - sbinfo->free_inodes++; - spin_unlock(&sbinfo->stat_lock); - } - clear_inode(inode); destroy_inode(inode); } @@ -251,7 +243,6 @@ static void hugetlbfs_delete_inode(struct inode *inode) static void hugetlbfs_forget_inode(struct inode *inode) { struct super_block *super_block = inode->i_sb; - struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(super_block); if (hlist_unhashed(&inode->i_hash)) goto out_truncate; @@ -278,12 +269,6 @@ out_truncate: if (inode->i_data.nrpages) truncate_hugepages(&inode->i_data, 0); - if (sbinfo->free_inodes >= 0) { - spin_lock(&sbinfo->stat_lock); - sbinfo->free_inodes++; - spin_unlock(&sbinfo->stat_lock); - } - clear_inode(inode); destroy_inode(inode); } @@ -375,17 +360,6 @@ static struct inode *hugetlbfs_get_inode(struct super_block *sb, uid_t uid, gid_t gid, int mode, dev_t dev) { struct inode *inode; - struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb); - - if (sbinfo->free_inodes >= 0) { - spin_lock(&sbinfo->stat_lock); - if (!sbinfo->free_inodes) { - spin_unlock(&sbinfo->stat_lock); - return NULL; - } - sbinfo->free_inodes--; - spin_unlock(&sbinfo->stat_lock); - } inode = new_inode(sb); if (inode) { @@ -527,29 +501,51 @@ static void hugetlbfs_put_super(struct super_block *sb) } } +static inline int hugetlbfs_dec_free_inodes(struct hugetlbfs_sb_info *sbinfo) +{ + if (sbinfo->free_inodes >= 0) { + spin_lock(&sbinfo->stat_lock); + if (unlikely(!sbinfo->free_inodes)) { + spin_unlock(&sbinfo->stat_lock); + return 0; + } + sbinfo->free_inodes--; + spin_unlock(&sbinfo->stat_lock); + } + + return 1; +} + +static void hugetlbfs_inc_free_inodes(struct hugetlbfs_sb_info *sbinfo) +{ + if (sbinfo->free_inodes >= 0) { + spin_lock(&sbinfo->stat_lock); + sbinfo->free_inodes++; + spin_unlock(&sbinfo->stat_lock); + } +} + + static kmem_cache_t *hugetlbfs_inode_cachep; static struct inode *hugetlbfs_alloc_inode(struct super_block *sb) { + struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb); struct hugetlbfs_inode_info *p; + if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo))) + return NULL; p = kmem_cache_alloc(hugetlbfs_inode_cachep, SLAB_KERNEL); - if (!p) + if (unlikely(!p)) { + hugetlbfs_inc_free_inodes(sbinfo); return NULL; + } return &p->vfs_inode; } -static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags) -{ - struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo; - - if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == - SLAB_CTOR_CONSTRUCTOR) - inode_init_once(&ei->vfs_inode); -} - static void hugetlbfs_destroy_inode(struct inode *inode) { + hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb)); mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy); kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode)); } @@ -561,6 +557,16 @@ static struct address_space_operations hugetlbfs_aops = { .set_page_dirty = hugetlbfs_set_page_dirty, }; + +static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags) +{ + struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo; + + if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == + SLAB_CTOR_CONSTRUCTOR) + inode_init_once(&ei->vfs_inode); +} + struct file_operations hugetlbfs_file_operations = { .mmap = hugetlbfs_file_mmap, .fsync = simple_sync_file, -- cgit v1.2.3 From 149f4211afda85743e3a3db3fa3abbd81506cf2b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 29 Oct 2005 18:16:43 -0700 Subject: [PATCH] hugetlbfs: clean up hugetlbfs_delete_inode Make hugetlbfs looks the same as generic_detelte_inode, fixing a bunch of missing updates to it at the same time. Rename it to hugetlbfs_do_delete_inode and add a real hugetlbfs_delete_inode that implements ->delete_inode. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 38 ++++++++++++++++++++++++++++++++------ 1 file changed, 32 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 8e9d4363336..2b9d1bee922 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -224,19 +224,44 @@ static void truncate_hugepages(struct address_space *mapping, loff_t lstart) static void hugetlbfs_delete_inode(struct inode *inode) { - hlist_del_init(&inode->i_hash); + if (inode->i_data.nrpages) + truncate_hugepages(&inode->i_data, 0); + clear_inode(inode); +} + +static void hugetlbfs_do_delete_inode(struct inode *inode) +{ + struct super_operations *op = inode->i_sb->s_op; + list_del_init(&inode->i_list); list_del_init(&inode->i_sb_list); inode->i_state |= I_FREEING; inodes_stat.nr_inodes--; spin_unlock(&inode_lock); - if (inode->i_data.nrpages) - truncate_hugepages(&inode->i_data, 0); - security_inode_delete(inode); - clear_inode(inode); + if (op->delete_inode) { + void (*delete)(struct inode *) = op->delete_inode; + if (!is_bad_inode(inode)) + DQUOT_INIT(inode); + /* Filesystems implementing their own + * s_op->delete_inode are required to call + * truncate_inode_pages and clear_inode() + * internally + */ + delete(inode); + } else { + truncate_inode_pages(&inode->i_data, 0); + clear_inode(inode); + } + + spin_lock(&inode_lock); + hlist_del_init(&inode->i_hash); + spin_unlock(&inode_lock); + wake_up_inode(inode); + if (inode->i_state != I_CLEAR) + BUG(); destroy_inode(inode); } @@ -276,7 +301,7 @@ out_truncate: static void hugetlbfs_drop_inode(struct inode *inode) { if (!inode->i_nlink) - hugetlbfs_delete_inode(inode); + hugetlbfs_do_delete_inode(inode); else hugetlbfs_forget_inode(inode); } @@ -594,6 +619,7 @@ static struct super_operations hugetlbfs_ops = { .alloc_inode = hugetlbfs_alloc_inode, .destroy_inode = hugetlbfs_destroy_inode, .statfs = hugetlbfs_statfs, + .delete_inode = hugetlbfs_delete_inode, .drop_inode = hugetlbfs_drop_inode, .put_super = hugetlbfs_put_super, }; -- cgit v1.2.3 From 6b09b9df05f319ec27e0dae1721efe097b8b23ad Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 29 Oct 2005 18:16:44 -0700 Subject: [PATCH] kill hugelbfs_do_delete_inode hugetlbfs_do_delete_inode is the same as generic_delete_inode now, so remove it in favour of the latter. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 38 +------------------------------------- 1 file changed, 1 insertion(+), 37 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 2b9d1bee922..ffdad4e6467 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -229,42 +229,6 @@ static void hugetlbfs_delete_inode(struct inode *inode) clear_inode(inode); } -static void hugetlbfs_do_delete_inode(struct inode *inode) -{ - struct super_operations *op = inode->i_sb->s_op; - - list_del_init(&inode->i_list); - list_del_init(&inode->i_sb_list); - inode->i_state |= I_FREEING; - inodes_stat.nr_inodes--; - spin_unlock(&inode_lock); - - security_inode_delete(inode); - - if (op->delete_inode) { - void (*delete)(struct inode *) = op->delete_inode; - if (!is_bad_inode(inode)) - DQUOT_INIT(inode); - /* Filesystems implementing their own - * s_op->delete_inode are required to call - * truncate_inode_pages and clear_inode() - * internally - */ - delete(inode); - } else { - truncate_inode_pages(&inode->i_data, 0); - clear_inode(inode); - } - - spin_lock(&inode_lock); - hlist_del_init(&inode->i_hash); - spin_unlock(&inode_lock); - wake_up_inode(inode); - if (inode->i_state != I_CLEAR) - BUG(); - destroy_inode(inode); -} - static void hugetlbfs_forget_inode(struct inode *inode) { struct super_block *super_block = inode->i_sb; @@ -301,7 +265,7 @@ out_truncate: static void hugetlbfs_drop_inode(struct inode *inode) { if (!inode->i_nlink) - hugetlbfs_do_delete_inode(inode); + generic_delete_inode(inode); else hugetlbfs_forget_inode(inode); } -- cgit v1.2.3 From 0b1533f67cc1a595457af6d05ab3510294e2ca9c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 29 Oct 2005 18:16:45 -0700 Subject: [PATCH] cleanup hugelbfs_forget_inode Reformat hugelbfs_forget_inode and add the missing but harmless write_inode_now call. It looks the same as generic_forget_inode now except for the call to truncate_hugepages instead of truncate_inode_pages. Signed-off-by: Christoph Hellwig Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 38 ++++++++++++++++++++------------------ 1 file changed, 20 insertions(+), 18 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index ffdad4e6467..8f94feb24c0 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -231,25 +231,28 @@ static void hugetlbfs_delete_inode(struct inode *inode) static void hugetlbfs_forget_inode(struct inode *inode) { - struct super_block *super_block = inode->i_sb; - - if (hlist_unhashed(&inode->i_hash)) - goto out_truncate; - - if (!(inode->i_state & (I_DIRTY|I_LOCK))) { - list_del(&inode->i_list); - list_add(&inode->i_list, &inode_unused); - } - inodes_stat.nr_unused++; - if (!super_block || (super_block->s_flags & MS_ACTIVE)) { + struct super_block *sb = inode->i_sb; + + if (!hlist_unhashed(&inode->i_hash)) { + if (!(inode->i_state & (I_DIRTY|I_LOCK))) + list_move(&inode->i_list, &inode_unused); + inodes_stat.nr_unused++; + if (!sb || (sb->s_flags & MS_ACTIVE)) { + spin_unlock(&inode_lock); + return; + } + inode->i_state |= I_WILL_FREE; spin_unlock(&inode_lock); - return; + /* + * write_inode_now is a noop as we set BDI_CAP_NO_WRITEBACK + * in our backing_dev_info. + */ + write_inode_now(inode, 1); + spin_lock(&inode_lock); + inode->i_state &= ~I_WILL_FREE; + inodes_stat.nr_unused--; + hlist_del_init(&inode->i_hash); } - - /* write_inode_now() ? */ - inodes_stat.nr_unused--; - hlist_del_init(&inode->i_hash); -out_truncate: list_del_init(&inode->i_list); list_del_init(&inode->i_sb_list); inode->i_state |= I_FREEING; @@ -257,7 +260,6 @@ out_truncate: spin_unlock(&inode_lock); if (inode->i_data.nrpages) truncate_hugepages(&inode->i_data, 0); - clear_inode(inode); destroy_inode(inode); } -- cgit v1.2.3 From 4c887265977213985091476be40ab11dfdcb4caf Mon Sep 17 00:00:00 2001 From: Adam Litke Date: Sat, 29 Oct 2005 18:16:46 -0700 Subject: [PATCH] hugetlb: demand fault handler Below is a patch to implement demand faulting for huge pages. The main motivation for changing from prefaulting to demand faulting is so that huge page memory areas can be allocated according to NUMA policy. Thanks to consolidated hugetlb code, switching the behavior requires changing only one fault handler. The bulk of the patch just moves the logic from hugelb_prefault() to hugetlb_pte_fault() and find_get_huge_page(). Signed-off-by: Adam Litke Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 8f94feb24c0..2627efe767c 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -48,7 +48,6 @@ int sysctl_hugetlb_shm_group; static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) { struct inode *inode = file->f_dentry->d_inode; - struct address_space *mapping = inode->i_mapping; loff_t len, vma_len; int ret; @@ -79,10 +78,8 @@ static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) if (!(vma->vm_flags & VM_WRITE) && len > inode->i_size) goto out; - ret = hugetlb_prefault(mapping, vma); - if (ret) - goto out; - + ret = 0; + hugetlb_prefault_arch_hook(vma->vm_mm); if (inode->i_size < len) inode->i_size = len; out: -- cgit v1.2.3 From 2e9b367c2273ed21c9852a04d90944d472c4f3e6 Mon Sep 17 00:00:00 2001 From: Adam Litke Date: Sat, 29 Oct 2005 18:16:47 -0700 Subject: [PATCH] hugetlb: overcommit accounting check Basic overcommit checking for hugetlb_file_map() based on an implementation used with demand faulting in SLES9. Since demand faulting can't guarantee the availability of pages at mmap time, this patch implements a basic sanity check to ensure that the number of huge pages required to satisfy the mmap are currently available. Despite the obvious race, I think it is a good start on doing proper accounting. I'd like to work towards an accounting system that mimics the semantics of normal pages (especially for the MAP_PRIVATE/COW case). That work is underway and builds on what this patch starts. Huge page shared memory segments are simpler and still maintain their commit on shmget semantics. Signed-off-by: Adam Litke Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/hugetlbfs/inode.c | 63 +++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 53 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 2627efe767c..e026c807e6b 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -45,9 +45,58 @@ static struct backing_dev_info hugetlbfs_backing_dev_info = { int sysctl_hugetlb_shm_group; +static void huge_pagevec_release(struct pagevec *pvec) +{ + int i; + + for (i = 0; i < pagevec_count(pvec); ++i) + put_page(pvec->pages[i]); + + pagevec_reinit(pvec); +} + +/* + * huge_pages_needed tries to determine the number of new huge pages that + * will be required to fully populate this VMA. This will be equal to + * the size of the VMA in huge pages minus the number of huge pages + * (covered by this VMA) that are found in the page cache. + * + * Result is in bytes to be compatible with is_hugepage_mem_enough() + */ +unsigned long +huge_pages_needed(struct address_space *mapping, struct vm_area_struct *vma) +{ + int i; + struct pagevec pvec; + unsigned long start = vma->vm_start; + unsigned long end = vma->vm_end; + unsigned long hugepages = (end - start) >> HPAGE_SHIFT; + pgoff_t next = vma->vm_pgoff; + pgoff_t endpg = next + ((end - start) >> PAGE_SHIFT); + + pagevec_init(&pvec, 0); + while (next < endpg) { + if (!pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) + break; + for (i = 0; i < pagevec_count(&pvec); i++) { + struct page *page = pvec.pages[i]; + if (page->index > next) + next = page->index; + if (page->index >= endpg) + break; + next++; + hugepages--; + } + huge_pagevec_release(&pvec); + } + return hugepages << HPAGE_SHIFT; +} + static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) { struct inode *inode = file->f_dentry->d_inode; + struct address_space *mapping = inode->i_mapping; + unsigned long bytes; loff_t len, vma_len; int ret; @@ -66,6 +115,10 @@ static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) if (vma->vm_end - vma->vm_start < HPAGE_SIZE) return -EINVAL; + bytes = huge_pages_needed(mapping, vma); + if (!is_hugepage_mem_enough(bytes)) + return -ENOMEM; + vma_len = (loff_t)(vma->vm_end - vma->vm_start); down(&inode->i_sem); @@ -168,16 +221,6 @@ static int hugetlbfs_commit_write(struct file *file, return -EINVAL; } -static void huge_pagevec_release(struct pagevec *pvec) -{ - int i; - - for (i = 0; i < pagevec_count(pvec); ++i) - put_page(pvec->pages[i]); - - pagevec_reinit(pvec); -} - static void truncate_huge_page(struct page *page) { clear_page_dirty(page); -- cgit v1.2.3