From 35e396cd100489dfe8f5a76e3613fb8049ffdff3 Mon Sep 17 00:00:00 2001 From: "xiphmont@xiph.org" Date: Fri, 22 Aug 2008 11:12:21 +0200 Subject: SG_IO block filter whitelist missing MMC SET READ AHEAD command I have another request for the block filter SG_IO command whitelist, specifically the MMC streaming command set SET READ AHEAD command. The command applies only to MMC CDROM/DVDROM drives with the streaming optional feature set. The command is useful to cdparanoia in that it allows explicit cache control side effects that are, on many drives, cdparanoia's most efficient way to flush/disable the media cache on cdrom drives. I am aware of no reason why it should not be accessible from usespace. Also note that the command is already fully accessible through the SCSI-native version of the SG_IO ioctl as well as the traditional SG interface. The command is only being refused on block devices. That means that on a typical stock distro, the command is available through /dev/sg* but not /dev/scd* although both are typically available and accessible. Filtering the command is not providing any protection, only a confusing inconsistency. Signed-off-by: Jens Axboe --- block/scsi_ioctl.c | 1 + 1 file changed, 1 insertion(+) (limited to 'block') diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c index ec4b7f23462..3aab80a4c48 100644 --- a/block/scsi_ioctl.c +++ b/block/scsi_ioctl.c @@ -185,6 +185,7 @@ void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter) __set_bit(GPCMD_PREVENT_ALLOW_MEDIUM_REMOVAL, filter->write_ok); __set_bit(GPCMD_LOAD_UNLOAD, filter->write_ok); __set_bit(GPCMD_SET_STREAMING, filter->write_ok); + __set_bit(GPCMD_SET_READ_AHEAD, filter->write_ok); } EXPORT_SYMBOL_GPL(blk_set_cmd_filter_defaults); -- cgit v1.2.3 From a9c701e594669dd49fed448c27c64f20cfacc8a7 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 8 Aug 2008 11:04:44 +0200 Subject: block: use bio_has_data() to check for data carrying bio Signed-off-by: Jens Axboe --- block/blk-core.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 2cba5ef97b2..54e442ba44a 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1490,10 +1490,7 @@ void submit_bio(int rw, struct bio *bio) * If it's a regular read/write or a barrier with data attached, * go through the normal accounting stuff before submission. */ - if (!bio_empty_barrier(bio)) { - - BIO_BUG_ON(!bio->bi_size); - BIO_BUG_ON(!bio->bi_io_vec); + if (bio_has_data(bio)) { if (rw & WRITE) { count_vm_events(PGPGOUT, count); -- cgit v1.2.3 From 051cc3952a8fb6fa875a4eff68d06cf42207dcf4 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 8 Aug 2008 11:06:45 +0200 Subject: block: use bio_has_data() in the IO completion path Signed-off-by: Jens Axboe --- block/blk-core.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 54e442ba44a..b5776c1fd52 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1885,7 +1885,7 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, struct request_queue *q = rq->q; unsigned long flags = 0UL; - if (blk_fs_request(rq) || blk_pc_request(rq)) { + if (bio_has_data(rq->bio)) { if (__end_that_request_first(rq, error, nr_bytes)) return 1; @@ -1943,10 +1943,9 @@ EXPORT_SYMBOL_GPL(blk_end_request); **/ int __blk_end_request(struct request *rq, int error, unsigned int nr_bytes) { - if (blk_fs_request(rq) || blk_pc_request(rq)) { - if (__end_that_request_first(rq, error, nr_bytes)) - return 1; - } + if (bio_has_data(rq->bio) && + __end_that_request_first(rq, error, nr_bytes)) + return 1; add_disk_randomness(rq->rq_disk); -- cgit v1.2.3 From d628eaef310533767ce68664873869c2d7f78f09 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sat, 9 Aug 2008 16:22:17 +0100 Subject: Fix up comments about matching flags between bio and rq Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/blk-core.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index b5776c1fd52..a496727df7e 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -624,10 +624,6 @@ blk_alloc_request(struct request_queue *q, int rw, int priv, gfp_t gfp_mask) blk_rq_init(q, rq); - /* - * first three bits are identical in rq->cmd_flags and bio->bi_rw, - * see bio.h and blkdev.h - */ rq->cmd_flags = rw | REQ_ALLOCED; if (priv) { @@ -2012,7 +2008,8 @@ EXPORT_SYMBOL_GPL(blk_end_request_callback); void blk_rq_bio_prep(struct request_queue *q, struct request *rq, struct bio *bio) { - /* first two bits are identical in rq->cmd_flags and bio->bi_rw */ + /* Bit 0 (R/W) is identical in rq->cmd_flags and bio->bi_rw, and + we want BIO_RW_AHEAD (bit 1) to imply REQ_FAILFAST (bit 1). */ rq->cmd_flags |= (bio->bi_rw & 3); rq->nr_phys_segments = bio_phys_segments(q, bio); -- cgit v1.2.3 From fb2dce862d9f9a68e6b9374579056ec9eca02a63 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Tue, 5 Aug 2008 18:01:53 +0100 Subject: Add 'discard' request handling Some block devices benefit from a hint that they can forget the contents of certain sectors. Add basic support for this to the block core, along with a 'blkdev_issue_discard()' helper function which issues such requests. The caller doesn't get to provide an end_io functio, since blkdev_issue_discard() will automatically split the request up into multiple bios if appropriate. Neither does the function wait for completion -- it's expected that callers won't care about when, or even _if_, the request completes. It's only a hint to the device anyway. By definition, the file system doesn't _care_ about these sectors any more. [With feedback from OGAWA Hirofumi and Jens Axboe Signed-off-by: Jens Axboe --- block/blk-barrier.c | 69 ++++++++++++++++++++++++++++++++++++++++++++++++++++ block/blk-core.c | 28 +++++++++++++++------ block/blk-settings.c | 17 +++++++++++++ 3 files changed, 106 insertions(+), 8 deletions(-) (limited to 'block') diff --git a/block/blk-barrier.c b/block/blk-barrier.c index a09ead19f9c..273121c0eb8 100644 --- a/block/blk-barrier.c +++ b/block/blk-barrier.c @@ -315,3 +315,72 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector) return ret; } EXPORT_SYMBOL(blkdev_issue_flush); + +static void blkdev_discard_end_io(struct bio *bio, int err) +{ + if (err) { + if (err == -EOPNOTSUPP) + set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); + clear_bit(BIO_UPTODATE, &bio->bi_flags); + } + + bio_put(bio); +} + +/** + * blkdev_issue_discard - queue a discard + * @bdev: blockdev to issue discard for + * @sector: start sector + * @nr_sects: number of sectors to discard + * + * Description: + * Issue a discard request for the sectors in question. Does not wait. + */ +int blkdev_issue_discard(struct block_device *bdev, sector_t sector, + unsigned nr_sects) +{ + struct request_queue *q; + struct bio *bio; + int ret = 0; + + if (bdev->bd_disk == NULL) + return -ENXIO; + + q = bdev_get_queue(bdev); + if (!q) + return -ENXIO; + + if (!q->prepare_discard_fn) + return -EOPNOTSUPP; + + while (nr_sects && !ret) { + bio = bio_alloc(GFP_KERNEL, 0); + if (!bio) + return -ENOMEM; + + bio->bi_end_io = blkdev_discard_end_io; + bio->bi_bdev = bdev; + + bio->bi_sector = sector; + + if (nr_sects > q->max_hw_sectors) { + bio->bi_size = q->max_hw_sectors << 9; + nr_sects -= q->max_hw_sectors; + sector += q->max_hw_sectors; + } else { + bio->bi_size = nr_sects << 9; + nr_sects = 0; + } + bio_get(bio); + submit_bio(WRITE_DISCARD, bio); + + /* Check if it failed immediately */ + if (bio_flagged(bio, BIO_EOPNOTSUPP)) + ret = -EOPNOTSUPP; + else if (!bio_flagged(bio, BIO_UPTODATE)) + ret = -EIO; + bio_put(bio); + } + return ret; +} +EXPORT_SYMBOL(blkdev_issue_discard); diff --git a/block/blk-core.c b/block/blk-core.c index a496727df7e..1e143c4f9d3 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1079,6 +1079,10 @@ void init_request_from_bio(struct request *req, struct bio *bio) */ if (unlikely(bio_barrier(bio))) req->cmd_flags |= (REQ_HARDBARRIER | REQ_NOMERGE); + if (unlikely(bio_discard(bio))) { + req->cmd_flags |= (REQ_SOFTBARRIER | REQ_DISCARD); + req->q->prepare_discard_fn(req->q, req); + } if (bio_sync(bio)) req->cmd_flags |= REQ_RW_SYNC; @@ -1095,7 +1099,7 @@ void init_request_from_bio(struct request *req, struct bio *bio) static int __make_request(struct request_queue *q, struct bio *bio) { struct request *req; - int el_ret, nr_sectors, barrier, err; + int el_ret, nr_sectors, barrier, discard, err; const unsigned short prio = bio_prio(bio); const int sync = bio_sync(bio); int rw_flags; @@ -1115,6 +1119,12 @@ static int __make_request(struct request_queue *q, struct bio *bio) goto end_io; } + discard = bio_discard(bio); + if (unlikely(discard) && !q->prepare_discard_fn) { + err = -EOPNOTSUPP; + goto end_io; + } + spin_lock_irq(q->queue_lock); if (unlikely(barrier) || elv_queue_empty(q)) @@ -1405,7 +1415,8 @@ end_io: if (bio_check_eod(bio, nr_sectors)) goto end_io; - if (bio_empty_barrier(bio) && !q->prepare_flush_fn) { + if ((bio_empty_barrier(bio) && !q->prepare_flush_fn) || + (bio_discard(bio) && !q->prepare_discard_fn)) { err = -EOPNOTSUPP; goto end_io; } @@ -1487,7 +1498,6 @@ void submit_bio(int rw, struct bio *bio) * go through the normal accounting stuff before submission. */ if (bio_has_data(bio)) { - if (rw & WRITE) { count_vm_events(PGPGOUT, count); } else { @@ -1881,7 +1891,7 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, struct request_queue *q = rq->q; unsigned long flags = 0UL; - if (bio_has_data(rq->bio)) { + if (bio_has_data(rq->bio) || blk_discard_rq(rq)) { if (__end_that_request_first(rq, error, nr_bytes)) return 1; @@ -1939,7 +1949,7 @@ EXPORT_SYMBOL_GPL(blk_end_request); **/ int __blk_end_request(struct request *rq, int error, unsigned int nr_bytes) { - if (bio_has_data(rq->bio) && + if ((bio_has_data(rq->bio) || blk_discard_rq(rq)) && __end_that_request_first(rq, error, nr_bytes)) return 1; @@ -2012,12 +2022,14 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq, we want BIO_RW_AHEAD (bit 1) to imply REQ_FAILFAST (bit 1). */ rq->cmd_flags |= (bio->bi_rw & 3); - rq->nr_phys_segments = bio_phys_segments(q, bio); - rq->nr_hw_segments = bio_hw_segments(q, bio); + if (bio_has_data(bio)) { + rq->nr_phys_segments = bio_phys_segments(q, bio); + rq->nr_hw_segments = bio_hw_segments(q, bio); + rq->buffer = bio_data(bio); + } rq->current_nr_sectors = bio_cur_sectors(bio); rq->hard_cur_sectors = rq->current_nr_sectors; rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio); - rq->buffer = bio_data(bio); rq->data_len = bio->bi_size; rq->bio = rq->biotail = bio; diff --git a/block/blk-settings.c b/block/blk-settings.c index dfc77012843..539d873c820 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -32,6 +32,23 @@ void blk_queue_prep_rq(struct request_queue *q, prep_rq_fn *pfn) } EXPORT_SYMBOL(blk_queue_prep_rq); +/** + * blk_queue_set_discard - set a discard_sectors function for queue + * @q: queue + * @dfn: prepare_discard function + * + * It's possible for a queue to register a discard callback which is used + * to transform a discard request into the appropriate type for the + * hardware. If none is registered, then discard requests are failed + * with %EOPNOTSUPP. + * + */ +void blk_queue_set_discard(struct request_queue *q, prepare_discard_fn *dfn) +{ + q->prepare_discard_fn = dfn; +} +EXPORT_SYMBOL(blk_queue_set_discard); + /** * blk_queue_merge_bvec - set a merge_bvec function for queue * @q: queue -- cgit v1.2.3 From 27b29e86bf3d4b3cf6641a0efd78ed11a9b633b2 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sun, 10 Aug 2008 11:21:57 +0100 Subject: blktrace: support discard requests Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/blktrace.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/blktrace.c b/block/blktrace.c index eb9651ccb24..7495a84353e 100644 --- a/block/blktrace.c +++ b/block/blktrace.c @@ -114,7 +114,13 @@ static u32 ddir_act[2] __read_mostly = { BLK_TC_ACT(BLK_TC_READ), BLK_TC_ACT(BLK /* * Bio action bits of interest */ -static u32 bio_act[9] __read_mostly = { 0, BLK_TC_ACT(BLK_TC_BARRIER), BLK_TC_ACT(BLK_TC_SYNC), 0, BLK_TC_ACT(BLK_TC_AHEAD), 0, 0, 0, BLK_TC_ACT(BLK_TC_META) }; +static u32 bio_act[17] __read_mostly = { + [1] = BLK_TC_ACT(BLK_TC_BARRIER), + [2] = BLK_TC_ACT(BLK_TC_SYNC), + [4] = BLK_TC_ACT(BLK_TC_AHEAD), + [8] = BLK_TC_ACT(BLK_TC_META), + [16] = BLK_TC_ACT(BLK_TC_DISCARD) +}; /* * More could be added as needed, taking care to increment the decrementer @@ -128,6 +134,8 @@ static u32 bio_act[9] __read_mostly = { 0, BLK_TC_ACT(BLK_TC_BARRIER), BLK_TC_AC (((rw) & (1 << BIO_RW_AHEAD)) << (2 - BIO_RW_AHEAD)) #define trace_meta_bit(rw) \ (((rw) & (1 << BIO_RW_META)) >> (BIO_RW_META - 3)) +#define trace_discard_bit(rw) \ + (((rw) & (1 << BIO_RW_DISCARD)) >> (BIO_RW_DISCARD - 4)) /* * The worker for the various blk_add_trace*() types. Fills out a @@ -151,6 +159,7 @@ void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes, what |= bio_act[trace_sync_bit(rw)]; what |= bio_act[trace_ahead_bit(rw)]; what |= bio_act[trace_meta_bit(rw)]; + what |= bio_act[trace_discard_bit(rw)]; pid = tsk->pid; if (unlikely(act_log_check(bt, what, sector, pid))) -- cgit v1.2.3 From 35ba8f7083e87602b695d6eaca38a6464d5b74db Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sun, 10 Aug 2008 12:33:00 +0100 Subject: blktrace: simplify flags handling in __blk_add_trace Let the compiler see what's going on, and it can all get a lot simpler. On PPC64 this reduces the size of the code calculating these bits by about 60%. On x86_64 it's less of a win -- only 40%. Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/blktrace.c | 38 ++++++++------------------------------ 1 file changed, 8 insertions(+), 30 deletions(-) (limited to 'block') diff --git a/block/blktrace.c b/block/blktrace.c index 7495a84353e..9e0212c90b2 100644 --- a/block/blktrace.c +++ b/block/blktrace.c @@ -111,31 +111,9 @@ static int act_log_check(struct blk_trace *bt, u32 what, sector_t sector, */ static u32 ddir_act[2] __read_mostly = { BLK_TC_ACT(BLK_TC_READ), BLK_TC_ACT(BLK_TC_WRITE) }; -/* - * Bio action bits of interest - */ -static u32 bio_act[17] __read_mostly = { - [1] = BLK_TC_ACT(BLK_TC_BARRIER), - [2] = BLK_TC_ACT(BLK_TC_SYNC), - [4] = BLK_TC_ACT(BLK_TC_AHEAD), - [8] = BLK_TC_ACT(BLK_TC_META), - [16] = BLK_TC_ACT(BLK_TC_DISCARD) -}; - -/* - * More could be added as needed, taking care to increment the decrementer - * to get correct indexing - */ -#define trace_barrier_bit(rw) \ - (((rw) & (1 << BIO_RW_BARRIER)) >> (BIO_RW_BARRIER - 0)) -#define trace_sync_bit(rw) \ - (((rw) & (1 << BIO_RW_SYNC)) >> (BIO_RW_SYNC - 1)) -#define trace_ahead_bit(rw) \ - (((rw) & (1 << BIO_RW_AHEAD)) << (2 - BIO_RW_AHEAD)) -#define trace_meta_bit(rw) \ - (((rw) & (1 << BIO_RW_META)) >> (BIO_RW_META - 3)) -#define trace_discard_bit(rw) \ - (((rw) & (1 << BIO_RW_DISCARD)) >> (BIO_RW_DISCARD - 4)) +/* The ilog2() calls fall out because they're constant */ +#define MASK_TC_BIT(rw, __name) ( (rw & (1 << BIO_RW_ ## __name)) << \ + (ilog2(BLK_TC_ ## __name) + BLK_TC_SHIFT - BIO_RW_ ## __name) ) /* * The worker for the various blk_add_trace*() types. Fills out a @@ -155,11 +133,11 @@ void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes, return; what |= ddir_act[rw & WRITE]; - what |= bio_act[trace_barrier_bit(rw)]; - what |= bio_act[trace_sync_bit(rw)]; - what |= bio_act[trace_ahead_bit(rw)]; - what |= bio_act[trace_meta_bit(rw)]; - what |= bio_act[trace_discard_bit(rw)]; + what |= MASK_TC_BIT(rw, BARRIER); + what |= MASK_TC_BIT(rw, SYNC); + what |= MASK_TC_BIT(rw, AHEAD); + what |= MASK_TC_BIT(rw, META); + what |= MASK_TC_BIT(rw, DISCARD); pid = tsk->pid; if (unlikely(act_log_check(bt, what, sector, pid))) -- cgit v1.2.3 From 2ebca85abcfcbaaf1c0b242e39fc88ad3da90090 Mon Sep 17 00:00:00 2001 From: OGAWA Hirofumi Date: Mon, 11 Aug 2008 17:07:08 +0100 Subject: Use WRITE_BARRIER in blkdev_issue_flush(), not (1< Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/blk-barrier.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'block') diff --git a/block/blk-barrier.c b/block/blk-barrier.c index 273121c0eb8..e5448131d4f 100644 --- a/block/blk-barrier.c +++ b/block/blk-barrier.c @@ -293,7 +293,7 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector) bio->bi_end_io = bio_end_empty_barrier; bio->bi_private = &wait; bio->bi_bdev = bdev; - submit_bio(1 << BIO_RW_BARRIER, bio); + submit_bio(WRITE_BARRIER, bio); wait_for_completion(&wait); -- cgit v1.2.3 From d30a2605be9d5132d95944916e8f578fcfe4f976 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Mon, 11 Aug 2008 15:58:42 +0100 Subject: Add BLKDISCARD ioctl to allow userspace to discard sectors We may well want mkfs tools to use this to mark the whole device as unwanted before they format it, for example. The ioctl takes a pair of uint64_ts, which are start offset and length in _bytes_. Although at the moment it might make sense for them both to be in 512-byte sectors, I don't want to limit the ABI to that. Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/compat_ioctl.c | 1 + block/ioctl.c | 76 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 77 insertions(+) (limited to 'block') diff --git a/block/compat_ioctl.c b/block/compat_ioctl.c index c23177e4623..1e559fba7bd 100644 --- a/block/compat_ioctl.c +++ b/block/compat_ioctl.c @@ -788,6 +788,7 @@ long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg) return compat_hdio_getgeo(disk, bdev, compat_ptr(arg)); case BLKFLSBUF: case BLKROSET: + case BLKDISCARD: /* * the ones below are implemented in blkdev_locked_ioctl, * but we call blkdev_ioctl, which gets the lock for us diff --git a/block/ioctl.c b/block/ioctl.c index 77185e5c026..342298bb608 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -111,6 +111,69 @@ static int blkdev_reread_part(struct block_device *bdev) return res; } +static void blk_ioc_discard_endio(struct bio *bio, int err) +{ + if (err) { + if (err == -EOPNOTSUPP) + set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); + clear_bit(BIO_UPTODATE, &bio->bi_flags); + } + complete(bio->bi_private); +} + +static int blk_ioctl_discard(struct block_device *bdev, uint64_t start, + uint64_t len) +{ + struct request_queue *q = bdev_get_queue(bdev); + int ret = 0; + + if (start & 511) + return -EINVAL; + if (len & 511) + return -EINVAL; + start >>= 9; + len >>= 9; + + if (start + len > (bdev->bd_inode->i_size >> 9)) + return -EINVAL; + + if (!q->prepare_discard_fn) + return -EOPNOTSUPP; + + while (len && !ret) { + DECLARE_COMPLETION_ONSTACK(wait); + struct bio *bio; + + bio = bio_alloc(GFP_KERNEL, 0); + if (!bio) + return -ENOMEM; + + bio->bi_end_io = blk_ioc_discard_endio; + bio->bi_bdev = bdev; + bio->bi_private = &wait; + bio->bi_sector = start; + + if (len > q->max_hw_sectors) { + bio->bi_size = q->max_hw_sectors << 9; + len -= q->max_hw_sectors; + start += q->max_hw_sectors; + } else { + bio->bi_size = len << 9; + len = 0; + } + submit_bio(WRITE_DISCARD, bio); + + wait_for_completion(&wait); + + if (bio_flagged(bio, BIO_EOPNOTSUPP)) + ret = -EOPNOTSUPP; + else if (!bio_flagged(bio, BIO_UPTODATE)) + ret = -EIO; + bio_put(bio); + } + return ret; +} + static int put_ushort(unsigned long arg, unsigned short val) { return put_user(val, (unsigned short __user *)arg); @@ -258,6 +321,19 @@ int blkdev_ioctl(struct inode *inode, struct file *file, unsigned cmd, set_device_ro(bdev, n); unlock_kernel(); return 0; + + case BLKDISCARD: { + uint64_t range[2]; + + if (!(file->f_mode & FMODE_WRITE)) + return -EBADF; + + if (copy_from_user(range, (void __user *)arg, sizeof(range))) + return -EFAULT; + + return blk_ioctl_discard(bdev, range[0], range[1]); + } + case HDIO_GETGEO: { struct hd_geometry geo; -- cgit v1.2.3 From e17fc0a1ccf88f6d4dcb363729f3141b0958c325 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sat, 9 Aug 2008 16:42:20 +0100 Subject: Allow elevators to sort/merge discard requests MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit But blkdev_issue_discard() still emits requests which are interpreted as soft barriers, because naïve callers might otherwise issue subsequent writes to those same sectors, which might cross on the queue (if they're reallocated quickly enough). Callers still _can_ issue non-barrier discard requests, but they have to take care of queue ordering for themselves. Signed-off-by: David Woodhouse Signed-off-by: Jens Axboe --- block/blk-barrier.c | 2 +- block/blk-core.c | 12 +++++++----- block/blk-merge.c | 27 +++++++++++++++++---------- block/elevator.c | 12 ++++++++++-- block/ioctl.c | 2 +- 5 files changed, 36 insertions(+), 19 deletions(-) (limited to 'block') diff --git a/block/blk-barrier.c b/block/blk-barrier.c index e5448131d4f..988b63479b2 100644 --- a/block/blk-barrier.c +++ b/block/blk-barrier.c @@ -372,7 +372,7 @@ int blkdev_issue_discard(struct block_device *bdev, sector_t sector, nr_sects = 0; } bio_get(bio); - submit_bio(WRITE_DISCARD, bio); + submit_bio(DISCARD_BARRIER, bio); /* Check if it failed immediately */ if (bio_flagged(bio, BIO_EOPNOTSUPP)) diff --git a/block/blk-core.c b/block/blk-core.c index 1e143c4f9d3..1261516dd42 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1077,12 +1077,13 @@ void init_request_from_bio(struct request *req, struct bio *bio) /* * REQ_BARRIER implies no merging, but lets make it explicit */ - if (unlikely(bio_barrier(bio))) - req->cmd_flags |= (REQ_HARDBARRIER | REQ_NOMERGE); if (unlikely(bio_discard(bio))) { - req->cmd_flags |= (REQ_SOFTBARRIER | REQ_DISCARD); + req->cmd_flags |= REQ_DISCARD; + if (bio_barrier(bio)) + req->cmd_flags |= REQ_SOFTBARRIER; req->q->prepare_discard_fn(req->q, req); - } + } else if (unlikely(bio_barrier(bio))) + req->cmd_flags |= (REQ_HARDBARRIER | REQ_NOMERGE); if (bio_sync(bio)) req->cmd_flags |= REQ_RW_SYNC; @@ -1114,7 +1115,8 @@ static int __make_request(struct request_queue *q, struct bio *bio) blk_queue_bounce(q, &bio); barrier = bio_barrier(bio); - if (unlikely(barrier) && (q->next_ordered == QUEUE_ORDERED_NONE)) { + if (unlikely(barrier) && bio_has_data(bio) && + (q->next_ordered == QUEUE_ORDERED_NONE)) { err = -EOPNOTSUPP; goto end_io; } diff --git a/block/blk-merge.c b/block/blk-merge.c index 5efc9e7a68b..6cf8f0c70a5 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -11,7 +11,7 @@ void blk_recalc_rq_sectors(struct request *rq, int nsect) { - if (blk_fs_request(rq)) { + if (blk_fs_request(rq) || blk_discard_rq(rq)) { rq->hard_sector += nsect; rq->hard_nr_sectors -= nsect; @@ -131,13 +131,17 @@ static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio, if (!test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags)) return 0; - if (!BIOVEC_PHYS_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt))) - return 0; if (bio->bi_size + nxt->bi_size > q->max_segment_size) return 0; + if (!bio_has_data(bio)) + return 1; + + if (!BIOVEC_PHYS_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt))) + return 0; + /* - * bio and nxt are contigous in memory, check if the queue allows + * bio and nxt are contiguous in memory; check if the queue allows * these two to be merged into one */ if (BIO_SEG_BOUNDARY(q, bio, nxt)) @@ -153,8 +157,9 @@ static int blk_hw_contig_segment(struct request_queue *q, struct bio *bio, blk_recount_segments(q, bio); if (!bio_flagged(nxt, BIO_SEG_VALID)) blk_recount_segments(q, nxt); - if (!BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)) || - BIOVEC_VIRT_OVERSIZE(bio->bi_hw_back_size + nxt->bi_hw_front_size)) + if (bio_has_data(bio) && + (!BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)) || + BIOVEC_VIRT_OVERSIZE(bio->bi_hw_back_size + nxt->bi_hw_front_size))) return 0; if (bio->bi_hw_back_size + nxt->bi_hw_front_size > q->max_segment_size) return 0; @@ -317,8 +322,9 @@ int ll_back_merge_fn(struct request_queue *q, struct request *req, if (!bio_flagged(bio, BIO_SEG_VALID)) blk_recount_segments(q, bio); len = req->biotail->bi_hw_back_size + bio->bi_hw_front_size; - if (BIOVEC_VIRT_MERGEABLE(__BVEC_END(req->biotail), __BVEC_START(bio)) - && !BIOVEC_VIRT_OVERSIZE(len)) { + if (!bio_has_data(bio) || + (BIOVEC_VIRT_MERGEABLE(__BVEC_END(req->biotail), __BVEC_START(bio)) + && !BIOVEC_VIRT_OVERSIZE(len))) { int mergeable = ll_new_mergeable(q, req, bio); if (mergeable) { @@ -356,8 +362,9 @@ int ll_front_merge_fn(struct request_queue *q, struct request *req, blk_recount_segments(q, bio); if (!bio_flagged(req->bio, BIO_SEG_VALID)) blk_recount_segments(q, req->bio); - if (BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(req->bio)) && - !BIOVEC_VIRT_OVERSIZE(len)) { + if (!bio_has_data(bio) || + (BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(req->bio)) && + !BIOVEC_VIRT_OVERSIZE(len))) { int mergeable = ll_new_mergeable(q, req, bio); if (mergeable) { diff --git a/block/elevator.c b/block/elevator.c index ed6f8f32d27..4f5127054e3 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -74,6 +74,12 @@ int elv_rq_merge_ok(struct request *rq, struct bio *bio) if (!rq_mergeable(rq)) return 0; + /* + * Don't merge file system requests and discard requests + */ + if (bio_discard(bio) != bio_discard(rq->bio)) + return 0; + /* * different data direction or already started, don't merge */ @@ -438,6 +444,8 @@ void elv_dispatch_sort(struct request_queue *q, struct request *rq) list_for_each_prev(entry, &q->queue_head) { struct request *pos = list_entry_rq(entry); + if (blk_discard_rq(rq) != blk_discard_rq(pos)) + break; if (rq_data_dir(rq) != rq_data_dir(pos)) break; if (pos->cmd_flags & stop_flags) @@ -607,7 +615,7 @@ void elv_insert(struct request_queue *q, struct request *rq, int where) break; case ELEVATOR_INSERT_SORT: - BUG_ON(!blk_fs_request(rq)); + BUG_ON(!blk_fs_request(rq) && !blk_discard_rq(rq)); rq->cmd_flags |= REQ_SORTED; q->nr_sorted++; if (rq_mergeable(rq)) { @@ -692,7 +700,7 @@ void __elv_add_request(struct request_queue *q, struct request *rq, int where, * this request is scheduling boundary, update * end_sector */ - if (blk_fs_request(rq)) { + if (blk_fs_request(rq) || blk_discard_rq(rq)) { q->end_sector = rq_end_sector(rq); q->boundary_rq = rq; } diff --git a/block/ioctl.c b/block/ioctl.c index 342298bb608..375c57922b0 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -161,7 +161,7 @@ static int blk_ioctl_discard(struct block_device *bdev, uint64_t start, bio->bi_size = len << 9; len = 0; } - submit_bio(WRITE_DISCARD, bio); + submit_bio(DISCARD_NOBARRIER, bio); wait_for_completion(&wait); -- cgit v1.2.3 From 63de428b139d3d31d86ebe25ae97b33f6540fb7e Mon Sep 17 00:00:00 2001 From: Aaron Carroll Date: Thu, 14 Aug 2008 18:17:13 +1000 Subject: deadline-iosched: allow non-sequential batching Deadline currently only batches sector-contiguous requests, so except for a few circumstances (e.g. requests in a single direction), it is essentially first come first served. This is bad for throughput, so change it to CSCAN, which means requests in a batch do not need to be sequential and are issued in increasing sector order. Signed-off-by: Aaron Carroll Signed-off-by: Jens Axboe --- block/deadline-iosched.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) (limited to 'block') diff --git a/block/deadline-iosched.c b/block/deadline-iosched.c index 342448c3d2d..07b80e4642f 100644 --- a/block/deadline-iosched.c +++ b/block/deadline-iosched.c @@ -258,17 +258,9 @@ static int deadline_dispatch_requests(struct request_queue *q, int force) else rq = dd->next_rq[READ]; - if (rq) { - /* we have a "next request" */ - - if (dd->last_sector != rq->sector) - /* end the batch on a non sequential request */ - dd->batching += dd->fifo_batch; - - if (dd->batching < dd->fifo_batch) - /* we are still entitled to batch */ - goto dispatch_request; - } + if (rq && dd->batching < dd->fifo_batch) + /* we have a next request are still entitled to batch */ + goto dispatch_request; /* * at this point we are not running a batch. select the appropriate -- cgit v1.2.3 From 4fb72f7646e86874eb2798256eaa6bf3fbe4edcf Mon Sep 17 00:00:00 2001 From: Aaron Carroll Date: Thu, 14 Aug 2008 18:17:14 +1000 Subject: deadline-iosched: non-functional fixes * convert goto to simpler while loop; * use rq_end_sector() instead of computing manually; * fix false comments; * remove spurious whitespace; * convert rq_rb_root macro to an inline function. Signed-off-by: Aaron Carroll Signed-off-by: Jens Axboe --- block/deadline-iosched.c | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) (limited to 'block') diff --git a/block/deadline-iosched.c b/block/deadline-iosched.c index 07b80e4642f..fd311179f44 100644 --- a/block/deadline-iosched.c +++ b/block/deadline-iosched.c @@ -33,7 +33,7 @@ struct deadline_data { */ struct rb_root sort_list[2]; struct list_head fifo_list[2]; - + /* * next in sort order. read, write or both are NULL */ @@ -53,7 +53,11 @@ struct deadline_data { static void deadline_move_request(struct deadline_data *, struct request *); -#define RQ_RB_ROOT(dd, rq) (&(dd)->sort_list[rq_data_dir((rq))]) +static inline struct rb_root * +deadline_rb_root(struct deadline_data *dd, struct request *rq) +{ + return &dd->sort_list[rq_data_dir(rq)]; +} /* * get the request after `rq' in sector-sorted order @@ -72,15 +76,11 @@ deadline_latter_request(struct request *rq) static void deadline_add_rq_rb(struct deadline_data *dd, struct request *rq) { - struct rb_root *root = RQ_RB_ROOT(dd, rq); + struct rb_root *root = deadline_rb_root(dd, rq); struct request *__alias; -retry: - __alias = elv_rb_add(root, rq); - if (unlikely(__alias)) { + while (unlikely(__alias = elv_rb_add(root, rq))) deadline_move_request(dd, __alias); - goto retry; - } } static inline void @@ -91,7 +91,7 @@ deadline_del_rq_rb(struct deadline_data *dd, struct request *rq) if (dd->next_rq[data_dir] == rq) dd->next_rq[data_dir] = deadline_latter_request(rq); - elv_rb_del(RQ_RB_ROOT(dd, rq), rq); + elv_rb_del(deadline_rb_root(dd, rq), rq); } /* @@ -106,7 +106,7 @@ deadline_add_request(struct request_queue *q, struct request *rq) deadline_add_rq_rb(dd, rq); /* - * set expire time (only used for reads) and add to fifo list + * set expire time and add to fifo list */ rq_set_fifo_time(rq, jiffies + dd->fifo_expire[data_dir]); list_add_tail(&rq->queuelist, &dd->fifo_list[data_dir]); @@ -162,7 +162,7 @@ static void deadline_merged_request(struct request_queue *q, * if the merge was a front merge, we need to reposition request */ if (type == ELEVATOR_FRONT_MERGE) { - elv_rb_del(RQ_RB_ROOT(dd, req), req); + elv_rb_del(deadline_rb_root(dd, req), req); deadline_add_rq_rb(dd, req); } } @@ -212,7 +212,7 @@ deadline_move_request(struct deadline_data *dd, struct request *rq) dd->next_rq[WRITE] = NULL; dd->next_rq[data_dir] = deadline_latter_request(rq); - dd->last_sector = rq->sector + rq->nr_sectors; + dd->last_sector = rq_end_sector(rq); /* * take it off the sort and fifo list, move @@ -222,7 +222,7 @@ deadline_move_request(struct deadline_data *dd, struct request *rq) } /* - * deadline_check_fifo returns 0 if there are no expired reads on the fifo, + * deadline_check_fifo returns 0 if there are no expired requests on the fifo, * 1 otherwise. Requires !list_empty(&dd->fifo_list[data_dir]) */ static inline int deadline_check_fifo(struct deadline_data *dd, int ddir) -- cgit v1.2.3 From b8b3e16cfe6435d961f6aaebcfd52a1ff2a988c5 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Fri, 15 Aug 2008 10:15:19 +0200 Subject: block: drop virtual merging accounting Remove virtual merge accounting. Signed-off-by: Mikulas Patocka Signed-off-by: Jens Axboe --- block/blk-merge.c | 79 +++++-------------------------------------------------- 1 file changed, 6 insertions(+), 73 deletions(-) (limited to 'block') diff --git a/block/blk-merge.c b/block/blk-merge.c index 6cf8f0c70a5..2c2a2ee716e 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -66,7 +66,7 @@ void blk_recalc_rq_segments(struct request *rq) */ high = page_to_pfn(bv->bv_page) > q->bounce_pfn; if (high || highprv) - goto new_hw_segment; + goto new_segment; if (cluster) { if (seg_size + bv->bv_len > q->max_segment_size) goto new_segment; @@ -74,8 +74,6 @@ void blk_recalc_rq_segments(struct request *rq) goto new_segment; if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bv)) goto new_segment; - if (BIOVEC_VIRT_OVERSIZE(hw_seg_size + bv->bv_len)) - goto new_hw_segment; seg_size += bv->bv_len; hw_seg_size += bv->bv_len; @@ -83,17 +81,11 @@ void blk_recalc_rq_segments(struct request *rq) continue; } new_segment: - if (BIOVEC_VIRT_MERGEABLE(bvprv, bv) && - !BIOVEC_VIRT_OVERSIZE(hw_seg_size + bv->bv_len)) - hw_seg_size += bv->bv_len; - else { -new_hw_segment: - if (nr_hw_segs == 1 && - hw_seg_size > rq->bio->bi_hw_front_size) - rq->bio->bi_hw_front_size = hw_seg_size; - hw_seg_size = BIOVEC_VIRT_START_SIZE(bv) + bv->bv_len; - nr_hw_segs++; - } + if (nr_hw_segs == 1 && + hw_seg_size > rq->bio->bi_hw_front_size) + rq->bio->bi_hw_front_size = hw_seg_size; + hw_seg_size = bv->bv_len; + nr_hw_segs++; nr_phys_segs++; bvprv = bv; @@ -150,23 +142,6 @@ static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio, return 0; } -static int blk_hw_contig_segment(struct request_queue *q, struct bio *bio, - struct bio *nxt) -{ - if (!bio_flagged(bio, BIO_SEG_VALID)) - blk_recount_segments(q, bio); - if (!bio_flagged(nxt, BIO_SEG_VALID)) - blk_recount_segments(q, nxt); - if (bio_has_data(bio) && - (!BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(nxt)) || - BIOVEC_VIRT_OVERSIZE(bio->bi_hw_back_size + nxt->bi_hw_front_size))) - return 0; - if (bio->bi_hw_back_size + nxt->bi_hw_front_size > q->max_segment_size) - return 0; - - return 1; -} - /* * map a request to scatterlist, return number of sg entries setup. Caller * must make sure sg can hold rq->nr_phys_segments entries @@ -304,7 +279,6 @@ int ll_back_merge_fn(struct request_queue *q, struct request *req, struct bio *bio) { unsigned short max_sectors; - int len; if (unlikely(blk_pc_request(req))) max_sectors = q->max_hw_sectors; @@ -321,20 +295,6 @@ int ll_back_merge_fn(struct request_queue *q, struct request *req, blk_recount_segments(q, req->biotail); if (!bio_flagged(bio, BIO_SEG_VALID)) blk_recount_segments(q, bio); - len = req->biotail->bi_hw_back_size + bio->bi_hw_front_size; - if (!bio_has_data(bio) || - (BIOVEC_VIRT_MERGEABLE(__BVEC_END(req->biotail), __BVEC_START(bio)) - && !BIOVEC_VIRT_OVERSIZE(len))) { - int mergeable = ll_new_mergeable(q, req, bio); - - if (mergeable) { - if (req->nr_hw_segments == 1) - req->bio->bi_hw_front_size = len; - if (bio->bi_hw_segments == 1) - bio->bi_hw_back_size = len; - } - return mergeable; - } return ll_new_hw_segment(q, req, bio); } @@ -343,7 +303,6 @@ int ll_front_merge_fn(struct request_queue *q, struct request *req, struct bio *bio) { unsigned short max_sectors; - int len; if (unlikely(blk_pc_request(req))) max_sectors = q->max_hw_sectors; @@ -357,24 +316,10 @@ int ll_front_merge_fn(struct request_queue *q, struct request *req, q->last_merge = NULL; return 0; } - len = bio->bi_hw_back_size + req->bio->bi_hw_front_size; if (!bio_flagged(bio, BIO_SEG_VALID)) blk_recount_segments(q, bio); if (!bio_flagged(req->bio, BIO_SEG_VALID)) blk_recount_segments(q, req->bio); - if (!bio_has_data(bio) || - (BIOVEC_VIRT_MERGEABLE(__BVEC_END(bio), __BVEC_START(req->bio)) && - !BIOVEC_VIRT_OVERSIZE(len))) { - int mergeable = ll_new_mergeable(q, req, bio); - - if (mergeable) { - if (bio->bi_hw_segments == 1) - bio->bi_hw_front_size = len; - if (req->nr_hw_segments == 1) - req->biotail->bi_hw_back_size = len; - } - return mergeable; - } return ll_new_hw_segment(q, req, bio); } @@ -406,18 +351,6 @@ static int ll_merge_requests_fn(struct request_queue *q, struct request *req, return 0; total_hw_segments = req->nr_hw_segments + next->nr_hw_segments; - if (blk_hw_contig_segment(q, req->biotail, next->bio)) { - int len = req->biotail->bi_hw_back_size + - next->bio->bi_hw_front_size; - /* - * propagate the combined length to the end of the requests - */ - if (req->nr_hw_segments == 1) - req->bio->bi_hw_front_size = len; - if (next->nr_hw_segments == 1) - next->biotail->bi_hw_back_size = len; - total_hw_segments--; - } if (total_hw_segments > q->max_hw_segments) return 0; -- cgit v1.2.3 From 5df97b91b5d7ed426034fcc84cb6e7cf682b8838 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Fri, 15 Aug 2008 10:20:02 +0200 Subject: drop vmerge accounting Remove hw_segments field from struct bio and struct request. Without virtual merge accounting they have no purpose. Signed-off-by: Mikulas Patocka Signed-off-by: Jens Axboe --- block/blk-core.c | 1 - block/blk-merge.c | 31 ++++--------------------------- block/elevator.c | 2 -- 3 files changed, 4 insertions(+), 30 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 1261516dd42..2616cdd049a 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -2026,7 +2026,6 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq, if (bio_has_data(bio)) { rq->nr_phys_segments = bio_phys_segments(q, bio); - rq->nr_hw_segments = bio_hw_segments(q, bio); rq->buffer = bio_data(bio); } rq->current_nr_sectors = bio_cur_sectors(bio); diff --git a/block/blk-merge.c b/block/blk-merge.c index 2c2a2ee716e..d81d91419ff 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -41,12 +41,9 @@ void blk_recalc_rq_sectors(struct request *rq, int nsect) void blk_recalc_rq_segments(struct request *rq) { int nr_phys_segs; - int nr_hw_segs; unsigned int phys_size; - unsigned int hw_size; struct bio_vec *bv, *bvprv = NULL; int seg_size; - int hw_seg_size; int cluster; struct req_iterator iter; int high, highprv = 1; @@ -56,8 +53,8 @@ void blk_recalc_rq_segments(struct request *rq) return; cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags); - hw_seg_size = seg_size = 0; - phys_size = hw_size = nr_phys_segs = nr_hw_segs = 0; + seg_size = 0; + phys_size = nr_phys_segs = 0; rq_for_each_segment(bv, rq, iter) { /* * the trick here is making sure that a high page is never @@ -76,30 +73,17 @@ void blk_recalc_rq_segments(struct request *rq) goto new_segment; seg_size += bv->bv_len; - hw_seg_size += bv->bv_len; bvprv = bv; continue; } new_segment: - if (nr_hw_segs == 1 && - hw_seg_size > rq->bio->bi_hw_front_size) - rq->bio->bi_hw_front_size = hw_seg_size; - hw_seg_size = bv->bv_len; - nr_hw_segs++; - nr_phys_segs++; bvprv = bv; seg_size = bv->bv_len; highprv = high; } - if (nr_hw_segs == 1 && - hw_seg_size > rq->bio->bi_hw_front_size) - rq->bio->bi_hw_front_size = hw_seg_size; - if (hw_seg_size > rq->biotail->bi_hw_back_size) - rq->biotail->bi_hw_back_size = hw_seg_size; rq->nr_phys_segments = nr_phys_segs; - rq->nr_hw_segments = nr_hw_segs; } void blk_recount_segments(struct request_queue *q, struct bio *bio) @@ -112,7 +96,6 @@ void blk_recount_segments(struct request_queue *q, struct bio *bio) blk_recalc_rq_segments(&rq); bio->bi_next = nxt; bio->bi_phys_segments = rq.nr_phys_segments; - bio->bi_hw_segments = rq.nr_hw_segments; bio->bi_flags |= (1 << BIO_SEG_VALID); } EXPORT_SYMBOL(blk_recount_segments); @@ -255,10 +238,9 @@ static inline int ll_new_hw_segment(struct request_queue *q, struct request *req, struct bio *bio) { - int nr_hw_segs = bio_hw_segments(q, bio); int nr_phys_segs = bio_phys_segments(q, bio); - if (req->nr_hw_segments + nr_hw_segs > q->max_hw_segments + if (req->nr_phys_segments + nr_phys_segs > q->max_hw_segments || req->nr_phys_segments + nr_phys_segs > q->max_phys_segments) { req->cmd_flags |= REQ_NOMERGE; if (req == q->last_merge) @@ -270,7 +252,6 @@ static inline int ll_new_hw_segment(struct request_queue *q, * This will form the start of a new hw segment. Bump both * counters. */ - req->nr_hw_segments += nr_hw_segs; req->nr_phys_segments += nr_phys_segs; return 1; } @@ -328,7 +309,6 @@ static int ll_merge_requests_fn(struct request_queue *q, struct request *req, struct request *next) { int total_phys_segments; - int total_hw_segments; /* * First check if the either of the requests are re-queued @@ -350,14 +330,11 @@ static int ll_merge_requests_fn(struct request_queue *q, struct request *req, if (total_phys_segments > q->max_phys_segments) return 0; - total_hw_segments = req->nr_hw_segments + next->nr_hw_segments; - - if (total_hw_segments > q->max_hw_segments) + if (total_phys_segments > q->max_hw_segments) return 0; /* Merge is OK... */ req->nr_phys_segments = total_phys_segments; - req->nr_hw_segments = total_hw_segments; return 1; } diff --git a/block/elevator.c b/block/elevator.c index 4f5127054e3..269615e6dbf 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -790,7 +790,6 @@ struct request *elv_next_request(struct request_queue *q) * device can handle */ rq->nr_phys_segments++; - rq->nr_hw_segments++; } if (!q->prep_rq_fn) @@ -813,7 +812,6 @@ struct request *elv_next_request(struct request_queue *q) * so that we don't add it again */ --rq->nr_phys_segments; - --rq->nr_hw_segments; } rq = NULL; -- cgit v1.2.3 From 710027a48ede75428cc68eaa8ae2269b1e356e2c Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Tue, 19 Aug 2008 20:13:11 +0200 Subject: Add some block/ source files to the kernel-api docbook. Fix kernel-doc notation in them as needed. Fix changed function parameter names. Fix typos/spellos. In comments, change REQ_SPECIAL to REQ_TYPE_SPECIAL and REQ_BLOCK_PC to REQ_TYPE_BLOCK_PC. Signed-off-by: Randy Dunlap Signed-off-by: Jens Axboe --- block/blk-core.c | 72 +++++++++++++++++++++++++-------------------------- block/blk-exec.c | 6 ++--- block/blk-integrity.c | 4 +-- block/blk-map.c | 16 ++++++------ block/blk-settings.c | 8 +++--- block/blk-tag.c | 8 +++--- block/genhd.c | 5 ++-- 7 files changed, 60 insertions(+), 59 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 2616cdd049a..86d22e7d65c 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -531,7 +531,7 @@ EXPORT_SYMBOL(blk_alloc_queue_node); * request queue; this lock will be taken also from interrupt context, so irq * disabling is needed for it. * - * Function returns a pointer to the initialized request queue, or NULL if + * Function returns a pointer to the initialized request queue, or %NULL if * it didn't succeed. * * Note: @@ -913,7 +913,7 @@ void blk_requeue_request(struct request_queue *q, struct request *rq) EXPORT_SYMBOL(blk_requeue_request); /** - * blk_insert_request - insert a special request in to a request queue + * blk_insert_request - insert a special request into a request queue * @q: request queue where request should be inserted * @rq: request to be inserted * @at_head: insert request at head or tail of queue @@ -923,8 +923,8 @@ EXPORT_SYMBOL(blk_requeue_request); * Many block devices need to execute commands asynchronously, so they don't * block the whole kernel from preemption during request execution. This is * accomplished normally by inserting aritficial requests tagged as - * REQ_SPECIAL in to the corresponding request queue, and letting them be - * scheduled for actual execution by the request queue. + * REQ_TYPE_SPECIAL in to the corresponding request queue, and letting them + * be scheduled for actual execution by the request queue. * * We have the option of inserting the head or the tail of the queue. * Typically we use the tail for new ioctls and so forth. We use the head @@ -1322,7 +1322,7 @@ static inline int bio_check_eod(struct bio *bio, unsigned int nr_sectors) } /** - * generic_make_request: hand a buffer to its device driver for I/O + * generic_make_request - hand a buffer to its device driver for I/O * @bio: The bio describing the location in memory and on the device. * * generic_make_request() is used to make I/O requests of block @@ -1480,13 +1480,13 @@ void generic_make_request(struct bio *bio) EXPORT_SYMBOL(generic_make_request); /** - * submit_bio: submit a bio to the block device layer for I/O + * submit_bio - submit a bio to the block device layer for I/O * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead) * @bio: The &struct bio which describes the I/O * * submit_bio() is very similar in purpose to generic_make_request(), and * uses that function to do most of the work. Both are fairly rough - * interfaces, @bio must be presetup and ready for I/O. + * interfaces; @bio must be presetup and ready for I/O. * */ void submit_bio(int rw, struct bio *bio) @@ -1524,7 +1524,7 @@ EXPORT_SYMBOL(submit_bio); /** * __end_that_request_first - end I/O on a request * @req: the request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete * * Description: @@ -1532,8 +1532,8 @@ EXPORT_SYMBOL(submit_bio); * for the next range of segments (if any) in the cluster. * * Return: - * 0 - we are done with this request, call end_that_request_last() - * 1 - still buffers pending for this request + * %0 - we are done with this request, call end_that_request_last() + * %1 - still buffers pending for this request **/ static int __end_that_request_first(struct request *req, int error, int nr_bytes) @@ -1544,7 +1544,7 @@ static int __end_that_request_first(struct request *req, int error, blk_add_trace_rq(req->q, req, BLK_TA_COMPLETE); /* - * for a REQ_BLOCK_PC request, we want to carry any eventual + * for a REQ_TYPE_BLOCK_PC request, we want to carry any eventual * sense key with us all the way through */ if (!blk_pc_request(req)) @@ -1810,11 +1810,11 @@ EXPORT_SYMBOL_GPL(blk_rq_cur_bytes); /** * end_queued_request - end all I/O on a queued request * @rq: the request being processed - * @uptodate: error value or 0/1 uptodate flag + * @uptodate: error value or %0/%1 uptodate flag * * Description: * Ends all I/O on a request, and removes it from the block layer queues. - * Not suitable for normal IO completion, unless the driver still has + * Not suitable for normal I/O completion, unless the driver still has * the request attached to the block layer. * **/ @@ -1827,7 +1827,7 @@ EXPORT_SYMBOL(end_queued_request); /** * end_dequeued_request - end all I/O on a dequeued request * @rq: the request being processed - * @uptodate: error value or 0/1 uptodate flag + * @uptodate: error value or %0/%1 uptodate flag * * Description: * Ends all I/O on a request. The request must already have been @@ -1845,14 +1845,14 @@ EXPORT_SYMBOL(end_dequeued_request); /** * end_request - end I/O on the current segment of the request * @req: the request being processed - * @uptodate: error value or 0/1 uptodate flag + * @uptodate: error value or %0/%1 uptodate flag * * Description: * Ends I/O on the current segment of a request. If that is the only * remaining segment, the request is also completed and freed. * - * This is a remnant of how older block drivers handled IO completions. - * Modern drivers typically end IO on the full request in one go, unless + * This is a remnant of how older block drivers handled I/O completions. + * Modern drivers typically end I/O on the full request in one go, unless * they have a residual value to account for. For that case this function * isn't really useful, unless the residual just happens to be the * full current segment. In other words, don't use this function in new @@ -1870,12 +1870,12 @@ EXPORT_SYMBOL(end_request); /** * blk_end_io - Generic end_io function to complete a request. * @rq: the request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete @rq * @bidi_bytes: number of bytes to complete @rq->next_rq * @drv_callback: function called between completion of bios in the request * and completion of the request. - * If the callback returns non 0, this helper returns without + * If the callback returns non %0, this helper returns without * completion of the request. * * Description: @@ -1883,8 +1883,8 @@ EXPORT_SYMBOL(end_request); * If @rq has leftover, sets it up for the next range of segments. * * Return: - * 0 - we are done with this request - * 1 - this request is not freed yet, it still has pending buffers. + * %0 - we are done with this request + * %1 - this request is not freed yet, it still has pending buffers. **/ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, unsigned int bidi_bytes, @@ -1919,7 +1919,7 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, /** * blk_end_request - Helper function for drivers to complete the request. * @rq: the request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete * * Description: @@ -1927,8 +1927,8 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, * If @rq has leftover, sets it up for the next range of segments. * * Return: - * 0 - we are done with this request - * 1 - still buffers pending for this request + * %0 - we are done with this request + * %1 - still buffers pending for this request **/ int blk_end_request(struct request *rq, int error, unsigned int nr_bytes) { @@ -1939,15 +1939,15 @@ EXPORT_SYMBOL_GPL(blk_end_request); /** * __blk_end_request - Helper function for drivers to complete the request. * @rq: the request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete * * Description: * Must be called with queue lock held unlike blk_end_request(). * * Return: - * 0 - we are done with this request - * 1 - still buffers pending for this request + * %0 - we are done with this request + * %1 - still buffers pending for this request **/ int __blk_end_request(struct request *rq, int error, unsigned int nr_bytes) { @@ -1966,7 +1966,7 @@ EXPORT_SYMBOL_GPL(__blk_end_request); /** * blk_end_bidi_request - Helper function for drivers to complete bidi request. * @rq: the bidi request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete @rq * @bidi_bytes: number of bytes to complete @rq->next_rq * @@ -1974,8 +1974,8 @@ EXPORT_SYMBOL_GPL(__blk_end_request); * Ends I/O on a number of bytes attached to @rq and @rq->next_rq. * * Return: - * 0 - we are done with this request - * 1 - still buffers pending for this request + * %0 - we are done with this request + * %1 - still buffers pending for this request **/ int blk_end_bidi_request(struct request *rq, int error, unsigned int nr_bytes, unsigned int bidi_bytes) @@ -1987,11 +1987,11 @@ EXPORT_SYMBOL_GPL(blk_end_bidi_request); /** * blk_end_request_callback - Special helper function for tricky drivers * @rq: the request being processed - * @error: 0 for success, < 0 for error + * @error: %0 for success, < %0 for error * @nr_bytes: number of bytes to complete * @drv_callback: function called between completion of bios in the request * and completion of the request. - * If the callback returns non 0, this helper returns without + * If the callback returns non %0, this helper returns without * completion of the request. * * Description: @@ -2004,10 +2004,10 @@ EXPORT_SYMBOL_GPL(blk_end_bidi_request); * Don't use this interface in other places anymore. * * Return: - * 0 - we are done with this request - * 1 - this request is not freed yet. - * this request still has pending buffers or - * the driver doesn't want to finish this request yet. + * %0 - we are done with this request + * %1 - this request is not freed yet. + * this request still has pending buffers or + * the driver doesn't want to finish this request yet. **/ int blk_end_request_callback(struct request *rq, int error, unsigned int nr_bytes, diff --git a/block/blk-exec.c b/block/blk-exec.c index 9bceff7674f..6af716d1e54 100644 --- a/block/blk-exec.c +++ b/block/blk-exec.c @@ -16,7 +16,7 @@ /** * blk_end_sync_rq - executes a completion event on a request * @rq: request to complete - * @error: end io status of the request + * @error: end I/O status of the request */ static void blk_end_sync_rq(struct request *rq, int error) { @@ -41,7 +41,7 @@ static void blk_end_sync_rq(struct request *rq, int error) * @done: I/O completion handler * * Description: - * Insert a fully prepared request at the back of the io scheduler queue + * Insert a fully prepared request at the back of the I/O scheduler queue * for execution. Don't wait for completion. */ void blk_execute_rq_nowait(struct request_queue *q, struct gendisk *bd_disk, @@ -72,7 +72,7 @@ EXPORT_SYMBOL_GPL(blk_execute_rq_nowait); * @at_head: insert request at head or tail of queue * * Description: - * Insert a fully prepared request at the back of the io scheduler queue + * Insert a fully prepared request at the back of the I/O scheduler queue * for execution and wait for completion. */ int blk_execute_rq(struct request_queue *q, struct gendisk *bd_disk, diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 3f1a8478cc3..d87606eaca1 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -109,8 +109,8 @@ EXPORT_SYMBOL(blk_rq_map_integrity_sg); /** * blk_integrity_compare - Compare integrity profile of two block devices - * @b1: Device to compare - * @b2: Device to compare + * @bd1: Device to compare + * @bd2: Device to compare * * Description: Meta-devices like DM and MD need to verify that all * sub-devices use the same integrity format before advertising to diff --git a/block/blk-map.c b/block/blk-map.c index af37e4ae62f..ea1bf53929e 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -85,17 +85,17 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, } /** - * blk_rq_map_user - map user data to a request, for REQ_BLOCK_PC usage + * blk_rq_map_user - map user data to a request, for REQ_TYPE_BLOCK_PC usage * @q: request queue where request should be inserted * @rq: request structure to fill * @ubuf: the user buffer * @len: length of user data * * Description: - * Data will be mapped directly for zero copy io, if possible. Otherwise + * Data will be mapped directly for zero copy I/O, if possible. Otherwise * a kernel bounce buffer is used. * - * A matching blk_rq_unmap_user() must be issued at the end of io, while + * A matching blk_rq_unmap_user() must be issued at the end of I/O, while * still in process context. * * Note: The mapped bio may need to be bounced through blk_queue_bounce() @@ -154,7 +154,7 @@ unmap_rq: EXPORT_SYMBOL(blk_rq_map_user); /** - * blk_rq_map_user_iov - map user data to a request, for REQ_BLOCK_PC usage + * blk_rq_map_user_iov - map user data to a request, for REQ_TYPE_BLOCK_PC usage * @q: request queue where request should be inserted * @rq: request to map data to * @iov: pointer to the iovec @@ -162,10 +162,10 @@ EXPORT_SYMBOL(blk_rq_map_user); * @len: I/O byte count * * Description: - * Data will be mapped directly for zero copy io, if possible. Otherwise + * Data will be mapped directly for zero copy I/O, if possible. Otherwise * a kernel bounce buffer is used. * - * A matching blk_rq_unmap_user() must be issued at the end of io, while + * A matching blk_rq_unmap_user() must be issued at the end of I/O, while * still in process context. * * Note: The mapped bio may need to be bounced through blk_queue_bounce() @@ -224,7 +224,7 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, * Description: * Unmap a rq previously mapped by blk_rq_map_user(). The caller must * supply the original rq->bio from the blk_rq_map_user() return, since - * the io completion may have changed rq->bio. + * the I/O completion may have changed rq->bio. */ int blk_rq_unmap_user(struct bio *bio) { @@ -250,7 +250,7 @@ int blk_rq_unmap_user(struct bio *bio) EXPORT_SYMBOL(blk_rq_unmap_user); /** - * blk_rq_map_kern - map kernel data to a request, for REQ_BLOCK_PC usage + * blk_rq_map_kern - map kernel data to a request, for REQ_TYPE_BLOCK_PC usage * @q: request queue where request should be inserted * @rq: request to fill * @kbuf: the kernel buffer diff --git a/block/blk-settings.c b/block/blk-settings.c index 539d873c820..d70692badcd 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -144,7 +144,7 @@ EXPORT_SYMBOL(blk_queue_make_request); * Different hardware can have different requirements as to what pages * it can do I/O directly to. A low level driver can call * blk_queue_bounce_limit to have lower memory pages allocated as bounce - * buffers for doing I/O to pages residing above @page. + * buffers for doing I/O to pages residing above @dma_addr. **/ void blk_queue_bounce_limit(struct request_queue *q, u64 dma_addr) { @@ -229,7 +229,7 @@ EXPORT_SYMBOL(blk_queue_max_phys_segments); * Description: * Enables a low level driver to set an upper limit on the number of * hw data segments in a request. This would be the largest number of - * address/length pairs the host adapter can actually give as once + * address/length pairs the host adapter can actually give at once * to the device. **/ void blk_queue_max_hw_segments(struct request_queue *q, @@ -410,7 +410,7 @@ EXPORT_SYMBOL(blk_queue_segment_boundary); * @mask: alignment mask * * description: - * set required memory and length aligment for direct dma transactions. + * set required memory and length alignment for direct dma transactions. * this is used when buiding direct io requests for the queue. * **/ @@ -426,7 +426,7 @@ EXPORT_SYMBOL(blk_queue_dma_alignment); * @mask: alignment mask * * description: - * update required memory and length aligment for direct dma transactions. + * update required memory and length alignment for direct dma transactions. * If the requested alignment is larger than the current alignment, then * the current queue alignment is updated to the new value, otherwise it * is left alone. The design of this is to allow multiple objects diff --git a/block/blk-tag.c b/block/blk-tag.c index ed5166fbc59..8a99688eb1b 100644 --- a/block/blk-tag.c +++ b/block/blk-tag.c @@ -29,7 +29,7 @@ EXPORT_SYMBOL(blk_queue_find_tag); * __blk_free_tags - release a given set of tag maintenance info * @bqt: the tag map to free * - * Tries to free the specified @bqt@. Returns true if it was + * Tries to free the specified @bqt. Returns true if it was * actually freed and false if there are still references using it */ static int __blk_free_tags(struct blk_queue_tag *bqt) @@ -78,7 +78,7 @@ void __blk_queue_free_tags(struct request_queue *q) * blk_free_tags - release a given set of tag maintenance info * @bqt: the tag map to free * - * For externally managed @bqt@ frees the map. Callers of this + * For externally managed @bqt frees the map. Callers of this * function must guarantee to have released all the queues that * might have been using this tag map. */ @@ -94,7 +94,7 @@ EXPORT_SYMBOL(blk_free_tags); * @q: the request queue for the device * * Notes: - * This is used to disabled tagged queuing to a device, yet leave + * This is used to disable tagged queuing to a device, yet leave * queue in function. **/ void blk_queue_free_tags(struct request_queue *q) @@ -271,7 +271,7 @@ EXPORT_SYMBOL(blk_queue_resize_tags); * @rq: the request that has completed * * Description: - * Typically called when end_that_request_first() returns 0, meaning + * Typically called when end_that_request_first() returns %0, meaning * all transfers have been done for a request. It's important to call * this function before end_that_request_last(), as that will put the * request back on the free list thus corrupting the internal tag list. diff --git a/block/genhd.c b/block/genhd.c index e0ce23ac2ec..c114a43052d 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -211,10 +211,11 @@ void unlink_gendisk(struct gendisk *disk) /** * get_gendisk - get partitioning information for a given device - * @dev: device to get partitioning information for + * @devt: device to get partitioning information for + * @part: returned partition index * * This function gets the structure containing partitioning - * information for the given device @dev. + * information for the given device @devt. */ struct gendisk *get_gendisk(dev_t devt, int *part) { -- cgit v1.2.3 From ac65ece4eee10b03ac29ee925cadc179dc810bab Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:30:12 +0900 Subject: block: fix partition info printouts Recent block_class iteration updates 5c6f35c5..27f3025 broke partition info printouts. * printk_all_partitions(): Partition print out stops when it meets a partition hole. Partition printing inner loop should continue instead of exiting on empty partition slot. * /proc/partitions and /proc/diskstats: If all information can't be read in single read(), the information is truncated. This is because find_start() doesn't actually update the counter containing the initial seek. It runs to the end and ends up always reporting EOF on the second read. This patch fixes both problems. Signed-off-by: Tejun Heo Cc: Greg Kroah-Hartman Signed-off-by: Jens Axboe --- block/genhd.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index c114a43052d..0be95135c40 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -236,7 +236,7 @@ static int printk_partition(struct device *dev, void *data) int n; if (dev->type != &disk_type) - goto exit; + return 0; sgp = dev_to_disk(dev); /* @@ -244,7 +244,7 @@ static int printk_partition(struct device *dev, void *data) */ if (get_capacity(sgp) == 0 || (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) - goto exit; + return 0; /* * Note, unlike /proc/partitions, I am showing the numbers in @@ -264,15 +264,15 @@ static int printk_partition(struct device *dev, void *data) /* now show the partitions */ for (n = 0; n < sgp->minors - 1; ++n) { if (sgp->part[n] == NULL) - goto exit; + continue; if (sgp->part[n]->nr_sects == 0) - goto exit; + continue; printk(" %02x%02x %10llu %s\n", sgp->major, n + 1 + sgp->first_minor, (unsigned long long)sgp->part[n]->nr_sects >> 1, disk_name(sgp, n + 1, buf)); } -exit: + return 0; } -- cgit v1.2.3 From 2ac3cee5298a247b2774f3319b28a05f588c3f0e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 08:53:37 +0200 Subject: block: don't grab block_class_lock unnecessarily block_class_lock protects major_names array and bdev_map and doesn't have anything to do with block class devices. Don't grab them while iterating over block class devices. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 28 +++++++++------------------- 1 file changed, 9 insertions(+), 19 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 0be95135c40..9eb8b3e212c 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -283,9 +283,7 @@ static int printk_partition(struct device *dev, void *data) */ void __init printk_all_partitions(void) { - mutex_lock(&block_class_lock); class_for_each_device(&block_class, NULL, NULL, printk_partition); - mutex_unlock(&block_class_lock); } #ifdef CONFIG_PROC_FS @@ -305,17 +303,15 @@ static int find_start(struct device *dev, void *data) static void *part_start(struct seq_file *part, loff_t *pos) { struct device *dev; - loff_t k = *pos; + loff_t n = *pos; - if (!k) + if (!n) part->private = (void *)1LU; /* tell show to print header */ - mutex_lock(&block_class_lock); - dev = class_find_device(&block_class, NULL, &k, find_start); - if (dev) { - put_device(dev); + dev = class_find_device(&block_class, NULL, &n, find_start); + if (dev) return dev_to_disk(dev); - } + return NULL; } @@ -341,7 +337,6 @@ static void *part_next(struct seq_file *part, void *v, loff_t *pos) static void part_stop(struct seq_file *part, void *v) { - mutex_unlock(&block_class_lock); } static int show_partition(struct seq_file *part, void *v) @@ -583,14 +578,12 @@ static struct device_type disk_type = { static void *diskstats_start(struct seq_file *part, loff_t *pos) { struct device *dev; - loff_t k = *pos; + loff_t n = *pos; - mutex_lock(&block_class_lock); - dev = class_find_device(&block_class, NULL, &k, find_start); - if (dev) { - put_device(dev); + dev = class_find_device(&block_class, NULL, &n, find_start); + if (dev) return dev_to_disk(dev); - } + return NULL; } @@ -610,7 +603,6 @@ static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos) static void diskstats_stop(struct seq_file *part, void *v) { - mutex_unlock(&block_class_lock); } static int diskstats_show(struct seq_file *s, void *v) @@ -729,7 +721,6 @@ dev_t blk_lookup_devt(const char *name, int part) dev_t devt = MKDEV(0, 0); struct find_block find; - mutex_lock(&block_class_lock); find.name = name; find.part = part; dev = class_find_device(&block_class, NULL, &find, match_id); @@ -738,7 +729,6 @@ dev_t blk_lookup_devt(const char *name, int part) devt = MKDEV(MAJOR(dev->devt), MINOR(dev->devt) + part); } - mutex_unlock(&block_class_lock); return devt; } -- cgit v1.2.3 From def4e38ddda9bef20b69bfa939195c2f79da7979 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 08:57:12 +0200 Subject: block: use class_dev_iterator instead of class_for_each_device() Recent block_class iteration updates 5c6f35c5..27f3025 converted all class device iteration to class_for_each_device() and class_find_device(), which are correct but pain in the ass to use. This pach converts them to newly introduced class_dev_iterator so that they can use more natural control structures instead of separate callbacks and struct to pass parameters to them. This results in smaller and easier code. This patch also restores the original behavior of not printing header in /proc/partitions if there's no partition to print. This is trivial but still user-visible behavior. Signed-off-by: Tejun Heo Cc: Greg Kroah-Hartman Signed-off-by: Jens Axboe --- block/genhd.c | 252 ++++++++++++++++++++++------------------------------------ 1 file changed, 97 insertions(+), 155 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 9eb8b3e212c..8b9a9ff1a84 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -225,57 +225,6 @@ struct gendisk *get_gendisk(dev_t devt, int *part) return kobj ? dev_to_disk(dev) : NULL; } -/* - * print a partitions - intended for places where the root filesystem can't be - * mounted and thus to give the victim some idea of what went wrong - */ -static int printk_partition(struct device *dev, void *data) -{ - struct gendisk *sgp; - char buf[BDEVNAME_SIZE]; - int n; - - if (dev->type != &disk_type) - return 0; - - sgp = dev_to_disk(dev); - /* - * Don't show empty devices or things that have been surpressed - */ - if (get_capacity(sgp) == 0 || - (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) - return 0; - - /* - * Note, unlike /proc/partitions, I am showing the numbers in - * hex - the same format as the root= option takes. - */ - printk("%02x%02x %10llu %s", - sgp->major, sgp->first_minor, - (unsigned long long)get_capacity(sgp) >> 1, - disk_name(sgp, 0, buf)); - if (sgp->driverfs_dev != NULL && - sgp->driverfs_dev->driver != NULL) - printk(" driver: %s\n", - sgp->driverfs_dev->driver->name); - else - printk(" (driver?)\n"); - - /* now show the partitions */ - for (n = 0; n < sgp->minors - 1; ++n) { - if (sgp->part[n] == NULL) - continue; - if (sgp->part[n]->nr_sects == 0) - continue; - printk(" %02x%02x %10llu %s\n", - sgp->major, n + 1 + sgp->first_minor, - (unsigned long long)sgp->part[n]->nr_sects >> 1, - disk_name(sgp, n + 1, buf)); - } - - return 0; -} - /* * print a full list of all partitions - intended for places where the root * filesystem can't be mounted and thus to give the victim some idea of what @@ -283,60 +232,108 @@ static int printk_partition(struct device *dev, void *data) */ void __init printk_all_partitions(void) { - class_for_each_device(&block_class, NULL, NULL, printk_partition); + struct class_dev_iter iter; + struct device *dev; + + class_dev_iter_init(&iter, &block_class, NULL, &disk_type); + while ((dev = class_dev_iter_next(&iter))) { + struct gendisk *disk = dev_to_disk(dev); + char buf[BDEVNAME_SIZE]; + int n; + + /* + * Don't show empty devices or things that have been + * surpressed + */ + if (get_capacity(disk) == 0 || + (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) + continue; + + /* + * Note, unlike /proc/partitions, I am showing the + * numbers in hex - the same format as the root= + * option takes. + */ + printk("%02x%02x %10llu %s", + disk->major, disk->first_minor, + (unsigned long long)get_capacity(disk) >> 1, + disk_name(disk, 0, buf)); + if (disk->driverfs_dev != NULL && + disk->driverfs_dev->driver != NULL) + printk(" driver: %s\n", + disk->driverfs_dev->driver->name); + else + printk(" (driver?)\n"); + + /* now show the partitions */ + for (n = 0; n < disk->minors - 1; ++n) { + if (disk->part[n] == NULL) + continue; + if (disk->part[n]->nr_sects == 0) + continue; + printk(" %02x%02x %10llu %s\n", + disk->major, n + 1 + disk->first_minor, + (unsigned long long)disk->part[n]->nr_sects >> 1, + disk_name(disk, n + 1, buf)); + } + } + class_dev_iter_exit(&iter); } #ifdef CONFIG_PROC_FS /* iterator */ -static int find_start(struct device *dev, void *data) +static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) { - loff_t *k = data; + loff_t skip = *pos; + struct class_dev_iter *iter; + struct device *dev; - if (dev->type != &disk_type) - return 0; - if (!*k) - return 1; - (*k)--; - return 0; + iter = kmalloc(GFP_KERNEL, sizeof(*iter)); + if (!iter) + return ERR_PTR(-ENOMEM); + + seqf->private = iter; + class_dev_iter_init(iter, &block_class, NULL, &disk_type); + do { + dev = class_dev_iter_next(iter); + if (!dev) + return NULL; + } while (skip--); + + return dev_to_disk(dev); } -static void *part_start(struct seq_file *part, loff_t *pos) +static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) { struct device *dev; - loff_t n = *pos; - - if (!n) - part->private = (void *)1LU; /* tell show to print header */ - dev = class_find_device(&block_class, NULL, &n, find_start); + (*pos)++; + dev = class_dev_iter_next(seqf->private); if (dev) return dev_to_disk(dev); return NULL; } -static int find_next(struct device *dev, void *data) +static void disk_seqf_stop(struct seq_file *seqf, void *v) { - if (dev->type == &disk_type) - return 1; - return 0; -} + struct class_dev_iter *iter = seqf->private; -static void *part_next(struct seq_file *part, void *v, loff_t *pos) -{ - struct gendisk *gp = v; - struct device *dev; - ++*pos; - dev = class_find_device(&block_class, &gp->dev, NULL, find_next); - if (dev) { - put_device(dev); - return dev_to_disk(dev); + /* stop is called even after start failed :-( */ + if (iter) { + class_dev_iter_exit(iter); + kfree(iter); } - return NULL; } -static void part_stop(struct seq_file *part, void *v) +static void *show_partition_start(struct seq_file *seqf, loff_t *pos) { + static void *p; + + p = disk_seqf_start(seqf, pos); + if (!IS_ERR(p) && p) + seq_puts(seqf, "major minor #blocks name\n\n"); + return p; } static int show_partition(struct seq_file *part, void *v) @@ -383,9 +380,9 @@ static int show_partition(struct seq_file *part, void *v) } const struct seq_operations partitions_op = { - .start = part_start, - .next = part_next, - .stop = part_stop, + .start = show_partition_start, + .next = disk_seqf_next, + .stop = disk_seqf_stop, .show = show_partition }; #endif @@ -567,44 +564,6 @@ static struct device_type disk_type = { }; #ifdef CONFIG_PROC_FS -/* - * aggregate disk stat collector. Uses the same stats that the sysfs - * entries do, above, but makes them available through one seq_file. - * - * The output looks suspiciously like /proc/partitions with a bunch of - * extra fields. - */ - -static void *diskstats_start(struct seq_file *part, loff_t *pos) -{ - struct device *dev; - loff_t n = *pos; - - dev = class_find_device(&block_class, NULL, &n, find_start); - if (dev) - return dev_to_disk(dev); - - return NULL; -} - -static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos) -{ - struct gendisk *gp = v; - struct device *dev; - - ++*pos; - dev = class_find_device(&block_class, &gp->dev, NULL, find_next); - if (dev) { - put_device(dev); - return dev_to_disk(dev); - } - return NULL; -} - -static void diskstats_stop(struct seq_file *part, void *v) -{ -} - static int diskstats_show(struct seq_file *s, void *v) { struct gendisk *gp = v; @@ -666,9 +625,9 @@ static int diskstats_show(struct seq_file *s, void *v) } const struct seq_operations diskstats_op = { - .start = diskstats_start, - .next = diskstats_next, - .stop = diskstats_stop, + .start = disk_seqf_start, + .next = disk_seqf_next, + .stop = disk_seqf_stop, .show = diskstats_show }; #endif /* CONFIG_PROC_FS */ @@ -696,40 +655,23 @@ void genhd_media_change_notify(struct gendisk *disk) EXPORT_SYMBOL_GPL(genhd_media_change_notify); #endif /* 0 */ -struct find_block { - const char *name; - int part; -}; - -static int match_id(struct device *dev, void *data) +dev_t blk_lookup_devt(const char *name, int part) { - struct find_block *find = data; + dev_t devt = MKDEV(0, 0); + struct class_dev_iter iter; + struct device *dev; - if (dev->type != &disk_type) - return 0; - if (strcmp(dev->bus_id, find->name) == 0) { + class_dev_iter_init(&iter, &block_class, NULL, &disk_type); + while ((dev = class_dev_iter_next(&iter))) { struct gendisk *disk = dev_to_disk(dev); - if (find->part < disk->minors) - return 1; - } - return 0; -} -dev_t blk_lookup_devt(const char *name, int part) -{ - struct device *dev; - dev_t devt = MKDEV(0, 0); - struct find_block find; - - find.name = name; - find.part = part; - dev = class_find_device(&block_class, NULL, &find, match_id); - if (dev) { - put_device(dev); - devt = MKDEV(MAJOR(dev->devt), - MINOR(dev->devt) + part); + if (!strcmp(dev->bus_id, name) && part < disk->minors) { + devt = MKDEV(MAJOR(dev->devt), + MINOR(dev->devt) + part); + break; + } } - + class_dev_iter_exit(&iter); return devt; } EXPORT_SYMBOL(blk_lookup_devt); -- cgit v1.2.3 From ec2cdedf798385a9397ac50dd0405dd658f8529c Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:30:15 +0900 Subject: block: allow deleting zero length partition delete_partition() was noop for zero length partition. As the addition code allows creating zero lenght partition and deletion is assumed to always succeed, this causes memory leak for zero length partitions. Allow zero length partitions to end their meaningless lives. While at it, allow deleting zero lenght partition via BLKPG_DEL_PARTITION ioctl too. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/ioctl.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'block') diff --git a/block/ioctl.c b/block/ioctl.c index 375c57922b0..c722de0ef2e 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -68,8 +68,6 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user case BLKPG_DEL_PARTITION: if (!disk->part[part-1]) return -ENXIO; - if (disk->part[part - 1]->nr_sects == 0) - return -ENXIO; bdevp = bdget_disk(disk, part); if (!bdevp) return -ENOMEM; -- cgit v1.2.3 From 88e341261ca4d39eec21b212961c77eff51105f7 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:30:16 +0900 Subject: block: update add_partition() error handling d805dda4 tried to fix error case handling in add_partition() but had a few problems. * disk->part[] entry is set early and left dangling if operation fails. * Once device initialized, the last put_device() is responsible for freeing all the resources. The failure path freed part_stats and p regardless of put_device() causing double free. * holders subdir holds reference to the disk device, so failure path should remove it to release resources properly which was missing. This patch fixes the above problems and while at it move partition slot busy check into add_partition() for completeness and inlines holders subdirectory creation. Using separate function for it just obfuscates the code. Signed-off-by: Tejun Heo Cc: Abdel Benamrouche Signed-off-by: Jens Axboe --- block/ioctl.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/ioctl.c b/block/ioctl.c index c722de0ef2e..eb046aeede8 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -43,12 +43,9 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user || pstart < 0 || plength < 0) return -EINVAL; } - /* partition number in use? */ + mutex_lock(&bdev->bd_mutex); - if (disk->part[part - 1]) { - mutex_unlock(&bdev->bd_mutex); - return -EBUSY; - } + /* overlap? */ for (i = 0; i < disk->minors - 1; i++) { struct hd_struct *s = disk->part[i]; -- cgit v1.2.3 From 310a2c1012934f590192377f65940cad4aa72b15 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:47:17 +0900 Subject: block: misc updates This patch makes the following misc updates in preparation for disk->part dereference fix and extended block devt support. * implment part_to_disk() * fix comment about gendisk->part indexing * rename get_part() to disk_map_sector() * don't use n which is always zero while printing disk information in diskstats_show() Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 7 ++++--- block/blk-merge.c | 4 ++-- block/genhd.c | 4 ++-- 3 files changed, 8 insertions(+), 7 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 86d22e7d65c..a0dc2e72fcb 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -60,7 +60,7 @@ static void drive_stat_acct(struct request *rq, int new_io) if (!blk_fs_request(rq) || !rq->rq_disk) return; - part = get_part(rq->rq_disk, rq->sector); + part = disk_map_sector(rq->rq_disk, rq->sector); if (!new_io) __all_stat_inc(rq->rq_disk, part, merges[rw], rq->sector); else { @@ -1557,7 +1557,8 @@ static int __end_that_request_first(struct request *req, int error, } if (blk_fs_request(req) && req->rq_disk) { - struct hd_struct *part = get_part(req->rq_disk, req->sector); + struct hd_struct *part = + disk_map_sector(req->rq_disk, req->sector); const int rw = rq_data_dir(req); all_stat_add(req->rq_disk, part, sectors[rw], @@ -1745,7 +1746,7 @@ static void end_that_request_last(struct request *req, int error) if (disk && blk_fs_request(req) && req != &req->q->bar_rq) { unsigned long duration = jiffies - req->start_time; const int rw = rq_data_dir(req); - struct hd_struct *part = get_part(disk, req->sector); + struct hd_struct *part = disk_map_sector(disk, req->sector); __all_stat_inc(disk, part, ios[rw], req->sector); __all_stat_add(disk, part, ticks[rw], duration, req->sector); diff --git a/block/blk-merge.c b/block/blk-merge.c index d81d91419ff..9b17da698d7 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -387,8 +387,8 @@ static int attempt_merge(struct request_queue *q, struct request *req, elv_merge_requests(q, req, next); if (req->rq_disk) { - struct hd_struct *part - = get_part(req->rq_disk, req->sector); + struct hd_struct *part = + disk_map_sector(req->rq_disk, req->sector); disk_round_stats(req->rq_disk); req->rq_disk->in_flight--; if (part) { diff --git a/block/genhd.c b/block/genhd.c index 8b9a9ff1a84..11038fbc75e 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -568,7 +568,7 @@ static int diskstats_show(struct seq_file *s, void *v) { struct gendisk *gp = v; char buf[BDEVNAME_SIZE]; - int n = 0; + int n; /* if (&gp->dev.kobj.entry == block_class.devices.next) @@ -582,7 +582,7 @@ static int diskstats_show(struct seq_file *s, void *v) disk_round_stats(gp); preempt_enable(); seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", - gp->major, n + gp->first_minor, disk_name(gp, n, buf), + gp->major, gp->first_minor, disk_name(gp, 0, buf), disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]), (unsigned long long)disk_stat_read(gp, sectors[0]), jiffies_to_msecs(disk_stat_read(gp, ticks[0])), -- cgit v1.2.3 From cf771cb5a7b716f3f9e532fd42a1e3a0a75adec5 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 09:01:09 +0200 Subject: block: make variable and argument names more consistent In hd_struct, @partno is used to denote partition number and a number of other places use @part to denote hd_struct. Functions use @part and @index instead. This causes confusion and makes it difficult to use consistent variable names for hd_struct. Always use @partno if a variable represents partition number. Also, print out functions use @f or @part for seq_file argument. Use @seqf uniformly instead. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 54 +++++++++++++++++++++++++----------------------------- block/ioctl.c | 15 ++++++++------- 2 files changed, 33 insertions(+), 36 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 11038fbc75e..dc9ad4c171e 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -43,14 +43,14 @@ static inline int major_to_index(int major) } #ifdef CONFIG_PROC_FS -void blkdev_show(struct seq_file *f, off_t offset) +void blkdev_show(struct seq_file *seqf, off_t offset) { struct blk_major_name *dp; if (offset < BLKDEV_MAJOR_HASH_SIZE) { mutex_lock(&block_class_lock); for (dp = major_names[offset]; dp; dp = dp->next) - seq_printf(f, "%3d %s\n", dp->major, dp->name); + seq_printf(seqf, "%3d %s\n", dp->major, dp->name); mutex_unlock(&block_class_lock); } } @@ -157,7 +157,7 @@ void blk_unregister_region(dev_t devt, unsigned long range) EXPORT_SYMBOL(blk_unregister_region); -static struct kobject *exact_match(dev_t devt, int *part, void *data) +static struct kobject *exact_match(dev_t devt, int *partno, void *data) { struct gendisk *p = data; @@ -217,9 +217,9 @@ void unlink_gendisk(struct gendisk *disk) * This function gets the structure containing partitioning * information for the given device @devt. */ -struct gendisk *get_gendisk(dev_t devt, int *part) +struct gendisk *get_gendisk(dev_t devt, int *partno) { - struct kobject *kobj = kobj_lookup(bdev_map, devt, part); + struct kobject *kobj = kobj_lookup(bdev_map, devt, partno); struct device *dev = kobj_to_dev(kobj); return kobj ? dev_to_disk(dev) : NULL; @@ -336,23 +336,12 @@ static void *show_partition_start(struct seq_file *seqf, loff_t *pos) return p; } -static int show_partition(struct seq_file *part, void *v) +static int show_partition(struct seq_file *seqf, void *v) { struct gendisk *sgp = v; int n; char buf[BDEVNAME_SIZE]; - /* - * Print header if start told us to do. This is to preserve - * the original behavior of not printing header if no - * partition exists. This hackery will be removed later with - * class iteration clean up. - */ - if (part->private) { - seq_puts(part, "major minor #blocks name\n\n"); - part->private = NULL; - } - /* Don't show non-partitionable removeable devices or empty devices */ if (!get_capacity(sgp) || (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE))) @@ -361,7 +350,7 @@ static int show_partition(struct seq_file *part, void *v) return 0; /* show the full disk and all non-0 size partitions of it */ - seq_printf(part, "%4d %4d %10llu %s\n", + seq_printf(seqf, "%4d %4d %10llu %s\n", sgp->major, sgp->first_minor, (unsigned long long)get_capacity(sgp) >> 1, disk_name(sgp, 0, buf)); @@ -370,7 +359,7 @@ static int show_partition(struct seq_file *part, void *v) continue; if (sgp->part[n]->nr_sects == 0) continue; - seq_printf(part, "%4d %4d %10llu %s\n", + seq_printf(seqf, "%4d %4d %10llu %s\n", sgp->major, n + 1 + sgp->first_minor, (unsigned long long)sgp->part[n]->nr_sects >> 1 , disk_name(sgp, n + 1, buf)); @@ -388,7 +377,7 @@ const struct seq_operations partitions_op = { #endif -static struct kobject *base_probe(dev_t devt, int *part, void *data) +static struct kobject *base_probe(dev_t devt, int *partno, void *data) { if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) /* Make old-style 2.4 aliases work */ @@ -564,7 +553,14 @@ static struct device_type disk_type = { }; #ifdef CONFIG_PROC_FS -static int diskstats_show(struct seq_file *s, void *v) +/* + * aggregate disk stat collector. Uses the same stats that the sysfs + * entries do, above, but makes them available through one seq_file. + * + * The output looks suspiciously like /proc/partitions with a bunch of + * extra fields. + */ +static int diskstats_show(struct seq_file *seqf, void *v) { struct gendisk *gp = v; char buf[BDEVNAME_SIZE]; @@ -572,7 +568,7 @@ static int diskstats_show(struct seq_file *s, void *v) /* if (&gp->dev.kobj.entry == block_class.devices.next) - seq_puts(s, "major minor name" + seq_puts(seqf, "major minor name" " rio rmerge rsect ruse wio wmerge " "wsect wuse running use aveq" "\n\n"); @@ -581,7 +577,7 @@ static int diskstats_show(struct seq_file *s, void *v) preempt_disable(); disk_round_stats(gp); preempt_enable(); - seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", + seq_printf(seqf, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", gp->major, gp->first_minor, disk_name(gp, 0, buf), disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]), (unsigned long long)disk_stat_read(gp, sectors[0]), @@ -603,7 +599,7 @@ static int diskstats_show(struct seq_file *s, void *v) preempt_disable(); part_round_stats(hd); preempt_enable(); - seq_printf(s, "%4d %4d %s %lu %lu %llu " + seq_printf(seqf, "%4d %4d %s %lu %lu %llu " "%u %lu %lu %llu %u %u %u %u\n", gp->major, n + gp->first_minor + 1, disk_name(gp, n + 1, buf), @@ -655,7 +651,7 @@ void genhd_media_change_notify(struct gendisk *disk) EXPORT_SYMBOL_GPL(genhd_media_change_notify); #endif /* 0 */ -dev_t blk_lookup_devt(const char *name, int part) +dev_t blk_lookup_devt(const char *name, int partno) { dev_t devt = MKDEV(0, 0); struct class_dev_iter iter; @@ -665,9 +661,9 @@ dev_t blk_lookup_devt(const char *name, int part) while ((dev = class_dev_iter_next(&iter))) { struct gendisk *disk = dev_to_disk(dev); - if (!strcmp(dev->bus_id, name) && part < disk->minors) { + if (!strcmp(dev->bus_id, name) && partno < disk->minors) { devt = MKDEV(MAJOR(dev->devt), - MINOR(dev->devt) + part); + MINOR(dev->devt) + partno); break; } } @@ -777,10 +773,10 @@ int bdev_read_only(struct block_device *bdev) EXPORT_SYMBOL(bdev_read_only); -int invalidate_partition(struct gendisk *disk, int index) +int invalidate_partition(struct gendisk *disk, int partno) { int res = 0; - struct block_device *bdev = bdget_disk(disk, index); + struct block_device *bdev = bdget_disk(disk, partno); if (bdev) { fsync_bdev(bdev); res = __invalidate_device(bdev); diff --git a/block/ioctl.c b/block/ioctl.c index eb046aeede8..d77f5e280a6 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -15,7 +15,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user struct blkpg_ioctl_arg a; struct blkpg_partition p; long long start, length; - int part; + int partno; int i; int err; @@ -28,8 +28,8 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user disk = bdev->bd_disk; if (bdev != bdev->bd_contains) return -EINVAL; - part = p.pno; - if (part <= 0 || part >= disk->minors) + partno = p.pno; + if (partno <= 0 || partno >= disk->minors) return -EINVAL; switch (a.op) { case BLKPG_ADD_PARTITION: @@ -59,13 +59,14 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user } } /* all seems OK */ - err = add_partition(disk, part, start, length, ADDPART_FLAG_NONE); + err = add_partition(disk, partno, start, length, + ADDPART_FLAG_NONE); mutex_unlock(&bdev->bd_mutex); return err; case BLKPG_DEL_PARTITION: - if (!disk->part[part-1]) + if (!disk->part[partno - 1]) return -ENXIO; - bdevp = bdget_disk(disk, part); + bdevp = bdget_disk(disk, partno); if (!bdevp) return -ENOMEM; mutex_lock(&bdevp->bd_mutex); @@ -79,7 +80,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user invalidate_bdev(bdevp); mutex_lock_nested(&bdev->bd_mutex, 1); - delete_partition(disk, part); + delete_partition(disk, partno); mutex_unlock(&bdev->bd_mutex); mutex_unlock(&bdevp->bd_mutex); bdput(bdevp); -- cgit v1.2.3 From f331c0296f2a9fee0d396a70598b954062603015 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 09:01:48 +0200 Subject: block: don't depend on consecutive minor space * Implement disk_devt() and part_devt() and use them to directly access devt instead of computing it from ->major and ->first_minor. Note that all references to ->major and ->first_minor outside of block layer is used to determine devt of the disk (the part0) and as ->major and ->first_minor will continue to represent devt for the disk, converting these users aren't strictly necessary. However, convert them for consistency. * Implement disk_max_parts() to avoid directly deferencing genhd->minors. * Update bdget_disk() such that it doesn't assume consecutive minor space. * Move devt computation from register_disk() to add_disk() and make it the only one (all other usages use the initially determined value). These changes clean up the code and will help disk->part dereference fix and extended block device numbers. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 107 +++++++++++++++++++++++++++++++++++++++++----------------- block/ioctl.c | 6 ++-- 2 files changed, 79 insertions(+), 34 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index dc9ad4c171e..fa32d09fda2 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -186,13 +186,14 @@ void add_disk(struct gendisk *disk) int retval; disk->flags |= GENHD_FL_UP; - blk_register_region(MKDEV(disk->major, disk->first_minor), - disk->minors, NULL, exact_match, exact_lock, disk); + disk->dev.devt = MKDEV(disk->major, disk->first_minor); + blk_register_region(disk_devt(disk), disk->minors, NULL, + exact_match, exact_lock, disk); register_disk(disk); blk_register_queue(disk); bdi = &disk->queue->backing_dev_info; - bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor)); + bdi_register_dev(bdi, disk_devt(disk)); retval = sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi"); WARN_ON(retval); } @@ -205,8 +206,7 @@ void unlink_gendisk(struct gendisk *disk) sysfs_remove_link(&disk->dev.kobj, "bdi"); bdi_unregister(&disk->queue->backing_dev_info); blk_unregister_queue(disk); - blk_unregister_region(MKDEV(disk->major, disk->first_minor), - disk->minors); + blk_unregister_region(disk_devt(disk), disk->minors); } /** @@ -225,6 +225,38 @@ struct gendisk *get_gendisk(dev_t devt, int *partno) return kobj ? dev_to_disk(dev) : NULL; } +/** + * bdget_disk - do bdget() by gendisk and partition number + * @disk: gendisk of interest + * @partno: partition number + * + * Find partition @partno from @disk, do bdget() on it. + * + * CONTEXT: + * Don't care. + * + * RETURNS: + * Resulting block_device on success, NULL on failure. + */ +extern struct block_device *bdget_disk(struct gendisk *disk, int partno) +{ + dev_t devt = MKDEV(0, 0); + + if (partno == 0) + devt = disk_devt(disk); + else { + struct hd_struct *part = disk->part[partno - 1]; + + if (part && part->nr_sects) + devt = part_devt(part); + } + + if (likely(devt != MKDEV(0, 0))) + return bdget(devt); + return NULL; +} +EXPORT_SYMBOL(bdget_disk); + /* * print a full list of all partitions - intended for places where the root * filesystem can't be mounted and thus to give the victim some idea of what @@ -255,7 +287,7 @@ void __init printk_all_partitions(void) * option takes. */ printk("%02x%02x %10llu %s", - disk->major, disk->first_minor, + MAJOR(disk_devt(disk)), MINOR(disk_devt(disk)), (unsigned long long)get_capacity(disk) >> 1, disk_name(disk, 0, buf)); if (disk->driverfs_dev != NULL && @@ -266,15 +298,15 @@ void __init printk_all_partitions(void) printk(" (driver?)\n"); /* now show the partitions */ - for (n = 0; n < disk->minors - 1; ++n) { - if (disk->part[n] == NULL) - continue; - if (disk->part[n]->nr_sects == 0) + for (n = 0; n < disk_max_parts(disk); ++n) { + struct hd_struct *part = disk->part[n]; + + if (!part || !part->nr_sects) continue; printk(" %02x%02x %10llu %s\n", - disk->major, n + 1 + disk->first_minor, - (unsigned long long)disk->part[n]->nr_sects >> 1, - disk_name(disk, n + 1, buf)); + MAJOR(part_devt(part)), MINOR(part_devt(part)), + (unsigned long long)part->nr_sects >> 1, + disk_name(disk, part->partno, buf)); } } class_dev_iter_exit(&iter); @@ -343,26 +375,27 @@ static int show_partition(struct seq_file *seqf, void *v) char buf[BDEVNAME_SIZE]; /* Don't show non-partitionable removeable devices or empty devices */ - if (!get_capacity(sgp) || - (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE))) + if (!get_capacity(sgp) || (!disk_max_parts(sgp) && + (sgp->flags & GENHD_FL_REMOVABLE))) return 0; if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) return 0; /* show the full disk and all non-0 size partitions of it */ seq_printf(seqf, "%4d %4d %10llu %s\n", - sgp->major, sgp->first_minor, + MAJOR(disk_devt(sgp)), MINOR(disk_devt(sgp)), (unsigned long long)get_capacity(sgp) >> 1, disk_name(sgp, 0, buf)); - for (n = 0; n < sgp->minors - 1; n++) { - if (!sgp->part[n]) + for (n = 0; n < disk_max_parts(sgp); n++) { + struct hd_struct *part = sgp->part[n]; + if (!part) continue; - if (sgp->part[n]->nr_sects == 0) + if (part->nr_sects == 0) continue; seq_printf(seqf, "%4d %4d %10llu %s\n", - sgp->major, n + 1 + sgp->first_minor, - (unsigned long long)sgp->part[n]->nr_sects >> 1 , - disk_name(sgp, n + 1, buf)); + MAJOR(part_devt(part)), MINOR(part_devt(part)), + (unsigned long long)part->nr_sects >> 1, + disk_name(sgp, part->partno, buf)); } return 0; @@ -578,7 +611,8 @@ static int diskstats_show(struct seq_file *seqf, void *v) disk_round_stats(gp); preempt_enable(); seq_printf(seqf, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", - gp->major, gp->first_minor, disk_name(gp, 0, buf), + MAJOR(disk_devt(gp)), MINOR(disk_devt(gp)), + disk_name(gp, 0, buf), disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]), (unsigned long long)disk_stat_read(gp, sectors[0]), jiffies_to_msecs(disk_stat_read(gp, ticks[0])), @@ -590,7 +624,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) jiffies_to_msecs(disk_stat_read(gp, time_in_queue))); /* now show all non-0 size partitions of it */ - for (n = 0; n < gp->minors - 1; n++) { + for (n = 0; n < disk_max_parts(gp); n++) { struct hd_struct *hd = gp->part[n]; if (!hd || !hd->nr_sects) @@ -601,8 +635,8 @@ static int diskstats_show(struct seq_file *seqf, void *v) preempt_enable(); seq_printf(seqf, "%4d %4d %s %lu %lu %llu " "%u %lu %lu %llu %u %u %u %u\n", - gp->major, n + gp->first_minor + 1, - disk_name(gp, n + 1, buf), + MAJOR(part_devt(hd)), MINOR(part_devt(hd)), + disk_name(gp, hd->partno, buf), part_stat_read(hd, ios[0]), part_stat_read(hd, merges[0]), (unsigned long long)part_stat_read(hd, sectors[0]), @@ -661,11 +695,22 @@ dev_t blk_lookup_devt(const char *name, int partno) while ((dev = class_dev_iter_next(&iter))) { struct gendisk *disk = dev_to_disk(dev); - if (!strcmp(dev->bus_id, name) && partno < disk->minors) { - devt = MKDEV(MAJOR(dev->devt), - MINOR(dev->devt) + partno); - break; + if (strcmp(dev->bus_id, name)) + continue; + if (partno < 0 || partno > disk_max_parts(disk)) + continue; + + if (partno == 0) + devt = disk_devt(disk); + else { + struct hd_struct *part = disk->part[partno - 1]; + + if (!part || !part->nr_sects) + continue; + + devt = part_devt(part); } + break; } class_dev_iter_exit(&iter); return devt; @@ -755,7 +800,7 @@ void set_disk_ro(struct gendisk *disk, int flag) { int i; disk->policy = flag; - for (i = 0; i < disk->minors - 1; i++) + for (i = 0; i < disk_max_parts(disk); i++) if (disk->part[i]) disk->part[i]->policy = flag; } diff --git a/block/ioctl.c b/block/ioctl.c index d77f5e280a6..403f7d7e0c2 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -29,7 +29,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user if (bdev != bdev->bd_contains) return -EINVAL; partno = p.pno; - if (partno <= 0 || partno >= disk->minors) + if (partno <= 0 || partno > disk_max_parts(disk)) return -EINVAL; switch (a.op) { case BLKPG_ADD_PARTITION: @@ -47,7 +47,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user mutex_lock(&bdev->bd_mutex); /* overlap? */ - for (i = 0; i < disk->minors - 1; i++) { + for (i = 0; i < disk_max_parts(disk); i++) { struct hd_struct *s = disk->part[i]; if (!s) @@ -96,7 +96,7 @@ static int blkdev_reread_part(struct block_device *bdev) struct gendisk *disk = bdev->bd_disk; int res; - if (disk->minors == 1 || bdev != bdev->bd_contains) + if (!disk_max_parts(disk) || bdev != bdev->bd_contains) return -EINVAL; if (!capable(CAP_SYS_ADMIN)) return -EACCES; -- cgit v1.2.3 From e71bf0d0ee89e51b92776391c5634938236977d5 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 09:03:02 +0200 Subject: block: fix disk->part[] dereferencing race disk->part[] is protected by its matching bdev's lock. However, non-critical accesses like collecting stats and printing out sysfs and proc information used to be performed without any locking. As partitions can come and go dynamically, partitions can go away underneath those non-critical accesses. As some of those accesses are writes, this theoretically can lead to silent corruption. This patch fixes the race by using RCU for the partition array and dev reference counter to hold partitions. * Rename disk->part[] to disk->__part[] to make sure no one outside genhd layer proper accesses it directly. * Use RCU for disk->__part[] dereferencing. * Implement disk_{get|put}_part() which can be used to get and put partitions from gendisk respectively. * Iterators are implemented to help iterate through all partitions safely. * Functions which require RCU readlock are marked with _rcu suffix. * Use disk_put_part() in __blkdev_put() instead of directly putting the contained kobject. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 20 ++++- block/blk-merge.c | 9 ++- block/genhd.c | 218 ++++++++++++++++++++++++++++++++++++++++++++++-------- block/ioctl.c | 26 ++++--- 4 files changed, 226 insertions(+), 47 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index a0dc2e72fcb..d6128d9ad60 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -60,7 +60,9 @@ static void drive_stat_acct(struct request *rq, int new_io) if (!blk_fs_request(rq) || !rq->rq_disk) return; - part = disk_map_sector(rq->rq_disk, rq->sector); + rcu_read_lock(); + + part = disk_map_sector_rcu(rq->rq_disk, rq->sector); if (!new_io) __all_stat_inc(rq->rq_disk, part, merges[rw], rq->sector); else { @@ -71,6 +73,8 @@ static void drive_stat_acct(struct request *rq, int new_io) part->in_flight++; } } + + rcu_read_unlock(); } void blk_queue_congestion_threshold(struct request_queue *q) @@ -1557,12 +1561,14 @@ static int __end_that_request_first(struct request *req, int error, } if (blk_fs_request(req) && req->rq_disk) { - struct hd_struct *part = - disk_map_sector(req->rq_disk, req->sector); const int rw = rq_data_dir(req); + struct hd_struct *part; + rcu_read_lock(); + part = disk_map_sector_rcu(req->rq_disk, req->sector); all_stat_add(req->rq_disk, part, sectors[rw], nr_bytes >> 9, req->sector); + rcu_read_unlock(); } total_bytes = bio_nbytes = 0; @@ -1746,7 +1752,11 @@ static void end_that_request_last(struct request *req, int error) if (disk && blk_fs_request(req) && req != &req->q->bar_rq) { unsigned long duration = jiffies - req->start_time; const int rw = rq_data_dir(req); - struct hd_struct *part = disk_map_sector(disk, req->sector); + struct hd_struct *part; + + rcu_read_lock(); + + part = disk_map_sector_rcu(disk, req->sector); __all_stat_inc(disk, part, ios[rw], req->sector); __all_stat_add(disk, part, ticks[rw], duration, req->sector); @@ -1756,6 +1766,8 @@ static void end_that_request_last(struct request *req, int error) part_round_stats(part); part->in_flight--; } + + rcu_read_unlock(); } if (req->end_io) diff --git a/block/blk-merge.c b/block/blk-merge.c index 9b17da698d7..eb2a3ca5830 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -387,14 +387,19 @@ static int attempt_merge(struct request_queue *q, struct request *req, elv_merge_requests(q, req, next); if (req->rq_disk) { - struct hd_struct *part = - disk_map_sector(req->rq_disk, req->sector); + struct hd_struct *part; + + rcu_read_lock(); + + part = disk_map_sector_rcu(req->rq_disk, req->sector); disk_round_stats(req->rq_disk); req->rq_disk->in_flight--; if (part) { part_round_stats(part); part->in_flight--; } + + rcu_read_unlock(); } req->ioprio = ioprio_best(req->ioprio, next->ioprio); diff --git a/block/genhd.c b/block/genhd.c index fa32d09fda2..b431d654394 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -26,6 +26,158 @@ struct kobject *block_depr; static struct device_type disk_type; +/** + * disk_get_part - get partition + * @disk: disk to look partition from + * @partno: partition number + * + * Look for partition @partno from @disk. If found, increment + * reference count and return it. + * + * CONTEXT: + * Don't care. + * + * RETURNS: + * Pointer to the found partition on success, NULL if not found. + */ +struct hd_struct *disk_get_part(struct gendisk *disk, int partno) +{ + struct hd_struct *part; + + if (unlikely(partno < 1 || partno > disk_max_parts(disk))) + return NULL; + rcu_read_lock(); + part = rcu_dereference(disk->__part[partno - 1]); + if (part) + get_device(&part->dev); + rcu_read_unlock(); + + return part; +} +EXPORT_SYMBOL_GPL(disk_get_part); + +/** + * disk_part_iter_init - initialize partition iterator + * @piter: iterator to initialize + * @disk: disk to iterate over + * @flags: DISK_PITER_* flags + * + * Initialize @piter so that it iterates over partitions of @disk. + * + * CONTEXT: + * Don't care. + */ +void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, + unsigned int flags) +{ + piter->disk = disk; + piter->part = NULL; + + if (flags & DISK_PITER_REVERSE) + piter->idx = disk_max_parts(piter->disk) - 1; + else + piter->idx = 0; + + piter->flags = flags; +} +EXPORT_SYMBOL_GPL(disk_part_iter_init); + +/** + * disk_part_iter_next - proceed iterator to the next partition and return it + * @piter: iterator of interest + * + * Proceed @piter to the next partition and return it. + * + * CONTEXT: + * Don't care. + */ +struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) +{ + int inc, end; + + /* put the last partition */ + disk_put_part(piter->part); + piter->part = NULL; + + rcu_read_lock(); + + /* determine iteration parameters */ + if (piter->flags & DISK_PITER_REVERSE) { + inc = -1; + end = -1; + } else { + inc = 1; + end = disk_max_parts(piter->disk); + } + + /* iterate to the next partition */ + for (; piter->idx != end; piter->idx += inc) { + struct hd_struct *part; + + part = rcu_dereference(piter->disk->__part[piter->idx]); + if (!part) + continue; + if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects) + continue; + + get_device(&part->dev); + piter->part = part; + piter->idx += inc; + break; + } + + rcu_read_unlock(); + + return piter->part; +} +EXPORT_SYMBOL_GPL(disk_part_iter_next); + +/** + * disk_part_iter_exit - finish up partition iteration + * @piter: iter of interest + * + * Called when iteration is over. Cleans up @piter. + * + * CONTEXT: + * Don't care. + */ +void disk_part_iter_exit(struct disk_part_iter *piter) +{ + disk_put_part(piter->part); + piter->part = NULL; +} +EXPORT_SYMBOL_GPL(disk_part_iter_exit); + +/** + * disk_map_sector_rcu - map sector to partition + * @disk: gendisk of interest + * @sector: sector to map + * + * Find out which partition @sector maps to on @disk. This is + * primarily used for stats accounting. + * + * CONTEXT: + * RCU read locked. The returned partition pointer is valid only + * while preemption is disabled. + * + * RETURNS: + * Found partition on success, NULL if there's no matching partition. + */ +struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) +{ + int i; + + for (i = 0; i < disk_max_parts(disk); i++) { + struct hd_struct *part = rcu_dereference(disk->__part[i]); + + if (part && part->start_sect <= sector && + sector < part->start_sect + part->nr_sects) + return part; + } + return NULL; +} +EXPORT_SYMBOL_GPL(disk_map_sector_rcu); + /* * Can be deleted altogether. Later. * @@ -245,10 +397,12 @@ extern struct block_device *bdget_disk(struct gendisk *disk, int partno) if (partno == 0) devt = disk_devt(disk); else { - struct hd_struct *part = disk->part[partno - 1]; + struct hd_struct *part; + part = disk_get_part(disk, partno); if (part && part->nr_sects) devt = part_devt(part); + disk_put_part(part); } if (likely(devt != MKDEV(0, 0))) @@ -270,8 +424,9 @@ void __init printk_all_partitions(void) class_dev_iter_init(&iter, &block_class, NULL, &disk_type); while ((dev = class_dev_iter_next(&iter))) { struct gendisk *disk = dev_to_disk(dev); + struct disk_part_iter piter; + struct hd_struct *part; char buf[BDEVNAME_SIZE]; - int n; /* * Don't show empty devices or things that have been @@ -298,16 +453,13 @@ void __init printk_all_partitions(void) printk(" (driver?)\n"); /* now show the partitions */ - for (n = 0; n < disk_max_parts(disk); ++n) { - struct hd_struct *part = disk->part[n]; - - if (!part || !part->nr_sects) - continue; + disk_part_iter_init(&piter, disk, 0); + while ((part = disk_part_iter_next(&piter))) printk(" %02x%02x %10llu %s\n", MAJOR(part_devt(part)), MINOR(part_devt(part)), (unsigned long long)part->nr_sects >> 1, disk_name(disk, part->partno, buf)); - } + disk_part_iter_exit(&piter); } class_dev_iter_exit(&iter); } @@ -371,7 +523,8 @@ static void *show_partition_start(struct seq_file *seqf, loff_t *pos) static int show_partition(struct seq_file *seqf, void *v) { struct gendisk *sgp = v; - int n; + struct disk_part_iter piter; + struct hd_struct *part; char buf[BDEVNAME_SIZE]; /* Don't show non-partitionable removeable devices or empty devices */ @@ -386,17 +539,14 @@ static int show_partition(struct seq_file *seqf, void *v) MAJOR(disk_devt(sgp)), MINOR(disk_devt(sgp)), (unsigned long long)get_capacity(sgp) >> 1, disk_name(sgp, 0, buf)); - for (n = 0; n < disk_max_parts(sgp); n++) { - struct hd_struct *part = sgp->part[n]; - if (!part) - continue; - if (part->nr_sects == 0) - continue; + + disk_part_iter_init(&piter, sgp, 0); + while ((part = disk_part_iter_next(&piter))) seq_printf(seqf, "%4d %4d %10llu %s\n", MAJOR(part_devt(part)), MINOR(part_devt(part)), (unsigned long long)part->nr_sects >> 1, disk_name(sgp, part->partno, buf)); - } + disk_part_iter_exit(&piter); return 0; } @@ -571,7 +721,7 @@ static void disk_release(struct device *dev) struct gendisk *disk = dev_to_disk(dev); kfree(disk->random); - kfree(disk->part); + kfree(disk->__part); free_disk_stats(disk); kfree(disk); } @@ -596,8 +746,9 @@ static struct device_type disk_type = { static int diskstats_show(struct seq_file *seqf, void *v) { struct gendisk *gp = v; + struct disk_part_iter piter; + struct hd_struct *hd; char buf[BDEVNAME_SIZE]; - int n; /* if (&gp->dev.kobj.entry == block_class.devices.next) @@ -624,12 +775,8 @@ static int diskstats_show(struct seq_file *seqf, void *v) jiffies_to_msecs(disk_stat_read(gp, time_in_queue))); /* now show all non-0 size partitions of it */ - for (n = 0; n < disk_max_parts(gp); n++) { - struct hd_struct *hd = gp->part[n]; - - if (!hd || !hd->nr_sects) - continue; - + disk_part_iter_init(&piter, gp, 0); + while ((hd = disk_part_iter_next(&piter))) { preempt_disable(); part_round_stats(hd); preempt_enable(); @@ -650,6 +797,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) jiffies_to_msecs(part_stat_read(hd, time_in_queue)) ); } + disk_part_iter_exit(&piter); return 0; } @@ -703,12 +851,16 @@ dev_t blk_lookup_devt(const char *name, int partno) if (partno == 0) devt = disk_devt(disk); else { - struct hd_struct *part = disk->part[partno - 1]; + struct hd_struct *part; - if (!part || !part->nr_sects) + part = disk_get_part(disk, partno); + if (!part || !part->nr_sects) { + disk_put_part(part); continue; + } devt = part_devt(part); + disk_put_part(part); } break; } @@ -735,9 +887,9 @@ struct gendisk *alloc_disk_node(int minors, int node_id) } if (minors > 1) { int size = (minors - 1) * sizeof(struct hd_struct *); - disk->part = kmalloc_node(size, + disk->__part = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, node_id); - if (!disk->part) { + if (!disk->__part) { free_disk_stats(disk); kfree(disk); return NULL; @@ -798,10 +950,14 @@ EXPORT_SYMBOL(set_device_ro); void set_disk_ro(struct gendisk *disk, int flag) { - int i; + struct disk_part_iter piter; + struct hd_struct *part; + disk->policy = flag; - for (i = 0; i < disk_max_parts(disk); i++) - if (disk->part[i]) disk->part[i]->policy = flag; + disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); + while ((part = disk_part_iter_next(&piter))) + part->policy = flag; + disk_part_iter_exit(&piter); } EXPORT_SYMBOL(set_disk_ro); diff --git a/block/ioctl.c b/block/ioctl.c index 403f7d7e0c2..a5f672ad55f 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -12,11 +12,12 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user { struct block_device *bdevp; struct gendisk *disk; + struct hd_struct *part; struct blkpg_ioctl_arg a; struct blkpg_partition p; + struct disk_part_iter piter; long long start, length; int partno; - int i; int err; if (!capable(CAP_SYS_ADMIN)) @@ -47,28 +48,33 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user mutex_lock(&bdev->bd_mutex); /* overlap? */ - for (i = 0; i < disk_max_parts(disk); i++) { - struct hd_struct *s = disk->part[i]; - - if (!s) - continue; - if (!(start+length <= s->start_sect || - start >= s->start_sect + s->nr_sects)) { + disk_part_iter_init(&piter, disk, + DISK_PITER_INCL_EMPTY); + while ((part = disk_part_iter_next(&piter))) { + if (!(start + length <= part->start_sect || + start >= part->start_sect + part->nr_sects)) { + disk_part_iter_exit(&piter); mutex_unlock(&bdev->bd_mutex); return -EBUSY; } } + disk_part_iter_exit(&piter); + /* all seems OK */ err = add_partition(disk, partno, start, length, ADDPART_FLAG_NONE); mutex_unlock(&bdev->bd_mutex); return err; case BLKPG_DEL_PARTITION: - if (!disk->part[partno - 1]) + part = disk_get_part(disk, partno); + if (!part) return -ENXIO; - bdevp = bdget_disk(disk, partno); + + bdevp = bdget(part_devt(part)); + disk_put_part(part); if (!bdevp) return -ENOMEM; + mutex_lock(&bdevp->bd_mutex); if (bdevp->bd_openers) { mutex_unlock(&bdevp->bd_mutex); -- cgit v1.2.3 From c9959059161ddd7bf4670cf47367033d6b2f79c4 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:47:21 +0900 Subject: block: fix diskstats access There are two variants of stat functions - ones prefixed with double underbars which don't care about preemption and ones without which disable preemption before manipulating per-cpu counters. It's unclear whether the underbarred ones assume that preemtion is disabled on entry as some callers don't do that. This patch unifies diskstats access by implementing disk_stat_lock() and disk_stat_unlock() which take care of both RCU (for partition access) and preemption (for per-cpu counter access). diskstats access should always be enclosed between the two functions. As such, there's no need for the versions which disables preemption. They're removed and double underbars ones are renamed to drop the underbars. As an extra argument is added, there's no danger of using the old version unconverted. disk_stat_lock() uses get_cpu() and returns the cpu index and all diskstat functions which access per-cpu counters now has @cpu argument to help RT. This change adds RCU or preemption operations at some places but also collapses several preemption ops into one at others. Overall, the performance difference should be negligible as all involved ops are very lightweight per-cpu ones. Signed-off-by: Tejun Heo Cc: Peter Zijlstra Signed-off-by: Jens Axboe --- block/blk-core.c | 52 +++++++++++++++++++++++++++------------------------- block/blk-merge.c | 11 ++++++----- block/genhd.c | 20 +++++++++++--------- 3 files changed, 44 insertions(+), 39 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index d6128d9ad60..e0a5ee36849 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -56,25 +56,26 @@ static void drive_stat_acct(struct request *rq, int new_io) { struct hd_struct *part; int rw = rq_data_dir(rq); + int cpu; if (!blk_fs_request(rq) || !rq->rq_disk) return; - rcu_read_lock(); - + cpu = disk_stat_lock(); part = disk_map_sector_rcu(rq->rq_disk, rq->sector); + if (!new_io) - __all_stat_inc(rq->rq_disk, part, merges[rw], rq->sector); + all_stat_inc(cpu, rq->rq_disk, part, merges[rw], rq->sector); else { - disk_round_stats(rq->rq_disk); + disk_round_stats(cpu, rq->rq_disk); rq->rq_disk->in_flight++; if (part) { - part_round_stats(part); + part_round_stats(cpu, part); part->in_flight++; } } - rcu_read_unlock(); + disk_stat_unlock(); } void blk_queue_congestion_threshold(struct request_queue *q) @@ -997,7 +998,7 @@ static inline void add_request(struct request_queue *q, struct request *req) * /proc/diskstats. This accounts immediately for all queue usage up to * the current jiffies and restarts the counters again. */ -void disk_round_stats(struct gendisk *disk) +void disk_round_stats(int cpu, struct gendisk *disk) { unsigned long now = jiffies; @@ -1005,15 +1006,15 @@ void disk_round_stats(struct gendisk *disk) return; if (disk->in_flight) { - __disk_stat_add(disk, time_in_queue, - disk->in_flight * (now - disk->stamp)); - __disk_stat_add(disk, io_ticks, (now - disk->stamp)); + disk_stat_add(cpu, disk, time_in_queue, + disk->in_flight * (now - disk->stamp)); + disk_stat_add(cpu, disk, io_ticks, (now - disk->stamp)); } disk->stamp = now; } EXPORT_SYMBOL_GPL(disk_round_stats); -void part_round_stats(struct hd_struct *part) +void part_round_stats(int cpu, struct hd_struct *part) { unsigned long now = jiffies; @@ -1021,9 +1022,9 @@ void part_round_stats(struct hd_struct *part) return; if (part->in_flight) { - __part_stat_add(part, time_in_queue, - part->in_flight * (now - part->stamp)); - __part_stat_add(part, io_ticks, (now - part->stamp)); + part_stat_add(cpu, part, time_in_queue, + part->in_flight * (now - part->stamp)); + part_stat_add(cpu, part, io_ticks, (now - part->stamp)); } part->stamp = now; } @@ -1563,12 +1564,13 @@ static int __end_that_request_first(struct request *req, int error, if (blk_fs_request(req) && req->rq_disk) { const int rw = rq_data_dir(req); struct hd_struct *part; + int cpu; - rcu_read_lock(); + cpu = disk_stat_lock(); part = disk_map_sector_rcu(req->rq_disk, req->sector); - all_stat_add(req->rq_disk, part, sectors[rw], - nr_bytes >> 9, req->sector); - rcu_read_unlock(); + all_stat_add(cpu, req->rq_disk, part, sectors[rw], + nr_bytes >> 9, req->sector); + disk_stat_unlock(); } total_bytes = bio_nbytes = 0; @@ -1753,21 +1755,21 @@ static void end_that_request_last(struct request *req, int error) unsigned long duration = jiffies - req->start_time; const int rw = rq_data_dir(req); struct hd_struct *part; + int cpu; - rcu_read_lock(); - + cpu = disk_stat_lock(); part = disk_map_sector_rcu(disk, req->sector); - __all_stat_inc(disk, part, ios[rw], req->sector); - __all_stat_add(disk, part, ticks[rw], duration, req->sector); - disk_round_stats(disk); + all_stat_inc(cpu, disk, part, ios[rw], req->sector); + all_stat_add(cpu, disk, part, ticks[rw], duration, req->sector); + disk_round_stats(cpu, disk); disk->in_flight--; if (part) { - part_round_stats(part); + part_round_stats(cpu, part); part->in_flight--; } - rcu_read_unlock(); + disk_stat_unlock(); } if (req->end_io) diff --git a/block/blk-merge.c b/block/blk-merge.c index eb2a3ca5830..d926a24bf1f 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -388,18 +388,19 @@ static int attempt_merge(struct request_queue *q, struct request *req, if (req->rq_disk) { struct hd_struct *part; + int cpu; - rcu_read_lock(); - + cpu = disk_stat_lock(); part = disk_map_sector_rcu(req->rq_disk, req->sector); - disk_round_stats(req->rq_disk); + + disk_round_stats(cpu, req->rq_disk); req->rq_disk->in_flight--; if (part) { - part_round_stats(part); + part_round_stats(cpu, part); part->in_flight--; } - rcu_read_unlock(); + disk_stat_unlock(); } req->ioprio = ioprio_best(req->ioprio, next->ioprio); diff --git a/block/genhd.c b/block/genhd.c index b431d654394..430626e440f 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -633,10 +633,11 @@ static ssize_t disk_stat_show(struct device *dev, struct device_attribute *attr, char *buf) { struct gendisk *disk = dev_to_disk(dev); + int cpu; - preempt_disable(); - disk_round_stats(disk); - preempt_enable(); + cpu = disk_stat_lock(); + disk_round_stats(cpu, disk); + disk_stat_unlock(); return sprintf(buf, "%8lu %8lu %8llu %8u " "%8lu %8lu %8llu %8u " @@ -749,6 +750,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) struct disk_part_iter piter; struct hd_struct *hd; char buf[BDEVNAME_SIZE]; + int cpu; /* if (&gp->dev.kobj.entry == block_class.devices.next) @@ -758,9 +760,9 @@ static int diskstats_show(struct seq_file *seqf, void *v) "\n\n"); */ - preempt_disable(); - disk_round_stats(gp); - preempt_enable(); + cpu = disk_stat_lock(); + disk_round_stats(cpu, gp); + disk_stat_unlock(); seq_printf(seqf, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", MAJOR(disk_devt(gp)), MINOR(disk_devt(gp)), disk_name(gp, 0, buf), @@ -777,9 +779,9 @@ static int diskstats_show(struct seq_file *seqf, void *v) /* now show all non-0 size partitions of it */ disk_part_iter_init(&piter, gp, 0); while ((hd = disk_part_iter_next(&piter))) { - preempt_disable(); - part_round_stats(hd); - preempt_enable(); + cpu = disk_stat_lock(); + part_round_stats(cpu, hd); + disk_stat_unlock(); seq_printf(seqf, "%4d %4d %s %lu %lu %llu " "%u %lu %lu %llu %u %u %u %u\n", MAJOR(part_devt(hd)), MINOR(part_devt(hd)), -- cgit v1.2.3 From bcce3de1be61e424deef35d1e86e86a35c4b6e65 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:47:22 +0900 Subject: block: implement extended dev numbers Implement extended device numbers. A block driver can tell block layer that it wants to use extended device numbers. After the usual minor space is used up, block layer automatically allocates devt's from EXT_BLOCK_MAJOR. Currently only one major number is allocated for this but as the allocation is strictly on-demand, ~1mil minor space under it should suffice unless the system actually has more than ~1mil partitions and if that ever happens adding more majors to the extended devt area is easy. Due to internal implementation issues, the first partition can't be allocated on the extended area. In other words, genhd->minors should at least be 1. This limitation will be lifted by later changes. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 120 +++++++++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 115 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 430626e440f..7bbfed05cec 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -16,6 +16,7 @@ #include #include #include +#include #include "blk.h" @@ -24,6 +25,15 @@ static DEFINE_MUTEX(block_class_lock); struct kobject *block_depr; #endif +/* for extended dynamic devt allocation, currently only one major is used */ +#define MAX_EXT_DEVT (1 << MINORBITS) + +/* For extended devt allocation. ext_devt_mutex prevents look up + * results from going away underneath its user. + */ +static DEFINE_MUTEX(ext_devt_mutex); +static DEFINE_IDR(ext_devt_idr); + static struct device_type disk_type; /** @@ -288,6 +298,74 @@ EXPORT_SYMBOL(unregister_blkdev); static struct kobj_map *bdev_map; +/** + * blk_alloc_devt - allocate a dev_t for a partition + * @part: partition to allocate dev_t for + * @gfp_mask: memory allocation flag + * @devt: out parameter for resulting dev_t + * + * Allocate a dev_t for block device. + * + * RETURNS: + * 0 on success, allocated dev_t is returned in *@devt. -errno on + * failure. + * + * CONTEXT: + * Might sleep. + */ +int blk_alloc_devt(struct hd_struct *part, dev_t *devt) +{ + struct gendisk *disk = part_to_disk(part); + int idx, rc; + + /* in consecutive minor range? */ + if (part->partno < disk->minors) { + *devt = MKDEV(disk->major, disk->first_minor + part->partno); + return 0; + } + + /* allocate ext devt */ + do { + if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL)) + return -ENOMEM; + rc = idr_get_new(&ext_devt_idr, part, &idx); + } while (rc == -EAGAIN); + + if (rc) + return rc; + + if (idx > MAX_EXT_DEVT) { + idr_remove(&ext_devt_idr, idx); + return -EBUSY; + } + + *devt = MKDEV(BLOCK_EXT_MAJOR, idx); + return 0; +} + +/** + * blk_free_devt - free a dev_t + * @devt: dev_t to free + * + * Free @devt which was allocated using blk_alloc_devt(). + * + * CONTEXT: + * Might sleep. + */ +void blk_free_devt(dev_t devt) +{ + might_sleep(); + + if (devt == MKDEV(0, 0)) + return; + + if (MAJOR(devt) == BLOCK_EXT_MAJOR) { + mutex_lock(&ext_devt_mutex); + idr_remove(&ext_devt_idr, MINOR(devt)); + mutex_unlock(&ext_devt_mutex); + } +} + /* * Register device numbers dev..(dev+range-1) * range must be nonzero @@ -371,10 +449,27 @@ void unlink_gendisk(struct gendisk *disk) */ struct gendisk *get_gendisk(dev_t devt, int *partno) { - struct kobject *kobj = kobj_lookup(bdev_map, devt, partno); - struct device *dev = kobj_to_dev(kobj); + struct gendisk *disk = NULL; + + if (MAJOR(devt) != BLOCK_EXT_MAJOR) { + struct kobject *kobj; + + kobj = kobj_lookup(bdev_map, devt, partno); + if (kobj) + disk = dev_to_disk(kobj_to_dev(kobj)); + } else { + struct hd_struct *part; + + mutex_lock(&ext_devt_mutex); + part = idr_find(&ext_devt_idr, MINOR(devt)); + if (part && get_disk(part_to_disk(part))) { + *partno = part->partno; + disk = part_to_disk(part); + } + mutex_unlock(&ext_devt_mutex); + } - return kobj ? dev_to_disk(dev) : NULL; + return disk; } /** @@ -877,18 +972,30 @@ struct gendisk *alloc_disk(int minors) } struct gendisk *alloc_disk_node(int minors, int node_id) +{ + return alloc_disk_ext_node(minors, 0, node_id); +} + +struct gendisk *alloc_disk_ext(int minors, int ext_minors) +{ + return alloc_disk_ext_node(minors, ext_minors, -1); +} + +struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) { struct gendisk *disk; disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL | __GFP_ZERO, node_id); if (disk) { + int tot_minors = minors + ext_minors; + if (!init_disk_stats(disk)) { kfree(disk); return NULL; } - if (minors > 1) { - int size = (minors - 1) * sizeof(struct hd_struct *); + if (tot_minors > 1) { + int size = (tot_minors - 1) * sizeof(struct hd_struct *); disk->__part = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, node_id); if (!disk->__part) { @@ -898,6 +1005,7 @@ struct gendisk *alloc_disk_node(int minors, int node_id) } } disk->minors = minors; + disk->ext_minors = ext_minors; rand_initialize_disk(disk); disk->dev.class = &block_class; disk->dev.type = &disk_type; @@ -910,6 +1018,8 @@ struct gendisk *alloc_disk_node(int minors, int node_id) EXPORT_SYMBOL(alloc_disk); EXPORT_SYMBOL(alloc_disk_node); +EXPORT_SYMBOL(alloc_disk_ext); +EXPORT_SYMBOL(alloc_disk_ext_node); struct kobject *get_disk(struct gendisk *disk) { -- cgit v1.2.3 From 1f0142905d4812966831613847db38a66da29eb8 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:47:23 +0900 Subject: block: adjust formatting for large minors and add ext_range sysfs attr With extended minors and the soon-to-follow debug feature, large minor numbers for block devices will be common. This patch does the followings to make printouts pretty. * Adapt print formats such that large minors don't break the formatting. * For extended MAJ:MIN, %02x%02x for MAJ:MIN used in printk_all_partitions() doesn't cut it anymore. Update it such that %03x:%05x is used if either MAJ or MIN doesn't fit in %02x. * Implement ext_range sysfs attribute which shows total minors the device can use including both conventional minor space and the extended one. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 7bbfed05cec..ee4b13520e5 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -366,6 +366,18 @@ void blk_free_devt(dev_t devt) } } +static char *bdevt_str(dev_t devt, char *buf) +{ + if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { + char tbuf[BDEVT_SIZE]; + snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); + snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); + } else + snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); + + return buf; +} + /* * Register device numbers dev..(dev+range-1) * range must be nonzero @@ -521,7 +533,8 @@ void __init printk_all_partitions(void) struct gendisk *disk = dev_to_disk(dev); struct disk_part_iter piter; struct hd_struct *part; - char buf[BDEVNAME_SIZE]; + char name_buf[BDEVNAME_SIZE]; + char devt_buf[BDEVT_SIZE]; /* * Don't show empty devices or things that have been @@ -536,10 +549,10 @@ void __init printk_all_partitions(void) * numbers in hex - the same format as the root= * option takes. */ - printk("%02x%02x %10llu %s", - MAJOR(disk_devt(disk)), MINOR(disk_devt(disk)), + printk("%s %10llu %s", + bdevt_str(disk_devt(disk), devt_buf), (unsigned long long)get_capacity(disk) >> 1, - disk_name(disk, 0, buf)); + disk_name(disk, 0, name_buf)); if (disk->driverfs_dev != NULL && disk->driverfs_dev->driver != NULL) printk(" driver: %s\n", @@ -550,10 +563,10 @@ void __init printk_all_partitions(void) /* now show the partitions */ disk_part_iter_init(&piter, disk, 0); while ((part = disk_part_iter_next(&piter))) - printk(" %02x%02x %10llu %s\n", - MAJOR(part_devt(part)), MINOR(part_devt(part)), + printk(" %s %10llu %s\n", + bdevt_str(part_devt(part), devt_buf), (unsigned long long)part->nr_sects >> 1, - disk_name(disk, part->partno, buf)); + disk_name(disk, part->partno, name_buf)); disk_part_iter_exit(&piter); } class_dev_iter_exit(&iter); @@ -630,14 +643,14 @@ static int show_partition(struct seq_file *seqf, void *v) return 0; /* show the full disk and all non-0 size partitions of it */ - seq_printf(seqf, "%4d %4d %10llu %s\n", + seq_printf(seqf, "%4d %7d %10llu %s\n", MAJOR(disk_devt(sgp)), MINOR(disk_devt(sgp)), (unsigned long long)get_capacity(sgp) >> 1, disk_name(sgp, 0, buf)); disk_part_iter_init(&piter, sgp, 0); while ((part = disk_part_iter_next(&piter))) - seq_printf(seqf, "%4d %4d %10llu %s\n", + seq_printf(seqf, "%4d %7d %10llu %s\n", MAJOR(part_devt(part)), MINOR(part_devt(part)), (unsigned long long)part->nr_sects >> 1, disk_name(sgp, part->partno, buf)); @@ -691,6 +704,14 @@ static ssize_t disk_range_show(struct device *dev, return sprintf(buf, "%d\n", disk->minors); } +static ssize_t disk_ext_range_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + + return sprintf(buf, "%d\n", disk_max_parts(disk) + 1); +} + static ssize_t disk_removable_show(struct device *dev, struct device_attribute *attr, char *buf) { @@ -780,6 +801,7 @@ static ssize_t disk_fail_store(struct device *dev, #endif static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); +static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL); @@ -792,6 +814,7 @@ static struct device_attribute dev_attr_fail = static struct attribute *disk_attrs[] = { &dev_attr_range.attr, + &dev_attr_ext_range.attr, &dev_attr_removable.attr, &dev_attr_ro.attr, &dev_attr_size.attr, @@ -858,7 +881,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) cpu = disk_stat_lock(); disk_round_stats(cpu, gp); disk_stat_unlock(); - seq_printf(seqf, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", + seq_printf(seqf, "%4d %7d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", MAJOR(disk_devt(gp)), MINOR(disk_devt(gp)), disk_name(gp, 0, buf), disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]), @@ -877,7 +900,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) cpu = disk_stat_lock(); part_round_stats(cpu, hd); disk_stat_unlock(); - seq_printf(seqf, "%4d %4d %s %lu %lu %llu " + seq_printf(seqf, "%4d %7d %s %lu %lu %llu " "%u %lu %lu %llu %u %u %u %u\n", MAJOR(part_devt(hd)), MINOR(part_devt(hd)), disk_name(gp, hd->partno, buf), -- cgit v1.2.3 From 870d6656126add8e383645732b03df2b7ccd4f94 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:47:25 +0900 Subject: block: implement CONFIG_DEBUG_BLOCK_EXT_DEVT Extended devt introduces non-contiguos device numbers. This patch implements a debug option which forces most devt allocations to be from the extended area and spreads them out. This is enabled by default if DEBUG_KERNEL is set and achieves... 1. Detects code paths in kernel or userland which expect predetermined consecutive device numbers. 2. When something goes wrong, avoid corruption as adding to the minor of earlier partition won't lead to the wrong but valid device. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 38 +++++++++++++++++++++++++++++++++++--- 1 file changed, 35 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index ee4b13520e5..67e5a59ced2 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -298,6 +298,38 @@ EXPORT_SYMBOL(unregister_blkdev); static struct kobj_map *bdev_map; +/** + * blk_mangle_minor - scatter minor numbers apart + * @minor: minor number to mangle + * + * Scatter consecutively allocated @minor number apart if MANGLE_DEVT + * is enabled. Mangling twice gives the original value. + * + * RETURNS: + * Mangled value. + * + * CONTEXT: + * Don't care. + */ +static int blk_mangle_minor(int minor) +{ +#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT + int i; + + for (i = 0; i < MINORBITS / 2; i++) { + int low = minor & (1 << i); + int high = minor & (1 << (MINORBITS - 1 - i)); + int distance = MINORBITS - 1 - 2 * i; + + minor ^= low | high; /* clear both bits */ + low <<= distance; /* swap the positions */ + high >>= distance; + minor |= low | high; /* and set */ + } +#endif + return minor; +} + /** * blk_alloc_devt - allocate a dev_t for a partition * @part: partition to allocate dev_t for @@ -339,7 +371,7 @@ int blk_alloc_devt(struct hd_struct *part, dev_t *devt) return -EBUSY; } - *devt = MKDEV(BLOCK_EXT_MAJOR, idx); + *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx)); return 0; } @@ -361,7 +393,7 @@ void blk_free_devt(dev_t devt) if (MAJOR(devt) == BLOCK_EXT_MAJOR) { mutex_lock(&ext_devt_mutex); - idr_remove(&ext_devt_idr, MINOR(devt)); + idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); mutex_unlock(&ext_devt_mutex); } } @@ -473,7 +505,7 @@ struct gendisk *get_gendisk(dev_t devt, int *partno) struct hd_struct *part; mutex_lock(&ext_devt_mutex); - part = idr_find(&ext_devt_idr, MINOR(devt)); + part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt))); if (part && get_disk(part_to_disk(part))) { *partno = part->partno; disk = part_to_disk(part); -- cgit v1.2.3 From ed9e1982347b36573cd622ee5f4e2a7ccd79b3fd Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:05 +0900 Subject: block: implement and use {disk|part}_to_dev() Implement {disk|part}_to_dev() and use them to access generic device instead of directly dereferencing {disk|part}->dev. To make sure no user is left behind, rename generic devices fields to __dev. This is in preparation of unifying partition 0 handling with other partitions. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-integrity.c | 5 +++-- block/blk-sysfs.c | 4 ++-- block/genhd.c | 27 ++++++++++++++------------- 3 files changed, 19 insertions(+), 17 deletions(-) (limited to 'block') diff --git a/block/blk-integrity.c b/block/blk-integrity.c index d87606eaca1..69023da6315 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -331,7 +331,8 @@ int blk_integrity_register(struct gendisk *disk, struct blk_integrity *template) return -1; if (kobject_init_and_add(&bi->kobj, &integrity_ktype, - &disk->dev.kobj, "%s", "integrity")) { + &disk_to_dev(disk)->kobj, + "%s", "integrity")) { kmem_cache_free(integrity_cachep, bi); return -1; } @@ -375,7 +376,7 @@ void blk_integrity_unregister(struct gendisk *disk) kobject_uevent(&bi->kobj, KOBJ_REMOVE); kobject_del(&bi->kobj); - kobject_put(&disk->dev.kobj); + kobject_put(&disk_to_dev(disk)->kobj); kmem_cache_free(integrity_cachep, bi); } EXPORT_SYMBOL(blk_integrity_unregister); diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 304ec73ab82..b9a6ed16664 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -310,7 +310,7 @@ int blk_register_queue(struct gendisk *disk) if (!q->request_fn) return 0; - ret = kobject_add(&q->kobj, kobject_get(&disk->dev.kobj), + ret = kobject_add(&q->kobj, kobject_get(&disk_to_dev(disk)->kobj), "%s", "queue"); if (ret < 0) return ret; @@ -339,6 +339,6 @@ void blk_unregister_queue(struct gendisk *disk) kobject_uevent(&q->kobj, KOBJ_REMOVE); kobject_del(&q->kobj); - kobject_put(&disk->dev.kobj); + kobject_put(&disk_to_dev(disk)->kobj); } } diff --git a/block/genhd.c b/block/genhd.c index 67e5a59ced2..0a2f16bd54b 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -59,7 +59,7 @@ struct hd_struct *disk_get_part(struct gendisk *disk, int partno) rcu_read_lock(); part = rcu_dereference(disk->__part[partno - 1]); if (part) - get_device(&part->dev); + get_device(part_to_dev(part)); rcu_read_unlock(); return part; @@ -130,7 +130,7 @@ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects) continue; - get_device(&part->dev); + get_device(part_to_dev(part)); piter->part = part; piter->idx += inc; break; @@ -435,7 +435,7 @@ static struct kobject *exact_match(dev_t devt, int *partno, void *data) { struct gendisk *p = data; - return &p->dev.kobj; + return &disk_to_dev(p)->kobj; } static int exact_lock(dev_t devt, void *data) @@ -460,7 +460,7 @@ void add_disk(struct gendisk *disk) int retval; disk->flags |= GENHD_FL_UP; - disk->dev.devt = MKDEV(disk->major, disk->first_minor); + disk_to_dev(disk)->devt = MKDEV(disk->major, disk->first_minor); blk_register_region(disk_devt(disk), disk->minors, NULL, exact_match, exact_lock, disk); register_disk(disk); @@ -468,7 +468,8 @@ void add_disk(struct gendisk *disk) bdi = &disk->queue->backing_dev_info; bdi_register_dev(bdi, disk_devt(disk)); - retval = sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi"); + retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, + "bdi"); WARN_ON(retval); } @@ -477,7 +478,7 @@ EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */ void unlink_gendisk(struct gendisk *disk) { - sysfs_remove_link(&disk->dev.kobj, "bdi"); + sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); bdi_unregister(&disk->queue->backing_dev_info); blk_unregister_queue(disk); blk_unregister_region(disk_devt(disk), disk->minors); @@ -903,7 +904,7 @@ static int diskstats_show(struct seq_file *seqf, void *v) int cpu; /* - if (&gp->dev.kobj.entry == block_class.devices.next) + if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) seq_puts(seqf, "major minor name" " rio rmerge rsect ruse wio wmerge " "wsect wuse running use aveq" @@ -972,7 +973,7 @@ static void media_change_notify_thread(struct work_struct *work) * set enviroment vars to indicate which event this is for * so that user space will know to go check the media status. */ - kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp); + kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); put_device(gd->driverfs_dev); } @@ -1062,9 +1063,9 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) disk->minors = minors; disk->ext_minors = ext_minors; rand_initialize_disk(disk); - disk->dev.class = &block_class; - disk->dev.type = &disk_type; - device_initialize(&disk->dev); + disk_to_dev(disk)->class = &block_class; + disk_to_dev(disk)->type = &disk_type; + device_initialize(disk_to_dev(disk)); INIT_WORK(&disk->async_notify, media_change_notify_thread); } @@ -1086,7 +1087,7 @@ struct kobject *get_disk(struct gendisk *disk) owner = disk->fops->owner; if (owner && !try_module_get(owner)) return NULL; - kobj = kobject_get(&disk->dev.kobj); + kobj = kobject_get(&disk_to_dev(disk)->kobj); if (kobj == NULL) { module_put(owner); return NULL; @@ -1100,7 +1101,7 @@ EXPORT_SYMBOL(get_disk); void put_disk(struct gendisk *disk) { if (disk) - kobject_put(&disk->dev.kobj); + kobject_put(&disk_to_dev(disk)->kobj); } EXPORT_SYMBOL(put_disk); -- cgit v1.2.3 From b5d0b9df0ba5d9a044f3a21e7544f53d90bd1465 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 3 Sep 2008 09:06:42 +0200 Subject: block: introduce partition 0 genhd and partition code handled disk and partitions separately. All information about the whole disk was in struct genhd and partitions in struct hd_struct. However, the whole disk (part0) and other partitions have a lot in common and the data structures end up having good number of common fields and thus separate code paths doing the same thing. Also, the partition array was indexed by partno - 1 which gets pretty confusing at times. This patch introduces partition 0 and makes the partition array indexed by partno. Following patches will unify the handling of disk and parts piece-by-piece. This patch also implements disk_partitionable() which tests whether a disk is partitionable. With coming dynamic partition array change, the most common usage of disk_max_parts() will be testing whether a disk is partitionable and the number of max partitions will become much less important. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 40 +++++++++++++++++++++++----------------- block/ioctl.c | 4 ++-- 2 files changed, 25 insertions(+), 19 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 0a2f16bd54b..65b7386c26d 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -54,10 +54,10 @@ struct hd_struct *disk_get_part(struct gendisk *disk, int partno) { struct hd_struct *part; - if (unlikely(partno < 1 || partno > disk_max_parts(disk))) + if (unlikely(partno < 0 || partno >= disk_max_parts(disk))) return NULL; rcu_read_lock(); - part = rcu_dereference(disk->__part[partno - 1]); + part = rcu_dereference(disk->__part[partno]); if (part) get_device(part_to_dev(part)); rcu_read_unlock(); @@ -85,8 +85,10 @@ void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, if (flags & DISK_PITER_REVERSE) piter->idx = disk_max_parts(piter->disk) - 1; - else + else if (flags & DISK_PITER_INCL_PART0) piter->idx = 0; + else + piter->idx = 1; piter->flags = flags; } @@ -114,7 +116,10 @@ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) /* determine iteration parameters */ if (piter->flags & DISK_PITER_REVERSE) { inc = -1; - end = -1; + if (piter->flags & DISK_PITER_INCL_PART0) + end = -1; + else + end = 0; } else { inc = 1; end = disk_max_parts(piter->disk); @@ -177,7 +182,7 @@ struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) { int i; - for (i = 0; i < disk_max_parts(disk); i++) { + for (i = 1; i < disk_max_parts(disk); i++) { struct hd_struct *part = rcu_dereference(disk->__part[i]); if (part && part->start_sect <= sector && @@ -669,7 +674,7 @@ static int show_partition(struct seq_file *seqf, void *v) char buf[BDEVNAME_SIZE]; /* Don't show non-partitionable removeable devices or empty devices */ - if (!get_capacity(sgp) || (!disk_max_parts(sgp) && + if (!get_capacity(sgp) || (!disk_partitionable(sgp) && (sgp->flags & GENHD_FL_REMOVABLE))) return 0; if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) @@ -742,7 +747,7 @@ static ssize_t disk_ext_range_show(struct device *dev, { struct gendisk *disk = dev_to_disk(dev); - return sprintf(buf, "%d\n", disk_max_parts(disk) + 1); + return sprintf(buf, "%d\n", disk_max_parts(disk)); } static ssize_t disk_removable_show(struct device *dev, @@ -998,7 +1003,7 @@ dev_t blk_lookup_devt(const char *name, int partno) if (strcmp(dev->bus_id, name)) continue; - if (partno < 0 || partno > disk_max_parts(disk)) + if (partno < 0 || partno >= disk_max_parts(disk)) continue; if (partno == 0) @@ -1045,21 +1050,22 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) GFP_KERNEL | __GFP_ZERO, node_id); if (disk) { int tot_minors = minors + ext_minors; + int size = tot_minors * sizeof(struct hd_struct *); if (!init_disk_stats(disk)) { kfree(disk); return NULL; } - if (tot_minors > 1) { - int size = (tot_minors - 1) * sizeof(struct hd_struct *); - disk->__part = kmalloc_node(size, - GFP_KERNEL | __GFP_ZERO, node_id); - if (!disk->__part) { - free_disk_stats(disk); - kfree(disk); - return NULL; - } + + disk->__part = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, + node_id); + if (!disk->__part) { + free_disk_stats(disk); + kfree(disk); + return NULL; } + disk->__part[0] = &disk->part0; + disk->minors = minors; disk->ext_minors = ext_minors; rand_initialize_disk(disk); diff --git a/block/ioctl.c b/block/ioctl.c index a5f672ad55f..64e7c67a64b 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -30,7 +30,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user if (bdev != bdev->bd_contains) return -EINVAL; partno = p.pno; - if (partno <= 0 || partno > disk_max_parts(disk)) + if (partno <= 0 || partno >= disk_max_parts(disk)) return -EINVAL; switch (a.op) { case BLKPG_ADD_PARTITION: @@ -102,7 +102,7 @@ static int blkdev_reread_part(struct block_device *bdev) struct gendisk *disk = bdev->bd_disk; int res; - if (!disk_max_parts(disk) || bdev != bdev->bd_contains) + if (!disk_partitionable(disk) || bdev != bdev->bd_contains) return -EINVAL; if (!capable(CAP_SYS_ADMIN)) return -EACCES; -- cgit v1.2.3 From 548b10eb2959c96cef6fc29fc96e0931eeb53bc5 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 29 Aug 2008 09:01:47 +0200 Subject: block: move __dev from disk to part0 Move disk->__dev to part0->__dev. This simplifies bdget_disk() and lookup_devt() and allows common sysfs attributes to be unified. part_to_disk() is updated to handle part0 -> disk. Updated to include a fix from Bartlomiej Zolnierkiewicz , he writes: "part0 is a "special" partition and doesn't need to have capacity set - this fixes regression caused by "block: move __dev from disk to part0" commit." Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 40 ++++++++++++---------------------------- 1 file changed, 12 insertions(+), 28 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 65b7386c26d..36b9f1bdd91 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -537,22 +537,15 @@ struct gendisk *get_gendisk(dev_t devt, int *partno) */ extern struct block_device *bdget_disk(struct gendisk *disk, int partno) { - dev_t devt = MKDEV(0, 0); + struct hd_struct *part; + struct block_device *bdev = NULL; - if (partno == 0) - devt = disk_devt(disk); - else { - struct hd_struct *part; + part = disk_get_part(disk, partno); + if (part && (part->nr_sects || partno == 0)) + bdev = bdget(part_devt(part)); + disk_put_part(part); - part = disk_get_part(disk, partno); - if (part && part->nr_sects) - devt = part_devt(part); - disk_put_part(part); - } - - if (likely(devt != MKDEV(0, 0))) - return bdget(devt); - return NULL; + return bdev; } EXPORT_SYMBOL(bdget_disk); @@ -1000,27 +993,18 @@ dev_t blk_lookup_devt(const char *name, int partno) class_dev_iter_init(&iter, &block_class, NULL, &disk_type); while ((dev = class_dev_iter_next(&iter))) { struct gendisk *disk = dev_to_disk(dev); + struct hd_struct *part; if (strcmp(dev->bus_id, name)) continue; - if (partno < 0 || partno >= disk_max_parts(disk)) - continue; - - if (partno == 0) - devt = disk_devt(disk); - else { - struct hd_struct *part; - - part = disk_get_part(disk, partno); - if (!part || !part->nr_sects) { - disk_put_part(part); - continue; - } + part = disk_get_part(disk, partno); + if (part && (part->nr_sects || partno == 0)) { devt = part_devt(part); disk_put_part(part); + break; } - break; + disk_put_part(part); } class_dev_iter_exit(&iter); return devt; -- cgit v1.2.3 From e56105214943ce5f0901d20e972a7cfd0d1d0656 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:09 +0900 Subject: block: unify sysfs size node handling Now that capacity and __dev are moved to part0, part0 and others can share the same method. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 36b9f1bdd91..c70db35076a 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -760,14 +760,6 @@ static ssize_t disk_ro_show(struct device *dev, return sprintf(buf, "%d\n", disk->policy ? 1 : 0); } -static ssize_t disk_size_show(struct device *dev, - struct device_attribute *attr, char *buf) -{ - struct gendisk *disk = dev_to_disk(dev); - - return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk)); -} - static ssize_t disk_capability_show(struct device *dev, struct device_attribute *attr, char *buf) { @@ -835,7 +827,7 @@ static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); -static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL); +static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL); #ifdef CONFIG_FAIL_MAKE_REQUEST -- cgit v1.2.3 From b7db9956e57c8151b930d5e5fe5c766e6aad3ff7 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:10 +0900 Subject: block: move policy from disk to part0 Move disk->policy to part0->policy. Implement and use get_disk_ro(). Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index c70db35076a..70358f3c742 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -757,7 +757,7 @@ static ssize_t disk_ro_show(struct device *dev, { struct gendisk *disk = dev_to_disk(dev); - return sprintf(buf, "%d\n", disk->policy ? 1 : 0); + return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); } static ssize_t disk_capability_show(struct device *dev, @@ -1090,10 +1090,7 @@ EXPORT_SYMBOL(put_disk); void set_device_ro(struct block_device *bdev, int flag) { - if (bdev->bd_contains != bdev) - bdev->bd_part->policy = flag; - else - bdev->bd_disk->policy = flag; + bdev->bd_part->policy = flag; } EXPORT_SYMBOL(set_device_ro); @@ -1103,8 +1100,8 @@ void set_disk_ro(struct gendisk *disk, int flag) struct disk_part_iter piter; struct hd_struct *part; - disk->policy = flag; - disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY); + disk_part_iter_init(&piter, disk, + DISK_PITER_INCL_EMPTY | DISK_PITER_INCL_PART0); while ((part = disk_part_iter_next(&piter))) part->policy = flag; disk_part_iter_exit(&piter); @@ -1116,10 +1113,7 @@ int bdev_read_only(struct block_device *bdev) { if (!bdev) return 0; - else if (bdev->bd_contains != bdev) - return bdev->bd_part->policy; - else - return bdev->bd_disk->policy; + return bdev->bd_part->policy; } EXPORT_SYMBOL(bdev_read_only); -- cgit v1.2.3 From 4c46501d1659475dc6c89554af6ce7fe6ecf615c Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:11 +0900 Subject: block: move holder_dir from disk to part0 Move disk->holder_dir to part0->holder_dir. Kill now mostly superflous bdev_get_holder(). While at it, kill superflous kobject_get/put() around holder_dir, slave_dir and cmd_filter creation and collapse disk_sysfs_add_subdirs() into register_disk(). These serve no purpose but obfuscating the code. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/cmd-filter.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) (limited to 'block') diff --git a/block/cmd-filter.c b/block/cmd-filter.c index 79c14996ac1..e669aed4c6b 100644 --- a/block/cmd-filter.c +++ b/block/cmd-filter.c @@ -211,14 +211,10 @@ int blk_register_filter(struct gendisk *disk) { int ret; struct blk_cmd_filter *filter = &disk->queue->cmd_filter; - struct kobject *parent = kobject_get(disk->holder_dir->parent); - if (!parent) - return -ENODEV; - - ret = kobject_init_and_add(&filter->kobj, &rcf_ktype, parent, + ret = kobject_init_and_add(&filter->kobj, &rcf_ktype, + &disk_to_dev(disk)->kobj, "%s", "cmd_filter"); - if (ret < 0) return ret; @@ -231,7 +227,6 @@ void blk_unregister_filter(struct gendisk *disk) struct blk_cmd_filter *filter = &disk->queue->cmd_filter; kobject_put(&filter->kobj); - kobject_put(disk->holder_dir->parent); } EXPORT_SYMBOL(blk_unregister_filter); #endif -- cgit v1.2.3 From 0762b8bde9729f10f8e6249809660ff2ec3ad735 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:12 +0900 Subject: block: always set bdev->bd_part Till now, bdev->bd_part is set only if the bdev was for parts other than part0. This patch makes bdev->bd_part always set so that code paths don't have to differenciate common handling. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index e0a5ee36849..a4a7c08d2f2 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1274,7 +1274,7 @@ __setup("fail_make_request=", setup_fail_make_request); static int should_fail_request(struct bio *bio) { if ((bio->bi_bdev->bd_disk->flags & GENHD_FL_FAIL) || - (bio->bi_bdev->bd_part && bio->bi_bdev->bd_part->make_it_fail)) + bio->bi_bdev->bd_part->make_it_fail) return should_fail(&fail_make_request, bio->bi_size); return 0; -- cgit v1.2.3 From eddb2e26b5ee3c5da68ba4bf1921ba20e2097bff Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:13 +0900 Subject: block: kill GENHD_FL_FAIL and use part0->make_it_fail GENHD_FL_FAIL for disk is what make_it_fail is for parts. Kill it and use part0->make_it_fail. Sysfs node handling is unified too. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 5 +++-- block/genhd.c | 30 +----------------------------- 2 files changed, 4 insertions(+), 31 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index a4a7c08d2f2..505ec61067d 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1273,8 +1273,9 @@ __setup("fail_make_request=", setup_fail_make_request); static int should_fail_request(struct bio *bio) { - if ((bio->bi_bdev->bd_disk->flags & GENHD_FL_FAIL) || - bio->bi_bdev->bd_part->make_it_fail) + struct hd_struct *part = bio->bi_bdev->bd_part; + + if (part_to_disk(part)->part0.make_it_fail || part->make_it_fail) return should_fail(&fail_make_request, bio->bi_size); return 0; diff --git a/block/genhd.c b/block/genhd.c index 70358f3c742..06a252f2b96 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -795,34 +795,6 @@ static ssize_t disk_stat_show(struct device *dev, jiffies_to_msecs(disk_stat_read(disk, time_in_queue))); } -#ifdef CONFIG_FAIL_MAKE_REQUEST -static ssize_t disk_fail_show(struct device *dev, - struct device_attribute *attr, char *buf) -{ - struct gendisk *disk = dev_to_disk(dev); - - return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0); -} - -static ssize_t disk_fail_store(struct device *dev, - struct device_attribute *attr, - const char *buf, size_t count) -{ - struct gendisk *disk = dev_to_disk(dev); - int i; - - if (count > 0 && sscanf(buf, "%d", &i) > 0) { - if (i == 0) - disk->flags &= ~GENHD_FL_FAIL; - else - disk->flags |= GENHD_FL_FAIL; - } - - return count; -} - -#endif - static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); @@ -832,7 +804,7 @@ static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL); #ifdef CONFIG_FAIL_MAKE_REQUEST static struct device_attribute dev_attr_fail = - __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store); + __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); #endif static struct attribute *disk_attrs[] = { -- cgit v1.2.3 From 074a7aca7afa6f230104e8e65eba3420263714a5 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:14 +0900 Subject: block: move stats from disk to part0 Move stats related fields - stamp, in_flight, dkstats - from disk to part0 and unify stat handling such that... * part_stat_*() now updates part0 together if the specified partition is not part0. ie. part_stat_*() are now essentially all_stat_*(). * {disk|all}_stat_*() are gone. * part_round_stats() is updated similary. It handles part0 stats automatically and disk_round_stats() is killed. * part_{inc|dec}_in_fligh() is implemented which automatically updates part0 stats for parts other than part0. * disk_map_sector_rcu() is updated to return part0 if no part matches. Combined with the above changes, this makes NULL special case handling in callers unnecessary. * Separate stats show code paths for disk are collapsed into part stats show code paths. * Rename disk_stat_lock/unlock() to part_stat_lock/unlock() While at it, reposition stat handling macros a bit and add missing parentheses around macro parameters. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 84 +++++++++++++++++++---------------------------- block/blk-merge.c | 12 +++---- block/genhd.c | 97 ++++++++++++++----------------------------------------- 3 files changed, 62 insertions(+), 131 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 505ec61067d..98138f00252 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -61,21 +61,17 @@ static void drive_stat_acct(struct request *rq, int new_io) if (!blk_fs_request(rq) || !rq->rq_disk) return; - cpu = disk_stat_lock(); + cpu = part_stat_lock(); part = disk_map_sector_rcu(rq->rq_disk, rq->sector); if (!new_io) - all_stat_inc(cpu, rq->rq_disk, part, merges[rw], rq->sector); + part_stat_inc(cpu, part, merges[rw]); else { - disk_round_stats(cpu, rq->rq_disk); - rq->rq_disk->in_flight++; - if (part) { - part_round_stats(cpu, part); - part->in_flight++; - } + part_round_stats(cpu, part); + part_inc_in_flight(part); } - disk_stat_unlock(); + part_stat_unlock(); } void blk_queue_congestion_threshold(struct request_queue *q) @@ -983,8 +979,22 @@ static inline void add_request(struct request_queue *q, struct request *req) __elv_add_request(q, req, ELEVATOR_INSERT_SORT, 0); } -/* - * disk_round_stats() - Round off the performance stats on a struct +static void part_round_stats_single(int cpu, struct hd_struct *part, + unsigned long now) +{ + if (now == part->stamp) + return; + + if (part->in_flight) { + __part_stat_add(cpu, part, time_in_queue, + part->in_flight * (now - part->stamp)); + __part_stat_add(cpu, part, io_ticks, (now - part->stamp)); + } + part->stamp = now; +} + +/** + * part_round_stats() - Round off the performance stats on a struct * disk_stats. * * The average IO queue length and utilisation statistics are maintained @@ -998,36 +1008,15 @@ static inline void add_request(struct request_queue *q, struct request *req) * /proc/diskstats. This accounts immediately for all queue usage up to * the current jiffies and restarts the counters again. */ -void disk_round_stats(int cpu, struct gendisk *disk) -{ - unsigned long now = jiffies; - - if (now == disk->stamp) - return; - - if (disk->in_flight) { - disk_stat_add(cpu, disk, time_in_queue, - disk->in_flight * (now - disk->stamp)); - disk_stat_add(cpu, disk, io_ticks, (now - disk->stamp)); - } - disk->stamp = now; -} -EXPORT_SYMBOL_GPL(disk_round_stats); - void part_round_stats(int cpu, struct hd_struct *part) { unsigned long now = jiffies; - if (now == part->stamp) - return; - - if (part->in_flight) { - part_stat_add(cpu, part, time_in_queue, - part->in_flight * (now - part->stamp)); - part_stat_add(cpu, part, io_ticks, (now - part->stamp)); - } - part->stamp = now; + if (part->partno) + part_round_stats_single(cpu, &part_to_disk(part)->part0, now); + part_round_stats_single(cpu, part, now); } +EXPORT_SYMBOL_GPL(part_round_stats); /* * queue lock must be held @@ -1567,11 +1556,10 @@ static int __end_that_request_first(struct request *req, int error, struct hd_struct *part; int cpu; - cpu = disk_stat_lock(); + cpu = part_stat_lock(); part = disk_map_sector_rcu(req->rq_disk, req->sector); - all_stat_add(cpu, req->rq_disk, part, sectors[rw], - nr_bytes >> 9, req->sector); - disk_stat_unlock(); + part_stat_add(cpu, part, sectors[rw], nr_bytes >> 9); + part_stat_unlock(); } total_bytes = bio_nbytes = 0; @@ -1758,19 +1746,15 @@ static void end_that_request_last(struct request *req, int error) struct hd_struct *part; int cpu; - cpu = disk_stat_lock(); + cpu = part_stat_lock(); part = disk_map_sector_rcu(disk, req->sector); - all_stat_inc(cpu, disk, part, ios[rw], req->sector); - all_stat_add(cpu, disk, part, ticks[rw], duration, req->sector); - disk_round_stats(cpu, disk); - disk->in_flight--; - if (part) { - part_round_stats(cpu, part); - part->in_flight--; - } + part_stat_inc(cpu, part, ios[rw]); + part_stat_add(cpu, part, ticks[rw], duration); + part_round_stats(cpu, part); + part_dec_in_flight(part); - disk_stat_unlock(); + part_stat_unlock(); } if (req->end_io) diff --git a/block/blk-merge.c b/block/blk-merge.c index d926a24bf1f..c77196d5589 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -390,17 +390,13 @@ static int attempt_merge(struct request_queue *q, struct request *req, struct hd_struct *part; int cpu; - cpu = disk_stat_lock(); + cpu = part_stat_lock(); part = disk_map_sector_rcu(req->rq_disk, req->sector); - disk_round_stats(cpu, req->rq_disk); - req->rq_disk->in_flight--; - if (part) { - part_round_stats(cpu, part); - part->in_flight--; - } + part_round_stats(cpu, part); + part_dec_in_flight(part); - disk_stat_unlock(); + part_stat_unlock(); } req->ioprio = ioprio_best(req->ioprio, next->ioprio); diff --git a/block/genhd.c b/block/genhd.c index 06a252f2b96..e1cb96fb883 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -176,7 +176,7 @@ EXPORT_SYMBOL_GPL(disk_part_iter_exit); * while preemption is disabled. * * RETURNS: - * Found partition on success, NULL if there's no matching partition. + * Found partition on success, part0 is returned if no partition matches */ struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) { @@ -189,7 +189,7 @@ struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) sector < part->start_sect + part->nr_sects) return part; } - return NULL; + return &disk->part0; } EXPORT_SYMBOL_GPL(disk_map_sector_rcu); @@ -580,24 +580,24 @@ void __init printk_all_partitions(void) * numbers in hex - the same format as the root= * option takes. */ - printk("%s %10llu %s", - bdevt_str(disk_devt(disk), devt_buf), - (unsigned long long)get_capacity(disk) >> 1, - disk_name(disk, 0, name_buf)); - if (disk->driverfs_dev != NULL && - disk->driverfs_dev->driver != NULL) - printk(" driver: %s\n", - disk->driverfs_dev->driver->name); - else - printk(" (driver?)\n"); + disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); + while ((part = disk_part_iter_next(&piter))) { + bool is_part0 = part == &disk->part0; - /* now show the partitions */ - disk_part_iter_init(&piter, disk, 0); - while ((part = disk_part_iter_next(&piter))) - printk(" %s %10llu %s\n", + printk("%s%s %10llu %s", is_part0 ? "" : " ", bdevt_str(part_devt(part), devt_buf), (unsigned long long)part->nr_sects >> 1, disk_name(disk, part->partno, name_buf)); + if (is_part0) { + if (disk->driverfs_dev != NULL && + disk->driverfs_dev->driver != NULL) + printk(" driver: %s\n", + disk->driverfs_dev->driver->name); + else + printk(" (driver?)\n"); + } else + printk("\n"); + } disk_part_iter_exit(&piter); } class_dev_iter_exit(&iter); @@ -674,12 +674,7 @@ static int show_partition(struct seq_file *seqf, void *v) return 0; /* show the full disk and all non-0 size partitions of it */ - seq_printf(seqf, "%4d %7d %10llu %s\n", - MAJOR(disk_devt(sgp)), MINOR(disk_devt(sgp)), - (unsigned long long)get_capacity(sgp) >> 1, - disk_name(sgp, 0, buf)); - - disk_part_iter_init(&piter, sgp, 0); + disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0); while ((part = disk_part_iter_next(&piter))) seq_printf(seqf, "%4d %7d %10llu %s\n", MAJOR(part_devt(part)), MINOR(part_devt(part)), @@ -768,40 +763,13 @@ static ssize_t disk_capability_show(struct device *dev, return sprintf(buf, "%x\n", disk->flags); } -static ssize_t disk_stat_show(struct device *dev, - struct device_attribute *attr, char *buf) -{ - struct gendisk *disk = dev_to_disk(dev); - int cpu; - - cpu = disk_stat_lock(); - disk_round_stats(cpu, disk); - disk_stat_unlock(); - return sprintf(buf, - "%8lu %8lu %8llu %8u " - "%8lu %8lu %8llu %8u " - "%8u %8u %8u" - "\n", - disk_stat_read(disk, ios[READ]), - disk_stat_read(disk, merges[READ]), - (unsigned long long)disk_stat_read(disk, sectors[READ]), - jiffies_to_msecs(disk_stat_read(disk, ticks[READ])), - disk_stat_read(disk, ios[WRITE]), - disk_stat_read(disk, merges[WRITE]), - (unsigned long long)disk_stat_read(disk, sectors[WRITE]), - jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])), - disk->in_flight, - jiffies_to_msecs(disk_stat_read(disk, io_ticks)), - jiffies_to_msecs(disk_stat_read(disk, time_in_queue))); -} - static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL); static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL); static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL); -static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL); +static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); #ifdef CONFIG_FAIL_MAKE_REQUEST static struct device_attribute dev_attr_fail = __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); @@ -836,7 +804,7 @@ static void disk_release(struct device *dev) kfree(disk->random); kfree(disk->__part); - free_disk_stats(disk); + free_part_stats(&disk->part0); kfree(disk); } struct class block_class = { @@ -873,28 +841,11 @@ static int diskstats_show(struct seq_file *seqf, void *v) "\n\n"); */ - cpu = disk_stat_lock(); - disk_round_stats(cpu, gp); - disk_stat_unlock(); - seq_printf(seqf, "%4d %7d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n", - MAJOR(disk_devt(gp)), MINOR(disk_devt(gp)), - disk_name(gp, 0, buf), - disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]), - (unsigned long long)disk_stat_read(gp, sectors[0]), - jiffies_to_msecs(disk_stat_read(gp, ticks[0])), - disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]), - (unsigned long long)disk_stat_read(gp, sectors[1]), - jiffies_to_msecs(disk_stat_read(gp, ticks[1])), - gp->in_flight, - jiffies_to_msecs(disk_stat_read(gp, io_ticks)), - jiffies_to_msecs(disk_stat_read(gp, time_in_queue))); - - /* now show all non-0 size partitions of it */ - disk_part_iter_init(&piter, gp, 0); + disk_part_iter_init(&piter, gp, DISK_PITER_INCL_PART0); while ((hd = disk_part_iter_next(&piter))) { - cpu = disk_stat_lock(); + cpu = part_stat_lock(); part_round_stats(cpu, hd); - disk_stat_unlock(); + part_stat_unlock(); seq_printf(seqf, "%4d %7d %s %lu %lu %llu " "%u %lu %lu %llu %u %u %u %u\n", MAJOR(part_devt(hd)), MINOR(part_devt(hd)), @@ -1000,7 +951,7 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) int tot_minors = minors + ext_minors; int size = tot_minors * sizeof(struct hd_struct *); - if (!init_disk_stats(disk)) { + if (!init_part_stats(&disk->part0)) { kfree(disk); return NULL; } @@ -1008,7 +959,7 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) disk->__part = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, node_id); if (!disk->__part) { - free_disk_stats(disk); + free_part_stats(&disk->part0); kfree(disk); return NULL; } -- cgit v1.2.3 From 540eed5637b766bb1e881ef744c42617760b4815 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:15 +0900 Subject: block: make partition array dynamic disk->__part used to be statically allocated to the maximum possible number of partitions. This patch makes partition array allocation dynamic. The added overhead is minimal as only real change is one memory dereference changed to RCU one. This saves both a bit of memory and cpu cycles iterating through unoccupied slots and makes increasing partition limit easier. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 129 +++++++++++++++++++++++++++++++++++++++++++++++++--------- block/ioctl.c | 2 +- 2 files changed, 110 insertions(+), 21 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index e1cb96fb883..c2b14aa69d5 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -52,14 +52,21 @@ static struct device_type disk_type; */ struct hd_struct *disk_get_part(struct gendisk *disk, int partno) { - struct hd_struct *part; + struct hd_struct *part = NULL; + struct disk_part_tbl *ptbl; - if (unlikely(partno < 0 || partno >= disk_max_parts(disk))) + if (unlikely(partno < 0)) return NULL; + rcu_read_lock(); - part = rcu_dereference(disk->__part[partno]); - if (part) - get_device(part_to_dev(part)); + + ptbl = rcu_dereference(disk->part_tbl); + if (likely(partno < ptbl->len)) { + part = rcu_dereference(ptbl->part[partno]); + if (part) + get_device(part_to_dev(part)); + } + rcu_read_unlock(); return part; @@ -80,17 +87,24 @@ EXPORT_SYMBOL_GPL(disk_get_part); void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk, unsigned int flags) { + struct disk_part_tbl *ptbl; + + rcu_read_lock(); + ptbl = rcu_dereference(disk->part_tbl); + piter->disk = disk; piter->part = NULL; if (flags & DISK_PITER_REVERSE) - piter->idx = disk_max_parts(piter->disk) - 1; + piter->idx = ptbl->len - 1; else if (flags & DISK_PITER_INCL_PART0) piter->idx = 0; else piter->idx = 1; piter->flags = flags; + + rcu_read_unlock(); } EXPORT_SYMBOL_GPL(disk_part_iter_init); @@ -105,13 +119,16 @@ EXPORT_SYMBOL_GPL(disk_part_iter_init); */ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) { + struct disk_part_tbl *ptbl; int inc, end; /* put the last partition */ disk_put_part(piter->part); piter->part = NULL; + /* get part_tbl */ rcu_read_lock(); + ptbl = rcu_dereference(piter->disk->part_tbl); /* determine iteration parameters */ if (piter->flags & DISK_PITER_REVERSE) { @@ -122,14 +139,14 @@ struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter) end = 0; } else { inc = 1; - end = disk_max_parts(piter->disk); + end = ptbl->len; } /* iterate to the next partition */ for (; piter->idx != end; piter->idx += inc) { struct hd_struct *part; - part = rcu_dereference(piter->disk->__part[piter->idx]); + part = rcu_dereference(ptbl->part[piter->idx]); if (!part) continue; if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects) @@ -180,10 +197,13 @@ EXPORT_SYMBOL_GPL(disk_part_iter_exit); */ struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector) { + struct disk_part_tbl *ptbl; int i; - for (i = 1; i < disk_max_parts(disk); i++) { - struct hd_struct *part = rcu_dereference(disk->__part[i]); + ptbl = rcu_dereference(disk->part_tbl); + + for (i = 1; i < ptbl->len; i++) { + struct hd_struct *part = rcu_dereference(ptbl->part[i]); if (part && part->start_sect <= sector && sector < part->start_sect + part->nr_sects) @@ -798,12 +818,86 @@ static struct attribute_group *disk_attr_groups[] = { NULL }; +static void disk_free_ptbl_rcu_cb(struct rcu_head *head) +{ + struct disk_part_tbl *ptbl = + container_of(head, struct disk_part_tbl, rcu_head); + + kfree(ptbl); +} + +/** + * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way + * @disk: disk to replace part_tbl for + * @new_ptbl: new part_tbl to install + * + * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The + * original ptbl is freed using RCU callback. + * + * LOCKING: + * Matching bd_mutx locked. + */ +static void disk_replace_part_tbl(struct gendisk *disk, + struct disk_part_tbl *new_ptbl) +{ + struct disk_part_tbl *old_ptbl = disk->part_tbl; + + rcu_assign_pointer(disk->part_tbl, new_ptbl); + if (old_ptbl) + call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb); +} + +/** + * disk_expand_part_tbl - expand disk->part_tbl + * @disk: disk to expand part_tbl for + * @partno: expand such that this partno can fit in + * + * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl + * uses RCU to allow unlocked dereferencing for stats and other stuff. + * + * LOCKING: + * Matching bd_mutex locked, might sleep. + * + * RETURNS: + * 0 on success, -errno on failure. + */ +int disk_expand_part_tbl(struct gendisk *disk, int partno) +{ + struct disk_part_tbl *old_ptbl = disk->part_tbl; + struct disk_part_tbl *new_ptbl; + int len = old_ptbl ? old_ptbl->len : 0; + int target = partno + 1; + size_t size; + int i; + + /* disk_max_parts() is zero during initialization, ignore if so */ + if (disk_max_parts(disk) && target > disk_max_parts(disk)) + return -EINVAL; + + if (target <= len) + return 0; + + size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]); + new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id); + if (!new_ptbl) + return -ENOMEM; + + INIT_RCU_HEAD(&new_ptbl->rcu_head); + new_ptbl->len = target; + + for (i = 0; i < len; i++) + rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]); + + disk_replace_part_tbl(disk, new_ptbl); + return 0; +} + static void disk_release(struct device *dev) { struct gendisk *disk = dev_to_disk(dev); kfree(disk->random); - kfree(disk->__part); + disk_replace_part_tbl(disk, NULL); free_part_stats(&disk->part0); kfree(disk); } @@ -948,22 +1042,16 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL | __GFP_ZERO, node_id); if (disk) { - int tot_minors = minors + ext_minors; - int size = tot_minors * sizeof(struct hd_struct *); - if (!init_part_stats(&disk->part0)) { kfree(disk); return NULL; } - - disk->__part = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, - node_id); - if (!disk->__part) { - free_part_stats(&disk->part0); + if (disk_expand_part_tbl(disk, 0)) { + free_part_stats(&disk->part0); kfree(disk); return NULL; } - disk->__part[0] = &disk->part0; + disk->part_tbl->part[0] = &disk->part0; disk->minors = minors; disk->ext_minors = ext_minors; @@ -973,6 +1061,7 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) device_initialize(disk_to_dev(disk)); INIT_WORK(&disk->async_notify, media_change_notify_thread); + disk->node_id = node_id; } return disk; } diff --git a/block/ioctl.c b/block/ioctl.c index 64e7c67a64b..38bee321e1f 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -30,7 +30,7 @@ static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user if (bdev != bdev->bd_contains) return -EINVAL; partno = p.pno; - if (partno <= 0 || partno >= disk_max_parts(disk)) + if (partno <= 0) return -EINVAL; switch (a.op) { case BLKPG_ADD_PARTITION: -- cgit v1.2.3 From 689d6fac40b41c7bf154f362deaf442548e4dc81 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:16 +0900 Subject: block: replace @ext_minors with GENHD_FL_EXT_DEVT With previous changes, it's meaningless to limit the number of partitions. Replace @ext_minors with GENHD_FL_EXT_DEVT such that setting the flag allows the disk to have maximum number of allowed partitions (only limited by the number of entries in parsed_partitions as determined by MAX_PART constant). This kills not-too-pretty alloc_disk_ext[_node]() functions and makes @minors parameter to alloc_disk[_node]() unnecessary. The parameter is left alone to avoid disturbing the users. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 16 +--------------- 1 file changed, 1 insertion(+), 15 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index c2b14aa69d5..eedab5b4685 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -1024,18 +1024,9 @@ struct gendisk *alloc_disk(int minors) { return alloc_disk_node(minors, -1); } +EXPORT_SYMBOL(alloc_disk); struct gendisk *alloc_disk_node(int minors, int node_id) -{ - return alloc_disk_ext_node(minors, 0, node_id); -} - -struct gendisk *alloc_disk_ext(int minors, int ext_minors) -{ - return alloc_disk_ext_node(minors, ext_minors, -1); -} - -struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) { struct gendisk *disk; @@ -1054,7 +1045,6 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) disk->part_tbl->part[0] = &disk->part0; disk->minors = minors; - disk->ext_minors = ext_minors; rand_initialize_disk(disk); disk_to_dev(disk)->class = &block_class; disk_to_dev(disk)->type = &disk_type; @@ -1065,11 +1055,7 @@ struct gendisk *alloc_disk_ext_node(int minors, int ext_minors, int node_id) } return disk; } - -EXPORT_SYMBOL(alloc_disk); EXPORT_SYMBOL(alloc_disk_node); -EXPORT_SYMBOL(alloc_disk_ext); -EXPORT_SYMBOL(alloc_disk_ext_node); struct kobject *get_disk(struct gendisk *disk) { -- cgit v1.2.3 From 3e1a7ff8a0a7b948f2684930166954f9e8e776fe Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 25 Aug 2008 19:56:17 +0900 Subject: block: allow disk to have extended device number Now that disk and partition handlings are mostly unified, it's easy to allow disk to have extended device number. This patch makes add_disk() use extended device number if disk->minors is zero. Both sd and ide-disk are updated to use this. * sd_format_disk_name() is implemented which can generically determine the drive name. This removes disk number restriction stemming from limited device names. * If sd index goes over SD_MAX_DISKS (which can be increased now BTW), sd simply doesn't initialize minors letting block layer choose extended device number. * If CONFIG_DEBUG_EXT_DEVT is set, both sd and ide-disk always set minors to 0 and use extended device numbers. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 25 ++++++++++++++++++++++++- 1 file changed, 24 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index eedab5b4685..d9de3e482d1 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -478,14 +478,37 @@ static int exact_lock(dev_t devt, void *data) * * This function registers the partitioning information in @disk * with the kernel. + * + * FIXME: error handling */ void add_disk(struct gendisk *disk) { struct backing_dev_info *bdi; + dev_t devt; int retval; + /* minors == 0 indicates to use ext devt from part0 and should + * be accompanied with EXT_DEVT flag. Make sure all + * parameters make sense. + */ + WARN_ON(disk->minors && !(disk->major || disk->first_minor)); + WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); + disk->flags |= GENHD_FL_UP; - disk_to_dev(disk)->devt = MKDEV(disk->major, disk->first_minor); + + retval = blk_alloc_devt(&disk->part0, &devt); + if (retval) { + WARN_ON(1); + return; + } + disk_to_dev(disk)->devt = devt; + + /* ->major and ->first_minor aren't supposed to be + * dereferenced from here on, but set them just in case. + */ + disk->major = MAJOR(devt); + disk->first_minor = MINOR(devt); + blk_register_region(disk_devt(disk), disk->minors, NULL, exact_match, exact_lock, disk); register_disk(disk); -- cgit v1.2.3 From 0835da67c11e879ed5dc23160934d8970470a2ce Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 26 Aug 2008 09:15:47 +0200 Subject: block: use linux/uaccess.h in elevator.c instead of asm variant Signed-off-by: Jens Axboe --- block/elevator.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'block') diff --git a/block/elevator.c b/block/elevator.c index 269615e6dbf..8e3fc3afc77 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -34,8 +34,7 @@ #include #include #include - -#include +#include static DEFINE_SPINLOCK(elv_list_lock); static LIST_HEAD(elv_list); -- cgit v1.2.3 From b646fc59b332ef307895558c9cd1359dc2d25813 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 28 Jul 2008 13:06:00 +0200 Subject: block: split softirq handling into blk-softirq.c Signed-off-by: Jens Axboe --- block/Makefile | 4 +- block/blk-core.c | 88 -------------------------------------------- block/blk-softirq.c | 103 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 105 insertions(+), 90 deletions(-) create mode 100644 block/blk-softirq.c (limited to 'block') diff --git a/block/Makefile b/block/Makefile index 208000b0750..0da976ce67d 100644 --- a/block/Makefile +++ b/block/Makefile @@ -4,8 +4,8 @@ obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \ blk-barrier.o blk-settings.o blk-ioc.o blk-map.o \ - blk-exec.o blk-merge.o ioctl.o genhd.o scsi_ioctl.o \ - cmd-filter.o + blk-exec.o blk-merge.o blk-softirq.o ioctl.o genhd.o \ + scsi_ioctl.o cmd-filter.o obj-$(CONFIG_BLK_DEV_BSG) += bsg.o obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o diff --git a/block/blk-core.c b/block/blk-core.c index 98138f00252..527b3382a61 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -26,8 +26,6 @@ #include #include #include -#include -#include #include #include @@ -50,8 +48,6 @@ struct kmem_cache *blk_requestq_cachep; */ static struct workqueue_struct *kblockd_workqueue; -static DEFINE_PER_CPU(struct list_head, blk_cpu_done); - static void drive_stat_acct(struct request *rq, int new_io) { struct hd_struct *part; @@ -1643,82 +1639,6 @@ static int __end_that_request_first(struct request *req, int error, return 1; } -/* - * splice the completion data to a local structure and hand off to - * process_completion_queue() to complete the requests - */ -static void blk_done_softirq(struct softirq_action *h) -{ - struct list_head *cpu_list, local_list; - - local_irq_disable(); - cpu_list = &__get_cpu_var(blk_cpu_done); - list_replace_init(cpu_list, &local_list); - local_irq_enable(); - - while (!list_empty(&local_list)) { - struct request *rq; - - rq = list_entry(local_list.next, struct request, donelist); - list_del_init(&rq->donelist); - rq->q->softirq_done_fn(rq); - } -} - -static int __cpuinit blk_cpu_notify(struct notifier_block *self, - unsigned long action, void *hcpu) -{ - /* - * If a CPU goes away, splice its entries to the current CPU - * and trigger a run of the softirq - */ - if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) { - int cpu = (unsigned long) hcpu; - - local_irq_disable(); - list_splice_init(&per_cpu(blk_cpu_done, cpu), - &__get_cpu_var(blk_cpu_done)); - raise_softirq_irqoff(BLOCK_SOFTIRQ); - local_irq_enable(); - } - - return NOTIFY_OK; -} - - -static struct notifier_block blk_cpu_notifier __cpuinitdata = { - .notifier_call = blk_cpu_notify, -}; - -/** - * blk_complete_request - end I/O on a request - * @req: the request being processed - * - * Description: - * Ends all I/O on a request. It does not handle partial completions, - * unless the driver actually implements this in its completion callback - * through requeueing. The actual completion happens out-of-order, - * through a softirq handler. The user must have registered a completion - * callback through blk_queue_softirq_done(). - **/ - -void blk_complete_request(struct request *req) -{ - struct list_head *cpu_list; - unsigned long flags; - - BUG_ON(!req->q->softirq_done_fn); - - local_irq_save(flags); - - cpu_list = &__get_cpu_var(blk_cpu_done); - list_add_tail(&req->donelist, cpu_list); - raise_softirq_irqoff(BLOCK_SOFTIRQ); - - local_irq_restore(flags); -} -EXPORT_SYMBOL(blk_complete_request); - /* * queue lock must be held */ @@ -2053,8 +1973,6 @@ EXPORT_SYMBOL(kblockd_flush_work); int __init blk_dev_init(void) { - int i; - kblockd_workqueue = create_workqueue("kblockd"); if (!kblockd_workqueue) panic("Failed to create kblockd\n"); @@ -2065,12 +1983,6 @@ int __init blk_dev_init(void) blk_requestq_cachep = kmem_cache_create("blkdev_queue", sizeof(struct request_queue), 0, SLAB_PANIC, NULL); - for_each_possible_cpu(i) - INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i)); - - open_softirq(BLOCK_SOFTIRQ, blk_done_softirq); - register_hotcpu_notifier(&blk_cpu_notifier); - return 0; } diff --git a/block/blk-softirq.c b/block/blk-softirq.c new file mode 100644 index 00000000000..9e1c43bff66 --- /dev/null +++ b/block/blk-softirq.c @@ -0,0 +1,103 @@ +/* + * Functions related to softirq rq completions + */ +#include +#include +#include +#include +#include +#include +#include + +#include "blk.h" + +static DEFINE_PER_CPU(struct list_head, blk_cpu_done); + +static int __cpuinit blk_cpu_notify(struct notifier_block *self, + unsigned long action, void *hcpu) +{ + /* + * If a CPU goes away, splice its entries to the current CPU + * and trigger a run of the softirq + */ + if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) { + int cpu = (unsigned long) hcpu; + + local_irq_disable(); + list_splice_init(&per_cpu(blk_cpu_done, cpu), + &__get_cpu_var(blk_cpu_done)); + raise_softirq_irqoff(BLOCK_SOFTIRQ); + local_irq_enable(); + } + + return NOTIFY_OK; +} + + +static struct notifier_block blk_cpu_notifier __cpuinitdata = { + .notifier_call = blk_cpu_notify, +}; + +/* + * splice the completion data to a local structure and hand off to + * process_completion_queue() to complete the requests + */ +static void blk_done_softirq(struct softirq_action *h) +{ + struct list_head *cpu_list, local_list; + + local_irq_disable(); + cpu_list = &__get_cpu_var(blk_cpu_done); + list_replace_init(cpu_list, &local_list); + local_irq_enable(); + + while (!list_empty(&local_list)) { + struct request *rq; + + rq = list_entry(local_list.next, struct request, donelist); + list_del_init(&rq->donelist); + rq->q->softirq_done_fn(rq); + } +} + +/** + * blk_complete_request - end I/O on a request + * @req: the request being processed + * + * Description: + * Ends all I/O on a request. It does not handle partial completions, + * unless the driver actually implements this in its completion callback + * through requeueing. The actual completion happens out-of-order, + * through a softirq handler. The user must have registered a completion + * callback through blk_queue_softirq_done(). + **/ + +void blk_complete_request(struct request *req) +{ + struct list_head *cpu_list; + unsigned long flags; + + BUG_ON(!req->q->softirq_done_fn); + + local_irq_save(flags); + + cpu_list = &__get_cpu_var(blk_cpu_done); + list_add_tail(&req->donelist, cpu_list); + raise_softirq_irqoff(BLOCK_SOFTIRQ); + + local_irq_restore(flags); +} +EXPORT_SYMBOL(blk_complete_request); + +int __init blk_softirq_init(void) +{ + int i; + + for_each_possible_cpu(i) + INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i)); + + open_softirq(BLOCK_SOFTIRQ, blk_done_softirq); + register_hotcpu_notifier(&blk_cpu_notifier); + return 0; +} +subsys_initcall(blk_softirq_init); -- cgit v1.2.3 From 18887ad910e56066233a07fd3cfb2fa11338b782 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 28 Jul 2008 13:08:45 +0200 Subject: block: make kblockd_schedule_work() take the queue as parameter Preparatory patch for checking queuing affinity. Signed-off-by: Jens Axboe --- block/as-iosched.c | 6 +++--- block/blk-core.c | 8 ++++---- block/cfq-iosched.c | 2 +- 3 files changed, 8 insertions(+), 8 deletions(-) (limited to 'block') diff --git a/block/as-iosched.c b/block/as-iosched.c index cf4eb0eefbb..80af9257e64 100644 --- a/block/as-iosched.c +++ b/block/as-iosched.c @@ -462,7 +462,7 @@ static void as_antic_stop(struct as_data *ad) del_timer(&ad->antic_timer); ad->antic_status = ANTIC_FINISHED; /* see as_work_handler */ - kblockd_schedule_work(&ad->antic_work); + kblockd_schedule_work(ad->q, &ad->antic_work); } } @@ -483,7 +483,7 @@ static void as_antic_timeout(unsigned long data) aic = ad->io_context->aic; ad->antic_status = ANTIC_FINISHED; - kblockd_schedule_work(&ad->antic_work); + kblockd_schedule_work(q, &ad->antic_work); if (aic->ttime_samples == 0) { /* process anticipated on has exited or timed out*/ @@ -844,7 +844,7 @@ static void as_completed_request(struct request_queue *q, struct request *rq) if (ad->changed_batch && ad->nr_dispatched == 1) { ad->current_batch_expires = jiffies + ad->batch_expire[ad->batch_data_dir]; - kblockd_schedule_work(&ad->antic_work); + kblockd_schedule_work(q, &ad->antic_work); ad->changed_batch = 0; if (ad->batch_data_dir == REQ_SYNC) diff --git a/block/blk-core.c b/block/blk-core.c index 527b3382a61..9c6f818d0c3 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -305,7 +305,7 @@ void blk_unplug_timeout(unsigned long data) blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_TIMER, NULL, q->rq.count[READ] + q->rq.count[WRITE]); - kblockd_schedule_work(&q->unplug_work); + kblockd_schedule_work(q, &q->unplug_work); } void blk_unplug(struct request_queue *q) @@ -346,7 +346,7 @@ void blk_start_queue(struct request_queue *q) queue_flag_clear(QUEUE_FLAG_REENTER, q); } else { blk_plug_device(q); - kblockd_schedule_work(&q->unplug_work); + kblockd_schedule_work(q, &q->unplug_work); } } EXPORT_SYMBOL(blk_start_queue); @@ -411,7 +411,7 @@ void __blk_run_queue(struct request_queue *q) queue_flag_clear(QUEUE_FLAG_REENTER, q); } else { blk_plug_device(q); - kblockd_schedule_work(&q->unplug_work); + kblockd_schedule_work(q, &q->unplug_work); } } } @@ -1959,7 +1959,7 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq, rq->rq_disk = bio->bi_bdev->bd_disk; } -int kblockd_schedule_work(struct work_struct *work) +int kblockd_schedule_work(struct request_queue *q, struct work_struct *work) { return queue_work(kblockd_workqueue, work); } diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 1e2aff812ee..5f6fd287c18 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -244,7 +244,7 @@ static inline void cfq_schedule_dispatch(struct cfq_data *cfqd) { if (cfqd->busy_queues) { cfq_log(cfqd, "schedule dispatch"); - kblockd_schedule_work(&cfqd->unplug_work); + kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work); } } -- cgit v1.2.3 From c7c22e4d5c1fdebfac4dba76de7d0338c2b0d832 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Sat, 13 Sep 2008 20:26:01 +0200 Subject: block: add support for IO CPU affinity This patch adds support for controlling the IO completion CPU of either all requests on a queue, or on a per-request basis. We export a sysfs variable (rq_affinity) which, if set, migrates completions of requests to the CPU that originally submitted it. A bio helper (bio_set_completion_cpu()) is also added, so that queuers can ask for completion on that specific CPU. In testing, this has been show to cut the system time by as much as 20-40% on synthetic workloads where CPU affinity is desired. This requires a little help from the architecture, so it'll only work as designed for archs that are using the new generic smp helper infrastructure. Signed-off-by: Jens Axboe --- block/blk-core.c | 46 +++++++++---------- block/blk-settings.c | 2 +- block/blk-softirq.c | 126 ++++++++++++++++++++++++++++++++++++++------------- block/blk-sysfs.c | 31 +++++++++++++ block/blk.h | 12 +++++ 5 files changed, 162 insertions(+), 55 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 9c6f818d0c3..5484838f46e 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -110,7 +110,7 @@ void blk_rq_init(struct request_queue *q, struct request *rq) memset(rq, 0, sizeof(*rq)); INIT_LIST_HEAD(&rq->queuelist); - INIT_LIST_HEAD(&rq->donelist); + rq->cpu = -1; rq->q = q; rq->sector = rq->hard_sector = (sector_t) -1; INIT_HLIST_NODE(&rq->hash); @@ -322,6 +322,21 @@ void blk_unplug(struct request_queue *q) } EXPORT_SYMBOL(blk_unplug); +static void blk_invoke_request_fn(struct request_queue *q) +{ + /* + * one level of recursion is ok and is much faster than kicking + * the unplug handling + */ + if (!queue_flag_test_and_set(QUEUE_FLAG_REENTER, q)) { + q->request_fn(q); + queue_flag_clear(QUEUE_FLAG_REENTER, q); + } else { + queue_flag_set(QUEUE_FLAG_PLUGGED, q); + kblockd_schedule_work(q, &q->unplug_work); + } +} + /** * blk_start_queue - restart a previously stopped queue * @q: The &struct request_queue in question @@ -336,18 +351,7 @@ void blk_start_queue(struct request_queue *q) WARN_ON(!irqs_disabled()); queue_flag_clear(QUEUE_FLAG_STOPPED, q); - - /* - * one level of recursion is ok and is much faster than kicking - * the unplug handling - */ - if (!queue_flag_test_and_set(QUEUE_FLAG_REENTER, q)) { - q->request_fn(q); - queue_flag_clear(QUEUE_FLAG_REENTER, q); - } else { - blk_plug_device(q); - kblockd_schedule_work(q, &q->unplug_work); - } + blk_invoke_request_fn(q); } EXPORT_SYMBOL(blk_start_queue); @@ -405,15 +409,8 @@ void __blk_run_queue(struct request_queue *q) * Only recurse once to avoid overrunning the stack, let the unplug * handling reinvoke the handler shortly if we already got there. */ - if (!elv_queue_empty(q)) { - if (!queue_flag_test_and_set(QUEUE_FLAG_REENTER, q)) { - q->request_fn(q); - queue_flag_clear(QUEUE_FLAG_REENTER, q); - } else { - blk_plug_device(q); - kblockd_schedule_work(q, &q->unplug_work); - } - } + if (!elv_queue_empty(q)) + blk_invoke_request_fn(q); } EXPORT_SYMBOL(__blk_run_queue); @@ -1056,6 +1053,7 @@ EXPORT_SYMBOL(blk_put_request); void init_request_from_bio(struct request *req, struct bio *bio) { + req->cpu = bio->bi_comp_cpu; req->cmd_type = REQ_TYPE_FS; /* @@ -1198,13 +1196,15 @@ get_rq: init_request_from_bio(req, bio); spin_lock_irq(q->queue_lock); + if (test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags) || + bio_flagged(bio, BIO_CPU_AFFINE)) + req->cpu = blk_cpu_to_group(smp_processor_id()); if (elv_queue_empty(q)) blk_plug_device(q); add_request(q, req); out: if (sync) __generic_unplug_device(q); - spin_unlock_irq(q->queue_lock); return 0; diff --git a/block/blk-settings.c b/block/blk-settings.c index d70692badcd..a60e959a12c 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -443,7 +443,7 @@ void blk_queue_update_dma_alignment(struct request_queue *q, int mask) } EXPORT_SYMBOL(blk_queue_update_dma_alignment); -static int __init blk_settings_init(void) +int __init blk_settings_init(void) { blk_max_low_pfn = max_low_pfn - 1; blk_max_pfn = max_pfn - 1; diff --git a/block/blk-softirq.c b/block/blk-softirq.c index 9e1c43bff66..3a1af551191 100644 --- a/block/blk-softirq.c +++ b/block/blk-softirq.c @@ -13,6 +13,70 @@ static DEFINE_PER_CPU(struct list_head, blk_cpu_done); +/* + * Softirq action handler - move entries to local list and loop over them + * while passing them to the queue registered handler. + */ +static void blk_done_softirq(struct softirq_action *h) +{ + struct list_head *cpu_list, local_list; + + local_irq_disable(); + cpu_list = &__get_cpu_var(blk_cpu_done); + list_replace_init(cpu_list, &local_list); + local_irq_enable(); + + while (!list_empty(&local_list)) { + struct request *rq; + + rq = list_entry(local_list.next, struct request, csd.list); + list_del_init(&rq->csd.list); + rq->q->softirq_done_fn(rq); + } +} + +#if defined(CONFIG_SMP) && defined(CONFIG_USE_GENERIC_SMP_HELPERS) +static void trigger_softirq(void *data) +{ + struct request *rq = data; + unsigned long flags; + struct list_head *list; + + local_irq_save(flags); + list = &__get_cpu_var(blk_cpu_done); + list_add_tail(&rq->csd.list, list); + + if (list->next == &rq->csd.list) + raise_softirq_irqoff(BLOCK_SOFTIRQ); + + local_irq_restore(flags); +} + +/* + * Setup and invoke a run of 'trigger_softirq' on the given cpu. + */ +static int raise_blk_irq(int cpu, struct request *rq) +{ + if (cpu_online(cpu)) { + struct call_single_data *data = &rq->csd; + + data->func = trigger_softirq; + data->info = rq; + data->flags = 0; + + __smp_call_function_single(cpu, data); + return 0; + } + + return 1; +} +#else /* CONFIG_SMP && CONFIG_USE_GENERIC_SMP_HELPERS */ +static int raise_blk_irq(int cpu, struct request *rq) +{ + return 1; +} +#endif + static int __cpuinit blk_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) { @@ -33,33 +97,10 @@ static int __cpuinit blk_cpu_notify(struct notifier_block *self, return NOTIFY_OK; } - -static struct notifier_block blk_cpu_notifier __cpuinitdata = { +static struct notifier_block __cpuinitdata blk_cpu_notifier = { .notifier_call = blk_cpu_notify, }; -/* - * splice the completion data to a local structure and hand off to - * process_completion_queue() to complete the requests - */ -static void blk_done_softirq(struct softirq_action *h) -{ - struct list_head *cpu_list, local_list; - - local_irq_disable(); - cpu_list = &__get_cpu_var(blk_cpu_done); - list_replace_init(cpu_list, &local_list); - local_irq_enable(); - - while (!list_empty(&local_list)) { - struct request *rq; - - rq = list_entry(local_list.next, struct request, donelist); - list_del_init(&rq->donelist); - rq->q->softirq_done_fn(rq); - } -} - /** * blk_complete_request - end I/O on a request * @req: the request being processed @@ -71,25 +112,48 @@ static void blk_done_softirq(struct softirq_action *h) * through a softirq handler. The user must have registered a completion * callback through blk_queue_softirq_done(). **/ - void blk_complete_request(struct request *req) { - struct list_head *cpu_list; + struct request_queue *q = req->q; unsigned long flags; + int ccpu, cpu, group_cpu; - BUG_ON(!req->q->softirq_done_fn); + BUG_ON(!q->softirq_done_fn); local_irq_save(flags); + cpu = smp_processor_id(); + group_cpu = blk_cpu_to_group(cpu); - cpu_list = &__get_cpu_var(blk_cpu_done); - list_add_tail(&req->donelist, cpu_list); - raise_softirq_irqoff(BLOCK_SOFTIRQ); + /* + * Select completion CPU + */ + if (test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags) && req->cpu != -1) + ccpu = req->cpu; + else + ccpu = cpu; + + if (ccpu == cpu || ccpu == group_cpu) { + struct list_head *list; +do_local: + list = &__get_cpu_var(blk_cpu_done); + list_add_tail(&req->csd.list, list); + + /* + * if the list only contains our just added request, + * signal a raise of the softirq. If there are already + * entries there, someone already raised the irq but it + * hasn't run yet. + */ + if (list->next == &req->csd.list) + raise_softirq_irqoff(BLOCK_SOFTIRQ); + } else if (raise_blk_irq(ccpu, req)) + goto do_local; local_irq_restore(flags); } EXPORT_SYMBOL(blk_complete_request); -int __init blk_softirq_init(void) +__init int blk_softirq_init(void) { int i; diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index b9a6ed16664..21e275d7eed 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -156,6 +156,30 @@ static ssize_t queue_nomerges_store(struct request_queue *q, const char *page, return ret; } +static ssize_t queue_rq_affinity_show(struct request_queue *q, char *page) +{ + unsigned int set = test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags); + + return queue_var_show(set != 0, page); +} + +static ssize_t +queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count) +{ + ssize_t ret = -EINVAL; +#if defined(CONFIG_USE_GENERIC_SMP_HELPERS) + unsigned long val; + + ret = queue_var_store(&val, page, count); + spin_lock_irq(q->queue_lock); + if (val) + queue_flag_set(QUEUE_FLAG_SAME_COMP, q); + else + queue_flag_clear(QUEUE_FLAG_SAME_COMP, q); + spin_unlock_irq(q->queue_lock); +#endif + return ret; +} static struct queue_sysfs_entry queue_requests_entry = { .attr = {.name = "nr_requests", .mode = S_IRUGO | S_IWUSR }, @@ -197,6 +221,12 @@ static struct queue_sysfs_entry queue_nomerges_entry = { .store = queue_nomerges_store, }; +static struct queue_sysfs_entry queue_rq_affinity_entry = { + .attr = {.name = "rq_affinity", .mode = S_IRUGO | S_IWUSR }, + .show = queue_rq_affinity_show, + .store = queue_rq_affinity_store, +}; + static struct attribute *default_attrs[] = { &queue_requests_entry.attr, &queue_ra_entry.attr, @@ -205,6 +235,7 @@ static struct attribute *default_attrs[] = { &queue_iosched_entry.attr, &queue_hw_sector_size_entry.attr, &queue_nomerges_entry.attr, + &queue_rq_affinity_entry.attr, NULL, }; diff --git a/block/blk.h b/block/blk.h index c79f30e1df5..de74254cb91 100644 --- a/block/blk.h +++ b/block/blk.h @@ -59,4 +59,16 @@ static inline int queue_congestion_off_threshold(struct request_queue *q) #endif /* BLK_DEV_INTEGRITY */ +static inline int blk_cpu_to_group(int cpu) +{ +#ifdef CONFIG_SCHED_MC + cpumask_t mask = cpu_coregroup_map(cpu); + return first_cpu(mask); +#elif defined(CONFIG_SCHED_SMT) + return first_cpu(per_cpu(cpu_sibling_map, cpu)); +#else + return cpu; +#endif +} + #endif -- cgit v1.2.3 From ab780f1ece0dc8d5e8e8e85435acc5e4747ccda3 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 26 Aug 2008 10:25:02 +0200 Subject: block: inherit CPU completion on bio->rq and rq->rq merges Somewhat incomplete, as we do allow merges of requests and bios that have different completion CPUs given. This is done on the assumption that a larger IO is still more beneficial than CPU locality. Signed-off-by: Jens Axboe --- block/blk-core.c | 4 ++++ block/blk-merge.c | 2 ++ 2 files changed, 6 insertions(+) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 5484838f46e..b9a252cae4d 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1134,6 +1134,8 @@ static int __make_request(struct request_queue *q, struct bio *bio) req->biotail = bio; req->nr_sectors = req->hard_nr_sectors += nr_sectors; req->ioprio = ioprio_best(req->ioprio, prio); + if (!blk_rq_cpu_valid(req)) + req->cpu = bio->bi_comp_cpu; drive_stat_acct(req, 0); if (!attempt_back_merge(q, req)) elv_merged_request(q, req, el_ret); @@ -1161,6 +1163,8 @@ static int __make_request(struct request_queue *q, struct bio *bio) req->sector = req->hard_sector = bio->bi_sector; req->nr_sectors = req->hard_nr_sectors += nr_sectors; req->ioprio = ioprio_best(req->ioprio, prio); + if (!blk_rq_cpu_valid(req)) + req->cpu = bio->bi_comp_cpu; drive_stat_acct(req, 0); if (!attempt_front_merge(q, req)) elv_merged_request(q, req, el_ret); diff --git a/block/blk-merge.c b/block/blk-merge.c index c77196d5589..908d3e11ac5 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -400,6 +400,8 @@ static int attempt_merge(struct request_queue *q, struct request *req, } req->ioprio = ioprio_best(req->ioprio, next->ioprio); + if (blk_rq_cpu_valid(next)) + req->cpu = next->cpu; __blk_put_request(q, next); return 1; -- cgit v1.2.3 From 605401618ce4409045bc4db86e88d4b38f2ad585 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 26 Aug 2008 13:34:34 +0200 Subject: block: don't use bio_has_data() in the completion path We should just check for rq->bio, as that is really the information we are looking for. Even if the bio attached doesn't carry data, we still need to do IO post processing on it. Signed-off-by: Jens Axboe --- block/blk-core.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index b9a252cae4d..5bf806adc77 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1817,7 +1817,7 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, struct request_queue *q = rq->q; unsigned long flags = 0UL; - if (bio_has_data(rq->bio) || blk_discard_rq(rq)) { + if (rq->bio) { if (__end_that_request_first(rq, error, nr_bytes)) return 1; @@ -1875,8 +1875,7 @@ EXPORT_SYMBOL_GPL(blk_end_request); **/ int __blk_end_request(struct request *rq, int error, unsigned int nr_bytes) { - if ((bio_has_data(rq->bio) || blk_discard_rq(rq)) && - __end_that_request_first(rq, error, nr_bytes)) + if (rq->bio && __end_that_request_first(rq, error, nr_bytes)) return 1; add_disk_randomness(rq->rq_disk); -- cgit v1.2.3 From 45333d5a31296d0af886d94f1d08f128231cab8e Mon Sep 17 00:00:00 2001 From: Aaron Carroll Date: Tue, 26 Aug 2008 15:52:36 +0200 Subject: cfq-iosched: fix queue depth detection CFQ's detection of queueing devices assumes a non-queuing device and detects if the queue depth reaches a certain threshold. Under some workloads (e.g. synchronous reads), CFQ effectively forces a unit queue depth, thus defeating the detection logic. This leads to poor performance on queuing hardware, since the idle window remains enabled. This patch inverts the sense of the logic: assume a queuing-capable device, and detect if the depth does not exceed the threshold. Signed-off-by: Aaron Carroll Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 47 ++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 38 insertions(+), 9 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 5f6fd287c18..494b6fdcb18 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -39,6 +39,7 @@ static int cfq_slice_idle = HZ / 125; #define CFQ_MIN_TT (2) #define CFQ_SLICE_SCALE (5) +#define CFQ_HW_QUEUE_MIN (5) #define RQ_CIC(rq) \ ((struct cfq_io_context *) (rq)->elevator_private) @@ -86,7 +87,14 @@ struct cfq_data { int rq_in_driver; int sync_flight; + + /* + * queue-depth detection + */ + int rq_queued; int hw_tag; + int hw_tag_samples; + int rq_in_driver_peak; /* * idle window management @@ -654,15 +662,6 @@ static void cfq_activate_request(struct request_queue *q, struct request *rq) cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d", cfqd->rq_in_driver); - /* - * If the depth is larger 1, it really could be queueing. But lets - * make the mark a little higher - idling could still be good for - * low queueing, and a low queueing number could also just indicate - * a SCSI mid layer like behaviour where limit+1 is often seen. - */ - if (!cfqd->hw_tag && cfqd->rq_in_driver > 4) - cfqd->hw_tag = 1; - cfqd->last_position = rq->hard_sector + rq->hard_nr_sectors; } @@ -686,6 +685,7 @@ static void cfq_remove_request(struct request *rq) list_del_init(&rq->queuelist); cfq_del_rq_rb(rq); + cfqq->cfqd->rq_queued--; if (rq_is_meta(rq)) { WARN_ON(!cfqq->meta_pending); cfqq->meta_pending--; @@ -1833,6 +1833,7 @@ cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq, { struct cfq_io_context *cic = RQ_CIC(rq); + cfqd->rq_queued++; if (rq_is_meta(rq)) cfqq->meta_pending++; @@ -1880,6 +1881,31 @@ static void cfq_insert_request(struct request_queue *q, struct request *rq) cfq_rq_enqueued(cfqd, cfqq, rq); } +/* + * Update hw_tag based on peak queue depth over 50 samples under + * sufficient load. + */ +static void cfq_update_hw_tag(struct cfq_data *cfqd) +{ + if (cfqd->rq_in_driver > cfqd->rq_in_driver_peak) + cfqd->rq_in_driver_peak = cfqd->rq_in_driver; + + if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN && + cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN) + return; + + if (cfqd->hw_tag_samples++ < 50) + return; + + if (cfqd->rq_in_driver_peak >= CFQ_HW_QUEUE_MIN) + cfqd->hw_tag = 1; + else + cfqd->hw_tag = 0; + + cfqd->hw_tag_samples = 0; + cfqd->rq_in_driver_peak = 0; +} + static void cfq_completed_request(struct request_queue *q, struct request *rq) { struct cfq_queue *cfqq = RQ_CFQQ(rq); @@ -1890,6 +1916,8 @@ static void cfq_completed_request(struct request_queue *q, struct request *rq) now = jiffies; cfq_log_cfqq(cfqd, cfqq, "complete"); + cfq_update_hw_tag(cfqd); + WARN_ON(!cfqd->rq_in_driver); WARN_ON(!cfqq->dispatched); cfqd->rq_in_driver--; @@ -2200,6 +2228,7 @@ static void *cfq_init_queue(struct request_queue *q) cfqd->cfq_slice[1] = cfq_slice_sync; cfqd->cfq_slice_async_rq = cfq_slice_async_rq; cfqd->cfq_slice_idle = cfq_slice_idle; + cfqd->hw_tag = 1; return cfqd; } -- cgit v1.2.3 From a3bce90edd8f6cafe3f63b1a943800792e830178 Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Thu, 28 Aug 2008 16:17:05 +0900 Subject: block: add gfp_mask argument to blk_rq_map_user and blk_rq_map_user_iov Currently, blk_rq_map_user and blk_rq_map_user_iov always do GFP_KERNEL allocation. This adds gfp_mask argument to blk_rq_map_user and blk_rq_map_user_iov so sg can use it (sg always does GFP_ATOMIC allocation). Signed-off-by: FUJITA Tomonori Signed-off-by: Douglas Gilbert Cc: Mike Christie Cc: James Bottomley Signed-off-by: Jens Axboe --- block/blk-map.c | 20 ++++++++++++-------- block/bsg.c | 5 +++-- block/scsi_ioctl.c | 5 +++-- 3 files changed, 18 insertions(+), 12 deletions(-) (limited to 'block') diff --git a/block/blk-map.c b/block/blk-map.c index ea1bf53929e..ac21b7397e1 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -41,7 +41,8 @@ static int __blk_rq_unmap_user(struct bio *bio) } static int __blk_rq_map_user(struct request_queue *q, struct request *rq, - void __user *ubuf, unsigned int len) + void __user *ubuf, unsigned int len, + gfp_t gfp_mask) { unsigned long uaddr; unsigned int alignment; @@ -57,9 +58,9 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, uaddr = (unsigned long) ubuf; alignment = queue_dma_alignment(q) | q->dma_pad_mask; if (!(uaddr & alignment) && !(len & alignment)) - bio = bio_map_user(q, NULL, uaddr, len, reading); + bio = bio_map_user(q, NULL, uaddr, len, reading, gfp_mask); else - bio = bio_copy_user(q, uaddr, len, reading); + bio = bio_copy_user(q, uaddr, len, reading, gfp_mask); if (IS_ERR(bio)) return PTR_ERR(bio); @@ -90,6 +91,7 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, * @rq: request structure to fill * @ubuf: the user buffer * @len: length of user data + * @gfp_mask: memory allocation flags * * Description: * Data will be mapped directly for zero copy I/O, if possible. Otherwise @@ -105,7 +107,7 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, * unmapping. */ int blk_rq_map_user(struct request_queue *q, struct request *rq, - void __user *ubuf, unsigned long len) + void __user *ubuf, unsigned long len, gfp_t gfp_mask) { unsigned long bytes_read = 0; struct bio *bio = NULL; @@ -132,7 +134,7 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq, if (end - start > BIO_MAX_PAGES) map_len -= PAGE_SIZE; - ret = __blk_rq_map_user(q, rq, ubuf, map_len); + ret = __blk_rq_map_user(q, rq, ubuf, map_len, gfp_mask); if (ret < 0) goto unmap_rq; if (!bio) @@ -160,6 +162,7 @@ EXPORT_SYMBOL(blk_rq_map_user); * @iov: pointer to the iovec * @iov_count: number of elements in the iovec * @len: I/O byte count + * @gfp_mask: memory allocation flags * * Description: * Data will be mapped directly for zero copy I/O, if possible. Otherwise @@ -175,7 +178,8 @@ EXPORT_SYMBOL(blk_rq_map_user); * unmapping. */ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, - struct sg_iovec *iov, int iov_count, unsigned int len) + struct sg_iovec *iov, int iov_count, unsigned int len, + gfp_t gfp_mask) { struct bio *bio; int i, read = rq_data_dir(rq) == READ; @@ -194,9 +198,9 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, } if (unaligned || (q->dma_pad_mask & len)) - bio = bio_copy_user_iov(q, iov, iov_count, read); + bio = bio_copy_user_iov(q, iov, iov_count, read, gfp_mask); else - bio = bio_map_user_iov(q, NULL, iov, iov_count, read); + bio = bio_map_user_iov(q, NULL, iov, iov_count, read, gfp_mask); if (IS_ERR(bio)) return PTR_ERR(bio); diff --git a/block/bsg.c b/block/bsg.c index 0aae8d7ba99..e7a142e9916 100644 --- a/block/bsg.c +++ b/block/bsg.c @@ -283,7 +283,8 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, int has_write_perm) next_rq->cmd_type = rq->cmd_type; dxferp = (void*)(unsigned long)hdr->din_xferp; - ret = blk_rq_map_user(q, next_rq, dxferp, hdr->din_xfer_len); + ret = blk_rq_map_user(q, next_rq, dxferp, hdr->din_xfer_len, + GFP_KERNEL); if (ret) goto out; } @@ -298,7 +299,7 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, int has_write_perm) dxfer_len = 0; if (dxfer_len) { - ret = blk_rq_map_user(q, rq, dxferp, dxfer_len); + ret = blk_rq_map_user(q, rq, dxferp, dxfer_len, GFP_KERNEL); if (ret) goto out; } diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c index 3aab80a4c48..f49d6a11a69 100644 --- a/block/scsi_ioctl.c +++ b/block/scsi_ioctl.c @@ -315,10 +315,11 @@ static int sg_io(struct file *file, struct request_queue *q, } ret = blk_rq_map_user_iov(q, rq, iov, hdr->iovec_count, - hdr->dxfer_len); + hdr->dxfer_len, GFP_KERNEL); kfree(iov); } else if (hdr->dxfer_len) - ret = blk_rq_map_user(q, rq, hdr->dxferp, hdr->dxfer_len); + ret = blk_rq_map_user(q, rq, hdr->dxferp, hdr->dxfer_len, + GFP_KERNEL); if (ret) goto out; -- cgit v1.2.3 From 152e283fdfea0cd11e297d982378b55937842dde Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Thu, 28 Aug 2008 16:17:06 +0900 Subject: block: introduce struct rq_map_data to use reserved pages This patch introduces struct rq_map_data to enable bio_copy_use_iov() use reserved pages. Currently, bio_copy_user_iov allocates bounce pages but drivers/scsi/sg.c wants to allocate pages by itself and use them. struct rq_map_data can be used to pass allocated pages to bio_copy_user_iov. The current users of bio_copy_user_iov simply passes NULL (they don't want to use pre-allocated pages). Signed-off-by: FUJITA Tomonori Cc: Jens Axboe Cc: Douglas Gilbert Cc: Mike Christie Cc: James Bottomley Signed-off-by: Jens Axboe --- block/blk-map.c | 26 ++++++++++++++++---------- block/bsg.c | 7 ++++--- block/scsi_ioctl.c | 4 ++-- 3 files changed, 22 insertions(+), 15 deletions(-) (limited to 'block') diff --git a/block/blk-map.c b/block/blk-map.c index ac21b7397e1..dad6a290783 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -41,8 +41,8 @@ static int __blk_rq_unmap_user(struct bio *bio) } static int __blk_rq_map_user(struct request_queue *q, struct request *rq, - void __user *ubuf, unsigned int len, - gfp_t gfp_mask) + struct rq_map_data *map_data, void __user *ubuf, + unsigned int len, gfp_t gfp_mask) { unsigned long uaddr; unsigned int alignment; @@ -57,10 +57,10 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, */ uaddr = (unsigned long) ubuf; alignment = queue_dma_alignment(q) | q->dma_pad_mask; - if (!(uaddr & alignment) && !(len & alignment)) + if (!(uaddr & alignment) && !(len & alignment) && !map_data) bio = bio_map_user(q, NULL, uaddr, len, reading, gfp_mask); else - bio = bio_copy_user(q, uaddr, len, reading, gfp_mask); + bio = bio_copy_user(q, map_data, uaddr, len, reading, gfp_mask); if (IS_ERR(bio)) return PTR_ERR(bio); @@ -89,6 +89,7 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, * blk_rq_map_user - map user data to a request, for REQ_TYPE_BLOCK_PC usage * @q: request queue where request should be inserted * @rq: request structure to fill + * @map_data: pointer to the rq_map_data holding pages (if necessary) * @ubuf: the user buffer * @len: length of user data * @gfp_mask: memory allocation flags @@ -107,7 +108,8 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, * unmapping. */ int blk_rq_map_user(struct request_queue *q, struct request *rq, - void __user *ubuf, unsigned long len, gfp_t gfp_mask) + struct rq_map_data *map_data, void __user *ubuf, + unsigned long len, gfp_t gfp_mask) { unsigned long bytes_read = 0; struct bio *bio = NULL; @@ -134,7 +136,8 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq, if (end - start > BIO_MAX_PAGES) map_len -= PAGE_SIZE; - ret = __blk_rq_map_user(q, rq, ubuf, map_len, gfp_mask); + ret = __blk_rq_map_user(q, rq, map_data, ubuf, map_len, + gfp_mask); if (ret < 0) goto unmap_rq; if (!bio) @@ -159,6 +162,7 @@ EXPORT_SYMBOL(blk_rq_map_user); * blk_rq_map_user_iov - map user data to a request, for REQ_TYPE_BLOCK_PC usage * @q: request queue where request should be inserted * @rq: request to map data to + * @map_data: pointer to the rq_map_data holding pages (if necessary) * @iov: pointer to the iovec * @iov_count: number of elements in the iovec * @len: I/O byte count @@ -178,8 +182,8 @@ EXPORT_SYMBOL(blk_rq_map_user); * unmapping. */ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, - struct sg_iovec *iov, int iov_count, unsigned int len, - gfp_t gfp_mask) + struct rq_map_data *map_data, struct sg_iovec *iov, + int iov_count, unsigned int len, gfp_t gfp_mask) { struct bio *bio; int i, read = rq_data_dir(rq) == READ; @@ -197,8 +201,9 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, } } - if (unaligned || (q->dma_pad_mask & len)) - bio = bio_copy_user_iov(q, iov, iov_count, read, gfp_mask); + if (unaligned || (q->dma_pad_mask & len) || map_data) + bio = bio_copy_user_iov(q, map_data, iov, iov_count, read, + gfp_mask); else bio = bio_map_user_iov(q, NULL, iov, iov_count, read, gfp_mask); @@ -220,6 +225,7 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, rq->buffer = rq->data = NULL; return 0; } +EXPORT_SYMBOL(blk_rq_map_user_iov); /** * blk_rq_unmap_user - unmap a request with user data diff --git a/block/bsg.c b/block/bsg.c index e7a142e9916..56cb343c76d 100644 --- a/block/bsg.c +++ b/block/bsg.c @@ -283,8 +283,8 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, int has_write_perm) next_rq->cmd_type = rq->cmd_type; dxferp = (void*)(unsigned long)hdr->din_xferp; - ret = blk_rq_map_user(q, next_rq, dxferp, hdr->din_xfer_len, - GFP_KERNEL); + ret = blk_rq_map_user(q, next_rq, NULL, dxferp, + hdr->din_xfer_len, GFP_KERNEL); if (ret) goto out; } @@ -299,7 +299,8 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, int has_write_perm) dxfer_len = 0; if (dxfer_len) { - ret = blk_rq_map_user(q, rq, dxferp, dxfer_len, GFP_KERNEL); + ret = blk_rq_map_user(q, rq, NULL, dxferp, dxfer_len, + GFP_KERNEL); if (ret) goto out; } diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c index f49d6a11a69..c34272a348f 100644 --- a/block/scsi_ioctl.c +++ b/block/scsi_ioctl.c @@ -314,11 +314,11 @@ static int sg_io(struct file *file, struct request_queue *q, goto out; } - ret = blk_rq_map_user_iov(q, rq, iov, hdr->iovec_count, + ret = blk_rq_map_user_iov(q, rq, NULL, iov, hdr->iovec_count, hdr->dxfer_len, GFP_KERNEL); kfree(iov); } else if (hdr->dxfer_len) - ret = blk_rq_map_user(q, rq, hdr->dxferp, hdr->dxfer_len, + ret = blk_rq_map_user(q, rq, NULL, hdr->dxferp, hdr->dxfer_len, GFP_KERNEL); if (ret) -- cgit v1.2.3 From 879040742cf09f2360a9ac41846288707e4e567c Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Thu, 28 Aug 2008 15:05:58 +0900 Subject: block: add blk_rq_aligned helper function This adds blk_rq_aligned helper function to see if alignment and padding requirement is satisfied for DMA transfer. This also converts blk_rq_map_kern and __blk_rq_map_user to use the helper function. Signed-off-by: FUJITA Tomonori Cc: Jens Axboe Signed-off-by: Jens Axboe --- block/blk-map.c | 12 ++---------- 1 file changed, 2 insertions(+), 10 deletions(-) (limited to 'block') diff --git a/block/blk-map.c b/block/blk-map.c index dad6a290783..572140cda5f 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -45,7 +45,6 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, unsigned int len, gfp_t gfp_mask) { unsigned long uaddr; - unsigned int alignment; struct bio *bio, *orig_bio; int reading, ret; @@ -56,8 +55,7 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, * direct dma. else, set up kernel bounce buffers */ uaddr = (unsigned long) ubuf; - alignment = queue_dma_alignment(q) | q->dma_pad_mask; - if (!(uaddr & alignment) && !(len & alignment) && !map_data) + if (blk_rq_aligned(q, ubuf, len) && !map_data) bio = bio_map_user(q, NULL, uaddr, len, reading, gfp_mask); else bio = bio_copy_user(q, map_data, uaddr, len, reading, gfp_mask); @@ -274,8 +272,6 @@ EXPORT_SYMBOL(blk_rq_unmap_user); int blk_rq_map_kern(struct request_queue *q, struct request *rq, void *kbuf, unsigned int len, gfp_t gfp_mask) { - unsigned long kaddr; - unsigned int alignment; int reading = rq_data_dir(rq) == READ; int do_copy = 0; struct bio *bio; @@ -285,11 +281,7 @@ int blk_rq_map_kern(struct request_queue *q, struct request *rq, void *kbuf, if (!len || !kbuf) return -EINVAL; - kaddr = (unsigned long)kbuf; - alignment = queue_dma_alignment(q) | q->dma_pad_mask; - do_copy = ((kaddr & alignment) || (len & alignment) || - object_is_on_stack(kbuf)); - + do_copy = !blk_rq_aligned(q, kbuf, len) || object_is_on_stack(kbuf); if (do_copy) bio = bio_copy_kern(q, kbuf, len, gfp_mask, reading); else -- cgit v1.2.3 From aeb3d3a81e81c6323a17fe914e91eb228b3f1aa1 Mon Sep 17 00:00:00 2001 From: Harvey Harrison Date: Thu, 28 Aug 2008 09:27:42 +0200 Subject: block: kmalloc args reversed, small function definition fixes Noticed by sparse: block/blk-softirq.c:156:12: warning: symbol 'blk_softirq_init' was not declared. Should it be static? block/genhd.c:583:28: warning: function 'bdget_disk' with external linkage has definition block/genhd.c:659:17: warning: incorrect type in argument 1 (different base types) block/genhd.c:659:17: expected unsigned int [unsigned] [usertype] size block/genhd.c:659:17: got restricted gfp_t block/genhd.c:659:29: warning: incorrect type in argument 2 (different base types) block/genhd.c:659:29: expected restricted gfp_t [usertype] flags block/genhd.c:659:29: got unsigned int block: kmalloc args reversed Signed-off-by: Harvey Harrison Signed-off-by: Jens Axboe --- block/blk-settings.c | 2 +- block/genhd.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/blk-settings.c b/block/blk-settings.c index a60e959a12c..d70692badcd 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -443,7 +443,7 @@ void blk_queue_update_dma_alignment(struct request_queue *q, int mask) } EXPORT_SYMBOL(blk_queue_update_dma_alignment); -int __init blk_settings_init(void) +static int __init blk_settings_init(void) { blk_max_low_pfn = max_low_pfn - 1; blk_max_pfn = max_pfn - 1; diff --git a/block/genhd.c b/block/genhd.c index d9de3e482d1..32ee73c6756 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -578,7 +578,7 @@ struct gendisk *get_gendisk(dev_t devt, int *partno) * RETURNS: * Resulting block_device on success, NULL on failure. */ -extern struct block_device *bdget_disk(struct gendisk *disk, int partno) +struct block_device *bdget_disk(struct gendisk *disk, int partno) { struct hd_struct *part; struct block_device *bdev = NULL; @@ -654,7 +654,7 @@ static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) struct class_dev_iter *iter; struct device *dev; - iter = kmalloc(GFP_KERNEL, sizeof(*iter)); + iter = kmalloc(sizeof(*iter), GFP_KERNEL); if (!iter) return ERR_PTR(-ENOMEM); -- cgit v1.2.3 From 2bbedcb4c1abac498f18e5770d62ae66ff235ada Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 29 Aug 2008 11:41:51 +0200 Subject: block: don't test for partition size in bdget_disk() and blk_lookup_devt() bdget_disk() and blk_lookup_devt() never cared whether the specified partition (or disk) is zero sized or not. I got confused while converting those not to depend on consecutive minor numbers in commit 5a6411b1178baf534aa9138052864dfa89d3eada and later when dev0 was added it broke callers which expected to get valid return for zero sized disk devices. So, they never needed nr_sects checks in the first place. Kill them. This problem was spotted and debugged by Bartlmoiej Zolnierkiewicz. Signed-off-by: Tejun Heo Cc: Bartlomiej Zolnierkiewicz Signed-off-by: Jens Axboe --- block/genhd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 32ee73c6756..ed926b760ca 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -584,7 +584,7 @@ struct block_device *bdget_disk(struct gendisk *disk, int partno) struct block_device *bdev = NULL; part = disk_get_part(disk, partno); - if (part && (part->nr_sects || partno == 0)) + if (part) bdev = bdget(part_devt(part)); disk_put_part(part); @@ -1031,7 +1031,7 @@ dev_t blk_lookup_devt(const char *name, int partno) continue; part = disk_get_part(disk, partno); - if (part && (part->nr_sects || partno == 0)) { + if (part) { devt = part_devt(part); disk_put_part(part); break; -- cgit v1.2.3 From 839e96afba87117befd39cf4e43f156edc8047a7 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 2 Sep 2008 09:25:21 +0200 Subject: block: update comment on end_request() It refers to functions that no longer exist after the IO completion changes. Signed-off-by: Jens Axboe --- block/blk-core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 5bf806adc77..f25eb9786d9 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1780,9 +1780,9 @@ EXPORT_SYMBOL(end_dequeued_request); * they have a residual value to account for. For that case this function * isn't really useful, unless the residual just happens to be the * full current segment. In other words, don't use this function in new - * code. Either use end_request_completely(), or the - * end_that_request_chunk() (along with end_that_request_last()) for - * partial completions. + * code. Use blk_end_request() or __blk_end_request() to end partial parts + * of a request, or end_dequeued_request() and end_queued_request() to + * completely end IO on a dequeued/queued request. * **/ void end_request(struct request *req, int uptodate) -- cgit v1.2.3 From 818827669d85b84241696ffef2de485db46b0b5e Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Tue, 2 Sep 2008 16:20:19 +0900 Subject: block: make blk_rq_map_user take a NULL user-space buffer This patch changes blk_rq_map_user to accept a NULL user-space buffer with a READ command if rq_map_data is not NULL. Thus a caller can pass page frames to lk_rq_map_user to just set up a request and bios with page frames propely. bio_uncopy_user (called via blk_rq_unmap_user) doesn't copy data to user space with such request. Signed-off-by: FUJITA Tomonori Signed-off-by: Jens Axboe --- block/blk-map.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'block') diff --git a/block/blk-map.c b/block/blk-map.c index 572140cda5f..4849fa36161 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -42,7 +42,7 @@ static int __blk_rq_unmap_user(struct bio *bio) static int __blk_rq_map_user(struct request_queue *q, struct request *rq, struct rq_map_data *map_data, void __user *ubuf, - unsigned int len, gfp_t gfp_mask) + unsigned int len, int null_mapped, gfp_t gfp_mask) { unsigned long uaddr; struct bio *bio, *orig_bio; @@ -63,6 +63,9 @@ static int __blk_rq_map_user(struct request_queue *q, struct request *rq, if (IS_ERR(bio)) return PTR_ERR(bio); + if (null_mapped) + bio->bi_flags |= (1 << BIO_NULL_MAPPED); + orig_bio = bio; blk_queue_bounce(q, &bio); @@ -111,12 +114,17 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq, { unsigned long bytes_read = 0; struct bio *bio = NULL; - int ret; + int ret, null_mapped = 0; if (len > (q->max_hw_sectors << 9)) return -EINVAL; - if (!len || !ubuf) + if (!len) return -EINVAL; + if (!ubuf) { + if (!map_data || rq_data_dir(rq) != READ) + return -EINVAL; + null_mapped = 1; + } while (bytes_read != len) { unsigned long map_len, end, start; @@ -135,7 +143,7 @@ int blk_rq_map_user(struct request_queue *q, struct request *rq, map_len -= PAGE_SIZE; ret = __blk_rq_map_user(q, rq, map_data, ubuf, map_len, - gfp_mask); + null_mapped, gfp_mask); if (ret < 0) goto unmap_rq; if (!bio) -- cgit v1.2.3 From 243294dae09c909c0442c8f04d470b69c3c19d6e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Thu, 4 Sep 2008 09:17:31 +0200 Subject: block: fix duplicate headers for /proc/partitions seqf can be started multiple times for a read and the header should be printed only for the initial one. Fix it. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index ed926b760ca..8acaff0154e 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -697,7 +697,7 @@ static void *show_partition_start(struct seq_file *seqf, loff_t *pos) static void *p; p = disk_seqf_start(seqf, pos); - if (!IS_ERR(p) && p) + if (!IS_ERR(p) && p && !*pos) seq_puts(seqf, "major minor #blocks name\n\n"); return p; } -- cgit v1.2.3 From 242f9dcb8ba6f68fcd217a119a7648a4f69290e9 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Sun, 14 Sep 2008 05:55:09 -0700 Subject: block: unify request timeout handling Right now SCSI and others do their own command timeout handling. Move those bits to the block layer. Instead of having a timer per command, we try to be a bit more clever and simply have one per-queue. This avoids the overhead of having to tear down and setup a timer for each command, so it will result in a lot less timer fiddling. Signed-off-by: Mike Anderson Signed-off-by: Jens Axboe --- block/Makefile | 4 +- block/blk-core.c | 7 +++ block/blk-settings.c | 12 ++++ block/blk-softirq.c | 30 ++++++---- block/blk-timeout.c | 155 +++++++++++++++++++++++++++++++++++++++++++++++++++ block/blk.h | 24 ++++++++ block/elevator.c | 8 +++ 7 files changed, 226 insertions(+), 14 deletions(-) create mode 100644 block/blk-timeout.c (limited to 'block') diff --git a/block/Makefile b/block/Makefile index 0da976ce67d..bfe73049f93 100644 --- a/block/Makefile +++ b/block/Makefile @@ -4,8 +4,8 @@ obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \ blk-barrier.o blk-settings.o blk-ioc.o blk-map.o \ - blk-exec.o blk-merge.o blk-softirq.o ioctl.o genhd.o \ - scsi_ioctl.o cmd-filter.o + blk-exec.o blk-merge.o blk-softirq.o blk-timeout.o \ + ioctl.o genhd.o scsi_ioctl.o cmd-filter.o obj-$(CONFIG_BLK_DEV_BSG) += bsg.o obj-$(CONFIG_IOSCHED_NOOP) += noop-iosched.o diff --git a/block/blk-core.c b/block/blk-core.c index f25eb9786d9..d768a8ddc17 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -110,6 +110,7 @@ void blk_rq_init(struct request_queue *q, struct request *rq) memset(rq, 0, sizeof(*rq)); INIT_LIST_HEAD(&rq->queuelist); + INIT_LIST_HEAD(&rq->timeout_list); rq->cpu = -1; rq->q = q; rq->sector = rq->hard_sector = (sector_t) -1; @@ -490,6 +491,8 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) } init_timer(&q->unplug_timer); + setup_timer(&q->timeout, blk_rq_timed_out_timer, (unsigned long) q); + INIT_LIST_HEAD(&q->timeout_list); kobject_init(&q->kobj, &blk_queue_ktype); @@ -897,6 +900,8 @@ EXPORT_SYMBOL(blk_start_queueing); */ void blk_requeue_request(struct request_queue *q, struct request *rq) { + blk_delete_timer(rq); + blk_clear_rq_complete(rq); blk_add_trace_rq(q, rq, BLK_TA_REQUEUE); if (blk_rq_tagged(rq)) @@ -1650,6 +1655,8 @@ static void end_that_request_last(struct request *req, int error) { struct gendisk *disk = req->rq_disk; + blk_delete_timer(req); + if (blk_rq_tagged(req)) blk_queue_end_tag(req->q, req); diff --git a/block/blk-settings.c b/block/blk-settings.c index d70692badcd..1d0330d0b40 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -77,6 +77,18 @@ void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn) } EXPORT_SYMBOL(blk_queue_softirq_done); +void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout) +{ + q->rq_timeout = timeout; +} +EXPORT_SYMBOL_GPL(blk_queue_rq_timeout); + +void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn) +{ + q->rq_timed_out_fn = fn; +} +EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out); + /** * blk_queue_make_request - define an alternate make_request function for a device * @q: the request queue for the device to be affected diff --git a/block/blk-softirq.c b/block/blk-softirq.c index 3a1af551191..7ab344afb16 100644 --- a/block/blk-softirq.c +++ b/block/blk-softirq.c @@ -101,18 +101,7 @@ static struct notifier_block __cpuinitdata blk_cpu_notifier = { .notifier_call = blk_cpu_notify, }; -/** - * blk_complete_request - end I/O on a request - * @req: the request being processed - * - * Description: - * Ends all I/O on a request. It does not handle partial completions, - * unless the driver actually implements this in its completion callback - * through requeueing. The actual completion happens out-of-order, - * through a softirq handler. The user must have registered a completion - * callback through blk_queue_softirq_done(). - **/ -void blk_complete_request(struct request *req) +void __blk_complete_request(struct request *req) { struct request_queue *q = req->q; unsigned long flags; @@ -151,6 +140,23 @@ do_local: local_irq_restore(flags); } + +/** + * blk_complete_request - end I/O on a request + * @req: the request being processed + * + * Description: + * Ends all I/O on a request. It does not handle partial completions, + * unless the driver actually implements this in its completion callback + * through requeueing. The actual completion happens out-of-order, + * through a softirq handler. The user must have registered a completion + * callback through blk_queue_softirq_done(). + **/ +void blk_complete_request(struct request *req) +{ + if (!blk_mark_rq_complete(req)) + __blk_complete_request(req); +} EXPORT_SYMBOL(blk_complete_request); __init int blk_softirq_init(void) diff --git a/block/blk-timeout.c b/block/blk-timeout.c new file mode 100644 index 00000000000..b36d07bf0af --- /dev/null +++ b/block/blk-timeout.c @@ -0,0 +1,155 @@ +/* + * Functions related to generic timeout handling of requests. + */ +#include +#include +#include + +#include "blk.h" + +/* + * blk_delete_timer - Delete/cancel timer for a given function. + * @req: request that we are canceling timer for + * + */ +void blk_delete_timer(struct request *req) +{ + struct request_queue *q = req->q; + + /* + * Nothing to detach + */ + if (!q->rq_timed_out_fn || !req->deadline) + return; + + list_del_init(&req->timeout_list); + + if (list_empty(&q->timeout_list)) + del_timer(&q->timeout); +} + +static void blk_rq_timed_out(struct request *req) +{ + struct request_queue *q = req->q; + enum blk_eh_timer_return ret; + + ret = q->rq_timed_out_fn(req); + switch (ret) { + case BLK_EH_HANDLED: + __blk_complete_request(req); + break; + case BLK_EH_RESET_TIMER: + blk_clear_rq_complete(req); + blk_add_timer(req); + break; + case BLK_EH_NOT_HANDLED: + /* + * LLD handles this for now but in the future + * we can send a request msg to abort the command + * and we can move more of the generic scsi eh code to + * the blk layer. + */ + break; + default: + printk(KERN_ERR "block: bad eh return: %d\n", ret); + break; + } +} + +void blk_rq_timed_out_timer(unsigned long data) +{ + struct request_queue *q = (struct request_queue *) data; + unsigned long flags, uninitialized_var(next), next_set = 0; + struct request *rq, *tmp; + + spin_lock_irqsave(q->queue_lock, flags); + + list_for_each_entry_safe(rq, tmp, &q->timeout_list, timeout_list) { + if (time_after_eq(jiffies, rq->deadline)) { + list_del_init(&rq->timeout_list); + + /* + * Check if we raced with end io completion + */ + if (blk_mark_rq_complete(rq)) + continue; + blk_rq_timed_out(rq); + } + if (!next_set) { + next = rq->deadline; + next_set = 1; + } else if (time_after(next, rq->deadline)) + next = rq->deadline; + } + + if (next_set && !list_empty(&q->timeout_list)) + mod_timer(&q->timeout, round_jiffies(next)); + + spin_unlock_irqrestore(q->queue_lock, flags); +} + +/** + * blk_abort_request -- Request request recovery for the specified command + * @req: pointer to the request of interest + * + * This function requests that the block layer start recovery for the + * request by deleting the timer and calling the q's timeout function. + * LLDDs who implement their own error recovery MAY ignore the timeout + * event if they generated blk_abort_req. Must hold queue lock. + */ +void blk_abort_request(struct request *req) +{ + blk_delete_timer(req); + blk_rq_timed_out(req); +} +EXPORT_SYMBOL_GPL(blk_abort_request); + +/** + * blk_add_timer - Start timeout timer for a single request + * @req: request that is about to start running. + * + * Notes: + * Each request has its own timer, and as it is added to the queue, we + * set up the timer. When the request completes, we cancel the timer. + */ +void blk_add_timer(struct request *req) +{ + struct request_queue *q = req->q; + unsigned long expiry; + + if (!q->rq_timed_out_fn) + return; + + BUG_ON(!list_empty(&req->timeout_list)); + BUG_ON(test_bit(REQ_ATOM_COMPLETE, &req->atomic_flags)); + + if (req->timeout) + req->deadline = jiffies + req->timeout; + else { + req->deadline = jiffies + q->rq_timeout; + /* + * Some LLDs, like scsi, peek at the timeout to prevent + * a command from being retried forever. + */ + req->timeout = q->rq_timeout; + } + list_add_tail(&req->timeout_list, &q->timeout_list); + + /* + * If the timer isn't already pending or this timeout is earlier + * than an existing one, modify the timer. Round to next nearest + * second. + */ + expiry = round_jiffies(req->deadline); + + /* + * We use ->deadline == 0 to detect whether a timer was added or + * not, so just increase to next jiffy for that specific case + */ + if (unlikely(!req->deadline)) + req->deadline = 1; + + if (!timer_pending(&q->timeout) || + time_before(expiry, q->timeout.expires)) + mod_timer(&q->timeout, expiry); +} diff --git a/block/blk.h b/block/blk.h index de74254cb91..a4f4a50aefa 100644 --- a/block/blk.h +++ b/block/blk.h @@ -17,6 +17,30 @@ void __blk_queue_free_tags(struct request_queue *q); void blk_unplug_work(struct work_struct *work); void blk_unplug_timeout(unsigned long data); +void blk_rq_timed_out_timer(unsigned long data); +void blk_delete_timer(struct request *); +void blk_add_timer(struct request *); + +/* + * Internal atomic flags for request handling + */ +enum rq_atomic_flags { + REQ_ATOM_COMPLETE = 0, +}; + +/* + * EH timer and IO completion will both attempt to 'grab' the request, make + * sure that only one of them suceeds + */ +static inline int blk_mark_rq_complete(struct request *rq) +{ + return test_and_set_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags); +} + +static inline void blk_clear_rq_complete(struct request *rq) +{ + clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags); +} struct io_context *current_io_context(gfp_t gfp_flags, int node); diff --git a/block/elevator.c b/block/elevator.c index 8e3fc3afc77..a91fc59edd0 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -36,6 +36,8 @@ #include #include +#include "blk.h" + static DEFINE_SPINLOCK(elv_list_lock); static LIST_HEAD(elv_list); @@ -771,6 +773,12 @@ struct request *elv_next_request(struct request_queue *q) */ rq->cmd_flags |= REQ_STARTED; blk_add_trace_rq(q, rq, BLK_TA_ISSUE); + + /* + * We are now handing the request to the hardware, + * add the timeout handler + */ + blk_add_timer(rq); } if (!q->boundary_rq || q->boundary_rq == rq) { -- cgit v1.2.3 From 11914a53d2ec2974a565311af327b8983d8c820d Mon Sep 17 00:00:00 2001 From: Mike Anderson Date: Sat, 13 Sep 2008 20:31:27 +0200 Subject: block: Add interface to abort queued requests Signed-off-by: Mike Anderson Signed-off-by: Jens Axboe --- block/blk-timeout.c | 22 ++++++++++++++++++++++ block/elevator.c | 13 +++++++++++++ 2 files changed, 35 insertions(+) (limited to 'block') diff --git a/block/blk-timeout.c b/block/blk-timeout.c index b36d07bf0af..6e5c781c5af 100644 --- a/block/blk-timeout.c +++ b/block/blk-timeout.c @@ -153,3 +153,25 @@ void blk_add_timer(struct request *req) time_before(expiry, q->timeout.expires)) mod_timer(&q->timeout, expiry); } + +/** + * blk_abort_queue -- Abort all request on given queue + * @queue: pointer to queue + * + */ +void blk_abort_queue(struct request_queue *q) +{ + unsigned long flags; + struct request *rq, *tmp; + + spin_lock_irqsave(q->queue_lock, flags); + + elv_abort_queue(q); + + list_for_each_entry_safe(rq, tmp, &q->timeout_list, timeout_list) + blk_abort_request(rq); + + spin_unlock_irqrestore(q->queue_lock, flags); + +} +EXPORT_SYMBOL_GPL(blk_abort_queue); diff --git a/block/elevator.c b/block/elevator.c index a91fc59edd0..8a74eedc353 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -914,6 +914,19 @@ int elv_may_queue(struct request_queue *q, int rw) return ELV_MQUEUE_MAY; } +void elv_abort_queue(struct request_queue *q) +{ + struct request *rq; + + while (!list_empty(&q->queue_head)) { + rq = list_entry_rq(q->queue_head.next); + rq->cmd_flags |= REQ_QUIET; + blk_add_trace_rq(q, rq, BLK_TA_ABORT); + end_queued_request(rq, 0); + } +} +EXPORT_SYMBOL(elv_abort_queue); + void elv_completed_request(struct request_queue *q, struct request *rq) { elevator_t *e = q->elevator; -- cgit v1.2.3 From 3e6053d76dcbd92b2f9f4ad5ece9bce83149523e Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Thu, 11 Sep 2008 10:57:55 +0200 Subject: block: adjust blkdev_issue_discard for swap Two mods to blkdev_issue_discard(), thinking ahead to its use on swap: 1. Add gfp_mask argument, so swap allocation can use it where GFP_KERNEL might deadlock but GFP_NOIO is safe. 2. Enlarge nr_sects argument from unsigned to sector_t: unsigned long is enough to cover a whole swap area, but sector_t suits any partition. Change sb_issue_discard()'s nr_blocks to sector_t too; but no need seen for a gfp_mask there, just pass GFP_KERNEL down to blkdev_issue_discard(). Signed-off-by: Hugh Dickins Signed-off-by: Jens Axboe --- block/blk-barrier.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/blk-barrier.c b/block/blk-barrier.c index 988b63479b2..5c99ff8d2db 100644 --- a/block/blk-barrier.c +++ b/block/blk-barrier.c @@ -332,12 +332,13 @@ static void blkdev_discard_end_io(struct bio *bio, int err) * @bdev: blockdev to issue discard for * @sector: start sector * @nr_sects: number of sectors to discard + * @gfp_mask: memory allocation flags (for bio_alloc) * * Description: * Issue a discard request for the sectors in question. Does not wait. */ -int blkdev_issue_discard(struct block_device *bdev, sector_t sector, - unsigned nr_sects) +int blkdev_issue_discard(struct block_device *bdev, + sector_t sector, sector_t nr_sects, gfp_t gfp_mask) { struct request_queue *q; struct bio *bio; @@ -354,7 +355,7 @@ int blkdev_issue_discard(struct block_device *bdev, sector_t sector, return -EOPNOTSUPP; while (nr_sects && !ret) { - bio = bio_alloc(GFP_KERNEL, 0); + bio = bio_alloc(gfp_mask, 0); if (!bio) return -ENOMEM; -- cgit v1.2.3 From 581d4e28d9195aa8b2231383dbabc288988d615e Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Sun, 14 Sep 2008 05:56:33 -0700 Subject: block: add fault injection mechanism for faking request timeouts Only works for the generic request timer handling. Allows one to sporadically ignore request completions, thus exercising the timeout handling. Signed-off-by: Jens Axboe --- block/blk-softirq.c | 2 ++ block/blk-timeout.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++++++++ block/blk.h | 12 +++++++++++ block/genhd.c | 8 ++++++++ 4 files changed, 81 insertions(+) (limited to 'block') diff --git a/block/blk-softirq.c b/block/blk-softirq.c index 7ab344afb16..e660d26ca65 100644 --- a/block/blk-softirq.c +++ b/block/blk-softirq.c @@ -154,6 +154,8 @@ do_local: **/ void blk_complete_request(struct request *req) { + if (unlikely(blk_should_fake_timeout(req->q))) + return; if (!blk_mark_rq_complete(req)) __blk_complete_request(req); } diff --git a/block/blk-timeout.c b/block/blk-timeout.c index 6e5c781c5af..9b4ad138bb3 100644 --- a/block/blk-timeout.c +++ b/block/blk-timeout.c @@ -4,9 +4,68 @@ #include #include #include +#include #include "blk.h" +#ifdef CONFIG_FAIL_IO_TIMEOUT + +static DECLARE_FAULT_ATTR(fail_io_timeout); + +static int __init setup_fail_io_timeout(char *str) +{ + return setup_fault_attr(&fail_io_timeout, str); +} +__setup("fail_io_timeout=", setup_fail_io_timeout); + +int blk_should_fake_timeout(struct request_queue *q) +{ + if (!test_bit(QUEUE_FLAG_FAIL_IO, &q->queue_flags)) + return 0; + + return should_fail(&fail_io_timeout, 1); +} + +static int __init fail_io_timeout_debugfs(void) +{ + return init_fault_attr_dentries(&fail_io_timeout, "fail_io_timeout"); +} + +late_initcall(fail_io_timeout_debugfs); + +ssize_t part_timeout_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + int set = test_bit(QUEUE_FLAG_FAIL_IO, &disk->queue->queue_flags); + + return sprintf(buf, "%d\n", set != 0); +} + +ssize_t part_timeout_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct gendisk *disk = dev_to_disk(dev); + int val; + + if (count) { + struct request_queue *q = disk->queue; + char *p = (char *) buf; + + val = simple_strtoul(p, &p, 10); + spin_lock_irq(q->queue_lock); + if (val) + queue_flag_set(QUEUE_FLAG_FAIL_IO, q); + else + queue_flag_clear(QUEUE_FLAG_FAIL_IO, q); + spin_unlock_irq(q->queue_lock); + } + + return count; +} + +#endif /* CONFIG_FAIL_IO_TIMEOUT */ + /* * blk_delete_timer - Delete/cancel timer for a given function. * @req: request that we are canceling timer for diff --git a/block/blk.h b/block/blk.h index a4f4a50aefa..e5c57976996 100644 --- a/block/blk.h +++ b/block/blk.h @@ -42,6 +42,18 @@ static inline void blk_clear_rq_complete(struct request *rq) clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags); } +#ifdef CONFIG_FAIL_IO_TIMEOUT +int blk_should_fake_timeout(struct request_queue *); +ssize_t part_timeout_show(struct device *, struct device_attribute *, char *); +ssize_t part_timeout_store(struct device *, struct device_attribute *, + const char *, size_t); +#else +static inline int blk_should_fake_timeout(struct request_queue *q) +{ + return 0; +} +#endif + struct io_context *current_io_context(gfp_t gfp_flags, int node); int ll_back_merge_fn(struct request_queue *q, struct request *req, diff --git a/block/genhd.c b/block/genhd.c index 8acaff0154e..4cd3433c99a 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -817,6 +817,11 @@ static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL); static struct device_attribute dev_attr_fail = __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store); #endif +#ifdef CONFIG_FAIL_IO_TIMEOUT +static struct device_attribute dev_attr_fail_timeout = + __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show, + part_timeout_store); +#endif static struct attribute *disk_attrs[] = { &dev_attr_range.attr, @@ -828,6 +833,9 @@ static struct attribute *disk_attrs[] = { &dev_attr_stat.attr, #ifdef CONFIG_FAIL_MAKE_REQUEST &dev_attr_fail.attr, +#endif +#ifdef CONFIG_FAIL_IO_TIMEOUT + &dev_attr_fail_timeout.attr, #endif NULL }; -- cgit v1.2.3 From 7ba1fbaa4a478f72fbaf5a56af9c82a77966b4c7 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 16 Sep 2008 09:54:11 -0700 Subject: block: use rq complete marking in blk_abort_request() We cannot abort a request if we raced with the timeout handler already, or with the IO completion. So make blk_abort_request() mark the request as complete, and only continue if we succeeded. Found and suggested by Mike Anderson Signed-off-by: Jens Axboe --- block/blk-timeout.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'block') diff --git a/block/blk-timeout.c b/block/blk-timeout.c index 9b4ad138bb3..972a63f848f 100644 --- a/block/blk-timeout.c +++ b/block/blk-timeout.c @@ -158,6 +158,8 @@ void blk_rq_timed_out_timer(unsigned long data) */ void blk_abort_request(struct request *req) { + if (blk_mark_rq_complete(req)) + return; blk_delete_timer(req); blk_rq_timed_out(req); } -- cgit v1.2.3 From e3335de94067dbebe22e3962632ead34e832cb60 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 18 Sep 2008 09:22:54 -0700 Subject: block: blk_cleanup_queue() should call blk_sync_queue() When a driver calls blk_cleanup_queue(), the device should be fully idle. However, the block layer may have pending plugging timers and the IO schedulers may have pending work in the work queues. So quisce the device by waiting for the timer and flushing the work queues. Signed-off-by: Jens Axboe --- block/blk-core.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index d768a8ddc17..37fba001bdc 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -436,6 +436,14 @@ void blk_put_queue(struct request_queue *q) void blk_cleanup_queue(struct request_queue *q) { + /* + * We know we have process context here, so we can be a little + * cautious and ensure that pending block actions on this device + * are done before moving on. Going into this function, we should + * not have processes doing IO to this device. + */ + blk_sync_queue(q); + mutex_lock(&q->sysfs_lock); queue_flag_set_unlocked(QUEUE_FLAG_DEAD, q); mutex_unlock(&q->sysfs_lock); -- cgit v1.2.3 From 32fab448e5e86694beade415e750363538ea5f49 Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Thu, 18 Sep 2008 10:45:09 -0400 Subject: block: add request update interface This patch adds blk_update_request(), which updates struct request with completing its data part, but doesn't complete the struct request itself. Though it looks like end_that_request_first() of older kernels, blk_update_request() should be used only by request stacking drivers. Request-based dm will use it in bio->bi_end_io callback to update the original request when a data part of a cloned request completes. Followings are additional background information of why request-based dm needs this interface. - Request stacking drivers can't use blk_end_request() directly from the lower driver's completion context (bio->bi_end_io or rq->end_io), because some device drivers (e.g. ide) may try to complete their request with queue lock held, and it may cause deadlock. See below for detailed description of possible deadlock: - To solve that, request-based dm offloads the completion of cloned struct request to softirq context (i.e. using blk_complete_request() from rq->end_io). - Though it is possible to use the same solution from bio->bi_end_io, it will delay the notification of bio completion to the original submitter. Also, it will cause inefficient partial completion, because the lower driver can't perform the cloned request anymore and request-based dm needs to requeue and redispatch it to the lower driver again later. That's not good. - So request-based dm needs blk_update_request() to perform the bio completion in the lower driver's completion context, which is more efficient. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Jens Axboe --- block/blk-core.c | 57 +++++++++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 48 insertions(+), 9 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 37fba001bdc..527d43e982b 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1806,6 +1806,22 @@ void end_request(struct request *req, int uptodate) } EXPORT_SYMBOL(end_request); +static int end_that_request_data(struct request *rq, int error, + unsigned int nr_bytes, unsigned int bidi_bytes) +{ + if (rq->bio) { + if (__end_that_request_first(rq, error, nr_bytes)) + return 1; + + /* Bidi request must be completed as a whole */ + if (blk_bidi_rq(rq) && + __end_that_request_first(rq->next_rq, error, bidi_bytes)) + return 1; + } + + return 0; +} + /** * blk_end_io - Generic end_io function to complete a request. * @rq: the request being processed @@ -1832,15 +1848,8 @@ static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, struct request_queue *q = rq->q; unsigned long flags = 0UL; - if (rq->bio) { - if (__end_that_request_first(rq, error, nr_bytes)) - return 1; - - /* Bidi request must be completed as a whole */ - if (blk_bidi_rq(rq) && - __end_that_request_first(rq->next_rq, error, bidi_bytes)) - return 1; - } + if (end_that_request_data(rq, error, nr_bytes, bidi_bytes)) + return 1; /* Special feature for tricky drivers */ if (drv_callback && drv_callback(rq)) @@ -1922,6 +1931,36 @@ int blk_end_bidi_request(struct request *rq, int error, unsigned int nr_bytes, } EXPORT_SYMBOL_GPL(blk_end_bidi_request); +/** + * blk_update_request - Special helper function for request stacking drivers + * @rq: the request being processed + * @error: %0 for success, < %0 for error + * @nr_bytes: number of bytes to complete @rq + * + * Description: + * Ends I/O on a number of bytes attached to @rq, but doesn't complete + * the request structure even if @rq doesn't have leftover. + * If @rq has leftover, sets it up for the next range of segments. + * + * This special helper function is only for request stacking drivers + * (e.g. request-based dm) so that they can handle partial completion. + * Actual device drivers should use blk_end_request instead. + */ +void blk_update_request(struct request *rq, int error, unsigned int nr_bytes) +{ + if (!end_that_request_data(rq, error, nr_bytes, 0)) { + /* + * These members are not updated in end_that_request_data() + * when all bios are completed. + * Update them so that the request stacking driver can find + * how many bytes remain in the request later. + */ + rq->nr_sectors = rq->hard_nr_sectors = 0; + rq->current_nr_sectors = rq->hard_cur_sectors = 0; + } +} +EXPORT_SYMBOL_GPL(blk_update_request); + /** * blk_end_request_callback - Special helper function for tricky drivers * @rq: the request being processed -- cgit v1.2.3 From 82124d60354846623a4b94af335717a5e142a074 Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Thu, 18 Sep 2008 10:45:38 -0400 Subject: block: add request submission interface This patch adds blk_insert_cloned_request(), a generic request submission interface for request stacking drivers. Request-based dm will use it to submit their clones to underlying devices. blk_rq_check_limits() is also added because it is possible that the lower queue has stronger limitations than the upper queue if multiple drivers are stacking at request-level. Not only for blk_insert_cloned_request()'s internal use, the function will be used by request-based dm when the queue limitation is modified (e.g. by replacing dm's table). Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Jens Axboe --- block/blk-core.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 81 insertions(+) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 527d43e982b..b8ffbfe85ca 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1529,6 +1529,87 @@ void submit_bio(int rw, struct bio *bio) } EXPORT_SYMBOL(submit_bio); +/** + * blk_rq_check_limits - Helper function to check a request for the queue limit + * @q: the queue + * @rq: the request being checked + * + * Description: + * @rq may have been made based on weaker limitations of upper-level queues + * in request stacking drivers, and it may violate the limitation of @q. + * Since the block layer and the underlying device driver trust @rq + * after it is inserted to @q, it should be checked against @q before + * the insertion using this generic function. + * + * This function should also be useful for request stacking drivers + * in some cases below, so export this fuction. + * Request stacking drivers like request-based dm may change the queue + * limits while requests are in the queue (e.g. dm's table swapping). + * Such request stacking drivers should check those requests agaist + * the new queue limits again when they dispatch those requests, + * although such checkings are also done against the old queue limits + * when submitting requests. + */ +int blk_rq_check_limits(struct request_queue *q, struct request *rq) +{ + if (rq->nr_sectors > q->max_sectors || + rq->data_len > q->max_hw_sectors << 9) { + printk(KERN_ERR "%s: over max size limit.\n", __func__); + return -EIO; + } + + /* + * queue's settings related to segment counting like q->bounce_pfn + * may differ from that of other stacking queues. + * Recalculate it to check the request correctly on this queue's + * limitation. + */ + blk_recalc_rq_segments(rq); + if (rq->nr_phys_segments > q->max_phys_segments || + rq->nr_phys_segments > q->max_hw_segments) { + printk(KERN_ERR "%s: over max segments limit.\n", __func__); + return -EIO; + } + + return 0; +} +EXPORT_SYMBOL_GPL(blk_rq_check_limits); + +/** + * blk_insert_cloned_request - Helper for stacking drivers to submit a request + * @q: the queue to submit the request + * @rq: the request being queued + */ +int blk_insert_cloned_request(struct request_queue *q, struct request *rq) +{ + unsigned long flags; + + if (blk_rq_check_limits(q, rq)) + return -EIO; + +#ifdef CONFIG_FAIL_MAKE_REQUEST + if (rq->rq_disk && rq->rq_disk->part0.make_it_fail && + should_fail(&fail_make_request, blk_rq_bytes(rq))) + return -EIO; +#endif + + spin_lock_irqsave(q->queue_lock, flags); + + /* + * Submitting request must be dequeued before calling this function + * because it will be linked to another request_queue + */ + BUG_ON(blk_queued_rq(rq)); + + drive_stat_acct(rq, 1); + __elv_add_request(q, rq, ELEVATOR_INSERT_BACK, 0); + + spin_unlock_irqrestore(q->queue_lock, flags); + + return 0; +} +EXPORT_SYMBOL_GPL(blk_insert_cloned_request); + /** * __end_that_request_first - end I/O on a request * @req: the request being processed -- cgit v1.2.3 From 4ee5eaf4516a60f8ef64d3c246c64c6be0cf8c3a Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Thu, 18 Sep 2008 10:46:13 -0400 Subject: block: add a queue flag for request stacking support This patch adds a queue flag to indicate the block device can be used for request stacking. Request stacking drivers need to stack their devices on top of only devices of which q->request_fn is functional. Since bio stacking drivers (e.g. md, loop) basically initialize their queue using blk_alloc_queue() and don't set q->request_fn, the check of (q->request_fn == NULL) looks enough for that purpose. However, dm will become both types of stacking driver (bio-based and request-based). And dm will always set q->request_fn even if the dm device is bio-based of which q->request_fn is not functional actually. So we need something else to distinguish the type of the device. Adding a queue flag is a solution for that. The reason why dm always sets q->request_fn is to keep the compatibility of dm user-space tools. Currently, all dm user-space tools are using bio-based dm without specifying the type of the dm device they use. To use request-based dm without changing such tools, the kernel must decide the type of the dm device automatically. The automatic type decision can't be done at the device creation time and needs to be deferred until such tools load a mapping table, since the actual type is decided by dm target type included in the mapping table. So a dm device has to be initialized using blk_init_queue() so that we can load either type of table. Then, all queue stuffs are set (e.g. q->request_fn) and we have no element to distinguish that it is bio-based or request-based, even after a table is loaded and the type of the device is decided. By the way, some stuffs of the queue (e.g. request_list, elevator) are needless when the dm device is used as bio-based. But the memory size is not so large (about 20[KB] per queue on ia64), so I hope the memory loss can be acceptable for bio-based dm users. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Jens Axboe --- block/blk-core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index b8ffbfe85ca..fa212348c4c 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -574,7 +574,8 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id) q->request_fn = rfn; q->prep_rq_fn = NULL; q->unplug_fn = generic_unplug_device; - q->queue_flags = (1 << QUEUE_FLAG_CLUSTER); + q->queue_flags = (1 << QUEUE_FLAG_CLUSTER | + 1 << QUEUE_FLAG_STACKABLE); q->queue_lock = lock; blk_queue_segment_boundary(q, 0xffffffff); -- cgit v1.2.3 From a68bbddba486020c9c74825ce90c4c1ec463e0e8 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 24 Sep 2008 13:03:33 +0200 Subject: block: add queue flag for SSD/non-rotational devices We don't want to idle in AS/CFQ if the device doesn't have a seek penalty. So add a QUEUE_FLAG_NONROT to indicate a non-rotational device, low level drivers should set this flag upon discovery of an SSD or similar device type. Signed-off-by: Jens Axboe --- block/as-iosched.c | 6 ++++++ block/cfq-iosched.c | 6 ++++++ 2 files changed, 12 insertions(+) (limited to 'block') diff --git a/block/as-iosched.c b/block/as-iosched.c index 80af9257e64..4c6fafbba93 100644 --- a/block/as-iosched.c +++ b/block/as-iosched.c @@ -745,6 +745,12 @@ static int as_can_break_anticipation(struct as_data *ad, struct request *rq) */ static int as_can_anticipate(struct as_data *ad, struct request *rq) { + /* + * SSD device without seek penalty, disable idling + */ + if (blk_queue_nonrot(ad->q)) + return 0; + if (!ad->io_context) /* * Last request submitted was a write diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 494b6fdcb18..03a5953bb5d 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -878,6 +878,12 @@ static void cfq_arm_slice_timer(struct cfq_data *cfqd) struct cfq_io_context *cic; unsigned long sl; + /* + * SSD device without seek penalty, disable idling + */ + if (blk_queue_nonrot(cfqd->queue)) + return; + WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); WARN_ON(cfq_cfqq_slice_new(cfqq)); -- cgit v1.2.3 From f7d7b7a7a3db6526a84ea755c1c54a051e9a52de Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 25 Sep 2008 11:37:50 +0200 Subject: block: as/cfq ssd idle check update We really need to know about the hardware tagging support as well, since if the SSD does not do tagging then we still want to idle. Otherwise have the same dependent sync IO vs flooding async IO problem as on rotational media. Signed-off-by: Jens Axboe --- block/as-iosched.c | 4 +++- block/cfq-iosched.c | 6 ++++-- 2 files changed, 7 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/as-iosched.c b/block/as-iosched.c index 4c6fafbba93..71f0abb219e 100644 --- a/block/as-iosched.c +++ b/block/as-iosched.c @@ -745,11 +745,13 @@ static int as_can_break_anticipation(struct as_data *ad, struct request *rq) */ static int as_can_anticipate(struct as_data *ad, struct request *rq) { +#if 0 /* disable for now, we need to check tag level as well */ /* * SSD device without seek penalty, disable idling */ - if (blk_queue_nonrot(ad->q)) + if (blk_queue_nonrot(ad->q)) axman return 0; +#endif if (!ad->io_context) /* diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 03a5953bb5d..6a062eebbd1 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -879,9 +879,11 @@ static void cfq_arm_slice_timer(struct cfq_data *cfqd) unsigned long sl; /* - * SSD device without seek penalty, disable idling + * SSD device without seek penalty, disable idling. But only do so + * for devices that support queuing, otherwise we still have a problem + * with sync vs async workloads. */ - if (blk_queue_nonrot(cfqd->queue)) + if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag) return; WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); -- cgit v1.2.3 From e3ba9ae58a5599226e3976b29c8093041ae7c332 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 25 Sep 2008 11:42:41 +0200 Subject: block: reserve some tags just for sync IO By only allowing async IO to consume 3/4 ths of the tag depth, we always have slots free to serve sync IO. This is important to avoid having writes fill the entire tag queue, thus starving reads. Original patch and idea from Linus Torvalds Signed-off-by: Jens Axboe --- block/blk-tag.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/blk-tag.c b/block/blk-tag.c index 8a99688eb1b..c0d419e84ce 100644 --- a/block/blk-tag.c +++ b/block/blk-tag.c @@ -337,6 +337,7 @@ EXPORT_SYMBOL(blk_queue_end_tag); int blk_queue_start_tag(struct request_queue *q, struct request *rq) { struct blk_queue_tag *bqt = q->queue_tags; + unsigned max_depth, offset; int tag; if (unlikely((rq->cmd_flags & REQ_QUEUED))) { @@ -350,10 +351,19 @@ int blk_queue_start_tag(struct request_queue *q, struct request *rq) /* * Protect against shared tag maps, as we may not have exclusive * access to the tag map. + * + * We reserve a few tags just for sync IO, since we don't want + * to starve sync IO on behalf of flooding async IO. */ + max_depth = bqt->max_depth; + if (rq_is_sync(rq)) + offset = 0; + else + offset = max_depth >> 2; + do { - tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth); - if (tag >= bqt->max_depth) + tag = find_next_zero_bit(bqt->tag_map, max_depth, offset); + if (tag >= max_depth) return 1; } while (test_and_set_bit_lock(tag, bqt->tag_map)); -- cgit v1.2.3 From 336c3d8ce771608815b65bcfa27a17a83b297328 Mon Sep 17 00:00:00 2001 From: Elias Oltmanns Date: Wed, 1 Oct 2008 16:02:33 +0200 Subject: block: Fix blk_start_queueing() to not kick a stopped queue blk_start_queueing() should act like the generic queue unplugging and kicking and ignore a stopped queue. Such a queue may not be run until after a call to blk_start_queue(). Signed-off-by: Elias Oltmanns Signed-off-by: Jens Axboe --- block/blk-core.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index fa212348c4c..c66333d8e48 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -890,9 +890,11 @@ EXPORT_SYMBOL(blk_get_request); */ void blk_start_queueing(struct request_queue *q) { - if (!blk_queue_plugged(q)) + if (!blk_queue_plugged(q)) { + if (unlikely(blk_queue_stopped(q))) + return; q->request_fn(q); - else + } else __generic_unplug_device(q); } EXPORT_SYMBOL(blk_start_queueing); -- cgit v1.2.3 From ef9e3facdf1fe1228721a7c295a76d1b7a0e57ec Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Wed, 1 Oct 2008 16:12:15 +0200 Subject: block: add lld busy state exporting interface This patch adds an new interface, blk_lld_busy(), to check lld's busy state from the block layer. blk_lld_busy() calls down into low-level drivers for the checking if the drivers set q->lld_busy_fn() using blk_queue_lld_busy(). This resolves a performance problem on request stacking devices below. Some drivers like scsi mid layer stop dispatching request when they detect busy state on its low-level device like host/target/device. It allows other requests to stay in the I/O scheduler's queue for a chance of merging. Request stacking drivers like request-based dm should follow the same logic. However, there is no generic interface for the stacked device to check if the underlying device(s) are busy. If the request stacking driver dispatches and submits requests to the busy underlying device, the requests will stay in the underlying device's queue without a chance of merging. This causes performance problem on burst I/O load. With this patch, busy state of the underlying device is exported via q->lld_busy_fn(). So the request stacking driver can check it and stop dispatching requests if busy. The underlying device driver must return the busy state appropriately: 1: when the device driver can't process requests immediately. 0: when the device driver can process requests immediately, including abnormal situations where the device driver needs to kill all requests. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Cc: Andrew Morton Signed-off-by: Jens Axboe --- block/blk-core.c | 28 ++++++++++++++++++++++++++++ block/blk-settings.c | 6 ++++++ 2 files changed, 34 insertions(+) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index c66333d8e48..b2d0ac8b760 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -2100,6 +2100,34 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq, rq->rq_disk = bio->bi_bdev->bd_disk; } +/** + * blk_lld_busy - Check if underlying low-level drivers of a device are busy + * @q : the queue of the device being checked + * + * Description: + * Check if underlying low-level drivers of a device are busy. + * If the drivers want to export their busy state, they must set own + * exporting function using blk_queue_lld_busy() first. + * + * Basically, this function is used only by request stacking drivers + * to stop dispatching requests to underlying devices when underlying + * devices are busy. This behavior helps more I/O merging on the queue + * of the request stacking driver and prevents I/O throughput regression + * on burst I/O load. + * + * Return: + * 0 - Not busy (The request stacking driver should dispatch request) + * 1 - Busy (The request stacking driver should stop dispatching request) + */ +int blk_lld_busy(struct request_queue *q) +{ + if (q->lld_busy_fn) + return q->lld_busy_fn(q); + + return 0; +} +EXPORT_SYMBOL_GPL(blk_lld_busy); + int kblockd_schedule_work(struct request_queue *q, struct work_struct *work) { return queue_work(kblockd_workqueue, work); diff --git a/block/blk-settings.c b/block/blk-settings.c index 1d0330d0b40..b21dcdb6415 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -89,6 +89,12 @@ void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn) } EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out); +void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn) +{ + q->lld_busy_fn = fn; +} +EXPORT_SYMBOL_GPL(blk_queue_lld_busy); + /** * blk_queue_make_request - define an alternate make_request function for a device * @q: the request queue for the device to be affected -- cgit v1.2.3 From 0497b345e7d067109e0dd9bf9f4978a6847ee13b Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 1 Oct 2008 16:16:25 +0200 Subject: blktrace: use BLKTRACE_BDEV_SIZE as the name size for setup structure Define as 32, which is is what BDEVNAME_SIZE is/was as well. This keeps the user interface the same and gets rid of the difference between kernel and user api here. Signed-off-by: Jens Axboe --- block/blktrace.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/blktrace.c b/block/blktrace.c index 9e0212c90b2..85049a7e7a1 100644 --- a/block/blktrace.c +++ b/block/blktrace.c @@ -369,7 +369,8 @@ int do_blk_trace_setup(struct request_queue *q, char *name, dev_t dev, if (!buts->buf_size || !buts->buf_nr) return -EINVAL; - strcpy(buts->name, name); + strncpy(buts->name, name, BLKTRACE_BDEV_SIZE); + buts->name[BLKTRACE_BDEV_SIZE - 1] = '\0'; /* * some device names have larger paths - convert the slashes -- cgit v1.2.3 From 99cd3386f290eaf61f2b7596d5a4cc2007771174 Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Wed, 1 Oct 2008 10:13:44 -0400 Subject: block: change elevator to use __blk_end_request() This patch converts elevator to use __blk_end_request() directly so that end_{queued|dequeued}_request() can be removed. Related 'uptodate' arguments is converted to 'error'. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Jens Axboe --- block/elevator.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/elevator.c b/block/elevator.c index 8a74eedc353..04518921db3 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -754,7 +754,7 @@ struct request *elv_next_request(struct request_queue *q) * not ever see it. */ if (blk_empty_barrier(rq)) { - end_queued_request(rq, 1); + __blk_end_request(rq, 0, blk_rq_bytes(rq)); continue; } if (!(rq->cmd_flags & REQ_STARTED)) { @@ -825,7 +825,7 @@ struct request *elv_next_request(struct request_queue *q) break; } else if (ret == BLKPREP_KILL) { rq->cmd_flags |= REQ_QUIET; - end_queued_request(rq, 0); + __blk_end_request(rq, -EIO, blk_rq_bytes(rq)); } else { printk(KERN_ERR "%s: bad return=%d\n", __func__, ret); break; @@ -922,7 +922,7 @@ void elv_abort_queue(struct request_queue *q) rq = list_entry_rq(q->queue_head.next); rq->cmd_flags |= REQ_QUIET; blk_add_trace_rq(q, rq, BLK_TA_ABORT); - end_queued_request(rq, 0); + __blk_end_request(rq, -EIO, blk_rq_bytes(rq)); } } EXPORT_SYMBOL(elv_abort_queue); -- cgit v1.2.3 From d00e29fd99dd63d1c51917604e35dee824ed567f Mon Sep 17 00:00:00 2001 From: Kiyoshi Ueda Date: Wed, 1 Oct 2008 10:14:46 -0400 Subject: block: remove end_{queued|dequeued}_request() This patch removes end_queued_request() and end_dequeued_request(), which are no longer used. As a results, users of __end_request() became only end_request(). So the actual code in __end_request() is moved to end_request() and __end_request() is removed. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Jens Axboe --- block/blk-core.c | 58 +++++++------------------------------------------------- 1 file changed, 7 insertions(+), 51 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index b2d0ac8b760..2d053b58441 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1790,17 +1790,6 @@ static void end_that_request_last(struct request *req, int error) } } -static inline void __end_request(struct request *rq, int uptodate, - unsigned int nr_bytes) -{ - int error = 0; - - if (uptodate <= 0) - error = uptodate ? uptodate : -EIO; - - __blk_end_request(rq, error, nr_bytes); -} - /** * blk_rq_bytes - Returns bytes left to complete in the entire request * @rq: the request being processed @@ -1830,41 +1819,6 @@ unsigned int blk_rq_cur_bytes(struct request *rq) } EXPORT_SYMBOL_GPL(blk_rq_cur_bytes); -/** - * end_queued_request - end all I/O on a queued request - * @rq: the request being processed - * @uptodate: error value or %0/%1 uptodate flag - * - * Description: - * Ends all I/O on a request, and removes it from the block layer queues. - * Not suitable for normal I/O completion, unless the driver still has - * the request attached to the block layer. - * - **/ -void end_queued_request(struct request *rq, int uptodate) -{ - __end_request(rq, uptodate, blk_rq_bytes(rq)); -} -EXPORT_SYMBOL(end_queued_request); - -/** - * end_dequeued_request - end all I/O on a dequeued request - * @rq: the request being processed - * @uptodate: error value or %0/%1 uptodate flag - * - * Description: - * Ends all I/O on a request. The request must already have been - * dequeued using blkdev_dequeue_request(), as is normally the case - * for most drivers. - * - **/ -void end_dequeued_request(struct request *rq, int uptodate) -{ - __end_request(rq, uptodate, blk_rq_bytes(rq)); -} -EXPORT_SYMBOL(end_dequeued_request); - - /** * end_request - end I/O on the current segment of the request * @req: the request being processed @@ -1879,14 +1833,16 @@ EXPORT_SYMBOL(end_dequeued_request); * they have a residual value to account for. For that case this function * isn't really useful, unless the residual just happens to be the * full current segment. In other words, don't use this function in new - * code. Use blk_end_request() or __blk_end_request() to end partial parts - * of a request, or end_dequeued_request() and end_queued_request() to - * completely end IO on a dequeued/queued request. - * + * code. Use blk_end_request() or __blk_end_request() to end a request. **/ void end_request(struct request *req, int uptodate) { - __end_request(req, uptodate, req->hard_cur_sectors << 9); + int error = 0; + + if (uptodate <= 0) + error = uptodate ? uptodate : -EIO; + + __blk_end_request(req, error, req->hard_cur_sectors << 9); } EXPORT_SYMBOL(end_request); -- cgit v1.2.3 From 0c032ab889e7b20b8a5a7d09313e4aca214a15f7 Mon Sep 17 00:00:00 2001 From: "Martin K. Petersen" Date: Wed, 1 Oct 2008 03:38:38 -0400 Subject: block: Fix double put in blk_integrity_unregister - kobject_del already puts the parent. - Set integrity profile to NULL to prevent stale data. Signed-off-by: Martin K. Petersen Signed-off-by: Jens Axboe --- block/blk-integrity.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'block') diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 69023da6315..e3817a016a1 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -376,7 +376,7 @@ void blk_integrity_unregister(struct gendisk *disk) kobject_uevent(&bi->kobj, KOBJ_REMOVE); kobject_del(&bi->kobj); - kobject_put(&disk_to_dev(disk)->kobj); kmem_cache_free(integrity_cachep, bi); + disk->integrity = NULL; } EXPORT_SYMBOL(blk_integrity_unregister); -- cgit v1.2.3 From ad7fce93147d32ae53d25d9ea1a8ba31a239deee Mon Sep 17 00:00:00 2001 From: "Martin K. Petersen" Date: Wed, 1 Oct 2008 03:38:39 -0400 Subject: block: Switch blk_integrity_compare from bdev to gendisk The DM and MD integrity support now depends on being able to use gendisks instead of block_devices when comparing integrity profiles. Change function parameters accordingly. Also update comparison logic so that two NULL profiles are a valid configuration. Signed-off-by: Martin K. Petersen Signed-off-by: Jens Axboe --- block/blk-integrity.c | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'block') diff --git a/block/blk-integrity.c b/block/blk-integrity.c index e3817a016a1..61a8e2f8fdd 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -108,51 +108,51 @@ new_segment: EXPORT_SYMBOL(blk_rq_map_integrity_sg); /** - * blk_integrity_compare - Compare integrity profile of two block devices - * @bd1: Device to compare - * @bd2: Device to compare + * blk_integrity_compare - Compare integrity profile of two disks + * @gd1: Disk to compare + * @gd2: Disk to compare * * Description: Meta-devices like DM and MD need to verify that all * sub-devices use the same integrity format before advertising to * upper layers that they can send/receive integrity metadata. This - * function can be used to check whether two block devices have + * function can be used to check whether two gendisk devices have * compatible integrity formats. */ -int blk_integrity_compare(struct block_device *bd1, struct block_device *bd2) +int blk_integrity_compare(struct gendisk *gd1, struct gendisk *gd2) { - struct blk_integrity *b1 = bd1->bd_disk->integrity; - struct blk_integrity *b2 = bd2->bd_disk->integrity; + struct blk_integrity *b1 = gd1->integrity; + struct blk_integrity *b2 = gd2->integrity; - BUG_ON(bd1->bd_disk == NULL); - BUG_ON(bd2->bd_disk == NULL); + if (!b1 && !b2) + return 0; if (!b1 || !b2) - return 0; + return -1; if (b1->sector_size != b2->sector_size) { printk(KERN_ERR "%s: %s/%s sector sz %u != %u\n", __func__, - bd1->bd_disk->disk_name, bd2->bd_disk->disk_name, + gd1->disk_name, gd2->disk_name, b1->sector_size, b2->sector_size); return -1; } if (b1->tuple_size != b2->tuple_size) { printk(KERN_ERR "%s: %s/%s tuple sz %u != %u\n", __func__, - bd1->bd_disk->disk_name, bd2->bd_disk->disk_name, + gd1->disk_name, gd2->disk_name, b1->tuple_size, b2->tuple_size); return -1; } if (b1->tag_size && b2->tag_size && (b1->tag_size != b2->tag_size)) { printk(KERN_ERR "%s: %s/%s tag sz %u != %u\n", __func__, - bd1->bd_disk->disk_name, bd2->bd_disk->disk_name, + gd1->disk_name, gd2->disk_name, b1->tag_size, b2->tag_size); return -1; } if (strcmp(b1->name, b2->name)) { printk(KERN_ERR "%s: %s/%s type %s != %s\n", __func__, - bd1->bd_disk->disk_name, bd2->bd_disk->disk_name, + gd1->disk_name, gd2->disk_name, b1->name, b2->name); return -1; } -- cgit v1.2.3