From 213d9417fec62ef4c3675621b9364a667954d4dd Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 6 Jan 2009 09:16:05 +0100 Subject: block: seperate bio/request unplug and sync bits Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/linux/blkdev.h') diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 044467ef7b1..75426e4b8cd 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -108,6 +108,7 @@ enum rq_flag_bits { __REQ_RW_META, /* metadata io request */ __REQ_COPY_USER, /* contains copies of user pages */ __REQ_INTEGRITY, /* integrity metadata has been remapped */ + __REQ_UNPLUG, /* unplug queue on submission */ __REQ_NR_BITS, /* stops here */ }; @@ -134,6 +135,7 @@ enum rq_flag_bits { #define REQ_RW_META (1 << __REQ_RW_META) #define REQ_COPY_USER (1 << __REQ_COPY_USER) #define REQ_INTEGRITY (1 << __REQ_INTEGRITY) +#define REQ_UNPLUG (1 << __REQ_UNPLUG) #define BLK_MAX_CDB 16 -- cgit v1.2.3-70-g09d2 From bc58ba9468d94d62c56ab9b47173583ec140b165 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 23 Jan 2009 10:54:44 +0100 Subject: block: add sysfs file for controlling io stats accounting This allows us to turn off disk stat accounting completely, for the cases where the 0.5-1% reduction in system time is important. Signed-off-by: Jens Axboe --- block/blk-core.c | 90 ++++++++++++++++++++++++++++++-------------------- block/blk-sysfs.c | 28 ++++++++++++++++ include/linux/blkdev.h | 6 ++++ 3 files changed, 88 insertions(+), 36 deletions(-) (limited to 'include/linux/blkdev.h') diff --git a/block/blk-core.c b/block/blk-core.c index ae75c047f45..ca69f3d9410 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -64,11 +64,12 @@ static struct workqueue_struct *kblockd_workqueue; static void drive_stat_acct(struct request *rq, int new_io) { + struct gendisk *disk = rq->rq_disk; struct hd_struct *part; int rw = rq_data_dir(rq); int cpu; - if (!blk_fs_request(rq) || !rq->rq_disk) + if (!blk_fs_request(rq) || !disk || !blk_queue_io_stat(disk->queue)) return; cpu = part_stat_lock(); @@ -599,8 +600,7 @@ 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 | - 1 << QUEUE_FLAG_STACKABLE); + q->queue_flags = QUEUE_FLAG_DEFAULT; q->queue_lock = lock; blk_queue_segment_boundary(q, BLK_SEG_BOUNDARY_MASK); @@ -1663,6 +1663,55 @@ void blkdev_dequeue_request(struct request *req) } EXPORT_SYMBOL(blkdev_dequeue_request); +static void blk_account_io_completion(struct request *req, unsigned int bytes) +{ + struct gendisk *disk = req->rq_disk; + + if (!disk || !blk_queue_io_stat(disk->queue)) + return; + + if (blk_fs_request(req)) { + const int rw = rq_data_dir(req); + struct hd_struct *part; + int cpu; + + cpu = part_stat_lock(); + part = disk_map_sector_rcu(req->rq_disk, req->sector); + part_stat_add(cpu, part, sectors[rw], bytes >> 9); + part_stat_unlock(); + } +} + +static void blk_account_io_done(struct request *req) +{ + struct gendisk *disk = req->rq_disk; + + if (!disk || !blk_queue_io_stat(disk->queue)) + return; + + /* + * Account IO completion. bar_rq isn't accounted as a normal + * IO on queueing nor completion. Accounting the containing + * request is enough. + */ + if (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; + int cpu; + + cpu = part_stat_lock(); + part = disk_map_sector_rcu(disk, req->sector); + + 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); + + part_stat_unlock(); + } +} + /** * __end_that_request_first - end I/O on a request * @req: the request being processed @@ -1698,16 +1747,7 @@ static int __end_that_request_first(struct request *req, int error, (unsigned long long)req->sector); } - if (blk_fs_request(req) && req->rq_disk) { - const int rw = rq_data_dir(req); - struct hd_struct *part; - int cpu; - - cpu = part_stat_lock(); - part = disk_map_sector_rcu(req->rq_disk, req->sector); - part_stat_add(cpu, part, sectors[rw], nr_bytes >> 9); - part_stat_unlock(); - } + blk_account_io_completion(req, nr_bytes); total_bytes = bio_nbytes = 0; while ((bio = req->bio) != NULL) { @@ -1787,8 +1827,6 @@ static int __end_that_request_first(struct request *req, int error, */ static void end_that_request_last(struct request *req, int error) { - struct gendisk *disk = req->rq_disk; - if (blk_rq_tagged(req)) blk_queue_end_tag(req->q, req); @@ -1800,27 +1838,7 @@ static void end_that_request_last(struct request *req, int error) blk_delete_timer(req); - /* - * Account IO completion. bar_rq isn't accounted as a normal - * IO on queueing nor completion. Accounting the containing - * request is enough. - */ - 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; - int cpu; - - cpu = part_stat_lock(); - part = disk_map_sector_rcu(disk, req->sector); - - 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); - - part_stat_unlock(); - } + blk_account_io_done(req); if (req->end_io) req->end_io(req, error); diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index b538ab8dbbd..e29ddfc73cf 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -197,6 +197,27 @@ queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count) return ret; } +static ssize_t queue_iostats_show(struct request_queue *q, char *page) +{ + return queue_var_show(blk_queue_io_stat(q), page); +} + +static ssize_t queue_iostats_store(struct request_queue *q, const char *page, + size_t count) +{ + unsigned long stats; + ssize_t ret = queue_var_store(&stats, page, count); + + spin_lock_irq(q->queue_lock); + if (stats) + queue_flag_set(QUEUE_FLAG_IO_STAT, q); + else + queue_flag_clear(QUEUE_FLAG_IO_STAT, q); + spin_unlock_irq(q->queue_lock); + + return ret; +} + static struct queue_sysfs_entry queue_requests_entry = { .attr = {.name = "nr_requests", .mode = S_IRUGO | S_IWUSR }, .show = queue_requests_show, @@ -249,6 +270,12 @@ static struct queue_sysfs_entry queue_rq_affinity_entry = { .store = queue_rq_affinity_store, }; +static struct queue_sysfs_entry queue_iostats_entry = { + .attr = {.name = "iostats", .mode = S_IRUGO | S_IWUSR }, + .show = queue_iostats_show, + .store = queue_iostats_store, +}; + static struct attribute *default_attrs[] = { &queue_requests_entry.attr, &queue_ra_entry.attr, @@ -259,6 +286,7 @@ static struct attribute *default_attrs[] = { &queue_nonrot_entry.attr, &queue_nomerges_entry.attr, &queue_rq_affinity_entry.attr, + &queue_iostats_entry.attr, NULL, }; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 75426e4b8cd..d08c4b8219a 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -451,6 +451,11 @@ struct request_queue #define QUEUE_FLAG_STACKABLE 13 /* supports request stacking */ #define QUEUE_FLAG_NONROT 14 /* non-rotational device (SSD) */ #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ +#define QUEUE_FLAG_IO_STAT 15 /* do IO stats */ + +#define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ + (1 << QUEUE_FLAG_CLUSTER) | \ + 1 << QUEUE_FLAG_STACKABLE) static inline int queue_is_locked(struct request_queue *q) { @@ -567,6 +572,7 @@ enum { #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) +#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) #define blk_queue_flushing(q) ((q)->ordseq) #define blk_queue_stackable(q) \ test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags) -- cgit v1.2.3-70-g09d2 From 0648e10d71c8e510d80772c4cb4220c97e9c34c7 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 2 Feb 2009 08:43:48 +0100 Subject: block: fix inconsistent parenthesisation of QUEUE_FLAG_DEFAULT Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include/linux/blkdev.h') diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index d08c4b8219a..dcaa0fd84b0 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -455,7 +455,7 @@ struct request_queue #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ (1 << QUEUE_FLAG_CLUSTER) | \ - 1 << QUEUE_FLAG_STACKABLE) + (1 << QUEUE_FLAG_STACKABLE)) static inline int queue_is_locked(struct request_queue *q) { -- cgit v1.2.3-70-g09d2 From 1e42807918d17e8c93bf14fbb74be84b141334c1 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 23 Feb 2009 09:03:10 +0100 Subject: block: reduce stack footprint of blk_recount_segments() blk_recalc_rq_segments() requires a request structure passed in, which we don't have from blk_recount_segments(). So the latter allocates one on the stack, using > 400 bytes of stack for that. This can cause us to spill over one page of stack from ext4 at least: 0) 4560 400 blk_recount_segments+0x43/0x62 1) 4160 32 bio_phys_segments+0x1c/0x24 2) 4128 32 blk_rq_bio_prep+0x2a/0xf9 3) 4096 32 init_request_from_bio+0xf9/0xfe 4) 4064 112 __make_request+0x33c/0x3f6 5) 3952 144 generic_make_request+0x2d1/0x321 6) 3808 64 submit_bio+0xb9/0xc3 7) 3744 48 submit_bh+0xea/0x10e 8) 3696 368 ext4_mb_init_cache+0x257/0xa6a [ext4] 9) 3328 288 ext4_mb_regular_allocator+0x421/0xcd9 [ext4] 10) 3040 160 ext4_mb_new_blocks+0x211/0x4b4 [ext4] 11) 2880 336 ext4_ext_get_blocks+0xb61/0xd45 [ext4] 12) 2544 96 ext4_get_blocks_wrap+0xf2/0x200 [ext4] 13) 2448 80 ext4_da_get_block_write+0x6e/0x16b [ext4] 14) 2368 352 mpage_da_map_blocks+0x7e/0x4b3 [ext4] 15) 2016 352 ext4_da_writepages+0x2ce/0x43c [ext4] 16) 1664 32 do_writepages+0x2d/0x3c 17) 1632 144 __writeback_single_inode+0x162/0x2cd 18) 1488 96 generic_sync_sb_inodes+0x1e3/0x32b 19) 1392 16 sync_sb_inodes+0xe/0x10 20) 1376 48 writeback_inodes+0x69/0xb3 21) 1328 208 balance_dirty_pages_ratelimited_nr+0x187/0x2f9 22) 1120 224 generic_file_buffered_write+0x1d4/0x2c4 23) 896 176 __generic_file_aio_write_nolock+0x35f/0x393 24) 720 80 generic_file_aio_write+0x6c/0xc8 25) 640 80 ext4_file_write+0xa9/0x137 [ext4] 26) 560 320 do_sync_write+0xf0/0x137 27) 240 48 vfs_write+0xb3/0x13c 28) 192 64 sys_write+0x4c/0x74 29) 128 128 system_call_fastpath+0x16/0x1b Split the segment counting out into a __blk_recalc_rq_segments() helper to avoid allocating an onstack request just for checking the physical segment count. Signed-off-by: Jens Axboe --- block/blk-merge.c | 94 ++++++++++++++++++++++++++++---------------------- include/linux/blkdev.h | 2 ++ 2 files changed, 55 insertions(+), 41 deletions(-) (limited to 'include/linux/blkdev.h') diff --git a/block/blk-merge.c b/block/blk-merge.c index b92f5b0866b..a104593e70c 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -38,72 +38,84 @@ void blk_recalc_rq_sectors(struct request *rq, int nsect) } } -void blk_recalc_rq_segments(struct request *rq) +static unsigned int __blk_recalc_rq_segments(struct request_queue *q, + struct bio *bio, + unsigned int *seg_size_ptr) { - int nr_phys_segs; unsigned int phys_size; struct bio_vec *bv, *bvprv = NULL; - int seg_size; - int cluster; - struct req_iterator iter; - int high, highprv = 1; - struct request_queue *q = rq->q; + int cluster, i, high, highprv = 1; + unsigned int seg_size, nr_phys_segs; + struct bio *fbio; - if (!rq->bio) - return; + if (!bio) + return 0; + fbio = bio; cluster = test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags); 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 - * considered part of another segment, since that might - * change with the bounce page. - */ - high = page_to_pfn(bv->bv_page) > q->bounce_pfn; - if (high || highprv) - goto new_segment; - if (cluster) { - if (seg_size + bv->bv_len > q->max_segment_size) - goto new_segment; - if (!BIOVEC_PHYS_MERGEABLE(bvprv, bv)) - goto new_segment; - if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bv)) + for_each_bio(bio) { + bio_for_each_segment(bv, bio, i) { + /* + * the trick here is making sure that a high page is + * never considered part of another segment, since that + * might change with the bounce page. + */ + high = page_to_pfn(bv->bv_page) > q->bounce_pfn; + if (high || highprv) goto new_segment; + if (cluster) { + if (seg_size + bv->bv_len > q->max_segment_size) + goto new_segment; + if (!BIOVEC_PHYS_MERGEABLE(bvprv, bv)) + goto new_segment; + if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bv)) + goto new_segment; + + seg_size += bv->bv_len; + bvprv = bv; + continue; + } +new_segment: + if (nr_phys_segs == 1 && seg_size > + fbio->bi_seg_front_size) + fbio->bi_seg_front_size = seg_size; - seg_size += bv->bv_len; + nr_phys_segs++; bvprv = bv; - continue; + seg_size = bv->bv_len; + highprv = high; } -new_segment: - if (nr_phys_segs == 1 && seg_size > rq->bio->bi_seg_front_size) - rq->bio->bi_seg_front_size = seg_size; - - nr_phys_segs++; - bvprv = bv; - seg_size = bv->bv_len; - highprv = high; } - if (nr_phys_segs == 1 && seg_size > rq->bio->bi_seg_front_size) + if (seg_size_ptr) + *seg_size_ptr = seg_size; + + return nr_phys_segs; +} + +void blk_recalc_rq_segments(struct request *rq) +{ + unsigned int seg_size = 0, phys_segs; + + phys_segs = __blk_recalc_rq_segments(rq->q, rq->bio, &seg_size); + + if (phys_segs == 1 && seg_size > rq->bio->bi_seg_front_size) rq->bio->bi_seg_front_size = seg_size; if (seg_size > rq->biotail->bi_seg_back_size) rq->biotail->bi_seg_back_size = seg_size; - rq->nr_phys_segments = nr_phys_segs; + rq->nr_phys_segments = phys_segs; } void blk_recount_segments(struct request_queue *q, struct bio *bio) { - struct request rq; struct bio *nxt = bio->bi_next; - rq.q = q; - rq.bio = rq.biotail = bio; + bio->bi_next = NULL; - blk_recalc_rq_segments(&rq); + bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio, NULL); bio->bi_next = nxt; - bio->bi_phys_segments = rq.nr_phys_segments; bio->bi_flags |= (1 << BIO_SEG_VALID); } EXPORT_SYMBOL(blk_recount_segments); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index dcaa0fd84b0..465d6babc84 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -708,6 +708,8 @@ struct req_iterator { }; /* This should not be used directly - use rq_for_each_segment */ +#define for_each_bio(_bio) \ + for (; _bio; _bio = _bio->bi_next) #define __rq_for_each_bio(_bio, rq) \ if ((rq->bio)) \ for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) -- cgit v1.2.3-70-g09d2