From c1e44756fdb7b363fd22cb5514dced40752e36c5 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Mon, 8 Nov 2010 15:01:02 +0100 Subject: cfq-iosched: do cleanup Some functions should return boolean. Signed-off-by: Shaohua Li Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 9eba291eb6fd..b8174bb4a6a1 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -637,11 +637,11 @@ cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) static inline bool cfq_slice_used(struct cfq_queue *cfqq) { if (cfq_cfqq_slice_new(cfqq)) - return 0; + return false; if (time_before(jiffies, cfqq->slice_end)) - return 0; + return false; - return 1; + return true; } /* @@ -1892,10 +1892,10 @@ static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq) * in their service tree. */ if (service_tree->count == 1 && cfq_cfqq_sync(cfqq)) - return 1; + return true; cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", service_tree->count); - return 0; + return false; } static void cfq_arm_slice_timer(struct cfq_data *cfqd) @@ -2359,12 +2359,12 @@ static inline bool cfq_slice_used_soon(struct cfq_data *cfqd, { /* the queue hasn't finished any request, can't estimate */ if (cfq_cfqq_slice_new(cfqq)) - return 1; + return true; if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched, cfqq->slice_end)) - return 1; + return true; - return 0; + return false; } static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq) -- cgit v1.2.1 From d2d59e18a1ea8ecdd1c0a52af320e9a7f5391cc4 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Mon, 8 Nov 2010 15:01:03 +0100 Subject: cfq-iosched: schedule dispatch for noidle queue A queue is idle at cfq_dispatch_requests(), but it gets noidle later. Unless other task explictly does unplug or all requests are drained, we will not deliever requests to the disk even cfq_arm_slice_timer doesn't make the queue idle. For example, cfq_should_idle() returns true because of service_tree->count == 1, and then other queues are added. Note, I didn't see obvious performance impacts so far with the patch, but just thought this could be a problem. Signed-off-by: Shaohua Li Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 23 ++++++++++++++++++++++- 1 file changed, 22 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index b8174bb4a6a1..986865e3fbc5 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -3255,6 +3255,10 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq, if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq)) return true; + /* An idle queue should not be idle now for some reason */ + if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq)) + return true; + if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq)) return false; @@ -3508,8 +3512,25 @@ static void cfq_completed_request(struct request_queue *q, struct request *rq) } } - if (!cfqd->rq_in_driver) + if (!cfqd->rq_in_driver) { + cfq_schedule_dispatch(cfqd); + return; + } + /* + * A queue is idle at cfq_dispatch_requests(), but it gets noidle + * later. We schedule a dispatch if the queue has no requests, + * otherwise the disk is actually in idle till all requests + * are finished even cfq_arm_slice_timer doesn't make the queue idle + * */ + cfqq = cfqd->active_queue; + if (!cfqq) + return; + + if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq) && + (!cfqd->cfq_group_idle || cfqq->cfqg->nr_cfqq > 1)) { + cfq_del_timer(cfqd, cfqq); cfq_schedule_dispatch(cfqd); + } } /* -- cgit v1.2.1 From 8e1ac6655104bc6e1e79d67e2df88cc8fa9b6e07 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Mon, 8 Nov 2010 15:01:04 +0100 Subject: cfq-iosched: don't idle if a deep seek queue is slow If a deep seek queue slowly deliver requests but disk is much faster, idle for the queue just wastes disk throughput. If the queue delevers all requests before half its slice is used, the patch disable idle for it. In my test, application delivers 32 requests one time, the disk can accept 128 requests at maxium and disk is fast. without the patch, the throughput is just around 30m/s, while with it, the speed is about 80m/s. The disk is a SSD, but is detected as a rotational disk. I can configure it as SSD, but I thought the deep seek queue logic should be fixed too, for example, considering a fast raid. Signed-off-by: Shaohua Li Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 986865e3fbc5..ca4d19907243 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -2285,6 +2285,17 @@ static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd) goto keep_queue; } + /* + * This is a deep seek queue, but the device is much faster than + * the queue can deliver, don't idle + **/ + if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) && + (cfq_cfqq_slice_new(cfqq) || + (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) { + cfq_clear_cfqq_deep(cfqq); + cfq_clear_cfqq_idle_window(cfqq); + } + if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) { cfqq = NULL; goto keep_queue; -- cgit v1.2.1 From 2b9408a45978dcda77407859148deeccf403c372 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Tue, 9 Nov 2010 14:51:13 +0100 Subject: cfq-iosched: don't schedule a dispatch for a non-idle queue Vivek suggests we don't need schedule a dispatch when an idle queue becomes nonidle. And he is right, cfq_should_preempt already covers the logic. Signed-off-by: Shaohua Li Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 19 +------------------ 1 file changed, 1 insertion(+), 18 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index ca4d19907243..f90519430be6 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -3523,25 +3523,8 @@ static void cfq_completed_request(struct request_queue *q, struct request *rq) } } - if (!cfqd->rq_in_driver) { + if (!cfqd->rq_in_driver) cfq_schedule_dispatch(cfqd); - return; - } - /* - * A queue is idle at cfq_dispatch_requests(), but it gets noidle - * later. We schedule a dispatch if the queue has no requests, - * otherwise the disk is actually in idle till all requests - * are finished even cfq_arm_slice_timer doesn't make the queue idle - * */ - cfqq = cfqd->active_queue; - if (!cfqq) - return; - - if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq) && - (!cfqd->cfq_group_idle || cfqq->cfqg->nr_cfqq > 1)) { - cfq_del_timer(cfqd, cfqq); - cfq_schedule_dispatch(cfqd); - } } /* -- cgit v1.2.1 From e525fd89d380c4a94c0d63913a1dd1a593ed25e7 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Sat, 13 Nov 2010 11:55:17 +0100 Subject: block: make blkdev_get/put() handle exclusive access Over time, block layer has accumulated a set of APIs dealing with bdev open, close, claim and release. * blkdev_get/put() are the primary open and close functions. * bd_claim/release() deal with exclusive open. * open/close_bdev_exclusive() are combination of open and claim and the other way around, respectively. * bd_link/unlink_disk_holder() to create and remove holder/slave symlinks. * open_by_devnum() wraps bdget() + blkdev_get(). The interface is a bit confusing and the decoupling of open and claim makes it impossible to properly guarantee exclusive access as in-kernel open + claim sequence can disturb the existing exclusive open even before the block layer knows the current open if for another exclusive access. Reorganize the interface such that, * blkdev_get() is extended to include exclusive access management. @holder argument is added and, if is @FMODE_EXCL specified, it will gain exclusive access atomically w.r.t. other exclusive accesses. * blkdev_put() is similarly extended. It now takes @mode argument and if @FMODE_EXCL is set, it releases an exclusive access. Also, when the last exclusive claim is released, the holder/slave symlinks are removed automatically. * bd_claim/release() and close_bdev_exclusive() are no longer necessary and either made static or removed. * bd_link_disk_holder() remains the same but bd_unlink_disk_holder() is no longer necessary and removed. * open_bdev_exclusive() becomes a simple wrapper around lookup_bdev() and blkdev_get(). It also has an unexpected extra bdev_read_only() test which probably should be moved into blkdev_get(). * open_by_devnum() is modified to take @holder argument and pass it to blkdev_get(). Most of bdev open/close operations are unified into blkdev_get/put() and most exclusive accesses are tested atomically at the open time (as it should). This cleans up code and removes some, both valid and invalid, but unnecessary all the same, corner cases. open_bdev_exclusive() and open_by_devnum() can use further cleanup - rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop special features. Well, let's leave them for another day. Most conversions are straight-forward. drbd conversion is a bit more involved as there was some reordering, but the logic should stay the same. Signed-off-by: Tejun Heo Acked-by: Neil Brown Acked-by: Ryusuke Konishi Acked-by: Mike Snitzer Acked-by: Philipp Reisner Cc: Peter Osterlund Cc: Martin Schwidefsky Cc: Heiko Carstens Cc: Jan Kara Cc: Andrew Morton Cc: Andreas Dilger Cc: "Theodore Ts'o" Cc: Mark Fasheh Cc: Joel Becker Cc: Alex Elder Cc: Christoph Hellwig Cc: dm-devel@redhat.com Cc: drbd-dev@lists.linbit.com Cc: Leo Chen Cc: Scott Branden Cc: Chris Mason Cc: Steven Whitehouse Cc: Dave Kleikamp Cc: Joern Engel Cc: reiserfs-devel@vger.kernel.org Cc: Alexander Viro --- block/ioctl.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'block') diff --git a/block/ioctl.c b/block/ioctl.c index d724ceb1d465..cc46d499fd27 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -294,11 +294,12 @@ int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, return -EINVAL; if (get_user(n, (int __user *) arg)) return -EFAULT; - if (!(mode & FMODE_EXCL) && bd_claim(bdev, &bdev) < 0) + if (!(mode & FMODE_EXCL) && + blkdev_get(bdev, mode | FMODE_EXCL, &bdev) < 0) return -EBUSY; ret = set_blocksize(bdev, n); if (!(mode & FMODE_EXCL)) - bd_release(bdev); + blkdev_put(bdev, mode | FMODE_EXCL); return ret; case BLKPG: ret = blkpg_ioctl(bdev, (struct blkpg_ioctl_arg __user *) arg); -- cgit v1.2.1 From bdc85df7a8417b9893443ff5520804699416b6f3 Mon Sep 17 00:00:00 2001 From: Vivek Goyal Date: Mon, 15 Nov 2010 19:37:36 +0100 Subject: blk-cgroup: Allow creation of hierarchical cgroups o Allow hierarchical cgroup creation for blkio controller o Currently we disallow it as both the io controller policies (throttling as well as proportion bandwidth) do not support hierarhical accounting and control. But the flip side is that blkio controller can not be used with libvirt as libvirt creates a cgroup hierarchy deeper than 1 level. //libvirt/qemu/ o So this patch will allow creation of cgroup hierarhcy but at the backend everything will be treated as flat. So if somebody created a an hierarchy like as follows. root / \ test1 test2 | test3 CFQ and throttling will practically treat all groups at same level. pivot / | \ \ root test1 test2 test3 o Once we have actual support for hierarchical accounting and control then we can introduce another cgroup tunable file "blkio.use_hierarchy" which will be 0 by default but if user wants to enforce hierarhical control then it can be set to 1. This way there should not be any ABI problems down the line. o The only not so pretty part is introduction of extra file "use_hierarchy" down the line. Kame-san had mentioned that hierarhical accounting is expensive in memory controller hence they keep it off by default. I suspect same will be the case for IO controller also as for each IO completion we shall have to account IO through hierarchy up to the root. if yes, then it probably is not a very bad idea to introduce this extra file so that it will be used only when somebody needs it and some people might enable hierarchy only in part of the hierarchy. o This is how basically memory controller also uses "use_hierarhcy" and they also allowed creation of hierarchies when actual backend support was not available. Signed-off-by: Vivek Goyal Acked-by: Balbir Singh Reviewed-by: Gui Jianfeng Reviewed-by: Ciju Rajan K Tested-by: Ciju Rajan K Signed-off-by: Jens Axboe --- block/blk-cgroup.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'block') diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c index b1febd0f6d2a..455768a3eb9e 100644 --- a/block/blk-cgroup.c +++ b/block/blk-cgroup.c @@ -1452,10 +1452,6 @@ blkiocg_create(struct cgroup_subsys *subsys, struct cgroup *cgroup) goto done; } - /* Currently we do not support hierarchy deeper than two level (0,1) */ - if (parent != cgroup->top_cgroup) - return ERR_PTR(-EPERM); - blkcg = kzalloc(sizeof(*blkcg), GFP_KERNEL); if (!blkcg) return ERR_PTR(-ENOMEM); -- cgit v1.2.1 From d07335e51df0c6dec202d315fc4f1f7e100eec4e Mon Sep 17 00:00:00 2001 From: Mike Snitzer Date: Tue, 16 Nov 2010 12:52:38 +0100 Subject: block: Rename "block_remap" tracepoint to "block_bio_remap" to clarify the event. Signed-off-by: Kiyoshi Ueda Signed-off-by: Jun'ichi Nomura Signed-off-by: Mike Snitzer Signed-off-by: Jens Axboe --- block/blk-core.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 4ce953f1b390..151070541e21 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -33,7 +33,7 @@ #include "blk.h" -EXPORT_TRACEPOINT_SYMBOL_GPL(block_remap); +EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); @@ -1329,9 +1329,9 @@ static inline void blk_partition_remap(struct bio *bio) bio->bi_sector += p->start_sect; bio->bi_bdev = bdev->bd_contains; - trace_block_remap(bdev_get_queue(bio->bi_bdev), bio, - bdev->bd_dev, - bio->bi_sector - p->start_sect); + trace_block_bio_remap(bdev_get_queue(bio->bi_bdev), bio, + bdev->bd_dev, + bio->bi_sector - p->start_sect); } } @@ -1500,7 +1500,7 @@ static inline void __generic_make_request(struct bio *bio) goto end_io; if (old_sector != -1) - trace_block_remap(q, bio, old_dev, old_sector); + trace_block_bio_remap(q, bio, old_dev, old_sector); old_sector = bio->bi_sector; old_dev = bio->bi_bdev->bd_dev; -- cgit v1.2.1 From b54ce60eb7f61f8e314b8b241b0469eda3bb1d42 Mon Sep 17 00:00:00 2001 From: Gui Jianfeng Date: Tue, 30 Nov 2010 20:52:46 +0100 Subject: cfq-iosched: Get rid of st->active When a cfq group is running, it won't be dequeued from service tree, so there's no need to store the active one in st->active. Just gid rid of it. Signed-off-by: Gui Jianfeng Acked-by: Vivek Goyal Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 15 +-------------- 1 file changed, 1 insertion(+), 14 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 73a58628f54a..e18d316ae652 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -87,7 +87,6 @@ struct cfq_rb_root { unsigned count; unsigned total_weight; u64 min_vdisktime; - struct rb_node *active; }; #define CFQ_RB_ROOT (struct cfq_rb_root) { .rb = RB_ROOT, .left = NULL, \ .count = 0, .min_vdisktime = 0, } @@ -563,11 +562,6 @@ static void update_min_vdisktime(struct cfq_rb_root *st) u64 vdisktime = st->min_vdisktime; struct cfq_group *cfqg; - if (st->active) { - cfqg = rb_entry_cfqg(st->active); - vdisktime = cfqg->vdisktime; - } - if (st->left) { cfqg = rb_entry_cfqg(st->left); vdisktime = min_vdisktime(vdisktime, cfqg->vdisktime); @@ -894,9 +888,6 @@ cfq_group_service_tree_del(struct cfq_data *cfqd, struct cfq_group *cfqg) { struct cfq_rb_root *st = &cfqd->grp_service_tree; - if (st->active == &cfqg->rb_node) - st->active = NULL; - BUG_ON(cfqg->nr_cfqq < 1); cfqg->nr_cfqq--; @@ -1095,7 +1086,7 @@ static void cfq_put_cfqg(struct cfq_group *cfqg) if (!atomic_dec_and_test(&cfqg->ref)) return; for_each_cfqg_st(cfqg, i, j, st) - BUG_ON(!RB_EMPTY_ROOT(&st->rb) || st->active != NULL); + BUG_ON(!RB_EMPTY_ROOT(&st->rb)); kfree(cfqg); } @@ -1687,9 +1678,6 @@ __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq, if (cfqq == cfqd->active_queue) cfqd->active_queue = NULL; - if (&cfqq->cfqg->rb_node == cfqd->grp_service_tree.active) - cfqd->grp_service_tree.active = NULL; - if (cfqd->active_cic) { put_io_context(cfqd->active_cic->ioc); cfqd->active_cic = NULL; @@ -2199,7 +2187,6 @@ static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd) if (RB_EMPTY_ROOT(&st->rb)) return NULL; cfqg = cfq_rb_first_group(st); - st->active = &cfqg->rb_node; update_min_vdisktime(st); return cfqg; } -- cgit v1.2.1 From 760701bfe14faee8ea0608a9cab2046071d98a39 Mon Sep 17 00:00:00 2001 From: Gui Jianfeng Date: Tue, 30 Nov 2010 20:52:47 +0100 Subject: cfq-iosched: Get rid of on_st flag It's able to check whether a CFQ group on a service tree by checking "cfqg->rb_node". There's no need to maintain an extra flag here. Signed-off-by: Gui Jianfeng Acked-by: Vivek Goyal Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index e18d316ae652..5d0349d602fe 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -179,7 +179,6 @@ struct cfq_group { /* group service_tree key */ u64 vdisktime; unsigned int weight; - bool on_st; /* number of cfqq currently on this group */ int nr_cfqq; @@ -863,7 +862,7 @@ cfq_group_service_tree_add(struct cfq_data *cfqd, struct cfq_group *cfqg) struct rb_node *n; cfqg->nr_cfqq++; - if (cfqg->on_st) + if (!RB_EMPTY_NODE(&cfqg->rb_node)) return; /* @@ -879,7 +878,6 @@ cfq_group_service_tree_add(struct cfq_data *cfqd, struct cfq_group *cfqg) cfqg->vdisktime = st->min_vdisktime; __cfq_group_service_tree_add(st, cfqg); - cfqg->on_st = true; st->total_weight += cfqg->weight; } @@ -896,7 +894,6 @@ cfq_group_service_tree_del(struct cfq_data *cfqd, struct cfq_group *cfqg) return; cfq_log_cfqg(cfqd, cfqg, "del_from_rr group"); - cfqg->on_st = false; st->total_weight -= cfqg->weight; if (!RB_EMPTY_NODE(&cfqg->rb_node)) cfq_rb_erase(&cfqg->rb_node, st); -- cgit v1.2.1 From e4ea0c16a85d221ebcc3a21f32e321440459e0fc Mon Sep 17 00:00:00 2001 From: Shaohua Li writes Date: Mon, 13 Dec 2010 14:32:22 +0100 Subject: block cfq: select new workload if priority changed If priority is changed, continuing to check workload_expires and service tree count of the previous workload does not make sense. We should always choose the workload with lowest key of new priority in such case. Signed-off-by: Shaohua Li Reviewed-by: Jeff Moyer Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 5d0349d602fe..9b186fd6bf47 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -2101,6 +2101,7 @@ static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg) unsigned count; struct cfq_rb_root *st; unsigned group_slice; + enum wl_prio_t original_prio = cfqd->serving_prio; if (!cfqg) { cfqd->serving_prio = IDLE_WORKLOAD; @@ -2119,6 +2120,9 @@ static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg) return; } + if (original_prio != cfqd->serving_prio) + goto new_workload; + /* * For RT and BE, we have to choose also the type * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload @@ -2133,6 +2137,7 @@ static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg) if (count && !time_after(jiffies, cfqd->workload_expires)) return; +new_workload: /* otherwise select new workload type */ cfqd->serving_type = cfq_choose_wl(cfqd, cfqg, cfqd->serving_prio); -- cgit v1.2.1 From dddd9dc340ae1a41d90e084529ca979c77c4ecfe Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Dec 2010 20:57:35 +0100 Subject: block: kill genhd_media_change_notify() There's no user of the facility. Kill it. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 25 ------------------------- 1 file changed, 25 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 5fa2b44a72ff..0905ab22c8cf 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -1110,29 +1110,6 @@ static int __init proc_genhd_init(void) module_init(proc_genhd_init); #endif /* CONFIG_PROC_FS */ -static void media_change_notify_thread(struct work_struct *work) -{ - struct gendisk *gd = container_of(work, struct gendisk, async_notify); - char event[] = "MEDIA_CHANGE=1"; - char *envp[] = { event, NULL }; - - /* - * 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(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); - put_device(gd->driverfs_dev); -} - -#if 0 -void genhd_media_change_notify(struct gendisk *disk) -{ - get_device(disk->driverfs_dev); - schedule_work(&disk->async_notify); -} -EXPORT_SYMBOL_GPL(genhd_media_change_notify); -#endif /* 0 */ - dev_t blk_lookup_devt(const char *name, int partno) { dev_t devt = MKDEV(0, 0); @@ -1198,8 +1175,6 @@ struct gendisk *alloc_disk_node(int minors, int node_id) 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); } return disk; } -- cgit v1.2.1 From d2bf1b6723ed0eab378363649d15b7893bf14e91 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Dec 2010 20:57:36 +0100 Subject: block: move register_disk() and del_gendisk() to block/genhd.c There's no reason for register_disk() and del_gendisk() to be in fs/partitions/check.c. Move both to genhd.c. While at it, collapse unlink_gendisk(), which was artificially in a separate function due to genhd.c / check.c split, into del_gendisk(). Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/genhd.c | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 87 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 0905ab22c8cf..2e5e4c0a1133 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -502,6 +502,64 @@ static int exact_lock(dev_t devt, void *data) return 0; } +void register_disk(struct gendisk *disk) +{ + struct device *ddev = disk_to_dev(disk); + struct block_device *bdev; + struct disk_part_iter piter; + struct hd_struct *part; + int err; + + ddev->parent = disk->driverfs_dev; + + dev_set_name(ddev, disk->disk_name); + + /* delay uevents, until we scanned partition table */ + dev_set_uevent_suppress(ddev, 1); + + if (device_add(ddev)) + return; + if (!sysfs_deprecated) { + err = sysfs_create_link(block_depr, &ddev->kobj, + kobject_name(&ddev->kobj)); + if (err) { + device_del(ddev); + return; + } + } + disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj); + disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); + + /* No minors to use for partitions */ + if (!disk_partitionable(disk)) + goto exit; + + /* No such device (e.g., media were just removed) */ + if (!get_capacity(disk)) + goto exit; + + bdev = bdget_disk(disk, 0); + if (!bdev) + goto exit; + + bdev->bd_invalidated = 1; + err = blkdev_get(bdev, FMODE_READ, NULL); + if (err < 0) + goto exit; + blkdev_put(bdev, FMODE_READ); + +exit: + /* announce disk after possible partitions are created */ + dev_set_uevent_suppress(ddev, 0); + kobject_uevent(&ddev->kobj, KOBJ_ADD); + + /* announce possible partitions */ + disk_part_iter_init(&piter, disk, 0); + while ((part = disk_part_iter_next(&piter))) + kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD); + disk_part_iter_exit(&piter); +} + /** * add_disk - add partitioning information to kernel list * @disk: per-device partitioning information @@ -552,17 +610,43 @@ void add_disk(struct gendisk *disk) "bdi"); WARN_ON(retval); } - EXPORT_SYMBOL(add_disk); -EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */ -void unlink_gendisk(struct gendisk *disk) +void del_gendisk(struct gendisk *disk) { + struct disk_part_iter piter; + struct hd_struct *part; + + /* invalidate stuff */ + disk_part_iter_init(&piter, disk, + DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); + while ((part = disk_part_iter_next(&piter))) { + invalidate_partition(disk, part->partno); + delete_partition(disk, part->partno); + } + disk_part_iter_exit(&piter); + + invalidate_partition(disk, 0); + blk_free_devt(disk_to_dev(disk)->devt); + set_capacity(disk, 0); + disk->flags &= ~GENHD_FL_UP; + 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); + + part_stat_set_all(&disk->part0, 0); + disk->part0.stamp = 0; + + kobject_put(disk->part0.holder_dir); + kobject_put(disk->slave_dir); + disk->driverfs_dev = NULL; + if (!sysfs_deprecated) + sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); + device_del(disk_to_dev(disk)); } +EXPORT_SYMBOL(del_gendisk); /** * get_gendisk - get partitioning information for a given device -- cgit v1.2.1 From 77ea887e433ad8389d416826936c110fa7910f80 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Dec 2010 20:57:37 +0100 Subject: implement in-kernel gendisk events handling Currently, media presence polling for removeable block devices is done from userland. There are several issues with this. * Polling is done by periodically opening the device. For SCSI devices, the command sequence generated by such action involves a few different commands including TEST_UNIT_READY. This behavior, while perfectly legal, is different from Windows which only issues single command, GET_EVENT_STATUS_NOTIFICATION. Unfortunately, some ATAPI devices lock up after being periodically queried such command sequences. * There is no reliable and unintrusive way for a userland program to tell whether the target device is safe for media presence polling. For example, polling for media presence during an on-going burning session can make it fail. The polling program can avoid this by opening the device with O_EXCL but then it risks making a valid exclusive user of the device fail w/ -EBUSY. * Userland polling is unnecessarily heavy and in-kernel implementation is lighter and better coordinated (workqueue, timer slack). This patch implements framework for in-kernel disk event handling, which includes media presence polling. * bdops->check_events() is added, which supercedes ->media_changed(). It should check whether there's any pending event and return if so. Currently, two events are defined - DISK_EVENT_MEDIA_CHANGE and DISK_EVENT_EJECT_REQUEST. ->check_events() is guaranteed not to be called parallelly. * gendisk->events and ->async_events are added. These should be initialized by block driver before passing the device to add_disk(). The former contains the mask of all supported events and the latter the mask of all events which the device can report without polling. /sys/block/*/events[_async] export these to userland. * Kernel parameter block.events_dfl_poll_msecs controls the system polling interval (default is 0 which means disable) and /sys/block/*/events_poll_msecs control polling intervals for individual devices (default is -1 meaning use system setting). Note that if a device can report all supported events asynchronously and its polling interval isn't explicitly set, the device won't be polled regardless of the system polling interval. * If a device is opened exclusively with write access, event checking is automatically disabled until all write exclusive accesses are released. * There are event 'clearing' events. For example, both of currently defined events are cleared after the device has been successfully opened. This information is passed to ->check_events() callback using @clearing argument as a hint. * Event checking is always performed from system_nrt_wq and timer slack is set to 25% for polling. * Nothing changes for drivers which implement ->media_changed() but not ->check_events(). Going forward, all drivers will be converted to ->check_events() and ->media_change() will be dropped. Signed-off-by: Tejun Heo Cc: Kay Sievers Cc: Jan Kara Signed-off-by: Jens Axboe --- block/genhd.c | 429 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 429 insertions(+) (limited to 'block') diff --git a/block/genhd.c b/block/genhd.c index 2e5e4c0a1133..5465a824d489 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -18,6 +18,7 @@ #include #include #include +#include #include "blk.h" @@ -35,6 +36,10 @@ static DEFINE_IDR(ext_devt_idr); static struct device_type disk_type; +static void disk_add_events(struct gendisk *disk); +static void disk_del_events(struct gendisk *disk); +static void disk_release_events(struct gendisk *disk); + /** * disk_get_part - get partition * @disk: disk to look partition from @@ -609,6 +614,8 @@ void add_disk(struct gendisk *disk) retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj, "bdi"); WARN_ON(retval); + + disk_add_events(disk); } EXPORT_SYMBOL(add_disk); @@ -617,6 +624,8 @@ void del_gendisk(struct gendisk *disk) struct disk_part_iter piter; struct hd_struct *part; + disk_del_events(disk); + /* invalidate stuff */ disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE); @@ -1089,6 +1098,7 @@ static void disk_release(struct device *dev) { struct gendisk *disk = dev_to_disk(dev); + disk_release_events(disk); kfree(disk->random); disk_replace_part_tbl(disk, NULL); free_part_stats(&disk->part0); @@ -1350,3 +1360,422 @@ int invalidate_partition(struct gendisk *disk, int partno) } EXPORT_SYMBOL(invalidate_partition); + +/* + * Disk events - monitor disk events like media change and eject request. + */ +struct disk_events { + struct list_head node; /* all disk_event's */ + struct gendisk *disk; /* the associated disk */ + spinlock_t lock; + + int block; /* event blocking depth */ + unsigned int pending; /* events already sent out */ + unsigned int clearing; /* events being cleared */ + + long poll_msecs; /* interval, -1 for default */ + struct delayed_work dwork; +}; + +static const char *disk_events_strs[] = { + [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change", + [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request", +}; + +static char *disk_uevents[] = { + [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1", + [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1", +}; + +/* list of all disk_events */ +static DEFINE_MUTEX(disk_events_mutex); +static LIST_HEAD(disk_events); + +/* disable in-kernel polling by default */ +static unsigned long disk_events_dfl_poll_msecs = 0; + +static unsigned long disk_events_poll_jiffies(struct gendisk *disk) +{ + struct disk_events *ev = disk->ev; + long intv_msecs = 0; + + /* + * If device-specific poll interval is set, always use it. If + * the default is being used, poll iff there are events which + * can't be monitored asynchronously. + */ + if (ev->poll_msecs >= 0) + intv_msecs = ev->poll_msecs; + else if (disk->events & ~disk->async_events) + intv_msecs = disk_events_dfl_poll_msecs; + + return msecs_to_jiffies(intv_msecs); +} + +static void __disk_block_events(struct gendisk *disk, bool sync) +{ + struct disk_events *ev = disk->ev; + unsigned long flags; + bool cancel; + + spin_lock_irqsave(&ev->lock, flags); + cancel = !ev->block++; + spin_unlock_irqrestore(&ev->lock, flags); + + if (cancel) { + if (sync) + cancel_delayed_work_sync(&disk->ev->dwork); + else + cancel_delayed_work(&disk->ev->dwork); + } +} + +static void __disk_unblock_events(struct gendisk *disk, bool check_now) +{ + struct disk_events *ev = disk->ev; + unsigned long intv; + unsigned long flags; + + spin_lock_irqsave(&ev->lock, flags); + + if (WARN_ON_ONCE(ev->block <= 0)) + goto out_unlock; + + if (--ev->block) + goto out_unlock; + + /* + * Not exactly a latency critical operation, set poll timer + * slack to 25% and kick event check. + */ + intv = disk_events_poll_jiffies(disk); + set_timer_slack(&ev->dwork.timer, intv / 4); + if (check_now) + queue_delayed_work(system_nrt_wq, &ev->dwork, 0); + else if (intv) + queue_delayed_work(system_nrt_wq, &ev->dwork, intv); +out_unlock: + spin_unlock_irqrestore(&ev->lock, flags); +} + +/** + * disk_block_events - block and flush disk event checking + * @disk: disk to block events for + * + * On return from this function, it is guaranteed that event checking + * isn't in progress and won't happen until unblocked by + * disk_unblock_events(). Events blocking is counted and the actual + * unblocking happens after the matching number of unblocks are done. + * + * Note that this intentionally does not block event checking from + * disk_clear_events(). + * + * CONTEXT: + * Might sleep. + */ +void disk_block_events(struct gendisk *disk) +{ + if (disk->ev) + __disk_block_events(disk, true); +} + +/** + * disk_unblock_events - unblock disk event checking + * @disk: disk to unblock events for + * + * Undo disk_block_events(). When the block count reaches zero, it + * starts events polling if configured. + * + * CONTEXT: + * Don't care. Safe to call from irq context. + */ +void disk_unblock_events(struct gendisk *disk) +{ + if (disk->ev) + __disk_unblock_events(disk, true); +} + +/** + * disk_check_events - schedule immediate event checking + * @disk: disk to check events for + * + * Schedule immediate event checking on @disk if not blocked. + * + * CONTEXT: + * Don't care. Safe to call from irq context. + */ +void disk_check_events(struct gendisk *disk) +{ + if (disk->ev) { + __disk_block_events(disk, false); + __disk_unblock_events(disk, true); + } +} +EXPORT_SYMBOL_GPL(disk_check_events); + +/** + * disk_clear_events - synchronously check, clear and return pending events + * @disk: disk to fetch and clear events from + * @mask: mask of events to be fetched and clearted + * + * Disk events are synchronously checked and pending events in @mask + * are cleared and returned. This ignores the block count. + * + * CONTEXT: + * Might sleep. + */ +unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask) +{ + const struct block_device_operations *bdops = disk->fops; + struct disk_events *ev = disk->ev; + unsigned int pending; + + if (!ev) { + /* for drivers still using the old ->media_changed method */ + if ((mask & DISK_EVENT_MEDIA_CHANGE) && + bdops->media_changed && bdops->media_changed(disk)) + return DISK_EVENT_MEDIA_CHANGE; + return 0; + } + + /* tell the workfn about the events being cleared */ + spin_lock_irq(&ev->lock); + ev->clearing |= mask; + spin_unlock_irq(&ev->lock); + + /* uncondtionally schedule event check and wait for it to finish */ + __disk_block_events(disk, true); + queue_delayed_work(system_nrt_wq, &ev->dwork, 0); + flush_delayed_work(&ev->dwork); + __disk_unblock_events(disk, false); + + /* then, fetch and clear pending events */ + spin_lock_irq(&ev->lock); + WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */ + pending = ev->pending & mask; + ev->pending &= ~mask; + spin_unlock_irq(&ev->lock); + + return pending; +} + +static void disk_events_workfn(struct work_struct *work) +{ + struct delayed_work *dwork = to_delayed_work(work); + struct disk_events *ev = container_of(dwork, struct disk_events, dwork); + struct gendisk *disk = ev->disk; + char *envp[ARRAY_SIZE(disk_uevents) + 1] = { }; + unsigned int clearing = ev->clearing; + unsigned int events; + unsigned long intv; + int nr_events = 0, i; + + /* check events */ + events = disk->fops->check_events(disk, clearing); + + /* accumulate pending events and schedule next poll if necessary */ + spin_lock_irq(&ev->lock); + + events &= ~ev->pending; + ev->pending |= events; + ev->clearing &= ~clearing; + + intv = disk_events_poll_jiffies(disk); + if (!ev->block && intv) + queue_delayed_work(system_nrt_wq, &ev->dwork, intv); + + spin_unlock_irq(&ev->lock); + + /* tell userland about new events */ + for (i = 0; i < ARRAY_SIZE(disk_uevents); i++) + if (events & (1 << i)) + envp[nr_events++] = disk_uevents[i]; + + if (nr_events) + kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); +} + +/* + * A disk events enabled device has the following sysfs nodes under + * its /sys/block/X/ directory. + * + * events : list of all supported events + * events_async : list of events which can be detected w/o polling + * events_poll_msecs : polling interval, 0: disable, -1: system default + */ +static ssize_t __disk_events_show(unsigned int events, char *buf) +{ + const char *delim = ""; + ssize_t pos = 0; + int i; + + for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++) + if (events & (1 << i)) { + pos += sprintf(buf + pos, "%s%s", + delim, disk_events_strs[i]); + delim = " "; + } + if (pos) + pos += sprintf(buf + pos, "\n"); + return pos; +} + +static ssize_t disk_events_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + + return __disk_events_show(disk->events, buf); +} + +static ssize_t disk_events_async_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + + return __disk_events_show(disk->async_events, buf); +} + +static ssize_t disk_events_poll_msecs_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct gendisk *disk = dev_to_disk(dev); + + return sprintf(buf, "%ld\n", disk->ev->poll_msecs); +} + +static ssize_t disk_events_poll_msecs_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct gendisk *disk = dev_to_disk(dev); + long intv; + + if (!count || !sscanf(buf, "%ld", &intv)) + return -EINVAL; + + if (intv < 0 && intv != -1) + return -EINVAL; + + __disk_block_events(disk, true); + disk->ev->poll_msecs = intv; + __disk_unblock_events(disk, true); + + return count; +} + +static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL); +static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL); +static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR, + disk_events_poll_msecs_show, + disk_events_poll_msecs_store); + +static const struct attribute *disk_events_attrs[] = { + &dev_attr_events.attr, + &dev_attr_events_async.attr, + &dev_attr_events_poll_msecs.attr, + NULL, +}; + +/* + * The default polling interval can be specified by the kernel + * parameter block.events_dfl_poll_msecs which defaults to 0 + * (disable). This can also be modified runtime by writing to + * /sys/module/block/events_dfl_poll_msecs. + */ +static int disk_events_set_dfl_poll_msecs(const char *val, + const struct kernel_param *kp) +{ + struct disk_events *ev; + int ret; + + ret = param_set_ulong(val, kp); + if (ret < 0) + return ret; + + mutex_lock(&disk_events_mutex); + + list_for_each_entry(ev, &disk_events, node) + disk_check_events(ev->disk); + + mutex_unlock(&disk_events_mutex); + + return 0; +} + +static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = { + .set = disk_events_set_dfl_poll_msecs, + .get = param_get_ulong, +}; + +#undef MODULE_PARAM_PREFIX +#define MODULE_PARAM_PREFIX "block." + +module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops, + &disk_events_dfl_poll_msecs, 0644); + +/* + * disk_{add|del|release}_events - initialize and destroy disk_events. + */ +static void disk_add_events(struct gendisk *disk) +{ + struct disk_events *ev; + + if (!disk->fops->check_events || !(disk->events | disk->async_events)) + return; + + ev = kzalloc(sizeof(*ev), GFP_KERNEL); + if (!ev) { + pr_warn("%s: failed to initialize events\n", disk->disk_name); + return; + } + + if (sysfs_create_files(&disk_to_dev(disk)->kobj, + disk_events_attrs) < 0) { + pr_warn("%s: failed to create sysfs files for events\n", + disk->disk_name); + kfree(ev); + return; + } + + disk->ev = ev; + + INIT_LIST_HEAD(&ev->node); + ev->disk = disk; + spin_lock_init(&ev->lock); + ev->block = 1; + ev->poll_msecs = -1; + INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn); + + mutex_lock(&disk_events_mutex); + list_add_tail(&ev->node, &disk_events); + mutex_unlock(&disk_events_mutex); + + /* + * Block count is initialized to 1 and the following initial + * unblock kicks it into action. + */ + __disk_unblock_events(disk, true); +} + +static void disk_del_events(struct gendisk *disk) +{ + if (!disk->ev) + return; + + __disk_block_events(disk, true); + + mutex_lock(&disk_events_mutex); + list_del_init(&disk->ev->node); + mutex_unlock(&disk_events_mutex); + + sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs); +} + +static void disk_release_events(struct gendisk *disk) +{ + /* the block count should be 1 from disk_del_events() */ + WARN_ON_ONCE(disk->ev && disk->ev->block != 1); + kfree(disk->ev); +} -- cgit v1.2.1 From 7278c9c19bd85cf33213a2e0b538a18d3ac8ad00 Mon Sep 17 00:00:00 2001 From: Gui Jianfeng Date: Fri, 17 Dec 2010 08:57:14 +0100 Subject: cfq-iosched: don't check cfqg in choose_service_tree() When cfq_choose_cfqg() is called in select_queue(), there must be at least one backlogged CFQ queue waiting for dispatching, hence there must be at least one backlogged CFQ group on service tree. So we never call choose_service_tree() with cfqg == NULL. Signed-off-by: Gui Jianfeng Reviewed-by: Jeff Moyer Acked-by: Vivek Goyal Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 9b186fd6bf47..c19d015ac5a5 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -2103,12 +2103,6 @@ static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg) unsigned group_slice; enum wl_prio_t original_prio = cfqd->serving_prio; - if (!cfqg) { - cfqd->serving_prio = IDLE_WORKLOAD; - cfqd->workload_expires = jiffies + 1; - return; - } - /* Choose next priority. RT > BE > IDLE */ if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg)) cfqd->serving_prio = RT_WORKLOAD; -- cgit v1.2.1 From b9f985b6e05ebd7af2aaef0eb3ae369390ef191f Mon Sep 17 00:00:00 2001 From: Yang Zhang Date: Fri, 17 Dec 2010 08:58:36 +0100 Subject: block: convert !IS_ERR(p) && p to !IS_ERR_NOR_NULL(p) Signed-off-by: Yang Zhang 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 5fa2b44a72ff..79b9e327f3cb 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -735,7 +735,7 @@ 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 && !*pos) + if (!IS_ERR_OR_NULL(p) && !*pos) seq_puts(seqf, "major minor #blocks name\n\n"); return p; } -- cgit v1.2.1 From e61eb2e93fe86931d46831752a82dab25a5335ca Mon Sep 17 00:00:00 2001 From: Yang Zhang Date: Fri, 17 Dec 2010 09:00:18 +0100 Subject: fs/block: type signature of major_to_index(int) to major_to_index(unsigned) The major/minor device numbers are always defined and used as `unsigned'. Signed-off-by: Yang Zhang 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 79b9e327f3cb..16ccc0d2d5d3 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -239,7 +239,7 @@ static struct blk_major_name { } *major_names[BLKDEV_MAJOR_HASH_SIZE]; /* index in the above - for now: assume no multimajor ranges */ -static inline int major_to_index(int major) +static inline int major_to_index(unsigned major) { return major % BLKDEV_MAJOR_HASH_SIZE; } -- cgit v1.2.1 From 27667c996f6a0bed4ad1e10ac0a0dbb6037968db Mon Sep 17 00:00:00 2001 From: Bart Van Assche Date: Tue, 21 Dec 2010 15:07:45 +0100 Subject: block: Clean up exit_io_context() source code. This patch fixes a spelling error in a source code comment and removes superfluous braces in the function exit_io_context(). Signed-off-by: Bart Van Assche Cc: Jens Axboe Signed-off-by: Jens Axboe --- block/blk-ioc.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'block') diff --git a/block/blk-ioc.c b/block/blk-ioc.c index 3c7a339fe381..b791022beef3 100644 --- a/block/blk-ioc.c +++ b/block/blk-ioc.c @@ -64,7 +64,7 @@ static void cfq_exit(struct io_context *ioc) rcu_read_unlock(); } -/* Called by the exitting task */ +/* Called by the exiting task */ void exit_io_context(struct task_struct *task) { struct io_context *ioc; @@ -74,10 +74,9 @@ void exit_io_context(struct task_struct *task) task->io_context = NULL; task_unlock(task); - if (atomic_dec_and_test(&ioc->nr_tasks)) { + if (atomic_dec_and_test(&ioc->nr_tasks)) cfq_exit(ioc); - } put_io_context(ioc); } -- cgit v1.2.1 From 89b90be2d877a904b1704e4029db65655bfc6282 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 3 Jan 2011 15:01:47 +0100 Subject: block: make kblockd_workqueue smarter kblockd is used for unplugging and may affect IO latency and throughput and the max number of concurrent work items are bound by the number of block devices. Make it HIGHPRI workqueue w/ default max concurrency. Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-core.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 151070541e21..3689319a5974 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -2606,7 +2606,9 @@ int __init blk_dev_init(void) BUILD_BUG_ON(__REQ_NR_BITS > 8 * sizeof(((struct request *)0)->cmd_flags)); - kblockd_workqueue = create_workqueue("kblockd"); + /* used for unplugging and affects IO latency/throughput - HIGHPRI */ + kblockd_workqueue = alloc_workqueue("kblockd", + WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); if (!kblockd_workqueue) panic("Failed to create kblockd\n"); -- cgit v1.2.1 From 09e099d4bafea3b15be003d548bdf94b4b6e0e17 Mon Sep 17 00:00:00 2001 From: Jerome Marchand Date: Wed, 5 Jan 2011 16:57:38 +0100 Subject: block: fix accounting bug on cross partition merges /proc/diskstats would display a strange output as follows. $ cat /proc/diskstats |grep sda 8 0 sda 90524 7579 102154 20464 0 0 0 0 0 14096 20089 8 1 sda1 19085 1352 21841 4209 0 0 0 0 4294967064 15689 4293424691 ~~~~~~~~~~ 8 2 sda2 71252 3624 74891 15950 0 0 0 0 232 23995 1562390 8 3 sda3 54 487 2188 92 0 0 0 0 0 88 92 8 4 sda4 4 0 8 0 0 0 0 0 0 0 0 8 5 sda5 81 2027 2130 138 0 0 0 0 0 87 137 Its reason is the wrong way of accounting hd_struct->in_flight. When a bio is merged into a request belongs to different partition by ELEVATOR_FRONT_MERGE. The detailed root cause is as follows. Assuming that there are two partition, sda1 and sda2. 1. A request for sda2 is in request_queue. Hence sda1's hd_struct->in_flight is 0 and sda2's one is 1. | hd_struct->in_flight --------------------------- sda1 | 0 sda2 | 1 --------------------------- 2. A bio belongs to sda1 is issued and is merged into the request mentioned on step1 by ELEVATOR_BACK_MERGE. The first sector of the request is changed from sda2 region to sda1 region. However the two partition's hd_struct->in_flight are not changed. | hd_struct->in_flight --------------------------- sda1 | 0 sda2 | 1 --------------------------- 3. The request is finished and blk_account_io_done() is called. In this case, sda2's hd_struct->in_flight, not a sda1's one, is decremented. | hd_struct->in_flight --------------------------- sda1 | -1 sda2 | 1 --------------------------- The patch fixes the problem by caching the partition lookup inside the request structure, hence making sure that the increment and decrement will always happen on the same partition struct. This also speeds up IO with accounting enabled, since it cuts down on the number of lookups we have to do. Also add a refcount to struct hd_struct to keep the partition in memory as long as users exist. We use kref_test_and_get() to ensure we don't add a reference to a partition which is going away. Signed-off-by: Jerome Marchand Signed-off-by: Yasuaki Ishimatsu Cc: stable@kernel.org Signed-off-by: Jens Axboe --- block/blk-core.c | 26 +++++++++++++++++++++----- block/blk-merge.c | 3 ++- block/genhd.c | 1 + 3 files changed, 24 insertions(+), 6 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 3689319a5974..500c080a6a6b 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -64,13 +64,27 @@ static void drive_stat_acct(struct request *rq, int new_io) return; cpu = part_stat_lock(); - part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); - if (!new_io) + if (!new_io) { + part = rq->part; part_stat_inc(cpu, part, merges[rw]); - else { + } else { + part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); + if (!kref_test_and_get(&part->ref)) { + /* + * The partition is already being removed, + * the request will be accounted on the disk only + * + * We take a reference on disk->part0 although that + * partition will never be deleted, so we can treat + * it as any other partition. + */ + part = &rq->rq_disk->part0; + kref_get(&part->ref); + } part_round_stats(cpu, part); part_inc_in_flight(part, rw); + rq->part = part; } part_stat_unlock(); @@ -128,6 +142,7 @@ void blk_rq_init(struct request_queue *q, struct request *rq) rq->ref_count = 1; rq->start_time = jiffies; set_start_time_ns(rq); + rq->part = NULL; } EXPORT_SYMBOL(blk_rq_init); @@ -1776,7 +1791,7 @@ static void blk_account_io_completion(struct request *req, unsigned int bytes) int cpu; cpu = part_stat_lock(); - part = disk_map_sector_rcu(req->rq_disk, blk_rq_pos(req)); + part = req->part; part_stat_add(cpu, part, sectors[rw], bytes >> 9); part_stat_unlock(); } @@ -1796,13 +1811,14 @@ static void blk_account_io_done(struct request *req) int cpu; cpu = part_stat_lock(); - part = disk_map_sector_rcu(req->rq_disk, blk_rq_pos(req)); + part = req->part; 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, rw); + kref_put(&part->ref, __delete_partition); part_stat_unlock(); } } diff --git a/block/blk-merge.c b/block/blk-merge.c index 77b7c26df6b5..b06b83b89d89 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -351,11 +351,12 @@ static void blk_account_io_merge(struct request *req) int cpu; cpu = part_stat_lock(); - part = disk_map_sector_rcu(req->rq_disk, blk_rq_pos(req)); + part = req->part; part_round_stats(cpu, part); part_dec_in_flight(part, rq_data_dir(req)); + kref_put(&part->ref, __delete_partition); part_stat_unlock(); } } diff --git a/block/genhd.c b/block/genhd.c index 16ccc0d2d5d3..85c150598830 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -1192,6 +1192,7 @@ struct gendisk *alloc_disk_node(int minors, int node_id) return NULL; } disk->part_tbl->part[0] = &disk->part0; + kref_init(&disk->part0.ref); disk->minors = minors; rand_initialize_disk(disk); -- cgit v1.2.1 From 6c23a9681c0fe7fb7dd331b39dda11926f43746e Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 7 Jan 2011 08:43:37 +0100 Subject: block: add internal hd part table references We can't use krefs since it's apparently restricted to very basic reference counting. This reverts commit e4a683c8. Signed-off-by: Jens Axboe --- block/blk-core.c | 6 +++--- block/blk-merge.c | 2 +- block/genhd.c | 3 ++- 3 files changed, 6 insertions(+), 5 deletions(-) (limited to 'block') diff --git a/block/blk-core.c b/block/blk-core.c index 500c080a6a6b..2f4002f79a24 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -70,7 +70,7 @@ static void drive_stat_acct(struct request *rq, int new_io) part_stat_inc(cpu, part, merges[rw]); } else { part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); - if (!kref_test_and_get(&part->ref)) { + if (!hd_struct_try_get(part)) { /* * The partition is already being removed, * the request will be accounted on the disk only @@ -80,7 +80,7 @@ static void drive_stat_acct(struct request *rq, int new_io) * it as any other partition. */ part = &rq->rq_disk->part0; - kref_get(&part->ref); + hd_struct_get(part); } part_round_stats(cpu, part); part_inc_in_flight(part, rw); @@ -1818,7 +1818,7 @@ static void blk_account_io_done(struct request *req) part_round_stats(cpu, part); part_dec_in_flight(part, rw); - kref_put(&part->ref, __delete_partition); + hd_struct_put(part); part_stat_unlock(); } } diff --git a/block/blk-merge.c b/block/blk-merge.c index b06b83b89d89..00b7d31b38a2 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -356,7 +356,7 @@ static void blk_account_io_merge(struct request *req) part_round_stats(cpu, part); part_dec_in_flight(part, rq_data_dir(req)); - kref_put(&part->ref, __delete_partition); + hd_struct_put(part); part_stat_unlock(); } } diff --git a/block/genhd.c b/block/genhd.c index 85c150598830..399d37ec7412 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -1192,7 +1192,8 @@ struct gendisk *alloc_disk_node(int minors, int node_id) return NULL; } disk->part_tbl->part[0] = &disk->part0; - kref_init(&disk->part0.ref); + + hd_ref_init(&disk->part0); disk->minors = minors; rand_initialize_disk(disk); -- cgit v1.2.1 From 30d7b9448f03f2c82d0fd44738674cc156a8ce0a Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Fri, 7 Jan 2011 08:46:59 +0100 Subject: block cfq: don't use atomic_t for cfq_queue cfq_queue->ref is used with queue_lock hold, so ref doesn't need to be an atomic and atomic operation is slower. Signed-off-by: Shaohua Li Reviewed-by: Jeff Moyer Acked-by: Vivek Goyal Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index c19d015ac5a5..4cb4cf73ac00 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -96,7 +96,7 @@ struct cfq_rb_root { */ struct cfq_queue { /* reference count */ - atomic_t ref; + int ref; /* various state flags, see below */ unsigned int flags; /* parent cfq_data */ @@ -2025,7 +2025,7 @@ static int cfqq_process_refs(struct cfq_queue *cfqq) int process_refs, io_refs; io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE]; - process_refs = atomic_read(&cfqq->ref) - io_refs; + process_refs = cfqq->ref - io_refs; BUG_ON(process_refs < 0); return process_refs; } @@ -2065,10 +2065,10 @@ static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq) */ if (new_process_refs >= process_refs) { cfqq->new_cfqq = new_cfqq; - atomic_add(process_refs, &new_cfqq->ref); + new_cfqq->ref += process_refs; } else { new_cfqq->new_cfqq = cfqq; - atomic_add(new_process_refs, &cfqq->ref); + cfqq->ref += new_process_refs; } } @@ -2532,9 +2532,10 @@ static void cfq_put_queue(struct cfq_queue *cfqq) struct cfq_data *cfqd = cfqq->cfqd; struct cfq_group *cfqg, *orig_cfqg; - BUG_ON(atomic_read(&cfqq->ref) <= 0); + BUG_ON(cfqq->ref <= 0); - if (!atomic_dec_and_test(&cfqq->ref)) + cfqq->ref--; + if (cfqq->ref) return; cfq_log_cfqq(cfqd, cfqq, "put_queue"); @@ -2837,7 +2838,7 @@ static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq, RB_CLEAR_NODE(&cfqq->p_node); INIT_LIST_HEAD(&cfqq->fifo); - atomic_set(&cfqq->ref, 0); + cfqq->ref = 0; cfqq->cfqd = cfqd; cfq_mark_cfqq_prio_changed(cfqq); @@ -2973,11 +2974,11 @@ cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct io_context *ioc, * pin the queue now that it's allocated, scheduler exit will prune it */ if (!is_sync && !(*async_cfqq)) { - atomic_inc(&cfqq->ref); + cfqq->ref++; *async_cfqq = cfqq; } - atomic_inc(&cfqq->ref); + cfqq->ref++; return cfqq; } @@ -3679,7 +3680,7 @@ new_queue: } cfqq->allocated[rw]++; - atomic_inc(&cfqq->ref); + cfqq->ref++; spin_unlock_irqrestore(q->queue_lock, flags); @@ -3860,6 +3861,10 @@ static void *cfq_init_queue(struct request_queue *q) if (!cfqd) return NULL; + /* + * Don't need take queue_lock in the routine, since we are + * initializing the ioscheduler, and nobody is using cfqd + */ cfqd->cic_index = i; /* Init root service tree */ @@ -3899,7 +3904,7 @@ static void *cfq_init_queue(struct request_queue *q) * will not attempt to free it. */ cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0); - atomic_inc(&cfqd->oom_cfqq.ref); + cfqd->oom_cfqq.ref++; cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, &cfqd->root_group); INIT_LIST_HEAD(&cfqd->cic_list); -- cgit v1.2.1 From 329a67815b596d23daf0caa588ae0800e925320f Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Fri, 7 Jan 2011 08:48:28 +0100 Subject: block cfq: don't use atomic_t for cfq_group cfq_group->ref is used with queue_lock hold, the only exception is cfq_set_request, which looks like a bug to me, so ref doesn't need to be an atomic and atomic operation is slower. Signed-off-by: Shaohua Li Reviewed-by: Jeff Moyer Acked-by: Vivek Goyal Signed-off-by: Jens Axboe --- block/cfq-iosched.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) (limited to 'block') diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 4cb4cf73ac00..f083bda30546 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -207,7 +207,7 @@ struct cfq_group { struct blkio_group blkg; #ifdef CONFIG_CFQ_GROUP_IOSCHED struct hlist_node cfqd_node; - atomic_t ref; + int ref; #endif /* number of requests that are on the dispatch list or inside driver */ int dispatched; @@ -1014,7 +1014,7 @@ cfq_find_alloc_cfqg(struct cfq_data *cfqd, struct cgroup *cgroup, int create) * elevator which will be dropped by either elevator exit * or cgroup deletion path depending on who is exiting first. */ - atomic_set(&cfqg->ref, 1); + cfqg->ref = 1; /* * Add group onto cgroup list. It might happen that bdi->dev is @@ -1059,7 +1059,7 @@ static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd, int create) static inline struct cfq_group *cfq_ref_get_cfqg(struct cfq_group *cfqg) { - atomic_inc(&cfqg->ref); + cfqg->ref++; return cfqg; } @@ -1071,7 +1071,7 @@ static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) cfqq->cfqg = cfqg; /* cfqq reference on cfqg */ - atomic_inc(&cfqq->cfqg->ref); + cfqq->cfqg->ref++; } static void cfq_put_cfqg(struct cfq_group *cfqg) @@ -1079,8 +1079,9 @@ static void cfq_put_cfqg(struct cfq_group *cfqg) struct cfq_rb_root *st; int i, j; - BUG_ON(atomic_read(&cfqg->ref) <= 0); - if (!atomic_dec_and_test(&cfqg->ref)) + BUG_ON(cfqg->ref <= 0); + cfqg->ref--; + if (cfqg->ref) return; for_each_cfqg_st(cfqg, i, j, st) BUG_ON(!RB_EMPTY_ROOT(&st->rb)); @@ -1188,7 +1189,7 @@ static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq, cfq_group_service_tree_del(cfqd, cfqq->cfqg); cfqq->orig_cfqg = cfqq->cfqg; cfqq->cfqg = &cfqd->root_group; - atomic_inc(&cfqd->root_group.ref); + cfqd->root_group.ref++; group_changed = 1; } else if (!cfqd->cfq_group_isolation && cfqq_type(cfqq) == SYNC_WORKLOAD && cfqq->orig_cfqg) { @@ -3681,12 +3682,12 @@ new_queue: cfqq->allocated[rw]++; cfqq->ref++; - - spin_unlock_irqrestore(q->queue_lock, flags); - rq->elevator_private = cic; rq->elevator_private2 = cfqq; rq->elevator_private3 = cfq_ref_get_cfqg(cfqq->cfqg); + + spin_unlock_irqrestore(q->queue_lock, flags); + return 0; queue_fail: @@ -3884,7 +3885,7 @@ static void *cfq_init_queue(struct request_queue *q) * Take a reference to root group which we never drop. This is just * to make sure that cfq_put_cfqg() does not try to kfree root group */ - atomic_set(&cfqg->ref, 1); + cfqg->ref = 1; rcu_read_lock(); cfq_blkiocg_add_blkio_group(&blkio_root_cgroup, &cfqg->blkg, (void *)cfqd, 0); -- cgit v1.2.1