From 11a3122f6cf2d988a77eb8883d0fc49cd013a6d5 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Tue, 7 Feb 2012 07:51:30 +0100 Subject: block: strip out locking optimization in put_io_context() put_io_context() performed a complex trylock dancing to avoid deferring ioc release to workqueue. It was also broken on UP because trylock was always assumed to succeed which resulted in unbalanced preemption count. While there are ways to fix the UP breakage, even the most pathological microbench (forced ioc allocation and tight fork/exit loop) fails to show any appreciable performance benefit of the optimization. Strip it out. If there turns out to be workloads which are affected by this change, simpler optimization from the discussion thread can be applied later. Signed-off-by: Tejun Heo LKML-Reference: <1328514611.21268.66.camel@sli10-conroe> Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 3 --- include/linux/iocontext.h | 5 ++--- 2 files changed, 2 insertions(+), 6 deletions(-) (limited to 'include/linux') diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 6c6a1f008065..606cf339bb56 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -399,9 +399,6 @@ struct request_queue { /* Throttle data */ struct throtl_data *td; #endif -#ifdef CONFIG_LOCKDEP - int ioc_release_depth; -#endif }; #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */ diff --git a/include/linux/iocontext.h b/include/linux/iocontext.h index 7e1371c4bccf..119773eebe31 100644 --- a/include/linux/iocontext.h +++ b/include/linux/iocontext.h @@ -133,7 +133,7 @@ static inline struct io_context *ioc_task_link(struct io_context *ioc) struct task_struct; #ifdef CONFIG_BLOCK -void put_io_context(struct io_context *ioc, struct request_queue *locked_q); +void put_io_context(struct io_context *ioc); void exit_io_context(struct task_struct *task); struct io_context *get_task_io_context(struct task_struct *task, gfp_t gfp_flags, int node); @@ -141,8 +141,7 @@ void ioc_ioprio_changed(struct io_context *ioc, int ioprio); void ioc_cgroup_changed(struct io_context *ioc); #else struct io_context; -static inline void put_io_context(struct io_context *ioc, - struct request_queue *locked_q) { } +static inline void put_io_context(struct io_context *ioc) { } static inline void exit_io_context(struct task_struct *task) { } #endif -- cgit v1.2.3 From 050c8ea80e3e90019d9e981c6a117ef614e882ed Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Feb 2012 09:19:38 +0100 Subject: block: separate out blk_rq_merge_ok() and blk_try_merge() from elevator functions blk_rq_merge_ok() is the elevator-neutral part of merge eligibility test. blk_try_merge() determines merge direction and expects the caller to have tested elv_rq_merge_ok() previously. elv_rq_merge_ok() now wraps blk_rq_merge_ok() and then calls elv_iosched_allow_merge(). elv_try_merge() is removed and the two callers are updated to call elv_rq_merge_ok() explicitly followed by blk_try_merge(). While at it, make rq_merge_ok() functions return bool. This is to prepare for plug merge update and doesn't introduce any behavior change. This is based on Jens' patch to skip elevator_allow_merge_fn() from plug merge. Signed-off-by: Tejun Heo LKML-Reference: <4F16F3CA.90904@kernel.dk> Original-patch-by: Jens Axboe Signed-off-by: Jens Axboe --- block/blk-core.c | 4 ++-- block/blk-merge.c | 37 ++++++++++++++++++++++++++++++++ block/blk.h | 2 ++ block/elevator.c | 55 ++++-------------------------------------------- include/linux/elevator.h | 3 +-- 5 files changed, 46 insertions(+), 55 deletions(-) (limited to 'include/linux') diff --git a/block/blk-core.c b/block/blk-core.c index 532b3a21b383..fa697bf691eb 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1282,10 +1282,10 @@ static bool attempt_plug_merge(struct request_queue *q, struct bio *bio, (*request_count)++; - if (rq->q != q) + if (rq->q != q || !elv_rq_merge_ok(rq, bio)) continue; - el_ret = elv_try_merge(rq, bio); + el_ret = blk_try_merge(rq, bio); if (el_ret == ELEVATOR_BACK_MERGE) { ret = bio_attempt_back_merge(q, rq, bio); if (ret) diff --git a/block/blk-merge.c b/block/blk-merge.c index cfcc37cb222b..160035f54882 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -471,3 +471,40 @@ int blk_attempt_req_merge(struct request_queue *q, struct request *rq, { return attempt_merge(q, rq, next); } + +bool blk_rq_merge_ok(struct request *rq, struct bio *bio) +{ + if (!rq_mergeable(rq)) + return false; + + /* don't merge file system requests and discard requests */ + if ((bio->bi_rw & REQ_DISCARD) != (rq->bio->bi_rw & REQ_DISCARD)) + return false; + + /* don't merge discard requests and secure discard requests */ + if ((bio->bi_rw & REQ_SECURE) != (rq->bio->bi_rw & REQ_SECURE)) + return false; + + /* different data direction or already started, don't merge */ + if (bio_data_dir(bio) != rq_data_dir(rq)) + return false; + + /* must be same device and not a special request */ + if (rq->rq_disk != bio->bi_bdev->bd_disk || rq->special) + return false; + + /* only merge integrity protected bio into ditto rq */ + if (bio_integrity(bio) != blk_integrity_rq(rq)) + return false; + + return true; +} + +int blk_try_merge(struct request *rq, struct bio *bio) +{ + if (blk_rq_pos(rq) + blk_rq_sectors(rq) == bio->bi_sector) + return ELEVATOR_BACK_MERGE; + else if (blk_rq_pos(rq) - bio_sectors(bio) == bio->bi_sector) + return ELEVATOR_FRONT_MERGE; + return ELEVATOR_NO_MERGE; +} diff --git a/block/blk.h b/block/blk.h index 7efd772336de..9c12f80882b0 100644 --- a/block/blk.h +++ b/block/blk.h @@ -137,6 +137,8 @@ int blk_attempt_req_merge(struct request_queue *q, struct request *rq, struct request *next); void blk_recalc_rq_segments(struct request *rq); void blk_rq_set_mixed_merge(struct request *rq); +bool blk_rq_merge_ok(struct request *rq, struct bio *bio); +int blk_try_merge(struct request *rq, struct bio *bio); void blk_queue_congestion_threshold(struct request_queue *q); diff --git a/block/elevator.c b/block/elevator.c index 91e18f8af9be..f016855a46b0 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -70,39 +70,9 @@ static int elv_iosched_allow_merge(struct request *rq, struct bio *bio) /* * can we safely merge with this request? */ -int elv_rq_merge_ok(struct request *rq, struct bio *bio) +bool 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->bi_rw & REQ_DISCARD) != (rq->bio->bi_rw & REQ_DISCARD)) - return 0; - - /* - * Don't merge discard requests and secure discard requests - */ - if ((bio->bi_rw & REQ_SECURE) != (rq->bio->bi_rw & REQ_SECURE)) - return 0; - - /* - * different data direction or already started, don't merge - */ - if (bio_data_dir(bio) != rq_data_dir(rq)) - return 0; - - /* - * must be same device and not a special request - */ - if (rq->rq_disk != bio->bi_bdev->bd_disk || rq->special) - return 0; - - /* - * only merge integrity protected bio into ditto rq - */ - if (bio_integrity(bio) != blk_integrity_rq(rq)) + if (!blk_rq_merge_ok(rq, bio)) return 0; if (!elv_iosched_allow_merge(rq, bio)) @@ -112,23 +82,6 @@ int elv_rq_merge_ok(struct request *rq, struct bio *bio) } EXPORT_SYMBOL(elv_rq_merge_ok); -int elv_try_merge(struct request *__rq, struct bio *bio) -{ - int ret = ELEVATOR_NO_MERGE; - - /* - * we can merge and sequence is ok, check if it's possible - */ - if (elv_rq_merge_ok(__rq, bio)) { - if (blk_rq_pos(__rq) + blk_rq_sectors(__rq) == bio->bi_sector) - ret = ELEVATOR_BACK_MERGE; - else if (blk_rq_pos(__rq) - bio_sectors(bio) == bio->bi_sector) - ret = ELEVATOR_FRONT_MERGE; - } - - return ret; -} - static struct elevator_type *elevator_find(const char *name) { struct elevator_type *e; @@ -478,8 +431,8 @@ int elv_merge(struct request_queue *q, struct request **req, struct bio *bio) /* * First try one-hit cache. */ - if (q->last_merge) { - ret = elv_try_merge(q->last_merge, bio); + if (q->last_merge && elv_rq_merge_ok(q->last_merge, bio)) { + ret = blk_try_merge(q->last_merge, bio); if (ret != ELEVATOR_NO_MERGE) { *req = q->last_merge; return ret; diff --git a/include/linux/elevator.h b/include/linux/elevator.h index c24f3d7fbf1e..d6dfb65c8885 100644 --- a/include/linux/elevator.h +++ b/include/linux/elevator.h @@ -122,7 +122,6 @@ extern void elv_dispatch_add_tail(struct request_queue *, struct request *); extern void elv_add_request(struct request_queue *, struct request *, int); extern void __elv_add_request(struct request_queue *, struct request *, int); extern int elv_merge(struct request_queue *, struct request **, struct bio *); -extern int elv_try_merge(struct request *, struct bio *); extern void elv_merge_requests(struct request_queue *, struct request *, struct request *); extern void elv_merged_request(struct request_queue *, struct request *, int); @@ -155,7 +154,7 @@ extern ssize_t elv_iosched_store(struct request_queue *, const char *, size_t); extern int elevator_init(struct request_queue *, char *); extern void elevator_exit(struct elevator_queue *); extern int elevator_change(struct request_queue *, const char *); -extern int elv_rq_merge_ok(struct request *, struct bio *); +extern bool elv_rq_merge_ok(struct request *, struct bio *); /* * Helper functions. -- cgit v1.2.3 From 07c2bd37350c9b1af71b35d05f16e300a6602948 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 8 Feb 2012 09:19:42 +0100 Subject: block: don't call elevator callbacks for plug merges Plug merge calls two elevator callbacks outside queue lock - elevator_allow_merge_fn() and elevator_bio_merged_fn(). Although attempt_plug_merge() suggests that elevator is guaranteed to be there through the existing request on the plug list, nothing prevents plug merge from calling into dying or initializing elevator. For regular merges, bypass ensures elvpriv count to reach zero, which in turn prevents merges as all !ELVPRIV requests get REQ_SOFTBARRIER from forced back insertion. Plug merge doesn't check ELVPRIV, and, as the requests haven't gone through elevator insertion yet, it doesn't have SOFTBARRIER set allowing merges on a bypassed queue. This, for example, leads to the following crash during elevator switch. BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 IP: [] cfq_allow_merge+0x49/0xa0 PGD 112cbc067 PUD 115d5c067 PMD 0 Oops: 0000 [#1] PREEMPT SMP CPU 1 Modules linked in: deadline_iosched Pid: 819, comm: dd Not tainted 3.3.0-rc2-work+ #76 Bochs Bochs RIP: 0010:[] [] cfq_allow_merge+0x49/0xa0 RSP: 0018:ffff8801143a38f8 EFLAGS: 00010297 RAX: 0000000000000000 RBX: ffff88011817ce28 RCX: ffff880116eb6cc0 RDX: 0000000000000000 RSI: ffff880118056e20 RDI: ffff8801199512f8 RBP: ffff8801143a3908 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000001 R11: 0000000000000000 R12: ffff880118195708 R13: ffff880118052aa0 R14: ffff8801143a3d50 R15: ffff880118195708 FS: 00007f19f82cb700(0000) GS:ffff88011fc80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000008 CR3: 0000000112c6a000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process dd (pid: 819, threadinfo ffff8801143a2000, task ffff880116eb6cc0) Stack: ffff88011817ce28 ffff880118195708 ffff8801143a3928 ffffffff81391bba ffff88011817ce28 ffff880118195708 ffff8801143a3948 ffffffff81391bf1 ffff88011817ce28 0000000000000000 ffff8801143a39a8 ffffffff81398e3e Call Trace: [] elv_rq_merge_ok+0x4a/0x60 [] elv_try_merge+0x21/0x40 [] blk_queue_bio+0x8e/0x390 [] generic_make_request+0xca/0x100 [] submit_bio+0x74/0x100 [] __blockdev_direct_IO+0x1ce2/0x3450 [] blkdev_direct_IO+0x57/0x60 [] generic_file_aio_read+0x6d5/0x760 [] do_sync_read+0xe2/0x120 [] vfs_read+0xc5/0x180 [] sys_read+0x51/0x90 [] system_call_fastpath+0x16/0x1b There are multiple ways to fix this including making plug merge check ELVPRIV; however, * Calling into elevator outside queue lock is confusing and error-prone. * Requests on plug list aren't known to the elevator. They aren't on the elevator yet, so there's no elevator specific state to update. * Given the nature of plug merges - collecting bio's for the same purpose from the same issuer - elevator specific restrictions aren't applicable. So, simply don't call into elevator methods from plug merge by moving elv_bio_merged() from bio_attempt_*_merge() to blk_queue_bio(), and using blk_try_merge() in attempt_plug_merge(). This is based on Jens' patch to skip elevator_allow_merge_fn() from plug merge. Note that this makes per-cgroup merged stats skip plug merging. Signed-off-by: Tejun Heo LKML-Reference: <4F16F3CA.90904@kernel.dk> Original-patch-by: Jens Axboe Signed-off-by: Jens Axboe --- block/blk-core.c | 19 +++++++++---------- block/cfq-iosched.c | 15 ++++----------- include/linux/elevator.h | 6 ------ 3 files changed, 13 insertions(+), 27 deletions(-) (limited to 'include/linux') diff --git a/block/blk-core.c b/block/blk-core.c index fa697bf691eb..3a78b00edd71 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1212,7 +1212,6 @@ static bool bio_attempt_back_merge(struct request_queue *q, struct request *req, req->ioprio = ioprio_best(req->ioprio, bio_prio(bio)); drive_stat_acct(req, 0); - elv_bio_merged(q, req, bio); return true; } @@ -1243,7 +1242,6 @@ static bool bio_attempt_front_merge(struct request_queue *q, req->ioprio = ioprio_best(req->ioprio, bio_prio(bio)); drive_stat_acct(req, 0); - elv_bio_merged(q, req, bio); return true; } @@ -1257,13 +1255,12 @@ static bool bio_attempt_front_merge(struct request_queue *q, * on %current's plugged list. Returns %true if merge was successful, * otherwise %false. * - * This function is called without @q->queue_lock; however, elevator is - * accessed iff there already are requests on the plugged list which in - * turn guarantees validity of the elevator. - * - * Note that, on successful merge, elevator operation - * elevator_bio_merged_fn() will be called without queue lock. Elevator - * must be ready for this. + * Plugging coalesces IOs from the same issuer for the same purpose without + * going through @q->queue_lock. As such it's more of an issuing mechanism + * than scheduling, and the request, while may have elvpriv data, is not + * added on the elevator at this point. In addition, we don't have + * reliable access to the elevator outside queue lock. Only check basic + * merging parameters without querying the elevator. */ static bool attempt_plug_merge(struct request_queue *q, struct bio *bio, unsigned int *request_count) @@ -1282,7 +1279,7 @@ static bool attempt_plug_merge(struct request_queue *q, struct bio *bio, (*request_count)++; - if (rq->q != q || !elv_rq_merge_ok(rq, bio)) + if (rq->q != q || !blk_rq_merge_ok(rq, bio)) continue; el_ret = blk_try_merge(rq, bio); @@ -1347,12 +1344,14 @@ void blk_queue_bio(struct request_queue *q, struct bio *bio) el_ret = elv_merge(q, &req, bio); if (el_ret == ELEVATOR_BACK_MERGE) { if (bio_attempt_back_merge(q, req, bio)) { + elv_bio_merged(q, req, bio); if (!attempt_back_merge(q, req)) elv_merged_request(q, req, el_ret); goto out_unlock; } } else if (el_ret == ELEVATOR_FRONT_MERGE) { if (bio_attempt_front_merge(q, req, bio)) { + elv_bio_merged(q, req, bio); if (!attempt_front_merge(q, req)) elv_merged_request(q, req, el_ret); goto out_unlock; diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index 5684df6848bc..d0ba50533668 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -1699,18 +1699,11 @@ static int cfq_allow_merge(struct request_queue *q, struct request *rq, /* * Lookup the cfqq that this bio will be queued with and allow - * merge only if rq is queued there. This function can be called - * from plug merge without queue_lock. In such cases, ioc of @rq - * and %current are guaranteed to be equal. Avoid lookup which - * requires queue_lock by using @rq's cic. + * merge only if rq is queued there. */ - if (current->io_context == RQ_CIC(rq)->icq.ioc) { - cic = RQ_CIC(rq); - } else { - cic = cfq_cic_lookup(cfqd, current->io_context); - if (!cic) - return false; - } + cic = cfq_cic_lookup(cfqd, current->io_context); + if (!cic) + return false; cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio)); return cfqq == RQ_CFQQ(rq); diff --git a/include/linux/elevator.h b/include/linux/elevator.h index d6dfb65c8885..7d4e0356f329 100644 --- a/include/linux/elevator.h +++ b/include/linux/elevator.h @@ -42,12 +42,6 @@ struct elevator_ops elevator_merged_fn *elevator_merged_fn; elevator_merge_req_fn *elevator_merge_req_fn; elevator_allow_merge_fn *elevator_allow_merge_fn; - - /* - * Used for both plugged list and elevator merging and in the - * former case called without queue_lock. Read comment on top of - * attempt_plug_merge() for details. - */ elevator_bio_merged_fn *elevator_bio_merged_fn; elevator_dispatch_fn *elevator_dispatch_fn; -- cgit v1.2.3 From cdccaa9467b982d57b139818d15e1e994feca372 Mon Sep 17 00:00:00 2001 From: Paolo Bonzini Date: Wed, 8 Feb 2012 20:03:14 +0100 Subject: cdrom: move shared static to cdrom_device_info The keeplocked variable in the cdrom driver is shared across multiple drives, but set in per-device ioctls. Move it to the per-device struct, avoiding that the setting on one drive affects the driver's behavior when closing another. [ Impact: limit udev's confusion to one drive when a CD burning program unlocks the CD door at the end of burning. ] Signed-off-by: Paolo Bonzini Signed-off-by: Jens Axboe --- drivers/cdrom/cdrom.c | 12 +++++------- include/linux/cdrom.h | 3 ++- 2 files changed, 7 insertions(+), 8 deletions(-) (limited to 'include/linux') diff --git a/drivers/cdrom/cdrom.c b/drivers/cdrom/cdrom.c index 8fefe59f52a7..d620b4495745 100644 --- a/drivers/cdrom/cdrom.c +++ b/drivers/cdrom/cdrom.c @@ -286,8 +286,6 @@ /* used to tell the module to turn on full debugging messages */ static bool debug; -/* used to keep tray locked at all times */ -static int keeplocked; /* default compatibility mode */ static bool autoclose=1; static bool autoeject; @@ -1204,7 +1202,7 @@ void cdrom_release(struct cdrom_device_info *cdi, fmode_t mode) cdinfo(CD_CLOSE, "Use count for \"/dev/%s\" now zero\n", cdi->name); cdrom_dvd_rw_close_write(cdi); - if ((cdo->capability & CDC_LOCK) && !keeplocked) { + if ((cdo->capability & CDC_LOCK) && !cdi->keeplocked) { cdinfo(CD_CLOSE, "Unlocking door!\n"); cdo->lock_door(cdi, 0); } @@ -1371,7 +1369,7 @@ static int cdrom_select_disc(struct cdrom_device_info *cdi, int slot) curslot = info->hdr.curslot; kfree(info); - if (cdi->use_count > 1 || keeplocked) { + if (cdi->use_count > 1 || cdi->keeplocked) { if (slot == CDSL_CURRENT) { return curslot; } else { @@ -2289,7 +2287,7 @@ static int cdrom_ioctl_eject(struct cdrom_device_info *cdi) if (!CDROM_CAN(CDC_OPEN_TRAY)) return -ENOSYS; - if (cdi->use_count != 1 || keeplocked) + if (cdi->use_count != 1 || cdi->keeplocked) return -EBUSY; if (CDROM_CAN(CDC_LOCK)) { int ret = cdi->ops->lock_door(cdi, 0); @@ -2316,7 +2314,7 @@ static int cdrom_ioctl_eject_sw(struct cdrom_device_info *cdi, if (!CDROM_CAN(CDC_OPEN_TRAY)) return -ENOSYS; - if (keeplocked) + if (cdi->keeplocked) return -EBUSY; cdi->options &= ~(CDO_AUTO_CLOSE | CDO_AUTO_EJECT); @@ -2447,7 +2445,7 @@ static int cdrom_ioctl_lock_door(struct cdrom_device_info *cdi, if (!CDROM_CAN(CDC_LOCK)) return -EDRIVE_CANT_DO_THIS; - keeplocked = arg ? 1 : 0; + cdi->keeplocked = arg ? 1 : 0; /* * Don't unlock the door on multiple opens by default, but allow diff --git a/include/linux/cdrom.h b/include/linux/cdrom.h index 35eae4b67503..7c48029dffe6 100644 --- a/include/linux/cdrom.h +++ b/include/linux/cdrom.h @@ -952,7 +952,8 @@ struct cdrom_device_info { char name[20]; /* name of the device type */ /* per-device flags */ __u8 sanyo_slot : 2; /* Sanyo 3 CD changer support */ - __u8 reserved : 6; /* not used yet */ + __u8 keeplocked : 1; /* CDROM_LOCKDOOR status */ + __u8 reserved : 5; /* not used yet */ int cdda_method; /* see flags */ __u8 last_sense; __u8 media_written; /* dirty flag, DVD+RW bookkeeping */ -- cgit v1.2.3