linux_dsm_epyc7002/block/blk.h

365 lines
11 KiB
C
Raw Normal View History

#ifndef BLK_INTERNAL_H
#define BLK_INTERNAL_H
#include <linux/idr.h>
#include <linux/blk-mq.h>
#include "blk-mq.h"
/* Amount of time in which a process may batch requests */
#define BLK_BATCH_TIME (HZ/50UL)
/* Number of requests a "batching" process may submit */
#define BLK_BATCH_REQ 32
/* Max future timer expiry for timeouts */
#define BLK_MAX_TIMEOUT (5 * HZ)
#ifdef CONFIG_DEBUG_FS
extern struct dentry *blk_debugfs_root;
#endif
struct blk_flush_queue {
unsigned int flush_queue_delayed:1;
unsigned int flush_pending_idx:1;
unsigned int flush_running_idx:1;
unsigned long flush_pending_since;
struct list_head flush_queue[2];
struct list_head flush_data_in_flight;
struct request *flush_rq;
blk-mq: fix race between timeout and freeing request Inside timeout handler, blk_mq_tag_to_rq() is called to retrieve the request from one tag. This way is obviously wrong because the request can be freed any time and some fiedds of the request can't be trusted, then kernel oops might be triggered[1]. Currently wrt. blk_mq_tag_to_rq(), the only special case is that the flush request can share same tag with the request cloned from, and the two requests can't be active at the same time, so this patch fixes the above issue by updating tags->rqs[tag] with the active request(either flush rq or the request cloned from) of the tag. Also blk_mq_tag_to_rq() gets much simplified with this patch. Given blk_mq_tag_to_rq() is mainly for drivers and the caller must make sure the request can't be freed, so in bt_for_each() this helper is replaced with tags->rqs[tag]. [1] kernel oops log [ 439.696220] BUG: unable to handle kernel NULL pointer dereference at 0000000000000158^M [ 439.697162] IP: [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M [ 439.700653] PGD 7ef765067 PUD 7ef764067 PMD 0 ^M [ 439.700653] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC ^M [ 439.700653] Dumping ftrace buffer:^M [ 439.700653] (ftrace buffer empty)^M [ 439.700653] Modules linked in: nbd ipv6 kvm_intel kvm serio_raw^M [ 439.700653] CPU: 6 PID: 2779 Comm: stress-ng-sigfd Not tainted 4.2.0-rc5-next-20150805+ #265^M [ 439.730500] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011^M [ 439.730500] task: ffff880605308000 ti: ffff88060530c000 task.ti: ffff88060530c000^M [ 439.730500] RIP: 0010:[<ffffffff812d89ba>] [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M [ 439.730500] RSP: 0018:ffff880819203da0 EFLAGS: 00010283^M [ 439.730500] RAX: ffff880811b0e000 RBX: ffff8800bb465f00 RCX: 0000000000000002^M [ 439.730500] RDX: 0000000000000000 RSI: 0000000000000202 RDI: 0000000000000000^M [ 439.730500] RBP: ffff880819203db0 R08: 0000000000000002 R09: 0000000000000000^M [ 439.730500] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000202^M [ 439.730500] R13: ffff880814104800 R14: 0000000000000002 R15: ffff880811a2ea00^M [ 439.730500] FS: 00007f165b3f5740(0000) GS:ffff880819200000(0000) knlGS:0000000000000000^M [ 439.730500] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b^M [ 439.730500] CR2: 0000000000000158 CR3: 00000007ef766000 CR4: 00000000000006e0^M [ 439.730500] Stack:^M [ 439.730500] 0000000000000008 ffff8808114eed90 ffff880819203e00 ffffffff812dc104^M [ 439.755663] ffff880819203e40 ffffffff812d9f5e 0000020000000000 ffff8808114eed80^M [ 439.755663] Call Trace:^M [ 439.755663] <IRQ> ^M [ 439.755663] [<ffffffff812dc104>] bt_for_each+0x6e/0xc8^M [ 439.755663] [<ffffffff812d9f5e>] ? blk_mq_rq_timed_out+0x6a/0x6a^M [ 439.755663] [<ffffffff812d9f5e>] ? blk_mq_rq_timed_out+0x6a/0x6a^M [ 439.755663] [<ffffffff812dc1b3>] blk_mq_tag_busy_iter+0x55/0x5e^M [ 439.755663] [<ffffffff812d88b4>] ? blk_mq_bio_to_request+0x38/0x38^M [ 439.755663] [<ffffffff812d8911>] blk_mq_rq_timer+0x5d/0xd4^M [ 439.755663] [<ffffffff810a3e10>] call_timer_fn+0xf7/0x284^M [ 439.755663] [<ffffffff810a3d1e>] ? call_timer_fn+0x5/0x284^M [ 439.755663] [<ffffffff812d88b4>] ? blk_mq_bio_to_request+0x38/0x38^M [ 439.755663] [<ffffffff810a46d6>] run_timer_softirq+0x1ce/0x1f8^M [ 439.755663] [<ffffffff8104c367>] __do_softirq+0x181/0x3a4^M [ 439.755663] [<ffffffff8104c76e>] irq_exit+0x40/0x94^M [ 439.755663] [<ffffffff81031482>] smp_apic_timer_interrupt+0x33/0x3e^M [ 439.755663] [<ffffffff815559a4>] apic_timer_interrupt+0x84/0x90^M [ 439.755663] <EOI> ^M [ 439.755663] [<ffffffff81554350>] ? _raw_spin_unlock_irq+0x32/0x4a^M [ 439.755663] [<ffffffff8106a98b>] finish_task_switch+0xe0/0x163^M [ 439.755663] [<ffffffff8106a94d>] ? finish_task_switch+0xa2/0x163^M [ 439.755663] [<ffffffff81550066>] __schedule+0x469/0x6cd^M [ 439.755663] [<ffffffff8155039b>] schedule+0x82/0x9a^M [ 439.789267] [<ffffffff8119b28b>] signalfd_read+0x186/0x49a^M [ 439.790911] [<ffffffff8106d86a>] ? wake_up_q+0x47/0x47^M [ 439.790911] [<ffffffff811618c2>] __vfs_read+0x28/0x9f^M [ 439.790911] [<ffffffff8117a289>] ? __fget_light+0x4d/0x74^M [ 439.790911] [<ffffffff811620a7>] vfs_read+0x7a/0xc6^M [ 439.790911] [<ffffffff8116292b>] SyS_read+0x49/0x7f^M [ 439.790911] [<ffffffff81554c17>] entry_SYSCALL_64_fastpath+0x12/0x6f^M [ 439.790911] Code: 48 89 e5 e8 a9 b8 e7 ff 5d c3 0f 1f 44 00 00 55 89 f2 48 89 e5 41 54 41 89 f4 53 48 8b 47 60 48 8b 1c d0 48 8b 7b 30 48 8b 53 38 <48> 8b 87 58 01 00 00 48 85 c0 75 09 48 8b 97 88 0c 00 00 eb 10 ^M [ 439.790911] RIP [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M [ 439.790911] RSP <ffff880819203da0>^M [ 439.790911] CR2: 0000000000000158^M [ 439.790911] ---[ end trace d40af58949325661 ]---^M Cc: <stable@vger.kernel.org> Signed-off-by: Ming Lei <ming.lei@canonical.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2015-08-09 14:41:51 +07:00
/*
* flush_rq shares tag with this rq, both can't be active
* at the same time
*/
struct request *orig_rq;
spinlock_t mq_flush_lock;
};
extern struct kmem_cache *blk_requestq_cachep;
blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li <shli@fusionio.com> Alexander Gordeev <agordeev@redhat.com> Christoph Hellwig <hch@infradead.org> Mike Christie <michaelc@cs.wisc.edu> Matias Bjorling <m@bjorling.me> Jeff Moyer <jmoyer@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-24 15:20:05 +07:00
extern struct kmem_cache *request_cachep;
extern struct kobj_type blk_queue_ktype;
extern struct ida blk_queue_ida;
static inline struct blk_flush_queue *blk_get_flush_queue(
struct request_queue *q, struct blk_mq_ctx *ctx)
{
if (q->mq_ops)
return blk_mq_map_queue(q, ctx->cpu)->fq;
return q->fq;
}
static inline void __blk_get_queue(struct request_queue *q)
{
kobject_get(&q->kobj);
}
struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
int node, int cmd_size);
void blk_free_flush_queue(struct blk_flush_queue *q);
int blk_init_rl(struct request_list *rl, struct request_queue *q,
gfp_t gfp_mask);
block: Avoid that blk_exit_rl() triggers a use-after-free Since the introduction of .init_rq_fn() and .exit_rq_fn() it is essential that the memory allocated for struct request_queue stays around until all blk_exit_rl() calls have finished. Hence make blk_init_rl() take a reference on struct request_queue. This patch fixes the following crash: general protection fault: 0000 [#2] SMP CPU: 3 PID: 28 Comm: ksoftirqd/3 Tainted: G D 4.12.0-rc2-dbg+ #2 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014 task: ffff88013a108040 task.stack: ffffc9000071c000 RIP: 0010:free_request_size+0x1a/0x30 RSP: 0018:ffffc9000071fd38 EFLAGS: 00010202 RAX: 6b6b6b6b6b6b6b6b RBX: ffff880067362a88 RCX: 0000000000000003 RDX: ffff880067464178 RSI: ffff880067362a88 RDI: ffff880135ea4418 RBP: ffffc9000071fd40 R08: 0000000000000000 R09: 0000000100180009 R10: ffffc9000071fd38 R11: ffffffff81110800 R12: ffff88006752d3d8 R13: ffff88006752d3d8 R14: ffff88013a108040 R15: 000000000000000a FS: 0000000000000000(0000) GS:ffff88013fd80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007fa8ec1edb00 CR3: 0000000138ee8000 CR4: 00000000001406e0 Call Trace: mempool_destroy.part.10+0x21/0x40 mempool_destroy+0xe/0x10 blk_exit_rl+0x12/0x20 blkg_free+0x4d/0xa0 __blkg_release_rcu+0x59/0x170 rcu_process_callbacks+0x260/0x4e0 __do_softirq+0x116/0x250 smpboot_thread_fn+0x123/0x1e0 kthread+0x109/0x140 ret_from_fork+0x31/0x40 Fixes: commit e9c787e65c0c ("scsi: allocate scsi_cmnd structures as part of struct request") Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Acked-by: Tejun Heo <tj@kernel.org> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Cc: Jan Kara <jack@suse.cz> Cc: <stable@vger.kernel.org> # v4.11+ Signed-off-by: Jens Axboe <axboe@fb.com>
2017-06-01 04:43:45 +07:00
void blk_exit_rl(struct request_queue *q, struct request_list *rl);
void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
struct bio *bio);
void blk_queue_bypass_start(struct request_queue *q);
void blk_queue_bypass_end(struct request_queue *q);
void __blk_queue_free_tags(struct request_queue *q);
block: generic request_queue reference counting Allow pmem, and other synchronous/bio-based block drivers, to fallback on a per-cpu reference count managed by the core for tracking queue live/dead state. The existing per-cpu reference count for the blk_mq case is promoted to be used in all block i/o scenarios. This involves initializing it by default, waiting for it to drop to zero at exit, and holding a live reference over the invocation of q->make_request_fn() in generic_make_request(). The blk_mq code continues to take its own reference per blk_mq request and retains the ability to freeze the queue, but the check that the queue is frozen is moved to generic_make_request(). This fixes crash signatures like the following: BUG: unable to handle kernel paging request at ffff880140000000 [..] Call Trace: [<ffffffff8145e8bf>] ? copy_user_handle_tail+0x5f/0x70 [<ffffffffa004e1e0>] pmem_do_bvec.isra.11+0x70/0xf0 [nd_pmem] [<ffffffffa004e331>] pmem_make_request+0xd1/0x200 [nd_pmem] [<ffffffff811c3162>] ? mempool_alloc+0x72/0x1a0 [<ffffffff8141f8b6>] generic_make_request+0xd6/0x110 [<ffffffff8141f966>] submit_bio+0x76/0x170 [<ffffffff81286dff>] submit_bh_wbc+0x12f/0x160 [<ffffffff81286e62>] submit_bh+0x12/0x20 [<ffffffff813395bd>] jbd2_write_superblock+0x8d/0x170 [<ffffffff8133974d>] jbd2_mark_journal_empty+0x5d/0x90 [<ffffffff813399cb>] jbd2_journal_destroy+0x24b/0x270 [<ffffffff810bc4ca>] ? put_pwq_unlocked+0x2a/0x30 [<ffffffff810bc6f5>] ? destroy_workqueue+0x225/0x250 [<ffffffff81303494>] ext4_put_super+0x64/0x360 [<ffffffff8124ab1a>] generic_shutdown_super+0x6a/0xf0 Cc: Jens Axboe <axboe@kernel.dk> Cc: Keith Busch <keith.busch@intel.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Suggested-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2015-10-22 00:20:12 +07:00
void blk_freeze_queue(struct request_queue *q);
static inline void blk_queue_enter_live(struct request_queue *q)
{
/*
* Given that running in generic_make_request() context
* guarantees that a live reference against q_usage_counter has
* been established, further references under that same context
* need not check that the queue has been frozen (marked dead).
*/
percpu_ref_get(&q->q_usage_counter);
}
#ifdef CONFIG_BLK_DEV_INTEGRITY
void blk_flush_integrity(void);
bool __bio_integrity_endio(struct bio *);
static inline bool bio_integrity_endio(struct bio *bio)
{
if (bio_integrity(bio))
return __bio_integrity_endio(bio);
return true;
}
#else
static inline void blk_flush_integrity(void)
{
}
static inline bool bio_integrity_endio(struct bio *bio)
{
return true;
}
#endif
void blk_timeout_work(struct work_struct *work);
unsigned long blk_rq_timeout(unsigned long timeout);
void blk_add_timer(struct request *req);
void blk_delete_timer(struct request *);
blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li <shli@fusionio.com> Alexander Gordeev <agordeev@redhat.com> Christoph Hellwig <hch@infradead.org> Mike Christie <michaelc@cs.wisc.edu> Matias Bjorling <m@bjorling.me> Jeff Moyer <jmoyer@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-24 15:20:05 +07:00
bool bio_attempt_front_merge(struct request_queue *q, struct request *req,
struct bio *bio);
bool bio_attempt_back_merge(struct request_queue *q, struct request *req,
struct bio *bio);
bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
struct bio *bio);
blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li <shli@fusionio.com> Alexander Gordeev <agordeev@redhat.com> Christoph Hellwig <hch@infradead.org> Mike Christie <michaelc@cs.wisc.edu> Matias Bjorling <m@bjorling.me> Jeff Moyer <jmoyer@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-24 15:20:05 +07:00
bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
unsigned int *request_count,
struct request **same_queue_rq);
unsigned int blk_plug_queued_count(struct request_queue *q);
blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li <shli@fusionio.com> Alexander Gordeev <agordeev@redhat.com> Christoph Hellwig <hch@infradead.org> Mike Christie <michaelc@cs.wisc.edu> Matias Bjorling <m@bjorling.me> Jeff Moyer <jmoyer@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-24 15:20:05 +07:00
void blk_account_io_start(struct request *req, bool new_io);
void blk_account_io_completion(struct request *req, unsigned int bytes);
void blk_account_io_done(struct request *req);
/*
* Internal atomic flags for request handling
*/
enum rq_atomic_flags {
REQ_ATOM_COMPLETE = 0,
blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li <shli@fusionio.com> Alexander Gordeev <agordeev@redhat.com> Christoph Hellwig <hch@infradead.org> Mike Christie <michaelc@cs.wisc.edu> Matias Bjorling <m@bjorling.me> Jeff Moyer <jmoyer@redhat.com> Acked-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-24 15:20:05 +07:00
REQ_ATOM_STARTED,
REQ_ATOM_POLL_SLEPT,
};
/*
* EH timer and IO completion will both attempt to 'grab' the request, make
* sure that only one of them succeeds
*/
static inline int blk_mark_rq_complete(struct request *rq)
{
return test_and_set_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
}
static inline void blk_clear_rq_complete(struct request *rq)
{
clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
}
/*
* Internal elevator interface
*/
#define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
block: reimplement FLUSH/FUA to support merge The current FLUSH/FUA support has evolved from the implementation which had to perform queue draining. As such, sequencing is done queue-wide one flush request after another. However, with the draining requirement gone, there's no reason to keep the queue-wide sequential approach. This patch reimplements FLUSH/FUA support such that each FLUSH/FUA request is sequenced individually. The actual FLUSH execution is double buffered and whenever a request wants to execute one for either PRE or POSTFLUSH, it queues on the pending queue. Once certain conditions are met, a flush request is issued and on its completion all pending requests proceed to the next sequence. This allows arbitrary merging of different type of flushes. How they are merged can be primarily controlled and tuned by adjusting the above said 'conditions' used to determine when to issue the next flush. This is inspired by Darrick's patches to merge multiple zero-data flushes which helps workloads with highly concurrent fsync requests. * As flush requests are never put on the IO scheduler, request fields used for flush share space with rq->rb_node. rq->completion_data is moved out of the union. This increases the request size by one pointer. As rq->elevator_private* are used only by the iosched too, it is possible to reduce the request size further. However, to do that, we need to modify request allocation path such that iosched data is not allocated for flush requests. * FLUSH/FUA processing happens on insertion now instead of dispatch. - Comments updated as per Vivek and Mike. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: "Darrick J. Wong" <djwong@us.ibm.com> Cc: Shaohua Li <shli@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-01-25 18:43:54 +07:00
void blk_insert_flush(struct request *rq);
static inline struct request *__elv_next_request(struct request_queue *q)
{
struct request *rq;
struct blk_flush_queue *fq = blk_get_flush_queue(q, NULL);
WARN_ON_ONCE(q->mq_ops);
while (1) {
block: reimplement FLUSH/FUA to support merge The current FLUSH/FUA support has evolved from the implementation which had to perform queue draining. As such, sequencing is done queue-wide one flush request after another. However, with the draining requirement gone, there's no reason to keep the queue-wide sequential approach. This patch reimplements FLUSH/FUA support such that each FLUSH/FUA request is sequenced individually. The actual FLUSH execution is double buffered and whenever a request wants to execute one for either PRE or POSTFLUSH, it queues on the pending queue. Once certain conditions are met, a flush request is issued and on its completion all pending requests proceed to the next sequence. This allows arbitrary merging of different type of flushes. How they are merged can be primarily controlled and tuned by adjusting the above said 'conditions' used to determine when to issue the next flush. This is inspired by Darrick's patches to merge multiple zero-data flushes which helps workloads with highly concurrent fsync requests. * As flush requests are never put on the IO scheduler, request fields used for flush share space with rq->rb_node. rq->completion_data is moved out of the union. This increases the request size by one pointer. As rq->elevator_private* are used only by the iosched too, it is possible to reduce the request size further. However, to do that, we need to modify request allocation path such that iosched data is not allocated for flush requests. * FLUSH/FUA processing happens on insertion now instead of dispatch. - Comments updated as per Vivek and Mike. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: "Darrick J. Wong" <djwong@us.ibm.com> Cc: Shaohua Li <shli@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-01-25 18:43:54 +07:00
if (!list_empty(&q->queue_head)) {
rq = list_entry_rq(q->queue_head.next);
block: reimplement FLUSH/FUA to support merge The current FLUSH/FUA support has evolved from the implementation which had to perform queue draining. As such, sequencing is done queue-wide one flush request after another. However, with the draining requirement gone, there's no reason to keep the queue-wide sequential approach. This patch reimplements FLUSH/FUA support such that each FLUSH/FUA request is sequenced individually. The actual FLUSH execution is double buffered and whenever a request wants to execute one for either PRE or POSTFLUSH, it queues on the pending queue. Once certain conditions are met, a flush request is issued and on its completion all pending requests proceed to the next sequence. This allows arbitrary merging of different type of flushes. How they are merged can be primarily controlled and tuned by adjusting the above said 'conditions' used to determine when to issue the next flush. This is inspired by Darrick's patches to merge multiple zero-data flushes which helps workloads with highly concurrent fsync requests. * As flush requests are never put on the IO scheduler, request fields used for flush share space with rq->rb_node. rq->completion_data is moved out of the union. This increases the request size by one pointer. As rq->elevator_private* are used only by the iosched too, it is possible to reduce the request size further. However, to do that, we need to modify request allocation path such that iosched data is not allocated for flush requests. * FLUSH/FUA processing happens on insertion now instead of dispatch. - Comments updated as per Vivek and Mike. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: "Darrick J. Wong" <djwong@us.ibm.com> Cc: Shaohua Li <shli@kernel.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-01-25 18:43:54 +07:00
return rq;
}
/*
* Flush request is running and flush request isn't queueable
* in the drive, we can hold the queue till flush request is
* finished. Even we don't do this, driver can't dispatch next
* requests and will requeue them. And this can improve
* throughput too. For example, we have request flush1, write1,
* flush 2. flush1 is dispatched, then queue is hold, write1
* isn't inserted to queue. After flush1 is finished, flush2
* will be dispatched. Since disk cache is already clean,
* flush2 will be finished very soon, so looks like flush2 is
* folded to flush1.
* Since the queue is hold, a flag is set to indicate the queue
* should be restarted later. Please see flush_end_io() for
* details.
*/
if (fq->flush_pending_idx != fq->flush_running_idx &&
!queue_flush_queueable(q)) {
fq->flush_queue_delayed = 1;
return NULL;
}
block: __elv_next_request() shouldn't call into the elevator if bypassing request_queue bypassing is used to suppress higher-level function of a request_queue so that they can be switched, reconfigured and shut down. A request_queue does the followings while bypassing. * bypasses elevator and io_cq association and queues requests directly to the FIFO dispatch queue. * bypasses block cgroup request_list lookup and always uses the root request_list. Once confirmed to be bypassing, specific elevator and block cgroup policy implementations can assume that nothing is in flight for them and perform various operations which would be dangerous otherwise. Such confirmation is acheived by short-circuiting all new requests directly to the dispatch queue and waiting for all the requests which were issued before to finish. Unfortunately, while the request allocating and draining sides were properly handled, we forgot to actually plug the request dispatch path. Even after bypassing mode is confirmed, if the attached driver tries to fetch a request and the dispatch queue is empty, __elv_next_request() would invoke the current elevator's elevator_dispatch_fn() callback. As all in-flight requests were drained, the elevator wouldn't contain any request but once bypass is confirmed we don't even know whether the elevator is even there. It might be in the process of being switched and half torn down. Frank Mayhar reports that this actually happened while switching elevators, leading to an oops. Let's fix it by making __elv_next_request() avoid invoking the elevator_dispatch_fn() callback if the queue is bypassing. It already avoids invoking the callback if the queue is dying. As a dying queue is guaranteed to be bypassing, we can simply replace blk_queue_dying() check with blk_queue_bypass(). Reported-by: Frank Mayhar <fmayhar@google.com> References: http://lkml.kernel.org/g/1390319905.20232.38.camel@bobble.lax.corp.google.com Cc: stable@vger.kernel.org Tested-by: Frank Mayhar <fmayhar@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2014-01-30 04:56:16 +07:00
if (unlikely(blk_queue_bypass(q)) ||
!q->elevator->type->ops.sq.elevator_dispatch_fn(q, 0))
return NULL;
}
}
static inline void elv_activate_rq(struct request_queue *q, struct request *rq)
{
struct elevator_queue *e = q->elevator;
if (e->type->ops.sq.elevator_activate_req_fn)
e->type->ops.sq.elevator_activate_req_fn(q, rq);
}
static inline void elv_deactivate_rq(struct request_queue *q, struct request *rq)
{
struct elevator_queue *e = q->elevator;
if (e->type->ops.sq.elevator_deactivate_req_fn)
e->type->ops.sq.elevator_deactivate_req_fn(q, rq);
}
struct hd_struct *__disk_get_part(struct gendisk *disk, int partno);
#ifdef CONFIG_FAIL_IO_TIMEOUT
int blk_should_fake_timeout(struct request_queue *);
ssize_t part_timeout_show(struct device *, struct device_attribute *, char *);
ssize_t part_timeout_store(struct device *, struct device_attribute *,
const char *, size_t);
#else
static inline int blk_should_fake_timeout(struct request_queue *q)
{
return 0;
}
#endif
int ll_back_merge_fn(struct request_queue *q, struct request *req,
struct bio *bio);
int ll_front_merge_fn(struct request_queue *q, struct request *req,
struct bio *bio);
struct request *attempt_back_merge(struct request_queue *q, struct request *rq);
struct request *attempt_front_merge(struct request_queue *q, struct request *rq);
int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
struct request *next);
void blk_recalc_rq_segments(struct request *rq);
block: implement mixed merge of different failfast requests Failfast has characteristics from other attributes. When issuing, executing and successuflly completing requests, failfast doesn't make any difference. It only affects how a request is handled on failure. Allowing requests with different failfast settings to be merged cause normal IOs to fail prematurely while not allowing has performance penalties as failfast is used for read aheads which are likely to be located near in-flight or to-be-issued normal IOs. This patch introduces the concept of 'mixed merge'. A request is a mixed merge if it is merge of segments which require different handling on failure. Currently the only mixable attributes are failfast ones (or lack thereof). When a bio with different failfast settings is added to an existing request or requests of different failfast settings are merged, the merged request is marked mixed. Each bio carries failfast settings and the request always tracks failfast state of the first bio. When the request fails, blk_rq_err_bytes() can be used to determine how many bytes can be safely failed without crossing into an area which requires further retrials. This allows request merging regardless of failfast settings while keeping the failure handling correct. This patch only implements mixed merge but doesn't enable it. The next one will update SCSI to make use of mixed merge. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Niel Lambrechts <niel.lambrechts@gmail.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
2009-07-03 15:48:17 +07:00
void blk_rq_set_mixed_merge(struct request *rq);
bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
void blk_queue_congestion_threshold(struct request_queue *q);
int blk_dev_init(void);
/*
* Return the threshold (number of used requests) at which the queue is
* considered to be congested. It include a little hysteresis to keep the
* context switch rate down.
*/
static inline int queue_congestion_on_threshold(struct request_queue *q)
{
return q->nr_congestion_on;
}
/*
* The threshold at which a queue is considered to be uncongested
*/
static inline int queue_congestion_off_threshold(struct request_queue *q)
{
return q->nr_congestion_off;
}
extern int blk_update_nr_requests(struct request_queue *, unsigned int);
/*
* Contribute to IO statistics IFF:
*
* a) it's attached to a gendisk, and
* b) the queue had IO stats enabled when this request was started, and
* c) it's a file system request
*/
static inline int blk_do_io_stat(struct request *rq)
{
return rq->rq_disk &&
(rq->rq_flags & RQF_IO_STAT) &&
!blk_rq_is_passthrough(rq);
}
static inline void req_set_nomerge(struct request_queue *q, struct request *req)
{
req->cmd_flags |= REQ_NOMERGE;
if (req == q->last_merge)
q->last_merge = NULL;
}
/*
* Internal io_context interface
*/
void get_io_context(struct io_context *ioc);
struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
gfp_t gfp_mask);
void ioc_clear_queue(struct request_queue *q);
int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
/**
* rq_ioc - determine io_context for request allocation
* @bio: request being allocated is for this bio (can be %NULL)
*
* Determine io_context to use for request allocation for @bio. May return
* %NULL if %current->io_context doesn't exist.
*/
static inline struct io_context *rq_ioc(struct bio *bio)
{
#ifdef CONFIG_BLK_CGROUP
if (bio && bio->bi_ioc)
return bio->bi_ioc;
#endif
return current->io_context;
}
/**
* create_io_context - try to create task->io_context
* @gfp_mask: allocation mask
* @node: allocation node
*
* If %current->io_context is %NULL, allocate a new io_context and install
* it. Returns the current %current->io_context which may be %NULL if
* allocation failed.
*
* Note that this function can't be called with IRQ disabled because
* task_lock which protects %current->io_context is IRQ-unsafe.
*/
static inline struct io_context *create_io_context(gfp_t gfp_mask, int node)
{
WARN_ON_ONCE(irqs_disabled());
if (unlikely(!current->io_context))
create_task_io_context(current, gfp_mask, node);
return current->io_context;
}
/*
* Internal throttling interface
*/
#ifdef CONFIG_BLK_DEV_THROTTLING
block: fix request_queue lifetime handling by making blk_queue_cleanup() properly shutdown request_queue is refcounted but actually depdends on lifetime management from the queue owner - on blk_cleanup_queue(), block layer expects that there's no request passing through request_queue and no new one will. This is fundamentally broken. The queue owner (e.g. SCSI layer) doesn't have a way to know whether there are other active users before calling blk_cleanup_queue() and other users (e.g. bsg) don't have any guarantee that the queue is and would stay valid while it's holding a reference. With delay added in blk_queue_bio() before queue_lock is grabbed, the following oops can be easily triggered when a device is removed with in-flight IOs. sd 0:0:1:0: [sdb] Stopping disk ata1.01: disabled general protection fault: 0000 [#1] PREEMPT SMP CPU 2 Modules linked in: Pid: 648, comm: test_rawio Not tainted 3.1.0-rc3-work+ #56 Bochs Bochs RIP: 0010:[<ffffffff8137d651>] [<ffffffff8137d651>] elv_rqhash_find+0x61/0x100 ... Process test_rawio (pid: 648, threadinfo ffff880019efa000, task ffff880019ef8a80) ... Call Trace: [<ffffffff8137d774>] elv_merge+0x84/0xe0 [<ffffffff81385b54>] blk_queue_bio+0xf4/0x400 [<ffffffff813838ea>] generic_make_request+0xca/0x100 [<ffffffff81383994>] submit_bio+0x74/0x100 [<ffffffff811c53ec>] dio_bio_submit+0xbc/0xc0 [<ffffffff811c610e>] __blockdev_direct_IO+0x92e/0xb40 [<ffffffff811c39f7>] blkdev_direct_IO+0x57/0x60 [<ffffffff8113b1c5>] generic_file_aio_read+0x6d5/0x760 [<ffffffff8118c1ca>] do_sync_read+0xda/0x120 [<ffffffff8118ce55>] vfs_read+0xc5/0x180 [<ffffffff8118cfaa>] sys_pread64+0x9a/0xb0 [<ffffffff81afaf6b>] system_call_fastpath+0x16/0x1b This happens because blk_queue_cleanup() destroys the queue and elevator whether IOs are in progress or not and DEAD tests are sprinkled in the request processing path without proper synchronization. Similar problem exists for blk-throtl. On queue cleanup, blk-throtl is shutdown whether it has requests in it or not. Depending on timing, it either oopses or throttled bios are lost putting tasks which are waiting for bio completion into eternal D state. The way it should work is having the usual clear distinction between shutdown and release. Shutdown drains all currently pending requests, marks the queue dead, and performs partial teardown of the now unnecessary part of the queue. Even after shutdown is complete, reference holders are still allowed to issue requests to the queue although they will be immmediately failed. The rest of teardown happens on release. This patch makes the following changes to make blk_queue_cleanup() behave as proper shutdown. * QUEUE_FLAG_DEAD is now set while holding both q->exit_mutex and queue_lock. * Unsynchronized DEAD check in generic_make_request_checks() removed. This couldn't make any meaningful difference as the queue could die after the check. * blk_drain_queue() updated such that it can drain all requests and is now called during cleanup. * blk_throtl updated such that it checks DEAD on grabbing queue_lock, drains all throttled bios during cleanup and free td when queue is released. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Vivek Goyal <vgoyal@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2011-10-19 19:42:16 +07:00
extern void blk_throtl_drain(struct request_queue *q);
extern int blk_throtl_init(struct request_queue *q);
extern void blk_throtl_exit(struct request_queue *q);
extern void blk_throtl_register_queue(struct request_queue *q);
#else /* CONFIG_BLK_DEV_THROTTLING */
block: fix request_queue lifetime handling by making blk_queue_cleanup() properly shutdown request_queue is refcounted but actually depdends on lifetime management from the queue owner - on blk_cleanup_queue(), block layer expects that there's no request passing through request_queue and no new one will. This is fundamentally broken. The queue owner (e.g. SCSI layer) doesn't have a way to know whether there are other active users before calling blk_cleanup_queue() and other users (e.g. bsg) don't have any guarantee that the queue is and would stay valid while it's holding a reference. With delay added in blk_queue_bio() before queue_lock is grabbed, the following oops can be easily triggered when a device is removed with in-flight IOs. sd 0:0:1:0: [sdb] Stopping disk ata1.01: disabled general protection fault: 0000 [#1] PREEMPT SMP CPU 2 Modules linked in: Pid: 648, comm: test_rawio Not tainted 3.1.0-rc3-work+ #56 Bochs Bochs RIP: 0010:[<ffffffff8137d651>] [<ffffffff8137d651>] elv_rqhash_find+0x61/0x100 ... Process test_rawio (pid: 648, threadinfo ffff880019efa000, task ffff880019ef8a80) ... Call Trace: [<ffffffff8137d774>] elv_merge+0x84/0xe0 [<ffffffff81385b54>] blk_queue_bio+0xf4/0x400 [<ffffffff813838ea>] generic_make_request+0xca/0x100 [<ffffffff81383994>] submit_bio+0x74/0x100 [<ffffffff811c53ec>] dio_bio_submit+0xbc/0xc0 [<ffffffff811c610e>] __blockdev_direct_IO+0x92e/0xb40 [<ffffffff811c39f7>] blkdev_direct_IO+0x57/0x60 [<ffffffff8113b1c5>] generic_file_aio_read+0x6d5/0x760 [<ffffffff8118c1ca>] do_sync_read+0xda/0x120 [<ffffffff8118ce55>] vfs_read+0xc5/0x180 [<ffffffff8118cfaa>] sys_pread64+0x9a/0xb0 [<ffffffff81afaf6b>] system_call_fastpath+0x16/0x1b This happens because blk_queue_cleanup() destroys the queue and elevator whether IOs are in progress or not and DEAD tests are sprinkled in the request processing path without proper synchronization. Similar problem exists for blk-throtl. On queue cleanup, blk-throtl is shutdown whether it has requests in it or not. Depending on timing, it either oopses or throttled bios are lost putting tasks which are waiting for bio completion into eternal D state. The way it should work is having the usual clear distinction between shutdown and release. Shutdown drains all currently pending requests, marks the queue dead, and performs partial teardown of the now unnecessary part of the queue. Even after shutdown is complete, reference holders are still allowed to issue requests to the queue although they will be immmediately failed. The rest of teardown happens on release. This patch makes the following changes to make blk_queue_cleanup() behave as proper shutdown. * QUEUE_FLAG_DEAD is now set while holding both q->exit_mutex and queue_lock. * Unsynchronized DEAD check in generic_make_request_checks() removed. This couldn't make any meaningful difference as the queue could die after the check. * blk_drain_queue() updated such that it can drain all requests and is now called during cleanup. * blk_throtl updated such that it checks DEAD on grabbing queue_lock, drains all throttled bios during cleanup and free td when queue is released. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Vivek Goyal <vgoyal@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2011-10-19 19:42:16 +07:00
static inline void blk_throtl_drain(struct request_queue *q) { }
static inline int blk_throtl_init(struct request_queue *q) { return 0; }
static inline void blk_throtl_exit(struct request_queue *q) { }
static inline void blk_throtl_register_queue(struct request_queue *q) { }
#endif /* CONFIG_BLK_DEV_THROTTLING */
#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
extern ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page);
extern ssize_t blk_throtl_sample_time_store(struct request_queue *q,
const char *page, size_t count);
blk-throttle: add a simple idle detection A cgroup gets assigned a low limit, but the cgroup could never dispatch enough IO to cross the low limit. In such case, the queue state machine will remain in LIMIT_LOW state and all other cgroups will be throttled according to low limit. This is unfair for other cgroups. We should treat the cgroup idle and upgrade the state machine to lower state. We also have a downgrade logic. If the state machine upgrades because of cgroup idle (real idle), the state machine will downgrade soon as the cgroup is below its low limit. This isn't what we want. A more complicated case is cgroup isn't idle when queue is in LIMIT_LOW. But when queue gets upgraded to lower state, other cgroups could dispatch more IO and this cgroup can't dispatch enough IO, so the cgroup is below its low limit and looks like idle (fake idle). In this case, the queue should downgrade soon. The key to determine if we should do downgrade is to detect if cgroup is truely idle. Unfortunately it's very hard to determine if a cgroup is real idle. This patch uses the 'think time check' idea from CFQ for the purpose. Please note, the idea doesn't work for all workloads. For example, a workload with io depth 8 has disk utilization 100%, hence think time is 0, eg, not idle. But the workload can run higher bandwidth with io depth 16. Compared to io depth 16, the io depth 8 workload is idle. We use the idea to roughly determine if a cgroup is idle. We treat a cgroup idle if its think time is above a threshold (by default 1ms for SSD and 100ms for HD). The idea is think time above the threshold will start to harm performance. HD is much slower so a longer think time is ok. The patch (and the latter patches) uses 'unsigned long' to track time. We convert 'ns' to 'us' with 'ns >> 10'. This is fast but loses precision, should not a big deal. Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2017-03-28 00:51:41 +07:00
extern void blk_throtl_bio_endio(struct bio *bio);
blk-throttle: add a mechanism to estimate IO latency User configures latency target, but the latency threshold for each request size isn't fixed. For a SSD, the IO latency highly depends on request size. To calculate latency threshold, we sample some data, eg, average latency for request size 4k, 8k, 16k, 32k .. 1M. The latency threshold of each request size will be the sample latency (I'll call it base latency) plus latency target. For example, the base latency for request size 4k is 80us and user configures latency target 60us. The 4k latency threshold will be 80 + 60 = 140us. To sample data, we calculate the order base 2 of rounded up IO sectors. If the IO size is bigger than 1M, it will be accounted as 1M. Since the calculation does round up, the base latency will be slightly smaller than actual value. Also if there isn't any IO dispatched for a specific IO size, we will use the base latency of smaller IO size for this IO size. But we shouldn't sample data at any time. The base latency is supposed to be latency where disk isn't congested, because we use latency threshold to schedule IOs between cgroups. If disk is congested, the latency is higher, using it for scheduling is meaningless. Hence we only do the sampling when block throttling is in the LOW limit, with assumption disk isn't congested in such state. If the assumption isn't true, eg, low limit is too high, calculated latency threshold will be higher. Hard disk is completely different. Latency depends on spindle seek instead of request size. Currently this feature is SSD only, we probably can use a fixed threshold like 4ms for hard disk though. Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2017-03-28 05:19:42 +07:00
extern void blk_throtl_stat_add(struct request *rq, u64 time);
blk-throttle: add a simple idle detection A cgroup gets assigned a low limit, but the cgroup could never dispatch enough IO to cross the low limit. In such case, the queue state machine will remain in LIMIT_LOW state and all other cgroups will be throttled according to low limit. This is unfair for other cgroups. We should treat the cgroup idle and upgrade the state machine to lower state. We also have a downgrade logic. If the state machine upgrades because of cgroup idle (real idle), the state machine will downgrade soon as the cgroup is below its low limit. This isn't what we want. A more complicated case is cgroup isn't idle when queue is in LIMIT_LOW. But when queue gets upgraded to lower state, other cgroups could dispatch more IO and this cgroup can't dispatch enough IO, so the cgroup is below its low limit and looks like idle (fake idle). In this case, the queue should downgrade soon. The key to determine if we should do downgrade is to detect if cgroup is truely idle. Unfortunately it's very hard to determine if a cgroup is real idle. This patch uses the 'think time check' idea from CFQ for the purpose. Please note, the idea doesn't work for all workloads. For example, a workload with io depth 8 has disk utilization 100%, hence think time is 0, eg, not idle. But the workload can run higher bandwidth with io depth 16. Compared to io depth 16, the io depth 8 workload is idle. We use the idea to roughly determine if a cgroup is idle. We treat a cgroup idle if its think time is above a threshold (by default 1ms for SSD and 100ms for HD). The idea is think time above the threshold will start to harm performance. HD is much slower so a longer think time is ok. The patch (and the latter patches) uses 'unsigned long' to track time. We convert 'ns' to 'us' with 'ns >> 10'. This is fast but loses precision, should not a big deal. Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2017-03-28 00:51:41 +07:00
#else
static inline void blk_throtl_bio_endio(struct bio *bio) { }
blk-throttle: add a mechanism to estimate IO latency User configures latency target, but the latency threshold for each request size isn't fixed. For a SSD, the IO latency highly depends on request size. To calculate latency threshold, we sample some data, eg, average latency for request size 4k, 8k, 16k, 32k .. 1M. The latency threshold of each request size will be the sample latency (I'll call it base latency) plus latency target. For example, the base latency for request size 4k is 80us and user configures latency target 60us. The 4k latency threshold will be 80 + 60 = 140us. To sample data, we calculate the order base 2 of rounded up IO sectors. If the IO size is bigger than 1M, it will be accounted as 1M. Since the calculation does round up, the base latency will be slightly smaller than actual value. Also if there isn't any IO dispatched for a specific IO size, we will use the base latency of smaller IO size for this IO size. But we shouldn't sample data at any time. The base latency is supposed to be latency where disk isn't congested, because we use latency threshold to schedule IOs between cgroups. If disk is congested, the latency is higher, using it for scheduling is meaningless. Hence we only do the sampling when block throttling is in the LOW limit, with assumption disk isn't congested in such state. If the assumption isn't true, eg, low limit is too high, calculated latency threshold will be higher. Hard disk is completely different. Latency depends on spindle seek instead of request size. Currently this feature is SSD only, we probably can use a fixed threshold like 4ms for hard disk though. Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Jens Axboe <axboe@fb.com>
2017-03-28 05:19:42 +07:00
static inline void blk_throtl_stat_add(struct request *rq, u64 time) { }
#endif
#ifdef CONFIG_BOUNCE
extern int init_emergency_isa_pool(void);
extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
#else
static inline int init_emergency_isa_pool(void)
{
return 0;
}
static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
{
}
#endif /* CONFIG_BOUNCE */
#endif /* BLK_INTERNAL_H */