linux/block
Muchun Song bb2f15b522 block: fix ordering between checking BLK_MQ_S_STOPPED request adding
commit 96a9fe64bf upstream.

Supposing first scenario with a virtio_blk driver.

CPU0                        CPU1

blk_mq_try_issue_directly()
  __blk_mq_issue_directly()
    q->mq_ops->queue_rq()
      virtio_queue_rq()
        blk_mq_stop_hw_queue()
                            virtblk_done()
  blk_mq_request_bypass_insert()  1) store
                              blk_mq_start_stopped_hw_queue()
                                clear_bit(BLK_MQ_S_STOPPED)       3) store
                                blk_mq_run_hw_queue()
                                  if (!blk_mq_hctx_has_pending()) 4) load
                                    return
                                  blk_mq_sched_dispatch_requests()
  blk_mq_run_hw_queue()
    if (!blk_mq_hctx_has_pending())
      return
    blk_mq_sched_dispatch_requests()
      if (blk_mq_hctx_stopped())  2) load
        return
      __blk_mq_sched_dispatch_requests()

Supposing another scenario.

CPU0                        CPU1

blk_mq_requeue_work()
  blk_mq_insert_request() 1) store
                            virtblk_done()
                              blk_mq_start_stopped_hw_queue()
  blk_mq_run_hw_queues()        clear_bit(BLK_MQ_S_STOPPED)       3) store
                                blk_mq_run_hw_queue()
                                  if (!blk_mq_hctx_has_pending()) 4) load
                                    return
                                  blk_mq_sched_dispatch_requests()
    if (blk_mq_hctx_stopped())  2) load
      continue
    blk_mq_run_hw_queue()

Both scenarios are similar, the full memory barrier should be inserted
between 1) and 2), as well as between 3) and 4) to make sure that either
CPU0 sees BLK_MQ_S_STOPPED is cleared or CPU1 sees dispatch list.
Otherwise, either CPU will not rerun the hardware queue causing
starvation of the request.

The easy way to fix it is to add the essential full memory barrier into
helper of blk_mq_hctx_stopped(). In order to not affect the fast path
(hardware queue is not stopped most of the time), we only insert the
barrier into the slow path. Actually, only slow path needs to care about
missing of dispatching the request to the low-level device driver.

Fixes: 320ae51fee ("blk-mq: new multi-queue block IO queueing mechanism")
Cc: stable@vger.kernel.org
Cc: Muchun Song <muchun.song@linux.dev>
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20241014092934.53630-4-songmuchun@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
History 2024-12-14 19:51:17 +01:00
..
partitions block: fix potential invalid pointer dereference in blk_add_partition 2024-10-17 15:10:48 +02:00
badblocks.c treewide: Use fallthrough pseudo-keyword 2020-08-23 17:36:59 -05:00
bdev.c block: move GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE to disk->event_flags 2023-09-19 12:23:02 +02:00
bfq-cgroup.c block, bfq: fix uaf for bfqq in bic_set_bfqq() 2023-02-09 11:26:36 +01:00
bfq-iosched.c block, bfq: fix procress reference leakage for bfqq in merge chain 2024-11-01 01:52:35 +01:00
bfq-iosched.h bfq: Get rid of __bio_blkcg() usage 2022-06-09 10:23:19 +02:00
bfq-wf2q.c block/bfq_wf2q: correct weight to ioprio 2022-04-08 14:23:55 +02:00
bio-integrity.c block: initialize integrity buffer to zero before writing it to media 2024-08-19 05:44:50 +02:00
bio.c block: prevent an integer overflow in bvec_try_merge_hw_page 2024-02-23 08:54:44 +01:00
blk-cgroup-rwstat.c blk-cgroup: Fix the recursive blkg rwstat 2021-03-05 11:32:15 -07:00
blk-cgroup-rwstat.h blk-cgroup: separate out blkg_rwstat under CONFIG_BLK_CGROUP_RWSTAT 2019-11-07 12:28:13 -07:00
blk-cgroup.c blk-cgroup: bypass blkcg_deactivate_policy after destroying 2023-12-20 15:17:39 +01:00
blk-core.c blk-core: use pr_warn_ratelimited() in bio_check_ro() 2023-11-20 11:08:29 +01:00
blk-crypto-fallback.c block: rename BIO_MAX_PAGES to BIO_MAX_VECS 2021-03-11 07:47:48 -07:00
blk-crypto-internal.h blk-mq: release crypto keyslot before reporting I/O complete 2023-05-11 23:00:16 +09:00
blk-crypto.c blk-crypto: make blk_crypto_evict_key() more robust 2023-05-11 23:00:16 +09:00
blk-exec.c block: return errors from blk_execute_rq() 2021-06-30 15:35:45 -06:00
blk-flush.c block: Fix fsync always failed if once failed 2022-01-27 11:05:25 +01:00
blk-integrity.c blk-integrity: register sysfs attributes on struct device 2024-10-17 15:11:33 +02:00
blk-ioc.c block: fix default IO priority handling again 2022-08-11 13:07:50 +02:00
blk-iocost.c blk_iocost: fix more out of bound shifts 2024-10-17 15:11:25 +02:00
blk-iolatency.c block: don't allow the same type rq_qos add more than once 2022-08-17 14:24:24 +02:00
blk-ioprio.c block: Introduce the ioprio rq-qos policy 2021-06-21 15:03:40 -06:00
blk-ioprio.h block: Introduce the ioprio rq-qos policy 2021-06-21 15:03:40 -06:00
blk-lib.c block: export blk_next_bio() 2021-06-17 15:51:20 +02:00
blk-map.c block-map: add __GFP_ZERO flag for alloc_page in function bio_copy_kern 2022-03-08 19:12:31 +01:00
blk-merge.c blk-mq: release crypto keyslot before reporting I/O complete 2023-05-11 23:00:16 +09:00
blk-mq-cpumap.c blk-mq: remove the calling of local_memory_node() 2020-10-20 07:08:17 -06:00
blk-mq-debugfs-zoned.c block: Cleanup license notice 2019-01-17 21:21:40 -07:00
blk-mq-debugfs.c blk-mq: don't create hctx debugfs dir until q->debugfs_dir is created 2022-08-17 14:23:12 +02:00
blk-mq-debugfs.h blk-mq: no need to check return value of debugfs_create functions 2019-06-13 03:00:30 -06:00
blk-mq-pci.c block: Fix blk_mq_*_map_queues() kernel-doc headers 2019-05-31 15:12:34 -06:00
blk-mq-rdma.c block: Fix blk_mq_*_map_queues() kernel-doc headers 2019-05-31 15:12:34 -06:00
blk-mq-sched.c blk-mq: correct stale comment of .get_budget 2023-03-10 09:39:04 +01:00
blk-mq-sched.h blk: Fix lock inversion between ioc lock and bfqd lock 2021-06-24 18:43:55 -06:00
blk-mq-sysfs.c blk-mq: fix possible memleak when register 'hctx' failed 2022-12-31 13:14:43 +01:00
blk-mq-tag.c blk-mq: avoid to iterate over stale request 2021-09-12 19:32:43 -06:00
blk-mq-tag.h blk-mq: Some tag allocation code refactoring 2021-05-24 06:47:22 -06:00
blk-mq-virtio.c blk-mq: Fix typo in comment 2020-03-17 20:55:21 +01:00
blk-mq.c block: fix ordering between checking BLK_MQ_S_STOPPED request adding 2024-12-14 19:51:17 +01:00
blk-mq.h block: fix ordering between checking BLK_MQ_S_STOPPED request adding 2024-12-14 19:51:17 +01:00
blk-pm.c scsi: block: pm: Always set request queue runtime active in blk_post_runtime_resume() 2022-01-27 11:04:15 +01:00
blk-pm.h block: Remove unused blk_pm_*() function definitions 2021-02-22 06:33:48 -07:00
blk-rq-qos.c blk-rq-qos: fix crash on rq_qos_wait vs. rq_qos_wake_function race 2024-10-22 15:40:43 +02:00
blk-rq-qos.h block: don't allow the same type rq_qos add more than once 2022-08-17 14:24:24 +02:00
blk-settings.c block: Clear zone limits for a non-zoned stacked queue 2024-04-10 16:18:35 +02:00
blk-stat.c block: prevent division by zero in blk_rq_stat_sum() 2024-04-13 13:01:45 +02:00
blk-stat.h block: deactivate blk_stat timer in wbt_disable_default() 2018-12-12 06:47:51 -07:00
blk-sysfs.c block: fix use-after-free of q->q_usage_counter 2023-10-10 21:59:05 +02:00
blk-throttle.c blk-throttle: fix lockdep warning of "cgroup_mutex or RCU read lock required!" 2023-12-20 15:17:39 +01:00
blk-timeout.c block: blk-timeout: delete duplicated word 2020-07-31 16:29:47 -06:00
blk-wbt.c blk-wbt: fix that 'rwb->wc' is always set to 1 in wbt_init() 2022-10-26 12:35:54 +02:00
blk-wbt.h blk-wbt: introduce a new disable state to prevent false positive by rwb_enabled() 2021-06-21 15:03:41 -06:00
blk-zoned.c block: Hold invalidate_lock in BLKRESETZONE ioctl 2021-11-18 19:17:15 +01:00
blk.h blk-integrity: register sysfs attributes on struct device 2024-10-17 15:11:33 +02:00
bounce.c block: use memcpy_from_bvec in __blk_queue_bounce 2021-08-02 13:37:28 -06:00
bsg-lib.c scsi: bsg-lib: Fix commands without data transfer in bsg_transport_sg_io_fn() 2021-08-01 13:21:40 -04:00
bsg.c scsi: bsg: Fix device unregistration 2021-09-14 00:22:15 -04:00
disk-events.c block: increment diskseq on all media change events 2023-07-23 13:47:20 +02:00
elevator.c block: Fix elevator_get_default() checking for NULL q->tag_set 2024-11-17 15:06:24 +01:00
fops.c block: Don't invalidate pagecache for invalid falloc modes 2024-01-15 18:51:07 +01:00
genhd.c blk-integrity: register sysfs attributes on struct device 2024-10-17 15:11:33 +02:00
holder.c block: drop unused includes in <linux/genhd.h> 2022-03-16 14:23:46 +01:00
ioctl.c block/ioctl: prefer different overflow check 2024-07-05 09:14:25 +02:00
ioprio.c block: fix default IO priority handling again 2022-08-11 13:07:50 +02:00
Kconfig SCSI misc on 20210902 2021-09-02 15:09:46 -07:00
Kconfig.iosched Revert "block/mq-deadline: Add cgroup support" 2021-08-11 13:47:26 -06:00
keyslot-manager.c blk-crypto: make blk_crypto_evict_key() more robust 2023-05-11 23:00:16 +09:00
kyber-iosched.c kyber: avoid q->disk dereferences in trace points 2021-10-15 21:02:57 -06:00
Makefile block: move CONFIG_BLOCK guard to top Makefile 2022-12-14 11:37:30 +01:00
mq-deadline.c block: mq-deadline: Rename deadline_is_seq_writes() 2023-01-24 07:22:49 +01:00
opal_proto.h block: sed-opal: handle empty atoms when parsing response 2024-03-26 18:21:13 -04:00
sed-opal.c block: sed-opal: handle empty atoms when parsing response 2024-03-26 18:21:13 -04:00
t10-pi.c block: use bvec_kmap_local in t10_pi_type1_{prepare,complete} 2021-08-02 13:37:28 -06:00