diff for duplicates of <9524396.xaYzBpxu59@natalenko.name> diff --git a/a/1.txt b/N1/1.txt index 92c3b6f..ed21414 100644 --- a/a/1.txt +++ b/N1/1.txt @@ -2,57 +2,56 @@ Again, Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name> -On sobota 2. z=C3=A1=C5=99=C3=AD 2017 15:08:32 CEST Ming Lei wrote: +On sobota 2. září 2017 15:08:32 CEST Ming Lei wrote: > Hi, ->=20 +> > The current SCSI quiesce isn't safe and easy to trigger I/O deadlock. ->=20 -> Once SCSI device is put into QUIESCE, no new request except for RQF_PREEM= -PT +> +> Once SCSI device is put into QUIESCE, no new request except for RQF_PREEMPT > can be dispatched to SCSI successfully, and scsi_device_quiesce() just > simply waits for completion of I/Os dispatched to SCSI stack. It isn't > enough at all. ->=20 +> > Because new request still can be allocated, but all the allocated > requests can't be dispatched successfully, so request pool can be > consumed up easily. ->=20 +> > Then request with RQF_PREEMPT can't be allocated, and system may > hang forever, such as during system suspend or SCSI domain alidation. ->=20 +> > Both IO hang inside system suspend[1] or SCSI domain validation > were reported before. ->=20 +> > This patch introduces preempt freez, and tries to solve the issue > by preempt freezing block queue during SCSI quiesce, and allows > to allocate request of RQF_PREEMPT when queue is preempt-frozen. ->=20 +> > Both SCSI and SCSI_MQ have this IO deadlock issue, this patch fixes > them all by introducing blk_freeze_queue_preempt() and > blk_unfreeze_queue_preempt(), also unifying current interfaces for > freezing queue between block legacy and blk-mq. ->=20 +> > Oleksandr has verified that this patchset V2 fixes his I/O hang > during suspend/resume cycle. ->=20 +> > V3: > - introduce q->preempt_unfreezing to fix one bug of preempt freeze > - call blk_queue_enter_live() only when queue is preempt frozen > - cleanup a bit on the implementation of preempt freeze > - only patch 6 and 7 are changed ->=20 +> > V2: > - drop the 1st patch in V1 because percpu_ref_is_dying() is > enough as pointed by Tejun > - introduce preempt version of blk_[freeze|unfreeze]_queue > - sync between preempt freeze and normal freeze > - fix warning from percpu-refcount as reported by Oleksandr ->=20 ->=20 -> [1] https://marc.info/?t=3D150340250100013&r=3D3&w=3D2 ->=20 ->=20 ->=20 +> +> +> [1] https://marc.info/?t=150340250100013&r=3&w=2 +> +> +> > Ming Lei (8): > blk-mq: rename blk_mq_unfreeze_queue as blk_unfreeze_queue > blk-mq: rename blk_mq_freeze_queue as blk_freeze_queue @@ -63,13 +62,12 @@ PT > block: allow to allocate req with REQF_PREEMPT when queue is preempt > frozen > SCSI: preempt freeze block queue when SCSI device is put into quiesce ->=20 +> > block/bfq-iosched.c | 2 +- > block/blk-cgroup.c | 8 +-- > block/blk-core.c | 53 ++++++++++++--- > block/blk-mq.c | 170 -> +++++++++++++++++++++++++++++++++++++++-------- block/blk-mq.h = -|=20 +> +++++++++++++++++++++++++++++++++++++++-------- block/blk-mq.h | > 1 - > block/blk.h | 17 +++++ > block/elevator.c | 4 +- diff --git a/a/content_digest b/N1/content_digest index 62577bf..e4cd981 100644 --- a/a/content_digest +++ b/N1/content_digest @@ -18,57 +18,56 @@ "\n" "Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>\n" "\n" - "On sobota 2. z=C3=A1=C5=99=C3=AD 2017 15:08:32 CEST Ming Lei wrote:\n" + "On sobota 2. z\303\241\305\231\303\255 2017 15:08:32 CEST Ming Lei wrote:\n" "> Hi,\n" - ">=20\n" + "> \n" "> The current SCSI quiesce isn't safe and easy to trigger I/O deadlock.\n" - ">=20\n" - "> Once SCSI device is put into QUIESCE, no new request except for RQF_PREEM=\n" - "PT\n" + "> \n" + "> Once SCSI device is put into QUIESCE, no new request except for RQF_PREEMPT\n" "> can be dispatched to SCSI successfully, and scsi_device_quiesce() just\n" "> simply waits for completion of I/Os dispatched to SCSI stack. It isn't\n" "> enough at all.\n" - ">=20\n" + "> \n" "> Because new request still can be allocated, but all the allocated\n" "> requests can't be dispatched successfully, so request pool can be\n" "> consumed up easily.\n" - ">=20\n" + "> \n" "> Then request with RQF_PREEMPT can't be allocated, and system may\n" "> hang forever, such as during system suspend or SCSI domain alidation.\n" - ">=20\n" + "> \n" "> Both IO hang inside system suspend[1] or SCSI domain validation\n" "> were reported before.\n" - ">=20\n" + "> \n" "> This patch introduces preempt freez, and tries to solve the issue\n" "> by preempt freezing block queue during SCSI quiesce, and allows\n" "> to allocate request of RQF_PREEMPT when queue is preempt-frozen.\n" - ">=20\n" + "> \n" "> Both SCSI and SCSI_MQ have this IO deadlock issue, this patch fixes\n" "> them all by introducing blk_freeze_queue_preempt() and\n" "> blk_unfreeze_queue_preempt(), also unifying current interfaces for\n" "> freezing queue between block legacy and blk-mq.\n" - ">=20\n" + "> \n" "> Oleksandr has verified that this patchset V2 fixes his I/O hang\n" "> during suspend/resume cycle.\n" - ">=20\n" + "> \n" "> V3:\n" "> \t- introduce q->preempt_unfreezing to fix one bug of preempt freeze\n" "> \t- call blk_queue_enter_live() only when queue is preempt frozen\n" "> \t- cleanup a bit on the implementation of preempt freeze\n" "> \t- only patch 6 and 7 are changed\n" - ">=20\n" + "> \n" "> V2:\n" "> \t- drop the 1st patch in V1 because percpu_ref_is_dying() is\n" "> \tenough as pointed by Tejun\n" "> \t- introduce preempt version of blk_[freeze|unfreeze]_queue\n" "> \t- sync between preempt freeze and normal freeze\n" "> \t- fix warning from percpu-refcount as reported by Oleksandr\n" - ">=20\n" - ">=20\n" - "> [1] https://marc.info/?t=3D150340250100013&r=3D3&w=3D2\n" - ">=20\n" - ">=20\n" - ">=20\n" + "> \n" + "> \n" + "> [1] https://marc.info/?t=150340250100013&r=3&w=2\n" + "> \n" + "> \n" + "> \n" "> Ming Lei (8):\n" "> blk-mq: rename blk_mq_unfreeze_queue as blk_unfreeze_queue\n" "> blk-mq: rename blk_mq_freeze_queue as blk_freeze_queue\n" @@ -79,13 +78,12 @@ "> block: allow to allocate req with REQF_PREEMPT when queue is preempt\n" "> frozen\n" "> SCSI: preempt freeze block queue when SCSI device is put into quiesce\n" - ">=20\n" + "> \n" "> block/bfq-iosched.c | 2 +-\n" "> block/blk-cgroup.c | 8 +--\n" "> block/blk-core.c | 53 ++++++++++++---\n" "> block/blk-mq.c | 170\n" - "> +++++++++++++++++++++++++++++++++++++++-------- block/blk-mq.h =\n" - "|=20\n" + "> +++++++++++++++++++++++++++++++++++++++-------- block/blk-mq.h | \n" "> 1 -\n" "> block/blk.h | 17 +++++\n" "> block/elevator.c | 4 +-\n" @@ -97,4 +95,4 @@ "> include/linux/blkdev.h | 21 +++++-\n" > 13 files changed, 273 insertions(+), 66 deletions(-) -571b18db0a70b803c45df65f71d7ee5daf00c87ca0cd37c17fd669760a082f40 +6abc1a89dd8be43380424823ad153468dbe17cfd6a3a0805fca9b451c5195565
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.