| Message ID | 20210402200841.347696-1-sagi@grimberg.me |
|---|---|
| State | New |
| Headers | show |
| Series | [stable/5.4..5.8] nvme-mpath: replace direct_make_request with generic_make_request | expand |
On Fri, Apr 02, 2021 at 01:08:41PM -0700, Sagi Grimberg wrote: > The below patches caused a regression in a multipath setup: > Fixes: 9f98772ba307 ("nvme-rdma: fix controller reset hang during traffic") > Fixes: 2875b0aecabe ("nvme-tcp: fix controller reset hang during traffic") > > These patches on their own are correct because they fixed a controller reset > regression. > > When we reset/teardown a controller, we must freeze and quiesce the namespaces > request queues to make sure that we safely stop inflight I/O submissions. > Freeze is mandatory because if our hctx map changed between reconnects, > blk_mq_update_nr_hw_queues will immediately attempt to freeze the queue, and > if it still has pending submissions (that are still quiesced) it will hang. > This is what the above patches fixed. > > However, by freezing the namespaces request queues, and only unfreezing them > when we successfully reconnect, inflight submissions that are running > concurrently can now block grabbing the nshead srcu until either we successfully > reconnect or ctrl_loss_tmo expired (or the user explicitly disconnected). > > This caused a deadlock [1] when a different controller (different path on the > same subsystem) became live (i.e. optimized/non-optimized). This is because > nvme_mpath_set_live needs to synchronize the nshead srcu before requeueing I/O > in order to make sure that current_path is visible to future (re)submisions. > However the srcu lock is taken by a blocked submission on a frozen request > queue, and we have a deadlock. > > In recent kernels (v5.9+) direct_make_request was replaced by submit_bio_noacct > which does not have this issue because it bio_list will be active when > nvme-mpath calls submit_bio_noacct on the bottom device (because it was > populated when submit_bio was triggered on it. > > Hence, we need to fix all the kernels that were before submit_bio_noacct was > introduced. Why can we not just add submit_bio_noacct to the 5.4 kernel to correct this? What commit id is that? thanks, greg k-h
>> Hence, we need to fix all the kernels that were before submit_bio_noacct was >> introduced. > > Why can we not just add submit_bio_noacct to the 5.4 kernel to correct > this? What commit id is that? Hey Greg, submit_bio_noacct was applied as part of a rework by Christoph that I didn't feel was suitable as a stable candidate. The commit-id is: ed00aabd5eb9fb44d6aff1173234a2e911b9fead
On Tue, Apr 06, 2021 at 06:04:09PM -0700, Sagi Grimberg wrote: > >>> Hence, we need to fix all the kernels that were before submit_bio_noacct was >>> introduced. >> >> Why can we not just add submit_bio_noacct to the 5.4 kernel to correct >> this? What commit id is that? > > Hey Greg, > > submit_bio_noacct was applied as part of a rework by Christoph that I > didn't feel was suitable as a stable candidate. The commit-id is: > ed00aabd5eb9fb44d6aff1173234a2e911b9fead submit_bio_noacct really is just a new name for generic_make_request, as the old one was horribly misleading. So this does use submit_bio_noacct, just with its old name.
>>>> Hence, we need to fix all the kernels that were before submit_bio_noacct was >>>> introduced. >>> >>> Why can we not just add submit_bio_noacct to the 5.4 kernel to correct >>> this? What commit id is that? >> >> Hey Greg, >> >> submit_bio_noacct was applied as part of a rework by Christoph that I >> didn't feel was suitable as a stable candidate. The commit-id is: >> ed00aabd5eb9fb44d6aff1173234a2e911b9fead > > submit_bio_noacct really is just a new name for generic_make_request, > as the old one was horribly misleading. So this does use > submit_bio_noacct, just with its old name. commit ed00aabd5eb9fb44d6aff1173234a2e911b9fead does not apply cleanly on any of these kernels, so I think it's better to take this small one-liner instead of trying to fit a commit that changes the name treewide. Greg, what is your preference? backporting ed00aabd5eb9fb44d6aff1173234a2e911b9fead to the various kernels or to take this isolated one?
On Wed, Apr 07, 2021 at 04:18:30PM -0700, Sagi Grimberg wrote: > > > > > > Hence, we need to fix all the kernels that were before submit_bio_noacct was > > > > > introduced. > > > > > > > > Why can we not just add submit_bio_noacct to the 5.4 kernel to correct > > > > this? What commit id is that? > > > > > > Hey Greg, > > > > > > submit_bio_noacct was applied as part of a rework by Christoph that I > > > didn't feel was suitable as a stable candidate. The commit-id is: > > > ed00aabd5eb9fb44d6aff1173234a2e911b9fead > > > > submit_bio_noacct really is just a new name for generic_make_request, > > as the old one was horribly misleading. So this does use > > submit_bio_noacct, just with its old name. > > commit ed00aabd5eb9fb44d6aff1173234a2e911b9fead does not apply > cleanly on any of these kernels, so I think it's better to take this > small one-liner instead of trying to fit a commit that changes the > name treewide. > > Greg, what is your preference? backporting > ed00aabd5eb9fb44d6aff1173234a2e911b9fead to the various kernels > or to take this isolated one? I took just this change now, thanks. greg k-h
diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 041a755f936a..0d9d0bebe645 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -339,7 +339,7 @@ static blk_qc_t nvme_ns_head_make_request(struct request_queue *q, trace_block_bio_remap(bio->bi_disk->queue, bio, disk_devt(ns->head->disk), bio->bi_iter.bi_sector); - ret = direct_make_request(bio); + ret = generic_make_request(bio); } else if (nvme_available_path(head)) { dev_warn_ratelimited(dev, "no usable path - requeuing I/O\n");
The below patches caused a regression in a multipath setup: Fixes: 9f98772ba307 ("nvme-rdma: fix controller reset hang during traffic") Fixes: 2875b0aecabe ("nvme-tcp: fix controller reset hang during traffic") These patches on their own are correct because they fixed a controller reset regression. When we reset/teardown a controller, we must freeze and quiesce the namespaces request queues to make sure that we safely stop inflight I/O submissions. Freeze is mandatory because if our hctx map changed between reconnects, blk_mq_update_nr_hw_queues will immediately attempt to freeze the queue, and if it still has pending submissions (that are still quiesced) it will hang. This is what the above patches fixed. However, by freezing the namespaces request queues, and only unfreezing them when we successfully reconnect, inflight submissions that are running concurrently can now block grabbing the nshead srcu until either we successfully reconnect or ctrl_loss_tmo expired (or the user explicitly disconnected). This caused a deadlock [1] when a different controller (different path on the same subsystem) became live (i.e. optimized/non-optimized). This is because nvme_mpath_set_live needs to synchronize the nshead srcu before requeueing I/O in order to make sure that current_path is visible to future (re)submisions. However the srcu lock is taken by a blocked submission on a frozen request queue, and we have a deadlock. In recent kernels (v5.9+) direct_make_request was replaced by submit_bio_noacct which does not have this issue because it bio_list will be active when nvme-mpath calls submit_bio_noacct on the bottom device (because it was populated when submit_bio was triggered on it. Hence, we need to fix all the kernels that were before submit_bio_noacct was introduced. [1]: Workqueue: nvme-wq nvme_tcp_reconnect_ctrl_work [nvme_tcp] Call Trace: __schedule+0x293/0x730 schedule+0x33/0xa0 schedule_timeout+0x1d3/0x2f0 wait_for_completion+0xba/0x140 __synchronize_srcu.part.21+0x91/0xc0 synchronize_srcu_expedited+0x27/0x30 synchronize_srcu+0xce/0xe0 nvme_mpath_set_live+0x64/0x130 [nvme_core] nvme_update_ns_ana_state+0x2c/0x30 [nvme_core] nvme_update_ana_state+0xcd/0xe0 [nvme_core] nvme_parse_ana_log+0xa1/0x180 [nvme_core] nvme_read_ana_log+0x76/0x100 [nvme_core] nvme_mpath_init+0x122/0x180 [nvme_core] nvme_init_identify+0x80e/0xe20 [nvme_core] nvme_tcp_setup_ctrl+0x359/0x660 [nvme_tcp] nvme_tcp_reconnect_ctrl_work+0x24/0x70 [nvme_tcp] Signed-off-by: Sagi Grimberg <sagi@grimberg.me> --- Note: This patch does not exist in upstream, it is a pure backport fix that was just now found. The reason for that is that this specific issue exists on on kernels 5.4-5.8 as it was fixed in 5.9, and the patches that caused this was only backported to linux-5.4.y (which are correct as mentioned in the patch description) drivers/nvme/host/multipath.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)