[PATCH 1/2] blk-mq: Fix failed allocation path when mapping queues

[PATCH 1/2] blk-mq: Fix failed allocation path when mapping queues

[Gabriel Krisman Bertazi]

In blk_mq_map_swqueue, there is a memory optimization that frees the
tags of a queue that has gone unmapped.  Later, if that hctx is remapped
after another topology change, the tags need to be reallocated.

If this allocation fails, a simple WARN_ON triggers, but the block layer
ends up with an active hctx without any corresponding set of tags.
Then, any income IO to that hctx can trigger an Oops.

I can reproduce it consistently by running IO, flipping CPUs on and off
and eventually injecting a memory allocation failure in that path.

In the fix below, if the system experiences a failed allocation of any
hctx's tags, we remap all the ctxs of that queue to the hctx_0, which
should always keep it's tags.  There is a minor performance hit, since
our mapping just got worse after the error path, but this is
the simplest solution to handle this error path.  The performance hit
will disappear after another successful remap.

I considered dropping the memory optimization all together, but it
seemed a bad trade-off to handle this very specific error case.

My one concern about this patch is if remapping an arbitrary queue to
hctx_0 could result in outstanding requests getting submitted to the
wrong hctx.  I couldn't observe this happening during tests, but I'm not
entirely sure it'll never happen.  I believe the queue will be empty if
we are trying to allocate tags for it, unless it was using another alive
hctx queue and for some reason got reassigned to this new hctx.  is this
possible at all?

This should apply cleanly on top of Jen's for-next branch.

The Oops is the one below:

SP (3fff935ce4d0) is in userspace
1:mon> e
cpu 0x1: Vector: 300 (Data Access) at [c000000fe99eb110]
    pc: c0000000005e868c: __sbitmap_queue_get+0x2c/0x180
    lr: c000000000575328: __bt_get+0x48/0xd0
    sp: c000000fe99eb390
   msr: 900000010280b033
   dar: 28
 dsisr: 40000000
  current = 0xc000000fe9966800
  paca    = 0xc000000007e80300   softe: 0        irq_happened: 0x01
    pid   = 11035, comm = aio-stress
Linux version 4.8.0-rc6+ (root@bean) (gcc version 5.4.0 20160609
(Ubuntu/IBM 5.4.0-6ubuntu1~16.04.2) ) #3 SMP Mon Oct 10 20:16:53 CDT 2016
1:mon> s
[c000000fe99eb3d0] c000000000575328 __bt_get+0x48/0xd0
[c000000fe99eb400] c000000000575838 bt_get.isra.1+0x78/0x2d0
[c000000fe99eb480] c000000000575cb4 blk_mq_get_tag+0x44/0x100
[c000000fe99eb4b0] c00000000056f6f4 __blk_mq_alloc_request+0x44/0x220
[c000000fe99eb500] c000000000570050 blk_mq_map_request+0x100/0x1f0
[c000000fe99eb580] c000000000574650 blk_mq_make_request+0xf0/0x540
[c000000fe99eb640] c000000000561c44 generic_make_request+0x144/0x230
[c000000fe99eb690] c000000000561e00 submit_bio+0xd0/0x200
[c000000fe99eb740] c0000000003ef740 ext4_io_submit+0x90/0xb0
[c000000fe99eb770] c0000000003e95d8 ext4_writepages+0x588/0xdd0
[c000000fe99eb910] c00000000025a9f0 do_writepages+0x60/0xc0
[c000000fe99eb940] c000000000246c88 __filemap_fdatawrite_range+0xf8/0x180
[c000000fe99eb9e0] c000000000246f90 filemap_write_and_wait_range+0x70/0xf0
[c000000fe99eba20] c0000000003dd844 ext4_sync_file+0x214/0x540
[c000000fe99eba80] c000000000364718 vfs_fsync_range+0x78/0x130
[c000000fe99ebad0] c0000000003dd46c ext4_file_write_iter+0x35c/0x430
[c000000fe99ebb90] c00000000038c280 aio_run_iocb+0x3b0/0x450
[c000000fe99ebce0] c00000000038dc28 do_io_submit+0x368/0x730
[c000000fe99ebe30] c000000000009404 system_call+0x38/0xec

Signed-off-by: Gabriel Krisman Bertazi <krisman@linux.vnet.ibm.com>
Cc: Brian King <brking@linux.vnet.ibm.com>
Cc: Douglas Miller <dougmill@linux.vnet.ibm.com>
Cc: linux-block@vger.kernel.org
Cc: linux-scsi@vger.kernel.org
---
 block/blk-mq.c | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 3c5aaead71bb..7f7c4ba91adf 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1862,7 +1862,7 @@ static void blk_mq_init_cpu_queues(struct request_queue *q,
 static void blk_mq_map_swqueue(struct request_queue *q,
 			       const struct cpumask *online_mask)
 {
-	unsigned int i;
+	unsigned int i, hctx_idx;
 	struct blk_mq_hw_ctx *hctx;
 	struct blk_mq_ctx *ctx;
 	struct blk_mq_tag_set *set = q->tag_set;
@@ -1885,6 +1885,15 @@ static void blk_mq_map_swqueue(struct request_queue *q,
 		if (!cpumask_test_cpu(i, online_mask))
 			continue;
 
+		hctx_idx = q->mq_map[i];
+		/* unmapped hw queue can be remapped after CPU topo changed */
+		if (!set->tags[hctx_idx]) {
+			set->tags[hctx_idx] = blk_mq_init_rq_map(set, hctx_idx);
+
+			if (!set->tags[hctx_idx])
+				q->mq_map[i] = 0;
+		}
+
 		ctx = per_cpu_ptr(q->queue_ctx, i);
 		hctx = blk_mq_map_queue(q, i);
 
@@ -1901,7 +1910,10 @@ static void blk_mq_map_swqueue(struct request_queue *q,
 		 * disable it and free the request entries.
 		 */
 		if (!hctx->nr_ctx) {
-			if (set->tags[i]) {
+			/* Never unmap queue 0.  We need it as a
+			 * fallback in case of a new remap fails
+			 * allocation. */
+			if (i && set->tags[i]) {
 				blk_mq_free_rq_map(set, set->tags[i], i);
 				set->tags[i] = NULL;
 			}
@@ -1909,11 +1921,8 @@ static void blk_mq_map_swqueue(struct request_queue *q,
 			continue;
 		}
 
-		/* unmapped hw queue can be remapped after CPU topo changed */
-		if (!set->tags[i])
-			set->tags[i] = blk_mq_init_rq_map(set, i);
 		hctx->tags = set->tags[i];
-		WARN_ON(!hctx->tags);
+		BUG_ON(!hctx->tags);
 
 		/*
 		 * Set the map size to the number of mapped software queues.
-- 
2.7.4

[PATCH 2/2] blk-mq: Avoid memory reclaim when remapping queues

[Gabriel Krisman Bertazi]

While stressing memory and IO at the same time we changed SMT settings,
we were able to consistently trigger deadlocks in the mm system, which
froze the entire machine.

I think that under memory stress conditions, the large allocations
performed by blk_mq_init_rq_map may trigger a reclaim, which stalls
waiting on the block layer remmaping completion, thus deadlocking the
system.  The trace below was collected after the machine stalled,
waiting for the hotplug event completion.

The simplest fix for this is to make allocations in this path
non-reclaimable, with GFP_NOWAIT.  With this patch, We couldn't hit the
issue anymore.

This should apply on top of Jen's for-next branch cleanly.

 Call Trace:
[c000000f0160aaf0] [c000000f0160ab50] 0xc000000f0160ab50 (unreliable)
[c000000f0160acc0] [c000000000016624] __switch_to+0x2e4/0x430
[c000000f0160ad20] [c000000000b1a880] __schedule+0x310/0x9b0
[c000000f0160ae00] [c000000000b1af68] schedule+0x48/0xc0
[c000000f0160ae30] [c000000000b1b4b0] schedule_preempt_disabled+0x20/0x30
[c000000f0160ae50] [c000000000b1d4fc] __mutex_lock_slowpath+0xec/0x1f0
[c000000f0160aed0] [c000000000b1d678] mutex_lock+0x78/0xa0
[c000000f0160af00] [d000000019413cac] xfs_reclaim_inodes_ag+0x33c/0x380 [xfs]
[c000000f0160b0b0] [d000000019415164] xfs_reclaim_inodes_nr+0x54/0x70 [xfs]
[c000000f0160b0f0] [d0000000194297f8] xfs_fs_free_cached_objects+0x38/0x60 [xfs]
[c000000f0160b120] [c0000000003172c8] super_cache_scan+0x1f8/0x210
[c000000f0160b190] [c00000000026301c] shrink_slab.part.13+0x21c/0x4c0
[c000000f0160b2d0] [c000000000268088] shrink_zone+0x2d8/0x3c0
[c000000f0160b380] [c00000000026834c] do_try_to_free_pages+0x1dc/0x520
[c000000f0160b450] [c00000000026876c] try_to_free_pages+0xdc/0x250
[c000000f0160b4e0] [c000000000251978] __alloc_pages_nodemask+0x868/0x10d0
[c000000f0160b6f0] [c000000000567030] blk_mq_init_rq_map+0x160/0x380
[c000000f0160b7a0] [c00000000056758c] blk_mq_map_swqueue+0x33c/0x360
[c000000f0160b820] [c000000000567904] blk_mq_queue_reinit+0x64/0xb0
[c000000f0160b850] [c00000000056a16c] blk_mq_queue_reinit_notify+0x19c/0x250
[c000000f0160b8a0] [c0000000000f5d38] notifier_call_chain+0x98/0x100
[c000000f0160b8f0] [c0000000000c5fb0] __cpu_notify+0x70/0xe0
[c000000f0160b930] [c0000000000c63c4] notify_prepare+0x44/0xb0
[c000000f0160b9b0] [c0000000000c52f4] cpuhp_invoke_callback+0x84/0x250
[c000000f0160ba10] [c0000000000c570c] cpuhp_up_callbacks+0x5c/0x120
[c000000f0160ba60] [c0000000000c7cb8] _cpu_up+0xf8/0x1d0
[c000000f0160bac0] [c0000000000c7eb0] do_cpu_up+0x120/0x150
[c000000f0160bb40] [c0000000006fe024] cpu_subsys_online+0x64/0xe0
[c000000f0160bb90] [c0000000006f5124] device_online+0xb4/0x120
[c000000f0160bbd0] [c0000000006f5244] online_store+0xb4/0xc0
[c000000f0160bc20] [c0000000006f0a68] dev_attr_store+0x68/0xa0
[c000000f0160bc60] [c0000000003ccc30] sysfs_kf_write+0x80/0xb0
[c000000f0160bca0] [c0000000003cbabc] kernfs_fop_write+0x17c/0x250
[c000000f0160bcf0] [c00000000030fe6c] __vfs_write+0x6c/0x1e0
[c000000f0160bd90] [c000000000311490] vfs_write+0xd0/0x270
[c000000f0160bde0] [c0000000003131fc] SyS_write+0x6c/0x110
[c000000f0160be30] [c000000000009204] system_call+0x38/0xec

Signed-off-by: Gabriel Krisman Bertazi <krisman@linux.vnet.ibm.com>
Cc: Brian King <brking@linux.vnet.ibm.com>
Cc: Douglas Miller <dougmill@linux.vnet.ibm.com>
Cc: linux-block@vger.kernel.org
Cc: linux-scsi@vger.kernel.org
---
 block/blk-mq.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 7f7c4ba91adf..3e44303646cb 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1597,7 +1597,7 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
 	INIT_LIST_HEAD(&tags->page_list);
 
 	tags->rqs = kzalloc_node(set->queue_depth * sizeof(struct request *),
-				 GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY,
+				 GFP_NOWAIT | __GFP_NOWARN | __GFP_NORETRY,
 				 set->numa_node);
 	if (!tags->rqs) {
 		blk_mq_free_tags(tags);
@@ -1623,7 +1623,7 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
 
 		do {
 			page = alloc_pages_node(set->numa_node,
-				GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO,
+				GFP_NOWAIT | __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO,
 				this_order);
 			if (page)
 				break;
@@ -1644,7 +1644,7 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
 		 * Allow kmemleak to scan these pages as they contain pointers
 		 * to additional allocations like via ops->init_request().
 		 */
-		kmemleak_alloc(p, order_to_size(this_order), 1, GFP_KERNEL);
+		kmemleak_alloc(p, order_to_size(this_order), 1, GFP_NOWAIT);
 		entries_per_page = order_to_size(this_order) / rq_size;
 		to_do = min(entries_per_page, set->queue_depth - i);
 		left -= to_do * rq_size;
-- 
2.7.4

Re: [PATCH 2/2] blk-mq: Avoid memory reclaim when remapping queues

[Bart Van Assche]

On 11/14/2016 11:24 AM, Gabriel Krisman Bertazi wrote:
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -1597,7 +1597,7 @@ static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
>  	INIT_LIST_HEAD(&tags->page_list);
>
>  	tags->rqs = kzalloc_node(set->queue_depth * sizeof(struct request *),
> -				 GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY,
> +				 GFP_NOWAIT | __GFP_NOWARN | __GFP_NORETRY,
>  				 set->numa_node);
>  	if (!tags->rqs) {
>  		blk_mq_free_tags(tags);

Hello Gabriel,

I don't think that GFP_NOWAIT is acceptable in this context. Have you 
tried GFP_NOIO instead of GFP_NOWAIT?

Bart.

Re: [PATCH 2/2] blk-mq: Avoid memory reclaim when remapping queues

[Gabriel Krisman Bertazi]

Bart Van Assche <bart.vanassche@sandisk.com> writes:

> I don't think that GFP_NOWAIT is acceptable in this context. Have you
> tried GFP_NOIO instead of GFP_NOWAIT?

At first I used GFP_NOIO, but after reviewing gfp.h I convinced myself
GFP_NOWAIT was what I wanted because I was concerned about FS accesses
that aren't restricted in GFP_NOIO.  For some reason, I assumed this was
an issue.  I'm ok with the change and can submit a v2 shortly, after
more tests.  in fact, this will make the change more compliant with the
rest of block layer critical allocations that use GFP_NOIO.

Thanks,

-- 
Gabriel Krisman Bertazi