[PATCH AUTOSEL 5.2 115/123] block: aoe: Fix kernel crash due to atomic sleep when exiting

[PATCH AUTOSEL 5.2 115/123] block: aoe: Fix kernel crash due to atomic sleep when exiting

[Sasha Levin]

From: He Zhe <zhe.he@windriver.com>

[ Upstream commit 430380b4637aec646996b4aef67ad417593923b2 ]

Since commit 3582dd291788 ("aoe: convert aoeblk to blk-mq"), aoedev_downdev
has had the possibility of sleeping and causing the following crash.

BUG: scheduling while atomic: rmmod/2242/0x00000003
Modules linked in: aoe
Preemption disabled at:
[<ffffffffc01d95e5>] flush+0x95/0x4a0 [aoe]
CPU: 7 PID: 2242 Comm: rmmod Tainted: G          I       5.2.3 #1
Hardware name: Intel Corporation S5520HC/S5520HC, BIOS S5500.86B.01.10.0025.030220091519 03/02/2009
Call Trace:
 dump_stack+0x4f/0x6a
 ? flush+0x95/0x4a0 [aoe]
 __schedule_bug.cold+0x44/0x54
 __schedule+0x44f/0x680
 schedule+0x44/0xd0
 blk_mq_freeze_queue_wait+0x46/0xb0
 ? wait_woken+0x80/0x80
 blk_mq_freeze_queue+0x1b/0x20
 aoedev_downdev+0x111/0x160 [aoe]
 flush+0xff/0x4a0 [aoe]
 aoedev_exit+0x23/0x30 [aoe]
 aoe_exit+0x35/0x948 [aoe]
 __se_sys_delete_module+0x183/0x210
 __x64_sys_delete_module+0x16/0x20
 do_syscall_64+0x4d/0x130
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7f24e0043b07
Code: 73 01 c3 48 8b 0d 89 73 0b 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f
1f 84 00 00 00 00 00 0f 1f 44 00 00 b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff
ff 73 01 c3 48 8b 0d 59 73 0b 00 f7 d8 64 89 01 48
RSP: 002b:00007ffe18f7f1e8 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f24e0043b07
RDX: 000000000000000a RSI: 0000000000000800 RDI: 0000555c3ecf87c8
RBP: 00007ffe18f7f1f0 R08: 0000000000000000 R09: 0000000000000000
R10: 00007f24e00b4ac0 R11: 0000000000000206 R12: 00007ffe18f7f238
R13: 00007ffe18f7f410 R14: 00007ffe18f80e73 R15: 0000555c3ecf8760

This patch, handling in the same way of pass two, unlocks the locks and
restart pass one after aoedev_downdev is done.

Fixes: 3582dd291788 ("aoe: convert aoeblk to blk-mq")
Signed-off-by: He Zhe <zhe.he@windriver.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 drivers/block/aoe/aoedev.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
index 5b49f1b33ebec..e2ea2356da061 100644
--- a/drivers/block/aoe/aoedev.c
+++ b/drivers/block/aoe/aoedev.c
@@ -323,10 +323,14 @@ flush(const char __user *str, size_t cnt, int exiting)
 	}
 
 	flush_scheduled_work();
-	/* pass one: without sleeping, do aoedev_downdev */
+	/* pass one: do aoedev_downdev, which might sleep */
+restart1:
 	spin_lock_irqsave(&devlist_lock, flags);
 	for (d = devlist; d; d = d->next) {
 		spin_lock(&d->lock);
+		if (d->flags & DEVFL_TKILL)
+			goto cont;
+
 		if (exiting) {
 			/* unconditionally take each device down */
 		} else if (specified) {
@@ -338,8 +342,11 @@ flush(const char __user *str, size_t cnt, int exiting)
 		|| d->ref)
 			goto cont;
 
+		spin_unlock(&d->lock);
+		spin_unlock_irqrestore(&devlist_lock, flags);
 		aoedev_downdev(d);
 		d->flags |= DEVFL_TKILL;
+		goto restart1;
 cont:
 		spin_unlock(&d->lock);
 	}
@@ -348,7 +355,7 @@ flush(const char __user *str, size_t cnt, int exiting)
 	/* pass two: call freedev, which might sleep,
 	 * for aoedevs marked with DEVFL_TKILL
 	 */
-restart:
+restart2:
 	spin_lock_irqsave(&devlist_lock, flags);
 	for (d = devlist; d; d = d->next) {
 		spin_lock(&d->lock);
@@ -357,7 +364,7 @@ flush(const char __user *str, size_t cnt, int exiting)
 			spin_unlock(&d->lock);
 			spin_unlock_irqrestore(&devlist_lock, flags);
 			freedev(d);
-			goto restart;
+			goto restart2;
 		}
 		spin_unlock(&d->lock);
 	}
-- 
2.20.1

[PATCH AUTOSEL 5.2 116/123] block, bfq: handle NULL return value by bfq_init_rq()

[Sasha Levin]

From: Paolo Valente <paolo.valente@linaro.org>

[ Upstream commit fd03177c33b287c6541f4048f1d67b7b45a1abc9 ]

As reported in [1], the call bfq_init_rq(rq) may return NULL in case
of OOM (in particular, if rq->elv.icq is NULL because memory
allocation failed in failed in ioc_create_icq()).

This commit handles this circumstance.

[1] https://lkml.org/lkml/2019/7/22/824

Cc: Hsin-Yi Wang <hsinyi@google.com>
Cc: Nicolas Boichat <drinkcat@chromium.org>
Cc: Doug Anderson <dianders@chromium.org>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Reported-by: Hsin-Yi Wang <hsinyi@google.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 block/bfq-iosched.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 404e776aa36d0..b528710364e9e 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -2085,9 +2085,14 @@ static void bfq_request_merged(struct request_queue *q, struct request *req,
 	    blk_rq_pos(container_of(rb_prev(&req->rb_node),
 				    struct request, rb_node))) {
 		struct bfq_queue *bfqq = bfq_init_rq(req);
-		struct bfq_data *bfqd = bfqq->bfqd;
+		struct bfq_data *bfqd;
 		struct request *prev, *next_rq;
 
+		if (!bfqq)
+			return;
+
+		bfqd = bfqq->bfqd;
+
 		/* Reposition request in its sort_list */
 		elv_rb_del(&bfqq->sort_list, req);
 		elv_rb_add(&bfqq->sort_list, req);
@@ -2134,6 +2139,9 @@ static void bfq_requests_merged(struct request_queue *q, struct request *rq,
 	struct bfq_queue *bfqq = bfq_init_rq(rq),
 		*next_bfqq = bfq_init_rq(next);
 
+	if (!bfqq)
+		return;
+
 	/*
 	 * If next and rq belong to the same bfq_queue and next is older
 	 * than rq, then reposition rq in the fifo (by substituting next
@@ -5061,12 +5069,12 @@ static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
 
 	spin_lock_irq(&bfqd->lock);
 	bfqq = bfq_init_rq(rq);
-	if (at_head || blk_rq_is_passthrough(rq)) {
+	if (!bfqq || at_head || blk_rq_is_passthrough(rq)) {
 		if (at_head)
 			list_add(&rq->queuelist, &bfqd->dispatch);
 		else
 			list_add_tail(&rq->queuelist, &bfqd->dispatch);
-	} else { /* bfqq is assumed to be non null here */
+	} else {
 		idle_timer_disabled = __bfq_insert_request(bfqd, rq);
 		/*
 		 * Update bfqq, because, if a queue merge has occurred
-- 
2.20.1