Skip to contentSkip to conversation

btrfs: fix BUG_ON condition in btrfs_cancel_balance

PATCH In review xiaoshoukui 2 participants 3 messages Aug 15, 2023 – Aug 18, 2023 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btrfs: fix BUG_ON condition in btrfs_cancel_balance

Pausing and canceling balance can race to intterupt balance lead to BUG_ON panic in btrfs_cancel_balance. The BUG_ON condition in btrfs_cancel_balance does not take this race scenario into account.

However, the race condition has no other side effects. We can fix that.

Reproducing it with panic trace like this:
kernel BUG at fs/btrfs/volumes.c:4618!

RIP: 0010:btrfs_cancel_balance+0x5cf/0x6a0
Call Trace:
 <TASK>
 ? do_nanosleep+0x60/0x120
 ? hrtimer_nanosleep+0xb7/0x1a0
 ? sched_core_clone_cookie+0x70/0x70
 btrfs_ioctl_balance_ctl+0x55/0x70
 btrfs_ioctl+0xa46/0xd20
 __x64_sys_ioctl+0x7d/0xa0
 do_syscall_64+0x38/0x80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

Race scenario as follows:

Quoted 11 lines

mutex_unlock(&fs_info->balance_mutex);
--------------------
.......issue pause and cancel req in another thread
--------------------
ret = __btrfs_balance(fs_info);

mutex_lock(&fs_info->balance_mutex);

if (ret == -ECANCELED && atomic_read(&fs_info->balance_pause_req)) {
        btrfs_info(fs_info, "balance: paused");
        btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
}
---
 fs/btrfs/volumes.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
fs/btrfs/volumes.c +1 -2
index 2ecb76cf3d91..886d667419ed 100644
@@ -4638,8 +4638,7 @@ int btrfs_cancel_balance(struct btrfs_fs_info *fs_info)
 		}
 	}
 
-	BUG_ON(fs_info->balance_ctl ||
-		test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags));
+	BUG_ON(test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags));
 	atomic_dec(&fs_info->balance_cancel_req);
 	mutex_unlock(&fs_info->balance_mutex);
 	return 0;
Signature · 1 line
2.34.1
#1 lore raw
Quoted 3 lines · xiaoshoukui

Pausing and canceling balance can race to intterupt balance lead to BUG_ON panic in btrfs_cancel_balance. The BUG_ON condition in btrfs_cancel_balance does not take this race scenario into account.

Seems that it's from times the balance was not cancellable the same way as now. Also it's a good time to switch the BUG_ON to an assertion or handle it properly.

Quoted 40 lines

However, the race condition has no other side effects. We can fix that.

Reproducing it with panic trace like this:
kernel BUG at fs/btrfs/volumes.c:4618!

RIP: 0010:btrfs_cancel_balance+0x5cf/0x6a0
Call Trace:
 <TASK>
 ? do_nanosleep+0x60/0x120
 ? hrtimer_nanosleep+0xb7/0x1a0
 ? sched_core_clone_cookie+0x70/0x70
 btrfs_ioctl_balance_ctl+0x55/0x70
 btrfs_ioctl+0xa46/0xd20
 __x64_sys_ioctl+0x7d/0xa0
 do_syscall_64+0x38/0x80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

Race scenario as follows:

Quoted 11 lines

mutex_unlock(&fs_info->balance_mutex);
--------------------
.......issue pause and cancel req in another thread
--------------------
ret = __btrfs_balance(fs_info);

mutex_lock(&fs_info->balance_mutex);

if (ret == -ECANCELED && atomic_read(&fs_info->balance_pause_req)) {
        btrfs_info(fs_info, "balance: paused");
        btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
}
---
 fs/btrfs/volumes.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
fs/btrfs/volumes.c +0 -0
index 2ecb76cf3d91..886d667419ed 100644

fs/btrfs/volumes.c int btrfs_cancel_balance(struct btrfs_fs_info *fs_info) · line 4641

Quoted 6 lines
fs/btrfs/volumes.c +1 -2
@@ -4638,8 +4638,7 @@ int btrfs_cancel_balance(struct btrfs_fs_info *fs_info)
 		}
 	}
 
-	BUG_ON(fs_info->balance_ctl ||
-		test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags));
+	BUG_ON(test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags));

I'll change to to ASSERT, this is really to verify that the state tracking works properly.

Quoted 5 lines
fs/btrfs/volumes.c +0 -0
 	atomic_dec(&fs_info->balance_cancel_req);
 	mutex_unlock(&fs_info->balance_mutex);
 	return 0;
Signature · 1 line
2.34.1
#2 lore raw
Quoted 3 lines

Seems that it's from times the balance was not cancellable the same way as now. Also it's a good time to switch the BUG_ON to an assertion or handle it properly.

That's the point. Canceling the balance only takes into account the normal scenarios. Replacing the BUG ON here with an assertion would make the code cleaner.

Quoted 2 lines

I'll change to to ASSERT, this is really to verify that the state tracking works properly.

The ASSERT and BUG ON macros have already helped us uncover many hidden issues.