[syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40

Unfortunately, I don't have any reproducer for this issue yet.

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/9705ecb6a595/disk-929beafb.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/dbdd1f64b9b8/vmlinux-929beafb.xz
kernel image: https://storage.googleapis.com/syzbot-assets/3f70d07a929b/bzImage-929beafb.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com

FAT-fs (loop3): error, invalid access to FAT (entry 0x0000616b)
======================================================
WARNING: possible circular locking dependency detected
6.12.0-rc6-next-20241108-syzkaller #0 Not tainted
------------------------------------------------------
syz.3.2125/17744 is trying to acquire lock:
ffff8880691980b0 (&sbi->fat_lock){+.+.}-{4:4}, at: lock_fat fs/fat/fatent.c:281 [inline]
ffff8880691980b0 (&sbi->fat_lock){+.+.}-{4:4}, at: fat_count_free_clusters+0x156/0xe70 fs/fat/fatent.c:724

but task is already holding lock:
ffff88802533deb0 (&q->limits_lock){+.+.}-{4:4}, at: queue_limits_start_update include/linux/blkdev.h:944 [inline]
ffff88802533deb0 (&q->limits_lock){+.+.}-{4:4}, at: loop_reconfigure_limits+0x287/0x9f0 drivers/block/loop.c:1003

which lock already depends on the new lock.


the existing dependency chain (in reverse order) is:

-> #2 (&q->limits_lock){+.+.}-{4:4}:
       lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5849
       __mutex_lock_common kernel/locking/mutex.c:585 [inline]
       __mutex_lock+0x1ac/0xee0 kernel/locking/mutex.c:735
       queue_limits_start_update include/linux/blkdev.h:944 [inline]
       loop_reconfigure_limits+0x287/0x9f0 drivers/block/loop.c:1003
       loop_set_block_size drivers/block/loop.c:1473 [inline]
       lo_simple_ioctl drivers/block/loop.c:1496 [inline]
       lo_ioctl+0x1351/0x1f50 drivers/block/loop.c:1559
       blkdev_ioctl+0x57d/0x6a0 block/ioctl.c:693
       vfs_ioctl fs/ioctl.c:51 [inline]
       __do_sys_ioctl fs/ioctl.c:906 [inline]
       __se_sys_ioctl+0xf5/0x170 fs/ioctl.c:892
       do_syscall_x64 arch/x86/entry/common.c:52 [inline]
       do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

-> #1 (&q->q_usage_counter(io)#17){++++}-{0:0}:
       lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5849
       bio_queue_enter block/blk.h:75 [inline]
       blk_mq_submit_bio+0x1510/0x2490 block/blk-mq.c:3095
       __submit_bio+0x2c2/0x560 block/blk-core.c:629
       __submit_bio_noacct_mq block/blk-core.c:710 [inline]
       submit_bio_noacct_nocheck+0x4d3/0xe30 block/blk-core.c:739
       submit_bh fs/buffer.c:2819 [inline]
       __bread_slow fs/buffer.c:1264 [inline]
       __bread_gfp+0x23c/0x400 fs/buffer.c:1488
       sb_bread include/linux/buffer_head.h:346 [inline]
       fat12_ent_bread+0x155/0x540 fs/fat/fatent.c:77
       fat_ent_read_block+0x3e4/0x530 fs/fat/fatent.c:445
       fat_alloc_clusters+0x4ee/0x11c0 fs/fat/fatent.c:493
       fat_add_cluster fs/fat/inode.c:107 [inline]
       __fat_get_block fs/fat/inode.c:154 [inline]
       fat_get_block+0x4c4/0xd00 fs/fat/inode.c:189
       __block_write_begin_int+0x50c/0x1a70 fs/buffer.c:2116
       block_write_begin fs/buffer.c:2226 [inline]
       cont_write_begin+0x6e2/0x9d0 fs/buffer.c:2577
       fat_write_begin+0x76/0x140 fs/fat/inode.c:228
       generic_perform_write+0x344/0x6d0 mm/filemap.c:4055
       generic_file_write_iter+0xae/0x310 mm/filemap.c:4182
       new_sync_write fs/read_write.c:586 [inline]
       vfs_write+0xaeb/0xd30 fs/read_write.c:679
       ksys_write+0x18f/0x2b0 fs/read_write.c:731
       do_syscall_x64 arch/x86/entry/common.c:52 [inline]
       do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

-> #0 (&sbi->fat_lock){+.+.}-{4:4}:
       check_prev_add kernel/locking/lockdep.c:3161 [inline]
       check_prevs_add kernel/locking/lockdep.c:3280 [inline]
       validate_chain+0x18ef/0x5920 kernel/locking/lockdep.c:3904
       __lock_acquire+0x1397/0x2100 kernel/locking/lockdep.c:5226
       lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5849
       __mutex_lock_common kernel/locking/mutex.c:585 [inline]
       __mutex_lock+0x1ac/0xee0 kernel/locking/mutex.c:735
       lock_fat fs/fat/fatent.c:281 [inline]
       fat_count_free_clusters+0x156/0xe70 fs/fat/fatent.c:724
       fat_statfs+0x139/0x450 fs/fat/inode.c:834
       statfs_by_dentry fs/statfs.c:66 [inline]
       vfs_statfs+0x13b/0x2c0 fs/statfs.c:90
       loop_config_discard drivers/block/loop.c:798 [inline]
       loop_reconfigure_limits+0x5fe/0x9f0 drivers/block/loop.c:1012
       loop_configure+0x77e/0xeb0 drivers/block/loop.c:1093
       lo_ioctl+0x846/0x1f50
       blkdev_ioctl+0x57d/0x6a0 block/ioctl.c:693
       vfs_ioctl fs/ioctl.c:51 [inline]
       __do_sys_ioctl fs/ioctl.c:906 [inline]
       __se_sys_ioctl+0xf5/0x170 fs/ioctl.c:892
       do_syscall_x64 arch/x86/entry/common.c:52 [inline]
       do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

other info that might help us debug this:

Chain exists of:
  &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&q->limits_lock);
                               lock(&q->q_usage_counter(io)#17);
                               lock(&q->limits_lock);
  lock(&sbi->fat_lock);

 *** DEADLOCK ***

2 locks held by syz.3.2125/17744:
 #0: ffff88802541fb60 (&lo->lo_mutex){+.+.}-{4:4}, at: loop_global_lock_killable drivers/block/loop.c:120 [inline]
 #0: ffff88802541fb60 (&lo->lo_mutex){+.+.}-{4:4}, at: loop_configure+0x1f7/0xeb0 drivers/block/loop.c:1044
 #1: ffff88802533deb0 (&q->limits_lock){+.+.}-{4:4}, at: queue_limits_start_update include/linux/blkdev.h:944 [inline]
 #1: ffff88802533deb0 (&q->limits_lock){+.+.}-{4:4}, at: loop_reconfigure_limits+0x287/0x9f0 drivers/block/loop.c:1003

stack backtrace:
CPU: 0 UID: 0 PID: 17744 Comm: syz.3.2125 Not tainted 6.12.0-rc6-next-20241108-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/30/2024
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
 print_circular_bug+0x13a/0x1b0 kernel/locking/lockdep.c:2074
 check_noncircular+0x36a/0x4a0 kernel/locking/lockdep.c:2206
 check_prev_add kernel/locking/lockdep.c:3161 [inline]
 check_prevs_add kernel/locking/lockdep.c:3280 [inline]
 validate_chain+0x18ef/0x5920 kernel/locking/lockdep.c:3904
 __lock_acquire+0x1397/0x2100 kernel/locking/lockdep.c:5226
 lock_acquire+0x1ed/0x550 kernel/locking/lockdep.c:5849
 __mutex_lock_common kernel/locking/mutex.c:585 [inline]
 __mutex_lock+0x1ac/0xee0 kernel/locking/mutex.c:735
 lock_fat fs/fat/fatent.c:281 [inline]
 fat_count_free_clusters+0x156/0xe70 fs/fat/fatent.c:724
 fat_statfs+0x139/0x450 fs/fat/inode.c:834
 statfs_by_dentry fs/statfs.c:66 [inline]
 vfs_statfs+0x13b/0x2c0 fs/statfs.c:90
 loop_config_discard drivers/block/loop.c:798 [inline]
 loop_reconfigure_limits+0x5fe/0x9f0 drivers/block/loop.c:1012
 loop_configure+0x77e/0xeb0 drivers/block/loop.c:1093
 lo_ioctl+0x846/0x1f50
 blkdev_ioctl+0x57d/0x6a0 block/ioctl.c:693
 vfs_ioctl fs/ioctl.c:51 [inline]
 __do_sys_ioctl fs/ioctl.c:906 [inline]
 __se_sys_ioctl+0xf5/0x170 fs/ioctl.c:892
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f9752d7e719
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f9753abd038 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00007f9752f36058 RCX: 00007f9752d7e719
RDX: 00000000200002c0 RSI: 0000000000004c0a RDI: 0000000000000008
RBP: 00007f9752df139e R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000000 R14: 00007f9752f36058 R15: 00007ffe36e679e8
 </TASK>


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzkaller@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

If the report is already addressed, let syzbot know by replying with:
#syz fix: exact-commit-title

If you want to overwrite report's subsystems, reply with:
#syz set subsystems: new-subsystem
(See the list of subsystem names on the web dashboard)

If the report is a duplicate of another one, reply with:
#syz dup: exact-subject-of-another-report

If you want to undo deduplication, reply with:
#syz undup

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[OGAWA Hirofumi]

Hi,

syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:

> syzbot found the following issue on:
>
> HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
> git tree:       linux-next
> console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
> dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
> compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40

This patch is to fix the above race. Please check this.

Thanks


From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Subject: [PATCH] loop: Fix ABBA locking race
Date: Mon, 11 Nov 2024 21:53:36 +0900

Current loop calls vfs_statfs() while holding the q->limits_lock. If
FS takes some locking in vfs_statfs callback, this may lead to ABBA
locking bug (at least, FAT fs has this issue actually).

So this patch calls vfs_statfs() outside q->limits_locks instead,
because looks like there is no reason to hold q->limits_locks while
getting discard configs.

Chain exists of:
  &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&q->limits_lock);
                               lock(&q->q_usage_counter(io)#17);
                               lock(&q->limits_lock);
  lock(&sbi->fat_lock);

 *** DEADLOCK ***

Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
---
 drivers/block/loop.c |   31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 78a7bb2..5f3ce51 100644
--- a/drivers/block/loop.c	2024-09-16 13:45:20.253220178 +0900
+++ b/drivers/block/loop.c	2024-11-11 21:51:00.910135443 +0900
@@ -770,12 +770,11 @@ static void loop_sysfs_exit(struct loop_
 				   &loop_attribute_group);
 }
 
-static void loop_config_discard(struct loop_device *lo,
-		struct queue_limits *lim)
+static void loop_get_discard_config(struct loop_device *lo,
+				    u32 *granularity, u32 *max_discard_sectors)
 {
 	struct file *file = lo->lo_backing_file;
 	struct inode *inode = file->f_mapping->host;
-	u32 granularity = 0, max_discard_sectors = 0;
 	struct kstatfs sbuf;
 
 	/*
@@ -788,8 +787,9 @@ static void loop_config_discard(struct l
 	if (S_ISBLK(inode->i_mode)) {
 		struct request_queue *backingq = bdev_get_queue(I_BDEV(inode));
 
-		max_discard_sectors = backingq->limits.max_write_zeroes_sectors;
-		granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
+		*max_discard_sectors =
+			backingq->limits.max_write_zeroes_sectors;
+		*granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
 			queue_physical_block_size(backingq);
 
 	/*
@@ -797,16 +797,9 @@ static void loop_config_discard(struct l
 	 * image a.k.a. discard.
 	 */
 	} else if (file->f_op->fallocate && !vfs_statfs(&file->f_path, &sbuf)) {
-		max_discard_sectors = UINT_MAX >> 9;
-		granularity = sbuf.f_bsize;
+		*max_discard_sectors = UINT_MAX >> 9;
+		*granularity = sbuf.f_bsize;
 	}
-
-	lim->max_hw_discard_sectors = max_discard_sectors;
-	lim->max_write_zeroes_sectors = max_discard_sectors;
-	if (max_discard_sectors)
-		lim->discard_granularity = granularity;
-	else
-		lim->discard_granularity = 0;
 }
 
 struct loop_worker {
@@ -992,6 +985,7 @@ static int loop_reconfigure_limits(struc
 	struct inode *inode = file->f_mapping->host;
 	struct block_device *backing_bdev = NULL;
 	struct queue_limits lim;
+	u32 granularity = 0, max_discard_sectors = 0;
 
 	if (S_ISBLK(inode->i_mode))
 		backing_bdev = I_BDEV(inode);
@@ -1001,6 +995,8 @@ static int loop_reconfigure_limits(struc
 	if (!bsize)
 		bsize = loop_default_blocksize(lo, backing_bdev);
 
+	loop_get_discard_config(lo, &granularity, &max_discard_sectors);
+
 	lim = queue_limits_start_update(lo->lo_queue);
 	lim.logical_block_size = bsize;
 	lim.physical_block_size = bsize;
@@ -1010,7 +1006,12 @@ static int loop_reconfigure_limits(struc
 		lim.features |= BLK_FEAT_WRITE_CACHE;
 	if (backing_bdev && !bdev_nonrot(backing_bdev))
 		lim.features |= BLK_FEAT_ROTATIONAL;
-	loop_config_discard(lo, &lim);
+	lim.max_hw_discard_sectors = max_discard_sectors;
+	lim.max_write_zeroes_sectors = max_discard_sectors;
+	if (max_discard_sectors)
+		lim.discard_granularity = granularity;
+	else
+		lim.discard_granularity = 0;
 	return queue_limits_commit_update(lo->lo_queue, &lim);
 }
 
_
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

[PATCH] loop: Fix ABBA locking race (Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters)

[OGAWA Hirofumi]

OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> writes:

ping?

> Hi,
>
> syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:
>
>> syzbot found the following issue on:
>>
>> HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
>> git tree:       linux-next
>> console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
>> kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
>> dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
>> compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
>
> This patch is to fix the above race. Please check this.
>
> Thanks
>
>
> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> Subject: [PATCH] loop: Fix ABBA locking race
> Date: Mon, 11 Nov 2024 21:53:36 +0900
>
> Current loop calls vfs_statfs() while holding the q->limits_lock. If
> FS takes some locking in vfs_statfs callback, this may lead to ABBA
> locking bug (at least, FAT fs has this issue actually).
>
> So this patch calls vfs_statfs() outside q->limits_locks instead,
> because looks like there is no reason to hold q->limits_locks while
> getting discard configs.
>
> Chain exists of:
>   &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock
>
>  Possible unsafe locking scenario:
>
>        CPU0                    CPU1
>        ----                    ----
>   lock(&q->limits_lock);
>                                lock(&q->q_usage_counter(io)#17);
>                                lock(&q->limits_lock);
>   lock(&sbi->fat_lock);
>
>  *** DEADLOCK ***
>
> Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
> Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> ---
>  drivers/block/loop.c |   31 ++++++++++++++++---------------
>  1 file changed, 16 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/block/loop.c b/drivers/block/loop.c
> index 78a7bb2..5f3ce51 100644
> --- a/drivers/block/loop.c	2024-09-16 13:45:20.253220178 +0900
> +++ b/drivers/block/loop.c	2024-11-11 21:51:00.910135443 +0900
> @@ -770,12 +770,11 @@ static void loop_sysfs_exit(struct loop_
>  				   &loop_attribute_group);
>  }
>  
> -static void loop_config_discard(struct loop_device *lo,
> -		struct queue_limits *lim)
> +static void loop_get_discard_config(struct loop_device *lo,
> +				    u32 *granularity, u32 *max_discard_sectors)
>  {
>  	struct file *file = lo->lo_backing_file;
>  	struct inode *inode = file->f_mapping->host;
> -	u32 granularity = 0, max_discard_sectors = 0;
>  	struct kstatfs sbuf;
>  
>  	/*
> @@ -788,8 +787,9 @@ static void loop_config_discard(struct l
>  	if (S_ISBLK(inode->i_mode)) {
>  		struct request_queue *backingq = bdev_get_queue(I_BDEV(inode));
>  
> -		max_discard_sectors = backingq->limits.max_write_zeroes_sectors;
> -		granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
> +		*max_discard_sectors =
> +			backingq->limits.max_write_zeroes_sectors;
> +		*granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
>  			queue_physical_block_size(backingq);
>  
>  	/*
> @@ -797,16 +797,9 @@ static void loop_config_discard(struct l
>  	 * image a.k.a. discard.
>  	 */
>  	} else if (file->f_op->fallocate && !vfs_statfs(&file->f_path, &sbuf)) {
> -		max_discard_sectors = UINT_MAX >> 9;
> -		granularity = sbuf.f_bsize;
> +		*max_discard_sectors = UINT_MAX >> 9;
> +		*granularity = sbuf.f_bsize;
>  	}
> -
> -	lim->max_hw_discard_sectors = max_discard_sectors;
> -	lim->max_write_zeroes_sectors = max_discard_sectors;
> -	if (max_discard_sectors)
> -		lim->discard_granularity = granularity;
> -	else
> -		lim->discard_granularity = 0;
>  }
>  
>  struct loop_worker {
> @@ -992,6 +985,7 @@ static int loop_reconfigure_limits(struc
>  	struct inode *inode = file->f_mapping->host;
>  	struct block_device *backing_bdev = NULL;
>  	struct queue_limits lim;
> +	u32 granularity = 0, max_discard_sectors = 0;
>  
>  	if (S_ISBLK(inode->i_mode))
>  		backing_bdev = I_BDEV(inode);
> @@ -1001,6 +995,8 @@ static int loop_reconfigure_limits(struc
>  	if (!bsize)
>  		bsize = loop_default_blocksize(lo, backing_bdev);
>  
> +	loop_get_discard_config(lo, &granularity, &max_discard_sectors);
> +
>  	lim = queue_limits_start_update(lo->lo_queue);
>  	lim.logical_block_size = bsize;
>  	lim.physical_block_size = bsize;
> @@ -1010,7 +1006,12 @@ static int loop_reconfigure_limits(struc
>  		lim.features |= BLK_FEAT_WRITE_CACHE;
>  	if (backing_bdev && !bdev_nonrot(backing_bdev))
>  		lim.features |= BLK_FEAT_ROTATIONAL;
> -	loop_config_discard(lo, &lim);
> +	lim.max_hw_discard_sectors = max_discard_sectors;
> +	lim.max_write_zeroes_sectors = max_discard_sectors;
> +	if (max_discard_sectors)
> +		lim.discard_granularity = granularity;
> +	else
> +		lim.discard_granularity = 0;
>  	return queue_limits_commit_update(lo->lo_queue, &lim);
>  }
>  
> _

-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[Ming Lei]

On Tue, Nov 12, 2024 at 12:44 AM OGAWA Hirofumi
<hirofumi@mail.parknet.co.jp> wrote:
>
> Hi,
>
> syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:
>
> > syzbot found the following issue on:
> >
> > HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
> > git tree:       linux-next
> > console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
> > dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
> > compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
>
> This patch is to fix the above race. Please check this.
>
> Thanks
>
>
> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> Subject: [PATCH] loop: Fix ABBA locking race
> Date: Mon, 11 Nov 2024 21:53:36 +0900
>
> Current loop calls vfs_statfs() while holding the q->limits_lock. If
> FS takes some locking in vfs_statfs callback, this may lead to ABBA
> locking bug (at least, FAT fs has this issue actually).
>
> So this patch calls vfs_statfs() outside q->limits_locks instead,
> because looks like there is no reason to hold q->limits_locks while
> getting discard configs.
>
> Chain exists of:
>   &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock
>
>  Possible unsafe locking scenario:
>
>        CPU0                    CPU1
>        ----                    ----
>   lock(&q->limits_lock);
>                                lock(&q->q_usage_counter(io)#17);
>                                lock(&q->limits_lock);
>   lock(&sbi->fat_lock);
>
>  *** DEADLOCK ***
>
> Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
> Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> ---
>  drivers/block/loop.c |   31 ++++++++++++++++---------------
>  1 file changed, 16 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/block/loop.c b/drivers/block/loop.c
> index 78a7bb2..5f3ce51 100644
> --- a/drivers/block/loop.c      2024-09-16 13:45:20.253220178 +0900
> +++ b/drivers/block/loop.c      2024-11-11 21:51:00.910135443 +0900
> @@ -770,12 +770,11 @@ static void loop_sysfs_exit(struct loop_
>                                    &loop_attribute_group);
>  }
>
> -static void loop_config_discard(struct loop_device *lo,
> -               struct queue_limits *lim)
> +static void loop_get_discard_config(struct loop_device *lo,
> +                                   u32 *granularity, u32 *max_discard_sectors)
>  {
>         struct file *file = lo->lo_backing_file;
>         struct inode *inode = file->f_mapping->host;
> -       u32 granularity = 0, max_discard_sectors = 0;
>         struct kstatfs sbuf;
>
>         /*
> @@ -788,8 +787,9 @@ static void loop_config_discard(struct l
>         if (S_ISBLK(inode->i_mode)) {
>                 struct request_queue *backingq = bdev_get_queue(I_BDEV(inode));
>
> -               max_discard_sectors = backingq->limits.max_write_zeroes_sectors;
> -               granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
> +               *max_discard_sectors =
> +                       backingq->limits.max_write_zeroes_sectors;
> +               *granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
>                         queue_physical_block_size(backingq);
>
>         /*
> @@ -797,16 +797,9 @@ static void loop_config_discard(struct l
>          * image a.k.a. discard.
>          */
>         } else if (file->f_op->fallocate && !vfs_statfs(&file->f_path, &sbuf)) {
> -               max_discard_sectors = UINT_MAX >> 9;
> -               granularity = sbuf.f_bsize;
> +               *max_discard_sectors = UINT_MAX >> 9;
> +               *granularity = sbuf.f_bsize;
>         }
> -
> -       lim->max_hw_discard_sectors = max_discard_sectors;
> -       lim->max_write_zeroes_sectors = max_discard_sectors;
> -       if (max_discard_sectors)
> -               lim->discard_granularity = granularity;
> -       else
> -               lim->discard_granularity = 0;
>  }
>
>  struct loop_worker {
> @@ -992,6 +985,7 @@ static int loop_reconfigure_limits(struc
>         struct inode *inode = file->f_mapping->host;
>         struct block_device *backing_bdev = NULL;
>         struct queue_limits lim;
> +       u32 granularity = 0, max_discard_sectors = 0;
>
>         if (S_ISBLK(inode->i_mode))
>                 backing_bdev = I_BDEV(inode);
> @@ -1001,6 +995,8 @@ static int loop_reconfigure_limits(struc
>         if (!bsize)
>                 bsize = loop_default_blocksize(lo, backing_bdev);
>
> +       loop_get_discard_config(lo, &granularity, &max_discard_sectors);
> +
>         lim = queue_limits_start_update(lo->lo_queue);
>         lim.logical_block_size = bsize;
>         lim.physical_block_size = bsize;
> @@ -1010,7 +1006,12 @@ static int loop_reconfigure_limits(struc
>                 lim.features |= BLK_FEAT_WRITE_CACHE;
>         if (backing_bdev && !bdev_nonrot(backing_bdev))
>                 lim.features |= BLK_FEAT_ROTATIONAL;
> -       loop_config_discard(lo, &lim);
> +       lim.max_hw_discard_sectors = max_discard_sectors;
> +       lim.max_write_zeroes_sectors = max_discard_sectors;
> +       if (max_discard_sectors)
> +               lim.discard_granularity = granularity;
> +       else
> +               lim.discard_granularity = 0;
>         return queue_limits_commit_update(lo->lo_queue, &lim);
>  }

Looks fine,

Reviewed-by: Ming Lei <ming.lei@redhat.com>

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[Jens Axboe]

On 11/19/24 5:10 AM, Ming Lei wrote:
> On Tue, Nov 12, 2024 at 12:44 AM OGAWA Hirofumi
> <hirofumi@mail.parknet.co.jp> wrote:
>>
>> Hi,
>>
>> syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:
>>
>>> syzbot found the following issue on:
>>>
>>> HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
>>> git tree:       linux-next
>>> console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
>>> dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
>>> compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
>>
>> This patch is to fix the above race. Please check this.
>>
>> Thanks
>>
>>
>> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>> Subject: [PATCH] loop: Fix ABBA locking race
>> Date: Mon, 11 Nov 2024 21:53:36 +0900
>>
>> Current loop calls vfs_statfs() while holding the q->limits_lock. If
>> FS takes some locking in vfs_statfs callback, this may lead to ABBA
>> locking bug (at least, FAT fs has this issue actually).
>>
>> So this patch calls vfs_statfs() outside q->limits_locks instead,
>> because looks like there is no reason to hold q->limits_locks while
>> getting discard configs.
>>
>> Chain exists of:
>>   &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock
>>
>>  Possible unsafe locking scenario:
>>
>>        CPU0                    CPU1
>>        ----                    ----
>>   lock(&q->limits_lock);
>>                                lock(&q->q_usage_counter(io)#17);
>>                                lock(&q->limits_lock);
>>   lock(&sbi->fat_lock);
>>
>>  *** DEADLOCK ***
>>
>> Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
>> Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>> ---
>>  drivers/block/loop.c |   31 ++++++++++++++++---------------
>>  1 file changed, 16 insertions(+), 15 deletions(-)
>>
>> diff --git a/drivers/block/loop.c b/drivers/block/loop.c
>> index 78a7bb2..5f3ce51 100644
>> --- a/drivers/block/loop.c      2024-09-16 13:45:20.253220178 +0900
>> +++ b/drivers/block/loop.c      2024-11-11 21:51:00.910135443 +0900
>> @@ -770,12 +770,11 @@ static void loop_sysfs_exit(struct loop_
>>                                    &loop_attribute_group);
>>  }
>>
>> -static void loop_config_discard(struct loop_device *lo,
>> -               struct queue_limits *lim)
>> +static void loop_get_discard_config(struct loop_device *lo,
>> +                                   u32 *granularity, u32 *max_discard_sectors)
>>  {
>>         struct file *file = lo->lo_backing_file;
>>         struct inode *inode = file->f_mapping->host;
>> -       u32 granularity = 0, max_discard_sectors = 0;
>>         struct kstatfs sbuf;
>>
>>         /*
>> @@ -788,8 +787,9 @@ static void loop_config_discard(struct l
>>         if (S_ISBLK(inode->i_mode)) {
>>                 struct request_queue *backingq = bdev_get_queue(I_BDEV(inode));
>>
>> -               max_discard_sectors = backingq->limits.max_write_zeroes_sectors;
>> -               granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
>> +               *max_discard_sectors =
>> +                       backingq->limits.max_write_zeroes_sectors;
>> +               *granularity = bdev_discard_granularity(I_BDEV(inode)) ?:
>>                         queue_physical_block_size(backingq);
>>
>>         /*
>> @@ -797,16 +797,9 @@ static void loop_config_discard(struct l
>>          * image a.k.a. discard.
>>          */
>>         } else if (file->f_op->fallocate && !vfs_statfs(&file->f_path, &sbuf)) {
>> -               max_discard_sectors = UINT_MAX >> 9;
>> -               granularity = sbuf.f_bsize;
>> +               *max_discard_sectors = UINT_MAX >> 9;
>> +               *granularity = sbuf.f_bsize;
>>         }
>> -
>> -       lim->max_hw_discard_sectors = max_discard_sectors;
>> -       lim->max_write_zeroes_sectors = max_discard_sectors;
>> -       if (max_discard_sectors)
>> -               lim->discard_granularity = granularity;
>> -       else
>> -               lim->discard_granularity = 0;
>>  }
>>
>>  struct loop_worker {
>> @@ -992,6 +985,7 @@ static int loop_reconfigure_limits(struc
>>         struct inode *inode = file->f_mapping->host;
>>         struct block_device *backing_bdev = NULL;
>>         struct queue_limits lim;
>> +       u32 granularity = 0, max_discard_sectors = 0;
>>
>>         if (S_ISBLK(inode->i_mode))
>>                 backing_bdev = I_BDEV(inode);
>> @@ -1001,6 +995,8 @@ static int loop_reconfigure_limits(struc
>>         if (!bsize)
>>                 bsize = loop_default_blocksize(lo, backing_bdev);
>>
>> +       loop_get_discard_config(lo, &granularity, &max_discard_sectors);
>> +
>>         lim = queue_limits_start_update(lo->lo_queue);
>>         lim.logical_block_size = bsize;
>>         lim.physical_block_size = bsize;
>> @@ -1010,7 +1006,12 @@ static int loop_reconfigure_limits(struc
>>                 lim.features |= BLK_FEAT_WRITE_CACHE;
>>         if (backing_bdev && !bdev_nonrot(backing_bdev))
>>                 lim.features |= BLK_FEAT_ROTATIONAL;
>> -       loop_config_discard(lo, &lim);
>> +       lim.max_hw_discard_sectors = max_discard_sectors;
>> +       lim.max_write_zeroes_sectors = max_discard_sectors;
>> +       if (max_discard_sectors)
>> +               lim.discard_granularity = granularity;
>> +       else
>> +               lim.discard_granularity = 0;
>>         return queue_limits_commit_update(lo->lo_queue, &lim);
>>  }
> 
> Looks fine,
> 
> Reviewed-by: Ming Lei <ming.lei@redhat.com>

The patch doesn't apply to the for-6.13/block tree, Ogawa can you send
an updated one please?

-- 
Jens Axboe

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[OGAWA Hirofumi]

Jens Axboe <axboe@kernel.dk> writes:

> On 11/19/24 5:10 AM, Ming Lei wrote:
>> On Tue, Nov 12, 2024 at 12:44 AM OGAWA Hirofumi
>> <hirofumi@mail.parknet.co.jp> wrote:
>>>
>>> Hi,
>>>
>>> syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:
>>>
>>>> syzbot found the following issue on:
>>>>
>>>> HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
>>>> git tree:       linux-next
>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
>>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
>>>> dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
>>>> compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40

[...]

>> 
>> Looks fine,
>> 
>> Reviewed-by: Ming Lei <ming.lei@redhat.com>
>
> The patch doesn't apply to the for-6.13/block tree, Ogawa can you send
> an updated one please?

Updated the patch for linux-block:for-6.13/block. Please apply.

Thanks.


From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Subject: [PATCH] loop: Fix ABBA locking race
Date: Tue, 19 Nov 2024 23:42:23 +0900

Current loop calls vfs_statfs() while holding the q->limits_lock. If
FS takes some locking in vfs_statfs callback, this may lead to ABBA
locking bug (at least, FAT fs has this issue actually).

So this patch calls vfs_statfs() outside q->limits_locks instead,
because looks like no reason to hold q->limits_locks while getting
discord configs.

Chain exists of:
  &sbi->fat_lock --> &q->q_usage_counter(io)#17 --> &q->limits_lock

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&q->limits_lock);
                               lock(&q->q_usage_counter(io)#17);
                               lock(&q->limits_lock);
  lock(&sbi->fat_lock);

 *** DEADLOCK ***

Reported-by: syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
---
 drivers/block/loop.c |   30 +++++++++++++++---------------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index fe9bb4f..8f6761c 100644
--- a/drivers/block/loop.c	2024-11-19 23:37:54.760751014 +0900
+++ b/drivers/block/loop.c	2024-11-19 23:38:55.645461107 +0900
@@ -770,12 +770,11 @@ static void loop_sysfs_exit(struct loop_
 				   &loop_attribute_group);
 }
 
-static void loop_config_discard(struct loop_device *lo,
-		struct queue_limits *lim)
+static void loop_get_discard_config(struct loop_device *lo,
+				    u32 *granularity, u32 *max_discard_sectors)
 {
 	struct file *file = lo->lo_backing_file;
 	struct inode *inode = file->f_mapping->host;
-	u32 granularity = 0, max_discard_sectors = 0;
 	struct kstatfs sbuf;
 
 	/*
@@ -788,24 +787,17 @@ static void loop_config_discard(struct l
 	if (S_ISBLK(inode->i_mode)) {
 		struct block_device *bdev = I_BDEV(inode);
 
-		max_discard_sectors = bdev_write_zeroes_sectors(bdev);
-		granularity = bdev_discard_granularity(bdev);
+		*max_discard_sectors = bdev_write_zeroes_sectors(bdev);
+		*granularity = bdev_discard_granularity(bdev);
 
 	/*
 	 * We use punch hole to reclaim the free space used by the
 	 * image a.k.a. discard.
 	 */
 	} else if (file->f_op->fallocate && !vfs_statfs(&file->f_path, &sbuf)) {
-		max_discard_sectors = UINT_MAX >> 9;
-		granularity = sbuf.f_bsize;
+		*max_discard_sectors = UINT_MAX >> 9;
+		*granularity = sbuf.f_bsize;
 	}
-
-	lim->max_hw_discard_sectors = max_discard_sectors;
-	lim->max_write_zeroes_sectors = max_discard_sectors;
-	if (max_discard_sectors)
-		lim->discard_granularity = granularity;
-	else
-		lim->discard_granularity = 0;
 }
 
 struct loop_worker {
@@ -991,6 +983,7 @@ static int loop_reconfigure_limits(struc
 	struct inode *inode = file->f_mapping->host;
 	struct block_device *backing_bdev = NULL;
 	struct queue_limits lim;
+	u32 granularity = 0, max_discard_sectors = 0;
 
 	if (S_ISBLK(inode->i_mode))
 		backing_bdev = I_BDEV(inode);
@@ -1000,6 +993,8 @@ static int loop_reconfigure_limits(struc
 	if (!bsize)
 		bsize = loop_default_blocksize(lo, backing_bdev);
 
+	loop_get_discard_config(lo, &granularity, &max_discard_sectors);
+
 	lim = queue_limits_start_update(lo->lo_queue);
 	lim.logical_block_size = bsize;
 	lim.physical_block_size = bsize;
@@ -1009,7 +1004,12 @@ static int loop_reconfigure_limits(struc
 		lim.features |= BLK_FEAT_WRITE_CACHE;
 	if (backing_bdev && !bdev_nonrot(backing_bdev))
 		lim.features |= BLK_FEAT_ROTATIONAL;
-	loop_config_discard(lo, &lim);
+	lim.max_hw_discard_sectors = max_discard_sectors;
+	lim.max_write_zeroes_sectors = max_discard_sectors;
+	if (max_discard_sectors)
+		lim.discard_granularity = granularity;
+	else
+		lim.discard_granularity = 0;
 	return queue_limits_commit_update(lo->lo_queue, &lim);
 }
 
_
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[Jens Axboe]

On 11/19/24 7:46 AM, OGAWA Hirofumi wrote:
> Jens Axboe <axboe@kernel.dk> writes:
> 
>> On 11/19/24 5:10 AM, Ming Lei wrote:
>>> On Tue, Nov 12, 2024 at 12:44?AM OGAWA Hirofumi
>>> <hirofumi@mail.parknet.co.jp> wrote:
>>>>
>>>> Hi,
>>>>
>>>> syzbot <syzbot+a5d8c609c02f508672cc@syzkaller.appspotmail.com> writes:
>>>>
>>>>> syzbot found the following issue on:
>>>>>
>>>>> HEAD commit:    929beafbe7ac Add linux-next specific files for 20241108
>>>>> git tree:       linux-next
>>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=1621bd87980000
>>>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=75175323f2078363
>>>>> dashboard link: https://syzkaller.appspot.com/bug?extid=a5d8c609c02f508672cc
>>>>> compiler:       Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
> 
> [...]
> 
>>>
>>> Looks fine,
>>>
>>> Reviewed-by: Ming Lei <ming.lei@redhat.com>
>>
>> The patch doesn't apply to the for-6.13/block tree, Ogawa can you send
>> an updated one please?
> 
> Updated the patch for linux-block:for-6.13/block. Please apply.

Applied, thanks.

FWIW, your outgoing mailer is mangling patches. I fixed it up manually,
but probably something you want to get sorted. Download the raw one from
lore and you can see what I mean.

-- 
Jens Axboe

Re: [syzbot] [exfat?] possible deadlock in fat_count_free_clusters

[OGAWA Hirofumi]

Jens Axboe <axboe@kernel.dk> writes:

> FWIW, your outgoing mailer is mangling patches. I fixed it up manually,
> but probably something you want to get sorted. Download the raw one from
> lore and you can see what I mean.

Looks like at Ming Lei's reply, unicode "NARROW NO-BREAK SPACE" was
included in ">>>> On Tue, Nov 12, 2024 at 12:44?AM OGAWA Hirofumi" line?
So my mailer may be encoded as utf-8, not raw.

I'll take more care next time if possible. However, this mistake (utf-8
whitespace) may hard to prevent without machinery check somehow.

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>