From 1e047eaab3bb5564f25b41e9cd3a053009f4e789 Mon Sep 17 00:00:00 2001 From: Tetsuo Handa Date: Fri, 6 Apr 2018 10:03:17 +0900 Subject: block/loop: fix deadlock after loop_set_status syzbot is reporting deadlocks at __blkdev_get() [1]. ---------------------------------------- [ 92.493919] systemd-udevd D12696 525 1 0x00000000 [ 92.495891] Call Trace: [ 92.501560] schedule+0x23/0x80 [ 92.502923] schedule_preempt_disabled+0x5/0x10 [ 92.504645] __mutex_lock+0x416/0x9e0 [ 92.510760] __blkdev_get+0x73/0x4f0 [ 92.512220] blkdev_get+0x12e/0x390 [ 92.518151] do_dentry_open+0x1c3/0x2f0 [ 92.519815] path_openat+0x5d9/0xdc0 [ 92.521437] do_filp_open+0x7d/0xf0 [ 92.527365] do_sys_open+0x1b8/0x250 [ 92.528831] do_syscall_64+0x6e/0x270 [ 92.530341] entry_SYSCALL_64_after_hwframe+0x42/0xb7 [ 92.931922] 1 lock held by systemd-udevd/525: [ 92.933642] #0: 00000000a2849e25 (&bdev->bd_mutex){+.+.}, at: __blkdev_get+0x73/0x4f0 ---------------------------------------- The reason of deadlock turned out that wait_event_interruptible() in blk_queue_enter() got stuck with bdev->bd_mutex held at __blkdev_put() due to q->mq_freeze_depth == 1. ---------------------------------------- [ 92.787172] a.out S12584 634 633 0x80000002 [ 92.789120] Call Trace: [ 92.796693] schedule+0x23/0x80 [ 92.797994] blk_queue_enter+0x3cb/0x540 [ 92.803272] generic_make_request+0xf0/0x3d0 [ 92.807970] submit_bio+0x67/0x130 [ 92.810928] submit_bh_wbc+0x15e/0x190 [ 92.812461] __block_write_full_page+0x218/0x460 [ 92.815792] __writepage+0x11/0x50 [ 92.817209] write_cache_pages+0x1ae/0x3d0 [ 92.825585] generic_writepages+0x5a/0x90 [ 92.831865] do_writepages+0x43/0xd0 [ 92.836972] __filemap_fdatawrite_range+0xc1/0x100 [ 92.838788] filemap_write_and_wait+0x24/0x70 [ 92.840491] __blkdev_put+0x69/0x1e0 [ 92.841949] blkdev_close+0x16/0x20 [ 92.843418] __fput+0xda/0x1f0 [ 92.844740] task_work_run+0x87/0xb0 [ 92.846215] do_exit+0x2f5/0xba0 [ 92.850528] do_group_exit+0x34/0xb0 [ 92.852018] SyS_exit_group+0xb/0x10 [ 92.853449] do_syscall_64+0x6e/0x270 [ 92.854944] entry_SYSCALL_64_after_hwframe+0x42/0xb7 [ 92.943530] 1 lock held by a.out/634: [ 92.945105] #0: 00000000a2849e25 (&bdev->bd_mutex){+.+.}, at: __blkdev_put+0x3c/0x1e0 ---------------------------------------- The reason of q->mq_freeze_depth == 1 turned out that loop_set_status() forgot to call blk_mq_unfreeze_queue() at error paths for info->lo_encrypt_type != NULL case. ---------------------------------------- [ 37.509497] CPU: 2 PID: 634 Comm: a.out Tainted: G W 4.16.0+ #457 [ 37.513608] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017 [ 37.518832] RIP: 0010:blk_freeze_queue_start+0x17/0x40 [ 37.521778] RSP: 0018:ffffb0c2013e7c60 EFLAGS: 00010246 [ 37.524078] RAX: 0000000000000000 RBX: ffff8b07b1519798 RCX: 0000000000000000 [ 37.527015] RDX: 0000000000000002 RSI: ffffb0c2013e7cc0 RDI: ffff8b07b1519798 [ 37.529934] RBP: ffffb0c2013e7cc0 R08: 0000000000000008 R09: 47a189966239b898 [ 37.532684] R10: dad78b99b278552f R11: 9332dca72259d5ef R12: ffff8b07acd73678 [ 37.535452] R13: 0000000000004c04 R14: 0000000000000000 R15: ffff8b07b841e940 [ 37.538186] FS: 00007fede33b9740(0000) GS:ffff8b07b8e80000(0000) knlGS:0000000000000000 [ 37.541168] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 37.543590] CR2: 00000000206fdf18 CR3: 0000000130b30006 CR4: 00000000000606e0 [ 37.546410] Call Trace: [ 37.547902] blk_freeze_queue+0x9/0x30 [ 37.549968] loop_set_status+0x67/0x3c0 [loop] [ 37.549975] loop_set_status64+0x3b/0x70 [loop] [ 37.549986] lo_ioctl+0x223/0x810 [loop] [ 37.549995] blkdev_ioctl+0x572/0x980 [ 37.550003] block_ioctl+0x34/0x40 [ 37.550006] do_vfs_ioctl+0xa7/0x6d0 [ 37.550017] ksys_ioctl+0x6b/0x80 [ 37.573076] SyS_ioctl+0x5/0x10 [ 37.574831] do_syscall_64+0x6e/0x270 [ 37.576769] entry_SYSCALL_64_after_hwframe+0x42/0xb7 ---------------------------------------- [1] https://syzkaller.appspot.com/bug?id=cd662bc3f6022c0979d01a262c318fab2ee9b56f Signed-off-by: Tetsuo Handa Reported-by: syzbot Fixes: ecdd09597a572513 ("block/loop: fix race between I/O and set_status") Cc: Ming Lei Cc: Dmitry Vyukov Cc: stable Cc: Jens Axboe Signed-off-by: Jens Axboe --- drivers/block/loop.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 264abaaff662..e5fc020cceda 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -1103,11 +1103,15 @@ loop_set_status(struct loop_device *lo, const struct loop_info64 *info) if (info->lo_encrypt_type) { unsigned int type = info->lo_encrypt_type; - if (type >= MAX_LO_CRYPT) - return -EINVAL; + if (type >= MAX_LO_CRYPT) { + err = -EINVAL; + goto exit; + } xfer = xfer_funcs[type]; - if (xfer == NULL) - return -EINVAL; + if (xfer == NULL) { + err = -EINVAL; + goto exit; + } } else xfer = NULL; -- cgit From bdac616db9bbadb90b7d6a406144571015e138f7 Mon Sep 17 00:00:00 2001 From: Omar Sandoval Date: Fri, 6 Apr 2018 09:57:03 -0700 Subject: loop: fix LOOP_GET_STATUS lock imbalance Commit 2d1d4c1e591f made loop_get_status() drop lo_ctx_mutex before returning, but the loop_get_status_old(), loop_get_status64(), and loop_get_status_compat() wrappers don't call loop_get_status() if the passed argument is NULL. The callers expect that the lock is dropped, so make sure we drop it in that case, too. Reported-by: syzbot+31e8daa8b3fc129e75f2@syzkaller.appspotmail.com Fixes: 2d1d4c1e591f ("loop: don't call into filesystem while holding lo_ctl_mutex") Signed-off-by: Omar Sandoval Signed-off-by: Jens Axboe --- drivers/block/loop.c | 33 ++++++++++++++++++--------------- 1 file changed, 18 insertions(+), 15 deletions(-) (limited to 'drivers/block') diff --git a/drivers/block/loop.c b/drivers/block/loop.c index e5fc020cceda..c9d04497a415 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -1287,12 +1287,13 @@ static int loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) { struct loop_info info; struct loop_info64 info64; - int err = 0; + int err; - if (!arg) - err = -EINVAL; - if (!err) - err = loop_get_status(lo, &info64); + if (!arg) { + mutex_unlock(&lo->lo_ctl_mutex); + return -EINVAL; + } + err = loop_get_status(lo, &info64); if (!err) err = loop_info64_to_old(&info64, &info); if (!err && copy_to_user(arg, &info, sizeof(info))) @@ -1304,12 +1305,13 @@ loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) { static int loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) { struct loop_info64 info64; - int err = 0; + int err; - if (!arg) - err = -EINVAL; - if (!err) - err = loop_get_status(lo, &info64); + if (!arg) { + mutex_unlock(&lo->lo_ctl_mutex); + return -EINVAL; + } + err = loop_get_status(lo, &info64); if (!err && copy_to_user(arg, &info64, sizeof(info64))) err = -EFAULT; @@ -1533,12 +1535,13 @@ loop_get_status_compat(struct loop_device *lo, struct compat_loop_info __user *arg) { struct loop_info64 info64; - int err = 0; + int err; - if (!arg) - err = -EINVAL; - if (!err) - err = loop_get_status(lo, &info64); + if (!arg) { + mutex_unlock(&lo->lo_ctl_mutex); + return -EINVAL; + } + err = loop_get_status(lo, &info64); if (!err) err = loop_info64_to_compat(&info64, arg); return err; -- cgit