projects / platform / kernel / linux-rpi.git / commitdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | patch | inline | side by side (parent: fd86fd7)
ext4: Fix BUG_ON in ext4_bread when write quota data
authorYe Bin <yebin10@huawei.com>
Thu, 23 Dec 2021 01:55:06 +0000 (09:55 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 27 Jan 2022 10:05:17 +0000 (11:05 +0100)
commit 380a0091cab482489e9b19e07f2a166ad2b76d5c upstream.

We got issue as follows when run syzkaller:
[  167.936972] EXT4-fs error (device loop0): __ext4_remount:6314: comm rep: Abort forced by user
[  167.938306] EXT4-fs (loop0): Remounting filesystem read-only
[  167.981637] Assertion failure in ext4_getblk() at fs/ext4/inode.c:847: '(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY) || handle != NULL || create == 0'
[  167.983601] ------------[ cut here ]------------
[  167.984245] kernel BUG at fs/ext4/inode.c:847!
[  167.984882] invalid opcode: 0000 [#1] PREEMPT SMP KASAN PTI
[  167.985624] CPU: 7 PID: 2290 Comm: rep Tainted: G    B             5.16.0-rc5-next-20211217+ #123
[  167.986823] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20190727_073836-buildvm-ppc64le-16.ppc.fedoraproject.org-3.fc31 04/01/2014
[  167.988590] RIP: 0010:ext4_getblk+0x17e/0x504
[  167.989189] Code: c6 01 74 28 49 c7 c0 a0 a3 5c 9b b9 4f 03 00 00 48 c7 c2 80 9c 5c 9b 48 c7 c6 40 b6 5c 9b 48 c7 c7 20 a4 5c 9b e8 77 e3 fd ff <0f> 0b 8b 04 244
[  167.991679] RSP: 0018:ffff8881736f7398 EFLAGS: 00010282
[  167.992385] RAX: 0000000000000094 RBX: 1ffff1102e6dee75 RCX: 0000000000000000
[  167.993337] RDX: 0000000000000001 RSI: ffffffff9b6e29e0 RDI: ffffed102e6dee66
[  167.994292] RBP: ffff88816a076210 R08: 0000000000000094 R09: ffffed107363fa09
[  167.995252] R10: ffff88839b1fd047 R11: ffffed107363fa08 R12: ffff88816a0761e8
[  167.996205] R13: 0000000000000000 R14: 0000000000000021 R15: 0000000000000001
[  167.997158] FS:  00007f6a1428c740(0000) GS:ffff88839b000000(0000) knlGS:0000000000000000
[  167.998238] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  167.999025] CR2: 00007f6a140716c8 CR3: 0000000133216000 CR4: 00000000000006e0
[  167.999987] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  168.000944] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  168.001899] Call Trace:
[  168.002235]  <TASK>
[  168.007167]  ext4_bread+0xd/0x53
[  168.007612]  ext4_quota_write+0x20c/0x5c0
[  168.010457]  write_blk+0x100/0x220
[  168.010944]  remove_free_dqentry+0x1c6/0x440
[  168.011525]  free_dqentry.isra.0+0x565/0x830
[  168.012133]  remove_tree+0x318/0x6d0
[  168.014744]  remove_tree+0x1eb/0x6d0
[  168.017346]  remove_tree+0x1eb/0x6d0
[  168.019969]  remove_tree+0x1eb/0x6d0
[  168.022128]  qtree_release_dquot+0x291/0x340
[  168.023297]  v2_release_dquot+0xce/0x120
[  168.023847]  dquot_release+0x197/0x3e0
[  168.024358]  ext4_release_dquot+0x22a/0x2d0
[  168.024932]  dqput.part.0+0x1c9/0x900
[  168.025430]  __dquot_drop+0x120/0x190
[  168.025942]  ext4_clear_inode+0x86/0x220
[  168.026472]  ext4_evict_inode+0x9e8/0xa22
[  168.028200]  evict+0x29e/0x4f0
[  168.028625]  dispose_list+0x102/0x1f0
[  168.029148]  evict_inodes+0x2c1/0x3e0
[  168.030188]  generic_shutdown_super+0xa4/0x3b0
[  168.030817]  kill_block_super+0x95/0xd0
[  168.031360]  deactivate_locked_super+0x85/0xd0
[  168.031977]  cleanup_mnt+0x2bc/0x480
[  168.033062]  task_work_run+0xd1/0x170
[  168.033565]  do_exit+0xa4f/0x2b50
[  168.037155]  do_group_exit+0xef/0x2d0
[  168.037666]  __x64_sys_exit_group+0x3a/0x50
[  168.038237]  do_syscall_64+0x3b/0x90
[  168.038751]  entry_SYSCALL_64_after_hwframe+0x44/0xae

In order to reproduce this problem, the following conditions need to be met:
1. Ext4 filesystem with no journal;
2. Filesystem image with incorrect quota data;
3. Abort filesystem forced by user;
4. umount filesystem;

As in ext4_quota_write:
...
         if (EXT4_SB(sb)->s_journal && !handle) {
                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
                         " cancelled because transaction is not started",
                         (unsigned long long)off, (unsigned long long)len);
                 return -EIO;
         }
...
We only check handle if NULL when filesystem has journal. There is need
check handle if NULL even when filesystem has no journal.

Signed-off-by: Ye Bin <yebin10@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20211223015506.297766-1-yebin10@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/ext4/super.c patch | blob | history

diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 6261aad8294a6671e597213beb3eb5f40e40e5d0..6fb635ebffb92982d40b50d3705ae75cbe74036e 100644 (file)
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -6470,7 +6470,7 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type,
        struct buffer_head *bh;
        handle_t *handle = journal_current_handle();
 
-       if (EXT4_SB(sb)->s_journal && !handle) {
+       if (!handle) {
                ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
                        " cancelled because transaction is not started",
                        (unsigned long long)off, (unsigned long long)len);
Domain: System / Kernel;