On 2021/7/22 22:05, Chao Yu wrote:
> On 2021/7/22 12:41, Jack Qiu wrote:
>> The root cause is shrink_dcache_sb after sbi has been freed.
>> So call shrink_dcache_sb before free sbi and other resources.
>>
>> ==================================================================
>> BUG: KASAN: use-after-free in f2fs_evict_inode+0x31c/0xde5
>> Read of size 4 at addr ffff8881d97f0d50 by task syz-executor.3/8729
>>
>> PU: 1 PID: 8729 Comm: syz-executor.3 Not tainted
>> 4.19.195-00002-g67dceea04431-dirty #31
> 
> Does this bug below to 4.19.195?
> 
Yes, I believe ed2e621a95d704e6a4e904cc00524e8cbddda0c2 causes this bug.

git describe --contains ed2e621a95d704e6a4e904cc00524e8cbddda0c2
v3.17-rc4~27^2~17

@@ -1126,6 +1130,13 @@ static int f2fs_fill_super(struct super_block *sb, void 
*data, int silent)
        brelse(raw_super_buf);
 free_sbi:
        kfree(sbi);    ---  free sbi
+
+       /* give only one another chance */
+       if (retry) {
+               retry = !retry;
+               shrink_dcache_sb(sb); --- call f2fs_evict_inode(root inode), it 
will access sbi
+               goto try_onemore;
+       }
        return err;
 }

>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
>> ?-20190727_073836-buildvm-ppc64le-16.ppc.fedoraproject.org-3.fc31
>> 04/01/2014
>> Call Trace:
>>   dump_stack+0xe5/0x14b
>>   ? f2fs_evict_inode+0x31c/0xde5
>>   print_address_description+0x6c/0x237
>>   ? f2fs_evict_inode+0x31c/0xde5
>>   kasan_report.cold+0x88/0x2a3
>>   f2fs_evict_inode+0x31c/0xde5
>>   ? full_proxy_open.cold+0x12/0x12
>>   evict+0x2cd/0x5f0
>>   iput+0x3d9/0x6f0
>>   dentry_unlink_inode+0x273/0x330
>>   __dentry_kill+0x340/0x5e0
>>   dentry_kill+0xb7/0x740
>>   shrink_dentry_list+0x256/0x660
>>   shrink_dcache_sb+0x11f/0x1d0
>>   ? shrink_dentry_list+0x660/0x660
>>   ? __kasan_slab_free+0x144/0x180
>>   f2fs_fill_super+0x2a34/0x4a80
>>   ? f2fs_sanity_check_ckpt.cold+0x2b6/0x2b6
>>   ? wait_for_completion+0x3c0/0x3c0
>>   ? set_blocksize+0x230/0x2b0
>>   mount_bdev+0x2c1/0x370
>>   ? f2fs_sanity_check_ckpt.cold+0x2b6/0x2b6
>>   mount_fs+0x4c/0x1c0
>>   vfs_kern_mount.part.0+0x60/0x3d0
>>   do_mount+0x367/0x2570
>>   ? kasan_unpoison_shadow+0x33/0x40
>>   ? copy_mount_string+0x40/0x40
>>   ? kmem_cache_alloc_trace+0x13f/0x2b0
>>   ? _copy_from_user+0x94/0x100
>>   ? copy_mount_options+0x1f1/0x2e0
>>   ksys_mount+0xa0/0x100
>>   __x64_sys_mount+0xbf/0x160
>>   do_syscall_64+0xc2/0x190
>>   entry_SYSCALL_64_after_hwframe+0x49/0xbe
>> RIP: 0033:0x47938e
>> Code: 48 c7 c1 b4 ff ff ff f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84
>> 00 00 00 00 00 90 f3 0f 1e fa 49 89 ca b8 a5 00 00 00 0f 05 <48> 3d 01
>> f0 ff ff 73 01 c3 48 c7 c1 b4 ff ff ff f7 d8 64 89 01 48
>> RSP: 002b:00007fed673b6a48 EFLAGS: 00000246 ORIG_RAX: 00000000000000a5
>> RAX: ffffffffffffffda RBX: 0000000020000200 RCX: 000000000047938e
>> RDX: 0000000020000000 RSI: 0000000020000100 RDI: 00007fed673b6aa0
>> RBP: 00007fed673b6ae0 R08: 00007fed673b6ae0 R09: 0000000020000000
>> R10: 0000000000000000 R11: 0000000000000246 R12: 0000000020000000
>> R13: 0000000020000100 R14: 00007fed673b6aa0 R15: 0000000020000b00
>>
>> Allocated by task 8729:
>>   kasan_kmalloc+0xc2/0xe0
>>   kmem_cache_alloc_trace+0x13f/0x2b0
>>   f2fs_fill_super+0x124/0x4a80
>>   mount_bdev+0x2c1/0x370
>>   mount_fs+0x4c/0x1c0
>>   vfs_kern_mount.part.0+0x60/0x3d0
>>   do_mount+0x367/0x2570
>>   ksys_mount+0xa0/0x100
>>   __x64_sys_mount+0xbf/0x160
>>   do_syscall_64+0xc2/0x190
>>   entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>
>> Freed by task 8729:
>>   __kasan_slab_free+0x12f/0x180
>>   kfree+0xfa/0x2a0
>>   f2fs_fill_super+0x2a00/0x4a80
>>   mount_bdev+0x2c1/0x370
>>   mount_fs+0x4c/0x1c0
>>   vfs_kern_mount.part.0+0x60/0x3d0
>>   do_mount+0x367/0x2570
>>   ksys_mount+0xa0/0x100
>>   __x64_sys_mount+0xbf/0x160
>>   do_syscall_64+0xc2/0x190
>>   entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>
>> Signed-off-by: Jack Qiu <[email protected]>
>> ---
>>   fs/f2fs/super.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
>> index 8fecd3050ccd..b041625e06ce 100644
>> --- a/fs/f2fs/super.c
>> +++ b/fs/f2fs/super.c
>> @@ -4229,6 +4229,8 @@ static int f2fs_fill_super(struct super_block *sb, 
>> void *data, int silent)
>>   free_compress_inode:
>>       f2fs_destroy_compress_inode(sbi);
>>   free_root_inode:
>> +    if (retry_cnt > 0 && skip_recovery)
>> +        shrink_dcache_sb(sb);
> 
> Compare to 4.19, last f2fs adds evict_inodes() before f2fs_unregister_sysfs(),
Can not find *evict_inodes*, could you please elaborate more detail?
> could you please check whether this can fix the issue?
I have run the test more than 1 day w/ this patch, it will fail in 1 hour 
before.
I believe this patch can fix *this* issue. But I'm not quite familiar with 
shrink_dcache_sb,
maybe I miss other scenario. If you have other comment, please let me know.

Thanks,
> 
> Thanks,
> 
>>       dput(sb->s_root);
>>       sb->s_root = NULL;
>>   free_node_inode:
>> @@ -4285,7 +4287,6 @@ static int f2fs_fill_super(struct super_block *sb, 
>> void *data, int silent)
>>       /* give only one another chance */
>>       if (retry_cnt > 0 && skip_recovery) {
>>           retry_cnt--;
>> -        shrink_dcache_sb(sb);
>>           goto try_onemore;
>>       }
>>       return err;
>> -- 
>> 2.17.1
>>
> .



_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Reply via email to