大家好，我遇到了一个奇怪的BTRFS相关问题，想请各位帮忙分析下：

我的服务器上，当使用内核版本≤4.9时，BTRFS文件系统一切正常；但只要升级到≥4.19的内核版本，整个系统就会直接崩溃，内核日志里会出现大量文件系统损坏相关的报错：

[Fri Oct 27 23:44:22 2023] BTRFS info (device dm-1): using crc32c (crc32c-intel) checksum algorithm

[Fri Oct 27 23:44:22 2023] BTRFS info (device dm-1): disk space caching is enabled

[Fri Oct 27 23:44:22 2023] BTRFS info (device dm-1): bdev /dev/mapper/vg0-root errs: wr 0, rd 4, flush 0, corrupt 0, gen 0

[Fri Oct 27 23:44:23 2023] BTRFS info (device dm-1): checking UUID tree

[Fri Oct 27 23:44:25 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=7800258560 slot=0 ino=22544639, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:25 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 7800258560 mirror 1

[Fri Oct 27 23:44:26 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=86444384256 slot=1 ino=8651007, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:26 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 86444384256 mirror 1

[Fri Oct 27 23:44:26 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=86032809984 slot=41 ino=24641791, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:26 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 86032809984 mirror 1

[Fri Oct 27 23:44:26 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=8081629184 slot=0 ino=8651599, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:26 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 8081629184 mirror 1

[Fri Oct 27 23:44:26 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=86032809984 slot=41 ino=24641791, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:26 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 86032809984 mirror 1

[Fri Oct 27 23:44:26 2023] BTRFS critical (device dm-1): corrupt leaf: root=5 block=7800258560 slot=0 ino=22544639, invalid inode transid: has 140710402240512 expect [0, 9501327]

[Fri Oct 27 23:44:26 2023] BTRFS error (device dm-1): read time tree block corruption detected on logical 7800258560 mirror 1

同时，使用内核≥5版本下的btrfs check工具检查该分区时，也会报出大量错误：

root@rescue ~ # btrfs check /dev/mapper/vg0-root

Opening filesystem to check...

Checking filesystem on /dev/mapper/vg0-root

UUID: 18c05fd2-0569-420b-bb74-676b2327d3d5

[1/7] checking root items

[2/7] checking extents

[3/7] checking free space cache

[4/7] checking fs roots

root 5 inode 8651007 errors 100000, invalid inode generation or transid

root 5 inode 8651008 errors 100000, invalid inode generation or transid

root 5 inode 8651599 errors 100000, invalid inode generation or transid

root 5 inode 8651600 errors 100000, invalid inode generation or transid

root 5 inode 22544639 errors 100000, invalid inode generation or transid

root 5 inode 24641791 errors 100000, invalid inode generation or transid

root 5 inode 40419460 errors 200, dir isize wrong

root 5 inode 44021300 errors 200, dir isize wrong

ERROR: errors found in fs roots

found 187456020480 bytes used, error(s) found

total csum bytes: 179106780

total tree bytes: 2667675648

total fs tree bytes: 2107199488

total extent tree bytes: 305291264

btree space waste bytes: 643077002

file data blocks allocated: 496942690304

referenced 175125786624

挂载分区后，目录内容也会出现大量异常：

root@rescue ~ # ll /mnt/

ls: cannot access '/mnt/proc': Input/output error

ls: cannot access '/mnt/run': Input/output error

ls: cannot access '/mnt/sys': Input/output error

ls: cannot access '/mnt/etc': Input/output error

ls: cannot access '/mnt/lib64': Input/output error

ls: cannot access '/mnt/usr': Input/output error

total 32K

drwxr-xr-x 1 root root 2.0K Oct 27 20:18 bin

drwxr-xr-x 1 root root    0 Sep 27  2014 boot

drwxr-xr-x 1 root root    0 Sep 27  2014 dev

d????????? ? ?    ?       ?            ? etc

drwxr-xr-x 1 root root   12 Dec 23  2014 home

lrwxrwxrwx 1 root root   31 Oct 26 11:04 initrd.img -> boot/initrd.img-5.10.0-26-amd64

lrwxrwxrwx 1 root root   30 Oct 26 17:24 initrd.img.old -> boot/initrd.img-4.9.0-13-amd64

-rw-r----- 1 root root  659 Sep 27  2014 installimage.conf

-rw-r----- 1 root root 9.8K Sep 27  2014 installimage.debug

drwxr-xr-x 1 root root  420 Oct 26 17:22 lib

d????????? ? ?    ?       ?            ? lib64

drwx------ 1 root root    0 Sep 12  2012 lost+found

drwxr-xr-x 1 root root   22 Sep 12  2012 media

drwxr-xr-x 1 root root    0 Apr  2  2019 mnt

drwxr-xr-x 1 root root   24 Dec 27  2019 opt

d????????? ? ?    ?       ?            ? proc

drwx------ 1 root root  560 Oct 27 22:47 root

d????????? ? ?    ?       ?            ? run

drwxr-xr-x 1 root root 3.1K Oct 27 20:18 sbin

drwxr-xr-x 1 root root    0 Sep 12  2012 srv

d????????? ? ?    ?       ?            ? sys

drwxrwxrwt 1 root root  106 Oct 27 23:35 tmp

d????????? ? ?    ?       ?            ? usr

drwxr-xr-x 1 root root  118 Sep 28  2014 var

lrwxrwxrwx 1 root root   28 Oct 26 11:04 vmlinuz -> boot/vmlinuz-5.10.0-26-amd64

lrwxrwxrwx 1 root root   27 Oct 26 17:24 vmlinuz.old -> boot/vmlinuz-4.9.0-13-amd64

但神奇的是，只要我切回4.9内核，所有这些报错都会消失，文件的读写操作完全正常——我反复验证过多次，文件可以正常读取、写入、重读和重写，没有任何问题。

另外，我还有其他使用BTRFS的服务器，升级到新内核后都没有出现类似问题，所以这应该不是普遍存在的兼容性问题。

希望有大佬能帮忙分析下原因，或者给出排查方向，非常感谢！

备注：内容来源于stack exchange，提问作者gmelis