[LU-16335] "lfs rm_entry" failed to remove broken directories

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0, Lustre 2.15.3
• Severity: 3
• Assignee: Lai Siyao
• Reporter: Lai Siyao

Description

In LU-16159, update logs are canceled upon recovery abort, which may leave some directories broken, and can't be removed by "lfs rm_entry". This is because "lfs rm_entry" failed some sanity check, but it will leave end-user a broken filesystem without any way to fix.

Issue Links

• is related to: LU-16159 remove update llog files after recovery abort
• is related to: LU-16398 ost-pools: FAIL: remove sub-test dirs failed
• is related to: LU-16336 LFSCK should fix inconsistencies caused by recovery abort

Comments

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/49329/
Subject: LU-16335 mdt: skip target check for rm_entry
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: ae98c5fdaaf37daeb328b7110cbcf42754752c9d

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/49335/
Subject: LU-16335 test: add fail_abort_cleanup()
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: d5fe41a02a6ed57bcbfc4a4c695bb509c9c7c313

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/49328/
Subject: LU-16335 build: remove _GNU_SOURCE dependency in lustre_user.h
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: efc5c8d4de60d394344506f7cfb188eaf04a4bac

[Lai Siyao]

In theory lfsck should fix these inconsistencies, but it's not fully tested, and LU-16159 test result shows some are not fixed, and it will be addressed in LU-16336.

[James A Simmons]

So we disable various test for the native client since lfs rm_dentry was removed upstream. Its considered a security risk. Does this mean lfsck can repair the file system?

[Gerrit Updater]

"Lai Siyao <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/49335
Subject: LU-16335 test: add fail_abort_cleanup()
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 243ffa00d0a4353c78ee8b73bdce0e1c60bdc5da

[Gerrit Updater]

"Lai Siyao <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/49334
Subject: LU-16335 header: implement native strchrnul()
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: bc03636059c5c2dc9a0f1030af83309aedc94ac0

[Gerrit Updater]

"Lai Siyao <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/49329
Subject: LU-16335 mdt: skip child check for rm_entry
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 2cc5c2bfed10f2d63ee767cf90d77fb16e08e366

[Gerrit Updater]

"Lai Siyao <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/49328
Subject: LU-16335 build: strchrnul() needs define _GNU_SOURCE
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f5f31fa51477c7091a5af7fbb95382cd87d97938

[Lai Siyao]

config.log show this:

configure:39672: result: no
configure:39682: checking if ioctl IOC_REMOVE_ENTRY' is supported
configure:39696: gcc -c -g -O2 -Werror -I/home/laisiyao/lustre/lnet/include/uapi -I/home/laisiyao/lustre/lustre/include/uapi -I/home/laisiyao/lustre/libcfs/include -I/home/laisiyao/lustre/lnet/utils/ -I/home/laisiyao/lustre/lustre/include  conftest.c >&5
In file included from /home/laisiyao/lustre/lustre/include/uapi/linux/lustre/lustre_idl.h:74,
                 from /home/laisiyao/lustre/lustre/include/uapi/linux/lustre/lustre_ioctl.h:34,
                 from conftest.c:253:
/home/laisiyao/lustre/lustre/include/uapi/linux/lustre/lustre_user.h: In function 'changelog_rec_sname':
/home/laisiyao/lustre/lustre/include/uapi/linux/lustre/lustre_user.h:2012:9: error: implicit declaration of function 'strchrnul'; did you mean 'strchr'? [-Werror=implicit-function-declaration]
  return strchrnul(changelog_rec_name(rec), '\0') + 1;
         ^~~~~~~~~
         strchr
/home/laisiyao/lustre/lustre/include/uapi/linux/lustre/lustre_user.h:2012:50: error: returning 'int' from a function with return type 'char *' makes pointer from integer without a cast [-Werror=int-conversion]
  return strchrnul(changelog_rec_name(rec), '\0') + 1;
         ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
cc1: all warnings being treated as errors
configure:39696: $? = 1
configure: failed program was:
 

This check is from https://review.whamcloud.com/39207, and it causes "lfs rm_entry" always returns -ENOTSUP, which should be the cause of broken directory unlink fail.

strchrnul() needs define _GNU_SOURCE, adding it in the check can solve the issue.