[LU-13523] zero link directory causes crash on 2.12 Created: 06/May/20  Updated: 06/May/20

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.12.3, Lustre 2.12.4
Fix Version/s: None

Type: Bug Priority: Minor
Reporter: Jesse Stroik Assignee: WC Triage
Resolution: Unresolved Votes: 0
Labels: None
Environment:

centos 7 / centos 8


Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

I have an older file system of roughly 400 million files that i'm about to move to new hardware. The FS is running lustre 2.8.0 on centos 6.

While traversing the file system, I found there exist some directories apparently with zero references. Here is an example such directory:

drwxr-s--- 0 <user> <group> 12288 Sep 11 2016 goes3.1980.286.1915

When attempting to list this directory on lustre 2.12.3 or 2.12.4 on centos 7/8 clients, it caused the system to crash. In one instance, I observed this LBUG:

 

 
kernel:LustreError: 19084:0:(mdc_locks.c:124:mdc_set_lock_data()) ASSERTION( old_inode->i_state & I_FREEING ) failed: Found existing inode ffff9de915cb7110/144115292816098530/0 state 0 in lock: setting data to ffff9de915cb6c90/144115292816098530/33554456Message from syslogd@s4-609-c1 at May 6 13:22:28 ...
kernel:LustreError: 19084:0:(mdc_locks.c:124:mdc_set_lock_data()) LBUG
This does not occur on lustre 2.10.6. I downgraded to 2.10.6 and now simply receive "no such file or directory" instead.

 



 Comments   
Comment by Jesse Stroik [ 06/May/20 ]

Looks like my formatting was lost.

I tested this on lustre-client-2.13.53_99 built from git and can confirm it has this issue also.

Generated at Sat Feb 10 03:02:00 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.