Bug 5077

Summary: unexpected "DT_GETPAGE: dtree page corrupt" errors
Product: File System Reporter: Max Alekseyev (relf)
Component: JFSAssignee: Dave Kleikamp (shaggy)
Status: RESOLVED PATCH_ALREADY_AVAILABLE    
Severity: high    
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.12 Subsystem:
Regression: --- Bisected commit-id:

Description Max Alekseyev 2005-08-16 16:19:13 UTC
Distribution: debian-amd64

I'm using JFS quite intensively and it happens that once in 2-3 weeks of uptime
the following error appears:
ERROR: (device hdb1): DT_GETPAGE: dtree page corrupt

Sometimes it is possible to fix the error "on-fly" with fsck.jfs (that found
some files cross-linked though), but sometimes attempt to use the volume fixed
with fsck.jfs causes more severe problem described in bug 4894 that forces
emergency reboot of the system.

Of course, there is no reason for the errors to appear. My system is doing just
regular file I/O operations, nothing harmful.
Comment 1 Dave Kleikamp 2005-08-17 12:16:33 UTC
Is this the same filesystem as in bug #5073?  If it is a case-insensitive
filesystem, I'm wondering if that bug may be the cause of the problems.  Having
bogus dentries in the dcache may possibly lead to some confusion.  I'd like to
know if I should follow that theory, or whether I could rule it out.  Of course,
if it happens again after applying the patch for 5073, I can rule it out as well.
Comment 2 Max Alekseyev 2005-08-18 15:13:18 UTC
Yes, it the same filesystem as in bug 5073. But I've caught bug 5073 just by
chance, so I could not say from my experience that the bug 5073 and the current
bug are related. 
Anyway I'm using a patched JFS driver now and we will see if it change anything.
Comment 3 Dave Kleikamp 2005-08-31 09:50:14 UTC
I'm going to assume this is fixed with bug 5073.  Please reopen it if you see it
again.