Bug 118963

Summary: [PATCH] bug fix in statfs support
Product: Red Hat Enterprise Linux 3 Reporter: Yoav Zach <yoav.zach>
Component: ia32elAssignee: Jakub Jelinek <jakub>
Status: CLOSED ERRATA QA Contact:
Severity: medium Docs Contact:
Priority: medium    
Version: 3.0CC: keve.a.gabbert, tao
Target Milestone: ---   
Target Release: ---   
Hardware: ia64   
OS: Linux   
Fixed In Version: 1.0-3 Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2004-10-02 18:36:52 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---
Description Flags
patch to fix the statfs problem
test case for reproducing the problem with statfs
libiai32toia64.so binary
Updated patch against build 4588 none

Description Yoav Zach 2004-03-23 09:32:25 UTC
Description of problem:
in the support for statfs 32 bit syscalls, the EL does not copy the 
f_spare fields back to the user.

Version-Release number of selected component (if applicable):

How reproducible:
Please see the attached test case.

Steps to Reproduce:
1. build and run the attached test case
2. check the output 
Actual results:
f_spare[0] = 20000001

Expected results:
f_spare[0] = 0

Additional info:
The problem is in libia32toia64.so. Attached please find a patch to 
fix this issue, and a new binary that was built with this patch.

Comment 1 Yoav Zach 2004-03-23 09:34:20 UTC
Created attachment 98773 [details]
patch to fix the statfs problem

this is a patch to the sources of libia32toia64.so

Comment 2 Yoav Zach 2004-03-23 09:36:34 UTC
Created attachment 98774 [details]
test case for reproducing the problem with statfs

please build and run this test with an existing file-system as argument. e.g.
-"./fstatfs /home".

Comment 3 Yoav Zach 2004-03-23 09:38:30 UTC
Created attachment 98775 [details]
libiai32toia64.so binary

this is a new version of libia32toia64.so, includes the patch that fixes the
statfs problem.

Comment 4 Jakub Jelinek 2004-03-25 22:06:10 UTC
I see your patch is changing build number for 4592 to 4593, but
our src.rpm has build 4588.  Are the 4588 .. 4592 differences

Comment 5 Tony Luck 2004-03-26 19:21:08 UTC
Comment from Israel team:

We recommend to take the latest version 4593 â which includes also 
the Veritas bug fix which is a high priority for them. We are  
applying full validation cycle to this version same as we did to the 
previous release candidate â and its stability is better. Full list 
of changes will be supplied early next week. 

It is important that this updated version of IA32-EL make it onto the 
3/31 beta release of update2.

Comment 6 Jakub Jelinek 2004-03-26 19:27:32 UTC
Are the 4593 sources publically available (if so where) or can you attach them here (if not)?

Comment 7 Arun Sharma 2004-03-27 00:46:15 UTC
Created attachment 98895 [details]
Updated patch against build 4588

This one applies cleanly.

Comment 8 Larry Troan 2004-03-29 13:13:19 UTC
Event posted 03-26-2004 08:20pm by tonyluck with duration of 0.00  	
To be a bit more specific about "some customers" ... the issue here is
a file system that supports snapshot mounts, and returns data about
the snapshot mount in the f_spare[] fields of the statfs structure. 
Without this data, all the maintenance tools for snapshot mounts fail,
so this critical feature is useless.

Event posted 03-29-2004 02:33am by arjanv with duration of 0.00 	
which filesystem is this ?

Comment 9 Larry Troan 2004-03-29 13:16:21 UTC

Note that depending on answer to fs question above, may be bug or
feature request (to support "shapshot"). 

Comment 10 Larry Troan 2004-03-29 13:20:02 UTC
Also, believe Version in this Bug should be "RHEL3-update" and from
Summary line that Platform in Bug should be "all" to reflect x86 and
ia64. Is it also a problem on x86_64?

Comment 11 Larry Troan 2004-03-29 13:25:11 UTC
Sorry, platform is ia64 and bug relates to "emulation layer". 

Comment 13 Yoav Zach 2004-04-14 11:11:19 UTC
The version of libia32toia64.so remained 4588, as before the fix, 
while it should have changed to 4593. This is not right - it may lead 
to confusion with versions, which will make customer support more 
You can see the version of libia32toia64.so by running libia32x.so 
with '-v' switch.
The patch that Arun sent changes the Makefile of ia32toia64 so that 
the version changes. Apparently, the patch was not applied fully.