xfs-masters
[Top] [All Lists]

[xfs-masters] [Bug 776] New: XFS hangs in SMP x86_64 kernel after many

To: xfs-master@xxxxxxxxxxx
Subject: [xfs-masters] [Bug 776] New: XFS hangs in SMP x86_64 kernel after many file creates
From: bugzilla-daemon@xxxxxxxxxxx
Date: Thu, 14 Feb 2008 15:38:29 -0800
Reply-to: xfs-masters@xxxxxxxxxxx
Sender: xfs-masters-bounce@xxxxxxxxxxx
http://oss.sgi.com/bugzilla/show_bug.cgi?id=776

           Summary: XFS hangs in SMP x86_64 kernel after many file creates
           Product: Linux XFS
           Version: Current
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: critical
          Priority: P2
         Component: XFS kernel code
        AssignedTo: xfs-master@xxxxxxxxxxx
        ReportedBy: spam_sgi@xxxxxxxxxxxxxxxxx


I am running Linux 2.6.22.15 x86_64, when running XFS on software RAID-6 and 
LVM, on a multiprocessor system, and after writing many files in rapid 
succession to a freshly formatted XFS filesystem, the md block device will 
eventually go into a mode where all I/O requests block indefinitely (including 
RAID reconstruction), even though accesses to the other RAIDs on the same 
disks continue to work.  Any process that is writing to files on that 
filesystem goes into iowait forever, including pdflush and xfsbufd (and 
sometimes xfssyncd).  I have found no resolution other than to reboot the 
machine.  This is repeatable 100% of the time on a freshly formatted volume by 
running the following bash script:

for ((x=0;x<10;x++)); do
        mkdir $x
        for ((y=0;y<40000;y++)); do
                dd if=/dev/zero of=$x/$y bs=1024 count=1024 2>/dev/null
        done
done

It usually will freeze while x==0, but I have seen it get to x==1.  Even after 
repeated tests, I have never seen it get to x>1.

This does not happen if the filesystem is ext3.  This also does not happen 
when running the kernel with the "nosmp" option.  Linux 2.6.18.1 does not seem 
to be affected.  The problem occurs whether or not the RAID is undergoing 
reconstruction.  Based on what I see in the kernel changelog, I suspect this 
may have been introduced in 2.6.19, but I have only tested the two kernel 
versions mentioned (these kernels are built in-house from unpatched kernel.org 
sources).

I have not shown if this problem is associated with the use of software RAID 
or LVM.  I have not tried another RAID level.  I have not tried a 32-bit 
kernel.

My platform is a dual Opteron 2.4 GHz machine with 4 GB RAM, 16 500 GB SATA 
disks, two 3ware 8-port SATA controllers (JBOD), LVM, and software RAID-6.  I 
am running CentOS 4.  The filesystem is 6.4 TB.  I have shown that this 
problem happens on another machine of similar specifications.

-- 
Configure bugmail: http://oss.sgi.com/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.


<Prev in Thread] Current Thread [Next in Thread>
  • [xfs-masters] [Bug 776] New: XFS hangs in SMP x86_64 kernel after many file creates, bugzilla-daemon <=