[ ssic-linux-Bugs-1700876 ] CFS oops on XFS failover

Bugs item #1700876, was opened at 2007-04-15 00:42
Message generated for change (Settings changed) made by rogertsang
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405834&aid=1700876&group_id=32541

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Filesystem
Group: v1.9.2
>Status: Closed
Resolution: Fixed
Priority: 5
Private: No
Submitted By: Roger Tsang (rogertsang)
Assigned to: Roger Tsang (rogertsang)
Summary: CFS oops on XFS failover

Initial Comment:
/abc is a hard mount.

[/sbin/fsck.xfs (1) -- /abc] fsck.xfs /dev/drbd3 
XFS mounting filesystem drbd3
Starting XFS recovery on filesystem: drbd3 (dev: drbd3)
Ending XFS recovery on filesystem: drbd3 (dev: drbd3)
Unable to handle kernel NULL pointer dereference at virtual address 00000000
 printing eip:
00000000
*pde = 00000000
Oops: 0000 [#1]
SMP 
Modules linked in: cpufreq_ondemand tun ipt_REJECT ipt_state ipt_multiport iptable_filter ipt_MASQUERADE iptable_nat ip_conntrack ip_tables softdog xfs exportfs dm_mod uhci_hcd ehci_hcd usbcore drbd via_rhine sk98lin r8169 forcedeth
CPU:    0
EIP:    0060:[<00000000>]    Not tainted VLI
EFLAGS: 00010202   (2.6.11-ssi5.12) 
EIP is at 0x0
eax: fab2c3a0   ebx: d357c1f4   ecx: f6026a8c   edx: c24dd7ec
esi: f6026a00   edi: cccf5ce0   ebp: cccf5ca4   esp: cccf5c68
ds: 007b   es: 007b   ss: 0068
Process mount (pid: 71515, threadinfo=cccf4000 task=f6734330)
Stack: c0286757 d357c1f4 7837c809 00000000 00000010 c04e8900 00000000 000000d0 
       00000000 00000000 f6f9c000 cccf4000 f4c34300 00000001 00000001 cccf5d08 
       c028d832 cccf5ce0 cccf5ccc f4903654 f490365c c277b840 00000010 cccf5d14 
Call Trace:
 [<c010623f>] show_stack+0x7f/0xa0
 [<c01063e6>] show_registers+0x166/0x230
 [<c0106786>] die+0xf6/0x1c0
 [<c011b34d>] do_page_fault+0x45d/0x652
 [<c0105e9f>] error_code+0x2b/0x30
 [<c028d832>] cfstok_start_svrcfstok+0xf2/0x100
 [<c028d8a8>] cfstok_rebuild_tokens+0x68/0x140
 [<c0293dc4>] cfs_rebuild_tokens+0x64/0x140
 [<c029566f>] cfs_sb_start_haves+0xdf/0x110
 [<c029632d>] cfsd_rb_phase_haves_0+0x4d/0xa0
 [<c02903d4>] cfs_send_rb_phase_haves+0x104/0x120
 [<c0296aa9>] cfs_rb_phase_haves+0x29/0x30
 [<c02967e1>] cfs_rb_do_phase+0x231/0x2d0
 [<c029601c>] cfsd_rb_0+0x3c/0xf0
 [<c028ff42>] cfs_send_rb+0xf2/0x100
 [<c02949f9>] cfs_do_kern_remount+0xd9/0x14e
 [<c0294912>] cfs_do_remount+0x52/0x60
 [<c0294841>] cfs_remount+0xc1/0x140
 [<c020e3ae>] ssisys_cfs_remount+0x7e/0xb0
 [<c020bc0b>] do_ssisys+0x9b/0x1f0
 [<c020bdae>] sys_ssisys+0x4e/0x70
 [<c0105305>] sysenter_past_esp+0x52/0x75
Code:  Bad EIP value.
 
Entering kdb (current=0xf6734330, pid 71515) on processor 0 Oops: Oops
due to oops @ 0x0
eax = 0xfab2c3a0 ebx = 0xd357c1f4 ecx = 0xf6026a8c edx = 0xc24dd7ec 
esi = 0xf6026a00 edi = 0xcccf5ce0 esp = 0xcccf5c68 eip = 0x00000000 
ebp = 0xcccf5ca4 xss = 0x00000068 xcs = 0x00000060 eflags = 0x00010202 
xds = 0x0000007b xes = 0x0000007b origeax = 0xffffffff &regs = 0xcccf5c34
[0]kdb> bt
Stack traceback for pid 71515
0xf6734330    71515    71505  1    0   R  0xf6734500 *mount
EBP        EIP        Function (args)
0xcccf5ca4 0x00000000 <unknown>
[0]kdb> 

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-06-17 14:50

Message:
Logged In: YES 
user_id=1246761
Originator: YES

...but CFS does O_SYNC for chard mounts.  Testing stability of this fix.

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-05-26 13:17

Message:
Logged In: YES 
user_id=1246761
Originator: YES

This oops is fixed, but XFS does not offer data consistency in failover
scenarios due to XFS data journaling is writeback.

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-04-21 19:52

Message:
Logged In: YES 
user_id=1246761
Originator: YES

Testing XFS failover without iget by rebuilding the exportfs file handle.

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-04-19 06:50

Message:
Logged In: YES 
user_id=1246761
Originator: YES

Oops occured while inside inline iget() call to read_inode() which is
depreciated in XFS.

Need to stop using iget() in CFS and instead use exportfs.

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-04-16 02:15

Message:
Logged In: YES 
user_id=1246761
Originator: YES

This oops occurs only when there is a CFS stacked XFS hard mount.

----------------------------------------------------------------------

Comment By: Roger Tsang (rogertsang)
Date: 2007-04-15 01:35

Message:
Logged In: YES 
user_id=1246761
Originator: YES

The title of this bug report may be misleading.  This might not be XFS
specific as rc.nodedown calls mount -a -d {down_node} after fsck.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=405834&aid=1700876&group_id=32541

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/