Bug 138 - Oops when mounting with krb5i flavor on linux-2.6.20-rc4-CITI_NFS4_ALL-1.diff
: Oops when mounting with krb5i flavor on linux-2.6.20-rc4-CITI_NFS4_ALL-1.diff
Status: RESOLVED FIXED
Product: kernel
client
: unspecified
: PC Linux
: P1 blocker
Assigned To: Trond Myklebust
:
:
:
:
  Show dependency treegraph
 
Reported: 2007-01-09 06:47 UTC by Aime Le Rouzic
Modified: 2007-03-27 02:01 UTC (History)
2 users (show)

See Also:


Attachments
NFS: Fix Oops in rpc_call_sync() (2.44 KB, patch)
2007-01-16 04:30 UTC, Trond Myklebust
Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Aime Le Rouzic 2007-01-09 06:47:11 UTC
Hi,
Using the last linux-2.6.20-rc4-CITI_NFS4_ALL-1.diff I get the following OOps
when mounting with krb5i .
Client and server are Intel Xeon 32 bits.

Hereis the trace:

Kernel BUG at net/sunrpc/sched.c:902!
invalid opcode: 0000 [#1]
SMP
Modules linked in: iptable_filter ip_tables x_tables ppdev autofs4 hidp rfc
omm l2cap bluetooth video button battery asus_acpi backlight ac lp parport_
pc parport floppy nvram uhci_hcd sg e1000 e100 mii e7xxx_edac pcspkr i2c_i8
01 edac_mc i2c_core dm_snapshot dm_zero dm_mirror dm_mod ext3 jbd aic79xx s
csi_transport_spi sd_mod scsi_mod
CPU:    0
EIP:    0060:[<c03b9574>]    Not tainted VLI
EFLAGS: 00010297   (2.6.20-rc4-CITI_NFS4_ALL-1 #0)
EIP is at rpc_release_task+0x11/0x9c
eax: f7f3d580   ebx: f7f3d580   ecx: f0f0e140   edx: c189c100
esi: fffffff3   edi: c8fa8a7c   ebp: c8fa8a3c   esp: c8fa8a30
ds: 007b   es: 007b   ss: 0068
Process mount (pid: 3513, ti=c8fa8000 task=f50c3570 task.ti=c8fa8000)
Stack: c03b50c5 f7f3d580 f7f3d580 c8fa8a5c c03b5220 00000000 00000000 00000
000
       c8fa8a7c f760a0c0 c8fa8a7c c8fa8ab0 c01e0cb0 c8fa8bec c8fa8d16 00000
000
       00000000 00000000 00000000 c04f1ca0 c8fa8aa0 c8fa8a8c 00000000 f760a
0c0
Call Trace:
 [<c01052dc>] show_trace_log_lvl+0x1a/0x2f
 [<c010538c>] show_stack_log_lvl+0x9b/0xa3
 [<c0105529>] show_registers+0x195/0x26b
 [<c010571b>] die+0x11c/0x27b
 [<c03cb1b7>] do_trap+0x79/0x91
 [<c0105d0a>] do_invalid_op+0x97/0xa1
 [<c03caf8c>] error_code+0x7c/0x84
 [<c03b5220>] rpc_call_sync+0x6b/0xa1
 [<c01e0cb0>] nfs4_proc_get_root+0x87/0xc5
 [<c01d1df9>] nfs4_path_walk+0x95/0x29b
 [<c01ce404>] nfs4_create_server+0x1ac/0x2f1
 [<c01d4f43>] nfs4_get_sb+0x1e0/0x2d1
[<c016c0e9>] vfs_kern_mount+0x81/0xf1
 [<c016c1a1>] do_kern_mount+0x30/0x42
 [<c017e71a>] do_mount+0x5fc/0x677
 [<c017e804>] sys_mount+0x6f/0xa4
 [<c01042c2>] sysenter_past_esp+0x5f/0x85
 =======================
Code: 43 70 ba 1a 93 3b c0 e8 83 d5 d7 ff 89 f2 89 f8 e8 7c ff ff ff 5b 5e
5f 5d c3 55 89 e5 53 89 c3 83 ec 08 81 38 aa 0b f0 00 74 04 <0f> 0b eb fe f
6 05 08 b6 62 c0 40 74 17 0f b7 80 94 00 00 00 c7
EIP: [<c03b9574>] rpc_release_task+0x11/0x9c SS:ESP 0068:c8fa8a30

Entering kdb (current=0xf50c3570, pid 3513) on processor 0 Oops: invalid op    
               code
due to oops @ 0xc03b9574
eax = 0xf7f3d580 ebx = 0xf7f3d580 ecx = 0xf0f0e140 edx = 0xc189c100
esi = 0xfffffff3 edi = 0xc8fa8a7c esp = 0xc8fa8a30 eip = 0xc03b9574
ebp = 0xc8fa8a3c xss = 0xc0320068 xcs = 0x00000060 eflags = 0x00010297
xds = 0x0000007b xes = 0x0000007b origeax = 0xffffffff &regs = 0xc8fa89f8
[0]kdb>


Cheers
Comment 1 Trond Myklebust 2007-01-09 08:47:09 UTC
Can it be reproduced on 2.6.20-rc4 without the CITI patch?
Comment 2 Aime Le Rouzic 2007-01-10 06:42:48 UTC
(In reply to comment #1)
> Can it be reproduced on 2.6.20-rc4 without the CITI patch?
> 

Hi,
I did several tries . I can reproduced it also on 2.6.20-rc4 without the CITI
patch but not each time!!!!

Best regards
Comment 3 Trond Myklebust 2007-01-16 04:30:43 UTC
Created attachment 51 [details]
NFS: Fix Oops in rpc_call_sync()

We shouldn't be calling rpc_release_task() for tasks that are not active.
Comment 4 Aime Le Rouzic 2007-01-16 08:57:46 UTC
(In reply to comment #3)
> Created an attachment (id=51)
> NFS: Fix Oops in rpc_call_sync()
> 
> We shouldn't be calling rpc_release_task() for tasks that are not active.
> 

Hi,
The patch in attachment fixes the problem.
This issue can be closed.
Thanks
Comment 5 Aime Le Rouzic 2007-03-27 02:01:43 UTC
Hi,

Let's close this issue since fixed.

Cheers