Bug#571980: /boot/vmlinuz-2.6.33-2-amd64: impi oops on intel motherboard

07-09-2012, 09:12 PM

Hi,

In 2010, ian@comtek.co.uk wrote:

> Any use of ipmi on a stock IBM 326m results, after a period of maybe a
> few minutes, in a kernel oops, after which I'm unable to run any
> programs or use anything really and the only option remaining is to restart.
>
> If I type:
> modprobe ipmi_devintf
> modprobe ipmi_si
> modprobe ipmi_poweroff
> modprobe ipmi_watchdog
>
> and do nothing else then it will 'oops'. I can use ipmitool before it
> dies.
[...]
> [ 502.946318] IPMI kcs interface initialized
> [ 1159.930234] BUG: unable to handle kernel paging request at f7277000
> [ 1159.934208] IP: [<c112eb8b>] find_next_zero_bit+0x43/0x83
[...]
> [ 1159.934208] Pid: 5247, comm: less Not tainted (2.6.32-trunk-686 #1) IBM eServer 326m -[796976Y]-
[...]
> [ 1159.934208] Call Trace:
> [ 1159.934208] [<f823f745>] ? ext3_new_inode+0x47e/0x906 [ext3]
> [ 1159.934208] [<c103b692>] ? mod_timer+0x18/0x1e
> [ 1159.934208] [<f82087a3>] ? start_this_handle+0x2c3/0x328 [jbd]
> [ 1159.934208] [<c13b6400>] ? intel_iommu_setup+0x5b/0x104
> [ 1159.934208] [<f820891b>] ? journal_start+0x81/0xac [jbd]
> [ 1159.934208] [<f824588d>] ? ext3_create+0x7d/0xe2 [ext3]
> [ 1159.934208] [<c10b979c>] ? vfs_create+0x5b/0x72
> [ 1159.934208] [<c10ba4a1>] ? do_filp_open+0x29d/0x7e0
> [ 1159.934208] [<c10b0353>] ? do_sys_open+0x4c/0xdf
> [ 1159.934208] [<c10b042a>] ? sys_open+0x1e/0x23
> [ 1159.934208] [<c10030fb>] ? sysenter_do_call+0x12/0x28
> [ 1159.934208] Code: ca 29 f3 83 e2 1f 74 33 b9 20 00 00 00 29 d1 83 ca ff d3 ea 0b 10 83 fb 1f 76 30 83 fa ff 75 3e 83 c0 04 83 eb 20 83 c6 20 eb 10 <8b> 10 83 fa ff 75 2c 83 c0 04 83 c6 20 83 eb 20 f7 c3 e0 ff ff
> [ 1159.934208] EIP: [<c112eb8b>] find_next_zero_bit+0x43/0x83 SS:ESP

In 2010, Xavier Bestel wrote:

> Here's the oops:
> [ 16.507977] IPMI System Interface driver.
> [ 16.508187] general protection fault: 0000 [#1] SMP
> [ 16.508189] last sysfs file: /sys/module/ipmi_msghandler/initstate
> [ 16.508191] CPU 0
> [ 16.508194] Pid: 1924, comm: modprobe Not tainted 2.6.33-2-amd64 #1 S975XBX2/
[...]
> [ 16.508233] Call Trace:
> [ 16.508237] [<ffffffff8120f5bb>] ? driver_register+0x8b/0x111
> [ 16.508243] [<ffffffffa03248df>] ? init_ipmi_si+0x5f1/0x79d [ipmi_si]
> [ 16.508246] [<ffffffff81182dc1>] ? __up_read+0x13/0x99
> [ 16.508251] [<ffffffffa03242ee>] ? init_ipmi_si+0x0/0x79d [ipmi_si]
> [ 16.508254] [<ffffffff81002065>] ? do_one_initcall+0x64/0x174
> [ 16.508258] [<ffffffff8107485e>] ? sys_init_module+0xc5/0x21a
> [ 16.508261] [<ffffffff8103146f>] ? sysenter_dispatch+0x7/0x2e
> [ 16.508262] Code: 00 e8 28 ff ff ff 48 85 c0 48 89 c3 74 0f 48 c7 c6 70 50 66 81 48 89 c7 e8 57 ff ff ff 48 89 d8 5b c3 48 85 ff 53 48 89 fb 74 37 <f6> 47 3c 01 75 20 48
> 8b 0f 49 89 f8 48 c7 c2 58 6e 4c 81 be 53
> [ 16.508278] RIP [<ffffffff8117fb95>] kobject_put+0x9/0x42
> [ 16.508281] RSP <ffff88010b209ea8>
> [ 16.508283] ---[ end trace 193d07c5a4c14e5e ]---

Thanks to both of you. Do you still have access to this hardware? If so,
what kernel do you use, and how does it behave?

Curious,
Jonathan

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: http://lists.debian.org/20120709211258.GA17780@burratino

07-10-2012, 09:30 AM

Le lundi 09 juillet 2012 à 16:12 -0500, Jonathan Nieder a écrit :
> Thanks to both of you. Do you still have access to this hardware? If so,
> what kernel do you use, and how does it behave?

Yes, it runs 3.2.0-2-amd64, it seems to behave like a champ. At least I
didn't spot an IPMI-related Oops recently (but I don't reboot it that
often).

Xav

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: http://lists.debian.org/1341912621.21667.10.camel@bip

07-10-2012, 09:30 PM

fixed 571980 linux/3.2.21-3
quit

Ian Crowther wrote:
> On 10/07/12 10:30, Xavier Bestel wrote:

>> Yes, it runs 3.2.0-2-amd64, it seems to behave like a champ. At least I
>> didn't spot an IPMI-related Oops recently (but I don't reboot it that
>> often).
>
> Ours are sitting on a shelf, powered off. I can use one to test a
> specific kernel if wanted.

Thanks, both.

If either of you gets a chance to try the latest 2.6.32.y kernel from
squeeze (it should work fine on a squeeze or wheezy/sid system), then
that would be interesting.

Hope that helps,
Jonathan

(A few upstream commits between 2.6.32.y and 3.0.y for reference:

9e368fa011d4 ipmi: add PNP discovery (ACPI namespace via PNPACPI)
c7df670bf702 sysfs: fix sysfs lockdep warning in ipmi code
56480287f977 ipmi: make sure drivers were registered before
unregistering them
561f8182dbf2 ipmi: fix hardcoded ipmi device exit path warning
de5e2ddf9bb3 ipmi: proper spinlock initialization
d2478521afc2 char/ipmi: fix OOPS caused by pnp_unregister_driver on
unregistered driver (as corrected by 0dcf334c44d9)
)

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: http://lists.debian.org/20120710213039.GA20050@burratino

07-12-2012, 02:44 PM

Thanks, both.

If either of you gets a chance to try the latest 2.6.32.y kernel from
squeeze (it should work fine on a squeeze or wheezy/sid system), then
that would be interesting.

I'm not sure what 2.6.32.y means. I've installed 2.6.32-5-686, which
seems to be recent. That has the same problem (errors after loading ipmi
modules), though.

Ian

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 4FFEE2BE.2030901@comtek.co.uk">http://lists.debian.org/4FFEE2BE.2030901@comtek.co.uk

07-12-2012, 08:21 PM

Ian Crowther wrote:

> I'm not sure what 2.6.32.y means. I've installed 2.6.32-5-686, which
> seems to be recent.

Am I correct in assuming you mean version 2.6.32-45? You can check
with

dpkg-query -W linux-image-$(uname -r)

to see the version number of the installed kernel, or

cat /proc/version

to see the version number of the running one in parentheses.

2.6.32-45 is close to 2.6.32.59 from kernel.org.

> That has the same problem (errors after loading
> ipmi modules), though.

Thanks for checking.

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: http://lists.debian.org/20120712202150.GA2956@burratino

07-13-2012, 04:00 PM

On 12/07/12 21:21, Jonathan Nieder wrote:

Ian Crowther wrote:

I'm not sure what 2.6.32.y means. I've installed 2.6.32-5-686, which
seems to be recent.

Am I correct in assuming you mean version 2.6.32-45? You can check
with

dpkg-query -W linux-image-$(uname -r)

to see the version number of the installed kernel, or

cat /proc/version

to see the version number of the running one in parentheses.

2.6.32-45 is close to 2.6.32.59 from kernel.org.

That has the same problem (errors after loading
ipmi modules), though.

Thanks for checking.

Sorry, yes, 2.6.32-45 retains the problem ('modprobe ipmi_si' generates
a bunch of messages and results in an unusable system).

3.2.0-3-686-pae (debian 3.2.21-3) works for at least an hour with all
the ipmi modules loaded.

Thanks,

Ian

--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 5000462B.2060800@comtek.co.uk">http://lists.debian.org/5000462B.2060800@comtek.co.uk