Re: Fw: [Bug 65131] New: kernel panic (BUG_ON raised) in SCTP function sctp_cmd_interpreter

[Vlad Yasevich <vyasevich@gmail.com>]

On 11/18/2013 12:46 PM, Vlad Yasevich wrote:
> On 11/18/2013 12:14 PM, Stephen Hemminger wrote:
>>
>>
>> Begin forwarded message:
>>
>> Date: Sun, 17 Nov 2013 19:38:56 -0800
>> From: "bugzilla-daemon@bugzilla.kernel.org" <bugzilla-daemon@bugzilla.kernel.org>
>> To: "stephen@networkplumber.org" <stephen@networkplumber.org>
>> Subject: [Bug 65131] New: kernel panic (BUG_ON raised) in SCTP function sctp_cmd_interpreter
>>
>>
>> https://bugzilla.kernel.org/show_bug.cgi?id=65131
>>
>>             Bug ID: 65131
>>            Summary: kernel panic (BUG_ON raised) in SCTP function
>>                     sctp_cmd_interpreter
>>            Product: Networking
>>            Version: 2.5
>>     Kernel Version: 3.11.8 custom build, repeated on 3.11.2
>>           Hardware: All
>>                 OS: Linux
>>               Tree: Mainline
>>             Status: NEW
>>           Severity: blocking
>>           Priority: P1
>>          Component: IPV4
>>           Assignee: shemminger@linux-foundation.org
>>           Reporter: yuras@uch.net
>>         Regression: No
>>
>> Created attachment 114991
>>   --> https://bugzilla.kernel.org/attachment.cgi?id=114991&action=edit
>> Screenshot of panic
>>
>> Two-node cluster configured using latest corosync (also DRBD 8.4.4, LVM2, and
>> GFS2 but this is unessential).
>> Steps to reproduce:
>> 1. Start corosync on both nodes.
>> 2. Start dlm_controld (version 4.0.2) on both nodes (used SCTP protocol as TCP
>> cannot be used on multi-homed hosts). Adds such lines to kern.log:
>>     kernel: [  580.428664] sctp: Hash tables configured (established 65536 bind
>> 65536)
>>     kernel: [  580.441779] DLM installed
>> 3. Start clvmd on either node. Adds such lines to kern.log:
>>     kernel: [ 1345.259502] dlm: Using SCTP for communications
>>     kernel: [ 1345.260699] dlm: clvmd: joining the lockspace group...
>>     kernel: [ 1345.262962] dlm: clvmd: dlm_recover 1
>>     kernel: [ 1345.262968] dlm: clvmd: group event done 0 0
>>     kernel: [ 1345.262992] dlm: clvmd: add member 1024
>>     kernel: [ 1345.262995] dlm: clvmd: dlm_recover_members 1 nodes
>>     kernel: [ 1345.262996] dlm: clvmd: join complete
>>     kernel: [ 1345.262998] dlm: clvmd: generation 1 slots 1 1:1024
>>     kernel: [ 1345.262999] dlm: clvmd: dlm_recover_directory
>>     kernel: [ 1345.263000] dlm: clvmd: dlm_recover_directory 0 in 0 new
>>     kernel: [ 1345.263002] dlm: clvmd: dlm_recover_directory 0 out 0 messages
>>     kernel: [ 1345.263019] dlm: clvmd: dlm_recover 1 generation 1 done: 0 ms
>> 4. Start clvmd on second node. With high probability one node or both nodes
>> panic in the similar way. Screenshot in attachment.
>>
>> Stack trace can differ slightly above EOI line, but RIP was always the same. I
>> suppose provided CPU codes correspond to one of BUG_ON macro inside
>> sctp_cmd_interpreter. So, this is a bug.
>>
>> Now this bug totally prevents me from using my cluster as DLM rejects to use
>> TCP for multi-homed hosts.
>>
> 
> Should be fixed by:
> commit 7926c1d5be0b7cbe5b8d5c788d7d39237e7b212c
> Author: Daniel Borkmann <dborkman@redhat.com>
> Date:   Thu Oct 31 09:13:32 2013 +0100
> 
>     net: sctp: do not trigger BUG_ON in sctp_cmd_delete_tcb
> 
> -vlad
> 

Just received confirmation that the above patch has been queued for 3.11.

-vlad