[Users] SRP Abort called: five out of 66 paths offline

[Bart Van Assche]

On 02/06/15 00:31, Tom Robinson wrote:
> Sorry for the long post (lots of logging) so first my questions: should I be concerned about this
> WARNING? And, how do I get my faulty paths back online?
> 
> We are using COMSTAR targets on OmniOS and OFED (3.12) stacked on CentOS 7. Last night we got this
> WARNING in our linux host logs:
> 
> Feb  5 22:28:33 daytona kernel: ------------[ cut here ]------------
> Feb  5 22:28:33 daytona kernel: WARNING: at lib/list_debug.c:59 __list_del_entry+0xa1/0xd0()
> Feb  5 22:28:33 daytona kernel: list_del corruption. prev->next should be ffff88343e85fd38, but was
> ffff881faf801e80
> [ ... ]
> 
> I'm not sure which side is at fault here (COMSTAR/OFED). I suspect COMSTAR but the linux WARNING
> doesn't look pretty either.

Any warning about list corruption indicates an initiator side kernel
bug. The OFED 3.12 SRP initiator is based on the Linux kernel 3.12 SRP
initiator. Later SRP initiator versions are much more robust with regard
to error handling. Since OFED 3.18 is not yet available, please consider
to switch to the SRP initiator available from
https://github.com/bvanassche/ib_srp-backport.

Bart.