Re: [Scst-devel] Conn reinst with outstanding commands

[John B. <ba...@sp...>]

Thanks!

Vladislav Bolkhovitin wrote:
> John Baboval, on 02/24/2010 07:42 PM wrote:
>   
>> I've been experiencing a deadlock with iscsi-scst with both [iscsird0]
>> and iscsi-scstd (in an ioctl) waiting on target_mutex. I've traced the
>> problem down:
>>
>> iscsi_task_mgmt_affected_cmds_done() calls
>> sess_reinst_finished() calls
>> conn_reinst_finished()...
>>
>> If there are items on the pending commands list, conn_reinst_finished
>> calls iscsi_restart_cmnd().
>> iscsi_restart_cmnd() does iscsi_extracheck_is_rd_thread(cmnd->conn)...
>> At this point, conn->rd_task is NULL, so the printk causes a NULL
>> pointer dereference, and the tm thread dies holding the target_mutex.
>> Even fixing the NULL pointer dereference, though, this code path is not
>> the rd_thread, so the check would BUG() and I'd still be dead in the
>> water...
>>
>> I've tried replacing iscsi_restart_cmnd() with req_cmnd_release_force()
>> and adding support for ISCSI_CMD_STATE_REINST_PENDING to cmnd_done as a
>> hack to try to workaround the problem, but that caused other issues.
>>
>> If somebody could provide me with some pointers as to the correct fix
>> for this issue, I'd appreciate it.
>>     
>
> Basically, iscsi_extracheck_is_rd_thread() is not needed in 
> iscsi_restart_cmnd(), it's an artifact. You can safely delete it. I did 
> it in r1520 together with some other small fixes and cleanups.
>
> Thanks for reporting,
> Vlad
>
>
> ------------------------------------------------------------------------------
> Download Intel® Parallel Studio Eval
> Try the new software tools for yourself. Speed compiling, find bugs
> proactively, and fine-tune applications for parallel performance.
> See why Intel Parallel Studio got high marks during beta.
> http://p.sf.net/sfu/intel-sw-dev
> _______________________________________________
> Scst-devel mailing list
> Scs...@li...
> https://lists.sourceforge.net/lists/listinfo/scst-devel
>