Re: EH action after scsi_remove_host, was: Re: [PATCH v2 12/12] IB/srp: Add multichannel support

[Bart Van Assche <bvanassche@acm.org>]

On 10/17/14 13:01, Christoph Hellwig wrote:
> On Tue, Oct 07, 2014 at 03:07:17PM +0200, Bart Van Assche wrote:
>> +	/*
>> +	 * Avoid that the SCSI error handler tries to use this channel after
>> +	 * it has been freed. The SCSI error handler can namely continue
>> +	 * trying to perform recovery actions after scsi_remove_host()
>> +	 * returned.
>> +	 */
>> +	ch->target = NULL;
>
> Do you have a reproducer for that?  I think we should fix the root
> cause.

Hello Christoph,

The above assignment statement has been reported to fix a kernel oops 
that could be triggered by cable pulling. Regarding fixing the root 
cause: some time ago I had posted a patch series that makes 
scsi_remove_host() wait until all error handler callback functions have 
finished and also that prevents that any new error handler function 
calls are initiated after scsi_remove_host() has finished 
(http://thread.gmane.org/gmane.linux.scsi/82572/focus=87985). Should I 
repost that patch series ?

Bart.