Re: [PATCH] scsi: restart list search after unlock in scsi_remove_target

[Dan Williams <dan.j.williams@intel.com>]

On Mon, Oct 19, 2015 at 8:56 AM, Christoph Hellwig <hch@lst.de> wrote:
> On Mon, Oct 19, 2015 at 08:36:23AM -0700, Bart Van Assche wrote:
>> Thanks for looking into this. However, I think we need a motivation in the
>> patch description why this patch does not reintroduce the soft lockup
>> documented in patch "scsi_remove_target: fix softlockup regression on hot
>> remove" (commit bc3f02a795d3).
>
> Interesting.  I tried to find the original report and what state
> changes would cause an endless loop here.  Dan, do you remember any
> details about this bug report?

I believe the issue I was seeing back then might have been fixed or at
least modulated by "f2495e228fce [SCSI] dual scan thread bug fix"
which came a few years later.  The original problem was hot-remove
racing hot-add and that scsi_target_reap() was not guaranteed to
advance the state of the target if it was in the process of being
scanned when a removal event arrived.  However the comment in that
change:

+       /*
+        * if we get here and the target is still in the CREATED state that
+        * means it was allocated but never made visible (because a scan
+        * turned up no LUNs), so don't call device_del() on it.
+        */

...is not what I was seeing.  The target was in the CREATED state
because it had not yet completed the initial scan before tear down was
initiated.