Re: [PATCH v6 10/13] Make scsi_remove_host() wait for device removal

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

On 12/03/12 17:15, Tejun Heo wrote:
> On Mon, Dec 03, 2012 at 09:23:59AM +0100, Bart Van Assche wrote:
>> On 12/02/12 14:45, Tejun Heo wrote:
>>> On Wed, Nov 28, 2012 at 01:51:13PM +0100, Bart Van Assche wrote:
>>>> +/* Test whether scsi_remove_host() may finish, and if so, wake it up. */
>>>> +void __scsi_check_remove_host_done(struct Scsi_Host *shost)
>>>> +{
>>>> +	lockdep_assert_held(shost->host_lock);
>>>> +
>>>> +	if (__scsi_remove_host_done(shost))
>>>> +		wake_up(&shost->remove_host);
>>>> +}
>>>
>>> This being an one-time thing.  Using completion could be simpler.  e.g.
>>
>> Sorry but I'm not sure that would work here. A user can e.g. delete
>> all SCSI devices associated with a SCSI host (echo 1
>>> /sys/class/scsi_host/host<n>/device/target<m>/<lun>/delete) and
>> then issue a rescan to re-add LUNs. When using a completion the
>> completion would be set as soon as the last SCSI device has been
>> deleted instead of only after scsi_remove_host() finished removing
>> the re-added SCSI devices.
>
> You can gate the whole thing with test on whether the host is being
> removed.  Once removal begins, no device should be allowed to be added
> and the completion triggers iff host removal is in progress.  That
> should work, right?

It is indeed possible to invoke complete() only if the device list 
became empty with the host state equal to SHOST_CANCEL, 
SHOST_CANCEL_RECOVERY, SHOST_DEL or SHOST_DEL_RECOVERY and in 
scsi_remove_host() to wait for that completion only if the device list 
was not empty before the host state was changed into one of the four 
mentioned states. Do you really prefer this approach over the approach 
in the patch at the start of this thread ?

Thanks,

Bart.