Re: [PATCH] scsi_debug: fix cmd_per_lun, set to max_queue

From: John Garry <john.garry@huawei.com>
To: Ming Lei <ming.lei@redhat.com>
Cc: Douglas Gilbert <dgilbert@interlog.com>,
	<linux-scsi@vger.kernel.org>, <martin.petersen@oracle.com>,
	<jejb@linux.vnet.ibm.com>, <kashyap.desai@broadcom.com>,
	<axboe@kernel.dk>
Subject: Re: [PATCH] scsi_debug: fix cmd_per_lun, set to max_queue
Date: Fri, 16 Apr 2021 10:07:39 +0100	[thread overview]
Message-ID: <23078a4f-5fab-2a21-8c3f-d293c8e05402@huawei.com> (raw)
In-Reply-To: <YHlP4J2MKSCM+phB@T590>

On 16/04/2021 09:50, Ming Lei wrote:
>>>> My concern here is that it is a common pattern in LLDDs to overwrite the
>>>> initial shost member values between scsi_host_alloc() and scsi_add_host().
>>> OK, then can we move the fix into beginning of scsi_add_host()?
>> I suppose that would be ok, but we don't do much sanitizing shost values at
>> that point. Apart from failing can_queue == 0.
> .can_queue has been finalized in scsi_add_host(), since it will be used for
> setting tagset, so .can_queue is reliable at that time.
> 
>> I suppose failing can_queue < cmd_per_lun could also be added.
> That will fail add host for scsi_debug simply.

But we still should have Doug's patch regardless.

Anyway, I'll prepare a patch, so we can discuss further on that thread.

Thanks,
John