Re: [PATCH RFC 00/24] scsi: enable reserved commands for LLDDs

From: Hannes Reinecke <hare@suse.de>
To: John Garry <john.garry@huawei.com>,
	"Martin K. Petersen" <martin.petersen@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>,
	James Bottomley <james.bottomley@hansenpartnership.com>,
	linux-scsi@vger.kernel.org, Ming Lei <ming.lei@redhat.com>
Subject: Re: [PATCH RFC 00/24] scsi: enable reserved commands for LLDDs
Date: Mon, 9 Sep 2019 17:11:22 +0200	[thread overview]
Message-ID: <600814b4-15dc-e4ec-ae00-ca5a72351012@suse.de> (raw)
In-Reply-To: <64d4b289-555d-8854-ca2b-ec26d397061e@huawei.com>

On 9/2/19 10:47 AM, John Garry wrote:
> On 26/08/2019 16:27, Hannes Reinecke wrote:
>> On 8/23/19 3:26 PM, John Garry wrote:
>>> On 29/05/2019 14:28, Hannes Reinecke wrote:
>>>> Hi all,
>>>>
>>>> quite some drivers use internal commands for various purposes, most
>>>> commonly sending TMFs or querying the HBA status.
>>>> While these commands use the same submission mechanism than normal
>>>> I/O commands, they will not be counted as outstanding commands,
>>>> requiring those drivers to implement their own mechanism to figure
>>>> out outstanding commands.
>>>> This patchset enables the use of reserved tags for the SCSI midlayer,
>>>> enabling LLDDs to rely on the block layer for tracking outstanding
>>>> commands.
>>>> More importantly, it allows LLDD to request a valid tag from the block
>>>> layer without having to implement some tracking mechanism within the
>>>> driver. This removes quite some hacks which were required for some
>>>> drivers (eg. fnic or snic).
>>>>
>>>> As usual, comments and reviews are welcome.
>>>>
>>>
>>> Hi Hannes,
>>>
>>> I was wondering if you have any plans to progress this series?
>>>
>>> I don't mind helping out...
>>>
>> Thanks for the reminder.
> 
> Hi Hannes,
> 
>> I'll need to re-evaluate this where we stand now with shared tags;
> 
> As Ming Lei mentioned in
> https://www.spinics.net/lists/linux-block/msg43779.html, the future for
> hostwide shared tags doesn't look bright. I would tend to agree.
> 
>> this patchset partially relies on them.
> 
> In which way? I didn't think/hope it did.
> 
>> Will be sending an updated patchset.
> 
> For me, the way I see forward is to upstream this series, in addition to
> Ming's, linked above.
> 
> As for LLDDs and the unique tag problem, I suggest that they use sbitmap
> to generate the tag internally if they want to expose multiple queues.
> From our testing, using managed interrupts + sbitmap still far
> outperforms non-managed interrupts.
> 
> This approach would mean that we can still revisit hostwide shared tags
> or other some other approach in future.
> 
Hmm. True. Will be checking.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      Teamlead Storage & Networking
hare@suse.de			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 247165 (AG München), GF: Felix Imendörffer