Message-ID: <d2e452cd-6a91-f25d-748e-94f560deb1bb@opensource.wdc.com>
Date:   Mon, 13 Jun 2022 18:43:59 +0900
From:   Damien Le Moal <damien.lemoal@...nsource.wdc.com>
To:     John Garry <john.garry@...wei.com>, axboe@...nel.dk,
        jejb@...ux.ibm.com, martin.petersen@...cle.com, brking@...ibm.com,
        hare@...e.de, hch@....de
Cc:     linux-block@...r.kernel.org, linux-ide@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-scsi@...r.kernel.org,
        chenxiang66@...ilicon.com
Subject: Re: [PATCH RFC v2 03/18] scsi: core: Implement reserved command
 handling

On 6/13/22 18:34, John Garry wrote:
> On 13/06/2022 10:06, Damien Le Moal wrote:
>>>> We cannot have more than 32 tags.
>>> We may have 32 regular tags and 1 reserved tag for SATA.
>> Right. But that is the messy part though. That extra 1 tag is actually not
>> a tag since all internal commands are non-NCQ commands that do not need a
>> tag...
> 
> But apart from SATA, libsas LLDDs do need a real tag for the libata 
> internal command.

Yes, but that is really to manage the LLD command descriptor and not for
the device to use in the case of non-NCQ commands. There is not
necessarily a 1:1 equivalence between the HBA command descriptor tag and
ATA command tag.

> 
>>
>> I am working on command duration limits support currently. This feature
>> set has a new horrendous "improvement": a command can be aborted by the
>> device if it fails its duration limit, but the abort is done with a good
>> status + sense data available bit set so that the device queue is not
>> aborted entirely like with a regular NCQ command error.
>>
>> For such aborted commands, the command sense data is set to
>> "COMPLETED/DATA UNAVAILABLE". In this case, the host needs to go read the
>> new "successful NCQ sense data log" to check that the command sense is
>> indeed "COMPLETED/DATA UNAVAILABLE". And to go read that log page without
>> stalling the device queue, we would need an internal NCQ (queuable) command.
>>
>> Currently, that is not possible to do cleanly as there are no guarantees
>> we can get a free tag (there is a race between block layer tag allocation
>> and libata internal tag counting). So a reserved tag for that would be
>> nice. We would end up with 31 IO tags at most + 1 reserved tag for NCQ
>> commands + ATA_TAG_INTERNAL for non-NCQ. That last one would be rendered
>> rather useless. But that also means that we kind-of go back to the days
>> when Linux showed ATA drives max QD of 31...
> 
> So must the ATA_TAG_INTERNAL qc always be available for non-NCQ action 
> like EH, and that is why you cannot reuse for this internal NCQ 
> (queuable) command?

Currently, ATA_TAG_INTERNAL is always used for non-NCQ commands. Seeing a
qc with that tag means it is *not* NCQ.

I am trying to see if I can reuse the tag from one of the commands that
completed with that weird good status/sense data available. The problem
though is that this all needs to be done *before* calling
qc->complete_fn() which will free the tag. So we endup with 2 qcs that
have the same tag, the second one (for the read log command) temporarily
using the tag but still going through the same completion path without the
original command fully completed yet. It is a real mess.

> 
>>
>> I am still struggling with this particular use case and trying to make it
>> fit with your series. Trying out different things right now.
>>
> 
> ok
> 
>>
>>>> I think keeping can_queue as the max queue depth with at most
>>>> nr_reserved_cmds tags reserved is better.
>>> Maybe the wording in the comment can be improved as it originally
>>> focused on SAS HBAs where there are no special rules for tagset depth or
>>> how the tagset should be carved up to handle regular and reserved commands.
>> Indeed. And that would be for HBAs that do*not*  use libsas/libata.
>> Otherwise, the NCQ vs non-NCQ reserved tag mess is there.
>>
> 
> Thanks,
> John


-- 
Damien Le Moal
Western Digital Research

Powered by blists - more mailing lists