Re: [PATCH v7 13/13] dm: add non power of 2 zoned target

From: Pankaj Raghav <p.raghav@samsung.com>
To: Damien Le Moal <damien.lemoal@opensource.wdc.com>, <hch@lst.de>,
	<snitzer@redhat.com>, <axboe@kernel.dk>
Cc: <bvanassche@acm.org>, <linux-kernel@vger.kernel.org>,
	<jiangbo.365@bytedance.com>, <hare@suse.de>,
	<pankydev8@gmail.com>, <dm-devel@redhat.com>,
	<jonathan.derrick@linux.dev>, <gost.dev@samsung.com>,
	<dsterba@suse.com>, <jaegeuk@kernel.org>,
	<linux-nvme@lists.infradead.org>, <Johannes.Thumshirn@wdc.com>,
	<linux-block@vger.kernel.org>
Subject: Re: [PATCH v7 13/13] dm: add non power of 2 zoned target
Date: Fri, 17 Jun 2022 10:03:09 +0200	[thread overview]
Message-ID: <f83eb255-89fe-90dc-3670-79b8684389f1@samsung.com> (raw)
In-Reply-To: <f4cf6348-dd94-aa82-7519-318248c51151@opensource.wdc.com>

On 2022-06-17 08:56, Damien Le Moal wrote:
>>
>> So we call this function device_not_matches_zone_sectors() from
>> validate_hardware_zoned_model() for each target and we let the validate
>> succeed even if the target's zone size is different from the underlying
>> device zone size if this feature flag is set. Let me know if I am
>> missing something and how this can be moved to
>> validate_hardware_zoned_model().
> 
> Your change does not match the function name
> device_not_matches_zone_sectors(), at all. So I think this is wrong.
> 
> The fact is that zone support in DM has been built under the following
> assumptions:
> 1) A zoned device can be used to create a *zoned* target (e.g. dm-linear,
> dm-flakey, dm-crypt). For this case, the target *must* use the same zone
> size as the underlying devices and all devices used for the target must
> have the same zone size.
> 2) A zoned device can be used to create a *regular* device target (e.g.
> dm-zoned). All zoned devices used for the target must have the same zone size.
> 
> This new target driver completely breaks (1) and does not fit with (2). I
> suspect this is why you are seeing problems with dm_revalidate_zones() as
> that one uses the underlying device instead of the target report zones.
> 
> Based on this analysis, validate_hardware_zoned_model() definitely needs
> to be changed. But device_not_matches_zone_sectors() is to check the
> assumptions (1) and (2) so changing it for your new case is wrong in my
> opinion. You need another set of assumptions (3) (define that well please)
> and modify validate_hardware_zoned_model() so that the defined constraints
> are checked. Using a target flag to indicate the type of zoned target is
> fine by me.
> 
Got it. Thanks for the explanation. Renaming
device_not_matches_zone_sectors() function to something meaningful with
my changes should address what you have pointed out to accommodate all
three types.

I see that something similar was done to dm_table_supports_zoned_model()
to accommodate type 2(dm-zoned) with different underlying zoned models
even though the initial impl. supported only type 1.