Re: [PATCH 01/15] nvme: Introduce namespace features flag

From: Max Gurtovoy <maxg@mellanox.com>
To: Keith Busch <kbusch@kernel.org>
Cc: axboe@kernel.dk, sagi@grimberg.me,
	"Martin K. Petersen" <martin.petersen@oracle.com>,
	shlomin@mellanox.com, israelr@mellanox.com,
	vladimirk@mellanox.com, linux-nvme@lists.infradead.org,
	idanb@mellanox.com, oren@mellanox.com, hch@lst.de
Subject: Re: [PATCH 01/15] nvme: Introduce namespace features flag
Date: Tue, 14 Jan 2020 18:04:20 +0200	[thread overview]
Message-ID: <2a237612-2114-ec1a-9cc8-91c0464591ad@mellanox.com> (raw)
In-Reply-To: <20200113203135.GB23750@redsun51.ssa.fujisawa.hgst.com>

On 1/13/2020 10:31 PM, Keith Busch wrote:
> On Sun, Jan 12, 2020 at 11:40:31AM +0200, Max Gurtovoy wrote:
>> On 1/9/2020 6:26 PM, Keith Busch wrote:
>>> On Thu, Jan 09, 2020 at 12:38:05PM +0200, Max Gurtovoy wrote:
>>>> On 1/9/2020 5:11 AM, Martin K. Petersen wrote:
>>>>> Max,
>>>>>
>>>>>> +#define NVME_NS_DIX_SUPPORTED	(1 << 1)
>>>>>> +#define NVME_NS_DIF_SUPPORTED	(1 << 2)
>>>>> Not so keen on the DIF/DIX terminology in this context.
>>>> Martin,
>>>>
>>>> how about:
>>>>
>>>> +#define NVME_NS_PI_HOST_SUPPORTED (1 << 1)
>>>> +#define NVME_NS_PI_CTRL_SUPPORTED (1 << 2)
>>> Well, I was trying to say earlier that nvme supports formats with metadata
>>> that's not used for protection information. The metadata, whether separate
>>> or interleaved, can be used for some proprietary non-pi related feature.
>> so how about:
>>
>> +#define NVME_NS_MD_HOST_SUPPORTED (1 << 1)
>> +#define NVME_NS_MD_CTRL_SUPPORTED (1 << 2)
> You're using these to report the in-memory format of the meta-data,
> right? These are either Extended or Separate from the block data, so
> I think the names should convey that. Something like the following
> maybe:
>   
>    #define NVME_NS_MD_SEP_SUPPORTED (1 << 1)
>    #define NVME_NS_MD_EXT_SUPPORTED (1 << 2)

No, we're setting it to know if we have host MD (aka DIX, memory-domain) 
and if we have ctrl MD (aka DIF, ctrl-domain).

for the MD format we use:

#define NVME_NS_EXT_LBAS	(1 << 0)

>   
>> We didn't change any logic for metadata support for non-PI format.
>>
>>> The nvme driver only leverages "blk-integrity" to facilitate allocating
>>> and managing the metdata payloads even when not used for integrity. It
>>> might make sense to give that block component a more generic name than
>>> "integrity".
>> Yup that True, But this patchset is already big enough by itself.
>>
>> We can do block layer refactoring in different series..
> Absolutely, didn't mean to imply you take that on. :)
>
> I was just trying to emphasize the current name doesn't always describe how
> that component is used.

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme