Re: [PATCH 3/3] nvme-core: mark internal passthru req REQ_QUIET

[Jonathan Derrick <jonathan.derrick@linux.dev>]

On 4/13/2022 11:09 AM, Keith Busch wrote:
> On Wed, Apr 13, 2022 at 10:57:19AM -0600, Jonathan Derrick wrote:
>>
>>
>> On 4/13/2022 10:52 AM, Christoph Hellwig wrote:
>>> On Wed, Apr 13, 2022 at 10:49:33AM -0600, Jonathan Derrick wrote:
>>>>>     +	req->rq_flags |= RQF_QUIET;
>>>> Any reason to not make it a part of flags in the argument list and let the
>>>> block layer build the req with it?
>>>>
>>>> eg:
>>>> flags |= REQ_QUIET;
>>>
>>> Because there is not REQ_QUIET, just RQF_QUIET and BIO_QUIET, and their
>>> relation isn't quite as straightforward as it seems.
>>
>> Well let's suppose I had written RQF_QUIET instead of the non-existent
>> REQ_QUIET
> 
> The "flags" value passed to blk_mq_alloc_request() is not the same as the
> values for rq_flags, though.
> 
> I guess you could add a new allocation flag, maybe called something like
> BLK_MQ_REQ_QUIET to tell the block layer to set RQF_QUIET. That is similiar to
> how BLK_MQ_REQ_PM is used for RQF_PM, but I'm not sure we want to add another
> 'if' check in the fast path.

Fair enough