[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <8f4997dd-f68b-cb57-14e2-379914dc14ee@collabora.com>
Date: Fri, 5 Jun 2020 14:49:07 -0300
From: André Almeida <andrealmeid@...labora.com>
To: Randy Dunlap <rdunlap@...radead.org>, axboe@...nel.dk,
corbet@....net, linux-block@...r.kernel.org,
linux-doc@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, kernel@...labora.com,
krisman@...labora.com
Subject: Re: [PATCH] docs: block: Create blk-mq documentation
Hello Randy,
On 5/27/20 6:58 PM, Randy Dunlap wrote:
> On 5/27/20 1:09 PM, André Almeida wrote:
>> Create a documentation providing a background and explanation around the
>> operation of the Multi-Queue Block IO Queueing Mechanism (blk-mq).
<snip>
>> +Hardware dispatch queues
>> +~~~~~~~~~~~~~~~~~~~~~~~~
>> +
>> +The hardware queue is a memory space shared with the block device (e.g. DMA)
>> +where the hardware can access and dispatch requests (represented by struct
>> +:c:type:`blk_mq_hw_ctx`). To run this queue, the block layer removes
>> +requests from the associated software queues and tries to dispatch to the
>> +hardware.
>
> This paragraph (above) says to me that these HW devices understand struct blk_mq_hw_ctx.
> Is that correct? Is is some kind of standard? If so, where?
>
This wasn't what I intended to say here. The struct blk_mq_hw_ctx is an
abstraction from the blk-mq API used by device drivers to map the
submission queue of the hardware. So no, HW devices can't understand
struct blk_mq_hw_ctx. This paragraph was reworked for v2.
>
> thanks for the documentation.
>
Thanks for the feedback.
Powered by blists - more mailing lists