public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Max Kanushin <rextuz@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: jmoyer@redhat.com
Subject: Re: Question about request queues in I/O scheduling
Date: Thu, 12 May 2016 08:33:16 +0300	[thread overview]
Message-ID: <5734159C.5070207@gmail.com> (raw)
In-Reply-To: <x494makcrl3.fsf@segfault.boston.devel.redhat.com>

Hello,

I was trying to use spin_trylock(q->queue_lock) on several queues when 
q->nr_sorted exceeds the per-defined number to block them and unlock the 
them later with spin_unlock(q->queue_lock), but I have faced the 
following problem: my system freezes when I am trying to test it by 
moving files around in my system. I thought it might be caused by 
blocking a queue of a system partition or swap, but I do not know how to 
check this as well.
Is it the right way to stop the queue/elevator? What are freezes might 
be caused by? Is there a simple way to find out which block device is a 
queue belong to?
I would really appreciate your reply.

Best regards,
   Max Kanushin.

On 04/29/2016 07:22 PM, Jeff Moyer wrote:
> Max Kanushin <rextuz@gmail.com> writes:
>
>> Thank you very much for the reply. My general idea is take control of all
>> request queues to block and unblock them manually depending on their load.
>> One of my steps is to find a length of a queue to decide if to block it.
>> Actually I think I've found the way. If I got it right, I can move from one
>> request to another within the list:
>> struct list_head queue_head;
>> So that I can calculate how many requests are there in the queue.
>
> Hi, Max,
>
> The queue_head is the dispatch list for I/O, so it does not represent
> all I/O queued for a request_queue.  The number of requests in the
> scheduler would be reflected by q->nr_sorted.  The number of dispatched
> requests is in an array, in_flight.
>
> Cheers,
> Jeff
>
>
>
>>
>> Best regards,
>>    Max Kanushin.
>> On Apr 29, 2016 6:32 PM, "Jeff Moyer" <jmoyer@redhat.com> wrote:
>>
>>> Max Kanushin <rextuz@gmail.com> writes:
>>>
>>>> Hello,
>>>>
>>>> I was searching for a way to find out the length of a request_queue
>>>> (that is defined as a structure in
>>>> linux/include/linux/blkdev.h). However I am new to the kernel
>>>> development and can't figure out where is the actual list of requests
>>>> to be processed by an elevator.
>>>> Is there a way to iterate requests in a queue or at least find a
>>>> number of them?
>>>
>>> Hi, Max,
>>>
>>> What exactly are you trying to accomplish?
>>>
>>> Cheers,
>>> Jeff
>>>

      reply	other threads:[~2016-05-12  5:33 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-24 18:42 Question about request queues in I/O scheduling Max Kanushin
2016-04-29 15:31 ` Jeff Moyer
     [not found]   ` <CAL+kshPDB8j3j7Gp4Zm5WK3W8HbO2pm3SQcB28p0LMF3KtdOBQ@mail.gmail.com>
2016-04-29 16:22     ` Jeff Moyer
2016-05-12  5:33       ` Max Kanushin [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=5734159C.5070207@gmail.com \
    --to=rextuz@gmail.com \
    --cc=jmoyer@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox