Lets make wifi fast again!
 help / color / mirror / Atom feed
From: "Toke Høiland-Jørgensen" <toke@toke.dk>
To: "Mohan\, Nitinder" <nitinder.mohan@helsinki.fi>
Cc: "make-wifi-fast\@lists.bufferbloat.net"
	<make-wifi-fast@lists.bufferbloat.net>,
	"bloat-devel\@lists.bufferbloat.net"
	<bloat-devel@lists.bufferbloat.net>
Subject: Re: [Make-wifi-fast] Get hardware queue length for wireless interface in linux kernel
Date: Tue, 16 May 2017 12:17:32 +0200	[thread overview]
Message-ID: <878tlxhzr7.fsf@alrua-kau> (raw)
In-Reply-To: <AM3PR07MB0661600959D3732708BF063E80E60@AM3PR07MB0661.eurprd07.prod.outlook.com> (Nitinder Mohan's message of "Tue, 16 May 2017 09:33:13 +0000")

"Mohan, Nitinder" <nitinder.mohan@helsinki.fi> writes:

> Hi Toke,
>
> Thank you for your quick reply.
>
> "Mohan, Nitinder" <nitinder.mohan@helsinki.fi> writes:
>
>>> Hi,
>>>
>>> I am a Ph.D. student working on a bufferbloat resistant scheduler for
>>> Multi-Path TCP. I was unsure of which list my question would be more
>>> suitable in so I am sending this in both make-wifi-fast and bloat-dev
>>> list.
>>
>> Sounds interesting! What, exactly, is such a scheduler going to be
>> doing? :)
>>
> The MPTCP scheduler takes into account lower layer buffer levels and
> schedules traffic over the interface to avoid HW bufferbloat.

Right. So how does this differ from using a low-prio CC (like LEDBAT) on
each of the MPTCP subflows? (If that makes sense; my knowledge of MPTCP
is somewhat limited).

>>> While implementing the scheduler in linux kernel, I was unable to get
>>> the current number of bytes in hardware queue for wlan interface. As
>>> currently, linux does not employ BQL for wifi devices, I could not get
>>> this value from dql->num_queued defined in dynamic_queue_limits.h. I
>>> also tried to get queue length from Queueing discipline structure i.e.
>>> qdisc->qstats.qlen defined in sch_generic.h yet it still gives me a
>>> zero value (I am getting zero values for other parameters in qstat as
>>> well so I am sure it is not because the queue length never becomes
>>> more than 0).
>>>
>>> If you have any idea for getting queue length for wireless interfaces,
>>> please do reply. Any help would be highly appreciated.
>>
>> There's no general interface to get the queue length for WiFi
>> interfaces. If you're using a driver that is using the mac80211
>> intermediate queues (i.e., ath9k, ath10k or mt76), there are
>> fq->backlock and fq->memory_usage which live inside the struct fq in
>> struct ieee80211_local. However, these are private state vars inside
>> mac80211, so not immediately accessible from other parts of the kernel.
>> You could add an access function to mac80211, though.
>
> I am using ath9k drivers so fq->backlog would work well for me. I was
> giving the mac80211 structure a look but cant seem to figure out an
> access point to it (say from netdevice.h). Do you have an idea of how
> to retrieve a struct in mac80211 that could point me to struct fq?

Yeah, that's what I meant with it being a private variable. What you can
do is add an accessor function somewhere in mac80211, which would take a
struct net_device and return the queue backlog.

Something like

u32 mac80211_get_backlog(struct net_device *dev)
{
	struct ieee80211_local *local = wdev_priv(dev->ieee80211_ptr);
        return local->fq->backlog;
}

> Thank you for your help.

You're welcome :)

-Toke

  reply	other threads:[~2017-05-16 10:17 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-05-15 15:40 Mohan, Nitinder
2017-05-15 17:55 ` Toke Høiland-Jørgensen
2017-05-16  9:33   ` Mohan, Nitinder
2017-05-16 10:17     ` Toke Høiland-Jørgensen [this message]
2017-05-16 10:45       ` Toke Høiland-Jørgensen

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

  List information: https://lists.bufferbloat.net/postorius/lists/make-wifi-fast.lists.bufferbloat.net/

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

  git send-email \
    --in-reply-to=878tlxhzr7.fsf@alrua-kau \
    --to=toke@toke.dk \
    --cc=bloat-devel@lists.bufferbloat.net \
    --cc=make-wifi-fast@lists.bufferbloat.net \
    --cc=nitinder.mohan@helsinki.fi \
    /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