netdev - Re: [PATCH 3/3] mac80211: Set lower memory limit for non-VHT devices

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 30 Sep 2016 14:41:42 +0200
From:   Toke Høiland-Jørgensen <toke@...e.dk>
To:     Johannes Berg <johannes@...solutions.net>
Cc:     make-wifi-fast@...ts.bufferbloat.net,
        linux-wireless@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: [PATCH 3/3] mac80211: Set lower memory limit for non-VHT devices

Johannes Berg <johannes@...solutions.net> writes:

> On Fri, 2016-09-23 at 21:59 +0200, Toke Høiland-Jørgensen wrote:
>> Small devices can run out of memory from queueing too many packets.
>> If VHT is not supported by the PHY, having more than 4 MBytes of
>> total queue in the TXQ intermediate queues is not needed, and so we
>> can safely limit the memory usage in these cases and avoid OOM.
>
> I kinda see the logic here - we really don't need to queue as much if
> we can't possibly transmit it out quickly - but it seems to me we
> should also throw in some kind of limit that's relative to the amount
> of memory you have on the system?

Yes, ideally. That goes for FQ-CoDel as well, BTW. LEDE currently
carries a patch for that which just changes the hard-coded default to
another hard-coded default. Not sure how to get a good value to use,
though; and deciding on how large a fraction of memory to use for
packets starts smelling an awful lot like setting policy in the kernel,
doesn't it?

> I've applied these anyway though. I just don't like your assumption (b)
> much as the rationale for it.

Right, thanks. I'll come up with a better rationale next time ;)

-Toke