2012-05-29 21:14:34

by Ben Greear

Subject: Is there no upper limit on mac80211 pending xmit queue size?

The tx pending queues doesn't seem to be limitted, allowing
run-away memory usage when something like pktgen is
sending packets faster than the system can actually
send them.

[root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
00: 0x00000000/0
01: 0x00000000/0
02: 0x00000001/40076
03: 0x00000000/0
[root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
00: 0x00000000/0
01: 0x00000000/0
02: 0x00000001/52649
03: 0x00000000/0
[root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
00: 0x00000000/0
01: 0x00000000/0
02: 0x00000001/107939
03: 0x00000000/0

I think this is what is causing my OOM issues when sending
with (my hacked version of) pktgen.

Shouldn't there be an upper limit on the number of pending frames
that mac80211 will queue?

Thanks,
Ben


--
Ben Greear <[email protected]>
Candela Technologies Inc http://www.candelatech.com

2012-05-29 21:37:06

by Dave Taht

Subject: Re: Is there no upper limit on mac80211 pending xmit queue size?

On Tue, May 29, 2012 at 2:14 PM, Ben Greear <[email protected]> wrote:
> The tx pending queues doesn't seem to be limitted, allowing
> run-away memory usage when something like pktgen is
> sending packets faster than the system can actually
> send them.
>
> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
> 00: 0x00000000/0
> 01: 0x00000000/0
> 02: 0x00000001/40076
> 03: 0x00000000/0
> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
> 00: 0x00000000/0
> 01: 0x00000000/0
> 02: 0x00000001/52649
> 03: 0x00000000/0
> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
> 00: 0x00000000/0
> 01: 0x00000000/0
> 02: 0x00000001/107939
> 03: 0x00000000/0
>
> I think this is what is causing my OOM issues when sending
> with (my hacked version of) pktgen.
>
> Shouldn't there be an upper limit on the number of pending frames
> that mac80211 will queue?

something bql-like would help a lot.

>
> Thanks,
> Ben
>
>
> --
> Ben Greear <[email protected]>
> Candela Technologies Inc ?http://www.candelatech.com
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to [email protected]
> More majordomo info at ?http://vger.kernel.org/majordomo-info.html



--
Dave T?ht
SKYPE: davetaht
US Tel: 1-239-829-5608
http://www.bufferbloat.net

2012-05-29 21:41:04

by Ben Greear

Subject: Re: Is there no upper limit on mac80211 pending xmit queue size?

On 05/29/2012 02:37 PM, Dave Taht wrote:
> On Tue, May 29, 2012 at 2:14 PM, Ben Greear<[email protected]> wrote:
>> The tx pending queues doesn't seem to be limitted, allowing
>> run-away memory usage when something like pktgen is
>> sending packets faster than the system can actually
>> send them.
>>
>> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
>> 00: 0x00000000/0
>> 01: 0x00000000/0
>> 02: 0x00000001/40076
>> 03: 0x00000000/0
>> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
>> 00: 0x00000000/0
>> 01: 0x00000000/0
>> 02: 0x00000001/52649
>> 03: 0x00000000/0
>> [root@ct520-2011 ~]# cat /debug/ieee80211/wiphy0/queues
>> 00: 0x00000000/0
>> 01: 0x00000000/0
>> 02: 0x00000001/107939
>> 03: 0x00000000/0
>>
>> I think this is what is causing my OOM issues when sending
>> with (my hacked version of) pktgen.
>>
>> Shouldn't there be an upper limit on the number of pending frames
>> that mac80211 will queue?
>
> something bql-like would help a lot.

Might help more righteous protocols, but I'm not sure it will help
with pktgen. I'm working on an upper limit implemented as a module
parameter. Hopefully that will at least keep the system from
going OOM.

Thanks,
Ben

--
Ben Greear <[email protected]>
Candela Technologies Inc http://www.candelatech.com