The net-next tree has this commit (which cherry-picks just fine on top
of 4.0, just like all the rhine crash fixes that follow it in the tree):

commit 92bf200881d978bc3c6a290991ae1f9ddc7b5411
Author: Tino Reichardt <milky-ker...@mcmilk.de>
Date:   Tue Feb 24 10:28:01 2015 -0800

    net: via-rhine: add BQL support

    Add Byte Queue Limits (BQL) support to via-rhine driver.

    [edumazet] tweaked patch and changed TX_RING_SIZE from 16 to 64

    Signed-off-by: Tino Reichardt <milky-ker...@mcmilk.de>
    Tested-by: Jamie Gloudon <jamie.glou...@gmail.com>
    Signed-off-by: Eric Dumazet <eduma...@google.com>
    Signed-off-by: David S. Miller <da...@davemloft.net>

This means that the codel and fq_codel queueing disciplines will
*finally* work! One addition to /etc/sysctl.conf later (and a bit of
manual fiddling to change the qdisc on my existing interfaces without
having to reboot) and I have:

[...]
4: adsl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state 
UNKNOWN group default qlen 10
5: bdsl: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state 
UNKNOWN group default qlen 10

[...]

qdisc fq_codel 0: dev adsl root refcnt 2 limit 10240p flows 1024 quantum 1514 
target 5.0ms interval 100.0ms ecn
qdisc fq_codel 0: dev bdsl root refcnt 2 limit 10240p flows 1024 quantum 1514 
target 5.0ms interval 100.0ms ecn

I can verify that even without any other traffic shapers to restrict the
rate of outgoing packets to the speed of the ADSL modem, even without
any other configuration at all, all my bufferbloat symptoms have gone :)
and my ADSL line was fairly bufferbloated and generally queue-damaged,
enough that I could barely use the line for other purposes once I
saturated it with something, up or down: ping times were 5s or higher,
with massive packet loss. Now, ping times are up on a saturated line,
yes -- they're up about, oh, 6ms, more or less the 5ms target above.
Actual noticeable difference: none, even doing remote ssh. (It might be
noticeable doing remote gaming if I had cyborg reflexes, I suppose.)

-- 
NULL && (void)
_______________________________________________
Soekris-tech mailing list
Soekris-tech@lists.soekris.com
http://lists.soekris.com/mailman/listinfo/soekris-tech

Reply via email to