linux-stable.git/net/sched, branch linux-3.6.y

pkt_sched: fix virtual-start-time update in QFQ

2012-09-19T20:23:53+00:00

If the old timestamps of a class, say cl, are stale when the class
becomes active, then QFQ may assign to cl a much higher start time
than the maximum value allowed. This may happen when QFQ assigns to
the start time of cl the finish time of a group whose classes are
characterized by a higher value of the ratio
max_class_pkt/weight_of_the_class with respect to that of
cl. Inserting a class with a too high start time into the bucket list
corrupts the data structure and may eventually lead to crashes.
This patch limits the maximum start time assigned to a class.

Signed-off-by: Paolo Valente 
Signed-off-by: David S. Miller

net_sched: gred: actually perform idling in WRED mode

2012-09-13T20:10:13+00:00

gred_dequeue() and gred_drop() do not seem to get called when the
queue is empty, meaning that we never start idling while in WRED
mode. And since qidlestart is not stored by gred_store_wred_set(),
we would never stop idling while in WRED mode if we ever started.
This messes up the average queue size calculation that influences
packet marking/dropping behavior.

Now, we start WRED mode idling as we are removing the last packet
from the queue. Also we now actually stop WRED mode idling when we
are enqueuing a packet.

Cc: Bruce Osler 
Signed-off-by: David Ward 
Acked-by: Jamal Hadi Salim 
Signed-off-by: David S. Miller

net_sched: gred: fix qave reporting via netlink

2012-09-13T20:10:13+00:00

q->vars.qavg is a Wlog scaled value, but q->backlog is not. In order
to pass q->vars.qavg as the backlog value, we need to un-scale it.
Additionally, the qave value returned via netlink should not be Wlog
scaled, so we need to un-scale the result of red_calc_qavg().

This caused artificially high values for "Average Queue" to be shown
by 'tc -s -d qdisc', but did not affect the actual operation of GRED.

Signed-off-by: David Ward 
Acked-by: Jamal Hadi Salim 
Signed-off-by: David S. Miller

net_sched: gred: eliminate redundant DP prio comparisons

2012-09-13T20:10:13+00:00

Each pair of DPs only needs to be compared once when searching for
a non-unique prio value.

Signed-off-by: David Ward 
Acked-by: Jamal Hadi Salim 
Signed-off-by: David S. Miller

net_sched: gred: correct comment about qavg calculation in RIO mode

2012-09-13T20:10:13+00:00

Signed-off-by: David Ward 
Acked-by: Jamal Hadi Salim 
Signed-off-by: David S. Miller

net-sched: sch_cbq: avoid infinite loop

2012-09-12T02:20:43+00:00

Its possible to setup a bad cbq configuration leading to
an infinite loop in cbq_classify()

DEV_OUT=eth0
ICMP="match ip protocol 1 0xff"
U32="protocol ip u32"
DST="match ip dst"
tc qdisc add dev $DEV_OUT root handle 1: cbq avpkt 1000 \
	bandwidth 100mbit
tc class add dev $DEV_OUT parent 1: classid 1:1 cbq \
	rate 512kbit allot 1500 prio 5 bounded isolated
tc filter add dev $DEV_OUT parent 1: prio 3 $U32 \
	$ICMP $DST 192.168.3.234 flowid 1:

Reported-by: Denys Fedoryschenko 
Tested-by: Denys Fedoryschenko 
Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

fq_codel: dont reinit flow state

2012-09-03T18:36:50+00:00

When fq_codel builds a new flow, it should not reset codel state.

Codel algo needs to get previous values (lastcount, drop_next) to get
proper behavior.

Signed-off-by: Dave Taht 
Signed-off-by: Eric Dumazet 
Acked-by: Dave Taht 
Signed-off-by: David S. Miller

act_mirred: do not drop packets when fails to mirror it

2012-08-16T21:54:44+00:00

We drop packet unconditionally when we fail to mirror it. This is not intended
in some cases. Consdier for kvm guest, we may mirror the traffic of the bridge
to a tap device used by a VM. When kernel fails to mirror the packet in
conditions such as when qemu crashes or stop polling the tap, it's hard for the
management software to detect such condition and clean the the mirroring
before. This would lead all packets to the bridge to be dropped and break the
netowrk of other virtual machines.

To solve the issue, the patch does not drop packets when kernel fails to mirror
it, and only drop the redirected packets.

Signed-off-by: Jason Wang 
Signed-off-by: Jamal Hadi Salim 
Signed-off-by: David S. Miller

sched: add missing group change to qfq_change_class

2012-08-08T23:02:05+00:00

[Resending again, as the text was corrupted by the email client]

To speed up operations, QFQ internally divides classes into
groups. Which group a class belongs to depends on the ratio between
the maximum packet length and the weight of the class. Unfortunately
the function qfq_change_class lacks the steps for changing the group
of a class when the ratio max_pkt_len/weight of the class changes.

For example, when the last of the following three commands is
executed, the group of class 1:1 is not correctly changed:

tc disc add dev XXX root handle 1: qfq
tc class add dev XXX parent 1: qfq classid 1:1 weight 1
tc class change dev XXX parent 1: classid 1:1 qfq weight 4

Not changing the group of a class does not affect the long-term
bandwidth guaranteed to the class, as the latter is independent of the
maximum packet length, and correctly changes (only) if the weight of
the class changes. In contrast, if the group of the class is not
updated, the class is still guaranteed the short-term bandwidth and
packet delay related to its old group, instead of the guarantees that
it should receive according to its new weight and/or maximum packet
length. This may also break service guarantees for other classes.
This patch adds the missing operations.

Signed-off-by: Paolo Valente 
Signed-off-by: David S. Miller

net_sched: act: Delete estimator in error path.

2012-08-06T20:30:01+00:00

Some action modules free struct tcf_common in their error path
while estimator is still active. This results in est_timer()
dereference freed memory.
Add gen_kill_estimator() in ipt, pedit and simple action.

Signed-off-by: Hiroaki SHIMODA 
Signed-off-by: David S. Miller