Reported-by: Stas Nichiporovich <stasn77@gmail.com>
Fixes: 2ccccf5fb4 ("net_sched: update hierarchical backlog too")
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Qdisc performance suffers when packets are dropped at enqueue()
time because drops (kfree_skb()) are done while qdisc lock is held,
delaying a dequeue() draining the queue.
Nominal throughput can be reduced by 50 % when this happens,
at a time we would like the dequeue() to proceed as fast as possible.
Even FQ is vulnerable to this problem, while one of FQ goals was
to provide some flow isolation.
This patch adds a 'struct sk_buff **to_free' parameter to all
qdisc->enqueue(), and in qdisc_drop() helper.
I measured a performance increase of up to 12 %, but this patch
is a prereq so that future batches in enqueue() can fly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
net/sched/act_police.c
net/sched/sch_drr.c
net/sched/sch_hfsc.c
net/sched/sch_prio.c
net/sched/sch_red.c
net/sched/sch_tbf.c
In net-next the drop methods of the packet schedulers got removed, so
the bug fixes to them in 'net' are irrelevant.
A packet action unload crash fix conflicts with the addition of the
new firstuse timestamp.
Signed-off-by: David S. Miller <davem@davemloft.net>
after removal of TCA_CBQ_OVL_STRATEGY from cbq scheduler, there are no
more callers of ->drop() outside of other ->drop functions, i.e.
nothing calls them.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we need to create a new aggregate to enqueue the skb we call kzalloc.
If that fails we returned ENOBUFS without freeing the skb.
Spotted during code review.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Large tc dumps (tc -s {qdisc|class} sh dev ethX) done by Google BwE host
agent [1] are problematic at scale :
For each qdisc/class found in the dump, we currently lock the root qdisc
spinlock in order to get stats. Sampling stats every 5 seconds from
thousands of HTB classes is a challenge when the root qdisc spinlock is
under high pressure. Not only the dumps take time, they also slow
down the fast path (queue/dequeue packets) by 10 % to 20 % in some cases.
An audit of existing qdiscs showed that sch_fq_codel is the only qdisc
that might need the qdisc lock in fq_codel_dump_stats() and
fq_codel_dump_class_stats()
In v2 of this patch, I now use the Qdisc running seqcount to provide
consistent reads of packets/bytes counters, regardless of 32/64 bit arches.
I also changed rate estimators to use the same infrastructure
so that they no longer need to lock root qdisc lock.
[1]
http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/43838.pdf
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Kevin Athey <kda@google.com>
Cc: Xiaotian Pei <xiaotian@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the bottom qdisc decides to, for example, drop some packet,
it calls qdisc_tree_decrease_qlen() to update the queue length
for all its ancestors, we need to update the backlog too to
keep the stats on root qdisc accurate.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove nearly duplicated code and prepare for the following patch.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For classifiers getting invoked via tc_classify(), we always need an
extra function call into tc_classify_compat(), as both are being
exported as symbols and tc_classify() itself doesn't do much except
handling of reclassifications when tp->classify() returned with
TC_ACT_RECLASSIFY.
CBQ and ATM are the only qdiscs that directly call into tc_classify_compat(),
all others use tc_classify(). When tc actions are being configured
out in the kernel, tc_classify() effectively does nothing besides
delegating.
We could spare this layer and consolidate both functions. pktgen on
single CPU constantly pushing skbs directly into the netif_receive_skb()
path with a dummy classifier on ingress qdisc attached, improves
slightly from 22.3Mpps to 23.1Mpps.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The member (u32) "num_active_agg" of struct qfq_sched has been unused
since its introduction in 462dbc9101
"pkt_sched: QFQ Plus: fair-queueing service at DRR cost" and (AFAICT)
there is no active plan to use it; this removes the member.
Signed-off-by: Andrea Parri <parri.andrea@gmail.com>
Acked-by: Paolo Valente <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
The control !hlist_unhashed() in qfq_destroy_agg() is unnecessary
because already performed in hlist_del_init(), so remove it.
Signed-off-by: Andrea Parri <parri.andrea@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After previous patches to simplify qstats the qstats can be
made per cpu with a packed union in Qdisc struct.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This removes the use of qstats->qlen variable from the classifiers
and makes it an explicit argument to gnet_stats_copy_queue().
The qlen represents the qdisc queue length and is packed into
the qstats at the last moment before passnig to user space. By
handling it explicitely we avoid, in the percpu stats case, having
to figure out which per_cpu variable to put it in.
It would probably be best to remove it from qstats completely
but qstats is a user space ABI and can't be broken. A future
patch could make an internal only qstats structure that would
avoid having to allocate an additional u32 variable on the
Qdisc struct. This would make the qstats struct 128bits instead
of 128+32.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds helpers to manipulate qstats logic and replaces locations
that touch the counters directly. This simplifies future patches
to push qstats onto per cpu counters.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to run qdisc's without locking statistics and estimators
need to be handled correctly.
To resolve bstats make the statistics per cpu. And because this is
only needed for qdiscs that are running without locks which is not
the case for most qdiscs in the near future only create percpu
stats when qdiscs set the TCQ_F_CPUSTATS flag.
Next because estimators use the bstats to calculate packets per
second and bytes per second the estimator code paths are updated
to use the per cpu statistics.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rcu'ify tcf_proto this allows calling tc_classify() without holding
any locks. Updaters are protected by RTNL.
This patch prepares the core net_sched infrastracture for running
the classifier/action chains without holding the qdisc lock however
it does nothing to ensure cls_xxx and act_xxx types also work without
locking. Additional patches are required to address the fall out.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
QFQ+ inherits from QFQ a design choice that may cause a high packet
delay/jitter and a severe short-term unfairness. As QFQ, QFQ+ uses a
special quantity, the system virtual time, to track the service
provided by the ideal system it approximates. When a packet is
dequeued, this quantity must be incremented by the size of the packet,
divided by the sum of the weights of the aggregates waiting to be
served. Tracking this sum correctly is a non-trivial task, because, to
preserve tight service guarantees, the decrement of this sum must be
delayed in a special way [1]: this sum can be decremented only after
that its value would decrease also in the ideal system approximated by
QFQ+. For efficiency, QFQ+ keeps track only of the 'instantaneous'
weight sum, increased and decreased immediately as the weight of an
aggregate changes, and as an aggregate is created or destroyed (which,
in its turn, happens as a consequence of some class being
created/destroyed/changed). However, to avoid the problems caused to
service guarantees by these immediate decreases, QFQ+ increments the
system virtual time using the maximum value allowed for the weight
sum, 2^10, in place of the dynamic, instantaneous value. The
instantaneous value of the weight sum is used only to check whether a
request of weight increase or a class creation can be satisfied.
Unfortunately, the problems caused by this choice are worse than the
temporary degradation of the service guarantees that may occur, when a
class is changed or destroyed, if the instantaneous value of the
weight sum was used to update the system virtual time. In fact, the
fraction of the link bandwidth guaranteed by QFQ+ to each aggregate is
equal to the ratio between the weight of the aggregate and the sum of
the weights of the competing aggregates. The packet delay guaranteed
to the aggregate is instead inversely proportional to the guaranteed
bandwidth. By using the maximum possible value, and not the actual
value of the weight sum, QFQ+ provides each aggregate with the worst
possible service guarantees, and not with service guarantees related
to the actual set of competing aggregates. To see the consequences of
this fact, consider the following simple example.
Suppose that only the following aggregates are backlogged, i.e., that
only the classes in the following aggregates have packets to transmit:
one aggregate with weight 10, say A, and ten aggregates with weight 1,
say B1, B2, ..., B10. In particular, suppose that these aggregates are
always backlogged. Given the weight distribution, the smoothest and
fairest service order would be:
A B1 A B2 A B3 A B4 A B5 A B6 A B7 A B8 A B9 A B10 A B1 A B2 ...
QFQ+ would provide exactly this optimal service if it used the actual
value for the weight sum instead of the maximum possible value, i.e.,
11 instead of 2^10. In contrast, since QFQ+ uses the latter value, it
serves aggregates as follows (easy to prove and to reproduce
experimentally):
A B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 A A A A A A A A A A B1 B2 ... B10 A A ...
By replacing 10 with N in the above example, and by increasing N, one
can increase at will the maximum packet delay and the jitter
experienced by the classes in aggregate A.
This patch addresses this issue by just using the above
'instantaneous' value of the weight sum, instead of the maximum
possible value, when updating the system virtual time. After the
instantaneous weight sum is decreased, QFQ+ may deviate from the ideal
service for a time interval in the order of the time to serve one
maximum-size packet for each backlogged class. The worst-case extent
of the deviation exhibited by QFQ+ during this time interval [1] is
basically the same as of the deviation described above (but, without
this patch, QFQ+ suffers from such a deviation all the time). Finally,
this patch modifies the comment to the function qfq_slot_insert, to
make it coherent with the fact that the weight sum used by QFQ+ can
now be lower than the maximum possible value.
[1] P. Valente, "Extending WF2Q+ to support a dynamic traffic mix",
Proceedings of AAA-IDEA'05, June 2005.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes the forward declaration of qfq_update_agg_ts, by moving
the definition of the function above its first call. This patch also
removes a useless forward declaration of qfq_schedule_agg.
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
In make_eligible, a mask is used to decide which groups must become eligible:
the i-th group becomes eligible only if the i-th bit of the mask (from the
right) is set. The mask is computed by left-shifting a 1 by a given number of
places, and decrementing the result. The shift is performed on a ULL to avoid
problems in case the number of places to shift is higher than 31. On a 32-bit
machine, this is more costly than working on an UL. This patch replaces such a
costly operation with two cheaper branches.
The trick is based on the following fact: in case of a shift of at least 32
places, the resulting mask has at least the 32 less significant bits set,
whereas the total number of groups is lower than 32. As a consequence, in this
case it is enough to just set the 32 less significant bits of the mask with a
cheaper ~0UL. In the other case, the shift can be safely performed on a UL.
Reported-by: David S. Miller <davem@davemloft.net>
Reported-by: David Laight <David.Laight@ACULAB.COM>
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct gnet_stats_rate_est contains u32 fields, so the bytes per second
field can wrap at 34360Mbit.
Add a new gnet_stats_rate_est64 structure to get 64bit bps/pps fields,
and switch the kernel to use this structure natively.
This structure is dumped to user space as a new attribute :
TCA_STATS_RATE_EST64
Old tc command will now display the capped bps (to 34360Mbit), instead
of wrapped values, and updated tc command will display correct
information.
Old tc command output, after patch :
eric:~# tc -s -d qd sh dev lo
qdisc pfifo 8001: root refcnt 2 limit 1000p
Sent 80868245400 bytes 1978837 pkt (dropped 0, overlimits 0 requeues 0)
rate 34360Mbit 189696pps backlog 0b 0p requeues 0
This patch carefully reorganizes "struct Qdisc" layout to get optimal
performance on SMP.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
QFQ+ can select for service only 'eligible' aggregates, i.e.,
aggregates that would have started to be served also in the emulated
ideal system. As a consequence, for QFQ+ to be work conserving, at
least one of the active aggregates must be eligible when it is time to
choose the next aggregate to serve.
The set of eligible aggregates is updated through the function
qfq_update_eligible(), which does guarantee that, after its
invocation, at least one of the active aggregates is eligible.
Because of this property, this function is invoked in
qfq_deactivate_agg() to guarantee that at least one of the active
aggregates is still eligible after an aggregate has been deactivated.
In particular, the critical case is when there are other active
aggregates, but the aggregate being deactivated happens to be the only
one eligible.
However, this precaution is not needed for QFQ+ to be work conserving,
because update_eligible() is always invoked also at the beginning of
qfq_choose_next_agg(). This patch removes the additional invocation of
update_eligible() in qfq_deactivate_agg().
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By definition of (the algorithm of) QFQ+, the system virtual time must
be pushed up only if there is no 'eligible' aggregate, i.e. no
aggregate that would have started to be served also in the ideal
system emulated by QFQ+. QFQ+ serves only eligible aggregates, hence
the aggregate currently in service is eligible. As a consequence, to
decide whether there is no eligible aggregate, QFQ+ must also check
whether there is no aggregate in service.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Aggregate budgets are computed so as to guarantee that, after an
aggregate has been selected for service, that aggregate has enough
budget to serve at least one maximum-size packet for the classes it
contains. For this reason, after a new aggregate has been selected
for service, its next packet is immediately dequeued, without any
further control.
The maximum packet size for a class, lmax, can be changed through
qfq_change_class(). In case the user sets lmax to a lower value than
the the size of some of the still-to-arrive packets, QFQ+ will
automatically push up lmax as it enqueues these packets. This
automatic push up is likely to happen with TSO/GSO.
In any case, if lmax is assigned a lower value than the size of some
of the packets already enqueued for the class, then the following
problem may occur: the size of the next packet to dequeue for the
class may happen to be larger than lmax, after the aggregate to which
the class belongs has been just selected for service. In this case,
even the budget of the aggregate, which is an unsigned value, may be
lower than the size of the next packet to dequeue. After dequeueing
this packet and subtracting its size from the budget, the latter would
wrap around.
This fix prevents the budget from wrapping around after any packet
dequeue.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If no aggregate is in service, then the function qfq_dequeue() does
not dequeue any packet. For this reason, to guarantee QFQ+ to be work
conserving, a just-activated aggregate must be set as in service
immediately if it happens to be the only active aggregate.
This is done by the function qfq_enqueue().
Unfortunately, the function qfq_add_to_agg(), used to add a class to
an aggregate, does not perform this important additional operation.
In particular, if: 1) qfq_add_to_agg() is invoked to complete the move
of a class from a source aggregate, becoming, for this move, inactive,
to a destination aggregate, becoming instead active, and 2) the
destination aggregate becomes the only active aggregate, then this
aggregate is not however set as in service. QFQ+ remains then in a
non-work-conserving state until a new invocation of qfq_enqueue()
recovers the situation.
This fix solves the problem by moving the logic for setting an
aggregate as in service directly into the function qfq_activate_agg().
Hence, from whatever point qfq_activate_aggregate() is invoked, QFQ+
remains work conserving. Since the more-complex logic of this new
version of activate_aggregate() is not necessary, in qfq_dequeue(), to
reschedule an aggregate that finishes its budget, then the aggregate
is now rescheduled by invoking directly the functions needed.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Between two invocations of make_eligible, the system virtual time may
happen to grow enough that, in its binary representation, a bit with
higher order than 31 flips. This happens especially with
TSO/GSO. Before this fix, the mask used in make_eligible was computed
as (1UL<<index_of_last_flipped_bit)-1, whose value is well defined on
a 64-bit architecture, because index_of_flipped_bit <= 63, but is in
general undefined on a 32-bit architecture if index_of_flipped_bit > 31.
The fix just replaces 1UL with 1ULL.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
QFQ+ schedules the active aggregates in a group using a bucket list
(one list per group). The bucket in which each aggregate is inserted
depends on the aggregate's timestamps, and the number
of buckets in a group is enough to accomodate the possible (range of)
values of the timestamps of all the aggregates in the group. For this
property to hold, timestamps must however be computed correctly. One
necessary condition for computing timestamps correctly is that the
number of bits dequeued for each aggregate, while the aggregate is in
service, does not exceed the maximum budget budgetmax assigned to the
aggregate.
For each aggregate, budgetmax is proportional to the number of classes
in the aggregate. If the number of classes of the aggregate is
decreased through qfq_change_class(), then budgetmax is decreased
automatically as well. Problems may occur if the aggregate is in
service when budgetmax is decreased, because the current remaining
budget of the aggregate and/or the service already received by the
aggregate may happen to be larger than the new value of budgetmax. In
this case, when the aggregate is eventually deselected and its
timestamps are updated, the aggregate may happen to have received an
amount of service larger than budgetmax. This may cause the aggregate
to be assigned a higher virtual finish time than the maximum
acceptable value for the last bucket in the bucket list of the group.
This fix introduces a cap that addresses this issue.
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Reviewed-by: Fabio Checconi <fchecconi@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch turns QFQ into QFQ+, a variant of QFQ that provides the
following two benefits: 1) QFQ+ is faster than QFQ, 2) differently
from QFQ, QFQ+ correctly schedules also non-leaves classes in a
hierarchical setting. A detailed description of QFQ+, plus a
performance comparison with DRR and QFQ, can be found in [1].
[1] P. Valente, "Reducing the Execution Time of Fair-Queueing Schedulers"
http://algo.ing.unimo.it/people/paolo/agg-sched/agg-sched.pdf
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the max packet size for some class (configured through tc) is
violated by the actual size of the packets of that class, then QFQ
would not schedule classes correctly, and the data structures
implementing the bucket lists may get corrupted. This problem occurs
with TSO/GSO even if the max packet size is set to the MTU, and is,
e.g., the cause of the failure reported in [1]. Two patches have been
proposed to solve this problem in [2], one of them is a preliminary
version of this patch.
This patch addresses the above issues by: 1) setting QFQ parameters to
proper values for supporting TSO/GSO (in particular, setting the
maximum possible packet size to 64KB), 2) automatically increasing the
max packet size for a class, lmax, when a packet with a larger size
than the current value of lmax arrives.
The drawback of the first point is that the maximum weight for a class
is now limited to 4096, which is equal to 1/16 of the maximum weight
sum.
Finally, this patch also forcibly caps the timestamps of a class if
they are too high to be stored in the bucket list. This capping, taken
from QFQ+ [3], handles the unfrequent case described in the comment to
the function slot_insert.
[1] http://marc.info/?l=linux-netdev&m=134968777902077&w=2
[2] http://marc.info/?l=linux-netdev&m=135096573507936&w=2
[3] http://marc.info/?l=linux-netdev&m=134902691421670&w=2
Signed-off-by: Paolo Valente <paolo.valente@unimore.it>
Tested-by: Cong Wang <amwang@redhat.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/team/team.c
drivers/net/usb/qmi_wwan.c
net/batman-adv/bat_iv_ogm.c
net/ipv4/fib_frontend.c
net/ipv4/route.c
net/l2tp/l2tp_netlink.c
The team, fib_frontend, route, and l2tp_netlink conflicts were simply
overlapping changes.
qmi_wwan and bat_iv_ogm were of the "use HEAD" variety.
With help from Antonio Quartulli.
Signed-off-by: David S. Miller <davem@davemloft.net>
GCC refuses to recognize that all error control flows do in fact
set err to something.
Add an explicit initialization to shut it up.
net/sched/sch_drr.c: In function ‘drr_enqueue’:
net/sched/sch_drr.c:359:11: warning: ‘err’ may be used uninitialized in this function [-Wmaybe-uninitialized]
net/sched/sch_qfq.c: In function ‘qfq_enqueue’:
net/sched/sch_qfq.c:885:11: warning: ‘err’ may be used uninitialized in this function [-Wmaybe-uninitialized]
Signed-off-by: David S. Miller <davem@davemloft.net>
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 <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
[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 <paolo.valente@unimore.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
These macros contain a hidden goto, and are thus extremely error
prone and make code hard to audit.
Signed-off-by: David S. Miller <davem@davemloft.net>
We can underestimate q->wsum in case of "tc class replace ... qfq"
and/or qdisc_create_dflt() error.
wsum is not really used in fast path, only at qfq qdisc/class setup,
to catch user error.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
grp->slot_shift is between 22 and 41, so using 32bit wide variables is
probably a typo.
This could explain QFQ hangs Dave reported to me, after 2^23 packets ?
(23 = 64 - 41)
Reported-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
CC: Dave Taht <dave.taht@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is an implementation of the Quick Fair Queue scheduler developed
by Fabio Checconi. The same algorithm is already implemented in ipfw
in FreeBSD. Fabio had an earlier version developed on Linux, I just
cleaned it up. Thanks to Eric Dumazet for testing this under load.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
