We currently have two levels of strict validation:
1) liberal (default)
- undefined (type >= max) & NLA_UNSPEC attributes accepted
- attribute length >= expected accepted
- garbage at end of message accepted
2) strict (opt-in)
- NLA_UNSPEC attributes accepted
- attribute length >= expected accepted
Split out parsing strictness into four different options:
* TRAILING - check that there's no trailing data after parsing
attributes (in message or nested)
* MAXTYPE - reject attrs > max known type
* UNSPEC - reject attributes with NLA_UNSPEC policy entries
* STRICT_ATTRS - strictly validate attribute size
The default for future things should be *everything*.
The current *_strict() is a combination of TRAILING and MAXTYPE,
and is renamed to _deprecated_strict().
The current regular parsing has none of this, and is renamed to
*_parse_deprecated().
Additionally it allows us to selectively set one of the new flags
even on old policies. Notably, the UNSPEC flag could be useful in
this case, since it can be arranged (by filling in the policy) to
not be an incompatible userspace ABI change, but would then going
forward prevent forgetting attribute entries. Similar can apply
to the POLICY flag.
We end up with the following renames:
* nla_parse -> nla_parse_deprecated
* nla_parse_strict -> nla_parse_deprecated_strict
* nlmsg_parse -> nlmsg_parse_deprecated
* nlmsg_parse_strict -> nlmsg_parse_deprecated_strict
* nla_parse_nested -> nla_parse_nested_deprecated
* nla_validate_nested -> nla_validate_nested_deprecated
Using spatch, of course:
@@
expression TB, MAX, HEAD, LEN, POL, EXT;
@@
-nla_parse(TB, MAX, HEAD, LEN, POL, EXT)
+nla_parse_deprecated(TB, MAX, HEAD, LEN, POL, EXT)
@@
expression NLH, HDRLEN, TB, MAX, POL, EXT;
@@
-nlmsg_parse(NLH, HDRLEN, TB, MAX, POL, EXT)
+nlmsg_parse_deprecated(NLH, HDRLEN, TB, MAX, POL, EXT)
@@
expression NLH, HDRLEN, TB, MAX, POL, EXT;
@@
-nlmsg_parse_strict(NLH, HDRLEN, TB, MAX, POL, EXT)
+nlmsg_parse_deprecated_strict(NLH, HDRLEN, TB, MAX, POL, EXT)
@@
expression TB, MAX, NLA, POL, EXT;
@@
-nla_parse_nested(TB, MAX, NLA, POL, EXT)
+nla_parse_nested_deprecated(TB, MAX, NLA, POL, EXT)
@@
expression START, MAX, POL, EXT;
@@
-nla_validate_nested(START, MAX, POL, EXT)
+nla_validate_nested_deprecated(START, MAX, POL, EXT)
@@
expression NLH, HDRLEN, MAX, POL, EXT;
@@
-nlmsg_validate(NLH, HDRLEN, MAX, POL, EXT)
+nlmsg_validate_deprecated(NLH, HDRLEN, MAX, POL, EXT)
For this patch, don't actually add the strict, non-renamed versions
yet so that it breaks compile if I get it wrong.
Also, while at it, make nla_validate and nla_parse go down to a
common __nla_validate_parse() function to avoid code duplication.
Ultimately, this allows us to have very strict validation for every
new caller of nla_parse()/nlmsg_parse() etc as re-introduced in the
next patch, while existing things will continue to work as is.
In effect then, this adds fully strict validation for any new command.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Even if the NLA_F_NESTED flag was introduced more than 11 years ago, most
netlink based interfaces (including recently added ones) are still not
setting it in kernel generated messages. Without the flag, message parsers
not aware of attribute semantics (e.g. wireshark dissector or libmnl's
mnl_nlmsg_fprintf()) cannot recognize nested attributes and won't display
the structure of their contents.
Unfortunately we cannot just add the flag everywhere as there may be
userspace applications which check nlattr::nla_type directly rather than
through a helper masking out the flags. Therefore the patch renames
nla_nest_start() to nla_nest_start_noflag() and introduces nla_nest_start()
as a wrapper adding NLA_F_NESTED. The calls which add NLA_F_NESTED manually
are rewritten to use nla_nest_start().
Except for changes in include/net/netlink.h, the patch was generated using
this semantic patch:
@@ expression E1, E2; @@
-nla_nest_start(E1, E2)
+nla_nest_start_noflag(E1, E2)
@@ expression E1, E2; @@
-nla_nest_start_noflag(E1, E2 | NLA_F_NESTED)
+nla_nest_start(E1, E2)
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The same code to flush qdisc tree and purge the qdisc queue
is duplicated in many places and in most cases it does not
respect NOLOCK qdisc: the global backlog len is used and the
per CPU values are ignored.
This change addresses the above, factoring-out the relevant
code and using the helpers introduced by the previous patch
to fetch the correct backlog len.
Fixes: c5ad119fb6 ("net: sched: pfifo_fast use skb_array")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Classful qdiscs can't access directly the child qdiscs backlog
length: if such qdisc is NOLOCK, per CPU values should be
accounted instead.
Most qdiscs no not respect the above. As a result, qstats fetching
for most classful qdisc is currently incorrect: if the child qdisc is
NOLOCK, it always reports 0 len backlog.
This change introduces a pair of helpers to safely fetch
both backlog and qlen and use them in stats class dumping
functions, fixing the above issue and cleaning a bit the code.
DRR needs also to access the child qdisc queue length, so it
needs custom handling.
Fixes: c5ad119fb6 ("net: sched: pfifo_fast use skb_array")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Several qdiscs check on enqueue whether the packet was enqueued to a class
with an empty queue, in which case the class is activated. This is done by
checking if the qlen is exactly 1 after enqueue. However, if GSO splitting
is enabled in the child qdisc, a single packet can result in a qlen longer
than 1. This means the activation check fails, leading to a stalled queue.
Fix this by checking if the queue is empty *before* enqueue, and running
the activation logic if this was the case.
Reported-by: Pete Heist <pete@heistp.net>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Parent qdiscs may dereference the pointer to the enqueued skb after
enqueue. However, both CAKE and TBF call consume_skb() on the original skb
when splitting GSO packets, leading to a potential use-after-free in the
parent. Fix this by avoiding dereferencing the skb pointer after enqueueing
to the child.
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current implementation of qdisc_destroy() decrements Qdisc reference
counter and only actually destroy Qdisc if reference counter value reached
zero. Rename qdisc_destroy() to qdisc_put() in order for it to better
describe the way in which this function currently implemented and used.
Extract code that deallocates Qdisc into new private qdisc_destroy()
function. It is intended to be shared between regular qdisc_put() and its
unlocked version that is introduced in next patch in this series.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for the function qdisc_create_dflt which is
a common used function in the tc subsystem. Callers which are interested
in the receiving error can assign extack to get a more detailed
information why qdisc_create_dflt failed. The function qdisc_create_dflt
will also call an init callback which can fail by any per-qdisc specific
handling.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for the function tcf_block_get which is
a common used function in the tc subsystem. Callers which are interested
in the receiving error can assign extack to get a more detailed
information why tcf_block_get failed.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for graft callback to prepare per-qdisc
specific changes for extack.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for block callback to prepare per-qdisc
specific changes for extack.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for class change callback api. This prepares
to handle extack support inside each specific class implementation.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds extack support for init callback to prepare per-qdisc
specific changes for extack.
Cc: David Ahern <dsahern@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prepare for removal of tp->q and store Qdisc pointer in the block
structure.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The commit 520ac30f45 ("net_sched: drop packets after root qdisc lock
is released) made a big change of tc for performance. There are two points
left in sch_prio and sch_qfq which are not changed with that commit. Now
enhance them now with __qdisc_drop.
Signed-off-by: Gao Feng <gfree.wind@vip.163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For TC classes, their ->get() and ->put() are always paired, and the
reference counting is completely useless, because:
1) For class modification and dumping paths, we already hold RTNL lock,
so all of these ->get(),->change(),->put() are atomic.
2) For filter bindiing/unbinding, we use other reference counter than
this one, and they should have RTNL lock too.
3) For ->qlen_notify(), it is special because it is called on ->enqueue()
path, but we already hold qdisc tree lock there, and we hold this
tree lock when graft or delete the class too, so it should not be gone
or changed until we release the tree lock.
Therefore, this patch removes ->get() and ->put(), but:
1) Adds a new ->find() to find the pointer to a class by classid, no
refcnt.
2) Move the original class destroy upon the last refcnt into ->delete(),
right after releasing tree lock. This is fine because the class is
already removed from hash when holding the lock.
For those who also use ->put() as ->unbind(), just rename them to reflect
this change.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This callback is used for deactivating class in parent qdisc.
This is cheaper to test queue length right here.
Also this allows to catch draining screwed backlog and prevent
second deactivation of already inactive parent class which will
crash kernel for sure. Kernel with print warning at destruction
of child qdisc where no packets but backlog is not zero.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is need to instruct the HW offloaded path to push certain matched
packets to cpu/kernel for further analysis. So this patch introduces a
new TRAP control action to TC.
For kernel datapath, this action does not make much sense. So with the
same logic as in HW, new TRAP behaves similar to STOLEN. The skb is just
dropped in the datapath (and virtually ejected to an upper level, which
does not exist in case of kernel).
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Yotam Gigi <yotamg@mellanox.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the filter chains are direcly put into the private structures
of qdiscs. In order to be able to have multiple chains per qdisc and to
allow filter chains sharing among qdiscs, there is a need for common
object that would hold the chains. This introduces such object and calls
it "tcf_block".
Helpers to get and put the blocks are provided to be called from
individual qdisc code. Also, the original filter_list pointers are left
in qdisc privs to allow the entry into tcf_block processing without any
added overhead of possible multiple pointer dereference on fast path.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move tc_classify function to cls_api.c where it belongs, rename it to
fit the namespace.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass the new extended ACK reporting struct to all of the generic
netlink parsing functions. For now, pass NULL in almost all callers
(except for some in the core.)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original reason [1] for having hidden qdiscs (potential scalability
issues in qdisc_match_from_root() with single linked list in case of large
amount of qdiscs) has been invalidated by 59cc1f61f0 ("net: sched: convert
qdisc linked list to hashtable").
This allows us for bringing more clarity and determinism into the dump by
making default pfifo qdiscs visible.
We're not turning this on by default though, at it was deemed [2] too
intrusive / unnecessary change of default behavior towards userspace.
Instead, TCA_DUMP_INVISIBLE netlink attribute is introduced, which allows
applications to request complete qdisc hierarchy dump, including the
ones that have always been implicit/invisible.
Singleton noop_qdisc stays invisible, as teaching the whole infrastructure
about singletons would require quite some surgery with very little gain
(seeing no qdisc or seeing noop qdisc in the dump is probably setting
the same user expectation).
[1] http://lkml.kernel.org/r/1460732328.10638.74.camel@edumazet-glaptop3.roam.corp.google.com
[2] http://lkml.kernel.org/r/20161021.105935.1907696543877061916.davem@davemloft.net
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) Old code was hard to maintain, due to complex lock chains.
(We probably will be able to remove some kfree_rcu() in callers)
2) Using a single timer to update all estimators does not scale.
3) Code was buggy on 32bit kernel (WRITE_ONCE() on 64bit quantity
is not supposed to work well)
In this rewrite :
- I removed the RB tree that had to be scanned in
gen_estimator_active(). qdisc dumps should be much faster.
- Each estimator has its own timer.
- Estimations are maintained in net_rate_estimator structure,
instead of dirtying the qdisc. Minor, but part of the simplification.
- Reading the estimator uses RCU and a seqcount to provide proper
support for 32bit kernels.
- We reduce memory need when estimators are not used, since
we store a pointer, instead of the bytes/packets counters.
- xt_rateest_mt() no longer has to grab a spinlock.
(In the future, xt_rateest_tg() could be switched to per cpu counters)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
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>
