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>
Some devices declare a high number of TX queues, then set a much
lower real_num_tx_queues
This cause setups using fq_codel, sfq or fq as the default qdisc to consume
more memory than really needed.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
qdisc_tree_decrease_qlen() suffers from two problems on multiqueue
devices.
One problem is that it updates sch->q.qlen and sch->qstats.drops
on the mq/mqprio root qdisc, while it should not : Daniele
reported underflows errors :
[ 681.774821] PAX: sch->q.qlen: 0 n: 1
[ 681.774825] PAX: size overflow detected in function qdisc_tree_decrease_qlen net/sched/sch_api.c:769 cicus.693_49 min, count: 72, decl: qlen; num: 0; context: sk_buff_head;
[ 681.774954] CPU: 2 PID: 19 Comm: ksoftirqd/2 Tainted: G O 4.2.6.201511282239-1-grsec #1
[ 681.774955] Hardware name: ASUSTeK COMPUTER INC. X302LJ/X302LJ, BIOS X302LJ.202 03/05/2015
[ 681.774956] ffffffffa9a04863 0000000000000000 0000000000000000 ffffffffa990ff7c
[ 681.774959] ffffc90000d3bc38 ffffffffa95d2810 0000000000000007 ffffffffa991002b
[ 681.774960] ffffc90000d3bc68 ffffffffa91a44f4 0000000000000001 0000000000000001
[ 681.774962] Call Trace:
[ 681.774967] [<ffffffffa95d2810>] dump_stack+0x4c/0x7f
[ 681.774970] [<ffffffffa91a44f4>] report_size_overflow+0x34/0x50
[ 681.774972] [<ffffffffa94d17e2>] qdisc_tree_decrease_qlen+0x152/0x160
[ 681.774976] [<ffffffffc02694b1>] fq_codel_dequeue+0x7b1/0x820 [sch_fq_codel]
[ 681.774978] [<ffffffffc02680a0>] ? qdisc_peek_dequeued+0xa0/0xa0 [sch_fq_codel]
[ 681.774980] [<ffffffffa94cd92d>] __qdisc_run+0x4d/0x1d0
[ 681.774983] [<ffffffffa949b2b2>] net_tx_action+0xc2/0x160
[ 681.774985] [<ffffffffa90664c1>] __do_softirq+0xf1/0x200
[ 681.774987] [<ffffffffa90665ee>] run_ksoftirqd+0x1e/0x30
[ 681.774989] [<ffffffffa90896b0>] smpboot_thread_fn+0x150/0x260
[ 681.774991] [<ffffffffa9089560>] ? sort_range+0x40/0x40
[ 681.774992] [<ffffffffa9085fe4>] kthread+0xe4/0x100
[ 681.774994] [<ffffffffa9085f00>] ? kthread_worker_fn+0x170/0x170
[ 681.774995] [<ffffffffa95d8d1e>] ret_from_fork+0x3e/0x70
mq/mqprio have their own ways to report qlen/drops by folding stats on
all their queues, with appropriate locking.
A second problem is that qdisc_tree_decrease_qlen() calls qdisc_lookup()
without proper locking : concurrent qdisc updates could corrupt the list
that qdisc_match_from_root() parses to find a qdisc given its handle.
Fix first problem adding a TCQ_F_NOPARENT qdisc flag that
qdisc_tree_decrease_qlen() can use to abort its tree traversal,
as soon as it meets a mq/mqprio qdisc children.
Second problem can be fixed by RCU protection.
Qdisc are already freed after RCU grace period, so qdisc_list_add() and
qdisc_list_del() simply have to use appropriate rcu list variants.
A future patch will add a per struct netdev_queue list anchor, so that
qdisc_tree_decrease_qlen() can have more efficient lookups.
Reported-by: Daniele Fucini <dfucini@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Cong Wang <cwang@twopensource.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.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>
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>
Commit 6da7c8fcbc ("qdisc: allow setting default queuing discipline")
added the ability to change default qdisc from pfifo_fast to say fq
But as most modern ethernet devices are multiqueue, we cant really
see all the statistics from "tc -s qdisc show", as the default root
qdisc is mq.
This patch adds the calls to qdisc_list_add() to mq and mqprio
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
By default, the pfifo_fast queue discipline has been used by default
for all devices. But we have better choices now.
This patch allow setting the default queueing discipline with sysctl.
This allows easy use of better queueing disciplines on all devices
without having to use tc qdisc scripts. It is intended to allow
an easy path for distributions to make fq_codel or sfq the default
qdisc.
This patch also makes pfifo_fast more of a first class qdisc, since
it is now possible to manually override the default and explicitly
use pfifo_fast. The behavior for systems who do not use the sysctl
is unchanged, they still get pfifo_fast
Also removes leftover random # in sysctl net core.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With BQL being deployed, we can more likely have following behavior :
We dequeue a packet from qdisc in dequeue_skb(), then we realize target
tx queue is in XOFF state in sch_direct_xmit(), and we have to hold the
skb into gso_skb for later.
This shows in stats (tc -s qdisc dev eth0) as requeues.
Problem of these requeues is that high priority packets can not be
dequeued as long as this (possibly low prio and big TSO packet) is not
removed from gso_skb.
At 1Gbps speed, a full size TSO packet is 500 us of extra latency.
In some cases, we know that all packets dequeued from a qdisc are
for a particular and known txq :
- If device is non multi queue
- For all MQ/MQPRIO slave qdiscs
This patch introduces a new qdisc flag, TCQ_F_ONETXQUEUE to mark
this capability, so that dequeue_skb() is allowed to dequeue a packet
only if the associated txq is not stopped.
This indeed reduce latencies for high prio packets (or improve fairness
with sfq/fq_codel), and almost remove qdisc 'requeues'.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These files are non modular, but need to export symbols using
the macros now living in export.h -- call out the include so
that things won't break when we remove the implicit presence
of module.h from everywhere.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Now qdisc stab is handled before TCQ_F_CAN_BYPASS test in
__dev_xmit_skb(), we can generalize TCQ_F_CAN_BYPASS to other qdiscs
than pfifo_fast : pfifo, bfifo, pfifo_head_drop and sfq
SFQ is special because it can have external classifiers, and in these
cases, we cannot bypass queue discipline (packet could be dropped by
classifier) without admin asking it, or further changes.
Its worth doing this, especially for SFQ, avoiding dirtying memory in
case no packets are already waiting in queue.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The first parameter dev isn't in use in qdisc_create_dflt().
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes from net/ (but not any netfilter files)
all the unnecessary return; statements that precede the
last closing brace of void functions.
It does not remove the returns that are immediately
preceded by a label as gcc doesn't like that.
Done via:
$ grep -rP --include=*.[ch] -l "return;\n}" net/ | \
xargs perl -i -e 'local $/ ; while (<>) { s/\n[ \t\n]+return;\n}/\n}/g; print; }'
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Some classful qdiscs miss qstats.qlen updating with q.qlen of their
child qdiscs in dump_stats methods.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After the recent mq change there is the new select_queue qdisc class
method used in tc_modify_qdisc, but it works OK only for direct child
qdiscs of mq qdisc. Grandchildren always get the first tx queue, which
would give wrong qdisc_root etc. results (e.g. for sch_htb as child of
sch_prio). This patch fixes it by using parent's dev_queue for such
grandchildren qdiscs. The select_queue method's return type is changed
BTW.
With feedback from: Patrick McHardy <kaber@trash.net>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When new child qdiscs are attached to the mq qdisc, they are actually
attached as root qdiscs to the device queues. The lock selection for
new estimators incorrectly picks the root lock of the existing and
to be replaced qdisc, which results in a use-after-free once the old
qdisc has been destroyed.
Mark mq qdisc instances with a new flag and treat qdiscs attached to
mq as children similar to regular root qdiscs.
Additionally prevent estimators from being attached to the mq qdisc
itself since it only updates its byte and packet counters during dumps.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a classful dummy scheduler which can be used as root qdisc
for multiqueue devices and exposes each device queue as a child class.
This allows to address queues individually and graft them similar to regular
classes. Additionally it presents an accumulated view of the statistics of
all real root qdiscs in the dummy root.
Two new callbacks are added to the qdisc_ops and qdisc_class_ops:
- cl_ops->select_queue selects the tx queue number for new child classes.
- qdisc_ops->attach() overrides root qdisc device grafting to attach
non-shared qdiscs to the queues.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet