dev_queue_xmit enqueue's a skb and calls qdisc_run which
dequeue's the skb and xmits it. In most cases, the skb that
is enqueue'd is the same one that is dequeue'd (unless the
queue gets stopped or multiple cpu's write to the same queue
and ends in a race with qdisc_run). For default qdiscs, we
can remove the redundant enqueue/dequeue and simply xmit the
skb since the default qdisc is work-conserving.
The patch uses a new flag - TCQ_F_CAN_BYPASS to identify the
default fast queue. The controversial part of the patch is
incrementing qlen when a skb is requeued - this is to avoid
checks like the second line below:
+ } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
>> !q->gso_skb &&
+ !test_and_set_bit(__QDISC_STATE_RUNNING, &q->state)) {
Results of a 2 hour testing for multiple netperf sessions (1,
2, 4, 8, 12 sessions on a 4 cpu system-X). The BW numbers are
aggregate Mb/s across iterations tested with this version on
System-X boxes with Chelsio 10gbps cards:
----------------------------------
Size | ORG BW NEW BW |
----------------------------------
128K | 156964 159381 |
256K | 158650 162042 |
----------------------------------
Changes from ver1:
1. Move sch_direct_xmit declaration from sch_generic.h to
pkt_sched.h
2. Update qdisc basic statistics for direct xmit path.
3. Set qlen to zero in qdisc_reset.
4. Changed some function names to more meaningful ones.
Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is the result of an automatic spatch transformation to convert
all ndo_start_xmit() return values of 0 to NETDEV_TX_OK.
Some occurences are missed by the automatic conversion, those will be
handled in a seperate patch.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 2b85a34e91
(net: No more expensive sock_hold()/sock_put() on each tx)
changed initial sk_wmem_alloc value.
We need to take into account this offset when reporting
sk_wmem_alloc to user, in PROC_FS files or various
ioctls (SIOCOUTQ/TIOCOUTQ)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Action police statistics could be misleading because drops are not
shown when expected.
With feedback from: Jamal Hadi Salim <hadi@cyberus.ca>
Reported-by: Pawel Staszewski <pstaszewski@itcare.pl>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let's use TICKS instead of US, so PSCHED_TICKS2NS and PSCHED_NS2TICKS
(like in PSCHED_TICKS_PER_SEC already) to avoid misleading.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert magic values 1 and -1 to NETDEV_TX_BUSY and NETDEV_TX_LOCKED respectively.
0 (NETDEV_TX_OK) is not changed to keep the noise down, except in very few cases
where its in direct proximity to one of the other values.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use PSCHED_SHIFT constant instead of '10' in PSCHED_US2NS() and
PSCHED_NS2US() macros to enable changing this value later.
Additionally use PSCHED_SHIFT in sch_hfsc SM_SHIFT and ISM_SHIFT
definitions. This part of the patch is based on feedback from
Patrick McHardy <kaber@trash.net>.
Reported-by: Antonio Almeida <vexwek@gmail.com>
Tested-by: Antonio Almeida <vexwek@gmail.com>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I found a bug in cls_cgroup_change() in cls_cgroup.c.
cls_cgroup_change() expected tca[TCA_OPTIONS] was set from user space properly,
but tc in iproute2-2.6.29-1 (which I used) didn't set it.
In the current source code of tc in git, it set tca[TCA_OPTIONS].
git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git
If we always use a newest iproute2 in git when we use cls_cgroup,
we don't face this oops probably.
But I think, kernel shouldn't panic regardless of use program's behaviour.
Signed-off-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define three accessors to get/set dst attached to a skb
struct dst_entry *skb_dst(const struct sk_buff *skb)
void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
void skb_dst_drop(struct sk_buff *skb)
This one should replace occurrences of :
dst_release(skb->dst)
skb->dst = NULL;
Delete skb->dst field
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define skb_rtable(const struct sk_buff *skb) accessor to get rtable from skb
Delete skb->rtable field
Setting rtable is not allowed, just set dst instead as rtable is an alias.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a bug which unconfigured struct tcf_proto keeps
chaining in tc_ctl_tfilter(), and avoids kernel panic in
cls_cgroup_classify() when we use cls_cgroup.
When we execute 'tc filter add', tcf_proto is allocated, initialized
by classifier's init(), and chained. After it's chained,
tc_ctl_tfilter() calls classifier's change(). When classifier's
change() fails, tc_ctl_tfilter() does not free and keeps tcf_proto.
In addition, cls_cgroup is initialized in change() not in init(). It
accesses unconfigured struct tcf_proto which is chained before
change(), then hits Oops.
Signed-off-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Tested-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid reading the unsynchronized value cs->classid multiple times,
since it could change concurrently from non-zero to zero; this would
result in the classifier returning a positive result with a bogus
(zero) classid.
Signed-off-by: Paul Menage <menage@google.com>
Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We would like to get rid of netdev->trans_start = jiffies; that about all net
drivers have to use in their start_xmit() function, and use txq->trans_start
instead.
This can be done generically in core network, as suggested by David.
Some devices, (particularly loopback) dont need trans_start update, because
they dont have transmit watchdog. We could add a new device flag, or rely
on fact that txq->tran_start can be updated is txq->xmit_lock_owner is
different than -1. Use a helper function to hide our choice.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can slightly reduce size of teqlN structure, not duplicating stats
structure in teql_master but using stats field from net_device.stats
for tx_errors and from netdev_queue for tx_bytes/tx_packets/tx_dropped
values.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is illegal to dereference a skb after a successful ndo_start_xmit()
call. We must store skb length in a local variable instead.
Bug was introduced in 2.6.27 by commit 0abf77e55a
(net_sched: Add accessor function for packet length for qdiscs)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct net_device trans_start field is a hot spot on SMP and high performance
devices, particularly multi queues ones, because every transmitter dirties
it. Is main use is tx watchdog and bonding alive checks.
But as most devices dont use NETIF_F_LLTX, we have to lock
a netdev_queue before calling their ndo_start_xmit(). So it makes
sense to move trans_start from net_device to netdev_queue. Its update
will occur on a already present (and in exclusive state) cache line, for
free.
We can do this transition smoothly. An old driver continue to
update dev->trans_start, while an updated one updates txq->trans_start.
Further patches could also put tx_bytes/tx_packets counters in
netdev_queue to avoid dirtying dev->stats (vlan device comes to mind)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can remove this lock here, since we are in cgroup write handler and
thus the cgrp is guaranteed to be valid, and no lock is needed when
writing a u32 variable.
Signed-off-by: Li Zefan <lizf@cn.fujitsuc.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When no limit is given, the bfifo uses a default of tx_queue_len * mtu.
Packets handled by qdiscs include the link layer header, so this should
be taken into account, similar to what other qdiscs do.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The kernel should only be using the high 16 bits of a kernel
generated priority. Filter priorities in all other cases only
use the upper 16 bits of the u32 'prio' field of 'struct tcf_proto',
but when the kernel generates the priority of a filter is saves all
32 bits which can result in incorrect lookup failures when a filter
needs to be deleted or modified.
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Alex Sidorenko reported:
"while experimenting with 'netem' we have found some strange behaviour. It
seemed that ingress delay as measured by 'ping' command shows up on some
hosts but not on others.
After some investigation I have found that the problem is that skbuff->tstamp
field value depends on whether there are any packet sniffers enabled. That
is:
- if any ptype_all handler is registered, the tstamp field is as expected
- if there are no ptype_all handlers, the tstamp field does not show the delay"
This patch prevents unnecessary update of tstamp in dev_queue_xmit_nit()
on ingress path (with act_mirred) adding a check, so minimal overhead on
the fast path, but only when sniffers etc. are active.
Since netem at ingress seems to logically emulate a network before a host,
tstamp is zeroed to trigger the update and pretend delays are from the
outside.
Reported-by: Alex Sidorenko <alexandre.sidorenko@hp.com>
Tested-by: Alex Sidorenko <alexandre.sidorenko@hp.com>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When vlan acceleration is used on receive, the vlan tag is maintained
outside of the skb data. The existing vlan tag match only works on TX
path because it uses vlan_get_tag which tests for VLAN_HW_TX_ACCEL.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_sack_swap seems unnecessary so I pushed swap to the caller.
Also removed comment that seemed then pointless, and added include
when not already there. Compile tested.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
While looking for a possible reason of bugzilla report on HTB oops:
http://bugzilla.kernel.org/show_bug.cgi?id=12858
I found the code in htb_delete calling htb_destroy_class on zero
refcount is very misleading: it can suggest this is a common path, and
destroy is called under sch_tree_lock. Actually, this can never happen
like this because before deletion cops->get() is done, and after
delete a class is still used by tclass_notify. The class destroy is
always called from cops->put(), so without sch_tree_lock.
This doesn't mean much now (since 2.6.27) because all vulnerable calls
were moved from htb_destroy_class to htb_delete, but there was a bug
in older kernels. The same change is done for other classful scheds,
which, it seems, didn't have similar locking problems here.
Reported-by: m0sia <m0sia@m0sia.ru>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A commit c1b56878fb "tc: policing requires
a rate estimator" introduced a test which invalidates previously working
configs, based on examples from iproute2: doc/actions/actions-general.
This is too rigorous: a rate estimator is needed only when police's
"avrate" option is used.
Reported-by: Joao Correia <joaomiguelcorreia@gmail.com>
Diagnosed-by: John Dykstra <john.dykstra1@gmail.com>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drr_change_class lacks a check for NULL of tca[TCA_OPTIONS], so oops
is possible.
Reported-by: Denys Fedoryschenko <denys@visp.net.lb>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current "RTNETLINK answers: Invalid argument" warning, while trying to
add multiq qdisc to non-multiqueue device, isn't very helpful and some
of these devs can be changed btw., so let's use a better errno.
With feedback from Stephen Hemminger <shemminger@vyatta.com>
Reported-by: Badalian Vyacheslav <slavon@bigtelecom.ru>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy <kaber@trash.net> suggested using a workqueue instead
of hrtimers to trigger netif_schedule() when there is a problem with
setting exact time of this event: 'The differnce - yeah, it shouldn't
make much, mainly wake up the qdisc earlier (but not too early) after
"too many events" occured _and_ no further enqueue events wake up the
qdisc anyways.'
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let's get some info on possible config problems. This patch brings
back an old warning, but is printed only once now.
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>
Patrick McHardy <kaber@trash.net> suggested:
> How about making this flag and the warning message (in a out-of-line
> function) globally available? Other qdiscs (f.i. HFSC) can't deal with
> inner non-work-conserving qdiscs as well.
This patch uses qdisc->flags field of "suspected" child qdisc.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently htb_do_events() breaks events recounting for a level after 2
jiffies, but there is no reason to repeat this for next levels and
increase delays even more (with softirqs disabled). htb_dequeue_tree()
can add to this too, btw. In such a case q->now time is invalid anyway.
Thanks to Patrick McHardy for spotting an error around earlier version
of this patch.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Next event time should consider jiffies used for recounting. Otherwise
qdisc_watchdog_schedule() triggers hrtimer immediately with the event
in the past, and may cause very high ksoftirqd cpu usage (if highres
is on).
There is also removed checking "event" for zero in htb_dequeue(): it's
always true in this place.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (84 commits)
wimax: fix kernel-doc for debufs_dentry member of struct wimax_dev
net: convert pegasus driver to net_device_ops
bnx2x: Prevent eeprom set when driver is down
net: switch kaweth driver to netdevops
pcnet32: round off carrier watch timer
i2400m/usb: wrap USB power saving in #ifdef CONFIG_PM
wimax: testing for rfkill support should also test for CONFIG_RFKILL_MODULE
wimax: fix kconfig interactions with rfkill and input layers
wimax: fix '#ifndef CONFIG_BUG' layout to avoid warning
r6040: bump release number to 0.20
r6040: warn about MAC address being unset
r6040: check PHY status when bringing interface up
r6040: make printks consistent with DRV_NAME
gianfar: Fixup use of BUS_ID_SIZE
mlx4_en: Returning real Max in get_ringparam
mlx4_en: Consider inline packets on completion
netdev: bfin_mac: enable bfin_mac net dev driver for BF51x
qeth: convert to net_device_ops
vlan: add neigh_setup
dm9601: warn on invalid mac address
...
New nodes are inserted in u32_change() under rtnl_lock() with wmb(),
so without tcf_tree_lock() like in other classifiers (e.g. cls_fw).
This isn't enough without rmb() on the read side, but on the other
hand adding such barriers doesn't give any savings, so the lock is
added instead.
Reported-by: m0sia <m0sia@plotinka.ru>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 22604c8668.
We can't fix this issue in this way, because we now can try
to take the dev_base_lock rwlock as a writer in software interrupt
context and that is not allowed without major surgery elsewhere.
This initial link state problem needs to be solved in some other
way.
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Michael Marineau <mike@marineau.org>
Commit b47300168e "Do not fire linkwatch
events until the device is registered." was made as a workaround for
drivers that call netif_carrier_off before registering the device.
Unfortunately this causes these drivers to incorrectly report their
link status as IF_OPER_UNKNOWN which can falsely set the IFF_RUNNING
flag when the interface is first brought up. This issues was
previously pointed out[1] but was dismissed saying that IFF_RUNNING is
not related to the link status. From my digging IFF_RUNNING, as
reported to userspace, is based on the link state. It is set based on
__LINK_STATE_START and IF_OPER_UP or IF_OPER_UNKNOWN. See [2], [3],
and [4]. (Whether or not the kernel has IFF_RUNNING set in flags is
not reported to user space so it may well be independent of the link,
I don't know if and when it may get set.)
The end result depends slightly depending on the driver. The the two I
tested were e1000e and b44. With e1000e if the system is booted
without a network cable attached the interface will falsely report
RUNNING when it is brought up causing NetworkManager to attempt to
start it and eventually time out. With b44 when the system is booted
with a network cable attached and brought up with dhcpcd it will time
out the first time.
The attached patch that will still set the operstate variable
correctly to IF_OPER_UP/DOWN/etc when linkwatch_fire_event is called
but then return rather than skipping the linkwatch_fire_event call
entirely as the previous fix did. (sorry it isn't inline, I don't have
a patch friendly email client at the moment)
Signed-off-by: David S. Miller <davem@davemloft.net>
cls_cgroup can't be compiled as a module, since it's not supported by
cgroup.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- It's better to use container_of() instead of casting cgroup_subsys_state *
to cgroup_cls_state *.
- Add helper function task_cls_state().
- Rename net_cls_state() to cgrp_cls_state().
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When removing a cgroup, an oops was triggered immediately. The cause
is wrong kfree() in cgrp_destroy().
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1429 commits)
net: Allow dependancies of FDDI & Tokenring to be modular.
igb: Fix build warning when DCA is disabled.
net: Fix warning fallout from recent NAPI interface changes.
gro: Fix potential use after free
sfc: If AN is enabled, always read speed/duplex from the AN advertising bits
sfc: When disabling the NIC, close the device rather than unregistering it
sfc: SFT9001: Add cable diagnostics
sfc: Add support for multiple PHY self-tests
sfc: Merge top-level functions for self-tests
sfc: Clean up PHY mode management in loopback self-test
sfc: Fix unreliable link detection in some loopback modes
sfc: Generate unique names for per-NIC workqueues
802.3ad: use standard ethhdr instead of ad_header
802.3ad: generalize out mac address initializer
802.3ad: initialize ports LACPDU from const initializer
802.3ad: remove typedef around ad_system
802.3ad: turn ports is_individual into a bool
802.3ad: turn ports is_enabled into a bool
802.3ad: make ntt bool
ixgbe: Fix set_ringparam in ixgbe to use the same memory pools.
...
Fixed trivial IPv4/6 address printing conflicts in fs/cifs/connect.c due
to the conversion to %pI (in this networking merge) and the addition of
doing IPv6 addresses (from the earlier merge of CIFS).
While implementing a TCQ_F_THROTTLED flag there was used an smp_wmb()
in qdisc_watchdog(), but since this flag is practically used only in
sch_netem(), and since it's not even clear what reordering is avoided
here (TCQ_F_THROTTLED vs. __QDISC_STATE_SCHED?) it seems the barrier
could be safely removed.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some gcc versions warn that ret may be used uninitialized in
sfq_enqueue(). It's a false positive, so let's annotate this.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
