netem checks PSCHED_TLESS(cb->time_to_send, now) to find out whether it is
allowed to send a packet, which is equivalent to cb->time_to_send < now.
Use !PSCHED_TLESS(now, cb->time_to_send) instead to properly handle
cb->time_to_send == now.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of some of my creative spelling.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If possible, avoid having to do a transmit softirq when a qdisc
watchdog decides to re-enable. The watchdog routine runs off
a timer, so it is already in the same effective context as
the softirq.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The netem code would call getnstimeofday() and dequeue/requeue after
every packet, even if it was waiting. Avoid this overhead by using
the throttled flag.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In most cases, the next packet will be sent after the
last one. So optimize that case.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The random number generator always generates 32 bit values.
The time values are limited by psched_tdiff_t
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If you setup netem to just delay packets; "tc qdisc ls" will report
the reordering as 100%. Well it's a lie, reorder isn't used unless
gap is set, so just set value to 0 so the output of utility
is correct.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
So that it is also an offset from skb->head, reduces its size from 8 to 4 bytes
on 64bit architectures, allowing us to combine the 4 bytes hole left by the
layer headers conversion, reducing struct sk_buff size to 256 bytes, i.e. 4
64byte cachelines, and since the sk_buff slab cache is SLAB_HWCACHE_ALIGN...
:-)
Many calculations that previously required that skb->{transport,network,
mac}_header be first converted to a pointer now can be done directly, being
meaningful as offsets or pointers.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/sched/sch_api.c: In function 'psched_show':
net/sched/sch_api.c:1219: warning: format '%08x' expects type 'unsigned int', but argument 6 has type 's64'
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
q->now is increased during dequeue and doesn't contain the current time
afterwards, resulting in a too large timeout value for the qdisc watchdog.
Use "now" instead, which still contains the current time.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The timer resolution exported in /proc/net/psched is used by userspace to
calculate HTB's burst values. Currently it is set to HZ, since we're now
using hrtimers, use KTIME_MONOTONIC_RES, which makes HTB use smaller burst
values.
This patch also affects libnl, which incorrectly uses this value for
the SFQ perturbation parameter, which is always in seconds, and some
routing cache values, which are in USER_HZ, so both cases are broken
anyway.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that all packet schedulers have been converted to hrtimers most users
of PSCHED_JIFFIE2US and PSCHED_US2JIFFIE are gone. The remaining users use
it to convert external time units to packet scheduler clock ticks, so use
PSCHED_TICKS_PER_SEC instead.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Switch delay_timer to hrtimer.
The class penalty parameter is changed to use psched ticks as units.
Since iproute never supported using this and the only existing user
(libnl) incorrectly assumes psched ticks as units anyway, this
shouldn't break anything.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
cbq_undelay_prio is supposed to return a time delta, but returns the
current time for non-active priorities, causing cbq_undelay to mark
the priority as active and schedule a timer for twice the current
time.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of the manual clock source selection mess and use ktime. Also
use a scalar representation, which allows to clean up pkt_sched.h a bit
more and results in less ktime_to_ns() calls in most cases.
The PSCHED_US2JIFFIE/PSCHED_JIFFIE2US macros are implemented quite
inefficient by this patch, following patches will convert all qdiscs
to hrtimers and get rid of them entirely.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now the skb->nh union has just one member, .raw, i.e. it is just like the
skb->mac union, strange, no? I'm just leaving it like that till the transport
layer is done with, when we'll rename skb->mac.raw to skb->mac_header (or
->mac_header_offset?), ditto for ->{h,nh}.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the places where we need a pointer to the network header, it is still legal
to touch skb->nh.raw directly if just adding to, subtracting from or setting it
to another layer header.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the quite common 'skb->nh.raw - skb->data' sequence.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Userspace uses an integer for TCA_TCINDEX_SHIFT, the kernel was changed
to expect and use a u16 value in 2.6.11, which broke compatibility on
big endian machines. Change back to use int.
Reported by Ole Reinartz <ole.reinartz@gmx.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The input_device pointer is not refcounted, which means the device may
disappear while packets are queued, causing a crash when ifb passes packets
with a stale skb->dev pointer to netif_rx().
Fix by storing the interface index instead and do a lookup where neccessary.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
During both HTB and HFSC class deletion the class is removed from the
class hash before calling qdisc_tree_decrease_qlen. This makes the
->get operation in qdisc_tree_decrease_qlen fail, so it passes a NULL
pointer to ->qlen_notify, causing an oops.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the worthless net/sched/Makefile entry for the non-existent
source file sch_hpfq.c.
Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
cls_basic doesn't allocate tp->root before it is linked into the
active classifier list, resulting in a NULL pointer dereference
when packets hit the classifier before its ->change function is
called.
Reported by Chris Madden <chris@reflexsecurity.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Turning up the warnings on gcc makes it emit warnings
about the placement of 'inline' in function declarations.
Here's everything that was under net/
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After Al Viro (finally) succeeded in removing the sched.h #include in module.h
recently, it makes sense again to remove other superfluous sched.h includes.
There are quite a lot of files which include it but don't actually need
anything defined in there. Presumably these includes were once needed for
macros that used to live in sched.h, but moved to other header files in the
course of cleaning it up.
To ease the pain, this time I did not fiddle with any header files and only
removed #includes from .c-files, which tend to cause less trouble.
Compile tested against 2.6.20-rc2 and 2.6.20-rc2-mm2 (with offsets) on alpha,
arm, i386, ia64, mips, powerpc, and x86_64 with allnoconfig, defconfig,
allmodconfig, and allyesconfig as well as a few randconfigs on x86_64 and all
configs in arch/arm/configs on arm. I also checked that no new warnings were
introduced by the patch (actually, some warnings are removed that were emitted
by unnecessarily included header files).
Signed-off-by: Tim Schmielau <tim@physik3.uni-rostock.de>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Many struct file_operations in the kernel can be "const". Marking them const
moves these to the .rodata section, which avoids false sharing with potential
dirty data. In addition it'll catch accidental writes at compile time to
these shared resources.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch introduces users of the round_jiffies() function in the
networking code.
These timers all were of the "about once a second" or "about once
every X seconds" variety and several showed up in the "what wakes the
cpu up" profiles that the tickless patches provide. Some timers are
highly dynamic based on network load; but even on low activity systems
they still show up so the rounding is done only in cases of low
activity, allowing higher frequency timers in the high activity case.
The various hardware watchdogs are an obvious case; they run every 2
seconds but aren't otherwise specific of exactly when they need to
run.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a dump_stats callback to enable
printing of basic statistics of prio classes.
(With help of Patrick McHardy).
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The x_tables patch broke target module autoloading in the ipt action
by replacing the ipt_find_target call (which does autoloading) by
xt_find_target (which doesn't do autoloading). Additionally xt_find_target
may return ERR_PTR values in case of an error, which are not handled.
Use xt_request_find_target, which does both autoloading and ERR_PTR
handling properly. Also don't forget to drop the target module reference
again when xt_check_target fails.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
- turn intermediate classes into leaves again when their
last child is deleted (struct htb_class changed)
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
- deactivating of active classes when q.qlen drops to zero
(cbq_drop)
- a redundant instruction removed from cbq_deactivate_class
PS: probably htb_deactivate in htb_delete and
cbq_deactivate_class in cbq_delete are also
redundant now.
Signed-off-by: Jarek Poplawski <jarkao2@o2.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the first fw classifier is initialized, there is a small window
between the ->init() and ->change() calls, during which the classifier
is active but not entirely set up and tp->root is still NULL (->init()
does nothing).
When a packet is queued during this window a NULL pointer dereference
occurs in fw_classify() when trying to dereference head->mask;
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Not returning -EINVAL, because someone might want to use the value
zero in some future gact_prob algorithm?
Signed-off-by: Kim Nordlund <kim.nordlund@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The tc actions increased the size of struct tc_police, which broke
compatibility with old iproute binaries since both the act_police
and the old NET_CLS_POLICE code check for an exact size match.
Since the new members are not even used, the simple fix is to also
accept the size of the old structure. Dumping is not affected since
old userspace will receive a bigger structure, which is handled fine.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes the following unused EXPORT_SYMBOL's:
- sch_api.c: qdisc_lookup
- sch_generic.c: __netdev_watchdog_up
- sch_generic.c: noop_qdisc_ops
- sch_generic.c: qdisc_alloc
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
When peeking at the next packet in a child qdisc by calling dequeue/requeue,
the upper qdisc qlen counter may get out of sync in case the requeue fails.
The qdisc and the child qdisc both have their counter decremented, but since
no packet is given to the upper qdisc it won't decrement its counter itself.
requeue should not fail, so this is mostly for "correctness".
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert HTB to use qdisc_tree_decrease_len() and add a callback
for deactivating a class when its child queue becomes empty.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert HFSC to use qdisc_tree_decrease_len() and add a callback
for deactivating a class when its child queue becomes empty.
All queue purging goes through hfsc_purge_queue(), which is used in
three cases: grafting, class creation (when a leaf class is turned
into an intermediate class by attaching a new class) and class
deletion. In all cases qdisc_tree_decrease_len() is needed.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert the "simple" qdiscs to use qdisc_tree_decrease_qlen() where
necessary:
- all graft operations
- destruction of old child qdiscs in prio, red and tbf change operation
- purging of queue in sfq change operation
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are multiple problems related to qlen adjustment that can lead
to an upper qdisc getting out of sync with the real number of packets
queued, leading to endless dequeueing attempts by the upper layer code.
All qdiscs must maintain an accurate q.qlen counter. There are basically
two groups of operations affecting the qlen: operations that propagate
down the tree (enqueue, dequeue, requeue, drop, reset) beginning at the
root qdisc and operations only affecting a subtree or single qdisc
(change, graft, delete class). Since qlen changes during operations from
the second group don't propagate to ancestor qdiscs, their qlen values
become desynchronized.
This patch adds a function to propagate qlen changes up the qdisc tree,
optionally calling a callback function to perform qdisc-internal
maintenance when the child qdisc becomes empty. The follow-up patches
will convert all qdiscs to use this function where necessary.
Noticed by Timo Steinbach <tsteinbach@astaro.com>.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set parent classids in default qdiscs to allow walking up the tree
from outside the qdiscs. This is needed by the next patch.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
qlen adjustment should happen immediately in ->delete and not in the
class destroy function because the reference count will not hit zero in
->delete (sch_api holds a reference) but in ->put. Since the qdisc
lock is released between deletion of the class and final destruction
this creates an externally visible error in the qlen counter.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on patch by Patrick McHardy.
Add a new option, NET_SCH_FIFO, which provides a simple fifo qdisc
without requiring CONFIG_NET_SCHED.
The d80211 stack needs a generic fifo qdisc for WME. At present it
uses net/d80211/fifo_qdisc.c which is functionally equivalent to
sch_fifo.c. This patch will allow the d80211 stack to remove
net/d80211/fifo_qdisc.c and use sch_fifo.c instead.
Signed-off-by: David Kimdon <david.kimdon@devicescape.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
nfmark is being used in various subsystems and has become
the defacto mark field for all kinds of packets. Therefore
it makes sense to rename it to `mark' and remove the
dependency on CONFIG_NETFILTER.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Otherwise we can hit paths that (legally) do multiple deletes on the
same node and OOPS with the HLIST poison values there instead of
NULL.
Signed-off-by: David S. Miller <davem@davemloft.net>
I don't want my code to downgraded to GPLv3 because of
cut-n-pasted the comments. These files which I hold copyright
on were started before it was clear what GPLv3 was going to be.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The networking emulator can queue SKBs for a very long
time, so if you're using netem on the sender side for
large bandwidth/delay product testing, the SKB socket
send queue sizes become artificially larger.
Correct this by calling skb_orphan() in netem_enqueue().
Signed-off-by: David S. Miller <davem@davemloft.net>
Use rb_first() to get first entry in rb tree.
Signed-off-by: Akinbou Mita <akinobu.mita@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove unused file, estimators live in net/core/gen_estimator.c now.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
With commit 10fd48f237 [1] , RB_EMPTY_NODE
changed behaviour so it returns true when the node is empty as expected.
Hence Patrick McHardy's fix for sched_htb.c should be reverted.
Signed-off-by: Ismail Donmez <ismail@pardus.org.tr>
ACKed-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The move of qdisc destruction to a rcu callback broke locking in the
entire qdisc layer by invalidating previously valid assumptions about
the context in which changes to the qdisc tree occur.
The two assumptions were:
- since changes only happen in process context, read_lock doesn't need
bottem half protection. Now invalid since destruction of inner qdiscs,
classifiers, actions and estimators happens in the RCU callback unless
they're manually deleted, resulting in dead-locks when read_lock in
process context is interrupted by write_lock_bh in bottem half context.
- since changes only happen under the RTNL, no additional locking is
necessary for data not used during packet processing (f.e. u32_list).
Again, since destruction now happens in the RCU callback, this assumption
is not valid anymore, causing races while using this data, which can
result in corruption or use-after-free.
Instead of "fixing" this by disabling bottem halfs everywhere and adding
new locks/refcounting, this patch makes these assumptions valid again by
moving destruction back to process context. Since only the dev->qdisc
pointer is protected by RCU, but ->enqueue and the qdisc tree are still
protected by dev->qdisc_lock, destruction of the tree can be performed
immediately and only the final free needs to happen in the rcu callback
to make sure dev_queue_xmit doesn't access already freed memory.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix incorrect use of RB_EMPTY_NODE in htb_safe_rb_erase, which makes it
skip nodes within the rbtree instead of nodes not in the tree, resulting
in crashes later on.
The root cause for this seems to be the very counter-intuitive behaviour
of the RB_EMPTY_NODE macro, which returns _false_ when the node is empty.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prevents filters from being added if the first generated
handle already exists.
Signed-off-by: Kim Nordlund <kim.nordlund@nokia.com>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
This patch makes the needlessly global struct simp_hash_info static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Support masking the nfmark value before the search. The mask value is
global for all filters contained in one instance. It can only be set
when a new instance is created, all filters must specify the same mask.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The size is verified by x_tables and isn't needed by the modules anymore.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This was simply making templates of functions and mostly causing a lot
of code duplication in the classifier action modules.
We solve this more cleanly by having a common "struct tcf_common" that
hash worker functions contained once in act_api.c can work with.
Callers work with real action objects that have the common struct
plus their module specific struct members. You go from a common
object to the higher level one using a "to_foo()" macro which makes
use of container_of() to do the dirty work.
This also kills off act_generic.h which was only used by act_simple.c
and keeping it around was more work than the it's value.
Signed-off-by: David S. Miller <davem@davemloft.net>
Add code to initialize rb tree nodes, and check for double deletion.
This is not a real fix, but I can make it trap sometimes and may
be a bandaid for: http://bugzilla.kernel.org/show_bug.cgi?id=6681
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use hlist instead of list for the hash list. This saves
space, and we can check for double delete better.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Code was a mess in terms of indentation. Run through Lindent
script, and cleanup the damage. Also, don't use, vim magic
comment, and substitute inline for __inline__.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the conditional compilation around HTB_HYSTERSIS
since code was splitting mid expression.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of the macro's being used to obscure the locking.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The HTB network scheduler had debug code that wouldn't compile
and confused and obfuscated the code, remove it.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace CHECKSUM_HW by CHECKSUM_PARTIAL (for outgoing packets, whose
checksum still needs to be completed) and CHECKSUM_COMPLETE (for
incoming packets, device supplied full checksum).
Patch originally from Herbert Xu, updated by myself for 2.6.18-rc3.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix lockdep warning with GRE, iptables and Speedtouch ADSL, PPP over ATM.
On Sat, Sep 02, 2006 at 08:39:28PM +0000, Krzysztof Halasa wrote:
>
> =======================================================
> [ INFO: possible circular locking dependency detected ]
> -------------------------------------------------------
> swapper/0 is trying to acquire lock:
> (&dev->queue_lock){-+..}, at: [<c02c8c46>] dev_queue_xmit+0x56/0x290
>
> but task is already holding lock:
> (&dev->_xmit_lock){-+..}, at: [<c02c8e14>] dev_queue_xmit+0x224/0x290
>
> which lock already depends on the new lock.
This turns out to be a genuine bug. The queue lock and xmit lock are
intentionally taken out of order. Two things are supposed to prevent
dead-locks from occuring:
1) When we hold the queue_lock we're supposed to only do try_lock on the
tx_lock.
2) We always drop the queue_lock after taking the tx_lock and before doing
anything else.
>
> the existing dependency chain (in reverse order) is:
>
> -> #1 (&dev->_xmit_lock){-+..}:
> [<c012e7b6>] lock_acquire+0x76/0xa0
> [<c0336241>] _spin_lock_bh+0x31/0x40
> [<c02d25a9>] dev_activate+0x69/0x120
This path obviously breaks assumption 1) and therefore can lead to ABBA
dead-locks.
I've looked at the history and there seems to be no reason for the lock
to be held at all in dev_watchdog_up. The lock appeared in day one and
even there it was unnecessary. In fact, people added __dev_watchdog_up
precisely in order to get around the tx lock there.
The function dev_watchdog_up is already serialised by rtnl_lock since
its only caller dev_activate is always called under it.
So here is a simple patch to remove the tx lock from dev_watchdog_up.
In 2.6.19 we can eliminate the unnecessary __dev_watchdog_up and
replace it with dev_watchdog_up.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
CONFIG_DEBUG_SLAB found the following bug:
netem_enqueue() in sch_netem.c gets a pointer inside a slab object:
struct netem_skb_cb *cb = (struct netem_skb_cb *)skb->cb;
But then, the slab object may be freed:
skb = skb_unshare(skb, GFP_ATOMIC)
cb is still pointing inside the freed skb, so here is a patch to
initialize cb later, and make it clear that initializing it sooner
is a bad idea.
[From Stephen Hemminger: leave cb unitialized in order to let gcc
complain in case of use before initialization]
Signed-off-by: Guillaume Chazarain <guichaz@yahoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/sched/sch_htb.c: In function 'htb_change_class':
net/sched/sch_htb.c:1605: error: expected ';' before 'do_gettimeofday'
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The upper bound for HTB time diff needs to be scaled to PSCHED
units rather than just assuming usecs. The field mbuffer is used
in TDIFF_SAFE(), as an upper bound.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Module reference needs to be given back if message header
construction fails.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
"return -err" and blindly inheriting the error code in the netlink
failure exception handler causes errors codes to be returned as
positive value therefore making them being ignored by the caller.
May lead to sending out incomplete netlink messages.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The TCA_ACT_KIND attribute is used without checking its
availability when dumping actions therefore leading to a
value of 0x4 being dereferenced.
The use of strcmp() in tc_lookup_action_n() isn't safe
when fed with string from an attribute without enforcing
proper NUL termination.
Both bugs can be triggered with malformed netlink message
and don't require any privileges.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the infrastructure for generic segmentation offload.
The idea is to tap into the potential savings of TSO without hardware
support by postponing the allocation of segmented skb's until just
before the entry point into the NIC driver.
The same structure can be used to support software IPv6 TSO, as well as
UFO and segmentation offload for other relevant protocols, e.g., DCCP.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The dev_deactivate function has bit-rotted since the introduction of
lockless drivers. In particular, the spin_unlock_wait call at the end
has no effect on the xmit routine of lockless drivers.
With a little bit of work, we can make it much more useful by providing
the guarantee that when it returns, no more calls to the xmit routine
of the underlying driver will be made.
The idea is simple. There are two entry points in to the xmit routine.
The first comes from dev_queue_xmit. That one is easily stopped by
using synchronize_rcu. This works because we set the qdisc to noop_qdisc
before the synchronize_rcu call. That in turn causes all subsequent
packets sent to dev_queue_xmit to be dropped. The synchronize_rcu call
also ensures all outstanding calls leave their critical section.
The other entry point is from qdisc_run. Since we now have a bit that
indicates whether it's running, all we have to do is to wait until the
bit is off.
I've removed the loop to wait for __LINK_STATE_SCHED to clear. This is
useless because netif_wake_queue can cause it to be set again. It is
also harmless because we've disarmed qdisc_run.
I've also removed the spin_unlock_wait on xmit_lock because its only
purpose of making sure that all outstanding xmit_lock holders have
exited is also given by dev_watchdog_down.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Having two or more qdisc_run's contend against each other is bad because
it can induce packet reordering if the packets have to be requeued. It
appears that this is an unintended consequence of relinquinshing the queue
lock while transmitting. That in turn is needed for devices that spend a
lot of time in their transmit routine.
There are no advantages to be had as devices with queues are inherently
single-threaded (the loopback device is not but then it doesn't have a
queue).
Even if you were to add a queue to a parallel virtual device (e.g., bolt
a tbf filter in front of an ipip tunnel device), you would still want to
process the queue in sequence to ensure that the packets are ordered
correctly.
The solution here is to steal a bit from net_device to prevent this.
BTW, as qdisc_restart is no longer used by anyone as a module inside the
kernel (IIRC it used to with netif_wake_queue), I have not exported the
new __qdisc_run function.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Various drivers use xmit_lock internally to synchronise with their
transmission routines. They do so without setting xmit_lock_owner.
This is fine as long as netpoll is not in use.
With netpoll it is possible for deadlocks to occur if xmit_lock_owner
isn't set. This is because if a printk occurs while xmit_lock is held
and xmit_lock_owner is not set can cause netpoll to attempt to take
xmit_lock recursively.
While it is possible to resolve this by getting netpoll to use
trylock, it is suboptimal because netpoll's sole objective is to
maximise the chance of getting the printk out on the wire. So
delaying or dropping the message is to be avoided as much as possible.
So the only alternative is to always set xmit_lock_owner. The
following patch does this by introducing the netif_tx_lock family of
functions that take care of setting/unsetting xmit_lock_owner.
I renamed xmit_lock to _xmit_lock to indicate that it should not be
used directly. I didn't provide irq versions of the netif_tx_lock
functions since xmit_lock is meant to be a BH-disabling lock.
This is pretty much a straight text substitution except for a small
bug fix in winbond. It currently uses
netif_stop_queue/spin_unlock_wait to stop transmission. This is
unsafe as an IRQ can potentially wake up the queue. So it is safer to
use netif_tx_disable.
The hamradio bits used spin_lock_irq but it is unnecessary as
xmit_lock must never be taken in an IRQ handler.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a potential jiffy wraparound bug in the transmit watchdog
that is easily avoided by using time_after().
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When deleting the last child the level of a class should drop to zero.
Noticed by Andreas Mueller <andreas@stapelspeicher.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The following one line fix is needed to make loss function of
netem work right when doing loss on the local host.
Otherwise, higher layers just recover.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The targets don't do the basic verification themselves anymore so
the ipt action needs to take care of it.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename policer specific _generic_ methods to be specific to
_act_police_
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
In both cases n can't be NULL without crashing anyway.
Coverity #78
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This option should IMHO no longer depend on EXPERIMENTAL.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
ACKed-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of the old __dev_put macro that is just a hold over from pre 2.6
kernel. And turn dev_hold into an inline instead of a macro.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert sch_red to a classful qdisc. All qdiscs that maintain accurate
backlog counters are eligible as child qdiscs. When a queue limit larger
than zero is given, a bfifo qdisc is used for backwards compatibility.
Current versions of tc enforce a limit larger than zero, other users
can avoid creating the default qdisc by using zero.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Keep backlog counter in SFQ qdisc to make it usable as child qdisc
with RED.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
When TBF was converted to a classful qdisc, the semantic of the limit
parameter was broken. On initilization an inner bfifo qdisc is created
for backwards compatibility, when changing parameters however the new
limit is ignored and the current child qdisc remains in place.
Always replace the child qdisc by the default bfifo when limit is above
zero, otherwise don't touch the inner qdisc. Current tc version enforce
a limit above zero, other users can avoid creating the inner qdisc by
using zero.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
A qdisc should set tcm_info to the child qdisc handle in its class
dump function.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The drop operation is optional and qdiscs must check if childs support it.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This allows to make decisions based on the revision (and address family
with a follow-up patch) at runtime.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The skb is allocated by the function, so it needs to be freed instead
of trimmed on overrun.
Coverity #614
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently when PRIO is configured to use N bands, it lets the packets be
directed to any of the bands 0..N-1. However, PRIO attaches a fifo qdisc
only to the bands that appear in the priomap; the rest of the N bands
remain with a noop qdisc attached. This patch changes PRIO's behavior so
that it attaches a fifo qdisc to all of the N bands.
Signed-off-by: Amnon Aaronsohn <bla@cs.huji.ac.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
The default of using jiffies is very bad and results in
underutilization except with very low bandwidth.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This monster-patch tries to do the best job for unifying the data
structures and backend interfaces for the three evil clones ip_tables,
ip6_tables and arp_tables. In an ideal world we would never have
allowed this kind of copy+paste programming... but well, our world
isn't (yet?) ideal.
o introduce a new x_tables module
o {ip,arp,ip6}_tables depend on this x_tables module
o registration functions for tables, matches and targets are only
wrappers around x_tables provided functions
o all matches/targets that are used from ip_tables and ip6_tables
are now implemented as xt_FOOBAR.c files and provide module aliases
to ipt_FOOBAR and ip6t_FOOBAR
o header files for xt_matches are in include/linux/netfilter/,
include/linux/netfilter_{ipv4,ipv6} contains compatibility wrappers
around the xt_FOOBAR.h headers
Based on this patchset we're going to further unify the code,
gradually getting rid of all the layer 3 specific assumptions.
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mapping between TC_ACTION_SHOT and the qdisc return codes is better
suited to NET_XMIT_BYPASS so as not to confuse TCP
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clean up the net/sched directory a bit by prefix all actions with act_.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Also make sure the legacy code is only built when CONFIG_NET_CLS_ACT
is not set.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcf_action_exec only gets a single skb pointer and doesn't own the skb,
but passes double skb pointers (to a local variable) to the action
functions. Change to use single skb pointers everywhere.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
To help in reducing the number of include dependencies, several files were
touched as they were getting needed headers indirectly for stuff they use.
Thanks also to Alan Menegotto for pointing out that net/dccp/proto.c had
linux/dccp.h include twice.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Here is a new feature for netem in 2.6.16. It adds the ability to
randomly corrupt packets with netem. A version was done by
Hagen Paul Pfeifer, but I redid it to handle the cases of backwards
compatibility with netlink interface and presence of hardware checksum
offload. It is useful for testing hardware offload in devices.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If two packets were queued to be sent at the same time in the future,
their order would be reversed. This would occur because the queue is
traversed back to front, and a position is found by checking whether
the new packet needs to be sent before the packet being examined. If
the new packet is to be sent at the same time of a previous packet, it
would end up before the old packet in the queue. This patch places
packets in the correct order when they are queued to be sent at a same
time in the future.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
On Thu, 17 Nov 2005, David Gmez wrote:
> I found out that if i select NET_CLS_ROUTE4, save my changes and exit
> menuconfig, execute again make menuconfig and go to QoS options, then the new
> available options are visible. So menuconfig has some problem refreshing
> contents :?
No, they were there before too, but you have to go up one level to see
them.
It's better in 2.6.15-rc1-git5, but the menu structure is still a little
messed up, the patch below properly indents all menu entries.
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Jesper Juhl <jesper.juhl@gmail.com>
This is the net/ part of the big kfree cleanup patch.
Remove pointless checks for NULL prior to calling kfree() in net/.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Arnaldo Carvalho de Melo <acme@conectiva.com.br>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Fixes an invalid memory reference when the basic classifier
is used without any ematches but just actions.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a version string to help support issues.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Change netem to support packets getting reordered because of variations in
delay. Introduce a special case version of FIFO that queues packets in order
based on the netem delay.
Since netem is classful, those users that don't want jitter based reordering
can just insert a pfifo instead of the default.
This required changes to generic skbuff code to allow finer grain manipulation
of sk_buff_head. Insertion into the middle and reverse walk.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduces a new flag TC_RED_HARDDROP which specifies that if ECN
marking is enabled packets should still be dropped once the
average queue length exceeds the maximum threshold.
This _may_ help to avoid global synchronisation during small
bursts of peers advertising but not caring about ECN. Use this
option very carefully, it does more harm than good if
(qth_max - qth_min) does not cover at least two average burst
cycles.
The difference to the current behaviour, in which we'd run into
the hard queue limit, is that due to the low pass filter of RED
short bursts are less likely to cause a global synchronisation.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Adds a new u8 flags in a unused padding area of the netlink
message. Adds ECN marking support to be used instead of dropping
packets immediately.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Removes unnecessary includes, initializers, and simplifies
the code a bit.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Since we are no longer depending on the default VQ to be always
allocated we can leave it up to the user to actually create it.
This gives the user the ability to leave it out on purpose and
enqueue packets directly to the device without applying the RED
algorithm.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduces a new red parameter set for use in equalize mode,
although only the qavg variable and the idle period marker are
being used for now this makes it possible to allow a separate
parameter set to be used for equalize later on.
The use of this separate parameter set fixes a bogus start of
an idle period in gred_drop() which did start an idle period
on the default VQ even if equalize mode was disabled.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
The case when the default VQ is not set up yet is already handled
in a less error prone way.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Try to enqueue packets if we cannot associate it with a VQ, this
basically means that the default VQ has not been set up yet.
We must check if the VQ still exists while requeueing, the VQ
might have been changed between dequeue and the requeue of the
underlying qdisc.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Adds a transformation function returning the DP index for a
given skb according to its tc_index.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Qdiscs are not supposed to reset statistics in reset() and while
changing parameters. My argumentation is that if the user wants
the counters to be reset he can simply remove and readd the
qdiscs, that's what most users do anyway.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Simplifies code a lot by separating the red algorithm and the
queueing logic. We now differentiate between probability marks
and forced marks but sum them together again to not break
backwards compatibility.
This brings GRED back to the level of RED and improves the
accuracy of the averge queue length calculations when stab
suggests a zero shift.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduces a function gred_change_vq() acting as a central point
to change VQ parameters. Fixes priority inheritance in rio mode
when the default DP equals 0. Adds proper locking during changes.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduces a function gred_change_table_def() acting as a central
point to change the table definition.
Adds missing validations for table definition: MAX_DPs > DPs > 0
and def_DP < DPs thus fixing possible invalid memory reference
oopses. Only root could do it but having a typo crashing the
machine is a bit hard.
Adds missing locking while changing the table definition, the
operation of changing the number of DPs and removing shadowed VQs
may not be interrupted by a dequeue.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Avoids the allocation of a buffer by appending the VQs directly
to the skb and simplifies the code by using the appropriate
message construction macros.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduces a flags variable using bitops and transforms eqp to use
it. Converts the conditions of the form (wred && rio) to (wred)
since wred can only be enabled in rio mode anyway.
The patch also improves WRED mode detection. The current behaviour
does not allow WRED mode to be turned off again without removing
the whole qdisc first. The new algorithm checks each VQ against
each other looking for equal priorities every time a VQ is changed
or added. The performance is poor, O(n**2), but it's used only
during administrative tasks and the number of VQs is strictly
limited.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Removes the skb trimming code which is not needed since we never
touch the skb upon failure. Removes unnecessary includes,
initializers, and simplifies the code a bit. Removes Jamal's
obsolete email addresses upon his own request.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
We should not interrupt and restart an idle period while idling already.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Simplifies code a lot by separating the red algorithm and the
queueing logic. We now differentiate between probability marks
and forced marks but sum them together again to not break
backwards compatibility.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Convert netem to use PSCHED_LESS and warn if requeue fails.
With some of the psched clock sources, the subtraction doesn't
work always work right without wrapping.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Make "QoS and/or fair queueing" have its own menu, it's too big to be
inlined into "Network options". Remove the obsolete NET_QOS option.
Automatically select NET_CLS if needed. Do the same for NET_ESTIMATOR
but allow it to be selected manually for statistical purposes. Add
comments to separate queueing from classification. Fix dependencies
and ordering of classifiers. Improve descriptions/help texts and
remove outdated pieces.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Opterons with frequency scaling have fully unsynchronized TSCs
running at different frequencies, so using TSCs there is not a good idea.
Also some other x86 boxes have this problem. gettimeofday should be good
enough, so just disable it.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Arnaldo and I agreed it could be applied now, because I have other
pending patches depending on this one (Thank you Arnaldo)
(The other important patch moves skc_refcnt in a separate cache line,
so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)
1) First some performance data :
--------------------------------
tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()
The most time critical code is :
sk_for_each(sk, node, &head->chain) {
if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
goto hit; /* You sunk my battleship! */
}
The sk_for_each() does use prefetch() hints but only the begining of
"struct sock" is prefetched.
As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
away from the begining of "struct sock", it has to bring into CPU
cache cold cache line. Each iteration has to use at least 2 cache
lines.
This can be problematic if some chains are very long.
2) The goal
-----------
The idea I had is to change things so that INET_MATCH() may return
FALSE in 99% of cases only using the data already in the CPU cache,
using one cache line per iteration.
3) Description of the patch
---------------------------
Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
filling a 32 bits hole on 64 bits platform.
struct sock_common {
unsigned short skc_family;
volatile unsigned char skc_state;
unsigned char skc_reuse;
int skc_bound_dev_if;
struct hlist_node skc_node;
struct hlist_node skc_bind_node;
atomic_t skc_refcnt;
+ unsigned int skc_hash;
struct proto *skc_prot;
};
Store in this 32 bits field the full hash, not masked by (ehash_size -
1) Using this full hash as the first comparison done in INET_MATCH
permits us immediatly skip the element without touching a second cache
line in case of a miss.
Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
sk_hash and tw_hash) already contains the slot number if we mask with
(ehash_size - 1)
File include/net/inet_hashtables.h
64 bits platforms :
#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash))
((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie)) && \
((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
32bits platforms:
#define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash)) && \
(inet_sk(__sk)->daddr == (__saddr)) && \
(inet_sk(__sk)->rcv_saddr == (__daddr)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
- Adds a prefetch(head->chain.first) in
__inet_lookup_established()/__tcp_v4_check_established() and
__inet6_lookup_established()/__tcp_v6_check_established() and
__dccp_v4_check_established() to bring into cache the first element of the
list, before the {read|write}_lock(&head->lock);
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clean up timer initialization by introducing DEFINE_TIMER a'la
DEFINE_SPINLOCK. Build and boot-tested on x86. A similar patch has been
been in the -RT tree for some time.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
# grep -r 'netif_carrier_o[nf]' linux-2.6.12 | wc -l
246
# size vmlinux.org vmlinux.carrier
text data bss dec hex filename
4339634 1054414 259296 5653344 564360 vmlinux.org
4337710 1054414 259296 5651420 563bdc vmlinux.carrier
And this ain't an allyesconfig kernel!
Signed-off-by: David S. Miller <davem@davemloft.net>
qdisc_create_dflt() is missing to destroy the newly allocated
default qdisc if the initialization fails resulting in leaks
of all kinds. The only caller in mainline which may trigger
this bug is sch_tbf.c in tbf_create_dflt_qdisc().
Note: qdisc_create_dflt() doesn't fulfill the official locking
requirements of qdisc_destroy() but since the qdisc could
never be seen by the outside world this doesn't matter
and it can stay as-is until the locking of pkt_sched
is cleaned up.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
More unusable TCF_META_* match types that need to get eliminated
before 2.6.13 goes out the door.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
It won't exist any longer when we shrink the SKB in 2.6.14,
and we should kill this off before anyone in userspace starts
using it.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
The current call to __qdisc_dequeue_head leads to a branch
misprediction for every loop iteration, the fact that the
most common priority is 2 makes this even worse. This issue
has been brought up by Eric Dumazet <dada1@cosmosbay.com>
but unlike his solution which was to manually unroll the loop,
this approach preserves the possibility to increase the number
of bands at compile time.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the protocol specific config options out to the specific protocols.
With this change net/Kconfig now starts to become readable and serve as a
good basis for further re-structuring.
The menu structure is left almost intact, except that indention is
fixed in most cases. Most visible are the INET changes where several
"depends on INET" are replaced with a single ifdef INET / endif pair.
Several new files were created to accomplish this change - they are
small but serve the purpose that config options are now distributed
out where they belongs.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is part of the grand scheme to eliminate the qlen
member of skb_queue_head, and subsequently remove the
'list' member of sk_buff.
Most users of skb_queue_len() want to know if the queue is
empty or not, and that's trivially done with skb_queue_empty()
which doesn't use the skb_queue_head->qlen member and instead
uses the queue list emptyness as the test.
Signed-off-by: David S. Miller <davem@davemloft.net>
Useful in combination with classful qdiscs to drop or
temporary disable certain flows, e.g. one could block
specific ds flows with dsmark.
Unlike the noop qdisc it can be controlled by the user and
statistic accounting is done.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current behaviour is to not report an error if a rate
estimator is created together with a qdisc and the
configuration of the rate estimator is bogus. This leads
to unexpected behaviour because the user is not notified.
New behaviour is to report the error and let the whole
qdisc creation operation fail so the user is able to fix
his mistake.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds qdisc_alloc() to share code between qdisc_create()
and qdisc_create_dflt(). Hides the qdisc alignment behind
macros and makes use of them.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Plug holes with padding fields and initialized them to zero.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mostly missing initialization of padding fields of 1 or 2 bytes length,
two instances of uninitialized nlmsgerr->msg of 16 bytes length.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Do not present these confusing new options to the user
unless he picked some facility that makes use of it,
such as NET_EMATCH_TEXT.
Signed-off-by: David S. Miller <davem@davemloft.net>
