One point of contention in high network loads is the dst_release() performed
when a transmited skb is freed. This is because NIC tx completion calls
dev_kree_skb() long after original call to dev_queue_xmit(skb).
CPU cache is cold and the atomic op in dst_release() stalls. On SMP, this is
quite visible if one CPU is 100% handling softirqs for a network device,
since dst_clone() is done by other cpus, involving cache line ping pongs.
It seems right place to release dst is in dev_hard_start_xmit(), for most
devices but ones that are virtual, and some exceptions.
David Miller suggested to define a new device flag, set in alloc_netdev_mq()
(so that most devices set it at init time), and carefuly unset in devices
which dont want a NULL skb->dst in their ndo_start_xmit().
List of devices that must clear this flag is :
- loopback device, because it calls netif_rx() and quoting Patrick :
"ip_route_input() doesn't accept loopback addresses, so loopback packets
already need to have a dst_entry attached."
- appletalk/ipddp.c : needs skb->dst in its xmit function
- And all devices that call again dev_queue_xmit() from their xmit function
(as some classifiers need skb->dst) : bonding, vlan, macvlan, eql, ifb, hdlc_fr
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch moves neigh_setup and hard_start_xmit into the network device ops
structure. For bisection, fix all the previously converted drivers as well.
Bonding driver took the biggest hit on this.
Added a prefetch of the hard_start_xmit in the fast path to try and reduce
any impact this would have.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert to new network device ops interface.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When support for multiple TX queues were added, the
netif_tx_lock() routines we converted to iterate over
all TX queues and grab each queue's spinlock.
This causes heartburn for lockdep and it's not a healthy
thing to do with lots of TX queues anyways.
So modify this to use a top-level lock and a "frozen"
state for the individual TX queues.
Signed-off-by: David S. Miller <davem@davemloft.net>
alloc_netdev_mq() now allocates an array of netdev_queue
structures for TX, based upon the queue_count argument.
Furthermore, all accesses to the TX queues are now vectored
through the netdev_get_tx_queue() and netdev_for_each_tx_queue()
interfaces. This makes it easy to grep the tree for all
things that want to get to a TX queue of a net device.
Problem spots which are not really multiqueue aware yet, and
only work with one queue, can easily be spotted by grepping
for all netdev_get_tx_queue() calls that pass in a zero index.
Signed-off-by: David S. Miller <davem@davemloft.net>
Every qdisc is assosciated with a queue, and in the case of ingress
qdiscs that will now be netdev->rx_queue so using that queue's lock is
the thing to do.
Signed-off-by: David S. Miller <davem@davemloft.net>
The lock is now an attribute of the device queue.
One thing to notice is that "suspicious" places
emerge which will need specific training about
multiple queue handling. They are so marked with
explicit "netdev->rx_queue" and "netdev->tx_queue"
references.
Signed-off-by: David S. Miller <davem@davemloft.net>
[ 10.536424] =======================================================
[ 10.536424] [ INFO: possible circular locking dependency detected ]
[ 10.536424] 2.6.25-rc3-devel #3
[ 10.536424] -------------------------------------------------------
[ 10.536424] swapper/0 is trying to acquire lock:
[ 10.536424] (&dev->queue_lock){-+..}, at: [<c0299b4a>]
dev_queue_xmit+0x175/0x2f3
[ 10.536424]
[ 10.536424] but task is already holding lock:
[ 10.536424] (&p->tcfc_lock){-+..}, at: [<f8a67154>] tcf_mirred+0x20/0x178
[act_mirred]
[ 10.536424]
[ 10.536424] which lock already depends on the new lock.
lockdep warns of locking order while using ifb with sch_ingress and
act_mirred: ingress_lock, tcfc_lock, queue_lock (usually queue_lock
is at the beginning). This patch is only to tell lockdep that ifb is
a different device (e.g. from eth) and has its own pair of queue
locks. (This warning is a false-positive in common scenario of using
ifb; yet there are possible situations, when this order could be
dangerous; lockdep should warn in such a case.) (With suggestions by
David S. Miller)
Reported-and-tested-by: Denys Fedoryshchenko <denys@visp.net.lb>
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>
We now have struct net_device_stats embedded in struct net_device,
and the default ->get_stats() hook does the obvious thing for us.
Run through drivers/net/* and remove the driver-local storage of
statistics, and driver-local ->get_stats() hook where applicable.
This was just the low-hanging fruit in drivers/net; plenty more drivers
remain to be updated.
[ Resolved conflicts with napi_struct changes and fix sunqe build
regression... -DaveM ]
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
It's been a useless no-op for long enough in 2.6 so I figured it's time to
remove it. The number of people that could object because they're
maintaining unified 2.4 and 2.6 drivers is probably rather small.
[ Handled drivers added by netdev tree and some missed IRDA cases... -DaveM ]
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch makes most of the generic device layer network
namespace safe. This patch makes dev_base_head a
network namespace variable, and then it picks up
a few associated variables. The functions:
dev_getbyhwaddr
dev_getfirsthwbytype
dev_get_by_flags
dev_get_by_name
__dev_get_by_name
dev_get_by_index
__dev_get_by_index
dev_ioctl
dev_ethtool
dev_load
wireless_process_ioctl
were modified to take a network namespace argument, and
deal with it.
vlan_ioctl_set and brioctl_set were modified so their
hooks will receive a network namespace argument.
So basically anthing in the core of the network stack that was
affected to by the change of dev_base was modified to handle
multiple network namespaces. The rest of the network stack was
simply modified to explicitly use &init_net the initial network
namespace. This can be fixed when those components of the network
stack are modified to handle multiple network namespaces.
For now the ifindex generator is left global.
Fundametally ifindex numbers are per namespace, or else
we will have corner case problems with migration when
we get that far.
At the same time there are assumptions in the network stack
that the ifindex of a network device won't change. Making
the ifindex number global seems a good compromise until
the network stack can cope with ifindex changes when
you change namespaces, and the like.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Drivers need to validate the initial addresses in their netlink attribute
validation function or manually reject them if they can't support this.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
All drivers need to unregister their devices in the module unload function.
While doing so they must hold the rtnl and atomically unregister the
rtnl_link ops as well. This makes the rtnl_link_unregister function that
takes the rtnl itself completely useless.
Provide default newlink/dellink functions, make __rtnl_link_unregister and
rtnl_link_unregister unregister all devices with matching rtnl_link_ops and
change the existing users to take advantage of that.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use a list instead of an array to allow creating new devices.
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>
This reverts commit 0c0b3ae68e.
Quoth David:
"Jeff, please revert
It's wrong. We had a lengthy analysis of this piece of code
several months ago, and it is correct.
Consider, if we run the loop and we get an error
the following happens:
1) attempt of ifb_init_one(i) fails, therefore we should
not try to "ifb_free_one()" on "i" since it failed
2) the loop iteration first increments "i", then it
check for error
Therefore we must decrement "i" twice before the first
free during the cleanup. One to "undo" the for() loop
increment, and one to "skip" the ifb_init_one() case which
failed."
Reported-by: David Miller <davem@davemloft.net>
Acked-by: Jeff Garzik <jgarzik@pobox.com>
Cc: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On error we should start freeing resources at [i-1] not [i-2].
Signed-off-by: Mariusz Kozlowski <m.kozlowski@tuxland.pl>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
ifb: replace missing comma to separate pr_debug arguments
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It before entering in the loop for freeing the other ifb devices.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
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>
The most usable number of ifb devices is 2. Change the default to 2.
Signed-off-by: Richard Lucassen <spamtrap@lucassen.org>
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
A new device to do intermidiate functional block in a system shared
manner. To use the new functionality, you need to turn on
qos/classifier actions.
The new functionality can be grouped as:
1) qdiscs/policies that are per device as opposed to system wide. ifb
allows for a device which can be redirected to thus providing an
impression of sharing.
2) Allows for queueing incoming traffic for shaping instead of
dropping.
Packets are redirected to this device using tc/action mirred redirect
construct. If they are sent to it by plain routing instead then they
will merely be dropped and the stats would indicate that.
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet