netif_needs_gso() is checked twice in the TX path once,
before submitting the skb to the qdisc and once after
it is dequeued from the qdisc just before calling
ndo_hard_start(). This opens a window for a user to
change the gso/tso or tx checksum settings that can
cause netif_needs_gso to be true in one check and false
in the other.
Specifically, changing TX checksum setting may cause
the warning in skb_gso_segment() to be triggered if
the checksum is calculated earlier.
This consolidates the netif_needs_gso() calls so that
the stack only checks if gso is needed in
dev_hard_start_xmit().
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Register net_bridge_port pointer as rx_handler data pointer. As br_port is
removed from struct net_device, another netdev priv_flag is added to indicate
the device serves as a bridge port. Also rcuized pointers are now correctly
dereferenced in br_fdb.c and in netfilter parts.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add possibility to register rx_handler data pointer along with a rx_handler.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use struct rtnl_link_stats64 as the statistics structure.
On 32-bit architectures, insert 32 bits of padding after/before each
field of struct net_device_stats to make its layout compatible with
struct rtnl_link_stats64. Add an anonymous union in net_device; move
stats into the union and add struct rtnl_link_stats64 stats64.
Add net_device_ops::ndo_get_stats64, implementations of which will
return a pointer to struct rtnl_link_stats64. Drivers that implement
this operation must not update the structure asynchronously.
Change dev_get_stats() to call ndo_get_stats64 if available, and to
return a pointer to struct rtnl_link_stats64. Change callers of
dev_get_stats() accordingly.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the accelerated receive path for VLAN's will
drop packets if the real device is an inactive slave and
is not one of the special pkts tested for in
skb_bond_should_drop(). This behavior is different then
the non-accelerated path and for pkts over a bonded vlan.
For example,
vlanx -> bond0 -> ethx
will be dropped in the vlan path and not delivered to any
packet handlers at all. However,
bond0 -> vlanx -> ethx
and
bond0 -> ethx
will be delivered to handlers that match the exact dev,
because the VLAN path checks the real_dev which is not a
slave and netif_recv_skb() doesn't drop frames but only
delivers them to exact matches.
This patch adds a sk_buff flag which is used for tagging
skbs that would previously been dropped and allows the
skb to continue to skb_netif_recv(). Here we add
logic to check for the deliver_no_wcard flag and if it
is set only deliver to handlers that match exactly. This
makes both paths above consistent and gives pkt handlers
a way to identify skbs that come from inactive slaves.
Without this patch in some configurations skbs will be
delivered to handlers with exact matches and in others
be dropped out right in the vlan path.
I have tested the following 4 configurations in failover modes
and load balancing modes.
# bond0 -> ethx
# vlanx -> bond0 -> ethx
# bond0 -> vlanx -> ethx
# bond0 -> ethx
|
vlanx -> --
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
BugLink: http://bugs.launchpad.net/bugs/591416
There are a number of network drivers (bridge, bonding, etc) that are not yet
receive multi-queue enabled and use alloc_netdev(), so don't print a
num_rx_queues imbalance warning in that case.
Also, only print the warning once for those drivers that _are_ multi-queue
enabled.
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
- dev_get_by_flags() changed to dev_get_by_flags_rcu()
- ipv6_sock_ac_join() dont touch dev & idev refcounts
- ipv6_sock_ac_drop() dont touch dev & idev refcounts
- ipv6_sock_ac_close() dont touch dev & idev refcounts
- ipv6_dev_ac_dec() dount touch idev refcount
- ipv6_chk_acast_addr() dont touch idev refcount
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The extra assertion to allow packet munging only when there are
no other ptypes listening which may have worked around an old bug
is unnecessary. It is sufficient to check if the skb is cloned before
trampling on it. Thanks to Herbert Xu for being persistent and patient
in getting this across.
[Note that cloning checks and assertions are the general rule used
by tc actions (documentation/networking/tc-actions-env-rules.txt)].
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can avoid an unecessary cache miss by checking if the skb is non-linear
before accessing gso_size/gso_type in skb_warn_if_lro, the same can also be
done to avoid a cache miss on nr_frags if data_len is 0.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
What this patch does is it removes two receive frame hooks (for bridge and for
macvlan) from __netif_receive_skb. These are replaced them with a single
hook for both. It only supports one hook per device because it makes no
sense to do bridging and macvlan on the same device.
Then a network driver (of virtual netdev like macvlan or bridge) can register
an rx_handler for needed net device.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When many cpus compete for sending frames on a given qdisc, the qdisc
spinlock suffers from very high contention.
The cpu owning __QDISC_STATE_RUNNING bit has same priority to acquire
the lock, and cannot dequeue packets fast enough, since it must wait for
this lock for each dequeued packet.
One solution to this problem is to force all cpus spinning on a second
lock before trying to get the main lock, when/if they see
__QDISC_STATE_RUNNING already set.
The owning cpu then compete with at most one other cpu for the main
lock, allowing for higher dequeueing rate.
Based on a previous patch from Alexander Duyck. I added the heuristic to
avoid the atomic in fast path, and put the new lock far away from the
cache line used by the dequeue worker. Also try to release the busylock
lock as late as possible.
Tests with following script gave a boost from ~50.000 pps to ~600.000
pps on a dual quad core machine (E5450 @3.00GHz), tg3 driver.
(A single netperf flow can reach ~800.000 pps on this platform)
for j in `seq 0 3`; do
for i in `seq 0 7`; do
netperf -H 192.168.0.1 -t UDP_STREAM -l 60 -N -T $i -- -m 6 &
done
done
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If a skb is received on an inactive bond that does not meet
the special cases checked for by skb_bond_should_drop it should
only be delivered to exact matches as the comment in
netif_receive_skb() says.
However because null_or_bond could also be null this is not
always true. This patch renames null_or_bond to orig_or_bond
and initializes it to orig_dev. This keeps the intent of
null_or_bond to pass frames received on VLAN interfaces stacked
on bonding interfaces without invalidating the statement for
null_or_orig.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define three helpers to manipulate QDISC_STATE_RUNNIG flag, that a
second patch will move on another location.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
netpoll does an interesting work in zap_completion_queue(), but this was
before we did skb orphaning before delivering packets to device.
It now makes sense to add a test in dev_kfree_skb_irq() to not queue a
skb if already orphaned, and to remove netpoll zap_completion_queue() as
a bonus.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch saves 224 bytes of text on my machine.
__this_cpu_inc() generates a single instruction, using no scratch
registers :
65 ff 04 25 a8 30 01 00 incl %gs:0x130a8
instead of :
48 c7 c2 80 30 01 00 mov $0x13080,%rdx
65 48 8b 04 25 88 ea 00 00 mov %gs:0xea88,%rax
83 44 10 28 01 addl $0x1,0x28(%rax,%rdx,1)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (63 commits)
drivers/net/usb/asix.c: Fix pointer cast.
be2net: Bug fix to avoid disabling bottom half during firmware upgrade.
proc_dointvec: write a single value
hso: add support for new products
Phonet: fix potential use-after-free in pep_sock_close()
ath9k: remove VEOL support for ad-hoc
ath9k: change beacon allocation to prefer the first beacon slot
sock.h: fix kernel-doc warning
cls_cgroup: Fix build error when built-in
macvlan: do proper cleanup in macvlan_common_newlink() V2
be2net: Bug fix in init code in probe
net/dccp: expansion of error code size
ath9k: Fix rx of mcast/bcast frames in PS mode with auto sleep
wireless: fix sta_info.h kernel-doc warnings
wireless: fix mac80211.h kernel-doc warnings
iwlwifi: testing the wrong variable in iwl_add_bssid_station()
ath9k_htc: rare leak in ath9k_hif_usb_alloc_tx_urbs()
ath9k_htc: dereferencing before check in hif_usb_tx_cb()
rt2x00: Fix rt2800usb TX descriptor writing.
rt2x00: Fix failed SLEEP->AWAKE and AWAKE->SLEEP transitions.
...
the commit:
commit d90310243f
Author: Octavian Purdila <opurdila@ixiacom.com>
Date: Wed Nov 18 02:36:59 2009 +0000
net: device name allocation cleanups
introduced a bug when there is a hash collision making impossible
to rename a device with eth%d. This bug is very hard to reproduce
and appears rarely.
The problem is coming from we don't pass a temporary buffer to
__dev_alloc_name but 'dev->name' which is modified by the function.
A detailed explanation is here:
http://marc.info/?l=linux-netdev&m=127417784011987&w=2
Changelog:
V2 : replaced strings comparison by pointers comparison
Signed-off-by: Daniel Lezcano <daniel.lezcano@free.fr>
Reviewed-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit aaf8cdc34d.
Drivers like the ipw2100 call device_create_group when they
are initialized and device_remove_group when they are shutdown.
Moving them between namespaces deletes their sysfs groups early.
In particular the following call chain results.
netdev_unregister_kobject -> device_del -> kobject_del -> sysfs_remove_dir
With sysfs_remove_dir recursively deleting all of it's subdirectories,
and nothing adding them back.
Ouch!
Therefore we need to call something that ultimate calls sysfs_mv_dir
as that sysfs function can move sysfs directories between namespaces
without deleting their subdirectories or their contents. Allowing
us to avoid placing extra boiler plate into every driver that does
something interesting with sysfs.
Currently the function that provides that capability is device_rename.
That is the code works without nasty side effects as originally written.
So remove the misguided fix for moving devices between namespaces. The
bug in the kobject layer that inspired it has now been recognized and
fixed.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix some issues introduced in batch skb dequeuing for input_pkt_queue.
The primary issue it that the queue head must be incremented only
after a packet has been processed, that is only after
__netif_receive_skb has been called. This is needed for the mechanism
to prevent OOO packet in RFS. Also when flushing the input_pkt_queue
and process_queue, the process queue should be done first to prevent
OOO packets.
Because the input_pkt_queue has been effectively split into two queues,
the calculation of the tail ptr is no longer correct. The correct value
would be head+input_pkt_queue->len+process_queue->len. To avoid
this calculation we added an explict input_queue_tail in softnet_data.
The tail value is simply incremented when queuing to input_pkt_queue.
Signed-off-by: Tom Herbert <therbert@google.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use low order bit of skb->_skb_dst to tell dst is not refcounted.
Change _skb_dst to _skb_refdst to make sure all uses are catched.
skb_dst() returns the dst, regardless of noref bit set or not, but
with a lockdep check to make sure a noref dst is not given if current
user is not rcu protected.
New skb_dst_set_noref() helper to set an notrefcounted dst on a skb.
(with lockdep check)
skb_dst_drop() drops a reference only if skb dst was refcounted.
skb_dst_force() helper is used to force a refcount on dst, when skb
is queued and not anymore RCU protected.
Use skb_dst_force() in __sk_add_backlog(), __dev_xmit_skb() if
!IFF_XMIT_DST_RELEASE or skb enqueued on qdisc queue, in
sock_queue_rcv_skb(), in __nf_queue().
Use skb_dst_force() in dev_requeue_skb().
Note: dst_use_noref() still dirties dst, we might transform it
later to do one dirtying per jiffies.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If CONFIG_SMP=y, then we own a queue spinlock, we can avoid the atomic
test_and_set_bit() from napi_schedule_prep().
We now have same number of atomic ops per netif_rx() calls than with
pre-RPS kernel.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With RPS inclusion, skb timestamping is not consistent in RX path.
If netif_receive_skb() is used, its deferred after RPS dispatch.
If netif_rx() is used, its done before RPS dispatch.
This can give strange tcpdump timestamps results.
I think timestamping should be done as soon as possible in the receive
path, to get meaningful values (ie timestamps taken at the time packet
was delivered by NIC driver to our stack), even if NAPI already can
defer timestamping a bit (RPS can help to reduce the gap)
Tom Herbert prefer to sample timestamps after RPS dispatch. In case
sampling is expensive (HPET/acpi_pm on x86), this makes sense.
Let admins switch from one mode to another, using a new
sysctl, /proc/sys/net/core/netdev_tstamp_prequeue
Its default value (1), means timestamps are taken as soon as possible,
before backlog queueing, giving accurate timestamps.
Setting a 0 value permits to sample timestamps when processing backlog,
after RPS dispatch, to lower the load of the pre-RPS cpu.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now there's null check here and also again in the hook. Looking at bridge bits
which are simmilar, port structure is rcu_dereferenced right away in
handle_bridge and passed to hook. Looks nicer.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce ____napi_schedule() helper for callers in irq disabled
contexts. rps_trigger_softirq() becomes a leaf function.
Use container_of() in process_backlog() instead of accessing per_cpu
address.
Use a custom inlined version of __napi_complete() in process_backlog()
to avoid one locked instruction :
only current cpu owns and manipulates this napi,
and NAPI_STATE_SCHED is the only possible flag set on backlog.
we can use a plain write instead of clear_bit(),
and we dont need an smp_mb() memory barrier, since RPS is on,
backlog is protected by a spinlock.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of congestion, netif_rx() frees the skb, so we must assume
dev_forward_skb() also consume skb.
Bug introduced by commit 445409602c
(veth: move loopback logic to common location)
We must change dev_forward_skb() to always consume skb, and veth to not
double free it.
Bug report : http://marc.info/?l=linux-netdev&m=127310770900442&w=3
Reported-by: Martín Ferrari <martin.ferrari@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Per cpu variable softnet_data.total was shared between IRQ and SoftIRQ context
without any protection. And enqueue_to_backlog should update the netdev_rx_stat
of the target CPU.
This patch renames softnet_data.total to softnet_data.processed: the number of
packets processed in uppper levels(IP stacks).
softnet_stat data is moved into softnet_data.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
include/linux/netdevice.h | 17 +++++++----------
net/core/dev.c | 26 ++++++++++++--------------
net/sched/sch_generic.c | 2 +-
3 files changed, 20 insertions(+), 25 deletions(-)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
batch skb dequeueing from softnet input_pkt_queue to reduce potential lock
contention when RPS is enabled.
Note: in the worst case, the number of packets in a softnet_data may
be double of netdev_max_backlog.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
reimplement softnet_data.output_queue as a FIFO queue to keep the
fairness among the qdiscs rescheduled.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
----
include/linux/netdevice.h | 1 +
net/core/dev.c | 22 ++++++++++++----------
2 files changed, 13 insertions(+), 10 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>
optimize rps_get_cpu().
don't initialize ports when we can get the ports. one memory access
for ports than two.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This way GSO packets don't get handled differently.
With help from Eric Dumazet.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
If some skb are queued to our backlog, we are delaying IPI sending at
the end of net_rx_action(), increasing latencies. This defeats the
queueing, since we want to quickly dispatch packets to the pool of
worker cpus, then eventually deeply process our packets.
It's better to send IPI before processing our packets in upper layers,
from process_backlog().
Change the _and_disable_irq suffix to _and_enable_irq(), since we enable
local irq in net_rps_action(), sorry for the confusion.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
At this point, skb->destructor is not the original one (stored in
DEV_GSO_CB(skb)->destructor)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since netdev_chain is guarded by rtnl_lock, ASSERT_RTNL should be
present here to make sure that all callers of call_netdevice_notifiers
does the locking properly.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case we compute a software skb->rxhash, we can generate a consistent
hash : Its value will be the same in both flow directions.
This helps some workloads, like conntracking, since the same state needs
to be accessed in both directions.
tbench + RFS + this patch gives better results than tbench with default
kernel configuration (no RPS, no RFS)
Also fixed some sparse warnings.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct softnet_data holds many queues, so consistent use "sd" name
instead of "queue" is better.
Adds a rps_ipi_queued() helper to cleanup enqueue_to_backlog()
Adds a _and_irq_disable suffix to net_rps_action() name, as David
suggested.
incr_input_queue_head() becomes input_queue_head_incr()
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net_rps_action() is a bit expensive on NR_CPUS=64..4096 kernels, even if
RPS is not active.
Tom Herbert used two bitmasks to hold information needed to send IPI,
but a single LIFO list seems more appropriate.
Move all RPS logic into net_rps_action() to cleanup net_rx_action() code
(remove two ifdefs)
Move rps_remote_softirq_cpus into softnet_data to share its first cache
line, filling an existing hole.
In a future patch, we could call net_rps_action() from process_backlog()
to make sure we send IPI before handling this cpu backlog.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Transmitted skb might be attached to a socket and a destructor, for
memory accounting purposes.
Traditionally, this destructor is called at tx completion time, when skb
is freed.
When tx completion is performed by another cpu than the sender, this
forces some cache lines to change ownership. XPS was an attempt to give
tx completion to initial cpu.
David idea is to call destructor right before giving skb to device (call
to ndo_start_xmit()). Because device queues are usually small, orphaning
skb before tx completion is not a big deal. Some drivers already do
this, we could do it in upper level.
There is one known exception to this early orphaning, called tx
timestamping. It needs to keep a reference to socket until device can
give a hardware or software timestamp.
This patch adds a skb_orphan_try() helper, to centralize all exceptions
to early orphaning in one spot, and use it in dev_hard_start_xmit().
"tbench 16" results on a Nehalem machine (2 X5570 @ 2.93GHz)
before: Throughput 4428.9 MB/sec 16 procs
after: Throughput 4448.14 MB/sec 16 procs
UDP should get even better results, its destructor being more complex,
since SOCK_USE_WRITE_QUEUE is not set (four atomic ops instead of one)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- There is no point to enforce a time limit in process_backlog(), since
other napi instances dont follow same rule. We can exit after only one
packet processed...
The normal quota of 64 packets per napi instance should be the norm, and
net_rx_action() already has its own time limit.
Note : /proc/net/core/dev_weight can be used to tune this 64 default
value.
- Use DEFINE_PER_CPU_ALIGNED for softnet_data definition.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements receive flow steering (RFS). RFS steers
received packets for layer 3 and 4 processing to the CPU where
the application for the corresponding flow is running. RFS is an
extension of Receive Packet Steering (RPS).
The basic idea of RFS is that when an application calls recvmsg
(or sendmsg) the application's running CPU is stored in a hash
table that is indexed by the connection's rxhash which is stored in
the socket structure. The rxhash is passed in skb's received on
the connection from netif_receive_skb. For each received packet,
the associated rxhash is used to look up the CPU in the hash table,
if a valid CPU is set then the packet is steered to that CPU using
the RPS mechanisms.
The convolution of the simple approach is that it would potentially
allow OOO packets. If threads are thrashing around CPUs or multiple
threads are trying to read from the same sockets, a quickly changing
CPU value in the hash table could cause rampant OOO packets--
we consider this a non-starter.
To avoid OOO packets, this solution implements two types of hash
tables: rps_sock_flow_table and rps_dev_flow_table.
rps_sock_table is a global hash table. Each entry is just a CPU
number and it is populated in recvmsg and sendmsg as described above.
This table contains the "desired" CPUs for flows.
rps_dev_flow_table is specific to each device queue. Each entry
contains a CPU and a tail queue counter. The CPU is the "current"
CPU for a matching flow. The tail queue counter holds the value
of a tail queue counter for the associated CPU's backlog queue at
the time of last enqueue for a flow matching the entry.
Each backlog queue has a queue head counter which is incremented
on dequeue, and so a queue tail counter is computed as queue head
count + queue length. When a packet is enqueued on a backlog queue,
the current value of the queue tail counter is saved in the hash
entry of the rps_dev_flow_table.
And now the trick: when selecting the CPU for RPS (get_rps_cpu)
the rps_sock_flow table and the rps_dev_flow table for the RX queue
are consulted. When the desired CPU for the flow (found in the
rps_sock_flow table) does not match the current CPU (found in the
rps_dev_flow table), the current CPU is changed to the desired CPU
if one of the following is true:
- The current CPU is unset (equal to RPS_NO_CPU)
- Current CPU is offline
- The current CPU's queue head counter >= queue tail counter in the
rps_dev_flow table. This checks if the queue tail has advanced
beyond the last packet that was enqueued using this table entry.
This guarantees that all packets queued using this entry have been
dequeued, thus preserving in order delivery.
Making each queue have its own rps_dev_flow table has two advantages:
1) the tail queue counters will be written on each receive, so
keeping the table local to interrupting CPU s good for locality. 2)
this allows lockless access to the table-- the CPU number and queue
tail counter need to be accessed together under mutual exclusion
from netif_receive_skb, we assume that this is only called from
device napi_poll which is non-reentrant.
This patch implements RFS for TCP and connected UDP sockets.
It should be usable for other flow oriented protocols.
There are two configuration parameters for RFS. The
"rps_flow_entries" kernel init parameter sets the number of
entries in the rps_sock_flow_table, the per rxqueue sysfs entry
"rps_flow_cnt" contains the number of entries in the rps_dev_flow
table for the rxqueue. Both are rounded to power of two.
The obvious benefit of RFS (over just RPS) is that it achieves
CPU locality between the receive processing for a flow and the
applications processing; this can result in increased performance
(higher pps, lower latency).
The benefits of RFS are dependent on cache hierarchy, application
load, and other factors. On simple benchmarks, we don't necessarily
see improvement and sometimes see degradation. However, for more
complex benchmarks and for applications where cache pressure is
much higher this technique seems to perform very well.
Below are some benchmark results which show the potential benfit of
this patch. The netperf test has 500 instances of netperf TCP_RR
test with 1 byte req. and resp. The RPC test is an request/response
test similar in structure to netperf RR test ith 100 threads on
each host, but does more work in userspace that netperf.
e1000e on 8 core Intel
No RFS or RPS 104K tps at 30% CPU
No RFS (best RPS config): 290K tps at 63% CPU
RFS 303K tps at 61% CPU
RPC test tps CPU% 50/90/99% usec latency Latency StdDev
No RFS/RPS 103K 48% 757/900/3185 4472.35
RPS only: 174K 73% 415/993/2468 491.66
RFS 223K 73% 379/651/1382 315.61
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When dev_pick_tx() caches tx queue_index on a socket, we must check
socket dst_entry matches skb one, or risk a crash later, as reported by
Denys Fedorysychenko, if old packets are in flight during a route
change, involving devices with different number of queues.
Bug introduced by commit a4ee3ce3
(net: Use sk_tx_queue_mapping for connected sockets)
Reported-by: Denys Fedorysychenko <nuclearcat@nuclearcat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Paris reported netif_rx() is calling smp_processor_id() from
preemptible context, in particular when caller is
ip_dev_loopback_xmit().
RPS commit added this smp_processor_id() call, this patch makes sure
preemption is disabled. rps_get_cpus() wants rcu_read_lock() anyway, we
can dot it a bit earlier.
Reported-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_bond_should_drop() is too big to be inlined.
This patch reduces kernel text size, and its compilation time as well
(shrinking include/linux/netdevice.h)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With latest CONFIG_PROVE_RCU stuff, I felt more comfortable to make this
work.
sk->sk_dst_cache is currently protected by a rwlock (sk_dst_lock)
This rwlock is readlocked for a very small amount of time, and dst
entries are already freed after RCU grace period. This calls for RCU
again :)
This patch converts sk_dst_lock to a spinlock, and use RCU for readers.
__sk_dst_get() is supposed to be called with rcu_read_lock() or if
socket locked by user, so use appropriate rcu_dereference_check()
condition (rcu_read_lock_held() || sock_owned_by_user(sk))
This patch avoids two atomic ops per tx packet on UDP connected sockets,
for example, and permits sk_dst_lock to be much less dirtied.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dont use netdev_warn() in dev_cap_txqueue() and get_rps_cpu() so that we
can catch following warnings without crash.
bond0.2240 received packet on queue 6, but number of RX queues is 1
bond0.2240 received packet on queue 11, but number of RX queues is 1
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
John Fastabend