tun_do_read always adds current thread to wait queue, even if a packet
is ready to read. This is inefficient because both sleeper and waker
want to acquire the wait queue spin lock when packet rate is high.
We restructure the read function and use common kernel networking
routines to handle receive, sleep and wakeup. With the change
available packets are checked first before the reading thread is added
to the wait queue.
Ran performance tests with the following configuration:
- my packet generator -> tap1 -> br0 -> tap0 -> my packet consumer
- sender pinned to one core and receiver pinned to another core
- sender send small UDP packets (64 bytes total) as fast as it can
- sandy bridge cores
- throughput are receiver side goodput numbers
The results are
baseline: 731k pkts/sec, cpu utilization at 1.50 cpus
changed: 783k pkts/sec, cpu utilization at 1.53 cpus
The performance difference is largely determined by packet rate and
inter-cpu communication cost. For example, if the sender and
receiver are pinned to different cpu sockets, the results are
baseline: 558k pkts/sec, cpu utilization at 1.71 cpus
changed: 690k pkts/sec, cpu utilization at 1.67 cpus
Co-authored-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Xi Wang <xii@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch replaces rcu_assign_pointer(x, NULL) with RCU_INIT_POINTER(x, NULL)
The rcu_assign_pointer() ensures that the initialization of a structure
is carried out before storing a pointer to that structure.
And in the case of the NULL pointer, there is no structure to initialize.
So, rcu_assign_pointer(p, NULL) can be safely converted to RCU_INIT_POINTER(p, NULL)
Signed-off-by: Monam Agarwal <monamagarwal123@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Even though only the outer vlan tag can be HW accelerated in the transmission
path, in the TUN/TAP driver vlan_features mirrors hw_features, which happens
to have the NETIF_F_HW_VLAN_?TAG_TX flags set. Because of this, during packet
tranmisssion through a stacked vlan device dev_hard_start_xmit, (incorrectly)
assuming that the vlan device supports hardware vlan acceleration, does not
add the vlan header to the skb payload and the inner vlan tags are lost
(vlan_tci contains the outer vlan tag when userspace reads the packet from
the tap device).
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a new argument for ndo_select_queue() callback that passes a
fallback handler. This gets invoked through netdev_pick_tx();
fallback handler is currently __netdev_pick_tx() as most drivers
invoke this function within their customized implementation in
case for skbs that don't need any special handling. This fallback
handler can then be replaced on other call-sites with different
queue selection methods (e.g. in packet sockets, pktgen etc).
This also has the nice side-effect that __netdev_pick_tx() is
then only invoked from netdev_pick_tx() and export of that
function to modules can be undone.
Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A file descriptor opened for /dev/net/tun and a tun device are
connected with ioctl. Though understanding the connection is
important for trouble shooting, no way is given to a user to know
the connected device for a given file descriptor at userland.
This patch adds a new fdinfo field for the device name connected to
a file descriptor opened for /dev/net/tun.
Here is an example of the field:
# lsof | grep tun
qemu-syst 4565 qemu 25u CHR 10,200 0t138 12921 /dev/net/tun
...
# cat /proc/4565/fdinfo/25
pos: 138
flags: 0104002
iff: vnet0
# ip link show dev vnet0
8: vnet0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 ...
changelog:
v2: indent iff just like the other fdinfo fields are.
v3: remove unused variable.
Both are suggested by David Miller <davem@davemloft.net>.
Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A patch for fixing a race between queue selection and changing queues
was introduced in commit 92bb73ea2("tuntap: fix a possible race between
queue selection and changing queues").
The fix was to prevent the driver from re-reading the tun->numqueues
more than once within tun_select_queue() using ACCESS_ONCE().
We have been experiancing 'Divide-by-zero' errors in tun_net_xmit()
since we moved from 3.6 to 3.10, and believe that they come from a
simular source where the value of tun->numqueues changes to zero
between the first and a subsequent read of tun->numqueues.
The fix is a simular use of ACCESS_ONCE(), as well as a multiply
instead of a divide in the if statement.
Signed-off-by: Dominic Curran <dominic.curran@citrix.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Maxim Krasnyansky <maxk@qti.qualcomm.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Max Krasnyansky <maxk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the tx queue were selected implicitly in ndo_dfwd_start_xmit(). The
will cause several issues:
- NETIF_F_LLTX were removed for macvlan, so txq lock were done for macvlan
instead of lower device which misses the necessary txq synchronization for
lower device such as txq stopping or frozen required by dev watchdog or
control path.
- dev_hard_start_xmit() was called with NULL txq which bypasses the net device
watchdog.
- dev_hard_start_xmit() does not check txq everywhere which will lead a crash
when tso is disabled for lower device.
Fix this by explicitly introducing a new param for .ndo_select_queue() for just
selecting queues in the case of l2 forwarding offload. netdev_pick_tx() was also
extended to accept this parameter and dev_queue_xmit_accel() was used to do l2
forwarding transmission.
With this fixes, NETIF_F_LLTX could be preserved for macvlan and there's no need
to check txq against NULL in dev_hard_start_xmit(). Also there's no need to keep
a dedicated ndo_dfwd_start_xmit() and we can just reuse the code of
dev_queue_xmit() to do the transmission.
In the future, it was also required for macvtap l2 forwarding support since it
provides a necessary synchronization method.
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: e1000-devel@lists.sourceforge.net
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code incorrectly save the queue index as the hash, so this patch
is fixing it with the hash received in the stack receive path.
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds support so that the rps_flow_tables (RFS) can be
programmed using the tun flows which are already set up to track flows
for the purposes of queue selection.
On the receive path (corresponding to select_queue and tun_net_xmit) the
rxhash is saved in the flow_entry. The original code only does flow
lookup in select_queue, so this patch adds a flow lookup in tun_net_xmit
if num_queues == 1 (select_queue is not called from
dev_queue_xmit->netdev_pick_tx in that case).
The flow is recorded (processing CPU) in tun_flow_update (TX path), and
reset when flow is deleted.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/intel/i40e/i40e_main.c
drivers/net/macvtap.c
Both minor merge hassles, simple overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
Changing name of function as part of making the hash in skbuff to be
generic property, not just for receive path.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 6680ec68ef
(tuntap: hardware vlan tx support) breaks the truncated packet signal by nev
return a length greater than iov length in tun_put_user(). This patch fixes
by always return the length of packet plus possible vlan header. Caller can
detect the truncated packet by comparing the return value and the size of io
length.
Cc: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 6680ec68ef
(tuntap: hardware vlan tx support) breaks the truncated packet signal by never
return a length greater than iov length in tun_put_user(). This patch fixes this
by always return the length of packet plus possible vlan header. Caller can
detect the truncated packet by comparing the return value and the size of iov
length.
Reported-by: Vlad Yasevich <vyasevich@gmail.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 73713357ab.
MSG_TRUNC handling was broken and is going to be fixed in
the 'net' tree, so revert this.
Signed-off-by: David S. Miller <davem@davemloft.net>
By checking related codes, it is impossible that ret > len or total_len,
so we should remove some useless codes in both above functions.
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Merge 'net' into 'net-next' to get the AF_PACKET bug fix that
Daniel's direct transmit changes depend upon.
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently use hdr_len as a hint of head length which is advertised by
guest. But when guest advertise a very big value, it can lead to an 64K+
allocating of kmalloc() which has a very high possibility of failure when host
memory is fragmented or under heavy stress. The huge hdr_len also reduce the
effect of zerocopy or even disable if a gso skb is linearized in guest.
To solves those issues, this patch introduces an upper limit (PAGE_SIZE) of the
head, which guarantees an order 0 allocation each time.
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We play with a wait queue even if socket is
non blocking. This is an obvious waste.
Besides, it will prevent calling the non blocking
variant when current is not valid.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit c8d68e6be1
(tuntap: multiqueue support) only call free_netdev() on error in
tun_set_iff(). This causes several issues:
- memory of tun security were leaked
- use after free since the flow gc timer was not deleted and the tfile
were not detached
This patch solves the above issues.
Reported-by: Wannes Rombouts <wannes.rombouts@epitech.eu>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sock_tx_timestamp() will clear all zerocopy flags of skb which may lead the
frags never to be orphaned. This will break guest to guest traffic when zerocopy
is enabled. Fix this by orphaning the frags before trying to set tx time stamp.
The issue were introduced by commit eda2977291
(tun: Support software transmit time stamping).
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Xi Wang
Monam Agarwal
Fernando Luis Vazquez Cao
Daniel Borkmann
Masatake YAMATO
Dominic Curran
David S. Miller
Jason Wang
Zhi Yong Wu
Tom Herbert
stephen hemminger
Michael S. Tsirkin