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>
We can update netdev_queue tx_bytes/tx_packets/tx_dropped counters instead
of dev->stats ones, to reduce number of cache lines dirtied in xmit path.
This fixes a performance problem on SMP when many different cpus take
vlan tx path.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the VLAN event handler does not adjust the VLAN
device's carrier state when the real device or the VLAN device is set
administratively up or down.
The following patch adds a transfer of operating state from the
real device to the VLAN device when the real device is administratively
set up or down, and sets the carrier state up or down during init, open
and close of the VLAN device.
This permits observers above the VLAN device that care about the
carrier state (bonding's link monitor, for example) to receive updates
for administrative changes by more closely mimicing the behavior of real
devices.
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Check whether the underlying device provides a set of ethtool ops before
checking for individual handlers to avoid NULL pointer dereferences.
Reported-by: Art van Breemen <ard@telegraafnet.nl>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
It turns out that copying a 16-byte area at ~800k times a second
can be really expensive :) This patch redesigns the frags GRO
interface to avoid copying that area twice.
The two disciples of the frags interface have been converted.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hi:
gro: Normalise skb before bypassing GRO on netpoll VLAN path
When we detect netpoll RX on the GRO VLAN path we bail out and
call the normal VLAN receive handler. However, the packet needs
to be normalised by calling eth_type_trans since that's what the
normal path expects (normally the GRO path does the fixup).
This patch adds the necessary call to vlan_gro_frags.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Thanks,
Signed-off-by: David S. Miller <davem@davemloft.net>
Jarek Poplawski pointed out that my previous fix is broken for
VLAN+netpoll as if netpoll is enabled we'd end up in the normal
receive path instead of the VLAN receive path.
This patch fixes it by calling the VLAN receive hook.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
As my netpoll fix for net doesn't really work for net-next, we
need this update to move the checks into the right place. As it
stands we may pass freed skbs to netpoll_receive_skb.
This patch also introduces a netpoll_rx_on function to avoid GRO
completely if we're invoked through netpoll. This might seem
paranoid but as netpoll may have an external receive hook it's
better to be safe than sorry. I don't think we need this for
2.6.29 though since there's nothing immediately broken by it.
This patch also moves the GRO_* return values to netdevice.h since
VLAN needs them too (I tried to avoid this originally but alas
this seems to be the easiest way out). This fixes a bug in VLAN
where it continued to use the old return value 2 instead of the
correct GRO_DROP.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Protocols that use packet_type can be __read_mostly section for better
locality. Elminate any unnecessary initializations of NULL.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As analyzed by Patrick McHardy, vlan needs to reset it's
netdev_ops pointer in it's ->init() function but this
leaves the compat method pointers stale.
Add a netdev_resync_ops() and call it from the vlan code.
Any other driver which changes ->netdev_ops after register_netdevice()
will need to call this new function after doing so too.
With help from Patrick McHardy.
Tested-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The netpoll entry checks are required to ensure that we don't
receive normal packets when invoked via netpoll. Unfortunately
it only ever worked for the netif_receive_skb/netif_rx entry
points. The VLAN (and subsequently GRO) entry point didn't
have the check and therefore can trigger all sorts of weird
problems.
This patch adds the netpoll check to all entry points.
I'm still uneasy with receiving at all under netpoll (which
apparently is only used by the out-of-tree kdump code). The
reason is it is perfectly legal to receive all data including
headers into highmem if netpoll is off, but if you try to do
that with netpoll on and someone gets a printk in an IRQ handler
you're going to get a nice BUG_ON.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch optimises the Ethernet header comparison to use 2-byte
and 4-byte xors instead of memcmp. In order to facilitate this,
the actual comparison is now carried out by the callers of the
shared dev_gro_receive function.
This has a significant impact when receiving 1500B packets through
10GbE.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Base versions handle constant folding now.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unfortunately simplicity isn't always the best. The fraginfo
interface turned out to be suboptimal. The problem was quite
obvious. For every packet, we have to copy the headers from
the frags structure into skb->head, even though for 99% of the
packets this part is immediately thrown away after the merge.
LRO didn't have this problem because it directly read the headers
from the frags structure.
This patch attempts to address this by creating an interface
that allows GRO to access the headers in the first frag without
having to copy it. Because all drivers that use frags place the
headers in the first frag this optimisation should be enough.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently VLAN still has a bit of common code handling the aftermath
of GRO that's shared with the common path. This patch moves them
into shared helpers to reduce code duplication.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
In previous kernels, any kernel module could get access to the
'real-device' and the VLAN-ID for a particular VLAN. In more recent
kernels, the code was restructured such that this is hard to do
without accessing private .h files for any module that cannot use
GPL-only symbols.
Attached is a patch to once again allow non-GPL modules the ability to
access the real-device and VLAN id for VLANs.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case the real device has a neigh_setup function, this
neigh_setup function should be used for the vlan device.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds GRO interfaces for hardware-assisted VLAN reception.
With this in place we're now at parity with LRO as far as the
interface is concerned. That is, you can now take any LRO driver
and convert it over to GRO.
As the CB memory clashes with GRO's use of CB, I've removed it
entirely by storing dev in skb->dev. This is OK because VLAN
gets called first thing in netif_receive_skb and skb->dev is
not used in between us calling netif_rx and netif_receive_skb
getting called.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 656299f706
(vlan: convert to net_device_ops) added a net_device_ops
with a NULL ndo_start_xmit field.
This gives a crash in dev_hard_start_xmit()
Fix it using two net_device_ops structures, one for hwaccel vlan,
one for non hwaccel vlan.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert vlan devices and function pointers to net_device_ops.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The changes to deliver hardware accelerated VLAN packets to packet
sockets (commit bc1d0411) caused a warning for non-NAPI drivers.
The __vlan_hwaccel_rx() function is called directly from the drivers
RX function, for non-NAPI drivers that means its still in RX IRQ
context:
[ 27.779463] ------------[ cut here ]------------
[ 27.779509] WARNING: at kernel/softirq.c:136 local_bh_enable+0x37/0x81()
...
[ 27.782520] [<c0264755>] netif_nit_deliver+0x5b/0x75
[ 27.782590] [<c02bba83>] __vlan_hwaccel_rx+0x79/0x162
[ 27.782664] [<f8851c1d>] atl1_intr+0x9a9/0xa7c [atl1]
[ 27.782738] [<c0155b17>] handle_IRQ_event+0x23/0x51
[ 27.782808] [<c015692e>] handle_edge_irq+0xc2/0x102
[ 27.782878] [<c0105fd5>] do_IRQ+0x4d/0x64
Split hardware accelerated VLAN reception into two parts to fix this:
- __vlan_hwaccel_rx just stores the VLAN TCI and performs the VLAN
device lookup, then calls netif_receive_skb()/netif_rx()
- vlan_hwaccel_do_receive(), which is invoked by netif_receive_skb()
in softirq context, performs the real reception and delivery to
packet sockets.
Reported-and-tested-by: Ramon Casellas <ramon.casellas@cttc.es>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The generic packet receive code takes care of setting
netdev->last_rx when necessary, for the sake of the
bonding ARP monitor.
Signed-off-by: David S. Miller <davem@davemloft.net>
This enables more ethtool information. The speed and settings of the
underlying device are propagated up. This makes services like SNMP that
use ethtool to get speed setting, work when managing a vlan, without adding
silly heurtistics into SNMP daemon.
For the driver info, just use existing driver strings.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vlan devices are not reading the gso max size of the parent device. As
a result devices that do not support 64K max gso size are currently
failing.
This issue is seen on 2.6.26 kernels as well and the same patch should be
able to be applied without any issues.
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>
The new address list lock needs to handle the same device layering
issues that the _xmit_lock one does.
This integrates work done by Patrick McHardy.
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>
Remove excessive comments and debugging, use NETDEV_TX codes,
remove some empty lines.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove some debugging and excessive comments, merge the two
dev_hard_header calls into one.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow to query LRO settings of underlying device when VLAN RX
acceleration is used.
Suggested by Ben Hutchings <bhutchings@solarflare.com>.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
When VLAN header stripping is used, packets currently bypass packet
sockets (and other network taps) completely. For locally existing
VLANs, they appear directly on the VLAN device, for unknown VLANs
they are silently dropped.
Add a new function netif_nit_deliver() to deliver incoming packets
to all network interface taps and use it in __vlan_hwaccel_rx() to
make VLAN packets visible on the underlying device.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
dev_set_promiscuity/allmulti might overflow.
Commit: "netdevice: Fix promiscuity and allmulti overflow" in net-next makes
dev_set_promiscuity/allmulti return error number if overflow happened.
Here, we check all positive increment for promiscuity and allmulti
to get error return.
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Accesses are mostly structured such that when there are multiple TX
queues the code transformations will be a little bit simpler.
Signed-off-by: David S. Miller <davem@davemloft.net>
- vlan_dev_reorder_header() is only called on the receive path after
calling skb_share_check(). This means we can use skb_cow() since
all we need is a writable header.
- vlan_dev_hard_header() includes a work-around for some apparently
broken out of tree MPLS code. The hard_header functions can expect
to always have a headroom of at least there own hard_header_len
available, so the reallocation check is unnecessary.
- __vlan_put_tag() can use skb_cow_head() to avoid the skb_unshare()
copy when the header is writable.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The VLAN code contains multiple spots that use tag, id and tci as
identifiers for arguments and variables incorrectly and they actually
contain or are expected to contain something different. Additionally
types are used inconsistently (unsigned short vs u16) and identifiers
are sometimes capitalized.
- consistently use u16 for storing TCI, ID or QoS values
- consistently use vlan_id and vlan_tci for storing the respective values
- remove capitalization
- add kdoc comment to netif_hwaccel_{rx,receive_skb}
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hide struct vlan_dev_info from drivers to prevent them from growing
more creative ways to use it. Provide accessors for the two drivers
that currently use it.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function is huge and included at least once in every VLAN acceleration
capable driver. Uninline it; to avoid having drivers depend on the VLAN
module, the function is always built in statically when VLAN is enabled.
With all VLAN acceleration capable drivers that build on x86_64 enabled,
this results in:
text data bss dec hex filename
6515227 854044 343968 7713239 75b1d7 vmlinux.inlined
6505637 854044 343968 7703649 758c61 vmlinux.uninlined
----------------------------------------------------------
-9590
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add ethtool support for querying the device for offload settings.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use simplified is_vlan_dev function.
Signed-off-by: Joonwoo Park <joonwpark81@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
