linux.git/drivers/net/macvlan.c, branch v2.6.31

drivers/net/macvlan.c: fix cloning of tagged VLAN interfaces

2009-06-11T09:32:39+00:00

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=13348

akpm: the reporter disappeared, so I typed it in again.

It is not possible to make clone of tagged VLAN interface to be used as
mac-based vlan interfave.

How reproducible:
Use any 802.1q tagged vlan interface, e.g. eth2.700 and clone it:

  ip link add link eth2.700 address 00:04:75:cb:38:09 macvlan0 type macvlan
  ip link set dev macvlan0 up
  ip addr add 10.195.1.1/24 dev macvlan0

So far, so good. Now try to ping anything via macvlan0:

  ping 10.195.1.2

Actual results:
For every attempted packet tx kernel writes to console:

------------[ cut here ]------------
WARNING: at net/8021q/vlan_dev.c:254 vlan_dev_hard_header+0x36/0x126 [8021q]()
Hardware name: M22ES
Modules linked in: arptable_filter arp_tables bridge veth macvlan arc4 ecb
ppp_mppe ppp_async crc_ccitt ppp_generic slhc autofs4 sunrpc 8021q garp stp
ipt_REJECT nf_conntrack_ipv4 nf_defrag_ipv4 xt_state nf_conntrack xt_tcpudp
x_tables dm_mirror dm_region_hash dm_log dm_multipath dm_mod sbs sbshc lp
floppy snd_intel8x0 joydev snd_seq_dummy snd_intel8x0m snd_ac97_codec
ide_cd_mod ac97_bus snd_seq_oss cdrom snd_seq_midi_event serio_raw snd_seq
snd_seq_device snd_pcm_oss snd_mixer_oss parport_pc snd_pcm parport battery
8139cp snd_timer i2c_sis96x ac button snd rtc_cmos rtc_core 8139too soundcore
rtc_lib mii i2c_core pcspkr snd_page_alloc pata_sis libata sd_mod scsi_mod ext3
jbd ehci_hcd ohci_hcd uhci_hcd [last unloaded: ip_tables]
Pid: 0, comm: swapper Tainted: G        W  2.6.29.3 #1
Call Trace:
 [] warn_slowpath+0x60/0x9f
 [] warn_slowpath+0x87/0x9f
 [] vlan_dev_hard_header+0x0/0x126 [8021q]
 [] vlan_dev_hard_header+0x36/0x126 [8021q]
 [] vlan_dev_hard_header+0x0/0x126 [8021q]
 [] macvlan_hard_header+0x3c/0x47 [macvlan]
 [] macvlan_hard_header+0x0/0x47 [macvlan]
 [] arp_create+0xef/0x1ff
 [] arp_send+0x3d/0x54
 [] arp_solicit+0x16c/0x177
 [] neigh_timer_handler+0x227/0x269
 [] neigh_timer_handler+0x0/0x269
 [] run_timer_softirq+0xf0/0x141
 [] __do_softirq+0x76/0xf8
 [] __do_softirq+0x0/0xf8
   [] handle_level_irq+0x0/0xad
 [] irq_exit+0x35/0x62
 [] do_IRQ+0xdf/0xf4
 [] common_interrupt+0x27/0x2c
 [] default_idle+0x2a/0x3d
 [] cpu_idle+0x57/0x70

Macvlan driver always uses standard ethernet header length for all types
of interface to which it is linked.  This patch fixes this problem.

Reported-by: 
Cc: Alan Cox 
Cc: "Rafael J. Wysocki" 
Cc: "David S. Miller" 
Cc: Stephen Hemminger 
Signed-off-by: Andrew Morton 
Signed-off-by: David S. Miller

net: convert unicast addr list

2009-05-30T05:12:32+00:00

This patch converts unicast address list to standard list_head using
previously introduced struct netdev_hw_addr. It also relaxes the
locking. Original spinlock (still used for multicast addresses) is not
needed and is no longer used for a protection of this list. All
reading and writing takes place under rtnl (with no changes).

I also removed a possibility to specify the length of the address
while adding or deleting unicast address. It's always dev->addr_len.

The convertion touched especially e1000 and ixgbe codes when the
change is not so trivial.

Signed-off-by: Jiri Pirko 

 drivers/net/bnx2.c               |   13 +--
 drivers/net/e1000/e1000_main.c   |   24 +++--
 drivers/net/ixgbe/ixgbe_common.c |   14 ++--
 drivers/net/ixgbe/ixgbe_common.h |    4 +-
 drivers/net/ixgbe/ixgbe_main.c   |    6 +-
 drivers/net/ixgbe/ixgbe_type.h   |    4 +-
 drivers/net/macvlan.c            |   11 +-
 drivers/net/mv643xx_eth.c        |   11 +-
 drivers/net/niu.c                |    7 +-
 drivers/net/virtio_net.c         |    7 +-
 drivers/s390/net/qeth_l2_main.c  |    6 +-
 drivers/scsi/fcoe/fcoe.c         |   16 ++--
 include/linux/netdevice.h        |   18 ++--
 net/8021q/vlan.c                 |    4 +-
 net/8021q/vlan_dev.c             |   10 +-
 net/core/dev.c                   |  195 +++++++++++++++++++++++++++-----------
 net/dsa/slave.c                  |   10 +-
 net/packet/af_packet.c           |    4 +-
 18 files changed, 227 insertions(+), 137 deletions(-)
Signed-off-by: David S. Miller

net: release dst entry in dev_hard_start_xmit()

2009-05-19T05:19:19+00:00

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 
Signed-off-by: David S. Miller

net: factor out ethtool invocation of vlan/macvlan drivers

2009-04-21T09:00:51+00:00

Signed-off-by: Patrick McHardy 
Signed-off-by: David S. Miller

vlan/macvlan: fix NULL pointer dereferences in ethtool handlers

2009-04-17T22:59:23+00:00

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 
Signed-off-by: Patrick McHardy 
Signed-off-by: David S. Miller

macvlan: Deterministic ingress packet delivery

2009-03-13T20:16:13+00:00

Changing the mac address when a macvlan device is up will leave the
device on the wrong hash chain making it impossible to receive
packets.

There is no checking of the mac address set on the macvlan.  Allowing
a misconfiguration to grab packets from the the underlying device or
another macvlan.

To resolve these problems I update the hash table of macvlans when the
mac address of a macvlan changes, and when updating the hash table
I verify that the new mac address is usable.

The result is well defined and predictable if not perfect handling of
mac vlan mac addresses.

To keep the code clear I have created a set of hash table maintenance
in macvlan so I am not open coding the hash function and the logic
needed to update the hash table all over the place.

Signed-off-by: Eric Biederman 
Acked-by: Patrick McHardy 
Signed-off-by: David S. Miller

macvlan: Support creating macvlans from macvlans

2009-03-13T20:15:37+00:00

When running in a network namespace whose only link to
the outside world is a macvlan device, not being
able to create another macvlan is a real pain.

So modify macvlan creation to allow automatically forward
a creation of a macvlan on a macvlan to become a creation
of a macvlan on the underlying network device.

Signed-off-by: Eric Biederman 
Acked-by: Patrick McHardy 
Signed-off-by: David S. Miller

Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

2008-12-03T03:50:27+00:00

Conflicts:

	drivers/net/ixgbe/ixgbe_main.c
	drivers/net/smc91x.c

macvlan: don't broadcast PAUSE frames to macvlan devices

2008-11-26T23:30:48+00:00

PAUSE frames are only relevant for the real device, broadcasting them
to all macvlan devices can cause a significant load increase.

Reported-by: Ben Greear 
Tested-by: Ben Greear 
Signed-off-by: Patrick McHardy 
Signed-off-by: David S. Miller

netdev: add more functions to netdevice ops

2008-11-21T04:14:53+00:00

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 
Signed-off-by: David S. Miller