An incoming datagram must bring into cpu cache *lot* of cache lines,
in particular : (other parts omitted (hash chains, ip route cache...))

On 32bit arches :

offsetof(struct sock, sk_rcvbuf)       =0x30    (read)
offsetof(struct sock, sk_lock)         =0x34   (rw)

offsetof(struct sock, sk_sleep)        =0x50 (read)
offsetof(struct sock, sk_rmem_alloc)   =0x64   (rw)
offsetof(struct sock, sk_receive_queue)=0x74   (rw)

offsetof(struct sock, sk_forward_alloc)=0x98   (rw)

offsetof(struct sock, sk_callback_lock)=0xcc    (rw)
offsetof(struct sock, sk_drops)        =0xd8 (read if we add dropcount support, rw if frame dropped)
offsetof(struct sock, sk_filter)       =0xf8    (read)

offsetof(struct sock, sk_socket)       =0x138 (read)

offsetof(struct sock, sk_data_ready)   =0x15c   (read)


We can avoid sk->sk_socket and socket->fasync_list referencing on sockets
with no fasync() structures. (socket->fasync_list ptr is probably already in cache
because it shares a cache line with socket->wait, ie location pointed by sk->sk_sleep)

This avoids one cache line load per incoming packet for common cases (no fasync())

We can leave (or even move in a future patch) sk->sk_socket in a cold location
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Version Update.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Allow multiple functions with INTA.
- Removed the condition to allow only one vpath with INTA
- Ensure that the alarm bit in titan_mask_all_int register is cleared when
  driver exits.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Added a function to check if FCS stripping is disabled by the firmware, if
  it is not disabled fail driver load.

- By default FCS stripping is disabled by the firmware. With this assumption
  driver decrements the indicated packet length by 4 bytes(FCS length).

- This patch ensures that FCS stripping is disabled during driver load time.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Removed the wrr_rebalance function

- This feature is not supported by the ASIC, hence removing the related code.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Fix a crash in PAE system due to wrong typecasting.

- On PAE system size_t is unsigned int which is 32bit. Avoid casting
  64 bit address to 32 bit
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Removed accessing GENDMA_INT register
 - This allowed the firmware to perform a generic DMA write to host memory.
   This feature is not supported by the ASIC, this patch removes access to
   GENDMA_INT register.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

- Update driver_config->vpath_per_dev for each function in probe.

- vpath_per_device specifies number of vpaths supported for each function/device. The
  current code was updating vpath_per_device only for physical device, however this has
  to be updated for each function also in case of a MF(Multi function) device.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxge: Modify __vxge_hw_device_is_privilaged() to not assume function-0 as the privileged function: Resubmit#1

- vxge driver was assuming function-0 is always the privilaged function. Now that
  restriction has been removed any function can act as a privilaged function.

- This patch modifies the __vxge_hw_device_is_privilaged routine to not assume
  function-0 as the privileged function.

- Recreated the patch by incorporating review comments from Dave Miller to
  remove double slash in path names.
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There are some switches that will do write combining when they see two
sequential regions written. In order to avoid any possible write combining
issues it is necessary to add a flush after writing each piece of a rar
register.
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>

This update adds additional exception handling to the phy code to handle
situations where it may be called incorrectly.  In addition it adds some
bounds checking to the cable length checks to prevent an array overrun in
the event that the hardware returned a different value than expected.
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>

This patch adds a retry to phy reads in the event of failure.  The original
code broke out of the loop on failure and this is a mistake as we should be
trying to do the read twice.
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>

This patch moves the generic portion of the copper link setup into a
seperate function in e1000_phy.c.  This helps to reduce the size of
copper_link_setup_82575 and make it a bit more readable.
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>

igb doesn't have any devices that use a microwire interface for NVM.  As
such the code related to this can be removed.
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>

This patch allows us to treat the alternate mac address as though it is the
physical address on the adapter.  This is accomplished by letting the
alt_mac_address function to only fail on an NVM error.  If no errors occur
and the alternate mac address is not present then RAR0 is read as the
default mac address.
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 mailbox timeout routines need to be updated as they were not correctly
handling the case of a mailbox timeout and could cause issues with long
delays when used.
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>

Both the read and write mailbox functions need to acquire the mailbox lock.
Since that is the case we might as well combine both of the procedures into
one function so it is easier to maintain.
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>

This patch adds the device ID necessary to support the 82576NS SerDes
adapter.
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>

In order to support functions such as vlan tag stripping when SR-IOV is
enabled any given packet must match at least one filter.  However in the
case of promiscous mode being enabled on the PF the traffic routed to it
may not match any filters and is just sent to the PF by default.  In order
to make certain that this traffic is processed we can set all bits in the
UTA registers to create a pseudo promiscous mode filter that accepts all
packets.
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>

This patch adds igb_rar_qsel which sets the mac address and pool bits for a
given mac address in the receive address register table.
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 current implementation of sgmii support isn't correctly locking the
interfaces for reads/writes.  This change pulls the read/write
functionality out of 82575.c and moves it to phy.c.  In addition it
replaces the implementation in 82575.c with one that uses locking around
the relocated i2c interface calls.
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 current approach is just using a ?: type mechanism to set the phy
locking bit.  This if fine for now but limits us to only 2.  Switch to a
nested if statement for future compatiblity with more than 2 phys.
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>

This update corrects the driver so that it handles duplex for serdes links
correctly instead of just forcing full duplex always.
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>

There is a temp variable in the stats clearing path that isn't needed since
the results from the stats read can be immediately discared.  Since it
isn't needed we might as well just drop it from the function call.
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>

This patch exports the link-speed (in Mbps) and duplex of an interface
via sysfs.  This eliminates the need to use ethtool just to check the
link-speed.  Not requiring 'ethtool' and not relying on the SIOCETHTOOL
ioctl should be helpful in an embedded environment where space is at a
premium as well.

NOTE: This patch also intentionally allows non-root users to check the link
speed and duplex -- something not possible with ethtool.

Here's some sample output:

# cat /sys/class/net/eth0/speed
100
# cat /sys/class/net/eth0/duplex
half
# ethtool eth0
Settings for eth0:
        Supported ports: [ TP ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
                                1000baseT/Half 1000baseT/Full
        Supports auto-negotiation: Yes
        Advertised link modes:  Not reported
        Advertised auto-negotiation: No
        Speed: 100Mb/s
        Duplex: Half
        Port: Twisted Pair
        PHYAD: 1
        Transceiver: internal
        Auto-negotiation: off
        Supports Wake-on: g
        Wake-on: g
        Current message level: 0x000000ff (255)
        Link detected: yes
Signed-off-by: David S. Miller <davem@davemloft.net>

Instead of having to modify every non-mac80211 for device type assignment,
do this inside the netdev notifier callback of cfg80211. So all drivers
that integrate with cfg80211 will export a proper device type.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

For various purposes including a wireless extensions
bugfix, we need to hook into the netdev creation before
before netdev_register_kobject(). This will also ease
doing the dev type assignment that Marcel was working
on for cfg80211 drivers w/o touching them all.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

For usbnet devices with FLAG_WLAN and FLAG_WWAN set the proper device
type so that uevent contains the correct value. This then allows an easy
identification of the actual underlying technology of the Ethernet device.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for usbnet based devices like CDC-Ether to indicate that they
are actually mobile broadband devices. In that case use wwan%d as default
interface name.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

We currently dirty a cache line to update tunnel device stats
(tx_packets/tx_bytes). We better use the txq->tx_bytes/tx_packets
counters that already are present in cpu cache, in the cache
line shared with txq->_xmit_lock

This patch extends IPTUNNEL_XMIT() macro to use txq pointer
provided by the caller.

Also &tunnel->dev->stats can be replaced by &dev->stats
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The FIB algorithim for IPV4 is set at compile time, but kernel goes through
the overhead of function call indirection at runtime. Save some
cycles by turning the indirect calls to direct calls to either
hash or trie code.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add Ancilliary data to better represent loss information

I've had a few requests recently to provide more detail regarding frame loss
during an AF_PACKET packet capture session.  Specifically the requestors want to
see where in a packet sequence frames were lost, i.e. they want to see that 40
frames were lost between frames 302 and 303 in a packet capture file.  In order
to do this we need:

1) The kernel to export this data to user space
2) The applications to make use of it

This patch addresses item (1).  It does this by doing the following:

A) Anytime we drop a frame for which we would increment po->stats.tp_drops, we
also no increment a stats called po->stats.tp_gap.

B) Every time we successfully enqueue a frame to sk_receive_queue, we record the
value of po->stats.tp_gap in skb->mark.  skb->cb would nominally be the place to
record this, but since all the space there is used up, we're overloading
skb->mark.  Its safe to do since any enqueued packet is guaranteed to be
unshared at this point, and skb->mark isn't used for anything else in the rx
path to the application.  After we record tp_gap in the skb, we zero
po->stats.tp_gap.  This allows us to keep a counter of the number of frames lost
between any two enqueued packets

C) When the application goes to dequeue a frame from the packet socket, we look
at skb->mark for that frame.  If it is non-zero, we add a cmsg chunk to the
msghdr of level SOL_PACKET and type PACKET_GAPDATA.  Its a 32 bit integer that
represents the number of frames lost between this packet and the last previous
frame received.

Note there is a chance that if there is frame loss after a receive, and then the
socket is closed, some gap data might be lost.  This is covered by the use of
the PACKET_AUXDATA socket option, which gives total loss data.  With a bit of
math, the final gap can be determined that way.

I've tested this patch myself, and it works well.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>

 include/linux/if_packet.h |    2 ++
 net/packet/af_packet.c    |   33 +++++++++++++++++++++++++++++++++
 2 files changed, 35 insertions(+)
Signed-off-by: David S. Miller <davem@davemloft.net>

Use BIT for macro definitions wherever possible, remove
unused and redundant macros.
Signed-off-by: Chaithrika U S <chaithrika@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can avoid two atomic ops on skb->users if packet is not going to be
sent to the device (because hardware txqueue is full)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can make icmp messages tx completion callback a litle bit faster.

Setting SOCK_USE_WRITE_QUEUE sk flag tells sock_wfree() to
not call sk_write_space() on a socket we know no thread is posssibly
waiting for write space. (on per cpu kernel internal icmp sockets only)

This avoids the sock_def_write_space() call and
read_lock(&sk->sk_callback_lock)/read_unlock(&sk->sk_callback_lock) calls
as well.

We avoid three atomic ops.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The in-tree implementations have all been converted to
get_sset_count().
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

These string query operations were supposed to be replaced by the
generic get_sset_count() starting in 2007.  Convert the remaining
implementations.

Also remove calls to these operations to initialise drvinfo->n_stats.
The ethtool core code already does that.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This string query operation was supposed to be replaced by the
generic get_sset_count() starting in 2007.  Convert tehuti's
implementation.

Also remove the dummy self-test name which was not used since tehuti
does not advertise any self-tests.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This string query operation was supposed to be replaced by the
generic get_sset_count() starting in 2007.  Convert qeth's
implementation.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

I'm skipping -rc2 because the -rc1 Makefile mistakenly said -rc2, so in
order to avoid confusion, I'm jumping from -rc1 to -rc3.  That way, when
'uname' (or an oops report) says 2.6.32-rc2, there's no confusion about
whether people perhaps meant -rc1 or -rc2.