This patch increases the "sequence gap" field of the LINK_PROTOCOL
message header from 8 bits to 13 bits (utilizing 5 previously
unused 0 bits).  This ensures that the field is big enough to
indicate the loss of up to 8191 consecutive messages on the link,
thereby accommodating the current worst-case scenario of 4000
lost messages.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch ensures that the display code that traverses the
publication lists belonging to a name table entry take its
associated spinlock, to protect against a possible change to
one of its "head of list" pointers caused by a simultaneous
name table lookup operation by another thread of control.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds a check to prevent TIPC's name table display code
from listing a name type entry if it exists only to hold subscription
info, rather than published names.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch eliminates the rarely-used "error code" argument
when initializing a TIPC message header, since the default
value of zero is the desired result in most cases; the few
exceptional cases now set the error code explicitly.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch eliminates a case where TIPC's link code could try reading
a field that is not present in a short message header.  (The random
value obtained was not being used, but the read operation could result
in an invalid memory access exception in extremely rare circumstances.)
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch enhances TIPC's handler for incoming messages in two
ways:
- the trivial, single-use routine for processing non-sequenced
  messages has been merged into the main handler
- the interface that received a message is now identified without
  having to access and/or modify the associated sk_buff
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces a new, out-of-range value to indicate that
a link endpoint does not have an existing session established
with its peer, eliminating the risk that the previously used
"invalid session number" value (i.e. zero) might eventually be
assigned as a valid session number and cause incorrect link
behavior.

The patch also introduces explicit bit masking when assigning a
new link session number to ensure it does not exceed 16 bits.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch corrects two problems in the display of error code
information in TIPC messages when debugging:
- no longer tries to display error code in NAME_DISTRIBUTOR
  messages, which don't have the error field
- now displays error code in 24 byte data messages, which do
  have the error field
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch re-orders & re-groups the error checks performed on
messages being delivered to native API ports, in order to clarify the
similarities and differences required for the various message types.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fixes a bug that prevented TIPC from receiving a
connection setup request message on a native TIPC port.
The revised connection setup logic ensures that validation
of the source of a connection-based message is skipped if
the port is not yet connected to a peer.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use netdev_alloc_skb. This sets skb->dev and allows arch specific
allocation.

Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Use netdev_alloc_skb for rx buffer allocation. This sets skb->dev
and can be overriden for NUMA machines.

This device is PowerPC only, so not tested or compiled.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: Ramkrishna Vepa <ram.vepa@neterion.com>
Cc: Rastapur Santosh <santosh.rastapur@neterion.com>
Cc: Sivakumar Subramani <sivakumar.subramani@neterion.com>
Cc: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Turn on special bits to save more power when device is shutdown.
Tested on a limited range of hardware, some of the bits are for hardware
that probably isn't even in production (like Yukon Supreme) and was ported
from the vendor driver.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Put PHY int sleep mode (from vendor sk98lin 10.50 driver) when the
network device is brought down.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Later changes add more code to PHY power changes so refactor now.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Match the suspend/resume code ordering in e100/e1000e more closely.
For example the configuration space should be saved on suspend even for
devices that are not up.
Signed-off-by: Tobias Diedrich <ranma+kernel@tdiedrich.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

The memory mapped device configuration space is lost during hibernate.
Save and restore it (fixes 'swapped mac' problem).
Signed-off-by: TTobias Diedrich <ranma+kernel@tdiedrich.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

When hibernating in 'shutdown' mode, after saving the image the suspend hook
is not called again.
However, if the device is in promiscous mode, wake-on-lan will not work.
This adds a shutdown hook to setup wake-on-lan before the final shutdown.
Signed-off-by: Tobias Diedrich <ranma+kernel@tdiedrich.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Port the fs_enet driver to support the MDIO on GPIO driver for PHY access
in addition to the mii-bitbang driver.
Signed-off-by: Laurent Pinchart <laurentp@cse-semaphore.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

This patch adds an MDIO bitbang driver that uses the GPIO library and its
OF bindings to access the bus I/Os.
Signed-off-by: Laurent Pinchart <laurentp@cse-semaphore.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Remove our own implementation of I2C bit-banging.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

cxgb3 uses lro_* functions and selects INET_LRO, but this doesn't help unless
INET is already enabled, so make the driver depend on INET also.

sge.c:(.text+0x9f09a): undefined reference to `lro_flush_all'
sge.c:(.text+0x9f62f): undefined reference to `lro_receive_skb'
sge.c:(.text+0x9f8a3): undefined reference to `lro_receive_frags'
sge.c:(.text+0x9fbe0): undefined reference to `lro_vlan_hwaccel_receive_skb'
sge.c:(.text+0x9ffcd): undefined reference to `lro_vlan_hwaccel_receive_frags'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Make driver more readable on standard 80 col windows.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Handle shared IRQ correctly. If IRQ is shared, it typically will show up
as an IRQ with an empty status field. So check in driver and handle it
without crapping out with invalid interrupt message.

Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Rx allocation failure at runtime is non-fatal. For normal Rx frame, it
just reuses the buffer, and during setup it just continues with a smaller
receive buffer pool.

Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Make this driver compile cleanly on 64 bit platforms.
Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Use new netdevice common stats area.

Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Use skb_padto to pad frames, this avoid allocation of separate buffer just
for dma of the extra bytes.

Compile tested only.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Add I/O AT DCA (Direct Cache Access) support.
Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Andrew Gallatin <gallatin@myri.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Add multi-slice/MSI-X support. By default, a single slice
(and the normal firmware) are used.  To enable msi-x, multi-slice
mode, one must load the driver with myri10ge_max_slices set to
either -1, or something larger than 1.
Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Andrew Gallatin <gallatin@myri.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Add several routines that multislices support will use.
Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Andrew Gallatin <gallatin@myri.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Merge branch 'upstream-next-davem' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/netdev-2.6

I tried to group recovery related fields nearby (non-CA_Open related
variables, to be more accurate) so that one to three cachelines would
not be necessary in CA_Open. These are now contiguously deployed:

  struct sk_buff_head        out_of_order_queue;   /*  1968    80 */
  /* --- cacheline 32 boundary (2048 bytes) --- */
  struct tcp_sack_block      duplicate_sack[1];    /*  2048     8 */
  struct tcp_sack_block      selective_acks[4];    /*  2056    32 */
  struct tcp_sack_block      recv_sack_cache[4];   /*  2088    32 */
  /* --- cacheline 33 boundary (2112 bytes) was 8 bytes ago --- */
  struct sk_buff *           highest_sack;         /*  2120     8 */
  int                        lost_cnt_hint;        /*  2128     4 */
  int                        retransmit_cnt_hint;  /*  2132     4 */
  u32                        lost_retrans_low;     /*  2136     4 */
  u8                         reordering;           /*  2140     1 */
  u8                         keepalive_probes;     /*  2141     1 */

  /* XXX 2 bytes hole, try to pack */

  u32                        prior_ssthresh;       /*  2144     4 */
  u32                        high_seq;             /*  2148     4 */
  u32                        retrans_stamp;        /*  2152     4 */
  u32                        undo_marker;          /*  2156     4 */
  int                        undo_retrans;         /*  2160     4 */
  u32                        total_retrans;        /*  2164     4 */

...and they're then followed by URG slowpath & keepalive related
variables.

Head of the out_of_order_queue always needed for empty checks, if
that's empty (and TCP is in CA_Open), following ~200 bytes (in 64-bit)
shouldn't be necessary for anything. If only OFO queue exists but TCP
is in CA_Open, selective_acks (and possibly duplicate_sack) are
necessary besides the out_of_order_queue but the rest of the block
again shouldn't be (ie., the other direction had losses).

As the cacheline boundaries depend on many factors in the preceeding
stuff, trying to align considering them doesn't make too much sense.

Commented one ordering hazard.

There are number of low utilized u8/16s that could be combined get 2
bytes less in total so that the hole could be made to vanish (includes
at least ecn_flags, urg_data, urg_mode, frto_counter, nonagle).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: David S. Miller <davem@davemloft.net>