Attached is the patch to implement module loadable parameters as per new API.
Following is the list of changes.

1. Used new module_param() API.
2. List of variables for tx_fifo_len and rx_ring_sz replaced with array.
3. Some of the module parameters which can be set thru setpci command have
   been removed, such as latency_timer, max_read_byte_cnt, max_split_transactions.
4. Other parameters which were felt to be not required, such as rx_prio,
   tx_prio have been removed.
5. Interrupt moderation parameters(such as tx_urange_*) are no longer module
   loadable parameters since they can be configured thru' a separate patch
   available to customers.
6. Changed default max_read_byte_count to 1024.
7. If scatter-gather is enabled, checksum is enabled too.
8. Not verifying if module loadable parameters are within valid range
   (verify_load_param() removed).
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

This patch addresses the comments by Chris Leech about
skb->mac.ethernet resulting in NULL dereference with the old method
of implementing 2 buffer mode. The new method performs a copy of the
MAC header to the head of the payload. This is a stop-gap measure till
the fragmented skb receive feature in the kernel is made functional.

Also, using GFP_KERNEL flag for buffer0, buffer1 memory allocation
instead of GFP_ATOMIC.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

The attached patch incorporates Jeff's comments related to creating
separate functions for restarting the NIC(without using close and
open entry points) and few other comments. Complete list of changes
are as follows:

1. Two new functions s2io_card_down() and s2io_card_up() are defined
and are called during reset procedure instead of close and open
routines.
2. tasklet_status field is now made as unsigned long.
3. On getting serious error, queue is stopped before resetting the card.
4. Removed the check for "queue stopped" in xmit routine.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

Attached is the patch for implementing 2-buffer mode on Rx path.

On certain systems when a DMA has to happen on an un-aligned memory
location performance will take a significant hit. It's standard
practice to offset the Rx buffer address by 2 (as Mac header is 14
bytes) so the IP header starts from an aligned location.  Obviously
using a single Rx buffer both cannot be achieved. Thus XFrame supports
something called 2 buffer Rx mode, where in the Rx'ed frame is split
into 2 parts, one is the Ethernet header and the other is the Ethernet
payload. So now we can allocate proper aligned memory for both buffers,
hence the DMA is not slowed down. Also, the Ethernet payload(starting
from L3 header) is on an aligned location so OS need not have to do
un-aligned accesses to process IP header. To achieve this, the kernel
function eth_type_trans functionality has to be partially implemented
in the driver itself.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. When processing Rx packets, making sure that get offset of ring
does not cross the put offset.

2. when NAPI is not in use a new spinlock(put_lock) is used to make
sure accessing put offset of ring is atomic.

3. Also introduced a new vaiable put_pos in nic_t to keep track of
absolute position of the put pointer of Rx ring.

4. When NAPI is used, fill_rx_buffer is not called from the interrupt
handler(s2io_isr) .

5. In s2io_poll, decrementing packets processed is done inside the
while loop unlike out side it as was being done last time.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

The attached patch contains a modified scheme for allocating Tx descriptor
blocks.
More description follows.

In the old scheme, the entire Tx descriptor space was allocated in
one go. This could cause driver load to fail on systems with low(or
scattered) memory. The Tx descriptor blocks are now allocated on
per-page basis. A new structure (list_info) has been introduced in
nic_t structure to keep track of the physical and virtual addresses
of every TxD allocated this way.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. Max Txds per List.

2. statistics refresh time

3. pause frame control parameters including gap between two successive
frames, threshold watermarks

4. RTI and TTI configuration parameters including ranges, packet
counts and timeout periods.  For further information please read the
section 3.5 of XFrame H/W spec.

5. PCI/PCI-X configuration variables latency_timer, MMRBC and OST.

6. OS offload features TSO (If support available) and checksum offload.

7. If NAPI is not in use, a variable indicate_max_pkts can be used
to limit number of Rx side packets processed for one call to Rx
Intr handler.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. Xena3's with a set of subsystem IDs had Link LED problems, fixed
that specifically for them.

2. To write into the Keyed Mac_Cfg register to enable broadcast,
writing two 32 bit writes into it along with a write to the key
register rather than a single write to key and a 64 bit write to
mac_cfg. This is necessary on 32 bit systems where a writeq(64 bit
write) is actually two writel (32 bit writes).

3. Writes to some special registers mentioned in UG is being done by
a special macro which defines which 32 bits of the 64 bit register
is to be written first. Again this applies only on 32 bit systems.

4. Configured pause frame related water marks and a shared_split
value which describes the Max TXDMA related split transaction that
can be used without giving room for the Rx transactions.

5. The mac_rmac_err_reg R1 register will be cleared in  the interrupt
handler itself rather than in the scheduled task as was being done
previously.

6. Even on PCC_FB_ECC error the card will be reset by disabling
adapter enable bit.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. Definitions of LOW and PANIC levels of the Rx buffers have changed.

2. In wait_for_cmd_complete there is no longer a writeq but just a
read and wait for strobe bit to reset.

3. In s2io_isr, the isr_lock has been done away with also the NICs
interrupt are no longer disabled explicitly on entering the interrupt
handler and re-enabled again before leaving it.

4. Also clearing the semaphore "tasklet_status" when exiting
erroneously from s2io_isr after failing fill_rx_buffer call.

5. The set/reset Tx Csum function through ethtool was added to the
ethtool_ops structure.

6. Added a Rx side error code in the rx_osm_handler function.

7. No longer stopping and waking Tx queue when link state changes in
s2io_link function.

8. removed the isr_lock spinlock from the s2io_nic structure.

9. changed parameters which determine thresholds(LOW and PANIC)
to replenish Rx buffers.
   This has been found to result in better performance.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. In free_rx_buffers clearing out RxDs not owned by Xena.

2. In alarm_intr_handler, when a serr error occurs, schedule a task
to reset the card rather than stopping Tx queue.

3. In s2io_close freeing IRQ before calling s2io_reset also added a
new call to flush queued tasks. This is not done if the s2io_close
itself is called from a queued task like s2io_restart_nic.

4. read_eeprom function has been changed such that data to be returned
is sent as an input argument and the return value represents a
pass/fail. The previous implementation as Randy had pointed out was
error prone as on failure it returned -1 which can be interpreted
as all ff's, so any data area which contained ff's in the eeprom was
likely to be treated as an error.

5. Added a flag "task_flag" to track if the call to s2io_close is
coming from the s2io_restart_nic function or from the ifconfig <I/F>
down called by user.

6. Moved register_netdev call from just after setting entry points
to the end of the s2io_init_nic function.

7. In s2io.h field added a new member into the s2io_nic structure
called "task_flag".
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

1. Indentation, change in comment styles, variable name changes etc.
2. Changed the value written to dtx_control register to force XGXS reset.
3. weight parameter(NAPI) changed to 90 for better performance.
Signed-off-by: Raghavendra Koushik <raghavendra.koushik@s2io.com>
Signed-off-by: Ravinandan Arakali <ravinandan.arakali@s2io.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

This patch kills the bogus radeonfb_read/write routines. In order to do so,
it adds a new member to fb_info, along with screen_base, which is screen_size,
indicating the mapped area. The default fb_read/write will now use that instead
of fix->smem_len if it is non-0, and radeonfb now sets it to the mapped size
of the framebuffer.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This is a resend of a patch sent in July and that got lost somewhat,
the "VSCR" register wasn't restored properly from the context on
load_up_altivec (typo), please apply the fix:
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into kernel.bkbits.net:/home/davem/net-2.6

into kernel.bkbits.net:/home/davem/sparc-2.6

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

Only send NETDEV_CHANGEADDR notifies for address and broadcast changes.
Notify is also sent out if only one of the 2 changes is successful.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

Am travelling, but this passed simple tests here.  If this isn't going
in, the current seqfile stuff should be ripped out; it's a mess.

/proc/net/ip_conntrack was changed over to seq_file.  However,
seq_file isn't a great fit (a linked list which is changing is not a
good candidate for seq file), and the conversion was done badly.

1) Don't do allocation: simply hand the pointer head of the correct chain.
2) Actually output the original tuple.
3) Lock only when actually traversing hash chain.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

With improvements by Francois Romieu and David
S. Miller.
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

	Appended is a 2nd version that uses nf_reset_debug.

- do not use skb_checksum_help in input path as ipvs can handle
incoming CHECKSUM_HW packets

- do not use skb_checksum_help in forwarding path

- claim that checksum is valid (CHECKSUM_NONE) when entering output
path for out->in packets

- do not reset/destroy the nfct in IP_VS_XMIT, the intention is to
reset the debugging field just to avoid log floods from nf_debug_ip_*
functions, it is known that the ipvs packets traverse other
hooks, eg. LOCAL_IN->LOCAL_OUT. Use nf_reset_debug instead of nf_reset.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>

Old debugging code we no longer need.
Signed-off-by: David S. Miller <davem@davemloft.net>

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

This fixes panic in rt6_device_match().

Well, rt->rt6i_idev is always set if it is dynamically allocated.
However, when we hit ip6_null_entry here, its rt6i_idev is NULL.
This patch is minimum fix to avoid the oops for now.
Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

From: Thomas Gleixner <tglx@linutronix.de>

1. If the fake 5513 id bit is not set by the BIOS we must have the 5518
id in the device table.

2. If the register remapping is not set by the BIOS then the enable bit
check in ide_pci_setup_ports will fail. It's safe to switch to the
remapping mode here. Keeping the not remapped mode would need quite big
changes AFAICS.

Forward ported to 2.6 by me.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@elka.pw.edu.pl>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This driver caused a _lot_ of warnings due to tons
of explicit casts to "uclong". Making all the types
sane not only removed the warnings, but got rid of
a lot of silly casting, since the types are now much
more natural to what the driver wanted to do in the
first place.

Forgot to add in the mmio base..

It disappeared somewhere in Al's cleanup patch..

into nuts.davemloft.net:/disk1/BK/sparc-2.6

le16_to_cpu() on 32bit field.
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>