Here is a rediff'd version of the patch to fix shared irq handling in pcnet_cs
and generic 8390 drivers.

Error path (EISA) probe was using device before set.

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
#	           ChangeSet	1.1567  -> 1.1568
#	  drivers/net/dgrs.c	1.23    -> 1.24
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 04/02/19	shemminger@osdl.org	1.1568
# fix bug in eisa error path.
# --------------------------------------------
#

* define dump stats complete constant
* add mwb() to keep order straight with HW when appending
  new resources to Rx and Tx queues
* increment stats for rx_dropped and rx_over_errors

* driver not handling failures of register_netdev
  [Stephen Hemminger (shemminger@osdl.org)]

* For half duplex, the IEEE says collision retry count should
  be 15 not 16.  Would cause really slow transfers on some hubs
  (i.e NetGear DS108).

* msec_delay macro wouldn't delay at all if x < 10, and
  wouldn't delay enough for x = 10.  Not a concern for
  2.6 (HZ=1000) but is for 2.4 (HZ=100).

* Multiple reports of system lock-ups with the CSA patch, so
  let's disable fix for now until we can understand why this
  is causing the hangs.  It was first thought that only pre-
  productions systems would hang, but that's not true.

* Disable TSO as default setting until some hangs/resets caused
  with TSO enabled are root-caused.  TSO can still be manually
  enabled using ethtool -K.

* Bug fix: flow control high/low watermark settings not
  within Rx FIFO range: math error.

Small patch to clean up 3c359_micrcode.h, no other drivers in the kernel
come anywhere near the file and the #if is superflous.

Mike Phillips

per-network-interface message logging standards for net drivers.

Written by Donald Becker.

into ppc970.osdl.org:/home/torvalds/v2.5/linux

From: Anton Blanchard <anton@samba.org>

There were a number of bugs in our cmd_line handling:

- We were looking at cmd_line after it had been tokenised. Use saved_command
  line instead
- Use strlcpy instead of memcpy/explicit NULL set.
- Remove magic numbers in iseries cmd_line parsing

From: Anton Blanchard <anton@samba.org>

UP compile fixes.

From: Anton Blanchard <anton@samba.org>

Restore the old cpu names in /proc/cpuinfo, some applications parse this
field.

From: Anton Blanchard <anton@samba.org>

Some trivial oops cleanups.

From: Anton Blanchard <anton@samba.org>

Set node 0 online, otherwise the recent cpu/memory topology stuff will oops
on a non NUMA machine.

From: Linda Xie <lxiep@us.ibm.com>

set err to -ENODEV when a new node doesn't have "interrupt" property.

From: Anton Blanchard <anton@samba.org>

__get_SP used to be a function call which meant we allocated a stack
frame before calling it. This meant the SP it returned was one frame
below the current function. Lets call that bogusSP (and the real one
SP).

The new dump_stack was being tail call optimised so it remained one
frame above bogusSP. dump_stack would then store below SP (as the ABI
allows us to) and would stomp over the back link that bogusSP pointed
to (__get_SP had set the back link up so it worked sometimes, just not
all the time).

Fix this by just making __get_SP an inline that returns the current SP.

From: Stephen Rothwell <sfr@canb.auug.org.au>

I think Ben introduced this new file and iSeries doesn't need it.

From: Anton Blanchard <anton@samba.org>

Another decent optimisation found in the ia64 port, dont take the irq
descriptor lock or do the note_interrupt stuff on PER_CPU irqs (ie IPIs).

From: Anton Blanchard <anton@samba.org>

Clean up per cpu usage in segment table code.

From: Anton Blanchard <anton@samba.org>

Make a number of segment table functions static.
Clean up one duplicated slbia.

From: Anton Blanchard <anton@samba.org>

Add cpus and NUMA memory nodes to sysfs. Also add cpu physical id.

From: Anton Blanchard <anton@samba.org>

The might_sleep infrastructure doesnt like our get_users in the backtrace
code, we often end up with might_sleep warnings inside might_sleep warnings.

Instead just be careful about pointers before dereferencing them.

Also remove the hack where we only printed the bottom 32bits of the WCHAN
value.

From: Anton Blanchard <anton@samba.org>

This adds stack overflow debugging much like x86. Its pretty agressive
and checks if > 8kB out of 16kB of stack is used.

From: Rusty Russell <rusty@rustcorp.com.au>

This cleans up the rtasd logic, and also makes it hotplug CPU safe.

From: Anton Blanchard <anton@samba.org>

smp_message_pass has a few args that are never used. Remove em.
Also add memory missing memory barrier in smp_call_function_interrupt.

From: Rusty Russell <rusty@rustcorp.com.au>

1) Check for cpuids >= NR_CPUS in prom_hold_cpus: we previously
   overflowed arrays when this happened, and checking it here ensures
   that it doesn't happen elsewhere.  Still move processors to
   secondary_hold (out of OF), but we won't release them.

2) smp_iSeries_message_pass and smp_xics_message_pass don't need to
   iterate if given a specific CPU: cleanup and fix.

3) kick_cpu variants don't need to check cpu range (caller is screwed
   anyway if this happened): replaced with BUG_ON for now.

4) smp_openpic_probe and smp_xics_probe can use cpus_weight(): it's
   faster and simpler.

5) User for_each_cpu/for_each_online_cpu as appropriate.

From: Hollis Blanchard <hollisb@us.ibm.com>

fix warning and compile error without CONFIG_SMP

From: Dave Boutcher <sleddog@us.ibm.com>

fix naming collision with asm-ppc64/vio.h. The define would be set there and
this file would not get pulled in.

From: Olaf Hering <olh@suse.de>

Fix for valid nvram rtas tokens.

From: Jake Moilanen <moilanen@austin.ibm.com>

Add 970FX entry into the cputable.

From: Anton Blanchard <anton@samba.org>

Be like x86 and enable -funit-at-a-time.
Use the check_gcc hook when enabling -mcpu=power4/-mtune=power4.