It finally dawned on me how to eliminate the race I've been narrowing
down with earlier patches: Instead of writing the command registers as
one word, write them one at a time (as bytes). The race was for settings
bits in ChipCmd and ChipCmd1 (0x09) against the chip clearing CmdTxOn
which is in ChipCmd.

In addition to writing single bytes, the fix requires a switch from
using bit 5 in ChipCmd0 to bit 5 in ChipCmd1 (which is equivalent)
to signal Tx demand.

Also, don't restart Rx engine "pre-emptively" in rhine_rx, that's a
sure way to race with the chip.

Introduce RHINE_WAIT_FOR, a macro for small busy loops with primitive
completion checking.
Signed-off-by: Roger Luethi <rl@hellgate.ch>

Nobody complained although media locking parameters were broken
forever. They were sort of fixed recently, but the code is still in a
bad shape.

More seriously, the options/full_duplex stuff has fundamental design
problems that have been discussed in-depth on the list (e.g. effect of
hotplugging, nameif, suspend/resume).

For those needing media locking with Linux 2.6+ via-rhine, ethtool(8)
is the replacement.
Signed-off-by: Roger Luethi <rl@hellgate.ch>

Instead of probing, set phy_id to 1 for Rhine III and read phy_id from
EEPROM-controlled register for Rhine I/II.

Remove code for handling anything other than 1 MII PHY. If it wasn't
unnecessary code to begin with, it would need to be fixed because it
wouldn't work.

Use mii_if_info.phy_id as the only container of phy_id. Not particulary
happy about those names (phy_id vs. MII_PHYSIDx), but being consequent
about it mitigates confusion.
Signed-off-by: Roger Luethi <rl@hellgate.ch>

All this code is broken (e.g. unconditionally programs all PHYs as if
they were the same model) and/or unused (IntrLinkChange never occurs
with driver as is).
Signed-off-by: Roger Luethi <rl@hellgate.ch>

A.J. from VIA Networking Technologies noticed that via-rhine is using
cpu_to_le32() when preparing mc_filter hashes. This breaks Rhine hardware
multicast filters on big-endian architectures.
Signed-off-by: Roger Luethi <rl@hellgate.ch>

Restructure code to make it easier to maintain.

rhine_hw_init: resets chip, reloads eeprom
rhine_chip_reset: chip reset + what used to be wait_for_reset
rhine_reload_eeprom: reload eeprom, re-enable MMIO, disable EEPROM-controlled
	WOL

Note: Chip reset clears a bunch of registers that should be reloaded
from EEPROM (which turns off MMIO). Deal with that later.
Signed-off-by: Roger Luethi <rl@hellgate.ch>

This contains hpfs fixes from my source ported to 2.6.7 kernel:

updates:
- allocator fragments files less
- OS/2 Warp Server filesystem can be mounted read/only
- added reschedule points so that it doesn't hog CPU

bug fixes:
- filesystem error message when syncing or fsyncing deleted file (or when
  system just writes it on its own)
- filesystem error on extremly fragmented files
- corrupted disk structures could possibly corrupt memory

drivers/message/fusion/mptctl.c:89:24: asm/compat.h: No such file or directory
Signed-off-by: Andrew Morton <akpm@osdl.org>

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

In base kernel 2.6.7, there is bug in function hugetlb_prefault().  The
call to free_huge_page() is incorrect and should be replaced with a call
to page_cache_release().
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

This patch moves the pfm_ovfl_arg_t structure from the kernel stack
into the pfm_context_t structure. This minimizes kernel stack space
consumption.
Signed-off-by: Stephane Eranian <eranian@hpl.hp.com>
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

This patch moves move_irq() from ioaspic.c to irq.c to make this common
for ia64 subarches.
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

- Handle the overlapping mmaps in the MAP_FIXED case
- Handle unmaps correctly i.e. call __ia32_delete_pp_range()
- There are now three clear entry points into the pp manipulation code
        - ia32_set_pp
        - ia32_unset_pp
        - ia32_compare_pp
- Rename internal functions to __ia32_*pp_*
- Grab mmap_sem just once at the main entry point
Signed-off-by: Gordon Jin <gordon.jin@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

A classic..

  lib/string.c:165:19: warning: assignment expression in conditional

From: Mika Kukkonen <mika@osdl.org>
Signed-off-by: Randy Dunlap <rddunlap@osdl.org>

Add __user annotation for !CONFIG_MODULE_UNLOAD case.

From: Mika Kukkonen <mika@osdl.org>
Signed-off-by: Randy Dunlap <rddunlap@osdl.org>

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

 - added define of __x86_64__ to sparse arguments on amd64
 - switched amd64 uaccess.h to __chk_user_ptr()

removed bogus externs from declarations in fs/udf/*

misc annotation in drivers/video/* (nowhere near complete)

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

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

This enhances the network simulation scheduler to do simple random loss.

The loss parameter is a simple 32 bit value such that 0 means no loss, and
0xffffffff is always drop.  I have a new version of the tc command which takes
care of conversion from percent to this value.

Same patch for 2.4 and 2.6
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>

Based upon a patch from Stephen Hemminger.
Signed-off-by: David S. Miller <davem@redhat.com>