more reliable in testing, and works for all cases as near as
everyone can tell.

Contributor: Arjan @ RedHat

version 2.0.19, plus boolean cleanups.
Bump version to 2.0.20-pre1.

Contributors: Eli Kupermann @ Intel, Amir Noam @ Intel

v1.27   improved CSR/PROM address detection, lots of cleanups,
       new pcnet32vlb module option, HP-PARISC support,
       added module parameter descriptions, 
       initial ethtool support - Helge Deller <deller@gmx.de>
v1.27a  Sun Feb 10 2002 Go Taniguchi <go@turbolinux.co.jp>
       use alloc_etherdev and register_netdev
       fix pci probe not increment cards_found
       FD auto negotiate error workaround for xSeries250
       clean up and using new mii module

for the following net drivers:

Several ham radio, several IrDA, lp4863, pcnet32, saa9730,
wireless orinoco.

Idea is that if there is a valid name in MODULE_DESCRIPTION("...")
then the name of the hardware/driver should not be also repeated
in each MODULE_PARM_DESC("...").  MODULE_DESCRIPTION has been
added to essentially all the 8390 drivers.

All of the drivers changed are 8390 based, with the exception of
eepro100 and 3c509.

the second port on dual-port SK-9844 NICs.

08/15/01 - Added ioctl() functionality for debugging, changed netif_*_queue
           calls and other incorrectness - Kent Yoder <yoder1@us.ibm.com>
11/05/01 - Restructured the interrupt function, added delays, reduced the 
           the number of TX descriptors to 1, which together can prevent 
           the card from locking up the box - <yoder1@us.ibm.com>

Remove duplicated ethtool ioctl handling code, fixing build.

with public namespace.

Add new PCI API functions pci_set_mwi(), pci_clear_mwi().

(a few csets later on this is auto-merged away)

* Sparc64 support and fixes.
* Better stats and error handling.

Contributor forgotten :(

into mandrakesoft.com:/spare/repo/net-drivers-2.5

This is actually an attempt to remove some stall code from
this driver. However if some *real* users complain (Not just
the usuall: "Hey - if someone!" but the "Hey I'm using this!")
I'm all open to reenable it. Since I prepared this patch
yerstoday it doesn't contain the ide_module.h fixup. This will
follow later.

- Don't use the convoluted byte type in ide-pci.c. Just use the proper
   u8instead.

- Move ide_get_or_set_dma_base to the only place where it's used and
   reorganize the code there by killing the unnecessary
   CONFIG_BLK_DEV_IDEDMA_FORCED configuration option.

- Remove unfunctional CONFIG_PKT_TASK_IOCTL code.

- Kill unused ALTSTAT_SCREW_UP code.

- Tons of dead code removed from ide-taskfile.c (#if 0 #endif and
   friends)

- Remove unused IDE_DEBUG macro as well as lots of other name space
   pollution from ide.h.

- Start using the ide_lock spin-lock for protecting access to data
   structures instead of the excessive interrupt disabling games.

- Shorten the proc ouput of the piix initialization module.

- Remove special /proc tape "name" output from ide-tape.c. This was
   redundant data which should only show up on syslog anyway.

- Kill the REALLY_FAST_IO undef from the ide.h. This was a mistake
   present since far too many years in this driver. The proper way to
   deal with broken systems is to define REALLY_SLOW_IO in system
   dependent headers or particular driver files.  We can always
   reintroduce it easy if real users will complain, since OUT_BYTE() and
   similar can be used as hooks. But I don't expect anybody reporting
   about this. Even on the most broken IDE chip in the world (cmd640
   at VLB) undefining this *always* worked for me. Nearly all the code
   pieces in the ide driver code *reverted* it's effects explicitly
   anyway.

- Remove the obsolete CONFIG_BLK_DEV_4DRIVES support. This was supposed
   to support 4 drivers attached at one channel on some older chipsets,
   in esp. Tekram 690CD, in the last century. They where all supposed to
   work at a register set starting at the base address 0x1f0.  Before
   complaining that this is removing functionality, please note that this
   must have been broken for already quite a long time, since the ide
   driver didn't contain the special device selection methods implicated
   by this any longer.  It didn't scan   this port too if PCI host chip
   support was enabled (as it is in all those distributions around
   there).  On the other hand this is the most prominent case of
   incoherent use of the mate member in the struct hwif_s. And please
   think about how big the probability is, that there are systems out
   there, where there are actually 4 drivers on such a channel?

- Streamline module initialization code by removing one shoot functions.

- Make the WAIT_READY value used in case of CONFIG_APM or
   CONFIG_APM_MODULE the default, since this is what really reflects the
   behavior of modern drives. It won't hurt any other case and finally
   removing it is reducing the necessary coverage for overall driver code
   testing/analysis.

- Move the IDE_LARGE_SEEK macro to the only place where it's actually
   used. Replace the IDE_MIN() and IDE_MAX() drivers with the obvious.
   Remove unused SPLIT_WORD and MAKE WORD from the local header.

- Remove CMD640_DUMP_REGS from global scope, since there is no
   development done on this any longer. Finally, the way the host chip
   initialization routines are called changed in the time between allows
   this to remain fully local to the host chip driver in question.

- Some spell checking of comments in the code. (Yeep I have extended my
   Vim to do this the "Word" way with nice undercurl lines... mozilla
   remains to be fixed...)

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

The fix for rename broke "link()". Fixed.

so we need to make init_idle() aware of it so that it gets the
preempt_count initialization right.

I've found what's going on there.  Basically, we should not use
__user_walk() with LOOKUP_PARENT - nd->last.name is set to the last
component of the name and freeing that name before we are done is not a
good idea. 

On SMP systems, preempt_count is erroneously set to 1 for idle task's on
all CPU besides CPU0.  This patch sets preempt_count properly.

	Robert Love

I see that this is fixed now in 2.2 and 2.4, but 2.5 still has broken
BSD partition handling.  Here a fix.

Andries

 - Fix bug in tw_aen_complete() where aen's could be lost.
   Fix tw_aen_drain_queue() to display useful info at init.
   Set tw_host->max_id for 12 port cards.
   Add ioctl support for raw command packet post from userspace
   with sglist fragments (parameter and io).
 - Fix read capacity to under report by 1 sector to fix get
   last sector ioctl.
 - Fix bug where more AEN codes weren't coming out during
   driver initialization.
   Improved handling of PCI aborts.
 - Fix bug in tw_findcards() where AEN code could be lost.
   Increase timeout in tw_aen_drain_queue() to 30 seconds.
 - Re-write raw command post with data ioctl method.
   Remove raid5 bounce buffers for raid5 for 6XXX for kernel 2.5
   Add tw_map/unmap_scsi_sg/single_data() for kernel 2.5
   Replace io_request_lock with host_lock for kernel 2.5
   Set max_cmd_len to 16 for 3dm for kernel 2.5
 - Set host->max_sectors back up to 256.
 - Modified pci parity error handling/clearing from config space
   during initialization.
 - Better handling of request sense opcode and sense information
   for failed commands.  Add tw_decode_sense().
   Replace all mdelay()'s with scsi_sleep().
 - Revert mdelay's and scsi_sleep's, this caused problems on
   some SMP systems.
 - Add pci_set_dma_mask(), rewrite kmalloc()/virt_to_bus() to
   pci_alloc/free_consistent().

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

 There is a problem with the through-8259A mode for IRQ 0 on I/O APIC
systems.  Depending on correctness of an MP table, IRQ 0 routing is either
not registered at all or registered at a wrong pin.  As a result the 8254
timer IRQ only works by an accident (it's edge-triggered and never
disabled/enabled so it happens to survive this incorrect configuration).
A visible effect is you can't change the affinity for IRQ 0.

 Following is a patch that fixes both cases referred to above.  The code
looks obvious but it was additionally run-time tested just in case.  The
issue is serious -- please apply the patch ASAP.  As no changes were done
to io_apic.c since the development fork, the patch applies cleanly both to
2.4 and to 2.5.

 Credit goes to Joe for discovering the affinity problem and providing a
fix proposal (incorporated in the final one).

  Maciej

There is no such a thing like a ide-clean-15. Never was.  But here comes
what has been done in ide-clean-16:

- Apply a patch for the initialization of the second PIIX channel.
   Found by Daniel Quinlan <quinlan@transmeta.com>

- Apply a patch for the DMA initialization of the serverworks chip.
   Ken Brownfield <brownfld@irridia.com>

- Make the ata_operations methods immune against device type drivers,
   which donot provide them by separating the access to them out. Audit
   them all.

- Resynchronize with 2.5.6-pre1.

- Remove unused IDE_DRIVE_CMD, IDE_DRIVE_TASK, IDE_DRIVE_TASK_MASK,
   IDE_DRIVE_TASKFILE macros.

- Disable configuration of the task file stuff. It is going to go away
   and will be replaced by a truly abstract interface based on
   functionality and *not* direct mess-up of hardware.

- Resync with 2.5.6-pre2.

- Add HPT entries to the fall-back list, since otherwise the driver
   won'trecognize the drives. We will have to make this the default
   behavior for allnot recognized host chip types.

- Fix compilation with no PCI host chip support enabled.

- Apply the overflow fixes for HPT366 by Vojtech Pavlik.

- Kill the one-shoot functions ide_wait_cmd_taks() ide_wait_cmd() by
   moving them to the places where they are actually used. Fix a
   potential buffer overflow on the way.

- Fix usage of ide.c as module. Thanks to Adam J. Richter for figuring
   out what was wrong.

- Various cleanups all along as well as removal of TONS of
   unfinished/dead code.

I think it's sometimes better to remove stuff, which isn't there,
instead of hoping for a "magical day" where it will be finished.

This is the Richard Henderson-approved, cleaner, brighter per-cpu patch.

Ok, I found it.  The problem was in the decompression code.

In inflate.c, in the routine zlib_inflate(), it reuses the passed-in "f"
parameter as a return value, but the PPP packet compression stuff needs
that return value later.

This patch adds two intermediate general cache sizes, 96 and 192 bytes.
On my system this saves about 34k.

size-256              63    180    256    5   12    1
size-192              95    120    192    5    6    1
size-128             213    240    128    8    8    1
size-96             1080   1120     96   28   28    1

During 2.5.5-pre schedule_tail was optimized away under UP.  We need it
for preempt-kernel, too, since it decrements the preempt_count to 0
coming off do_fork.

Without this patch, CONFIG_PREEMPT + !CONFIG_SMP does not gain one very
much.  

	Robert Love

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

	Below is a followup to bd_claim patch - it is the last one
from the current series.  Removes SWP_BLOCKDEV flag - it isn't needed
anymore.

*	new functions - bd_claim(bdev, holder) and bd_release(bdev).
	bd_claim(bdev, holder) fails is device is already claimed by
	somebody else; bd_release(bdev) gives device up.

*	get_sb_bdev() claims device for fs_type; it means that we don't need
	to look through entire least of superblocks anymore - just through
	the list of superblocks belonging to that type (i.e. the same thing
	we do for non-block filesystems; that will allow to merge quite a
	bit of code afterwards).

*	sys_swapon claims device for itself;  free exclusion with mounting,
	end of problems with bogus set_blocksize().

*	is_mounted() and is_swap_partition() are gone - what we actually
	wanted was "try to claim device for ourselves".  Which we can do
	now - without races inherent to is_mounted()/is_swap_partition().

*	RAID lock_rdev() claims device for itself.  I.e. we get rid of
	is_mounted() in there (BTW, is_swap_partition() was missing) and
	we get protection both ways - not only RAID won't take an already
	mounted device, but mount won't stomp on a device claimed by RAID.

There are other places that would benefit from the same (e.g. ext3 with
external journal almost definitely wants to claim device for itself).

Notice that it's a cooperative thing - neither open() nor raw device stuff
claim the block device, so they don't care if device is mounted, etc.  So
we don't break fsck and friends - exclusion is between those who know that
they want that exclusion.