It looks like sk->num became inet_sk(sk)->num in 2.5.7-pre1,
but one of these changes was missed in net/ipv4/ipmr.c.  Here is
the patch.

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

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

In the FIT macro in ide-timing.h the argument got swapped because of a
typo. All timings generated for VIA and AMD chips are wrong because of
that. Safe, though, but slow.

This fixes it.

If I was to give this patch a name it would be:

"Vojtech Pavlik unleashed from the chains".

So credit where credit is due :-).

Anyway here follows the change log:

Mon Mar 11 23:48:28 CET 2002 ide-clean-21

- Swallow rewritten amd74xx host chip setup code from Vojtech Pavlik.  We can
   revert it easly if it turns out to be a bad thing. However the code looks
   quite sane to me. In esp. it doesn't containg that many magic numbers.

- Clean stale white spaces in ide-timing.h tirvial fix.

- Make ide_release_dma return void. It's value is never used anyway.

- Swallow more timing setup code cleanup by Vojtech Pavlik. Apply some
   cosmetics to it. Port opti621 to the new setup code.

- Kill abuse of ide_do_reset() on error return paths for atapi floppy tape and
   cd-rom devices. Just stop them. This gives better changes that defect
   removable media will not cause suddenly broken timings on hard discs
   containing system data! Even then comments in ide_do_reset() admit, that
   resetting the whole channel can have adverse effects on the second interface
   on this channel. And I have too frequently observed linux struggling on
   defect cd-rom for a far too long time to wish it to continue.

   Oh did I forget to say that the corresponding "how can I break my system fast
   and reliable" ioctl is gone as well?

   Removing it recovered the fact that the CONFIG_BLK_DEV_IDEDMA_TIMEOUT is
   completely bogous. I have removed this option therefore as well, because it's
   playing the same wrack havoc on the devices if enabled. This cat has been in
   an unfinished and *unfunctional* state anyway.

- Actually add physical suspend code to the power handling code.  Still the
   resume code isn't finished just jet. This is all subject to change at the
   point in time when we get to proper command queueing.
   I think however that Pavel will be interrested in tidding this bit up...

- Resync with 2.5.7-pre1.

Contributor: Eli Kupermann @ Intel

into mandrakesoft.com:/home/jgarzik/repo/net-drivers-2.5

coalescing, ring params, pause params, hw csum disable/enable,
scatter-gather enable/disable

* Merge support for hardware vlan accel
* Better wakeup behavior on TX completion
* dev->mtu setting fixes

* Merge several bug fixes from vger cvs
* Merge h/w VLAN support, now that API is in the main tree
* Add support for TX/RX coalescing

Separates efs_sb_info from struct super_block.

Add nfs_compare_super() and teaches nfs_get_sb() to look for existing
superblocks.

Switch NFS to separate allocation of private part of superblock,
uss explicit sget() instead of get_sb_nodev()

Export sget(9), deactivate_super(9) and set_anon_sb(9)

Switch NFS to use of NFS_SB(sb) instead of sb->u.nfs_sb.s_server

I forgot to zero out the newly allocated memory in the previous patches
for ext2 and ncpfs.

I forgot to zero out the newly allocated memory in the previous patches
for cramfs and minixfs.

Complete the ext2 superblock seperation.

Abstract access to ext2_sb_info.

Seperates ncp_sb_info from struct super_block.

  The following patch resyncs 2.5.6 with the 2.4.x series w.r.t. the
handling of the EJUKEBOX error. The latter is an NFS-specific error
that is returned by servers that support hierarchical storage: it
notifies the client that the request cannot be completed in a timely
fashion (Imagine for instance a situation where you have a cdrom
jukebox system, and the user has just requested a file on another cd).

Under these circumstances, the RFC specifies that the request should
be retried after suitable timeout during which the server will attempt
to perform whatever action is required to make the file available
again.

  The following patch fixes a bug in NFS file creation. Recently (not
sure exactly when), open_namei() was changed so that it expects
vfs_create() to always return a fully instantiated dentry for the new
file.

The following patch ensures this is done in the cases where the RPC
CREATE call does not return valid attributes/filehandles. This is
always the case for NFSv2, and can sometimes be the case for v3...

  The following fixes up a couple of bugs that resulted from the fix
in 2.5.4 for ETXTBSY: Since the READ requests now only store RPC
credentials and not the struct file, we need to be careful when
deciding to coalesce requests on different pages into 1 RPC call that
we compare the credentials instead of the struct file.

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

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

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

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

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

This patch fixes the es1370 build problems due to virt_to_bus()
being gone.

This patch is a redesign for videodev.[ch].  Changes:

- drop the function pointers (read/write/mmap/poll/...) from struct
  video_device, use struct file_operations directly instead.
  Dispatching to different drivers by minor number is done the same way
  soundcore.o handles this: swap file->f_fops at open() time.

- also drop the now obsolete video_red/write/mmap/poll/...  functions
  from videodev.c

- Stop using the BKL, use a mutex to protect open,register+unregister
  calls against races.

- provide a video_generic_ioctl() function which can (and should) be
  used by v4l drivers to handle copying from and to userspace.

- provide video_exclusive_open/release functions which can be used by
  v4l drivers to make sure only one process at a time opens the
  device.  They can be hooked directly into struct file_operations if
  some driver has nothing to initialize at open time (which is true
  for many drivers in drivers/media/radio/).

The move from function pointers in struct video_device to struct
file_operations does break all existing v4l drivers.  Thus I have a
large number of patches for the drivers in the kernel tree.  Most of it
is just the adoption to the videodev.[ch] changes, but I've also fixed a
small bug there and there while walking througth the source files.

This patch fixes the build failure of the miro radio driver due to
the new location of the sound drivers in the tree (alsa merge).

This patch updates/fixes the video4linux API documantation.  The current
description for mmap() based capture is unclear and somewhat misleading.

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

 - Preliminary version of NAND flash support.
 - Locking documentation and fixes (including BKL removal because it's superfluous).
 - Performance improvements - especially for mount time. 
 - Annoying stuff like i_nlink on directories fixed.
 - Portability cleanups.

Seperates cramfs_sb_info from struct super_block.