More merging from zlib-1.1.4

force windowBits > 8 to avoid a bug in the encoder for a window size
of 256 bytes. (A complete fix will be available in 1.1.5).

James Carlson:

        The problem is that s->strstart gets set to a very large
        positive integer when wsize (local copy of s->w_size) is
        subtracted in deflate.c:fill_window().  This happens because
        MAX_DIST(s) resolves as a negative number when the window size
        is 8 -- MAX_DIST(s) is defined as s->w_size-MIN_LOOKAHEAD in
        deflate.h.  MIN_LOOKAHEAD is MAX_MATCH+MIN_MATCH+1, and that
        is 258+3+1 or 262.  Since a window size of 8 gives s->w_size
        256, MAX_DIST(s) is 256-262 or -6.

        This results in read_buf() writing over memory outside of
        s->window, and a crash.

Don't think anyone actually bothers to check specific error codes, but
it shouldn't hurt either.

This removes unnecessary NULL casting.

s/Z_NULL/NULL/g.

Simple s/local/static/.

They may have some new HP address, unknown for now.

Most of it is reformatting, but the functional bits should fix real
problems.  A loop is introduced, just like in the turboc patch and one
of the three condition bodies has been expanded.

This is the first bit of the missing merge towards 1.1.4.  Applies on
top of the previous cleanups.

This one rips out an ugly #ifdef and seems to catch a theoretical
error possibility.  Always thought that they fixed more than they
officially admitted.

We don't use any cplusplus in the kernel.

The old 26-bit ARM support was long since dropped out of the regular ARM
support, since it was different enough to not make sense to maintain as
one port.

This re-introduces arm26 as an architecture of its own.

lib-y is the new way to define what objects belongs to a library.  The
implementation was not made backwards compatible and therefore an update
to all architectures are needed.

This is a simple replacement of obj-* to lib-* and deletion of L_TARGET.
The new mechanish where lib.a can be mixed with built-in.o is not
utilised.

- rewrite taskfile PIO handlers
  (they now comply with ide state machine and use bio walking)
- switch ide-disk.c to use *only* taskfile IO
- swicth pdc4030.c to use *only* taskfile IO (untested)
- remove old cruft (>600 lines)

From: Davide Libenzi <davidel@xmailserver.org>

After the ep_remove() the "epi" is given back to the cache, so "epi->ep"
might become invalid.  It was not cought by my tests because the element
wasn't immediately reused (and because I was using a single epoll fd, so
the "ep" item remained the same).

From: Dave Hansen <haveblue@us.ibm.com>

__alloc_bootmem_core() has a couple of BUG_ON()'s.  Since the handlers
aren't set up this early, if you hit it, you just get along stream of
"Unknown Interrupt" messages.  It would be very nice to have a little
bit more information when something has decided to BUG() out this
early.

Christoph says this undef is not correct.

From: Matt Dobson, via Martin Bligh

This parses the machine's BIOS tables to populate the
mp_bus_id_to_node[bus] array.  Only affects Summit machines, safe, boring.
Has been in -mjb tree for ages, and works fine.

From: "Martin J. Bligh" <mbligh@aracnet.com>

The "magic" switching in subarch was ugly when I put it there, and nobody
liked it then (including me).  It hasn't got any prettier since.  Andi's
generic arch stuff is a cleaner solution for now, so we can remove the old
hacky stuff, and significantly simplify the code.  All this does is replace
"(x86_summit ?  A : B)" with "A" everywhere.

From: "Martin J. Bligh" <mbligh@aracnet.com>

All this fancy stuff in cpu_mask_to_apicid doesn't work for NUMA-Q, because
it's based on logical apicids, and we use physical.  Drop back to just
always returning 0xF instead, which is the broadcast physical ID, and has
been working fine since the dawn of time.

From: Adrian Bunk <bunk@fs.tum.de>

Compilation fails if JFFS_PROC_FS and !PROC.  The following dependency in
the Kconfig file is needed.

From Bartlomiej Zolnierkiewicz

With !CONFIG_PROC_FS, ieee1394_core fails to compile because the argument to
this inline is still evaluated.  But it doesn't exist.

A general fix is to not evaluate the arg at all.

From: Pavel Machek <pavel@suse.cz>

Convert x86_64 to use the common ioctl code.

From: Pavel Machek <pavel@suse.cz>

Make sparc64 use generic ioctl32 code.

From: Pavel Machek <pavel@suse.cz>

Various 64-bit architectures are duplicating a ton of 32-bit compat code.

Pavel's patch creates a generic 32-bit ioctl file in fs/compat_ioctl.c which
architectures will #include from within their arch/ layer.

Has been reviewed by everyone and tested on sparc64, x86_64, ppc64 and ia32.

From: Adrian Bunk <bunk@fs.tum.de>

Fix the ppc64 build for the great socket member renaming.

From: "David S. Miller" <davem@redhat.com>

add_kallsyms() doesn't exist if !CONFIG_KALLSYMS.

New makefile variable introduced: lib-y

The lib-y syntax allows you to do the usual tricks such as:

	lib-$(CONFIG_SMP) += percpu_counter.o

A built-in.o is always present in a directory that list .o files in
either obj-* or lib-*.

In contrast, lib.a is made only when lib-y is defined.

I also updated lib/Makefile, so that crc32.o is now always built-in
if selected. 

This brings the x86-64 port uptodate for 2.5.70. Just various bugfixes
and a few merges with other people.

Only changes architecture specific files.

- Fix compiling with CONFIG_IA32_EMULATION on
- Readd lost apic power management patch from Pavel (fixes oprofile too)
- Increase max IOAPICs to 16
- Fix compiling with CONFIG_IA32_EMULATION off
- Compile fix for suspend (Pavel Machek)
- Support boxes with APIC disabled
- Remove code to forcibly enable APIC
- Small fix for APIC timer calibration.
- Fix deadlock in SMP reboot
- Some warning fixes
- Save edid info at boot (Bryan O'Sullivan)
- Add better locking to oops printing and support it for page faults.
- Don't printk handled signals.
- Update defconfig
- Add copy_in_user

cpu_relax is on i386 and x86-64 in processor.h, not system.h
This makes CONFIG_DEBUG_SPINLOCK compile for x86-64

This patch moves the definitions of BUG, BUG_ON and WARN_ON from
<linux/kernel.h> to <asm/bug.h> (which <linux/kernel.h> includes), and
supplies a new implementation for PPC which uses a conditional trap
instruction for BUG_ON and WARN_ON, thus avoiding a conditional
branch.  This patch trims over 50kB from the size of the kernel that I
use on powermacs.

With this patch, on PPC we have a __bug_table section in the vmlinux
binary, and also in modules if they use BUG, BUG_ON or WARN_ON.  The
__bug_table section has one entry for each BUG/BUG_ON/WARN_ON, giving
the address of the trap instruction and the corresponding line number,
filename and function name.  This information is used in the exception
handler for the exception that the trap instruction produces.  The
arch-specific module code handles the __bug_table section so that
BUG/BUG_ON/WARN_ON work correctly in modules.

Several architecture maintainers have acked this change.  It should be
completely benign for all of the other architectures (though they may
decide to do something similar if they have a conditional trap
instruction available).

This patch removes the warnings that the check program came up with in
the PPP code: ppp_async.c, ppp_deflate.c and ppp_generic.c.  This
involved adding __user and converting K&R-style function definitions
to ANSI-style.  I also took the time to add some extra comments to
ppp_deflate.c explaining in more detail what each function does and
what its arguments are.

This patch removes the warnings that the `check' program came up with
in drivers/macintosh.  This involves adding __user in various places
and fixing some non-ANSI function definitions for functions that take
no arguments.

Here is another final K&R to ANSI C cleanup patch for zlib.

The members of struct sock got a prefix of 'sk_' recently.

This patch updates smbfs to match the new world order.

into home.transmeta.com:/home/torvalds/v2.5/linux

This changeset:

1. Implements hlist_add_after
2. uses prefetch in hlist_for_each, using a trick that ends up being
   equivalent to having the prefetch instruction in the first block
   of the hlist_for_each for block, the compiler optimizes the second
   "test" away, as its result is constant
3. implements hlist_for_each_entry and hlist_for_each_entry safe,
   using a struct hlist_node as iterator to avoid the extra branches a
   similar implementation to list_for_each_entry would have if used
   a typed iterator, but while avoiding having to have the explicit
   hlist_entry as in hlist_for_each.

4. Converts the hlist_for_each users that had explicit prefetches, i.e.
   removed the explicit prefetch

5. fix a harmless list_entry use in a hlist_for_each in inode.c
   

into home.transmeta.com:/home/torvalds/v2.5/linux

powernow-k7 needs dmi_broken

This patch works around a gcc 3.3 bug on AMD64.  On AMD64 the
get_current() function expands to a switch on sizeof which has to be
optimized away by the compiler, finally yielding an inline assembler
statement.  In some cases it seems to get that wrong and forgets to
reference and use the argument.  I have only seen it happen in
binfmt_elf so far.

Work around by just computing "current" once.