This is the second of 3 patches from David Woodhouse which fix various
problems with signal handling on ppc64.

The problem that this patch fixes is that a process being single-stepped
will execute one instruction too many in certain cases: when a signal is
delivered, when a signal handler returns, and when a system call
instruction is single-stepped.  This patch fixes it by checking if the
process is being single-stepped in the syscall exit path and on signal
delivery, and stopping it if so.  To avoid slowing down the syscall exit
path in the normal case, we use a bit in the thread_info flags, which can
be tested along with the other bits we already test in the syscall exit
path.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This is the first of 3 patches from David Woodhouse which fix various
problems with signal handling on ppc64.  (This is David's patch plus a
couple of comment fixes from me.)

Without this patch, a signal handler that is called as a result of delivery
of a signal while a process is waiting in sigsuspend() will have wrong
values given to it for the second and third arguments.  The reason is that
we were returning to userspace via the syscall return path, which doesn't
bother to restore r4 and r5.  This patch arranges for the return to
userspace to be done via ret_from_except in this case, which restores all
registers.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Gerald Schaefer <geraldsc@de.ibm.com>

 - Remove ESPIPE logic, use nonseekable_open() instead.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Gerald Schaefer <geraldsc@de.ibm.com>

 - Use nonseekable_open() to disable seeking and pread()/pwrite().
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Klaus Dittrich observed this bug and posted a test case for it.

This patch fixes both that failure mode and some others possible.  What
Klaus saw was a false negative (i.e.  ECHILD when there was a child)
when the group leader was a zombie but delayed because other children
live; in the test program this happens in a race between the two threads
dying on a signal.

The change to the TASK_TRACED case avoids a potential false positive
(blocking, or WNOHANG returning 0, when there are really no children
left), in the race condition where my_ptrace_child returns zero.
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Don't try autosuspend if we think the hardware won't resume correctly
from the OHCI suspend state.  This makes the RWC bit serve double duty,
but that appears to work OK, and the only penalty is increased power
consumption (from OHCI clocks) on boards/chips that don't work right.

For example, the amd756 erratum 4 workaround needs this logic; and at
least one ServerWorks box issues spurious resume IRQs (~3x/second!)
in the suspend state.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This patch fixes a bug in the usbfs code.  The driver is too zealous about
checking for disconnected devices before doing things.  In particular, it
is necessary to reap all outstanding asynchronous URBs and unbind from
interfaces when the device file is closed, even if the device is no longer
connected.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

Patch from Richard Purdie

The base PCMCIA code for PXA2xx doesn't work over a suspend/resume 
cycle as the MECR register isn't preserved. This patch restores 
the register.

Signed-off-by: Richard Purdie
Signed-off-by: Russell King

Patch from Catalin Marinas

The CType field in the cache type CP15 register does not always
specify whether the cache is VIVT or VIPT. The dump_cpu_info()
function should use the cache_is_(vivt|vipt) macros instead.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

Patch from Catalin Marinas

The CPU Id register changed with the ARM1176 processor and the
ARMv6 architecture description field can be either 7 or 0xF.
This patch also detects the first byte of the part number since
0xF in the description field will also be used for future ARM
architectures.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

Patch from Deepak Saxena

Add support for Intel's new IXP46x CPU family.

Signed-off-by: Deepak Saxena
Signed-off-by: Russell King

Patch from Catalin Marinas

CONFIG_FPE_NWFPE was enabled and CONFIG_FPE_FASTFPE disabled in the
(integrator|versatile)_defconfig files because the latter is only
available as a separate patch.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

Patch from Dave Jiang

Update all IOP def configs to utilize the MTD physmap driver instead of IOP specific MTD maps driver.

Signed-off-by: Dave Jiang
Signed-off-by: Russell King

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

Here is trivial patch to enable the early console setup for UP as well.
Tested on zx1.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

into intel.com:/data/home/aegl/BK/linux-ia64-release-2.6.10

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Tony Luck <tony.luck@intel.com>

into bkbits.net:/repos/c/cifs/linux-2.5cifs

Error mapping workaround for NT4 bug of reporting oldstyle DOS error for ERR invalid level (shows up on attempts to do SetPathInfo to NT4).
Call posix_lock_wait to avoid fs/locks.c panic per Trond's suggestion

Signed-off-by: Steve French (sfrench@us.ibm.com)

Patch from Dave Jiang

remove iop331-mm.c in arch/arm/mach-iop3xx

Signed-off-by: Dave Jiang
Signed-off-by: Russell King

Patch from Catalin Marinas

The Versatile/AB platform is defined as a MACH (not ARCH) in the
arch/arm/tools/mach-types file.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

Patch from Catalin Marinas

CONFIG_ARCH_VERSATILE=y was added to the versatile_defconfig file.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

into sambaltcdom.austin.ibm.com:/home/sfrench/linux-2.5cifs

incorrectly limited to just over 512)

Pointed out by Shaggy.

Signed-off-by: Steve French (sfrench@us.ibm.com)

We notiecd proc_net_fops_create() failing, and cleaned up, but returned
0 anyway.

Found by the --failtest flag to nfsim-testsuite.  --failtest rocks my
world.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Found by nfsim (-Wmissing-prototypes is your friend).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

The documentation says dma_set_api() returns 1 on success and
0 on error, which is plain wrong, standard kernel return codes
are used instead (0 on ok, -ESOMETHING on error).

The meye hardware needs to access the main memory for DMA using
32 bit addresses. The previous version of the meye driver used
dma_addr_t types to build those addresses and ensured that
sizeof(dma_addr_t) = 4 by disabling HIGHMEM64G in Kconfig.

However, this way of doing it also makes meye unavailable on
some kernel configurations. As Arjan said previously, future Fedora
kernels may have HIGHMEM64G activated by default (davej says it won't
happen, at least for now...). Other distributions may do the same and
this will require meye users to recompile the whole kernel.

The attached patch makes the meye driver use dma_addr_t addresses 
internally, but converts them to u32 before giving them to the
hardware.
Signed-off-by: Stelian Pop <stelian@popies.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This patch applies to the kernel 2.6.9 and fixes the initialisation of
the DMA channel polarity in the function ppc4xx_init_dma_channel() for
the PPC 4XX processor.
Signed-off-by: Colin P Wernham <cwernham@airspan.com>
Signed-off-by: Matt Porter <mporter@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Newer binutils (2.15) when they know they aren't assembling for a
classic target (say e500 instead of 750) disallow certain opcodes,
causing the compile to fail.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Tom Rini <trini@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into bkbits.net:/repos/c/cifs/linux-2.5cifs

servers to avoid mount oops.  Fix case in which tcp stack only returns 3 bytes
of larger peek of smb header (which would otherwise kill the
session and cause us to have to reconnect)

Signed-of-by: Steve French (sfrench@us.ibm.com)

Unfortunately the restoring of SPE state was causing data corruption
since we were restoring based on the size of the altivec context and not
the SPE context.  Also, fixed setting of last_task_used_spe on
start_thread, flush_thread, and exit_thread. 
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

It's "const volatile __iomem *". The "const" was missing.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

When adding or deleting from the source list make sure to find matches
by comparing against the new source address, not the group address.
Also, check each addr in the list rather than just the first one.
And, finally, only delete from list when there's a match rather than
vice-versa.  Drop the effort to keep list sorted, since it's not done
on full-state api and can create an sl_addr entry that the delta api
won't be able to delete. Without these fixes sl_count can be corrupted
which can allow for kernel memory corruption.
Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>