into nuts.davemloft.net:/disk1/davem/BK/sparc-2.6

From: "H. Peter Anvin" <hpa@zytor.com>

Apparently, on x86-64, the stack isn't always aligned properly (16 bytes)
in the kernel at the moment.  This causes the RAID-6 code to crash the
system.  This patch is a workaround for that; the right thing is to muck
with the assembly entrypoints to enforce proper stack alignment.  However,
that's not anything I feel comfortable doing in an evening, especially
since I don't have a machine on which I can test the resulting kernels.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

SCLP console/tty fixes:

- Fix incorrect state change of SCLP_RUNNING flag in interrupt handler

- Suppress emission of empty buffers to prevent stack overflow

- Fix off by one error in sclp_write (used to return # of chars written + 1)

- Prevent sclp_tty_write_string from waiting in interrupt (during flush)

- Fix deadlock after TIOCSCLPSNL ioctl

- Fix sclp_tty_chars_in_buffer calculation

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

s390 inline assembly bug-fixes:

- Add memory barriers to spinlocks, atomic variable functions and to
  test_and_{set,clear,change}_bit.

- Add "=m" and "m" contraints to tell gcc that the content of a variable
  is in fact used.

- Replace "+m" constraints by "=m" on the output and "m" on the input list.

- Use c-implemtation for ffz and __ffs.

- Use generic c-implemtation for ffs and fls.

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

s390 general update:

- Add archhelp for "make image".

- 32-bit execve fix.

- Export some symbols: sys_ioctl, put_files_struct & ccw_device_work.

- Minor cleanups in arch/s390/mm/init.c & drivers/s390/net/ctctty.c.

From: David Gibson <david@gibson.dropbear.id.au>

These are some trivial cleanups to the hugepage ppc64 support

From: Anton Blanchard <anton@samba.org>

From: Anton Blanchard <anton@samba.org>

The current SLB handling code has a number of problems:

- We loop trying to find an empty SLB entry before deciding to cast one
  out.  On large working sets this really hurts since the SLB is always full
  and we end up looping through all 64 entries unnecessarily.

- During castout we currently invalidate the entry we are replacing.  This
  is to avoid a nasty race where the entry is in the ERAT but not the SLB and
  another cpu does a tlbie that removes the ERAT at a critical point.  If
  this race is fixed the SLB can be removed.

- The SLB prefault code doesnt work properly

The following patch addresses all the above concerns and adds some more
optimisations:

- feature nop out some segment table only code

- slb invalidate the kernel segment on context switch (avoids us having to
  slb invalidate at each cast out)

- optimise flush on context switch, the lazy tlb stuff avoids it being
  called when going from userspace to kernel thread, but it gets called when
  going to kernel thread to userspace.  In many cases we are returning to the
  same userspace task, we now check for this and avoid the flush

- use the optimised POWER4 mtcrf where possible

From: Anton Blanchard <anton@samba.org>

From: Anton Blanchard <anton@samba.org>

From: Anton Blanchard <anton@samba.org>

From: Anton Blanchard <anton@samba.org>

Lets make a decent attempt to find out where your console is.  The new
preferred_console stuff is pretty nice.

From: Anton Blanchard <anton@samba.org>

From: Anton Blanchard <anton@samba.org>

Small obvious fix to ppc32_timer_create.  Since sys_timer_create access
structures we pass on the stack, we need set_fs(KERNEL_DS).

From: Anton Blanchard <anton@samba.org>

- nop out proc_device_tree_add_node if CONFIG_PROC_DEVICETREE=n

- stubs for procfs-related functions when CONFIG_PROC_DEVICETREE=n

From: Anton Blanchard <anton@samba.org>

Add lparcfg_write() for changing SPLPAR system parameters

From: Anton Blanchard <anton@samba.org>

add missing include guards, from Nathan Lynch

From: Tom Rini <trini@kernel.crashing.org>

Correct the defines for MC146818 RTCs, reg c is interrupts, reg d is not
watchdog (it's valid time & nvram).

into home.osdl.org:/home/torvalds/v2.5/linux

SGI Modid: xfs-linux:xfs-kern:165722a

SGI Modid: xfs-linux:xfs-kern:165673a

SGI Modid: xfs-linux:xfs-kern:165648a

SGI Modid: xfs-linux:xfs-kern:165638a

SGI Modid: xfs-linux:xfs-kern:165631a

SGI Modid: xfs-linux:xfs-kern:165500a

SGI Modid: xfs-linux:xfs-kern:164115a

SGI Modid: xfs-linux:xfs-kern:165646a

SGI Modid: xfs-linux:xfs-kern:165037a

This forward ports a few important fixes from the 2.4 driver.  This
changes have been well tested.

Changelog:
 - Update maintainer email address
 - Mark pci_device_id list with __devinitdata.
 - Set correct protocol type on packet receive (this caused the kernel to
   drop all packets received)
 - Add #ifdef DEBUG around debug printk()
 - ioctl: Add missing size checks before
   copying data from userspace.

into home.osdl.org:/home/torvalds/v2.5/linux

We used to have code to allow binaries linked against glibc to run,
when glibc itself contained some FP instructions (for PCS stack
frames for functions like printf and scanf) thereby allowing FP
emulators like nwfpe to be built as modules.

This has proved to be unreliable with later compilers, so support
for this was dropped a while ago.  Since no one complained, we can
finally remove the dead code.

(NB. a klibc based module-init-tools shouldn't suffer from this
problem.)

According to x86, the pointers for bitops are supposed to be
qualified with volatile.  Make ARM bitops reflect this.

Remove usage of tsk->used_math on ARM, moving the status to an array
of co-processor usage.  (ARM can have up to 15 co-processors
providing various extra facilities such as SIMD, VFP or FP.)

Jakub points out that having an automatic array is not only
bad for performance (and stack usage), gcc has also historically
had lots of bugs here, and gcc-3.2.3 seems to miscompile it
otherwise.