Joel Soete points out that we refer to pa_tlb_lock but only define it if
CONFIG_SMP which breaks a uniprocessor build with CONFIG_DEBUG_SPINLOCK
enabled.  No module refers to pa_tlb_lock, so we can delete the export.
Signed-off-by: Matthew Wilcox <willy@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

GCC no longer allows a cast as lvalue; fix the same way fs/readdir.c was
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

iounmap's argument needs to be both const and volatile, otherwise we'll
get warnings that we're discarding pointer qualifiers
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

header. This will allow the use of more constants in the
agpgart driver.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Pull out struct sba_device and struct lba_device into a
common ropes.h header. Also fold the parisc portion of
iosapic.h into this file. (Then delete the useless portion
of iosapic.h)
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

IS_ELROY. Also remove the useless caching of the
parisc_device_id in struct sba_device.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

(it's pretty useless telling a user they have a whole bunch of
 TR%d.%d LBAs)
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Our prior mode of operation didn't allow nested interrupts
because it makes the interrupt code much simpler.  However,
nested interrupts are better for latency.

This code uses the EIEM register to simulate level interrupts
and thus achieve nesting.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

This version (relative to the current tree):
o eliminates "while (ticks_elapsed)" loop. It's not needed.
o drop "ticks_elapsed" completely from timer_interrupt().
o Estimates elapsed cycles (based on HZ) to see which kind of
  math we want to use to calculate "cycles_remainder".
o Fixes a bug where we would loose a tick if we decided
  we wanted to skip one interrupt.
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

It's just a bit easier to follow and timer code is complex enough.

So far, only tested on A500-5x (64-bit SMP), ie: gettimeoffset() code
hasn't been tested at all.
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Use the __raw_spin_lock_flags routine so we can take an interrupt while
spinning.  This re-fixes a bug jejb found on 2005-10-20:

CPU0 does a flush_tlb_all holding the vmlist_lock for write.
CPU1 tries a cat of /proc/meminfo which tries to acquire vmlist_lock for read
CPU1 is now spinning with interrupts disabled
CPU0 tries to execute a smp_call_function to flush the local tlb caches
This is now a deadlock because CPU1 is spinning with interrupts disabled and
can never receive the IPI
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

I couldn't find where the itimer was getting started for slave CPUs.
CPU 0 (master) itimer was started in time_init() (arch/parisc/kernel/time.c).
start_cpu_itimer() code was striped from time_init().
Slaves now start their itimer in smp_cpu_init().

This is a first step towards making gettimeoffset() work for SMP.
Next step will be to determine the CR16 (cycle counter)
offsets for each CPU relative to the master (CPU 0).
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

processor_probe() shouldn't clobber cpu_data[0]
cpu_data[0].it_value (used by timer_interrupt()) is already set.
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

This isn't likely to be causing problems for other bits of
kernel code. I can't find any other user of CONFIG_HZ outside
of arch specific code.
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Implement trivial futex_atomic_cmpxchg_inatomic for testing.
Signed-off-by: Carlos O'Donell <carlos@systemhalted.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Rewrite rwlock implementation to avoid various deadlocks in the current
scheme.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

intr_restore in intr_do_preempt. This prevents the execution
of an unwanted insn...
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Russell King pointed out that asm/serial.h is anachronistic and we were
misusing BASE_BAUD.  So fix BASE_BAUD for PCI 16550 UARTs, move LASI_BASE_BAUD
into 8250_gsc, and fix the obsolete comment about reserving serial port slots.
Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

And now suddenly, linux32 works on parisc...
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Stop using PER_LINUX32 to designate processes needing
compaterizing. Convert is_compat_task to use TIF_32BIT and
set TIF_32BIT in binfmt_elf32.c
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

This patch fixes the pa8800 at a gross level (there are still other
subtle incoherency issues which can still cause crashes and HPMCs).

What it does is try to force eject inequivalent aliases before they
become visible to the L2 cache (which is where we get the incoherence
problems).

A new function (parisc_requires_coherency) is introduced in
asm/processor.h to identify the pa8x00 processors (8800 and 8900)
which have the issue.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Carlos O'Donell <carlos@systemhalted.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Clean up enough to get things compiling again in the interim.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Add header for McKinley bus related code. Remove extern decl
of proc_mckinley_root in drivers/parisc/sba_iommu.c
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Abstract existing shift register left macros as shift register
right are. This lends itself to a nice clean up of some #ifdef
blocks in entry.S
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

It seems PA7200 processors also suppress traps on loads to
%r0. This means we can prefetch for read on these cpus. Of course,
we can't support prefetch for write, since that requires
LOAD DOUBLEWORD which was added with PA2.0
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

asm/processor.h on parisc wants spinlocks for cpuinfo, but
linux/spinlock_types.h needs lockdep, and lockdep wants prefetch.

This leads to a horrible circular dependancy, because <asm/processor.h>
is including something which depends on things which are not defined
until the end of the file.

Kludge around this by moving prefetch related code into <asm/prefetch.h>
and including it before <linux/spinlock_types.h>, however this is just
a temporary solution until this mess can be cleaned up.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

This allows numaq to properly align cpus to their given node during
boot.  Pass logical apicid to apicid_to_node and allow the summit
sub-arch to use physical apicid (hard_smp_processor_id()).

Tested against numaq and summit based systems with no issues.
Signed-off-by: Keith Mannthey <kmannth@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial: (39 commits)
  Add missing maintainer countries in CREDITS
  Fix bytes <-> kilobytes  typo in Kconfig for ramdisk
  fix a typo in Documentation/pi-futex.txt
  BUG_ON conversion for fs/xfs/
  BUG_ON() conversion in fs/nfsd/
  BUG_ON conversion for fs/reiserfs
  BUG_ON cleanups in arch/i386
  BUG_ON cleanup in drivers/net/tokenring/
  BUG_ON cleanup for drivers/md/
  kerneldoc-typo in led-class.c
  debugfs: spelling fix
  rcutorture: Fix incorrect description of default for nreaders parameter
  parport: Remove space in function calls
  Michal Wronski: update contact info
  Spelling fix: "control" instead of "cotrol"
  reboot parameter in Documentation/kernel-parameters.txt
  Fix copy&waste bug in comment in scripts/kernel-doc
  remove duplicate "until" from kernel/workqueue.c
  ite_gpio fix tabbage
  fix file specification in comments
  ...

Fixed trivial path conflicts due to removed files:
   arch/mips/dec/boot/decstation.c, drivers/char/ite_gpio.c

Add missing maintainer countries.
Signed-off-by: Samuel Tardieu <sam@rfc1149.net>
Signed-off-by: Adrian Bunk <bunk@stusta.de>

This is a small fix for a typo in Kconfig. The default value for the block
size is 1024 bytes not 1024 kilobytes.
Signed-off-by: Christian Borntraeger <borntrae@de.ibm.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>

Signed-off-by: Riccardo Magliocchetti <riccardo@datahost.it>
Signed-off-by: Adrian Bunk <bunk@stusta.de>