Brian Murphy says:
    this fixes Lasat pci to work with multi-function devices by assigning
    the correct values based on pin number (instead of ignoring them).

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Thiemo Seufer <ths@networkno.de>

Code cleanup, thanks Brian Murphy.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Thiemo Seufer <ths@networkno.de>

Use physical addresses at the interface level, letting drivers remap
them as appropriate.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Prevent 64-bit constants from being cropped to 32 bits when used in C code.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Avoid tlbw* hazards for the R4600/R4700/R5000.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Inline ioremap() calls for constant addresses that map to KSEG1.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Avoid defining variables in the middle of a block which breaks older
compilers.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

More .set to keep 32-bit processors happy.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix the diagnostic dump for the XTLB refill handler.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix a diagnostic message.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Conversion to plat_setup() for TX4927 also.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Using get_nasid() to find the console node will blow up nicely if
called on a another node than the console node, so use the master_nasid
instead and in the unlikely case that one isn't initialized yet, fall
back to get_nasid().

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Always use ".set mips3" rather than select between "mips2" or "mips3"
for assembling ll/sc sequences to avoid problems with 64-bit
configurations.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Actual handlers for bus errors for Pmax and 3min.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
diff --git a/arch/mips/dec/kn01-berr.c b/arch/mips/dec/kn01-berr.c
new file mode 100644

System-specific handling of bus errors for DECstation variations
supporting parity errors only for memory (Pmax/3min/Maxine).
Fixes for resources decoded by the KN04/KN05 MB ASIC.  Additional
clean-ups for the ECC handler.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Use correct names for bits in the R3k cp0.status register.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Get the thing to compile again ...

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Use macros for the RM7k cp0.config bits instead of magic numbers.
Minor clean-ups.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix types for firmware arguments.  Don't define unneeded messages.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Remove left-over unused bits.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Mark __die() "noreturn" for real.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix function types to ones appropriate for initcalls.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

GCC 4.0.0 broke `attribute(("alias"))' -- resort to an assembly variant.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Deal with the bloody KSEG vs CKSEG horror...

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

There is NO port I/O space on the DECstation.  Minor clean-ups.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Redo RM9000 workaround which along with other DSP ASE changes was
causing some headache for debuggers knowing about signal frames.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Enable a suitable ISA for the assembler around ll/sc so that code
builds even for processors that don't support the instructions.
Plus minor formatting fixes.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Optimize R3k TLB Load/Store/Modified handlers, by scheduling
delay slots properly and avoiding an unnecessary jump to a jump.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fill R3k load delay slots properly.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Only dump instructions actually emitted.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

dec_esp: Use physical addresses

These should really be addresses obtained with ioremap() or some
bus-specific backend, but for now...

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

This interrupt is *always* handled -- MIPS_BE_DISCARD just means
no further action wanted.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix dependencies for DECstation framebuffers.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Remove dead code which was causing warnings.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

For MIPS32/MIPS64 cp0.config.mt == 1 implies a standard (R4k-style)
TLB, so no need to set it separately for each implementation.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

The DbAu1500 board also support big endian. Gee, imagine that.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix build with CONFIG_PRINTK disabled.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Support the MIPS32 / MIPS64 DSP ASE.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

64-bit fixes for Alchemy code ;)

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix build for CONFIG_BUG=n.  Yes, bugs are now a compile time option ;-)

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

When building for Atlas, Malta or SEAD convert the kernel to srecs by
default.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix tasteless #ifdef mess in audit_arch(), minor cleanups.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Bugs are now a configuration option.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

On MIPS the struct sigev preamble is only 8 bytes.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Now that a struct is the only member left in struct
mips_fpu_emulator_stats cleanup that unnecessary nesting of structs.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Get rid of the eir struct mips_fpu_emulator_private member.  It's
never initialized been initialized anywhere, just saved to and
restored from signal frames so nonsense anyway.  As neat side effect
of being shared between all processors it was also abusable as a
nice covert channel between processes.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

__compute_return_epc() uses CFC1 instruction which might result in a
coprocessor unusable exception since the process can lose its fpu
context by preemption.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

sys_nfsservctl() needs translation.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

No point in checking cpu_has_tlb before we've computed the CPU options.
So for now we just unconditionally set the option - Linux wouldn't
work without a TLB anyway.

Setting MIPS_CPU_4KTLB was missing for Alchemy and Sandcraft, add that
back.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Cleanup decoding of MIPSxx config registers.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix preemption and SMP problems in the FP emulator code.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Handle _PAGE_DIRTY correctly for CONFIG_64BIT_PHYS_ADDR on 32bit CPUs.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Better interface to run uncached cache setup code.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

* use 'unsigned long' as address supplied to au_write[bwl]()
* remove two already unused and commented structures
* added an ULL suffix to several address constants that use bits 35-32

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix D-cache aliasing problem in the PIO IDE driver potencially resulting
in the kernel or userspace seeing stale data.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix a few build warnings.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Need to include smp.h for the definition of smp_processor_id().

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Ustat needs a wrapper on n32.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Move sync into the delay slot here also.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Useless includes of everything, the kitchen sink and version.h ...

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Arrested for multiple offences of header file inclusion.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

We pass a kernel pointer to do_sigaltstack in sys32_sigaltstack, so
we need to do the set_fs(KERNEL_DS) thing around this call.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

NPTL, round one.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Build fix for certain configurations.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix BogoMIPS display on UP and some minor cosmetical things.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

fcntl64 needs to be wrapped for n32.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix one more case of computing the return EPC after the registers have
already been modified.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

When simulating ll/sc compute the return EPC before modifying the
registers.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Use fixed up pfn.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Detect the 4KEcR2 and for now detect handle it like the 4KEc.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fixed buglet with previous patch that broke non au1x builds.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Remove CONFIG_PM dependency from au1x wait in cpu_probe.
Additional work necessary to completely remove that config option.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Au1100 FB driver uplift for 2.6.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Antonino Daplas <adaplas@pol.net>

Fix race conditions for read_c0_entryhi. Remove broken ASID masks in
tlb-sb1.c. Make tlb-r4k.c and tlb-sb1.c more similiar and more efficient.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Remove useless casts.  Fix formatting.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix 64bit SMP TLB handler and stack frame handling, optimize 32bit SMP
TLB handlers a bit, match definitions in pgtable-{32,64}.h better.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

HUB interrupts are allocated per node, not per slice.  Make
manipulation of the interrupt mask register atomic by disabling
interrupts.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

HUB interrupts are allocated per node, not per slice.  Make manipulation
of the interrupt mask register atomic by disabling interrupts.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

R4300 delay slot.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Use compat_sigval_t in struct compat_siginfo.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

SECCOMP for MIPS.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Reformat; cosmetic cleanups.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Reformatting, remove debugging code.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Get rid of the the remains of 2.4-style ramdisk support.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Multithreaded core dumps.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

When CONFIG_PM is enabled, it uses the TOY_MATCH2 interrupt as the system
timer tick. Prior to this patch, if IDE IRQ probing occured, then the
TOY_MATCH2 interrupt would be permanently disabled, and no system timer
tick occurs. This patch corrects this situation by correctly registering
the TOY_MATCH2 interrupt so that IDE IRQ probing doesn't have adverse
side effects.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

sys_futex has 6 arguments.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

qtronix.c: Handle kmalloc failure.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Export shm_align_mask and flush_data_cache_page.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Put salone.o back into arclib.  It's a lib so doesn't harm if unused atm.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

It helps to not use a _mem_ function for requesting I/O space.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Gcc 4.0 fixes.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Use new txx9 serial driver.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Fix preemption bug.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Int controller fixes.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

One definition of back_to_back_c0_hazard too much.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Sparseify MIPS.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

On 24K we did always disable cache parity protection - obviously not
the greatest thing to do.  Try to enable parity protection, check if
we actually succeeded and print a message about the outcome of this.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

New hazard handling function back_to_back_c0_hazard() to handle back to
back mtc0 / mfc0 pairs from the same coprocessor register.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

ISOify.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>