pandora-kernel.git
18 years ago[PATCH] x86-64: Improve error handling for overlapping PXMs in SRAT.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Improve error handling for overlapping PXMs in SRAT.

- Report PXMs instead of nodes
- Report the correct PXM, not always the one of node 1.
- Only warn for the case of a PXM overlapping by itself

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Prevent gcc 4 from optimizing away vsyscalls
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Prevent gcc 4 from optimizing away vsyscalls

They were previously static.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Delivery mode should be APIC_DM_FIXED when using physical mode.
Ashok Raj [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Delivery mode should be APIC_DM_FIXED when using physical mode.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Replace extern inline with static inline in asm-x86_64/*
Adrian Bunk [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] Replace extern inline with static inline in asm-x86_64/*

They should be identical in the kernel now, but this
makes it consistent with other code.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove freeing of SMP trampoline pages
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove freeing of SMP trampoline pages

Nick points out it never worked because PageReserved was
set and it might cause problems later on. Also HOTPLUG_CPU
is much more common now so let's care not too much
about the !hotplug case.

Cc: nickpiggin@yahoo.com.au
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix typo CONFIG_CPU_HOTPLUG -> CONFIG_HOTPLUG_CPU in genapic.c
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix typo CONFIG_CPU_HOTPLUG -> CONFIG_HOTPLUG_CPU in genapic.c

Noted by Ashok Raj

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Remove unnecessary BUG_ON in irq.c
Alexander Nyberg [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] Remove unnecessary BUG_ON in irq.c

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Increase nodemap hash.
Nakul Saraiya [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] Increase nodemap hash.

Needed for some newer Opteron systems with E stepping and memory
relocation enabled. The node addresses are different in lower
bits now so the nodemap hash function needs to be enlarged.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix off by one in pfn_valid
Jim Paradis [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix off by one in pfn_valid

When I gave proposed the fix to pfn_valid() for RHEL4, Stephen Tweedie's
sharp eyes caught this:

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] i386: add memory clobbers to syscall macros
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] i386: add memory clobbers to syscall macros

As noted by matz@suse.de

The problem is, that on i386 the syscallN
macro is defined like so:

  long __res; \
  __asm__ volatile ("int $0x80" \
        : "=a" (__res) \
        : "0" (__NR_##name),"b" ((long)(arg1)),"c" ((long)(arg2)), \
          "d" ((long)(arg3)),"S" ((long)(arg4)),"D" ((long)(arg5))); \

If one of the arguments (in the _llseek syscall it's the arg4) is a pointer
which the syscall is expected to write to (to the memory pointed to by this
ptr), then this side-effect is not captured in the asm.

If anyone uses this macro to define it's own version of the syscall
(sometimes necessary when not using glibc) and it's inlined, then GCC
doesn't know that this asm write to "*dest", when called like so for instance:

  out = 1;
  llseek (fd, bla, blubb, &out, trara)
  use (out);

Here nobody tells GCC that "out" actually is written to (just a pointer to it
is passed to the asm).  Hence GCC might (and in the above bug did)
copy-propagate "1" into the second use of "out".

The easiest solution would be to add a "memory" clobber to the definition
of this syscall macro.  As this is a syscall, it shouldn't inhibit too many
optimizations.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix show_mem a little bit
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix show_mem a little bit

- Add KERN_INFO to printks (from i386)
- Use longs instead of ints to accumulate pages.
- Fix broken indenting.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove disable_tsc code in context switch
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove disable_tsc code in context switch

It only offers extremly dubious security advantages and
is not worth the overhead in this critical path.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Print version at end of kernel build
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Print version at end of kernel build

(from i386)

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix (harmless) typo in head.S early level2 page table
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix (harmless) typo in head.S early level2 page table

The global bit  was not set in the first 2MB page, instead
it had a bit in the free AVL section which is useless.
Fixed thus.

Noticed by Eric Biederman

Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix idle=poll
Hugh Dickins [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix idle=poll

x86_64 idle=poll might be a little less responsive than it should: unlike
mwait_idle, and unlike i386, its poll_idle left TIF_POLLING_NRFLAG set.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Only allocate per cpu data for possible CPUs, not compiled in CPUs.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Only allocate per cpu data for possible CPUs, not compiled in CPUs.

Saves some memory except for hotplug situations.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Increase TLB flush array size
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Increase TLB flush array size

The generic TLB flush functions kept upto 506 pages per
CPU to avoid too frequent IPIs.

This value was done for the L1 cache of older x86 CPUs,
but with modern CPUs it does not make much sense anymore.
TLB flushing is slow enough that using the L2 cache is fine.

This patch increases the flush array on x86-64 to cache
5350 pages. That is roughly 20MB with 4K pages. It speeds
up large munmaps in multithreaded processes on SMP considerably.

The cost is roughly 42k of memory per CPU, which is reasonable.

I only increased it on x86-64 for now, but it would probably
make sense to increase it everywhere. Embedded architectures
with SMP may keep it smaller to save some memory per CPU.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't include config.h in asm/timex.h
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't include config.h in asm/timex.h

asm-x86-64/timex.h does not reference CONFIG constants.
Do not need to include config.h.

Signed-off-by: Grant Grundler <iod00d@hp.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Some cleanup and optimization to the processor data area.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Some cleanup and optimization to the processor data area.

- Remove unused irqrsp field
- Remove pda->me
- Optimize set_softirq_pending slightly

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Add simnow console
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Add simnow console

This adds console and earlyprintk support for a host file
on AMD's SimNow simulator.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Make remote TLB flush more scalable
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Make remote TLB flush more scalable

Instead of using a global spinlock to protect the state
of the remote TLB flush use a lock and state for each sending CPU.

To tell the receiver where to look for the state use 8 different
call vectors.  Each CPU uses a specific vector to trigger flushes on other
CPUs. Depending on the received vector the target CPUs look into
the right per cpu variable for the flush data.

When the system has more than 8 CPUs they are hashed to the 8 available
vectors. The limited global vector space forces us to this right now.
In future when interrupts are split into per CPU domains this could be
fixed, at the cost of needing more IPIs in flat mode.

Also some minor cleanup in the smp flush code and remove some outdated
debug code.

Requires patch to move cpu_possible_map setup earlier.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix 32bit sendfile
Tsuneo.Yoshioka@f-secure.com [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix 32bit sendfile

If we use 64bit kernel on ia64/x86_64/s390 architecture, and we run
32bit binary on 32bit compatibility mode, sendfile system call seems be
not set offset argument.

This is because sendfile's return value is not zero but the code regards
the result by return value is zero or not.

This problem will be affect to ia64/x86_64/s390 and not affect to other
architecture does not affect other architecture (mips/parisc/ppc64/sparc64).

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Include build number in oops output
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Include build number in oops output

Include build number in oops output

Helps me to match oopses to correct kernel.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Use ACPI PXM to parse PCI<->node assignments
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Use ACPI PXM to parse PCI<->node assignments

Since this is shared code I had to implement it for i386 too

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove code to resume machine check state of other CPUs.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove code to resume machine check state of other CPUs.

The resume code uses CPU hotplug now so at resume time
we only ever see one CPU.

Pointed out by Yu Luming.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove redundant max_mapnr and replace with end_pfn
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove redundant max_mapnr and replace with end_pfn

The FLATMEM people added it, but there doesn't seem a good reason
because end_pfn is identical.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Use e820_find_hole to compute reserved pages
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Use e820_find_hole to compute reserved pages

Avoids a very dumb loop

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix harmless off by one in e820 code
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix harmless off by one in e820 code

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Micro optimization to dma_alloc_coherent node lookup
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Micro optimization to dma_alloc_coherent node lookup

Use pcibus_to_node directly

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't trust boot_cpu_id in the mptable.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't trust boot_cpu_id in the mptable.

It could be wrong for kexec or other cases. Read it from
the CPU instead.

Signed-off-by: Murali <muralim@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Add command line option to set machine check tolerance level
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Add command line option to set machine check tolerance level

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Make lockless machine check record passing a bit more robust.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Make lockless machine check record passing a bit more robust.

One machine is constantly throwing NMI watchdog timeouts in mce_log

This was one attempt to fix it.

(AK: this doesn't actually fix the bug I'm seeing unfortunately, probably
drop.  I don't like it that the reader can spin forever now waiting
for a writer)

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove redundant swiotlb=force handling
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove redundant swiotlb=force handling

It's already handled in the main swiotlb code.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't allocate aperture when swiotlb is enabled and no AGP found
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't allocate aperture when swiotlb is enabled and no AGP found

No-one needs it then

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Reverse order of bootmem lists
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Reverse order of bootmem lists

This leads to bootmem allocating first from node 0 instead
of from the last node.  This avoids swiotlb allocating on the last node, which
doesn't really work on a machine with >4GB.

Note: there is a better patch around from someone else that gets
rid of the pgdat list completely.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Avoid unnecessary double bouncing for swiotlb
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Avoid unnecessary double bouncing for swiotlb

PCI_DMA_BUS_IS_PHYS has to be zero even when the GART IOMMU is disabled
and the swiotlb is used. Otherwise the block layer does unnecessary
double bouncing.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't assume APIC for boot processor has an APIC ID of zero
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't assume APIC for boot processor has an APIC ID of zero

Originally from Stuart Hayes.

When setting up the APIC for the Uniprocessor kernel don't
assume the CPU has an APIC ID of zero.

This fixes boot with the UP kernel on Dell PowerEdge 6800/6850 4way systems.

Cc: Stuart.Hayes@dell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Use SRAT data on Intel systems too.
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Use SRAT data on Intel systems too.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Support dualcore and 8 socket systems in k8 fallback node parsing
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Support dualcore and 8 socket systems in k8 fallback node parsing

In particular on systems where the local APIC space and node space
is very different from the Linux CPU number space.

Previously the older NUMA setup code directly parsing the K8
northbridge registers had some issues on 8 socket or dual core
systems. This patch fixes them.

This is mainly done by fixing some confusion between Linux
CPU numbers and local APIC ids. We now pass the local APIC IDs
to later code, which avoids mismatches.

Also add some heuristics to detect cases where the Hypertransport
nodeids and the local APIC IDs don't match, but are shifted
by a constant offset.

This is still all quite hackish, hopefully BIOS writers fill
in correct SRATs instead.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't cache align PDA on UP builds
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't cache align PDA on UP builds

Suggested by someone I forgot who sorry.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Fix the apic version that gets printed during boot
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Fix the apic version that gets printed during boot

Signed-off-by: Suresh Sidda <suresh.b.siddha@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Don't assign CPU numbers in SRAT parsing
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Don't assign CPU numbers in SRAT parsing

Do that later when the CPU boots. SRAT just stores the APIC<->Node
mapping node. This fixes problems on systems where the order
of SRAT entries does not match the MADT.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: White space and comment fixes for smp_call_function_single
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: White space and comment fixes for smp_call_function_single

No functional changes

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Enable interrupts during delay calibration on APs
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Enable interrupts during delay calibration on APs

We used to disable them to work around a bug, but that
is not needed anymore. Keeping them enabled avoids the NMI
watchdog triggering in some cases.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Use largest APIC number, not number of CPUs to decide on physflat...
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Use largest APIC number, not number of CPUs to decide on physflat mode

Handles case where BIOS gives CPUs very large APIC numbers correctly.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove code for outdated APICs
Andi Kleen [Mon, 12 Sep 2005 16:49:24 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove code for outdated APICs

No x86-64 chipset has these APICs.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove esr disable hack in APIC code
Andi Kleen [Mon, 12 Sep 2005 16:49:23 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove esr disable hack in APIC code

This was just needed for the Numasaurus, which fortunately
doesn't support x86-64 CPUs.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove obsolete APIC "write around" bug workaround
Andi Kleen [Mon, 12 Sep 2005 16:49:23 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove obsolete APIC "write around" bug workaround

No x86-64 chipset has this bug

Generated code doesn't change because it was always disabled.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Remove apic_write_around from smpboot.c
Andi Kleen [Mon, 12 Sep 2005 16:49:23 +0000 (18:49 +0200)]
[PATCH] x86-64: Remove apic_write_around from smpboot.c

We don't do workarounds for ancient hardware bugs.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86-64: Update defconfig
Andi Kleen [Mon, 12 Sep 2005 16:49:23 +0000 (18:49 +0200)]
[PATCH] x86-64: Update defconfig

Allow diskless booting.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] x86{-64}: Remove old hack that disabled mmconfig support on AMD systems.
Andi Kleen [Mon, 12 Sep 2005 16:49:23 +0000 (18:49 +0200)]
[PATCH] x86{-64}: Remove old hack that disabled mmconfig support on AMD systems.

Now that Greg implemented MCFG/_SEG support this shouldn't be needed
anymore

Cc: gregkh@suse.de
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] cpuset semaphore depth check optimize
Paul Jackson [Mon, 12 Sep 2005 11:30:30 +0000 (04:30 -0700)]
[PATCH] cpuset semaphore depth check optimize

Optimize the deadlock avoidance check on the global cpuset
semaphore cpuset_sem.  Instead of adding a depth counter to the
task struct of each task, rather just two words are enough, one
to store the depth and the other the current cpuset_sem holder.

Thanks to Nikita Danilov for the idea.

Signed-off-by: Paul Jackson <pj@sgi.com>
[ We may want to change this further, but at least it's now
  a totally internal decision to the cpusets code ]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] crc16: remove w1 specific comments.
Evgeniy Polyakov [Mon, 12 Sep 2005 13:12:43 +0000 (17:12 +0400)]
[PATCH] crc16: remove w1 specific comments.

Remove w1 comments from crc16.h and move specific constants into
w1_ds2433.c where they are used.

Replace %d with %zd.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/paulus/ppc64-2.6
Linus Torvalds [Mon, 12 Sep 2005 15:33:53 +0000 (08:33 -0700)]
Merge /pub/scm/linux/kernel/git/paulus/ppc64-2.6

18 years agoMark ia64-specific MCA/INIT scheduler hooks as dangerous
Linus Torvalds [Mon, 12 Sep 2005 14:59:21 +0000 (07:59 -0700)]
Mark ia64-specific MCA/INIT scheduler hooks as dangerous

..and only enable them for ia64. The functions are only valid
when the whole system has been totally stopped and no scheduler
activity is ongoing on any CPU, and interrupts are globally
disabled.

In other words, they aren't useful for anything else. So make
sure that nobody can use them by mistake.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] i386: Don't miss pending signals returning to user mode after signal processing
Roland McGrath [Sun, 11 Sep 2005 08:44:45 +0000 (01:44 -0700)]
[PATCH] i386: Don't miss pending signals returning to user mode after signal processing

Signed-off-by: Roland McGrath <roland@redhat.com>
Acked-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] ppc64: Remove unused code
Anton Blanchard [Mon, 12 Sep 2005 03:19:51 +0000 (13:19 +1000)]
[PATCH] ppc64: Remove unused code

ppc64_attention_msg and ppc64_dump_msg are not used so remove them.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: Fix for missing start-cpu rtas token
Anton Blanchard [Mon, 12 Sep 2005 03:17:27 +0000 (13:17 +1000)]
[PATCH] ppc64: Fix for missing start-cpu rtas token

If the rtas start-cpu token doesnt exist then presume the cpu is already
spinning. If it isnt we will catch it later on when the cpu doesnt
respond.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: indent pci code
Anton Blanchard [Mon, 12 Sep 2005 03:14:26 +0000 (13:14 +1000)]
[PATCH] ppc64: indent pci code

Fix up some badly indented code.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: xics cleanup
Anton Blanchard [Mon, 12 Sep 2005 03:12:11 +0000 (13:12 +1000)]
[PATCH] ppc64: xics cleanup

A few xics cleanups:

- Make some things static.
- Be more consistent with error printing - interrupts are unsigned,
  error values are signed.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: Add ptrace data breakpoint support
Anton Blanchard [Sat, 10 Sep 2005 06:01:11 +0000 (16:01 +1000)]
[PATCH] ppc64: Add ptrace data breakpoint support

Add hardware data breakpoint support.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: Add definitions for new PTRACE calls
Anton Blanchard [Sat, 10 Sep 2005 06:01:10 +0000 (16:01 +1000)]
[PATCH] ppc64: Add definitions for new PTRACE calls

- Add PTRACE_GET_DEBUGREG/PTRACE_SET_DEBUGREG. The definition is
  as follows:

/*
 * Get or set a debug register. The first 16 are DABR registers and the
 * second 16 are IABR registers.
 */
#define PTRACE_GET_DEBUGREG    25
#define PTRACE_SET_DEBUGREG    26

  DABR == data breakpoint and IABR = instruction breakpoint in IBM
  speak. We could split out the IABR into 2 more ptrace calls but I
  figured there was no need and 16 DABR registers should be more
  than enough (POWER4/POWER5 have one).

- Add 2 new SIGTRAP si_codes: TRAP_HWBKPT and TRAP_BRANCH. I couldnt
  find any standards on either of these so I copied what ia64 is
  doing. Again this might be better placed in
  include/asm-generic/siginfo.h

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: Fix up some whitespace issues in ptrace32.c
Anton Blanchard [Sat, 10 Sep 2005 06:01:09 +0000 (16:01 +1000)]
[PATCH] ppc64: Fix up some whitespace issues in ptrace32.c

Fix up some whitespace issues in ptrace32.c

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: ptrace cleanups
Anton Blanchard [Sat, 10 Sep 2005 06:01:08 +0000 (16:01 +1000)]
[PATCH] ppc64: ptrace cleanups

- Remove the PPC_REG* defines
- Wrap some more stuff with ifdef __KERNEL__
- Add missing PT_TRAP, PT_DAR, PT_DSISR defines
- Add PTRACE_GETEVRREGS/PTRACE_SETEVRREGS, even though we dont use it on
  ppc64 we dont want to allocate them for something else.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] ppc64: Add PTRACE_{GET|SET}VRREGS
Robert Jennings [Sat, 10 Sep 2005 06:01:07 +0000 (16:01 +1000)]
[PATCH] ppc64: Add PTRACE_{GET|SET}VRREGS

The ptrace get and set methods for VMX/Altivec registers present in the
ppc tree were missing for ppc64.  This patch adds the 32-bit and
64-bit methods.  Updated with the suggestions from Anton following the lines
of his code snippet.

Added:
 - flush_altivec_to_thread calls as suggested by Anton
 - piecewise copy of structure to preserve 32-bit vrsave data as per
   Anton

(I consolidated the 32 and 64bit versions with 2 helper macros - Anton)

Signed-off-by: Robert C Jennings <rcjenn@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years agoppc64: Set up PCI tree from Open Firmware device tree
Paul Mackerras [Mon, 12 Sep 2005 07:17:36 +0000 (17:17 +1000)]
ppc64: Set up PCI tree from Open Firmware device tree

This adds code which gives us the option on ppc64 of instantiating the
PCI tree (the tree of pci_bus and pci_dev structs) from the Open
Firmware device tree rather than by probing PCI configuration space.
The OF device tree has a node for each PCI device and bridge in the
system, with properties that tell us what addresses the firmware has
configured for them and other details.

There are a couple of reasons why this is needed.  First, on systems
with a hypervisor, there is a PCI-PCI bridge per slot under the PCI
host bridges.  These PCI-PCI bridges have special isolation features
for virtualization.  We can't write to their config space, and we are
not supposed to be reading their config space either.  The firmware
tells us about the address ranges that they pass in the OF device
tree.

Secondly, on powermacs, the interrupt controller is in a PCI device
that may be behind a PCI-PCI bridge.  If we happened to take an
interrupt just at the point when the device or a bridge on the path to
it was disabled for probing, we would crash when we try to access the
interrupt controller.

I have implemented a platform-specific function which is called for
each PCI bridge (host or PCI-PCI) to say whether the code should look
in the device tree or use normal PCI probing for the devices under
that bridge.  On pSeries machines we use the device tree if we're
running under a hypervisor, otherwise we use normal probing.  On
powermacs we use normal probing for the AGP bridge, since the device
for the AGP bridge itself isn't shown in the device tree (at least on
my G5), and the device tree for everything else.

This has been tested on a dual G5 powermac, a partition on a POWER5
machine (running under the hypervisor), and a legacy iSeries
partition.

Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[PATCH] i810fb: Update i810fb documentation
Antonino A. Daplas [Mon, 12 Sep 2005 01:18:12 +0000 (09:18 +0800)]
[PATCH] i810fb: Update i810fb documentation

Update i810fb documentation to describe new features and configuration
changes.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] i810fb: Change option ext_vga to extvga to match documentation
Antonino A. Daplas [Mon, 12 Sep 2005 01:16:47 +0000 (09:16 +0800)]
[PATCH] i810fb: Change option ext_vga to extvga to match documentation

Reported by: Manuel Lauss <mano@roarinelk.homelinux.net>

He was getting random initial video modes depending on the kernel
configuration.  His option line includes 'extvga'.

The i810fb documentation describes the option 'extvga', however the
driver accepts 'ext_vga'.  Besides 'extvga' being ignored by i810fb,
it also confuses the option parser of i810fb and assigns 'extvga' to
'mode_option'.  This leads to an incorrect video mode at boot time.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] i810fb: Restore xres and yres option parameters
Antonino A. Daplas [Mon, 12 Sep 2005 01:15:16 +0000 (09:15 +0800)]
[PATCH] i810fb: Restore xres and yres option parameters

If i810fb successfully probed for the EDID, it will disregard the
boot option parameters 'xres' and 'yres'.  Fix this regression.

Excellent testing done by Manuel Lauss <mano@roarinelk.homelinux.net>.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] i810fb: Fix oops if CONFIG_FB_I810_I2C is set to no
Antonino A. Daplas [Mon, 12 Sep 2005 01:13:32 +0000 (09:13 +0800)]
[PATCH] i810fb: Fix oops if CONFIG_FB_I810_I2C is set to no

Reported by: Manuel Lauss <mano@roarinelk.homelinux.net>

compiled with CONFIG_FB_I810_I2C = n and CONFIG_FB_I810 = y
it oopses at boot in file drivers/video/i810/i810_main.c:1884

...
Unable to handle kernel NULL pointer dereference at virtual address 00000054
printing eip:
c02543c0
*pde = 00000000
Oops: 0000 [#1]
last sysfs file:
Modules linked in:
CPU:    0
EIP:    0060:[<c02543c0>]    Not tainted VLI
EFLAGS: 00010286   (2.6.13-mm2)
EIP is at i810fb_find_init_mode+0x53/0x93
eax: c113ddd4   ebx: c1194000   ecx: c04be2dd   edx: c1194000
esi: c1194008   edi: c113ddd4   ebp: c1194240   esp: c113ddcc
ds: 007b   es: 007b   ss: 0068

struct fb_monspecs *specs is initialized to NULL causing the oops.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: common RAM based 68360 startup code
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: common RAM based 68360 startup code

Create common start code for all 68360 based platforms that are
loaded and run directly from RAM (as opposed to running from
flash/ROM). This replaces the old specific startup code for
each board.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: allow for SDRAM and GPIO differences on 5270/1 and 5274/5 processors
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: allow for SDRAM and GPIO differences on 5270/1 and 5274/5 processors

Allow for differences in the SDRAM controller setup and GPIO pin setup
of the 5270/1 and 5274/5 parts. With separate config options for each
now this no longer needs to be board specific.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: FEC eth definitions support for the 523x Coldfire processor family
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: FEC eth definitions support for the 523x Coldfire processor family

Add support for the FEC ethernet driver of the Freescale 523x processor
family to the FEC header definitions.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: add SPI register definitions for 528x processors
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: add SPI register definitions for 528x processors

Add QSPI register definitions of ColdFIre 528x processor SPI controller.

Patch originally submitted by Derek Cheung <derek.cheung@sympatico.ca>

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uclinux: add NULL check, 0 end valid check and some more exports to nommu.c
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] uclinux: add NULL check, 0 end valid check and some more exports to nommu.c

Move call to get_mm_counter() in update_mem_hiwater() to be
inside the check for tsk->mm being null. Otherwise you can be
following a null pointer here. This patch submitted by
Javier Herrero <jherrero@hvsistemas.es>.

Modify the end check for munmap regions to allow for the
legacy behavior of 0 being valid. Pretty much all current
uClinux system libc malloc's pass in 0 as the end point.
A hard check will fail on these, so change the check so
that if it is non-zero it must be valid otherwise it fails.
A passed in value will always succeed (as it used too).

Also export a few more mm system functions - to be consistent
with the VM code exports.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: dma support for 523x processors
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: dma support for 523x processors

Support the DMA unit of the ColdFire 523x processor family.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uclinux: remove use of mtd_put_device() in uclinux MTD map driver
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] uclinux: remove use of mtd_put_device() in uclinux MTD map driver

We should not call mtd_put_device() in the uclinux MTD map driver.
Also consistently use phys/virt fields of maps map_info struct,
instead of mixing it with map_priv_1.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: add ColdFire serial driver support for 523x processor family
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: add ColdFire serial driver support for 523x processor family

Add support for the new Freescale 523x processor family to ColdFire
serial driver. Also set different default baud rate for MOD5272
board.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: change addr arg to const in bitops.h/find_next_zero_bit()
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: change addr arg to const in bitops.h/find_next_zero_bit()

Change addr arg to find_next_zero_bit to be a const.
Cleans up compiler warning.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: correct prototype args in checksum.h
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: correct prototype args in checksum.h

Bring arg types for csum_partial_copy and csum_paritial_copy_from_user
prototypes into line with their actual implementation.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: common ROM/flash based 68360 startup code
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: common ROM/flash based 68360 startup code

Create common start code for all 68360 based platforms that are
loaded and run directly from ROM/flash (as opposed to running from
RAM). This replaces the old specific startup code for each board.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: cache support for 523x/528x processors
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: cache support for 523x/528x processors

Add support for the cache of the ColdFIre 523x family of processors.
Enable the 528x cache by default now, all final shipping silicon
has the cache bug fixed.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: ColdFire FEC eth driver improvements
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: ColdFire FEC eth driver improvements

A few improvements to the Freescale/ColdFire FEC driver:

. some formatting cleanups
. add support for the FEC device in the ColdFire 523x processor family
. add support for MAC address setting on MOD5272 and M5272C3 boards
. don't re-read the PHY status register many times
. ack status interrupt before reading status register
. move printing init message to after full init (so that the
  ethX name is filled out for printing)

Some parts of this patch submitted by Philippe De Muyter <phdm@macqel.be>

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] m68knommu: config support for FEC eth of 523x Coldfire processor family
Greg Ungerer [Mon, 12 Sep 2005 01:18:10 +0000 (11:18 +1000)]
[PATCH] m68knommu: config support for FEC eth of 523x Coldfire processor family

Add configuration support for the FEC ethernet controller in the
Freescale 523x processor family. Also add and option to configure
the second FEC controller on some Freescale processors.

Signed-off-by: Greg Ungerer <gerg@uclinux.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge branch 'release' of master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Sun, 11 Sep 2005 22:51:40 +0000 (15:51 -0700)]
Merge branch 'release' of /linux/kernel/git/aegl/linux-2.6

18 years agoPull sn-features into release branch
Tony Luck [Sun, 11 Sep 2005 21:34:23 +0000 (14:34 -0700)]
Pull sn-features into release branch

18 years agoPull sim-fixes into release branch
Tony Luck [Sun, 11 Sep 2005 21:27:15 +0000 (14:27 -0700)]
Pull sim-fixes into release branch

18 years ago[IA64] MCA/INIT: remove obsolete unwind code
Keith Owens [Sun, 11 Sep 2005 07:24:42 +0000 (17:24 +1000)]
[IA64] MCA/INIT: remove obsolete unwind code

Delete the special case unwind code that was only used by the old
MCA/INIT handler.

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years ago[IA64] MCA/INIT: remove the physical mode path from minstate.h
Keith Owens [Sun, 11 Sep 2005 07:23:42 +0000 (17:23 +1000)]
[IA64] MCA/INIT: remove the physical mode path from minstate.h

Remove the physical mode path from minstate.h.

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years ago[PATCH] MCA/INIT: use per cpu stacks
Keith Owens [Sun, 11 Sep 2005 07:22:53 +0000 (17:22 +1000)]
[PATCH] MCA/INIT: use per cpu stacks

The bulk of the change.  Use per cpu MCA/INIT stacks.  Change the SAL
to OS state (sos) to be per process.  Do all the assembler work on the
MCA/INIT stacks, leaving the original stack alone.  Pass per cpu state
data to the C handlers for MCA and INIT, which also means changing the
mca_drv interfaces slightly.  Lots of verification on whether the
original stack is usable before converting it to a sleeping process.

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years ago[IA64] MCA/INIT: avoid reading INIT record during INIT event
Keith Owens [Sun, 11 Sep 2005 07:21:46 +0000 (17:21 +1000)]
[IA64] MCA/INIT: avoid reading INIT record during INIT event

Reading the INIT record from SAL during the INIT event has proved to be
unreliable, and a source of hangs during INIT processing.  The new
MCA/INIT handlers remove the need to get the INIT record from SAL.
Change salinfo.c so mca.c can just flag that a new record is available,
without having to read the record during INIT processing.  This patch
can be applied without the new MCA/INIT handlers.

Also clean up some usage of NR_CPUS which should have been using
cpu_online().

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years ago[IA64] MCA/INIT: add an extra thread_info flag
Keith Owens [Sun, 11 Sep 2005 07:20:14 +0000 (17:20 +1000)]
[IA64] MCA/INIT: add an extra thread_info flag

Add an extra thread_info flag to indicate the special MCA/INIT stacks.
Mainly for debuggers.

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years ago[PATCH] MCA/INIT: scheduler hooks
Keith Owens [Sun, 11 Sep 2005 07:19:06 +0000 (17:19 +1000)]
[PATCH] MCA/INIT: scheduler hooks

Scheduler hooks to see/change which process is deemed to be on a cpu.

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
18 years agomips: rename offsets.c to asm-offsets.c
Sam Ravnborg [Sun, 11 Sep 2005 20:35:18 +0000 (22:35 +0200)]
mips: rename offsets.c to asm-offsets.c

Cannot build MIPS now.
We need to change offset.c to asm-offsets.c

Signed-off-by: Yoichi Yuasa <yuasa@hh.iij4u.or.jp>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
18 years agokbuild: fix silentoldconfig with make O=
Sam Ravnborg [Sun, 11 Sep 2005 20:32:57 +0000 (22:32 +0200)]
kbuild: fix silentoldconfig with make O=

Al Viro reported that sometimes silentoldconfig failed because
output directory was missing.
So create it unconditionally before executing conf

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
18 years agokbuild: rename prepare to archprepare to fix dependency chain
Sam Ravnborg [Sun, 11 Sep 2005 20:30:22 +0000 (22:30 +0200)]
kbuild: rename prepare to archprepare to fix dependency chain

When introducing the generic asm-offsets.h support the dependency
chain for the prepare targets was changed. All build scripts expecting
include/asm/asm-offsets.h to be made when using the prepare target would broke.
With the limited number of prepare targets left in arch Makefiles
the trivial solution was to introduce a new arch specific target: archprepare

The dependency chain looks like this now:

prepare
  |
  +--> prepare0
         |
         +--> archprepare
                |
+--> scripts_basic
                +--> prepare1
                       |
                       +---> prepare2
                               |
                               +--> prepare3

So prepare 3 is processed before prepare2 etc.
This guaantees that the asm symlink, version.h, scripts_basic
are all updated before archprepare is processed.

prepare0 which build the asm-offsets.h file will need the
actions performed by archprepare.

The head target is now named prepare, because users scripts will most
likely use that target, but prepare-all has been kept for compatibility.
Updated Documentation/kbuild/makefiles.txt.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband
Linus Torvalds [Sun, 11 Sep 2005 17:16:07 +0000 (10:16 -0700)]
Merge /pub/scm/linux/kernel/git/roland/infiniband

18 years agoMerge branch 'for-linus' from kernel.org:/.../shaggy/jfs-2.6 manually
Linus Torvalds [Sun, 11 Sep 2005 17:14:54 +0000 (10:14 -0700)]
Merge branch 'for-linus' from /.../shaggy/jfs-2.6 manually

Clash due to new delete_inode behavior (the filesystem now needs to do
the truncate_inode_pages() call itself).

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Sun HME: enable and map PCI ROM properly
Willy Tarreau [Sun, 11 Sep 2005 07:04:07 +0000 (09:04 +0200)]
[PATCH] Sun HME: enable and map PCI ROM properly

This ports the Sun GEM ROM mapping/enable fixes it sunhme (which used
the same PCI ROM mapping code).

Without this, I get NULL MAC addresses for all 4 ports (it's a SUN QFE).
With it, I get the correct addresses (the ones printed on the label on
the card).

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoSun GEM ethernet: enable and map PCI ROM properly
Linus Torvalds [Sun, 11 Sep 2005 16:26:20 +0000 (09:26 -0700)]
Sun GEM ethernet: enable and map PCI ROM properly

This same patch was reported to fix the MAC address detection on sunhme
(next patch).  Most people seem to be running this on Sparcs or PPC
machines, where we get the MAC address from their respective firmware
rather than from the (previously broken) ROM mapping routines.

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