pandora-kernel.git
15 years agosh: ioremap_prot support.
Paul Mundt [Fri, 12 Sep 2008 11:41:05 +0000 (20:41 +0900)]
sh: ioremap_prot support.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: ptrace: Introduce user_regset interface for gp regs.
Paul Mundt [Fri, 12 Sep 2008 10:52:36 +0000 (19:52 +0900)]
sh: ptrace: Introduce user_regset interface for gp regs.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Provide a fixed UTS_MACHINE definition for sh64.
Paul Mundt [Fri, 12 Sep 2008 09:58:28 +0000 (18:58 +0900)]
sh: Provide a fixed UTS_MACHINE definition for sh64.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: update edosk7760 defconfig for physmap-flash.
Luca Santini [Fri, 12 Sep 2008 09:08:01 +0000 (18:08 +0900)]
sh: update edosk7760 defconfig for physmap-flash.

Signed-off-by: Luca Santini <luca.santini@spesonline.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: edosk7760 physmap-flash support.
Luca Santini [Fri, 12 Sep 2008 09:07:16 +0000 (18:07 +0900)]
sh: edosk7760 physmap-flash support.

Signed-off-by: Luca Santini <luca.santini@spesonline.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agortc-sh: Unsigned rtc->{periodic,carry,alarm}_irq cannot be negative
roel kluin [Wed, 10 Sep 2008 17:34:44 +0000 (19:34 +0200)]
rtc-sh: Unsigned rtc->{periodic,carry,alarm}_irq cannot be negative

possibly since commit b420b1a7a17ea88531d0e12b2f2679a0c8365803

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh_eth: unsigned ndev->irq cannot be negative
roel kluin [Wed, 10 Sep 2008 17:22:44 +0000 (19:22 +0200)]
sh_eth: unsigned ndev->irq cannot be negative

unsigned ndev->irq cannot be negative

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: intc_prio_data() test before subtraction on unsigned
roel kluin [Tue, 9 Sep 2008 21:02:43 +0000 (23:02 +0200)]
sh: intc_prio_data() test before subtraction on unsigned

bit is unsigned, so test before subtraction

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: kretprobe_trampoline needs to be global.
Paul Mundt [Mon, 8 Sep 2008 23:13:28 +0000 (08:13 +0900)]
sh: kprobes: kretprobe_trampoline needs to be global.

Needed by CONFIG_TRACING.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add R2D+ defconfig for qemu system emulator.
Paul Mundt [Mon, 8 Sep 2008 21:15:47 +0000 (06:15 +0900)]
sh: Add R2D+ defconfig for qemu system emulator.

This adds a defconfig for the R2D+ target in the qemu system emulator.
Eventually it will be possible to simply use the r2d+ defconfig as it is.

Provided by Shin-ichiro KAWASAKI.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Derive calibrate_delay lpj from clk fwk.
Paul Mundt [Mon, 8 Sep 2008 11:47:42 +0000 (20:47 +0900)]
sh: Derive calibrate_delay lpj from clk fwk.

All CPUs must have a sensible cpu_clk definition these days, which we can
safely use for deriving the preset loops_per_jiffy. The only odd one out
is SH-5, which hasn't been hammered in to the framework yet.

Based on the ST patch.

Signed-off-by: Francesco Virlinzi <francesco.virlinzi@st.com>
Signed-off-by: Carl Shaw <carl.shaw@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: __kprobes annotations and formatting cleanups.
Paul Mundt [Mon, 8 Sep 2008 09:22:47 +0000 (18:22 +0900)]
sh: kprobes: __kprobes annotations and formatting cleanups.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: Fix up race against probe point removal.
Paul Mundt [Mon, 8 Sep 2008 09:15:55 +0000 (18:15 +0900)]
sh: kprobes: Fix up race against probe point removal.

Handle a corner case where another CPU or debugger removes the probe
point from underneath us.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: Fix up a preemption imbalance on jprobe return.
Paul Mundt [Mon, 8 Sep 2008 09:14:50 +0000 (18:14 +0900)]
sh: kprobes: Fix up a preemption imbalance on jprobe return.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: Default to NOTIFY_DONE for unhandled debug traps.
Paul Mundt [Mon, 8 Sep 2008 09:12:33 +0000 (18:12 +0900)]
sh: kprobes: Default to NOTIFY_DONE for unhandled debug traps.

Presently this is doing a force_sig() SIGTRAP, which is already taken
care of in the generic code if no one asserts NOTIFY_STOP. Switch the
default return to NOTIFY_DONE in the case of unhandled traps, so that
the same trap may pass through to other users on the same die chain.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: Use trapa #0x3a for breakpoint trap.
Paul Mundt [Mon, 8 Sep 2008 09:10:10 +0000 (18:10 +0900)]
sh: kprobes: Use trapa #0x3a for breakpoint trap.

Not all parts support trapa #0xff, so use something within the debug trap
range that's accessible on all parts.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: kprobes: Hook up kprobe_fault_handler() in the page fault path.
Paul Mundt [Mon, 8 Sep 2008 03:22:47 +0000 (12:22 +0900)]
sh: kprobes: Hook up kprobe_fault_handler() in the page fault path.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Disable seccomp support by default.
Paul Mundt [Mon, 8 Sep 2008 03:10:35 +0000 (12:10 +0900)]
sh: Disable seccomp support by default.

This was initially checked in with a stupid default of y, while most
everyone is going to want to have this disabled anyways.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Mark kretprobe_trampoline_holder static and __used.
Paul Mundt [Mon, 8 Sep 2008 03:02:17 +0000 (12:02 +0900)]
sh: Mark kretprobe_trampoline_holder static and __used.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add EDOSK7760 mach type.
Paul Mundt [Mon, 8 Sep 2008 03:01:55 +0000 (12:01 +0900)]
sh: Add EDOSK7760 mach type.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add Renesas EDOSK7760 board support.
Luca Santini [Mon, 8 Sep 2008 03:01:15 +0000 (12:01 +0900)]
sh: Add Renesas EDOSK7760 board support.

This adds support for the Renesas (RTE) EDOSK7760 board. Currently
supported devices are:

 - ramdisk support
 - ethernet support
 - nfs support
 - ext2/ext3 support
 - i2c support
 - fb support (M)

Signed-off-by: Luca Santini <luca.santini@spesonline.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Enable IRLM mode for SH7760 IRQ_MODE_IRQ.
Luca Santini [Mon, 8 Sep 2008 02:54:56 +0000 (11:54 +0900)]
sh: Enable IRLM mode for SH7760 IRQ_MODE_IRQ.

Follows the same setting as SH7750.

Signed-off-by: Luca Santini <luca.santini@spesonline.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: uninline flush_icache_all().
Paul Mundt [Fri, 5 Sep 2008 09:00:29 +0000 (18:00 +0900)]
sh: uninline flush_icache_all().

This uses jump_to_uncached() which is now given the noinline attribute
due to the special section mapping. Kill off the inline attribute to
fix up compilation failure.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add the rest of the boot targets to arch/sh/boot/.gitignore.
Paul Mundt [Fri, 5 Sep 2008 08:27:37 +0000 (17:27 +0900)]
sh: Add the rest of the boot targets to arch/sh/boot/.gitignore.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add kprobes support.
Chris Smith [Fri, 5 Sep 2008 08:15:39 +0000 (17:15 +0900)]
sh: Add kprobes support.

Initial support for kprobes/kretprobes for 32-bit SH platforms.

[ General cleanup and some rework for the kretprobe hash lock. -- PFM ]

Signed-off-by: Chris Smith <chris.smith@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: remove unnecessary memset after alloc_bootmem_low_pages
Marek Skuczynski [Fri, 5 Sep 2008 07:42:58 +0000 (16:42 +0900)]
sh: remove unnecessary memset after alloc_bootmem_low_pages

Because alloc_bootmem functions return the allocated memory always
zeroed, an additional call of memset on allocated memory is
unnecessary.

Signed-off-by: Marek Skuczynski <M.Skuczynski@adbglobal.com>
Signed-off-by: Carl Shaw <carl.shaw@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: fix the TMU code to allow a fully running NO_HZ system
Francesco Virlinzi [Fri, 5 Sep 2008 07:40:22 +0000 (16:40 +0900)]
sh: fix the TMU code to allow a fully running NO_HZ system

This patch fixes the TMU code to allow NO_HZ to work on sh

Signed-off-by: Francesco Virlinzi <francesco.virlinzi@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Subnormal double to float conversion
Carl Shaw [Fri, 5 Sep 2008 07:36:19 +0000 (16:36 +0900)]
sh: Subnormal double to float conversion

This patch adds support for the SH4 to convert a subnormal double
into a float by catching the FPE and implementing the FCNVDS
instruction in software.

Signed-off-by: Carl Shaw <carl.shaw@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Fix an unusual memory initialisation error.
Stuart Menefy [Fri, 5 Sep 2008 07:29:40 +0000 (16:29 +0900)]
sh: Fix an unusual memory initialisation error.

This fixes a problems with the set up of Linux memory:

 - When reserving memory at boot time, the code previously reserved
   the bottom page of memory, and then from one page up to the end of
   the bootmap. This had the desired effect, but was strictly speaking
   wrong, as the one page was actually whatever CONFIG_ZERO_PAGE_OFFSET
   had been set to.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Fix uImage load address in 32-bit mode.
Chris Smith [Fri, 5 Sep 2008 07:24:13 +0000 (16:24 +0900)]
sh: Fix uImage load address in 32-bit mode.

Fix "make uImage" load and entry addresses in 32-bit mode.

Signed-off-by: Chris Smith <chris.smith@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: vmalloc pgtable sync fix.
Stuart Menefy [Fri, 5 Sep 2008 07:17:15 +0000 (16:17 +0900)]
sh: vmalloc pgtable sync fix.

This fixes a problem in the code which copies the vmalloc portion of the
kernel's page table into the current user space page table. The addition
of the four level page table code breaks on folded page tables, because
the pud level is always present (although folded). This updates the code
to use the same style of updates for the pud as is used for the pgd
level.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Fix up broken 32-bit initrd support.
Stuart Menefy [Fri, 5 Sep 2008 07:14:17 +0000 (16:14 +0900)]
sh: Fix up broken 32-bit initrd support.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: early cached_to_uncached initialization.
Stuart Menefy [Fri, 5 Sep 2008 07:06:42 +0000 (16:06 +0900)]
sh: early cached_to_uncached initialization.

statically initialise the cached_to_uncached offset, so that we can use
it immediatly.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Add support for memory hot-remove.
Paul Mundt [Fri, 5 Sep 2008 06:39:12 +0000 (15:39 +0900)]
sh: Add support for memory hot-remove.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Don't enable clockevents broadcasting on UP SH-X3 builds.
Paul Mundt [Fri, 5 Sep 2008 06:36:39 +0000 (15:36 +0900)]
sh: Don't enable clockevents broadcasting on UP SH-X3 builds.

Fixes up compile errors with missing timer definitions. It's pointless to
have this enabled anyways if CONFIG_SMP=n.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: fixup many sparse errors.
Paul Mundt [Thu, 4 Sep 2008 09:53:58 +0000 (18:53 +0900)]
sh: fixup many sparse errors.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Display CPU information in show_regs().
Paul Mundt [Thu, 7 Aug 2008 16:23:34 +0000 (01:23 +0900)]
sh: Display CPU information in show_regs().

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Early dummy clockevent registration on boot CPU.
Paul Mundt [Thu, 7 Aug 2008 16:18:48 +0000 (01:18 +0900)]
sh: Early dummy clockevent registration on boot CPU.

The dummy timer needs to be registered on the boot CPU before the
system timer clockevent is registered, or broadcasting doesn't work
as advertized.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: smp: shove a cpu_relax() in the plat_start_cpu() busy loop.
Paul Mundt [Wed, 6 Aug 2008 09:39:32 +0000 (18:39 +0900)]
sh: smp: shove a cpu_relax() in the plat_start_cpu() busy loop.

Without this, certain versions of GCC will happily optimize the entire
loop out.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: generic clockevent broadcast support.
Paul Mundt [Wed, 6 Aug 2008 09:37:07 +0000 (18:37 +0900)]
sh: generic clockevent broadcast support.

This hooks up GENERIC_CLOCKEVENTS_BROADCAST and a dummy local timer,
which we call in to from the timer IPI when no other local timer is
provided.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: smp: Hook up a timer IPI stub.
Paul Mundt [Wed, 6 Aug 2008 09:21:03 +0000 (18:21 +0900)]
sh: smp: Hook up a timer IPI stub.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: smp: Hook in to the generic IPI handler for SH-X3 SMP.
Paul Mundt [Wed, 6 Aug 2008 09:05:09 +0000 (18:05 +0900)]
sh: smp: Hook in to the generic IPI handler for SH-X3 SMP.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: smp: Provide a generic IPI handler.
Paul Mundt [Wed, 6 Aug 2008 09:02:48 +0000 (18:02 +0900)]
sh: smp: Provide a generic IPI handler.

This provides a generic smp_message_recv() routine (based on the PPC
one), that IPI IRQs can wrap in to.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Kill off unused defines from asm/smp.h.
Paul Mundt [Wed, 6 Aug 2008 08:29:14 +0000 (17:29 +0900)]
sh: Kill off unused defines from asm/smp.h.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Provide movli.l/movco.l-based cmpxchg.
Paul Mundt [Thu, 7 Aug 2008 09:01:43 +0000 (18:01 +0900)]
sh: Provide movli.l/movco.l-based cmpxchg.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: Provide movli.l/movco.l-based bitops.
Paul Mundt [Thu, 7 Aug 2008 08:36:12 +0000 (17:36 +0900)]
sh: Provide movli.l/movco.l-based bitops.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 7 Sep 2008 04:47:30 +0000 (21:47 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: i8042 - make Lenovo 3000 N100 blacklist entry more specific
  Input: bcm5974 - add BTN_TOUCH event for mousedev benefit
  Input: bcm5974 - improve finger tracking and counting
  Input: bcm5974 - small formatting cleanup
  Input: bcm5974 - add maintainer entry

15 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Sep 2008 02:36:23 +0000 (19:36 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: cpu_init(): fix memory leak when using CPU hotplug
  x86: pda_init(): fix memory leak when using CPU hotplug
  x86, xen: Use native_pte_flags instead of native_pte_val for .pte_flags
  x86: move mtrr cpu cap setting early in early_init_xxxx
  x86: delay early cpu initialization until cpuid is done
  x86: use X86_FEATURE_NOPL in alternatives
  x86: add NOPL as a synthetic CPU feature bit
  x86: boot: stub out unimplemented CPU feature words

15 years agoMerge branch 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Sep 2008 02:33:26 +0000 (19:33 -0700)]
Merge branch 'timers-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  clocksource, acpi_pm.c: check for monotonicity
  clocksource, acpi_pm.c: use proper read function also in errata mode
  ntp: fix calculation of the next jiffie to trigger RTC sync
  x86: HPET: read back compare register before reading counter
  x86: HPET fix moronic 32/64bit thinko
  clockevents: broadcast fixup possible waiters
  HPET: make minimum reprogramming delta useful
  clockevents: prevent endless loop lockup
  clockevents: prevent multiple init/shutdown
  clockevents: enforce reprogram in oneshot setup
  clockevents: prevent endless loop in periodic broadcast handler
  clockevents: prevent clockevent event_handler ending up handler_noop

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Sun, 7 Sep 2008 02:33:10 +0000 (19:33 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  Fix CONFIG_AC97_BUS dependency

15 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Sun, 7 Sep 2008 02:32:21 +0000 (19:32 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] Probe initrd header only if explicitly specified
  [MIPS] TX39xx: Add missing local_flush_icache_range initialization
  [MIPS] TXx9: Fix txx9_pcode initialization
  [MIPS] Fix WARNING: at kernel/smp.c:290
  [MIPS] Fix data bus error recovery

15 years agox86: cpu_init(): fix memory leak when using CPU hotplug
Andreas Herrmann [Wed, 6 Aug 2008 08:29:37 +0000 (10:29 +0200)]
x86: cpu_init(): fix memory leak when using CPU hotplug

Exception stacks are allocated each time a CPU is set online.
But the allocated space is never freed. Thus with one CPU hotplug
offline/online cycle there is a memory leak of 24K (6 pages) for
a CPU.

Fix is to allocate exception stacks only once -- when the CPU is
set online for the first time.

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: akpm@linux-foundation.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: pda_init(): fix memory leak when using CPU hotplug
Andreas Herrmann [Wed, 6 Aug 2008 08:27:30 +0000 (10:27 +0200)]
x86: pda_init(): fix memory leak when using CPU hotplug

pda->irqstackptr is allocated whenever a CPU is set online.
But it is never freed. This results in a memory leak of 16K
for each CPU offline/online cycle.

Fix is to allocate pda->irqstackptr only once.

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: akpm@linux-foundation.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86, xen: Use native_pte_flags instead of native_pte_val for .pte_flags
Eduardo Habkost [Thu, 24 Jul 2008 15:15:45 +0000 (12:15 -0300)]
x86, xen: Use native_pte_flags instead of native_pte_val for .pte_flags

Using native_pte_val triggers the BUG_ON() in the paravirt_ops
version of pte_flags().

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move mtrr cpu cap setting early in early_init_xxxx
Yinghai Lu [Thu, 4 Sep 2008 19:09:43 +0000 (21:09 +0200)]
x86: move mtrr cpu cap setting early in early_init_xxxx

Krzysztof Helt found MTRR is not detected on k6-2

root cause:
we moved mtrr_bp_init() early for mtrr trimming,
and in early_detect we only read the CPU capability from cpuid,
so some cpu doesn't have that bit in cpuid.

So we need to add early_init_xxxx to preset those bit before mtrr_bp_init
for those earlier cpus.

this patch is for v2.6.27

Reported-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: delay early cpu initialization until cpuid is done
Krzysztof Helt [Thu, 4 Sep 2008 19:09:43 +0000 (21:09 +0200)]
x86: delay early cpu initialization until cpuid is done

Move early cpu initialization after cpu early get cap so the
early cpu initialization can fix up cpu caps.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclocksource, acpi_pm.c: check for monotonicity
Dominik Brodowski [Fri, 5 Sep 2008 21:05:35 +0000 (14:05 -0700)]
clocksource, acpi_pm.c: check for monotonicity

The current check for monotonicity is way too weak: Andreas Mohr reports (
http://lkml.org/lkml/2008/8/10/77 ) that on one of his test systems the
current check only triggers in 50% of all cases, leading to catastrophic
timer behaviour.  To fix this issue, expand the check for monotonicity by
doing ten consecutive tests instead of one.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclocksource, acpi_pm.c: use proper read function also in errata mode
Dominik Brodowski [Fri, 5 Sep 2008 21:05:33 +0000 (14:05 -0700)]
clocksource, acpi_pm.c: use proper read function also in errata mode

On all hardware (some Intel ICH4, PIIX4 and PIIX4E chipsets) affected by a
hardware errata there's about a 4.2% chance that initialization of the
ACPI PMTMR fails.  On those chipsets, we need to read out the timer value
at least three times to get a correct result, for every once in a while
(i.e.  within a 3 ns window every 69.8 ns) the read returns a bogus
result.  During normal operation we work around this issue, but during
initialization reading a bogus value may lead to -EINVAL even though the
hardware is usable.

Thanks to Andreas Mohr for spotting this issue.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agontp: fix calculation of the next jiffie to trigger RTC sync
Maciej W. Rozycki [Fri, 5 Sep 2008 21:05:31 +0000 (14:05 -0700)]
ntp: fix calculation of the next jiffie to trigger RTC sync

We have a bug in the calculation of the next jiffie to trigger the RTC
synchronisation.  The aim here is to run sync_cmos_clock() as close as
possible to the middle of a second.  Which means we want this function to
be called less than or equal to half a jiffie away from when now.tv_nsec
equals 5e8 (500000000).

If this is not the case for a given call to the function, for this purpose
instead of updating the RTC we calculate the offset in nanoseconds to the
next point in time where now.tv_nsec will be equal 5e8.  The calculated
offset is then converted to jiffies as these are the unit used by the
timer.

Hovewer timespec_to_jiffies() used here uses a ceil()-type rounding mode,
where the resulting value is rounded up.  As a result the range of
now.tv_nsec when the timer will trigger is from 5e8 to 5e8 + TICK_NSEC
rather than the desired 5e8 - TICK_NSEC / 2 to 5e8 + TICK_NSEC / 2.

As a result if for example sync_cmos_clock() happens to be called at the
time when now.tv_nsec is between 5e8 + TICK_NSEC / 2 and 5e8 to 5e8 +
TICK_NSEC, it will simply be rescheduled HZ jiffies later, falling in the
same range of now.tv_nsec again.  Similarly for cases offsetted by an
integer multiple of TICK_NSEC.

This change addresses the problem by subtracting TICK_NSEC / 2 from the
nanosecond offset to the next point in time where now.tv_nsec will be
equal 5e8, effectively shifting the following rounding in
timespec_to_jiffies() so that it produces a rounded-to-nearest result.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoFix CONFIG_AC97_BUS dependency
Takashi Iwai [Sat, 6 Sep 2008 09:43:41 +0000 (11:43 +0200)]
Fix CONFIG_AC97_BUS dependency

CONFIG_AC97_BUS is used from both sound and ucb1400 drivers.
The recent change in Kconfig introduced the exclusive dependency on
CONFIG_SOUND, and disabled the ucb1400 build without sound.
This patch makes CONFIG_AC97_BUS independent.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Tested-by: Randy Dunlap <randy.dunlap@oracle.com>
15 years agox86: HPET: read back compare register before reading counter
Thomas Gleixner [Sat, 6 Sep 2008 01:06:08 +0000 (03:06 +0200)]
x86: HPET: read back compare register before reading counter

After fixing the u32 thinko I sill had occasional hickups on ATI chipsets
with small deltas. There seems to be a delay between writing the compare
register and the transffer to the internal register which triggers the
interrupt. Reading back the value makes sure, that it hit the internal
match register befor we compare against the counter value.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
15 years agox86: HPET fix moronic 32/64bit thinko
Thomas Gleixner [Sat, 6 Sep 2008 01:03:32 +0000 (03:03 +0200)]
x86: HPET fix moronic 32/64bit thinko

We use the HPET only in 32bit mode because:
1) some HPETs are 32bit only
2) on i386 there is no way to read/write the HPET atomic 64bit wide

The HPET code unification done by the "moron of the year" did
not take into account that unsigned long is different on 32 and
64 bit.

This thinko results in a possible endless loop in the clockevents
code, when the return comparison fails due to the 64bit/332bit
unawareness.

unsigned long cnt = (u32) hpet_read() + delta can wrap over 32bit.
but the final compare will fail and return -ETIME causing endless
loops.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
15 years agoclockevents: broadcast fixup possible waiters
Thomas Gleixner [Sat, 6 Sep 2008 01:01:45 +0000 (03:01 +0200)]
clockevents: broadcast fixup possible waiters

Until the C1E patches arrived there where no users of periodic broadcast
before switching to oneshot mode. Now we need to trigger a possible
waiter for a periodic broadcast when switching to oneshot mode.
Otherwise we can starve them for ever.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
15 years agox86: use X86_FEATURE_NOPL in alternatives
H. Peter Anvin [Tue, 19 Aug 2008 00:50:33 +0000 (17:50 -0700)]
x86: use X86_FEATURE_NOPL in alternatives

Use X86_FEATURE_NOPL to determine if it is safe to use P6 NOPs in
alternatives.  Also, replace table and loop with simple if statement.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
15 years agox86: add NOPL as a synthetic CPU feature bit
H. Peter Anvin [Tue, 19 Aug 2008 00:39:32 +0000 (17:39 -0700)]
x86: add NOPL as a synthetic CPU feature bit

The long noops ("NOPL") are supposed to be detected by family >= 6.
Unfortunately, several non-Intel x86 implementations, both hardware
and software, don't obey this dictum.  Instead, probe for NOPL
directly by executing a NOPL instruction and see if we get #UD.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
15 years agox86: boot: stub out unimplemented CPU feature words
H. Peter Anvin [Fri, 15 Aug 2008 22:36:31 +0000 (15:36 -0700)]
x86: boot: stub out unimplemented CPU feature words

The CPU feature detection code in the boot code is somewhat minimal,
and doesn't include all possible CPUID words.  In particular, it
doesn't contain the code for CPU feature words 2 (Transmeta),
3 (Linux-specific), 5 (VIA), or 7 (scattered).  Zero them out, so we
can still set those bits as known at compile time; in particular, this
allows creating a Linux-specific NOPL flag and have it required (and
therefore resolvable at compile time) in 64-bit mode.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
15 years agodrivers/mmc/card/block.c: fix refcount leak in mmc_block_open()
Andrew Morton [Fri, 5 Sep 2008 21:00:24 +0000 (14:00 -0700)]
drivers/mmc/card/block.c: fix refcount leak in mmc_block_open()

mmc_block_open() increments md->usage although it returns with -EROFS when
default mounting a MMC/SD card with write protect switch on.  This
reference counting bug prevents /dev/mmcblkX from being released on card
removal, and situation worsen with reinsertion until the minor number
range runs out.

Reported-by: <sasin@solomon-systech.com>
Acked-by: Pierre Ossman <drzeus-list@drzeus.cx>
Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotracehook: comment pasto fixes
Roland McGrath [Fri, 5 Sep 2008 21:00:23 +0000 (14:00 -0700)]
tracehook: comment pasto fixes

Fix some pasto's in comments in the new linux/tracehook.h and
asm-generic/syscall.h files.

Reported-by: Wenji Huang <wenji.huang@oracle.com>
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoatmel_lcdfb: fix oops in rmmod when framebuffer fails to register
Stanislaw Gruszka [Fri, 5 Sep 2008 21:00:22 +0000 (14:00 -0700)]
atmel_lcdfb: fix oops in rmmod when framebuffer fails to register

If framebuffer registration failed in platform driver ->probe() callback,
dev_get_drvdata() points to freed memory region, but ->remove() function
try to use it and the following oops occurs:

Unable to handle kernel NULL pointer dereference at virtual address 00000228
pgd = c3a20000
[00000228] *pgd=23a2b031, *pte=00000000, *ppte=00000000
Internal error: Oops: 17 [#1]
Modules linked in: atmel_lcdfb(-) cfbcopyarea cfbimgblt cfbfillrect [last unloaded: atmel_lcdfb]
CPU: 0    Not tainted  (2.6.27-rc2 #116)
PC is at atmel_lcdfb_remove+0x14/0xf8 [atmel_lcdfb]
LR is at platform_drv_remove+0x20/0x24
pc : [<bf006bc4>]    lr : [<c0157d28>]    psr: a0000013
sp : c3a45e84  ip : c3a45ea0  fp : c3a45e9c
r10: 00000002  r9 : c3a44000  r8 : c0026c04
r7 : 00000880  r6 : c02bb228  r5 : 00000000  r4 : c02bb230
r3 : bf007e3c  r2 : c02bb230  r1 : 00000004  r0 : c02bb228
Flags: NzCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
Control: 0005317f  Table: 23a20000  DAC: 00000015
Process rmmod (pid: 6799, stack limit = 0xc3a44260)
Stack: (0xc3a45e84 to 0xc3a46000)
5e80:          c02bb230 bf007e3c bf007e3c c3a45eac c3a45ea0 c0157d28 bf006bc0
5ea0: c3a45ec4 c3a45eb0 c0156d20 c0157d18 c02bb230 c02bb2d8 c3a45ee0 c3a45ec8
5ec0: c0156da8 c0156cb8 bf007e3c bf007ee0 c02c8e14 c3a45efc c3a45ee4 c0156018
5ee0: c0156d50 bf007e3c bf007ee0 00000000 c3a45f18 c3a45f00 c0157220 c0155f9c
5f00: 00000000 bf007ee0 bf008000 c3a45f28 c3a45f1c c0157e34 c01571ec c3a45f38
5f20: c3a45f2c bf006ba8 c0157e30 c3a45fa4 c3a45f3c c005772c bf006ba4 656d7461
5f40: 636c5f6c 00626664 c004c988 c3a45f80 c3a45f5c 00000000 c3a45fb0 00000000
5f60: ffffffff becaccd8 00000880 00000000 000a5e80 00000001 bf007ee0 00000880
5f80: c3a45f84 00000000 becaccd4 00000002 000003df 00000081 00000000 c3a45fa8
5fa0: c0026a60 c0057584 00000002 000003df 00900081 000a5e80 00000880 00000000
5fc0: becaccd4 00000002 000003df 00000000 000a5e80 00000001 00000002 0000005f
5fe0: 4004f5ec becacbe8 0001a158 4004f5fc 20000010 00900081 f9ffbadf 7bbfb2bb
Backtrace:
[<bf006bb0>] (atmel_lcdfb_remove+0x0/0xf8 [atmel_lcdfb]) from [<c0157d28>] (platform_drv_remove+0x20/0x24)
 r6:bf007e3c r5:bf007e3c r4:c02bb230
[<c0157d08>] (platform_drv_remove+0x0/0x24) from [<c0156d20>] (__device_release_driver+0x78/0x98)
[<c0156ca8>] (__device_release_driver+0x0/0x98) from [<c0156da8>] (driver_detach+0x68/0x90)
 r5:c02bb2d8 r4:c02bb230
[<c0156d40>] (driver_detach+0x0/0x90) from [<c0156018>] (bus_remove_driver+0x8c/0xb4)
 r6:c02c8e14 r5:bf007ee0 r4:bf007e3c
[<c0155f8c>] (bus_remove_driver+0x0/0xb4) from [<c0157220>] (driver_unregister+0x44/0x48)
 r6:00000000 r5:bf007ee0 r4:bf007e3c
[<c01571dc>] (driver_unregister+0x0/0x48) from [<c0157e34>] (platform_driver_unregister+0x14/0x18)
 r6:bf008000 r5:bf007ee0 r4:00000000
[<c0157e20>] (platform_driver_unregister+0x0/0x18) from [<bf006ba8>] (atmel_lcdfb_exit+0x14/0x1c [atmel_lcdfb])
[<bf006b94>] (atmel_lcdfb_exit+0x0/0x1c [atmel_lcdfb]) from [<c005772c>] (sys_delete_module+0x1b8/0x22c)
[<c0057574>] (sys_delete_module+0x0/0x22c) from [<c0026a60>] (ret_fast_syscall+0x0/0x2c)
 r7:00000081 r6:000003df r5:00000002 r4:becaccd4
Code: e92dd870 e24cb004 e59050c4 e1a06000 (e5954228)
---[ end trace 85476b184d9e68d8 ]---

This patch fixes the oops.

Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoforcedeth: fix kexec regression
Rafael J. Wysocki [Fri, 5 Sep 2008 21:00:19 +0000 (14:00 -0700)]
forcedeth: fix kexec regression

Fix regression tracked as http://bugzilla.kernel.org/show_bug.cgi?id=11361
and caused by commit f735a2a1a4f2a0f5cd823ce323e82675990469e2 ("[netdrvr]
forcedeth: setup wake-on-lan before shutting down") that makes network
adapters integrated into the NVidia MCP55 chipsets fail to work in kexeced
kernels.  The problem appears to be that if the adapter is put into D3_hot
during ->shutdown(), it cannot be brought back into D0 after kexec (ref.
http://marc.info/?l=linux-kernel&m=121900062814967&w=4).  Therefore, only
put forcedeth into D3 during ->shutdown() if the system is to be powered
off.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Tested-by: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Ayaz Abdulla <aabdulla@nvidia.com>
Acked-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agores_counter: fix off-by-one bug in setting limit
Li Zefan [Fri, 5 Sep 2008 21:00:18 +0000 (14:00 -0700)]
res_counter: fix off-by-one bug in setting limit

I found we can no longer set limit to 0 with 2.6.27-rcX:
 # mount -t cgroup -omemory xxx /mnt
 # mkdir /mnt/0
 # echo 0 > /mnt/0/memory.limit_in_bytes
 bash: echo: write error: Device or resource busy

It turned out 'limit' can't be set to 'usage', which is wrong IMO.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 5 Sep 2008 21:37:15 +0000 (14:37 -0700)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: fix process time monotonicity
  sched_clock: fix NOHZ interaction

15 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 5 Sep 2008 21:36:21 +0000 (14:36 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: add io delay quirk for Presario F700

15 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx
Linus Torvalds [Fri, 5 Sep 2008 21:35:56 +0000 (14:35 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/djbw/async_tx

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
  async_tx: fix the bug in async_tx_run_dependencies

15 years agoMerge git://git.infradead.org/~dwmw2/dwmw2-2.6.27
Linus Torvalds [Fri, 5 Sep 2008 21:31:54 +0000 (14:31 -0700)]
Merge git://git.infradead.org/~dwmw2/dwmw2-2.6.27

* git://git.infradead.org/~dwmw2/dwmw2-2.6.27:
  Revert "[ARM] use the new byteorder headers"
  Fix conditional export of kvh.h and a.out.h to userspace.
  [MTD] [NAND] tmio_nand: fix base address programming

15 years agoMerge branch 'sh/for-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/lethal...
Linus Torvalds [Fri, 5 Sep 2008 21:30:58 +0000 (14:30 -0700)]
Merge branch 'sh/for-2.6.27' of git://git./linux/kernel/git/lethal/sh-2.6

* 'sh/for-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6:
  i2c: fix i2c-sh_mobile timing issues
  sh64: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.
  sh: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.
  sh: fix semtimedop syscall
  sh: update AP325RXA defconfig
  sh: update Migo-R defconfig
  sh: fix platform_resource_setup_memory() section mismatch
  sh: fix kexec entry point for crash kernels
  sh: crash kernel resource fix
  sh: fix ptrace_64.c:user_disable_single_step()
  sh64: re-add the __strnlen_user() prototype

15 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb
Linus Torvalds [Fri, 5 Sep 2008 21:29:50 +0000 (14:29 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/mchehab/v4l-dvb

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/v4l-dvb: (98 commits)
  V4L/DVB (8881): gspca: After 'while (retry--) {...}', retry will be -1 but not 0.
  V4L/DVB (8880): PATCH: Fix parents on some webcam drivers
  V4L/DVB (8877): b2c2 and bt8xx: udelay to mdelay
  V4L/DVB (8876): budget: udelay changed to mdelay
  V4L/DVB (8874): gspca: Adjust hstart for sn9c103/ov7630 and update usb-id's.
  V4L/DVB (8873): gspca: Bad image offset with rev012a of spca561 and adjust exposure.
  V4L/DVB (8872): gspca: Bad image format and offset with rev072a of spca561.
  V4L/DVB (8870): gspca: Fix dark room problem with sonixb.
  V4L/DVB (8869): gspca: Move the Sonix webcams with TAS5110C1B from sn9c102 to gspca.
  V4L/DVB (8868): gspca: Support for vga modes with sif sensors in sonixb.
  V4L/DVB (8844): dabusb_fpga_download(): fix a memory leak
  V4L/DVB (8843): tda10048_firmware_upload(): fix a memory leak
  V4L/DVB (8842): vivi_release(): fix use-after-free
  V4L/DVB (8840): dib0700: add basic support for Hauppauge Nova-TD-500 (84xxx)
  V4L/DVB (8839): dib0700: add comment to identify 35th USB id pair
  V4L/DVB (8837): dvb: fix I2C adapters name size
  V4L/DVB (8835): gspca: Same pixfmt as the sn9c102 driver and raw Bayer added in sonixb.
  V4L/DVB (8834): gspca: Have a bigger buffer for sn9c10x compressed images.
  V4L/DVB (8833): gspca: Cleanup the sonixb code.
  V4L/DVB (8832): gspca: Bad pixelformat of vc0321 webcams.
  ...

15 years agoMerge branch 'core/debugobjects' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 5 Sep 2008 21:28:19 +0000 (14:28 -0700)]
Merge branch 'core/debugobjects' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core/debugobjects' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  debugobjects: fix lockdep warning

15 years agoMerge branch 'release-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/ak...
Linus Torvalds [Fri, 5 Sep 2008 21:27:12 +0000 (14:27 -0700)]
Merge branch 'release-2.6.27' of git://git./linux/kernel/git/ak/linux-acpi-2.6

* 'release-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-acpi-2.6:
  acer-wmi: remove debugfs entries upon unloading
  ACPI: Avoid bogus timeout about SMbus check
  fujitsu-laptop: fix regression for P8010 in 2.6.27-rc
  ACPI: Make Len Brown the ACPI maintainer again
  ACPI: thinkpad-acpi: wan radio control is not experimental
  PNPACPI: ignore the producer/consumer bit for extended IRQ descriptors
  acpi: add checking for NULL early param
  ACPI: Fix typo in "Disable MWAIT via DMI on broken Compal board"
  ACPI: Fix now signed module parameter.
  ACPI: Change package length error to warning
  ACPI: Fix now signed module parameter.

15 years ago[MIPS] Probe initrd header only if explicitly specified
Atsushi Nemoto [Tue, 26 Aug 2008 13:34:57 +0000 (22:34 +0900)]
[MIPS] Probe initrd header only if explicitly specified

Currently init_initrd() probes initrd header at the last page of kernel
image, but it is valid only if addinitrd was used.  If addinitrd was not
used, the area contains garbage so probing there might misdetect initrd
header (magic number is not strictly robust).

This patch introduces CONFIG_PROBE_INITRD_HEADER to explicitly enable this
probing.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years ago[MIPS] TX39xx: Add missing local_flush_icache_range initialization
Atsushi Nemoto [Tue, 26 Aug 2008 13:30:41 +0000 (22:30 +0900)]
[MIPS] TX39xx: Add missing local_flush_icache_range initialization

Commmit 59e39ecd933ba49eb6efe84cbfa5597a6c9ef18a ("Fix WARNING: at
kernel/smp.c:290") introduced local_flush_icache_range but lacks
initialization for some TX39 case.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years ago[MIPS] TXx9: Fix txx9_pcode initialization
Atsushi Nemoto [Tue, 26 Aug 2008 12:29:58 +0000 (21:29 +0900)]
[MIPS] TXx9: Fix txx9_pcode initialization

The txx9_pcode variable was introduced in commit
fe1c2bc64f65003b39f331a8e4b0d15b235a4afd ("TXx9: Add 64-bit support")
but was not initialized properly.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years ago[MIPS] Fix WARNING: at kernel/smp.c:290
Thomas Bogendoerfer [Mon, 4 Aug 2008 18:53:57 +0000 (20:53 +0200)]
[MIPS] Fix WARNING: at kernel/smp.c:290

trap_init issues flush_icache_range(), which uses ipi functions to
get icache flushing done on all cpus. But this is done before interrupts
are enabled and caused WARN_ON messages. This changeset introduces
a new local_flush_icache_range() and uses it before interrupts (and
additional CPUs) are enabled to avoid this problem.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years ago[MIPS] Fix data bus error recovery
Thomas Bogendoerfer [Mon, 4 Aug 2008 17:44:34 +0000 (19:44 +0200)]
[MIPS] Fix data bus error recovery

With -ffunction-section the entries in __dbe_table aren't no longer
sorted, so the lookup of exception addresses in do_be() failed for
some addresses. To avoid this we now sort __dbe_table.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoRevert "mac80211: Use IWEVASSOCREQIE instead of IWEVCUSTOM"
Linus Torvalds [Fri, 5 Sep 2008 19:38:09 +0000 (12:38 -0700)]
Revert "mac80211: Use IWEVASSOCREQIE instead of IWEVCUSTOM"

This reverts commit 087d833e5a9f67ba933cb32eaf5a2279c1a5b47c, which was
reported to break wireless at least in some combinations with 32bit user
space and a 64bit kernel.  Alex Williamnson bisected it to this commit.

Reported-and-bisected-by: Alex Williamson <alex.williamson@hp.com>
Acked-by: John W. Linville <linville@tuxdriver.com>
Cc: David Miller <davem@davemloft.net>
Cc: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosched: fix process time monotonicity
Balbir Singh [Fri, 5 Sep 2008 16:12:23 +0000 (18:12 +0200)]
sched: fix process time monotonicity

Spencer reported a problem where utime and stime were going negative despite
the fixes in commit b27f03d4bdc145a09fb7b0c0e004b29f1ee555fa. The suspected
reason for the problem is that signal_struct maintains it's own utime and
stime (of exited tasks), these are not updated using the new task_utime()
routine, hence sig->utime can go backwards and cause the same problem
to occur (sig->utime, adds tsk->utime and not task_utime()). This patch
fixes the problem

TODO: using max(task->prev_utime, derived utime) works for now, but a more
generic solution is to implement cputime_max() and use the cputime_gt()
function for comparison.

Reported-by: spencer@bluehost.com
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agosched_clock: fix NOHZ interaction
Peter Zijlstra [Mon, 1 Sep 2008 14:44:23 +0000 (16:44 +0200)]
sched_clock: fix NOHZ interaction

If HLT stops the TSC, we'll fail to account idle time, thereby inflating the
actual process times. Fix this by re-calibrating the clock against GTOD when
leaving nohz mode.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Tested-by: Avi Kivity <avi@qumranet.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoRevert "[ARM] use the new byteorder headers"
David Woodhouse [Fri, 5 Sep 2008 15:01:15 +0000 (16:01 +0100)]
Revert "[ARM] use the new byteorder headers"

This reverts commit ae82cbfc8beaa69007aa09966d3983ac938c3577. It
needs the new byteorder headers to be exported to userspace, and
they aren't yet -- and probably shouldn't be, at this point in the
2.6.27 release cycle (or ever, for that matter).

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
15 years agoasync_tx: fix the bug in async_tx_run_dependencies
Yuri Tikhonov [Fri, 5 Sep 2008 15:15:47 +0000 (08:15 -0700)]
async_tx: fix the bug in async_tx_run_dependencies

Should clear the next pointer of the TX if we are sure that the
next TX (say NXT) will be submitted to the channel too. Overwise,
we break the chain of descriptors, because we lose the information
about the next descriptor to run. So next time, when invoke
async_tx_run_dependencies() with TX, it's TX->next will be NULL, and
NXT will be never submitted.

Cc: <stable@kernel.org> [2.6.26]
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
15 years agoFix conditional export of kvh.h and a.out.h to userspace.
Khem Raj [Fri, 5 Sep 2008 06:11:01 +0000 (23:11 -0700)]
Fix conditional export of kvh.h and a.out.h to userspace.

Some architectures have moved the asm/ into arch/ and some have not.
This patch checks for a.out.h and kvh.h in both places before exporting
the corresponding file from linux/

[dwmw2: simplified a little]
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
15 years ago[MTD] [NAND] tmio_nand: fix base address programming
Dmitry Baryshkov [Thu, 4 Sep 2008 09:28:33 +0000 (13:28 +0400)]
[MTD] [NAND] tmio_nand: fix base address programming

Fix offset of second word used for programming base address of memory
window. Also program tmio with offset of the FCR, not with physical
memory location.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Cc: Ian Molton <spyro@f2s.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
15 years agoHPET: make minimum reprogramming delta useful
Thomas Gleixner [Wed, 3 Sep 2008 21:37:24 +0000 (21:37 +0000)]
HPET: make minimum reprogramming delta useful

The minimum reprogramming delta was hardcoded in HPET ticks,
which is stupid as it does not work with faster running HPETs.
The C1E idle patches made this prominent on AMD/RS690 chipsets,
where the HPET runs with 25MHz. Set it to 5us which seems to be
a reasonable value and fixes the problems on the bug reporters
machines. We have a further sanity check now in the clock events,
which increases the delta when it is not sufficient.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br>
Tested-by: Dmitry Nezhevenko <dion@inhex.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclockevents: prevent endless loop lockup
Thomas Gleixner [Wed, 3 Sep 2008 21:37:14 +0000 (21:37 +0000)]
clockevents: prevent endless loop lockup

The C1E/HPET bug reports on AMDX2/RS690 systems where tracked down to a
too small value of the HPET minumum delta for programming an event.

The clockevents code needs to enforce an interrupt event on the clock event
device in some cases. The enforcement code was stupid and naive, as it just
added the minimum delta to the current time and tried to reprogram the device.
When the minimum delta is too small, then this loops forever.

Add a sanity check. Allow reprogramming to fail 3 times, then print a warning
and double the minimum delta value to make sure, that this does not happen again.
Use the same function for both tick-oneshot and tick-broadcast code.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclockevents: prevent multiple init/shutdown
Thomas Gleixner [Wed, 3 Sep 2008 21:37:08 +0000 (21:37 +0000)]
clockevents: prevent multiple init/shutdown

While chasing the C1E/HPET bugreports I went through the clock events
code inch by inch and found that the broadcast device can be initialized
and shutdown multiple times. Multiple shutdowns are not critical, but
useless waste of time. Multiple initializations are simply broken. Another
CPU might have the device in use already after the first initialization and
the second init could just render it unusable again.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclockevents: enforce reprogram in oneshot setup
Thomas Gleixner [Wed, 3 Sep 2008 21:37:03 +0000 (21:37 +0000)]
clockevents: enforce reprogram in oneshot setup

In tick_oneshot_setup we program the device to the given next_event,
but we do not check the return value. We need to make sure that the
device is programmed enforced so the interrupt handler engine starts
working. Split out the reprogramming function from tick_program_event()
and call it with the device, which was handed in to tick_setup_oneshot().
Set the force argument, so the devices is firing an interrupt.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclockevents: prevent endless loop in periodic broadcast handler
Thomas Gleixner [Wed, 3 Sep 2008 21:36:57 +0000 (21:36 +0000)]
clockevents: prevent endless loop in periodic broadcast handler

The reprogramming of the periodic broadcast handler was broken,
when the first programming returned -ETIME. The clockevents code
stores the new expiry value in the clock events device next_event field
only when the programming time has not been elapsed yet. The loop in
question calculates the new expiry value from the next_event value
and therefor never increases.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoclockevents: prevent clockevent event_handler ending up handler_noop
Venkatesh Pallipadi [Wed, 3 Sep 2008 21:36:50 +0000 (21:36 +0000)]
clockevents: prevent clockevent event_handler ending up handler_noop

There is a ordering related problem with clockevents code, due to which
clockevents_register_device() called after tickless/highres switch
will not work. The new clockevent ends up with clockevents_handle_noop as
event handler, resulting in no timer activity.

The problematic path seems to be

* old device already has hrtimer_interrupt as the event_handler
* new clockevent device registers with a higher rating
* tick_check_new_device() is called
  * clockevents_exchange_device() gets called
    * old->event_handler is set to clockevents_handle_noop
  * tick_setup_device() is called for the new device
    * which sets new->event_handler using the old->event_handler which is noop.

Change the ordering so that new device inherits the proper handler.

This does not have any issue in normal case as most likely all the clockevent
devices are setup before the highres switch. But, can potentially be affecting
some corner case where HPET force detect happens after the highres switch.
This was a problem with HPET in MSI mode code that we have been experimenting
with.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoi2c: fix i2c-sh_mobile timing issues
Magnus Damm [Wed, 27 Aug 2008 09:33:56 +0000 (18:33 +0900)]
i2c: fix i2c-sh_mobile timing issues

This patch teaches the i2c-sh_mobile driver to make use of wait irqs.
Without this patch only dte irqs are used which may lead to overruns
and cases of missing stop and extra bytes being read on the i2c bus.

Use of wait irqs forces the hardware to pause and wait until the cpu
is ready. Polling is also reworked in this patch to fix ms delay issues.

Verified with bus analyzer and tested on MigoR and AP325RXA boards.

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh64: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.
Paul Mundt [Fri, 5 Sep 2008 05:51:28 +0000 (14:51 +0900)]
sh64: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.

Follows the SH change.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
15 years agosh: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.
Carmelo Amoroso [Fri, 5 Sep 2008 05:42:16 +0000 (14:42 +0900)]
sh: resume_kernel fix for kernel oops built with CONFIG_BKL_PREEMPT=y.

This patch fixes a problem within the SH implementation of resume_kernel code,
that implements in assembly the bulk of preempt_schedule_irq function without
taking care of the extra code needed to handle the BKL preemptible.

The patch basically consists of removing this asm code and calling the common
C implementation (see kernel/sched.c) as other archs do.

Another change is the missing 'cli' macro invocation at the beginning of
the resume_kernel.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>