GitHub/mt8127/android_kernel_alcatel_ttab.git
15 years agoagp: add user mapping support to ATI AGP bridge.
Dave Airlie [Fri, 19 Jun 2009 00:52:57 +0000 (10:52 +1000)]
agp: add user mapping support to ATI AGP bridge.

This should fix TTM/KMS on some of the original ATI IGP chipsets.
(rs100/rs200)

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/i915: enable GEM on PAE.
Dave Airlie [Fri, 19 Jun 2009 00:29:20 +0000 (10:29 +1000)]
drm/i915: enable GEM on PAE.

In theory now that the AGP subsystem is using struct page, we should
have on problems enabling GEM on PAE systems.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: fix unused variables warning
Dave Airlie [Fri, 19 Jun 2009 00:22:21 +0000 (10:22 +1000)]
drm/radeon: fix unused variables warning

just remove i variable left over from previous code.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoagp: switch AGP to use page array instead of unsigned long array
Dave Airlie [Fri, 12 Jun 2009 04:11:41 +0000 (14:11 +1000)]
agp: switch AGP to use page array instead of unsigned long array

This switches AGP to use an array of pages for tracking the
pages allocated to the GART. This should enable GEM on PAE to work
a lot better as we can pass highmem pages to the PAT code and it will
do the right thing with them.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoagpgart: detected ALi M???? chipset with M1621
Ondrej Zary [Wed, 10 Jun 2009 19:41:11 +0000 (12:41 -0700)]
agpgart: detected ALi M???? chipset with M1621

Add M1621 chipset name to ali-agp, preventing "Detected ALi M???? chipset"
message.

Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: command stream checker for r3xx-r5xx hardware
Jerome Glisse [Wed, 17 Jun 2009 11:28:30 +0000 (13:28 +0200)]
drm/radeon: command stream checker for r3xx-r5xx hardware

For security purpose we want to make sure the userspace process doesn't
access memory beyond buffer it owns. To achieve this we need to check
states the userspace program. For color buffer and zbuffer we check that
the clipping register will discard access beyond buffers set as color
or zbuffer. For vertex buffer we check that no vertex fetch will happen
beyond buffer end. For texture we check various texture states (number
of mipmap level, texture size, texture depth, ...) to compute the amount
of memory the texture fetcher might access.

The command stream checking impact the performances so far quick benchmark
shows an average of 3% decrease in fps of various applications. It can
be optimized a bit more by caching result of checking and thus avoid a
full recheck if no states changed since last check.

Note that this patch is still incomplete on checking side as it doesn't
check 2d rendering states.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: Fully initialize LVDS info also when we can't get it from the ROM.
Michel Dänzer [Wed, 17 Jun 2009 16:28:38 +0000 (18:28 +0200)]
drm/radeon: Fully initialize LVDS info also when we can't get it from the ROM.

This makes the panel on my PowerBook light up.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: Fix CP byte order on big endian architectures with KMS.
Michel Dänzer [Tue, 16 Jun 2009 15:29:06 +0000 (17:29 +0200)]
radeon: Fix CP byte order on big endian architectures with KMS.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoagp/uninorth: Handle user memory types.
Michel Dänzer [Mon, 15 Jun 2009 14:56:15 +0000 (16:56 +0200)]
agp/uninorth: Handle user memory types.

This adds support for TTM to the uninorth AGP bridge.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/ttm: Add some powerpc cache flush code.
Michel Dänzer [Mon, 15 Jun 2009 14:56:13 +0000 (16:56 +0200)]
drm/ttm: Add some powerpc cache flush code.

Optimise the powerpc flushing path for TTM.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: Enable modesetting on non-x86.
Michel Dänzer [Mon, 15 Jun 2009 14:56:09 +0000 (16:56 +0200)]
radeon: Enable modesetting on non-x86.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: Respect AGP cant_use_aperture flag.
Michel Dänzer [Mon, 15 Jun 2009 14:56:11 +0000 (16:56 +0200)]
drm/radeon: Respect AGP cant_use_aperture flag.

Some AGP devices can't map the aperture, radeon needs to tell TTM this.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: EDID endianness fixes.
Michel Dänzer [Mon, 15 Jun 2009 14:56:07 +0000 (16:56 +0200)]
drm: EDID endianness fixes.

Mostly replacing bitfields with explicit masks and shifts.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: this VRAM vs aperture test is wrong, just remove it.
Dave Airlie [Thu, 18 Jun 2009 08:08:33 +0000 (18:08 +1000)]
drm/radeon: this VRAM vs aperture test is wrong, just remove it.

Its quite valid to have VRAM < aperture size.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/ttm: fix an error path to exit function correctly
Thomas Hellstrom [Wed, 17 Jun 2009 10:29:57 +0000 (12:29 +0200)]
drm/ttm: fix an error path to exit function correctly

Just a goto instead of a direct exit.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: Apply "Memory fragmentation from lost alignment blocks"
Thomas Hellstrom [Wed, 17 Jun 2009 10:29:56 +0000 (12:29 +0200)]
drm: Apply "Memory fragmentation from lost alignment blocks"

also for the atomic path by using a common code-path.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agottm: Return -ERESTART when a signal interrupts bo eviction.
Thomas Hellstrom [Wed, 17 Jun 2009 10:29:55 +0000 (12:29 +0200)]
ttm: Return -ERESTART when a signal interrupts bo eviction.

A bug caused the ttm code to just terminate the wait when a signal
was received while waiting for the GPU to release a buffer object that
was to be evicted.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: Remove memory debugging infrastructure.
Eric Anholt [Tue, 24 Mar 2009 19:23:04 +0000 (12:23 -0700)]
drm: Remove memory debugging infrastructure.

It hasn't been used in ages, and having the user tell your how much
memory is being freed at free time is a recipe for disaster even if it
was ever used.

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Clear fence register on tiling stride change.
Chris Wilson [Sat, 6 Jun 2009 08:46:01 +0000 (09:46 +0100)]
drm/i915: Clear fence register on tiling stride change.

The fence register value also depends upon the stride of the object, so we
need to clear the fence if that is changed as well.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[anholt: Added 8xx and 965 paths, and renamed the confusing
i915_gem_object_tiling_ok function to i915_gem_object_fence_offset_ok]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Install fence register for tiled scanout on i915
Chris Wilson [Wed, 17 Jun 2009 21:08:52 +0000 (22:08 +0100)]
drm/i915: Install fence register for tiled scanout on i915

With the work by Jesse Barnes to eliminate allocation of fences during
execbuffer, it becomes possible to write to the scan-out buffer with it
never acquiring a fence (simply by only ever writing to the object using
tiled GPU commands and never writing to it via the GTT). So for pre-i965
chipsets which require fenced access for tiled scan-out buffers, we need
to obtain a fence register.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: detach/attach get/put pages symmetry
Chris Wilson [Wed, 17 Jun 2009 20:52:49 +0000 (21:52 +0100)]
drm/i915: detach/attach get/put pages symmetry

After performing an operation over the page list for a buffer retrieved by
i915_gem_object_get_pages() the pages need to be returned with
i915_gem_object_put_pages(). This was not being observed for the phys
objects which were thus leaking references to their backing pages.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
CC: Dave Airlie <airlied@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: A few debugfs formatting fixes
Ben Gamari [Wed, 10 Jun 2009 22:26:20 +0000 (18:26 -0400)]
drm/i915: A few debugfs formatting fixes

Signed-Off-By: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Warn when inteldrmfb fails to restore its framebuffer config
Ben Gamari [Thu, 11 Jun 2009 04:44:26 +0000 (00:44 -0400)]
drm/i915: Warn when inteldrmfb fails to restore its framebuffer config

While sifting through the inteldrmfb code trying to solve #22040 I found that
the fb restore path doesn't check the return value of
drm_crtc_helper_set_config(), which seems to have all sorts of potential
failure modes. We should warn someone if one of these is triggered.

Signed-Off-By: Ben Gamari <bgamari.foss@gmail.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[anholt: hand-applied, failures are mine]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agoMerge branch 'kmemleak' of git://linux-arm.org/linux-2.6
Linus Torvalds [Wed, 17 Jun 2009 17:42:21 +0000 (10:42 -0700)]
Merge branch 'kmemleak' of git://linux-arm.org/linux-2.6

* 'kmemleak' of git://linux-arm.org/linux-2.6:
  kmemleak: Fix some typos in comments
  kmemleak: Rename kmemleak_panic to kmemleak_stop
  kmemleak: Only use GFP_KERNEL|GFP_ATOMIC for the internal allocations

15 years agokmemleak: Fix some typos in comments
Catalin Marinas [Wed, 17 Jun 2009 17:29:04 +0000 (18:29 +0100)]
kmemleak: Fix some typos in comments

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
15 years agokmemleak: Rename kmemleak_panic to kmemleak_stop
Catalin Marinas [Wed, 17 Jun 2009 17:29:03 +0000 (18:29 +0100)]
kmemleak: Rename kmemleak_panic to kmemleak_stop

This is to avoid the confusion created by the "panic" word.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
15 years agokmemleak: Only use GFP_KERNEL|GFP_ATOMIC for the internal allocations
Catalin Marinas [Wed, 17 Jun 2009 17:29:02 +0000 (18:29 +0100)]
kmemleak: Only use GFP_KERNEL|GFP_ATOMIC for the internal allocations

Kmemleak allocates memory for pointer tracking and it tries to avoid
using GFP_ATOMIC if the caller doesn't require it. However other gfp
flags may be passed by the caller which aren't required by kmemleak.
This patch filters the gfp flags so that only GFP_KERNEL | GFP_ATOMIC
are used.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
15 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Wed, 17 Jun 2009 16:51:50 +0000 (09:51 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] cpumask: new cpumask operators for arch/x86/kernel/cpu/cpufreq/powernow-k8.c
  [CPUFREQ] cpumask: avoid playing with cpus_allowed in powernow-k8.c
  [CPUFREQ] cpumask: avoid cpumask games in arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
  [CPUFREQ] cpumask: avoid playing with cpus_allowed in speedstep-ich.c
  [CPUFREQ] powernow-k8: get drv data for correct CPU
  [CPUFREQ] powernow-k8: read P-state from HW
  [CPUFREQ] reduce scope of ACPI_PSS_BIOS_BUG_MSG[]
  [CPUFREQ] Clean up convoluted code in arch/x86/kernel/tsc.c:time_cpufreq_notifier()
  [CPUFREQ] minor correction to cpu-freq documentation
  [CPUFREQ] powernow-k8.c: mess cleanup
  [CPUFREQ] Only set sampling_rate_max deprecated, sampling_rate_min is useful
  [CPUFREQ] powernow-k8: Set transition latency to 1 if ACPI tables export 0
  [CPUFREQ] ondemand: Uncouple minimal sampling rate from HZ in NO_HZ case

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6
Linus Torvalds [Wed, 17 Jun 2009 16:50:44 +0000 (09:50 -0700)]
Merge git://git./linux/kernel/git/jejb/scsi-misc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6:
  [SCSI] aic79xx: make driver respect nvram for IU and QAS settings
  [SCSI] don't attach ULD to Dell Universal Xport
  [SCSI] lpfc 8.3.3 : Update driver version to 8.3.3
  [SCSI] lpfc 8.3.3 : Add support for Target Reset handler entrypoint
  [SCSI] lpfc 8.3.3 : Fix a couple of spin_lock and memory issues and a crash
  [SCSI] lpfc 8.3.3 : FC/FCOE discovery fixes
  [SCSI] lpfc 8.3.3 : Fix various SLI-3 vs SLI-4 differences
  [SCSI] qla2xxx: Resolve a performance issue in interrupt
  [SCSI] cnic, bnx2i: Fix build failure when CONFIG_PCI is not set.
  [SCSI] nsp_cs: time_out reaches -1
  [SCSI] qla2xxx: fix printk format warnings
  [SCSI] ncr53c8xx: div reaches -1
  [SCSI] compat: don't perform unneeded copy in sg_io code
  [SCSI] zfcp: Update FC pass-through support
  [SCSI] zfcp: Add FC pass-through support
  [SCSI] FC Pass Thru support

15 years agoMerge branch 'linux-next' of git://git.infradead.org/ubi-2.6
Linus Torvalds [Wed, 17 Jun 2009 16:48:30 +0000 (09:48 -0700)]
Merge branch 'linux-next' of git://git.infradead.org/ubi-2.6

* 'linux-next' of git://git.infradead.org/ubi-2.6: (21 commits)
  UBI: add reboot notifier
  UBI: handle more error codes
  UBI: fix multiple spelling typos
  UBI: fix kmem_cache_free on error patch
  UBI: print amount of reserved PEBs
  UBI: improve messages in the WL worker
  UBI: make gluebi a separate module
  UBI: remove built-in gluebi
  UBI: add notification API
  UBI: do not switch to R/O mode on read errors
  UBI: fix and clean-up error paths in WL worker
  UBI: introduce new constants
  UBI: fix race condition
  UBI: minor serialization fix
  UBI: do not panic if volume check fails
  UBI: add dump_stack in checking code
  UBI: fix races in I/O debugging checks
  UBI: small debugging code optimization
  UBI: improve debugging messages
  UBI: re-name volumes_mutex to device_mutex
  ...

15 years agoMerge branch 'linux-next' of git://git.infradead.org/ubifs-2.6
Linus Torvalds [Wed, 17 Jun 2009 16:46:33 +0000 (09:46 -0700)]
Merge branch 'linux-next' of git://git.infradead.org/ubifs-2.6

* 'linux-next' of git://git.infradead.org/ubifs-2.6:
  UBIFS: start using hrtimers
  hrtimer: export ktime_add_safe
  UBIFS: do not forget to register BDI device
  UBIFS: allow sync option in rootflags
  UBIFS: remove dead code
  UBIFS: use anonymous device
  UBIFS: return proper error code if the compr is not present
  UBIFS: return error if link and unlink race
  UBIFS: reset no_space flag after inode deletion

15 years agoDocumentation/vm/Makefile: don't try to build slqbinfo
Andrew Morton [Wed, 17 Jun 2009 05:38:29 +0000 (22:38 -0700)]
Documentation/vm/Makefile: don't try to build slqbinfo

For it is only in linux-next at this stage.

Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrm/radeon/kms: remove the _DRM_DRIVER from the KMS paths.
Dave Airlie [Wed, 17 Jun 2009 07:21:13 +0000 (17:21 +1000)]
drm/radeon/kms: remove the _DRM_DRIVER from the KMS paths.

This causes an issue since we fixed the drm mappings to do the right thing,
so its just a copy and pasto.

Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg...
Linus Torvalds [Wed, 17 Jun 2009 16:41:25 +0000 (09:41 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/penberg/slab-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6:
  SLUB: Fix early boot GFP_DMA allocations
  SLUB: Don't print out OOM warning for __GFP_NOFAIL
  SLUB: fix build when !SLUB_DEBUG
  SLUB: Out-of-memory diagnostics
  slab: document kzfree() zeroing behavior
  slab: fix generic PAGE_POISONING conflict with SLAB_RED_ZONE
  slob: use PG_slab for identifying SLOB pages

15 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Wed, 17 Jun 2009 16:13:52 +0000 (09:13 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (47 commits)
  MIPS: Add hibernation support
  MIPS: Move Cavium CP0 hwrena impl bits to cpu-feature-overrides.h
  MIPS: Allow CPU specific overriding of CP0 hwrena impl bits.
  MIPS: Kconfig Add SYS_SUPPORTS_HUGETLBFS and enable it for some systems.
  Hugetlbfs: Enable hugetlbfs for more systems in Kconfig.
  MIPS: TLB support for hugetlbfs.
  MIPS: Add hugetlbfs page defines.
  MIPS: Add support files for hugetlbfs.
  MIPS: Remove unused parameters from iPTE_LW.
  Staging: Add octeon-ethernet driver files.
  MIPS: Export erratum function needed by octeon-ethernet driver.
  MIPS: Cavium-Octeon: Add more chip specific feature tests.
  MIPS: Cavium-Octeon: Add more board type constants.
  MIPS: Export cvmx_sysinfo_get needed by octeon-ethernet driver.
  MIPS: Add named alloc functions to OCTEON boot monitor memory allocator.
  MIPS: Alchemy: devboards: Convert to gpio calls.
  MIPS: Alchemy: xxs1500: use linux gpio api.
  MIPS: Alchemy: MTX-1: Use linux gpio api.
  MIPS: Alchemy: Rewrite GPIO support.
  MIPS: Alchemy: Remove unused au1000_gpio.h header
  ...

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Wed, 17 Jun 2009 15:46:57 +0000 (08:46 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  get rid of BKL in fs/sysv
  get rid of BKL in fs/minix
  get rid of BKL in fs/efs
  befs ->pust_super() doesn't need BKL
  Cleanup of adfs headers
  9P doesn't need BKL in ->umount_begin()
  fuse doesn't need BKL in ->umount_begin()
  No instance of ->bmap() needs BKL
  remove unlock_kernel() left accidentally
  ext4: avoid unnecessary spinlock in critical POSIX ACL path
  ext3: avoid unnecessary spinlock in critical POSIX ACL path

15 years agoMIPS: Add hibernation support
Wu Zhangjin [Thu, 4 Jun 2009 12:27:10 +0000 (20:27 +0800)]
MIPS: Add hibernation support

[Ralf: SMP support requires CPU hotplugging which MIPS currently doesn't
support.  As implemented in this patch cache and tlb flushing will also be
invoked with interrupts disabled so smp_call_function() will blow up in
charming ways.  So limit to !SMP.]

Reviewed-by: Pavel Machek <pavel@ucw.cz>
Reviewed-by: Yan Hua <yanh@lemote.com>
Reviewed-by: Arnaud Patard <apatard@mandriva.com>
Reviewed-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Wu Zhangjin <wuzj@lemote.com>
Signed-off-by: Hu Hongbing <huhb@lemote.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Move Cavium CP0 hwrena impl bits to cpu-feature-overrides.h
David Daney [Wed, 13 May 2009 22:59:56 +0000 (15:59 -0700)]
MIPS: Move Cavium CP0 hwrena impl bits to cpu-feature-overrides.h

We had an ugly #ifdef for Cavium Octeon hwrena bits in traps.c, remove
it to mach-cavium-octeon/cpu-feature-overrides.h

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Allow CPU specific overriding of CP0 hwrena impl bits.
David Daney [Wed, 13 May 2009 22:59:55 +0000 (15:59 -0700)]
MIPS: Allow CPU specific overriding of CP0 hwrena impl bits.

Some CPUs have implementation dependent rdhwr registers.  Allow them
to be enabled on a per CPU basis.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Kconfig Add SYS_SUPPORTS_HUGETLBFS and enable it for some systems.
David Daney [Thu, 28 May 2009 00:47:46 +0000 (17:47 -0700)]
MIPS: Kconfig Add SYS_SUPPORTS_HUGETLBFS and enable it for some systems.

Add new kconfig variables SYS_SUPPORTS_HUGETLBFS and
CPU_SUPPORTS_HUGEPAGES.  They are enabled for systems that are known
to support huge pages.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoHugetlbfs: Enable hugetlbfs for more systems in Kconfig.
David Daney [Thu, 28 May 2009 00:47:45 +0000 (17:47 -0700)]
Hugetlbfs: Enable hugetlbfs for more systems in Kconfig.

As part of adding hugetlbfs support for MIPS, I am adding a new
kconfig variable 'SYS_SUPPORTS_HUGETLBFS'.  Since some mips cpu
varients don't yet support it, we can enable selection of HUGETLBFS on
a system by system basis from the arch/mips/Kconfig.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
CC: William Irwin <wli@holomorphy.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TLB support for hugetlbfs.
David Daney [Thu, 28 May 2009 00:47:44 +0000 (17:47 -0700)]
MIPS: TLB support for hugetlbfs.

The TLB handlers need to check for huge pages and give them special
handling.  Huge pages consist of two contiguous sub-pages of physical
memory.

* Loading entrylo0 and entrylo1 need to be handled specially.

* The page mask must be set for huge pages and then restored after
  writing the TLB entries.

* The PTE for huge pages resides in the PMD, we halt traversal of the
  tables there.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add hugetlbfs page defines.
David Daney [Thu, 28 May 2009 00:47:43 +0000 (17:47 -0700)]
MIPS: Add hugetlbfs page defines.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add support files for hugetlbfs.
David Daney [Thu, 28 May 2009 00:47:42 +0000 (17:47 -0700)]
MIPS: Add support files for hugetlbfs.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Remove unused parameters from iPTE_LW.
David Daney [Fri, 8 May 2009 22:10:50 +0000 (15:10 -0700)]
MIPS: Remove unused parameters from iPTE_LW.

The l parameter to iPTE_LW() is unused. Remove it and from some of its
callers as well.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoStaging: Add octeon-ethernet driver files.
David Daney [Wed, 6 May 2009 00:35:21 +0000 (17:35 -0700)]
Staging: Add octeon-ethernet driver files.

The octeon-ethernet driver supports the sgmii, rgmii, spi, and xaui
ports present on the Cavium OCTEON family of SOCs.  These SOCs are
multi-core mips64 processors with existing support over in arch/mips.

The driver files can be categorized into three basic groups:

1) Register definitions, these are named cvmx-*-defs.h

2) Main driver code, these have names that don't start cvmx-.

3) Interface specific functions and other utility code, names starting
with cvmx-

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Export erratum function needed by octeon-ethernet driver.
David Daney [Wed, 6 May 2009 00:35:20 +0000 (17:35 -0700)]
MIPS: Export erratum function needed by octeon-ethernet driver.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Cavium-Octeon: Add more chip specific feature tests.
David Daney [Wed, 6 May 2009 00:35:19 +0000 (17:35 -0700)]
MIPS: Cavium-Octeon: Add more chip specific feature tests.

The octeon-ethernet driver needs to check for additional chip specific
features, we add them to the octeon_has_feature() framework.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Cavium-Octeon: Add more board type constants.
David Daney [Wed, 6 May 2009 00:35:18 +0000 (17:35 -0700)]
MIPS: Cavium-Octeon: Add more board type constants.

The bootloader now uses additional board type constants.  The
octeon-ethernet driver needs some of the new values.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Export cvmx_sysinfo_get needed by octeon-ethernet driver.
David Daney [Wed, 6 May 2009 00:35:17 +0000 (17:35 -0700)]
MIPS: Export cvmx_sysinfo_get needed by octeon-ethernet driver.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add named alloc functions to OCTEON boot monitor memory allocator.
David Daney [Wed, 6 May 2009 00:35:16 +0000 (17:35 -0700)]
MIPS: Add named alloc functions to OCTEON boot monitor memory allocator.

The various Octeon ethernet drivers use these new functions.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Alchemy: devboards: Convert to gpio calls.
Manuel Lauss [Sat, 6 Jun 2009 12:09:58 +0000 (14:09 +0200)]
MIPS: Alchemy: devboards: Convert to gpio calls.

Replace a few open-coded GPIO register accesses with gpio calls.

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Alchemy: xxs1500: use linux gpio api.
Manuel Lauss [Sat, 6 Jun 2009 12:09:57 +0000 (14:09 +0200)]
MIPS: Alchemy: xxs1500: use linux gpio api.

Replace a few GPIO register accesses in the board init code with calls to
the gpio api.

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Alchemy: MTX-1: Use linux gpio api.
Manuel Lauss [Sat, 6 Jun 2009 12:09:56 +0000 (14:09 +0200)]
MIPS: Alchemy: MTX-1: Use linux gpio api.

Replace a few GPIO register accesses in the board init code with calls
to the gpio api.

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Alchemy: Rewrite GPIO support.
Manuel Lauss [Sat, 6 Jun 2009 12:09:55 +0000 (14:09 +0200)]
MIPS: Alchemy: Rewrite GPIO support.

The current in-kernel Alchemy GPIO support is far too inflexible for
all my use cases.  To address this, the following changes are made:

* create generic functions which deal with manipulating the on-chip
  GPIO1/2 blocks.  Such functions are universally useful.
* Macros for GPIO2 shared interrupt management and block control.
* support for both built-in CONFIG_GPIOLIB and fast, inlined GPIO macros.

  If CONFIG_GPIOLIB is not enabled, provide linux gpio framework
  compatibility by directly inlining the GPIO1/2 functions.  GPIO access
  is limited to on-chip ones and they can be accessed as documented in
  the datasheets (GPIO0-31 and 200-215).

  If CONFIG_GPIOLIB is selected, two (2) gpio_chip-s, one for GPIO1 and
  one for GPIO2, are registered.  GPIOs can still be accessed by using
  the numberspace established in the databooks.

  However this is not yet flexible enough for my uses:  My Alchemy
  systems have a documented "external" gpio interface (fixed, different
  numberspace) and can support a variety of baseboards, some of which
  are equipped with I2C gpio expanders.  I want to be able to provide
  the default 16 GPIOs of the CPU board numbered as 0..15 and also
  support gpio expanders, if present, starting as gpio16.

  To achieve this, a new Kconfig symbol for Alchemy is introduced,
  CONFIG_ALCHEMY_GPIO_INDIRECT, which boards can enable to signal
  that they don't want the Alchemy numberspace exposed to the outside
  world, but instead want to provide their own.  Boards are now respon-
  sible for providing the linux gpio interface glue code (either in a
  custom gpio.h header (in board include directory) or with gpio_chips).

  To make the board-specific inlined gpio functions work, the MIPS
  Makefile must be changed so that the mach-au1x00/gpio.h header is
  included _after_ the board headers, by moving the inclusion of
  the mach-au1x00/ to the end of the header list.

  See arch/mips/include/asm/mach-au1x00/gpio.h for more info.

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Alchemy: Remove unused au1000_gpio.h header
Manuel Lauss [Sat, 6 Jun 2009 12:09:54 +0000 (14:09 +0200)]
MIPS: Alchemy: Remove unused au1000_gpio.h header

Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: BCM47xx: Fix gpio_direction_output
Matthieu Castet [Sun, 24 May 2009 17:48:51 +0000 (19:48 +0200)]
MIPS: BCM47xx: Fix gpio_direction_output

gpio_direction_output should also set an output value according to the API.

Signed-off-by: Matthieu CASTET <castet.matthieu@free.fr>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: ioctl.h: Cleanup.
Ralf Baechle [Wed, 17 Jun 2009 10:06:28 +0000 (11:06 +0100)]
MIPS: ioctl.h: Cleanup.

 o Rewrite to use <asm-generic/ioctl.h>.  Cuts down the file from 40 to
   16 lines.
 o Delete _IOC_VOID, _IOC_OUT, _IOC_IN and _IOC_INOUT.  They were added
   for 2.1.14 but I was not able to find any user - not even historical
   ones.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: Add TX4939 RNG support
Atsushi Nemoto [Tue, 2 Jun 2009 14:54:22 +0000 (23:54 +0900)]
MIPS: TXx9: Add TX4939 RNG support

Add platform support for RNG of TX4939 SoC.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: hwrng: Add TX4939 RNG driver
Atsushi Nemoto [Tue, 2 Jun 2009 14:54:21 +0000 (23:54 +0900)]
MIPS: hwrng: Add TX4939 RNG driver

This patch adds support for the integrated RNG of the TX4939 SoC.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: Add SRAMC support
Atsushi Nemoto [Mon, 25 May 2009 13:04:02 +0000 (22:04 +0900)]
MIPS: TXx9: Add SRAMC support

Add a sysdev to access SRAM in TXx9 SoCs via sysfs.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Sibyte: Remove standalone kernel support
Imre Kaloz [Tue, 2 Jun 2009 12:22:06 +0000 (14:22 +0200)]
MIPS: Sibyte: Remove standalone kernel support

CFE is the only supported and used bootloader on the SiByte boards,
the standalone kernel support has been never used outside Broadcom.
Remove it and make the kernel use CFE by default.

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Sibyte: Remove simulator option
Imre Kaloz [Tue, 2 Jun 2009 12:22:00 +0000 (14:22 +0200)]
MIPS: Sibyte: Remove simulator option

This patch removes the SiByte simulation Kconfig option, which only modified
a printk.

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: RB532: Check irq number when handling GPIO interrupts
Florian Fainelli [Thu, 21 May 2009 17:49:47 +0000 (19:49 +0200)]
MIPS: RB532: Check irq number when handling GPIO interrupts

This patch makes sure that we are not going to clear
or change the interrupt status of a GPIO interrupt
superior to 13 as this is the maximum number of GPIO
interrupt source (p.232 of the RC32434 reference manual).

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: RB532: Cleanup cpu-features-overrides
Florian Fainelli [Thu, 21 May 2009 17:49:39 +0000 (19:49 +0200)]
MIPS: RB532: Cleanup cpu-features-overrides

Remove commented out definitions.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Remove dead case label.
David Daney [Tue, 12 May 2009 19:41:55 +0000 (12:41 -0700)]
MIPS: Remove dead case label.

CPU_CAVIUM_OCTEON is mips_r2 which is handled before the switch.  This
label in the switch statement is dead code, so we remove it.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Reviewed by: David VomLehn <dvomlehn@cisco.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Remove execution hazard barriers for Octeon.
David Daney [Tue, 12 May 2009 19:41:54 +0000 (12:41 -0700)]
MIPS: Remove execution hazard barriers for Octeon.

The Octeon has no execution hazards, so we can remove them and save an
instruction per TLB handler invocation.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Reviewed by: David VomLehn <dvomlehn@cisco.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Allow R2 CPUs to turn off generation of 'ehb' instructions.
David Daney [Tue, 12 May 2009 19:41:53 +0000 (12:41 -0700)]
MIPS: Allow R2 CPUs to turn off generation of 'ehb' instructions.

Some CPUs do not need ehb instructions after writing CP0 registers.
By allowing ehb generation to be overridden in
cpu-feature-overrides.h, we can save a few instructions in the TLB
handler hot paths.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Fold the TLB refill at the vmalloc path if possible.
David Daney [Wed, 20 May 2009 18:40:59 +0000 (11:40 -0700)]
MIPS: Fold the TLB refill at the vmalloc path if possible.

Try to fold the 64-bit TLB refill handler opportunistically at the
beginning of the vmalloc path so as to avoid splitting execution flow in
half and wasting cycles for a branch required at that point then.  Resort
to doing the split if either of the newly created parts would not fit into
its designated slot.

Original-patch-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Replace some magic numbers with symbolic values in tlbex.c
David Daney [Wed, 20 May 2009 18:40:58 +0000 (11:40 -0700)]
MIPS: Replace some magic numbers with symbolic values in tlbex.c

The logic used to split the r4000 refill handler is liberally
sprinkled with magic numbers.  We attempt to explain what they are and
normalize them against a new symbolic value (MIPS64_REFILL_INSNS).

CC: David VomLehn <dvomlehn@cisco.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: Add ACLC support
Atsushi Nemoto [Tue, 19 May 2009 13:12:22 +0000 (22:12 +0900)]
MIPS: TXx9: Add ACLC support

Add platform support for ACLC of TXx9 SoCs.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add Cavium OCTEON PCI support.
David Daney [Fri, 24 Apr 2009 00:44:38 +0000 (17:44 -0700)]
MIPS: Add Cavium OCTEON PCI support.

This patch adds support for PCI and PCIe to the base Cavium OCTEON
processor support.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add register definitions for PCI.
David Daney [Fri, 24 Apr 2009 00:44:37 +0000 (17:44 -0700)]
MIPS: Add register definitions for PCI.

Here we add the register definitions for the processor blocks used by
the following PCI support patch.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: Add DMAC support
Atsushi Nemoto [Wed, 22 Apr 2009 15:40:31 +0000 (00:40 +0900)]
MIPS: TXx9: Add DMAC support

Add platform support for DMAC of TXx9 SoCs.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoDMA: TXx9 Soc DMA Controller driver
Atsushi Nemoto [Wed, 22 Apr 2009 15:40:30 +0000 (00:40 +0900)]
DMA: TXx9 Soc DMA Controller driver

This patch adds support for the integrated DMAC of the TXx9 family.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: micro optimization for clocksource and clock_event
Atsushi Nemoto [Thu, 23 Apr 2009 15:10:36 +0000 (00:10 +0900)]
MIPS: TXx9: micro optimization for clocksource and clock_event

Use container structure for clocksource, clock_event_device and hold a
pointer to txx9_tmr_reg in it.

This saves a few instructions in clocksource and clock_event handlers.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Support 64-byte D-cache line size
Kevin Cernekee [Fri, 24 Apr 2009 00:36:53 +0000 (17:36 -0700)]
MIPS: Support 64-byte D-cache line size

Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Pass struct device to plat_dma_addr_to_phys()
Kevin Cernekee [Fri, 24 Apr 2009 00:25:12 +0000 (17:25 -0700)]
MIPS: Pass struct device to plat_dma_addr_to_phys()

Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Add size and direction arguments to plat_unmap_dma_mem()
Kevin Cernekee [Fri, 24 Apr 2009 00:03:43 +0000 (17:03 -0700)]
MIPS: Add size and direction arguments to plat_unmap_dma_mem()

Signed-off-by: Kevin Cernekee <cernekee@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: SB1250: Sort out merge mistake.
Ralf Baechle [Wed, 17 Jun 2009 10:06:24 +0000 (11:06 +0100)]
MIPS: SB1250: Sort out merge mistake.

A wrong resolution of a merge conflict made the recently deleted wrong
error check in sb1250_set_affinity.  Send the zombie back to the empire
of the undead.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Fix __ndelay build error and add 'ull' suffix for 32-bit kernel
Atsushi Nemoto [Tue, 9 Jun 2009 02:12:48 +0000 (11:12 +0900)]
MIPS: Fix __ndelay build error and add 'ull' suffix for 32-bit kernel

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: SMTC: Fix formatting difference to linux-mips.org code
Ralf Baechle [Wed, 17 Jun 2009 10:06:24 +0000 (11:06 +0100)]
MIPS: SMTC: Fix formatting difference to linux-mips.org code

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Fix typo resulting in far too long ndelay times.
Ralf Baechle [Fri, 12 Jun 2009 16:28:00 +0000 (17:28 +0100)]
MIPS: Fix typo resulting in far too long ndelay times.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMerge branch 'slub/earlyboot' into for-linus
Pekka Enberg [Wed, 17 Jun 2009 05:30:54 +0000 (08:30 +0300)]
Merge branch 'slub/earlyboot' into for-linus

Conflicts:
mm/slub.c

15 years agoMerge branches 'slab/documentation', 'slab/fixes', 'slob/cleanups' and 'slub/fixes...
Pekka Enberg [Wed, 17 Jun 2009 05:30:15 +0000 (08:30 +0300)]
Merge branches 'slab/documentation', 'slab/fixes', 'slob/cleanups' and 'slub/fixes' into for-linus

15 years agoget rid of BKL in fs/sysv
Al Viro [Wed, 17 Jun 2009 03:59:37 +0000 (23:59 -0400)]
get rid of BKL in fs/sysv

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoget rid of BKL in fs/minix
Al Viro [Wed, 17 Jun 2009 03:47:45 +0000 (23:47 -0400)]
get rid of BKL in fs/minix

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoget rid of BKL in fs/efs
Al Viro [Wed, 17 Jun 2009 03:35:46 +0000 (23:35 -0400)]
get rid of BKL in fs/efs

Only readdir() really needed it, and that's easily fixable by switch to
generic_file_llseek()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agobefs ->pust_super() doesn't need BKL
Al Viro [Wed, 17 Jun 2009 03:24:50 +0000 (23:24 -0400)]
befs ->pust_super() doesn't need BKL

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoCleanup of adfs headers
Al Viro [Tue, 16 Jun 2009 18:52:13 +0000 (14:52 -0400)]
Cleanup of adfs headers

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years ago9P doesn't need BKL in ->umount_begin()
Al Viro [Tue, 16 Jun 2009 18:17:21 +0000 (14:17 -0400)]
9P doesn't need BKL in ->umount_begin()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agofuse doesn't need BKL in ->umount_begin()
Al Viro [Tue, 16 Jun 2009 18:15:00 +0000 (14:15 -0400)]
fuse doesn't need BKL in ->umount_begin()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoNo instance of ->bmap() needs BKL
Al Viro [Tue, 16 Jun 2009 17:35:01 +0000 (13:35 -0400)]
No instance of ->bmap() needs BKL

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoremove unlock_kernel() left accidentally
J. R. Okajima [Tue, 16 Jun 2009 16:16:50 +0000 (01:16 +0900)]
remove unlock_kernel() left accidentally

commit 337eb00a2c3a421999c39c94ce7e33545ee8baa7
Push BKL down into ->remount_fs()
and
commit 4aa98cf768b6f2ea4b204620d949a665959214f6
Push BKL down into do_remount_sb()

were uncorrectly merged.
The former removes one pair of lock/unlock_kernel(), but the latter adds
several unlock_kernel(). Finally a few unlock_kernel() calls left.

Signed-off-by: J. R. Okajima <hooanon05@yahoo.co.jp>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoext4: avoid unnecessary spinlock in critical POSIX ACL path
Theodore Ts'o [Mon, 8 Jun 2009 19:22:25 +0000 (15:22 -0400)]
ext4: avoid unnecessary spinlock in critical POSIX ACL path

If a filesystem supports POSIX ACL's, the VFS layer expects the filesystem
to do POSIX ACL checks on any files not owned by the caller, and it does
this for every single pathname component that it looks up.

That obviously can be pretty expensive if the filesystem isn't careful
about it, especially with locking. That's doubly sad, since the common
case tends to be that there are no ACL's associated with the files in
question.

ext4 already caches the ACL data so that it doesn't have to look it up
over and over again, but it does so by taking the inode->i_lock spinlock
on every lookup. Which is a noticeable overhead even if it's a private
lock, especially on CPU's where the serialization is expensive (eg Intel
Netburst aka 'P4').

For the special case of not actually having any ACL's, all that locking is
unnecessary. Even if somebody else were to be changing the ACL's on
another CPU, we simply don't care - if we've seen a NULL ACL, we might as
well use it.

So just load the ACL speculatively without any locking, and if it was
NULL, just use it. If it's non-NULL (either because we had a cached
entry, or because the cache hasn't been filled in at all), it means that
we'll need to get the lock and re-load it properly.

(This commit was ported from a patch originally authored by Linus for
ext3.)

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoext3: avoid unnecessary spinlock in critical POSIX ACL path
Linus Torvalds [Mon, 8 Jun 2009 19:22:24 +0000 (15:22 -0400)]
ext3: avoid unnecessary spinlock in critical POSIX ACL path

If a filesystem supports POSIX ACL's, the VFS layer expects the filesystem
to do POSIX ACL checks on any files not owned by the caller, and it does
this for every single pathname component that it looks up.

That obviously can be pretty expensive if the filesystem isn't careful
about it, especially with locking. That's doubly sad, since the common
case tends to be that there are no ACL's associated with the files in
question.

ext3 already caches the ACL data so that it doesn't have to look it up
over and over again, but it does so by taking the inode->i_lock spinlock
on every lookup. Which is a noticeable overhead even if it's a private
lock, especially on CPU's where the serialization is expensive (eg Intel
Netburst aka 'P4').

For the special case of not actually having any ACL's, all that locking is
unnecessary. Even if somebody else were to be changing the ACL's on
another CPU, we simply don't care - if we've seen a NULL ACL, we might as
well use it.

So just load the ACL speculatively without any locking, and if it was
NULL, just use it. If it's non-NULL (either because we had a cached
entry, or because the cache hasn't been filled in at all), it means that
we'll need to get the lock and re-load it properly.

This is noticeable even on Nehalem, which does locking quite well (much
better than P4). From lmbench:

Processor, Processes - times in microseconds - smaller is better
--------------------------------------------------------------------
Host                 OS  Mhz null null      open slct fork exec sh
                             call  I/O stat clos TCP  proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ---- ---- ---- ----
 - before:
nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.95 1.45 2.18 69.1 273. 1141
nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.95 1.48 2.28 69.9 253. 1140
nehalem.l Linux 2.6.30- 3193 0.04 0.10 0.95 1.42 2.19 68.6 284. 1141
 - after:
nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.44 2.12 68.3 282. 1094
nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.39 2.20 67.0 308. 1123
nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.39 2.36 67.4 293. 1148

where you can see what appears to be a roughly 3% improvement in stat
and open/close latencies from just the removal of the locking overhead.

Of course, this only matters for files you don't own (the owner never
needs to do the ACL checks), but that's the common case for libraries,
header files, and executables. As well as for the base components of any
absolute pathname, even if you are the owner of the final file.

[ At some point we probably want to move this ACL caching logic entirely
  into the VFS layer (and only call down to the filesystem when
  uncached), but in the meantime this improves ext3 a bit.

  A similar fix to btrfs makes a much bigger difference (15x improvement
  in lmbench) due to broken caching. ]

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Acked-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoMerge branch 'next-i2c' of git://aeryn.fluff.org.uk/bjdooks/linux
Linus Torvalds [Wed, 17 Jun 2009 04:26:42 +0000 (21:26 -0700)]
Merge branch 'next-i2c' of git://aeryn.fluff.org.uk/bjdooks/linux

* 'next-i2c' of git://aeryn.fluff.org.uk/bjdooks/linux:
  i2c-stu300: Make driver depend on MACH_U300
  i2c-s3c2410: use resource_size()
  i2c: Use resource_size macro
  i2c: ST DDC I2C U300 bus driver v3
  i2c-bfin-twi: pull in io.h for ioremap()

15 years agoMerge branch 'drm-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Wed, 17 Jun 2009 04:20:39 +0000 (21:20 -0700)]
Merge branch 'drm-linus' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/radeon: switch to using late_initcall
  radeon legacy chips: tv dac bg/dac adj updates
  drm/radeon: introduce kernel modesetting for radeon hardware
  drm: Add the TTM GPU memory manager subsystem.
  drm: Memory fragmentation from lost alignment blocks
  drm/radeon: fix mobility flags on new PCI IDs.

15 years agoAFS: Correctly translate auth error aborts and don't failover in such cases
David Howells [Tue, 16 Jun 2009 20:36:49 +0000 (21:36 +0100)]
AFS: Correctly translate auth error aborts and don't failover in such cases

Authentication error abort codes should be translated to appropriate
Linux error codes, rather than all being translated to EREMOTEIO - which
indicates that the server had internal problems.

Additionally, a server shouldn't be marked unavailable and the next
server tried if an authentication error occurs.  This will quickly make
all the servers unavailable to the client.  Instead the error should be
returned straight to the user.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoRxRPC: Don't attempt to reuse aborted connections
David Howells [Tue, 16 Jun 2009 20:36:44 +0000 (21:36 +0100)]
RxRPC: Don't attempt to reuse aborted connections

Connections that have seen a connection-level abort should not be reused
as the far end will just abort them again; instead a new connection
should be made.

Connection-level aborts occur due to such things as authentication
failures.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>