firefly-linux-kernel-4.4.55.git
15 years agodrm/radeon: RS600: fix interrupt handling
Alex Deucher [Fri, 6 Mar 2009 16:47:54 +0000 (11:47 -0500)]
drm/radeon: RS600: fix interrupt handling

the checks weren't updated when RS600 support
was added.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm/r600: fix rptr address along lines of previous fixes to radeon.
Dave Airlie [Thu, 26 Feb 2009 00:15:24 +0000 (10:15 +1000)]
drm/r600: fix rptr address along lines of previous fixes to radeon.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/r600: fixup r600 gart table accessor like ati_pcigart.c
Dave Airlie [Thu, 26 Feb 2009 00:14:40 +0000 (10:14 +1000)]
drm/r600: fixup r600 gart table accessor like ati_pcigart.c

This attempts to fixup the r600 GART accessors so they work on other arches.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/ati_pcigart: use memset_io to reset the memory
Dave Airlie [Thu, 26 Feb 2009 00:13:47 +0000 (10:13 +1000)]
drm/ati_pcigart: use memset_io to reset the memory

Also don't setup pci_gart if we aren't going to need it.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: add DRM_READ/WRITE64 wrappers around readq/writeq.
Dave Airlie [Thu, 26 Feb 2009 00:12:10 +0000 (10:12 +1000)]
drm: add DRM_READ/WRITE64 wrappers around readq/writeq.

The readq/writeq stuff is from Dave Miller, and he
warns users to be careful about using these. Plans are only
r600 to use it so far.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: add RS600 pci ids
Alex Deucher [Wed, 25 Feb 2009 22:02:19 +0000 (17:02 -0500)]
radeon: add RS600 pci ids

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: add support for rs600 GPUs
Alex Deucher [Wed, 25 Feb 2009 21:57:49 +0000 (16:57 -0500)]
radeon: add support for rs600 GPUs

RS600s are an AMD IGP for Intel CPUs, that look like RS690s from
a lot of perspectives but look like r600s from a memory controller
point of view.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: fix r600 AGP support
Alex Deucher [Wed, 25 Feb 2009 20:55:01 +0000 (15:55 -0500)]
radeon: fix r600 AGP support

This fixes the ioremap issues with r600 AGP.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeon: add R6xx/R7xx pci ids
Alex Deucher [Tue, 24 Feb 2009 22:13:42 +0000 (17:13 -0500)]
radeon: add R6xx/R7xx pci ids

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: add initial support for R6xx/R7xx GPUs
Alex Deucher [Tue, 24 Feb 2009 21:22:29 +0000 (16:22 -0500)]
drm/radeon: add initial support for R6xx/R7xx GPUs

This adds support for 2D/Xv acceleration in the X.org 2D driver,
to the drm. It doesn't yet provide any 3D support hooks.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: add r6xx/r7xx microcode
Alex Deucher [Tue, 24 Feb 2009 19:28:34 +0000 (14:28 -0500)]
drm/radeon: add r6xx/r7xx microcode

This uses the same microcode system as the current radeon code.

It should be converted to the new microcode loader I suppose,
though really I need a lot more proof of the worth of me maintaining
firmware blobs externally.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: prep for r6xx/r7xx support
Alex Deucher [Tue, 24 Feb 2009 19:02:13 +0000 (14:02 -0500)]
drm/radeon: prep for r6xx/r7xx support

- add r6xx/r7xx regs and macros
- add r6xx/r7xx chip families
- fix register access for regs with offsets >= 0x10000

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoi915/drm: Remove two redundant agp_chipset_flushes
Owain G. Ainsworth [Fri, 20 Feb 2009 08:30:19 +0000 (08:30 +0000)]
i915/drm: Remove two redundant agp_chipset_flushes

agp_chipset_flush() is for flushing the intel GMCH write cache via the
IFP, these two uses are for when we're getting the object into the cpu
READ domain, and thus should not be needed. This confused me when I was
getting my head around the code.

With thanks to airlied for helping me check my mental picture of how the
flushes and clflushes are supposed to be used.

Signed-off-by: Owain G. Ainsworth <oga@openbsd.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/i915: Display fence register state in debugfs i915_gem_fence_regs node.
Chris Wilson [Wed, 11 Feb 2009 14:26:38 +0000 (14:26 +0000)]
drm/i915: Display fence register state in debugfs i915_gem_fence_regs node.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/i915: Add information on pinning and fencing to the i915 list debug.
Eric Anholt [Wed, 18 Feb 2009 07:53:41 +0000 (23:53 -0800)]
drm/i915: Add information on pinning and fencing to the i915 list debug.

This was inspired by a patch by Chris Wilson, though none of it applied in any
way due to the debugfs work and I decided to change the formatting of the
new information anyway.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/i915: Consolidate gem object list dumping
Ben Gamari [Wed, 18 Feb 2009 01:08:51 +0000 (20:08 -0500)]
drm/i915: Consolidate gem object list dumping

Here we eliminate a few functions in favor of using a single function
to dump from all of the object lists.

Signed-Off-By: Ben Gamari <bgamari@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: Convert proc files to seq_file and introduce debugfs
Ben Gamari [Wed, 18 Feb 2009 01:08:49 +0000 (20:08 -0500)]
drm: Convert proc files to seq_file and introduce debugfs

The old mechanism to formatting proc files is extremely ugly. The
seq_file API was designed specifically for cases like this and greatly
simplifies the process.

Also, most of the files in /proc really don't belong there. This patch
introduces the infrastructure for putting these into debugfs and exposes
all of the proc files in debugfs as well.

This contains the i915 hooks rewrite as well, to make bisectability better.

Signed-off-by: Ben Gamari <bgamari@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: split busmaster enable out to a separate function
Dave Airlie [Fri, 20 Feb 2009 03:28:59 +0000 (13:28 +1000)]
drm/radeon: split busmaster enable out to a separate function

this is just a code cleanup from the kms tree.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: align ring writes to 16 dwords boundaries.
Dave Airlie [Fri, 20 Feb 2009 03:28:34 +0000 (13:28 +1000)]
drm/radeon: align ring writes to 16 dwords boundaries.

On some radeon GPUs this appears to introduce another level of
stability around interacting with the ring.

Its pretty much what fglrx appears to do.

Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm/radeon: Print PCI ID of cards when probing
Benjamin Herrenschmidt [Fri, 6 Feb 2009 05:46:27 +0000 (16:46 +1100)]
drm/radeon: Print PCI ID of cards when probing

This is usedul when you have multiple cards to figure out which
one is which minor.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: Only use DRM_IOCTL_UPDATE_DRAW compat wrapper for compat X86.
David Miller [Wed, 18 Feb 2009 09:35:21 +0000 (01:35 -0800)]
drm: Only use DRM_IOCTL_UPDATE_DRAW compat wrapper for compat X86.

Only X86 32-bit uses a different alignment for "unsigned long long"
than it's 64-bit counterpart.

Therefore this compat translation is only correct, and only needed,
when either CONFIG_X86 or CONFIG_IA64.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: radeon: Fix unaligned access in r300_scratch().
David Miller [Wed, 18 Feb 2009 09:35:23 +0000 (01:35 -0800)]
drm: radeon: Fix unaligned access in r300_scratch().

In compat mode, the cmdbuf->buf 64-bit address cookie can
potentially be only 32-bit aligned.  Dereferencing this as
64-bit causes expensive unaligned traps on platforms like
sparc64.

Use get_unaligned() to fix.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: Preserve SHMLBA bits in hash key for _DRM_SHM mappings.
David Miller [Wed, 18 Feb 2009 23:41:02 +0000 (15:41 -0800)]
drm: Preserve SHMLBA bits in hash key for _DRM_SHM mappings.

Platforms such as sparc64 have D-cache aliasing issues.  We
cannot allow virtual mappings in different contexts to be such
that two cache lines can be loaded for the same backing data.
Updates to one cache line won't be seen by accesses to the other
cache line.

Code in sparc64 and other architectures solve this problem by
making sure that all userland mappings of MAP_SHARED objects have
the same virtual address base.  They implement this by keying
off of the page offset, and using that to choose a suitably
consistent virtual address for mmap() requests.

Making things even worse, getting this wrong on sparc64 can result
in hangs during DRM lock acquisition.  This is because, at least on
UltraSPARC-III, normal loads consult the D-cache but atomics such
as 'cas' (which is what cmpxchg() is implement using) only consult
the L2 cache.  So if a D-cache alias is inserted, the load can
see different data than the atomic, and we'll loop forever because
the atomic compare-and-exchange will never complete successfully.

So to make this all work properly, we need to make sure that the
hash address computed by drm_map_handle() preserves the SHMLBA
relevant bits, and that's what this patch does for _DRM_SHM mappings.

As a historical note, many years ago this bug didn't exist because we
used to just use the low 32-bits of the address as the hash and just
hope for the best.  This preserved the SHMLBA bits properly.  But when
the hashtab code was added to DRM, this was no longer the case.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agodrm: ati_pcigart: Fix limit check in drm_ati_pcigart_init().
David Miller [Sun, 15 Feb 2009 09:08:07 +0000 (01:08 -0800)]
drm: ati_pcigart: Fix limit check in drm_ati_pcigart_init().

The variable 'max_pages' is ambiguous.  There are two concepts
of "pages" being used in this function.

First, we have ATI GART pages which are always 4096 bytes.
Then, we have system pages which are of size PAGE_SIZE.

Eliminate the confusion by creating max_ati_pages and
max_real_pages.  Calculate and use them as appropriate.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: radeon: Use surface for PCI GART table.
David Miller [Sat, 14 Feb 2009 09:51:07 +0000 (01:51 -0800)]
drm: radeon: Use surface for PCI GART table.

This allocates a physical surface for the PCI GART table, this way no
matter what other surface configurations exist the GART table will
always be seen by the hardware properly.

We encode the file pointer of the virtual surface allocate using a
special cookie value, called PCIGART_FILE_PRIV.  On the last close, we
release that surface.

Just to be doubly safe, we run the pcigart table setup with the main
surface control register clear.

Based upon ideas from David Airlie and Ben Benjamin Herrenschmidt.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: radeon: Fix calculation of RB_RPTR_ADDR in non-AGP case.
David Miller [Thu, 12 Feb 2009 10:15:44 +0000 (02:15 -0800)]
drm: radeon: Fix calculation of RB_RPTR_ADDR in non-AGP case.

The address needs to be a GART relative address, rather than a PCI
DMA address.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: radeon: Fix RADEON_*_EMITED defines.
David Miller [Thu, 12 Feb 2009 10:15:39 +0000 (02:15 -0800)]
drm: radeon: Fix RADEON_*_EMITED defines.

These are not supposed to be booleans, they are
supposed to be bit masks.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: radeon: Fix ring_rptr accesses.
David Miller [Thu, 12 Feb 2009 10:15:37 +0000 (02:15 -0800)]
drm: radeon: Fix ring_rptr accesses.

The memory behind ring_rptr can either be in ioremapped memory
or a vmalloc() normal kernel memory buffer.

However, the code unconditionally uses DRM_{READ,WRITE}32() (and thus
readl() and writel()) to access it.

Basically, if RADEON_IS_AGP then it's ioremap()'d memory else it's
vmalloc'd memory.

Adjust all of the ring_rptr access code as needed.

While we're here, kill the 'scratch' pointer in drm_radeon_private.
It's only used in the one place where it is initialized.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: ati_pcigart: Need to use PCI_DMA_BIDIRECTIONAL.
David Miller [Thu, 12 Feb 2009 10:15:34 +0000 (02:15 -0800)]
drm: ati_pcigart: Need to use PCI_DMA_BIDIRECTIONAL.

The buffers mapped by the PCI GART can be written to by the device,
not just read.

For example, this happens via the RB_RPTR writeback on Radeon.

So we can't use PCI_DMA_TODEVICE else we'll get protection faults
on IOMMU platforms.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: ati_pcigart: Do not access I/O MEM space using pointer derefs.
David Miller [Thu, 12 Feb 2009 10:15:27 +0000 (02:15 -0800)]
drm: ati_pcigart: Do not access I/O MEM space using pointer derefs.

The PCI GART table initialization code treats the GART table mapping
unconditionally as a kernel virtual address.

But it could be in the framebuffer, for example, and thus we're
dealing with a PCI MEM space ioremap() cookie.  Treating that as a
virtual address is illegal and will crash some system types (such as
sparc64 where the ioremap() return value is actually a physical I/O
address).

So access the area correctly, using gart_info->gart_table_location as
our guide.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: Drop unused and broken dri_library_name sysfs attribute.
Kristian Høgsberg [Mon, 5 Jan 2009 21:10:05 +0000 (16:10 -0500)]
drm: Drop unused and broken dri_library_name sysfs attribute.

The kernel shouldn't be in the business of telling user space which
driver to load.  The kernel defers mapping PCI IDs to module names
to user space and we should do the same for DRI drivers.

And in fact, that's how it does work today.  Nothing uses the
dri_library_name attribute, and the attribute is in fact broken.
For intel devices, it falls back to the default behaviour of returning
the kernel module name as the DRI driver name, which doesn't work for
i965 devices.  Nobody has ever hit this problem or filed a bug about this.

Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: claim PCI device when running in modesetting mode.
Kristian Høgsberg [Sun, 4 Jan 2009 21:55:33 +0000 (16:55 -0500)]
drm: claim PCI device when running in modesetting mode.

Under kernel modesetting, we manage the device at all times, regardless
of VT switching and X servers, so the only decent thing to do is to
claim the PCI device.  In that case, we call the suspend/resume hooks
directly from the pci driver hooks instead of the current class device detour.

Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: Make drm_local_map use a resource_size_t offset
Benjamin Herrenschmidt [Mon, 2 Feb 2009 05:55:47 +0000 (16:55 +1100)]
drm: Make drm_local_map use a resource_size_t offset

This changes drm_local_map to use a resource_size for its "offset"
member instead of an unsigned long, thus allowing 32-bit machines
with a >32-bit physical address space to be able to store there
their register or framebuffer addresses when those are above 4G,
such as when using a PCI video card on a recent AMCC 440 SoC.

This patch isn't as "trivial" as it sounds: A few functions needed
to have some unsigned long/int changed to resource_size_t and a few
printk's had to be adjusted.

But also, because userspace isn't capable of passing such offsets,
I had to modify drm_find_matching_map() to ignore the offset passed
in for maps of type _DRM_FRAMEBUFFER or _DRM_REGISTERS.

If we ever support multiple _DRM_FRAMEBUFFER or _DRM_REGISTERS maps
for a given device, we might have to change that trick, but I don't
think that happens on any current driver.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: Split drm_map and drm_local_map
Benjamin Herrenschmidt [Mon, 2 Feb 2009 05:55:46 +0000 (16:55 +1100)]
drm: Split drm_map and drm_local_map

Once upon a time, the DRM made the distinction between the drm_map
data structure exchanged with user space and the drm_local_map used
in the kernel.

For some reasons, while the BSD port still has that "feature", the
linux part abused drm_map for kernel internal usage as the local
map only existed as a typedef of the struct drm_map.

This patch fixes it by declaring struct drm_local_map separately
(though its content is currently identical to the userspace variant),
and changing the kernel code to only use that, except when it's a
user<->kernel interface (ie. ioctl).

This allows subsequent changes to the in-kernel format

I've also replaced the use of drm_local_map_t with struct drm_local_map
in a couple of places. Mostly by accident but they are the same (the
former is a typedef of the later) and I have some remote plans and
half finished patch to completely kill the drm_local_map_t typedef
so I left those bits in.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agodrm: Use resource_size_t for drm_get_resource_{start, len}
Benjamin Herrenschmidt [Mon, 2 Feb 2009 05:55:45 +0000 (16:55 +1100)]
drm: Use resource_size_t for drm_get_resource_{start, len}

The DRM uses its own wrappers to obtain resources from PCI devices,
which currently convert the resource_size_t into an unsigned long.

This is broken on 32-bit platforms with >32-bit physical address
space.

This fixes them, along with a few occurences of unsigned long used
to store such a resource in drivers.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Dave Airlie <airlied@linux.ie>
15 years agoLinus 2.6.29-rc8
Linus Torvalds [Fri, 13 Mar 2009 02:39:28 +0000 (19:39 -0700)]
Linus 2.6.29-rc8

15 years agobitmap: fix end condition in bitmap_find_free_region
Linus Torvalds [Fri, 13 Mar 2009 02:32:51 +0000 (19:32 -0700)]
bitmap: fix end condition in bitmap_find_free_region

Guennadi Liakhovetski noticed that the end condition for the loop in
bitmap_find_free_region() is wrong, and the "return if error" was also
using the wrong conditional that would only trigger if the bitmap was an
exact multiple of the allocation size, which is not necessarily the case
with dma_alloc_from_coherent().

Such a failure would end up in bitmap_find_free_region() accessing
beyond the end of the bitmap.

Reported-by: Guennadi Liakhovetski <lg@denx.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-fixes
Linus Torvalds [Thu, 12 Mar 2009 23:35:26 +0000 (16:35 -0700)]
Merge git://git./linux/kernel/git/sam/kbuild-fixes

* git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-fixes:
  kbuild: remove unused -r option for module-init-tool depmod
  kbuild: fix 'make rpm' when CONFIG_LOCALVERSION_AUTO=y and using SCM tree
  kbuild: fix mkspec to cleanup RPM_BUILD_ROOT
  kbuild: fix C libary confusion in unifdef.c due to getline()

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
Linus Torvalds [Thu, 12 Mar 2009 23:34:59 +0000 (16:34 -0700)]
Merge git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  cpumask: mm_cpumask for accessing the struct mm_struct's cpu_vm_mask.
  cpumask: tsk_cpumask for accessing the struct task_struct's cpus_allowed.

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pkl/squashfs-linus
Linus Torvalds [Thu, 12 Mar 2009 23:32:36 +0000 (16:32 -0700)]
Merge git://git./linux/kernel/git/pkl/squashfs-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/pkl/squashfs-linus:
  Squashfs: Valid filesystems are flagged as bad by the corrupted fs patch

15 years agoMerge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6
Linus Torvalds [Thu, 12 Mar 2009 23:25:04 +0000 (16:25 -0700)]
Merge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6

* 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6:
  hwmon: (f75375s) Remove unnecessary and confusing initialization
  hwmon: (it87) Properly decode -128 degrees C temperature
  hwmon: (lm90) Document support for the MAX6648/6692 chips
  hwmon: (abituguru3) Fix I/O error handling

15 years agotrivial: fix bad links in the ext2 and ext3 documentation
Jody McIntyre [Thu, 12 Mar 2009 21:39:23 +0000 (17:39 -0400)]
trivial: fix bad links in the ext2 and ext3 documentation

Trivial patch to fix bad links in the ext2 and ext3 documentation.

Signed-off-by: Jody McIntyre <scjody@sun.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'fixes-20090312' of git://git.kernel.org/pub/scm/linux/kernel/git/willy/pci
Linus Torvalds [Thu, 12 Mar 2009 23:22:51 +0000 (16:22 -0700)]
Merge branch 'fixes-20090312' of git://git./linux/kernel/git/willy/pci

* 'fixes-20090312' of git://git.kernel.org/pub/scm/linux/kernel/git/willy/pci:
  PCIe: portdrv: call pci_disable_device during remove
  pci: Fix typo in message while disabling HT MSI mapping
  pci: don't disable too many HT MSI mapping
  powerpc/pseries: The RPA PCI hotplug driver depends on EEH
  PCIe: AER: during disable, check subordinate before walking
  PCI: Add PCI quirk to disable L0s ASPM state for 82575 and 82598

15 years agoRDMA/nes: Don't allow userspace QPs to use STag zero
Faisal Latif [Thu, 12 Mar 2009 21:34:59 +0000 (14:34 -0700)]
RDMA/nes: Don't allow userspace QPs to use STag zero

STag zero is a special STag that allows consumers to access any bus
address without registering memory.  The nes driver unfortunately
allows STag zero to be used even with QPs created by unprivileged
userspace consumers, which means that any process with direct verbs
access to the nes device can read and write any memory accessible to
the underlying PCI device (usually any memory in the system).  Such
access is usually given for cluster software such as MPI to use, so
this is a local privilege escalation bug on most systems running this
driver.

The driver was using STag zero to receive the last streaming mode
data; to allow STag zero to be disabled for unprivileged QPs, the
driver now registers a special MR for this data.

Cc: <stable@kernel.org>
Signed-off-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofs: new inode i_state corruption fix
Nick Piggin [Thu, 12 Mar 2009 21:31:38 +0000 (14:31 -0700)]
fs: new inode i_state corruption fix

There was a report of a data corruption
http://lkml.org/lkml/2008/11/14/121.  There is a script included to
reproduce the problem.

During testing, I encountered a number of strange things with ext3, so I
tried ext2 to attempt to reduce complexity of the problem.  I found that
fsstress would quickly hang in wait_on_inode, waiting for I_LOCK to be
cleared, even though instrumentation showed that unlock_new_inode had
already been called for that inode.  This points to memory scribble, or
synchronisation problme.

i_state of I_NEW inodes is not protected by inode_lock because other
processes are not supposed to touch them until I_LOCK (and I_NEW) is
cleared.  Adding WARN_ON(inode->i_state & I_NEW) to sites where we modify
i_state revealed that generic_sync_sb_inodes is picking up new inodes from
the inode lists and passing them to __writeback_single_inode without
waiting for I_NEW.  Subsequently modifying i_state causes corruption.  In
my case it would look like this:

CPU0                            CPU1
unlock_new_inode()              __sync_single_inode()
 reg <- inode->i_state
 reg -> reg & ~(I_LOCK|I_NEW)   reg <- inode->i_state
 reg -> inode->i_state          reg -> reg | I_SYNC
                                reg -> inode->i_state

Non-atomic RMW on CPU1 overwrites CPU0 store and sets I_LOCK|I_NEW again.

Fix for this is rather than wait for I_NEW inodes, just skip over them:
inodes concurrently being created are not subject to data integrity
operations, and should not significantly contribute to dirty memory
either.

After this change, I'm unable to reproduce any of the added warnings or
hangs after ~1hour of running.  Previously, the new warnings would start
immediately and hang would happen in under 5 minutes.

I'm also testing on ext3 now, and so far no problems there either.  I
don't know whether this fixes the problem reported above, but it fixes a
real problem for me.

Cc: "Jorge Boncompte [DTI2]" <jorge@dti2.net>
Reported-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
Cc: Jan Kara <jack@suse.cz>
Cc: <stable@kernel.org>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomemcg: use correct scan number at reclaim
KOSAKI Motohiro [Thu, 12 Mar 2009 21:31:36 +0000 (14:31 -0700)]
memcg: use correct scan number at reclaim

Even when page reclaim is under mem_cgroup, # of scan page is determined by
status of global LRU. Fix that.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomfd: add support for WM8351 revision B
Mark Brown [Thu, 12 Mar 2009 21:31:36 +0000 (14:31 -0700)]
mfd: add support for WM8351 revision B

No software visible difference from revision A.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: Samuel Ortiz <sameo@openedhand.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoacer-wmi: fix regression in backlight detection
Michael Spang [Thu, 12 Mar 2009 21:31:34 +0000 (14:31 -0700)]
acer-wmi: fix regression in backlight detection

Currently we disable the Acer WMI backlight device if there is no ACPI
backlight device.  As a result, we end up with no backlight device at all.
 We should instead disable it if there is an ACPI device, as the other
laptop drivers do.  This regression was introduced in febf2d9 ("Acer-WMI:
fingers off backlight if video.ko is serving this functionality").

Each laptop driver with backlight support got a similar change around
febf2d9.  The changes to the other drivers look correct; see e.g.
a598c82f for a similar but correct change.  The regression is also in
2.6.28.

Signed-off-by: Michael Spang <mspang@csclub.uwaterloo.ca>
Acked-by: Thomas Renninger <trenn@suse.de>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Carlos Corbacho <carlos@strangeworlds.co.uk>
Cc: Len Brown <len.brown@intel.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: <stable@kernel.org> [2.6.28.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agommc: s3cmci: fix s3c2410_dma_config() arguments.
Ben Dooks [Thu, 12 Mar 2009 21:31:33 +0000 (14:31 -0700)]
mmc: s3cmci: fix s3c2410_dma_config() arguments.

The s3cmci driver is calling s3c2410_dma_config with incorrect data for
the DCON register.  The S3C2410_DCON_HWTRIG is implicit in the channel
configuration and the device selection of S3C2410_DCON_CH0_SDI is
incorrect as the DMA system may not select channel 0.

Signed-off-by: Ben Dooks <ben@simtec.co.uk>
Acked-by: Pierre Ossman <drzeus@drzeus.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMAINTAINERS: downgrade support for man-pages
Michael Kerrisk [Thu, 12 Mar 2009 21:31:32 +0000 (14:31 -0700)]
MAINTAINERS: downgrade support for man-pages

Unfortunately, Linux Foundation funding for my work on
man-pages/testing/doc under the auspices of the LF documentation
fellowship unfortunately ran out a short while ago (after earlier attempts
to seek funding, only Google stepped forward with a bit of further funding
for the position), so the patch below acknowledges something closer to
reality.

Unfortunately, there will (probably very) soon be a further downgrade from
"Maintained" to "Odd Fixes" or "Orphan", unless some funding miracle
occurs.  So, if anyone is looking to become man-pages maintainer, there
may soon be an opening (okay, don't trample me in the rush ;-).)

Signed-off-by: Michael Kerrisk <mtk.manpages@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agods2760_battery.c: fix division by zero
Daniel Mack [Thu, 12 Mar 2009 21:31:30 +0000 (14:31 -0700)]
ds2760_battery.c: fix division by zero

The 'battery remaining capacity' calculation in
drivers/power/ds2760_battery.c lacks a parameter check to a division
operation which causes the kernel to oops on my board.

[   21.233750] Division by zero in kernel.
[   21.237646] [<c002955c>] (__div0+0x0/0x20) from [<c012561c>] (Ldiv0+0x8/0x10)
[   21.244816] [<c01bef34>] (ds2760_battery_read_status+0x0/0x2a4) from [<c01bf3a4>] (ds2760_battery_get_property+0x30/0xdc)
[   21.255803]  r8:c03a22c0 r7:c7886100 r6:00000009 r5:c782fe7c r4:c7886084
[   21.262518] [<c01bf374>] (ds2760_battery_get_property+0x0/0xdc) from [<c01bde98>] (power_supply_show_property+0x48/0x114)
[   21.273480]  r6:c7996000 r5:00000009 r4:00000000
[   21.278111] [<c01bde50>] (power_supply_show_property+0x0/0x114) from [<c01be158>] (power_supply_uevent+0x188/0x280)
[   21.288537]  r8:00000001 r7:c7886100 r6:c7996000 r5:000000b4 r4:00000000
[   21.295222] [<c01bdfd0>] (power_supply_uevent+0x0/0x280) from [<c015c664>] (dev_uevent+0xd4/0x10c)
[   21.304199] [<c015c590>] (dev_uevent+0x0/0x10c) from [<c0128440>] (kobject_uevent_env+0x180/0x390)
[   21.313170]  r5:00000000 r4:c78860ac
[   21.316725] [<c01282c0>] (kobject_uevent_env+0x0/0x390) from [<c0128664>] (kobject_uevent+0x14/0x18)
[   21.325850] [<c0128650>] (kobject_uevent+0x0/0x18) from [<c01bdc34>] (power_supply_changed_work+0x5c/0x70)
[   21.335506] [<c01bdbd8>] (power_supply_changed_work+0x0/0x70) from [<c004d290>] (run_workqueue+0xbc/0x144)
[   21.345167]  r4:c7812040
[   21.347716] [<c004d1d4>] (run_workqueue+0x0/0x144) from [<c004d94c>] (worker_thread+0xa8/0xbc)
[   21.356296]  r7:c7812040 r6:c7820b00 r5:c782ffa4 r4:c7812048
[   21.361957] [<c004d8a4>] (worker_thread+0x0/0xbc) from [<c0051008>] (kthread+0x5c/0x94)
[   21.369971]  r7:00000000 r6:c004d8a4 r5:c7812040 r4:c782e000
[   21.375612] [<c0050fac>] (kthread+0x0/0x94) from [<c00403d0>] (do_exit+0x0/0x688)

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Cc: Szabolcs Gyurko <szabolcs.gyurko@tlt.hu>
Acked-by: Matt Reimer <mreimer@vpop.net>
Acked-by: Anton Vorontsov <cbou@mail.ru>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovfs: add missing unlock in sget()
Li Zefan [Thu, 12 Mar 2009 21:31:29 +0000 (14:31 -0700)]
vfs: add missing unlock in sget()

In sget(), destroy_super(s) is called with s->s_umount held, which makes
lockdep unhappy.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopipe_rdwr_fasync: fix the error handling to prevent the leak/crash
Oleg Nesterov [Thu, 12 Mar 2009 21:31:28 +0000 (14:31 -0700)]
pipe_rdwr_fasync: fix the error handling to prevent the leak/crash

If the second fasync_helper() fails, pipe_rdwr_fasync() returns the error
but leaves the file on ->fasync_readers.

This was always wrong, but since 233e70f4228e78eb2f80dc6650f65d3ae3dbf17c
"saner FASYNC handling on file close" we have the new problem.  Because in
this case setfl() doesn't set FASYNC bit, __fput() will not do
->fasync(0), and we leak fasync_struct with ->fa_file pointing to the
freed file.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/w1/masters/w1-gpio.c: fix read_bit()
Daniel Mack [Thu, 12 Mar 2009 21:31:25 +0000 (14:31 -0700)]
drivers/w1/masters/w1-gpio.c: fix read_bit()

W1 master implementations are expected to return 0 or 1 from their
read_bit() function.  However, not all platforms do return these values
from gpio_get_value() - namely PXAs won't.  Hence the w1 gpio-master needs
to break the result down to 0 or 1 itself.

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Cc: Ville Syrjala <syrjala@sci.fi>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouml: fix WARNING: vmlinux: 'memcpy' exported twice
akpm@linux-foundation.org [Thu, 12 Mar 2009 21:31:24 +0000 (14:31 -0700)]
uml: fix WARNING: vmlinux: 'memcpy' exported twice

Fix the following warning on x86_64:

LD vmlinux.o
MODPOST vmlinux.o
WARNING: vmlinux: 'memcpy' exported twice. Previous export was in vmlinux

For x86_64, this symbol is already exported from arch/um/sys-x86_64/ksyms.c.

Reported-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Tested-by: Boaz Harrosh <bharrosh@panasas.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoUML on UML fixed: it did not start
Renzo Davoli [Thu, 12 Mar 2009 21:31:23 +0000 (14:31 -0700)]
UML on UML fixed: it did not start

It is currently impossible to run a user-mode linux machine inside another
user-mode linux (UML on UML).  It breaks after a few instructions.  When
it tries to check whether SYSEMU is installed (the inner) UML receives an
inconsistent result (from the outer UML).

This is the output of a broken attempt:
$ ./linux mem=256m ubd0=cow
Locating the bottom of the address space ... 0x0
Locating the top of the address space ... 0xc0000000
Core dump limits :
        soft - 0
        hard - NONE
Checking that ptrace can change system call numbers...OK
Checking ptrace new tags for syscall emulation...unsupported
Checking syscall emulation patch for ptrace...check_sysemu : expected SIGTRAP, got status = 256
$

The problem is the following:

PTRACE_SYSCALL/SINGLESTEP is currently managed inside arch_ptrace for ARCH=um.

PTRACE_SYSEMU/SUSEMU_SINGLESTEP is not captured in arch_ptrace's switch,
therefore it is erroneously passed back to ptrace_request (in
kernel/ptrace).

This simple patch simply forces ptrace to return an error on
PTRACE_SYSEMU/SUSEMU_SINGLESTEP as it is unsupported on ARCH=um, and fixes
the problem.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Renzo Davoli <renzo@cs.unibo.it>
Reviewed-by: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoPCIe: portdrv: call pci_disable_device during remove
Alex Chiang [Sun, 8 Mar 2009 02:35:47 +0000 (19:35 -0700)]
PCIe: portdrv: call pci_disable_device during remove

The PCIe port driver calls pci_enable_device() during probe but
never calls pci_disable_device() during remove.

Cc: stable@kernel.org
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agopci: Fix typo in message while disabling HT MSI mapping
Prakash Punnoor [Fri, 6 Mar 2009 09:10:35 +0000 (10:10 +0100)]
pci: Fix typo in message while disabling HT MSI mapping

"Enabling" should read "Disabling"

Signed-off-by: Prakash Punnoor <prakash@punnoor.de>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agopci: don't disable too many HT MSI mapping
Prakash Punnoor [Thu, 5 Mar 2009 23:45:12 +0000 (00:45 +0100)]
pci: don't disable too many HT MSI mapping

Prakash's system needs MSI disabled on some bridges, but not all.
This seems to be the minimal fix for 2.6.29, but should be replaced
during 2.6.30.

Signed-off-by: Prakash Punnoor <prakash@punnoor.de>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agopowerpc/pseries: The RPA PCI hotplug driver depends on EEH
Michael Ellerman [Fri, 6 Mar 2009 03:39:14 +0000 (14:39 +1100)]
powerpc/pseries: The RPA PCI hotplug driver depends on EEH

The RPA PCI hotplug driver calls EEH routines, so should depend on
EEH. Also PPC_PSERIES implies PPC64, so remove that.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agoPCIe: AER: during disable, check subordinate before walking
Alex Chiang [Fri, 6 Mar 2009 02:28:40 +0000 (19:28 -0700)]
PCIe: AER: during disable, check subordinate before walking

Commit 47a8b0cc (Enable PCIe AER only after checking firmware
support) wants to walk the PCI bus in the remove path to disable
AER, and calls pci_walk_bus for downstream bridges.

Unfortunately, in the remove path, we remove devices and bridges
in a depth-first manner, starting with the furthest downstream
bridge and working our way backwards.

The furthest downstream bridges will not have a dev->subordinate,
and we hit a NULL deref in pci_walk_bus.

Check for dev->subordinate first before attempting to walk the
PCI hierarchy below us.

Acked-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agoPCI: Add PCI quirk to disable L0s ASPM state for 82575 and 82598
Alexander Duyck [Thu, 5 Mar 2009 18:57:28 +0000 (13:57 -0500)]
PCI: Add PCI quirk to disable L0s ASPM state for 82575 and 82598

This patch is intended to disable L0s ASPM link state for 82598 (ixgbe)
parts due to the fact that it is possible to corrupt TX data when coming
back out of L0s on some systems.  The workaround had been added for 82575
(igb) previously, but did not use the ASPM api.  This quirk uses the ASPM
api to prevent the ASPM subsystem from re-enabling the L0s state.

Instead of adding the fix in igb to the ixgbe driver as well it was
decided to move it into a pci quirk.  It is necessary to move the fix out
of the driver and into a pci quirk in order to prevent the issue from
occuring prior to driver load to handle the possibility of the device being
passed to a VM via direct assignment.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Thu, 12 Mar 2009 16:27:53 +0000 (09:27 -0700)]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sunhme: Fix qfe parent detection.
  sparc64: Fix lost interrupts on sun4u.
  sparc64: wait_event_interruptible_timeout may return -ERESTARTSYS
  jsflash: stop defining MAJOR_NR

15 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Thu, 12 Mar 2009 16:25:10 +0000 (09:25 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  MIPS: IP27: Enable RAID5 module
  MIPS: TXx9: update defconfigs
  MIPS: NEC VR5500 processor support fixup
  MIPS: Fix build of non-CONFIG_SYSVIPC version of sys_32_ipc

15 years agohwmon: (f75375s) Remove unnecessary and confusing initialization
Andrew Klossner [Thu, 12 Mar 2009 12:36:39 +0000 (13:36 +0100)]
hwmon: (f75375s) Remove unnecessary and confusing initialization

f75375_probe calls i2c_get_clientdata to initialize the data pointer,
but there isn't yet any client data to get, and the value is never
used before the variable is assigned a new value seven lines later.

The call doesn't hurt anything and wastes only a couple of cycles.
The reason to fix it is because this module serves as an example to
hackers writing new hwmon drivers, and this part of the example is
confusing.

Signed-off-by: Andrew Klossner <andrew@cesa.opbu.xerox.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
15 years agohwmon: (it87) Properly decode -128 degrees C temperature
Jean Delvare [Thu, 12 Mar 2009 12:36:39 +0000 (13:36 +0100)]
hwmon: (it87) Properly decode -128 degrees C temperature

The it87 driver is reporting -128 degrees C as +128 degrees C.
That's not a terribly likely temperature value but let's still
get it right, especially when it simplifies the code.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
15 years agohwmon: (lm90) Document support for the MAX6648/6692 chips
Darrick J. Wong [Thu, 12 Mar 2009 12:36:38 +0000 (13:36 +0100)]
hwmon: (lm90) Document support for the MAX6648/6692 chips

Update documentation to prevent further confusion/duplication.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
15 years agohwmon: (abituguru3) Fix I/O error handling
Jean Delvare [Thu, 12 Mar 2009 12:36:38 +0000 (13:36 +0100)]
hwmon: (abituguru3) Fix I/O error handling

Fix a logic bug reported by Roel Kluin, by rewriting the error
handling code in a clearer way.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Alistair John Strachan <alistair@devzero.co.uk>
Acked-by: Hans de Goede <hdegoede@redhat.com>
15 years agocpumask: mm_cpumask for accessing the struct mm_struct's cpu_vm_mask.
Rusty Russell [Thu, 12 Mar 2009 20:35:44 +0000 (14:35 -0600)]
cpumask: mm_cpumask for accessing the struct mm_struct's cpu_vm_mask.

This allows us to change the representation (to a dangling bitmap or
cpumask_var_t) without breaking all the callers: they can use
mm_cpumask() now and won't see a difference as the changes roll into
linux-next.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
15 years agocpumask: tsk_cpumask for accessing the struct task_struct's cpus_allowed.
Rusty Russell [Thu, 12 Mar 2009 20:35:43 +0000 (14:35 -0600)]
cpumask: tsk_cpumask for accessing the struct task_struct's cpus_allowed.

This allows us to change the representation (to a dangling bitmap or
cpumask_var_t) without breaking all the callers: they can use
tsk_cpumask() now and won't see a difference as the changes roll into
linux-next.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
15 years agoSquashfs: Valid filesystems are flagged as bad by the corrupted fs patch
Phillip Lougher [Thu, 12 Mar 2009 03:23:48 +0000 (03:23 +0000)]
Squashfs: Valid filesystems are flagged as bad by the corrupted fs patch

The corrupted filesystem patch added a check against zlib trying to
output too much data in the presence of data corruption.  This check
triggered if zlib_inflate asked to be called again (Z_OK) with
avail_out == 0 and no more output buffers available.  This check proves
to be rather dumb, as it incorrectly catches the case where zlib has
generated all the output, but there are still input bytes to be processed.

This patch does a number of things.  It removes the original check and
replaces it with code to not move to the next output buffer if there
are no more output buffers available, relying on zlib to error if it
wants an extra output buffer in the case of data corruption.  It
also replaces the Z_NO_FLUSH flag with the more correct Z_SYNC_FLUSH
flag, and makes the error messages more understandable to
non-technical users.

Signed-off-by: Phillip Lougher <phillip@lougher.demon.co.uk>
Reported-by: Stefan Lippers-Hollmann <s.L-H@gmx.de>
15 years agoMerge branch 'for-linus' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Wed, 11 Mar 2009 21:29:03 +0000 (14:29 -0700)]
Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: only issues a cache flush on unmount if barriers are enabled
  xfs: prevent lockdep false positive in xfs_iget_cache_miss
  xfs: prevent kernel crash due to corrupted inode log format

15 years agoMIPS: IP27: Enable RAID5 module
Ralf Baechle [Wed, 11 Mar 2009 20:08:50 +0000 (21:08 +0100)]
MIPS: IP27: Enable RAID5 module

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: TXx9: update defconfigs
Atsushi Nemoto [Wed, 4 Mar 2009 14:45:44 +0000 (23:45 +0900)]
MIPS: TXx9: update defconfigs

Enable following features:
* MTD (PHYSMAP)
* LED (LEDS_GPIO)
* RBTX4939
* 7SEGLED
* IDE (IDE_TX4938, IDE_TX4939)
* SMC91X
* RTC_DRV_TX4939

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: NEC VR5500 processor support fixup
Shinya Kuribayashi [Tue, 3 Mar 2009 09:05:51 +0000 (18:05 +0900)]
MIPS: NEC VR5500 processor support fixup

Current VR5500 processor support lacks of some functions which are
expected to be configured/synthesized on arch initialization.

Here're some VR5500A spec notes:

* All execution hazards are handled in hardware.

* Once VR5500A stops the operation of the pipeline by WAIT instruction,
  it could return from the standby mode only when either a reset, NMI
  request, or all enabled interrupts is/are detected.  In other words,
  if interrupts are disabled by Status.IE=0, it keeps in standby mode
  even when interrupts are internally asserted.

  Notes on WAIT: The operation of the processor is undefined if WAIT
  insn is in the branch delay slot.  The operation is also undefined
  if WAIT insn is executed when Status.EXL and Status.ERL are set to 1.

* VR5500A core only implements the Load prefetch.

With these changes, it boots fine.

Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi@necel.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMIPS: Fix build of non-CONFIG_SYSVIPC version of sys_32_ipc
Xiaotian Feng [Mon, 9 Mar 2009 01:45:12 +0000 (09:45 +0800)]
MIPS: Fix build of non-CONFIG_SYSVIPC version of sys_32_ipc

Signed-off-by: Xiaotian Feng <xiaotian.feng@windriver.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
15 years agoMerge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Wed, 11 Mar 2009 19:14:55 +0000 (12:14 -0700)]
Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm: fix EDID parser problem with positive/negative hsync/vsync

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 11 Mar 2009 19:14:04 +0000 (12:14 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  radeonfb/aty128fb: Disable broken early resume hook for PowerBooks
  hvc_console: Remove tty->low_latency on pseries backends
  powerpc: fix linkstation and storcenter compilation breakage
  powerpc/4xx: Enable SERIAL_OF support by default for Virtex platforms

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt...
Linus Torvalds [Wed, 11 Mar 2009 19:09:45 +0000 (12:09 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/anholt/drm-intel

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel:
  drm/i915: fix 945 fence register writes for fence 8 and above.
  drm/i915: Protect active fences on i915
  drm/i915: Check to see if we've pinned all available fences
  drm/i915: Check fence status on every pin.
  drm/i915: First recheck for an empty fence register.
  drm/i915: Fix bad \n in MTRR failure notice.
  drm/i915: Don't restore palettes through VGA registers.
  i915: add newline to i915_gem_object_pin failure msg
  drm: Return EINVAL on duplicate objects in execbuffer object list

15 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 11 Mar 2009 19:04:51 +0000 (12:04 -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: work around Fedora-11 x86-32 kernel failures on Intel Atom CPUs

15 years agoFix _fat_bmap() locking
OGAWA Hirofumi [Wed, 11 Mar 2009 17:03:23 +0000 (02:03 +0900)]
Fix _fat_bmap() locking

On swapon() path, it has already i_mutex. So, this uses i_alloc_sem
instead of it.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Reported-by: Laurent GUERBY <laurent@guerby.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrm/i915: fix 945 fence register writes for fence 8 and above.
Eric Anholt [Wed, 11 Mar 2009 05:34:49 +0000 (22:34 -0700)]
drm/i915: fix 945 fence register writes for fence 8 and above.

The last 8 fence registers sit at a different offset, so when we went to set
fence number 8 in the lower offset, we instead set PGETBL_CTL, and the GPU
got all sorts of angry at us.

fd.o bug #20567.  Easily reproducible by running glxgears and killing it about
6 times.

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Protect active fences on i915
Chris Wilson [Wed, 11 Feb 2009 14:26:47 +0000 (14:26 +0000)]
drm/i915: Protect active fences on i915

The i915 also uses the fence registers for GPU access to tiled buffers so
we cannot reallocate one whilst it is on the active list. By performing a
LRU scan of the fenced buffers we also avoid waiting the possibility of
waiting on a pinned, or otherwise unusable, buffer.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agox86: work around Fedora-11 x86-32 kernel failures on Intel Atom CPUs
Ingo Molnar [Tue, 10 Mar 2009 21:31:03 +0000 (22:31 +0100)]
x86: work around Fedora-11 x86-32 kernel failures on Intel Atom CPUs

Impact: work around boot crash

Work around Intel Atom erratum AAH41 (probabilistically) - it's triggering
in the field.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Tested-by: Kyle McMartin <kyle@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoproc: fix kflags to uflags copying in /proc/kpageflags
Wu Fengguang [Wed, 11 Mar 2009 01:00:04 +0000 (09:00 +0800)]
proc: fix kflags to uflags copying in /proc/kpageflags

Fix kpf_copy_bit(src,dst) to be kpf_copy_bit(dst,src) to match the
actual call patterns, e.g. kpf_copy_bit(kflags, KPF_LOCKED, PG_locked).

This misplacement of src/dst only affected reporting of PG_writeback,
PG_reclaim and PG_buddy. For others kflags==uflags so not affected.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrm/i915: Check to see if we've pinned all available fences
Chris Wilson [Wed, 11 Feb 2009 14:26:46 +0000 (14:26 +0000)]
drm/i915: Check to see if we've pinned all available fences

We need to check and report if there are no available fences - or else we
spin endlessly waiting for a buffer to magically unpin itself.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Check fence status on every pin.
Chris Wilson [Wed, 11 Feb 2009 14:26:45 +0000 (14:26 +0000)]
drm/i915: Check fence status on every pin.

As we may steal the fence register of an unpinned buffer for another,
every time we repin the buffer we need to recheck whether it needs to be
allocated a fence.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: First recheck for an empty fence register.
Chris Wilson [Wed, 11 Feb 2009 14:26:44 +0000 (14:26 +0000)]
drm/i915: First recheck for an empty fence register.

If we wait upon a request and successfully unbind a buffer occupying a
fence register, then that slot will be freed and cause a NULL derefrence
upon rescanning.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm: fix EDID parser problem with positive/negative hsync/vsync
Pantelis Koukousoulas [Tue, 10 Mar 2009 11:16:14 +0000 (13:16 +0200)]
drm: fix EDID parser problem with positive/negative hsync/vsync

Comparing the layouts of struct detail_pixel_timing with
x.org's struct detailed_timings and how those are handled,
it appears that the hsync_positive and vsync_positive
fields are backwards.

This patch fixes https://bugs.freedesktop.org/show_bug.cgi?id=20019
for me. It was tested on 2 monitors, LG FLATRON L225WS 22" and
a YAKUMO 17" for which more details are unknown.

Signed-off-by: Pantelis Koukousoulas <pktoss@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
15 years agoradeonfb/aty128fb: Disable broken early resume hook for PowerBooks
Benjamin Herrenschmidt [Tue, 10 Mar 2009 23:45:17 +0000 (10:45 +1100)]
radeonfb/aty128fb: Disable broken early resume hook for PowerBooks

radeonfb and aty128fb have a special hook called by the PowerMac platform
code very very early on resume from sleep to bring the screen back. This
is useful for debugging wakup problems, but unfortunately, this also became
a source of problems of its own.

The hook is called extremely early, with interrupts still off, and the code
path involved with that code nowadays rely on things like taking mutexes,
GFP_KERNEL allocations, etc...

In addition, the driver now relies on the PCI core to restore the standard
config space before calling resume which doesn't happen with this early
code path.

I'm keeping the code in but commented out along with a fixup call to
pci_restore_state(). The reason is that I still want to make it easy to
re-enable temporarily to track wake up problems, and it's possible that
I can revive it at some stage if we make sleeping things save to call
in early resume using a system state.

In the meantime, this should fix several reported regressions.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agohvc_console: Remove tty->low_latency on pseries backends
Benjamin Herrenschmidt [Mon, 9 Mar 2009 14:36:15 +0000 (14:36 +0000)]
hvc_console: Remove tty->low_latency on pseries backends

The hvcs and hvsi backends both set tty->low_latency to one, along
with more or less scary comments regarding bugs or races that would
happen if not doing so.

However, they also both call tty_flip_buffer_push() in conexts where
it's illegal to do so since some recent tty changes (or at least it
may have been illegal always but it nows blows) when low_latency is
set (ie, hard interrupt or with spinlock held and irqs disabled).

This removes the setting for now to get them back to working condition,
we'll have to address the races described in the comments separately
if they are still an issue (some of this might have been fixed already).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoMerge commit 'gcl/merge' into merge
Benjamin Herrenschmidt [Tue, 10 Mar 2009 23:40:29 +0000 (10:40 +1100)]
Merge commit 'gcl/merge' into merge

15 years agoi810: fix kernel crash fix when struct fb_var_screeninfo is supplied
Samuel CUELLA [Tue, 10 Mar 2009 19:56:00 +0000 (12:56 -0700)]
i810: fix kernel crash fix when struct fb_var_screeninfo is supplied

Prevent the kernel from being crashed by a divide-by-zero operation when
supplied an incorrectly filled 'struct fb_var_screeninfo' from userland.

Previously i810_main.c:1005 (i810_check_params) was using the global
'yres' symbol previously defined at i810_main.c:145 as a module parameter
value holder (i810_main.c:2174).  If i810fb is compiled-in or if this
param doesn't get a default value, this direct usage leads to a
divide-by-zero at i810_main.c:1005 (i810_check_params).  The patch simply
replace the 'yres' global, perhaps undefined symbol usage by a given
parameter structure lookup.

This problem occurs with directfb, mplayer -vo fbdev, SDL library.
It was also reported ( but non solved ) at:

http://mail.directfb.org/pipermail/directfb-dev/2008-March/004050.html

Signed-off-by: Samuel CUELLA <samuel.cuella@supinfo.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agom68knommu: m528x build fix
Steven King [Tue, 10 Mar 2009 19:55:58 +0000 (12:55 -0700)]
m68knommu: m528x build fix

There isn't any mcfqspi.h in the tree, and without it everything inside the
#ifdef CONFIG_SPI is uncompilable.

Signed-off-by: Steven King <sfking@fdwdc.com>
Acked-by: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agom68knommu: m5206e build fix
Steven King [Tue, 10 Mar 2009 19:55:57 +0000 (12:55 -0700)]
m68knommu: m5206e build fix

Signed-off-by: Steven King <sfking@fdwdc.com>
Acked-by: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agorcu: documentation 1Q09 update
Paul E. McKenney [Tue, 10 Mar 2009 19:55:57 +0000 (12:55 -0700)]
rcu: documentation 1Q09 update

Update the RCU documentation to call out the need for callers of
primitives like call_rcu() and synchronize_rcu() to prevent subsequent RCU
readers from hazard.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agokernel/user.c: fix a memory leak when freeing up non-init usernamespaces users
Dhaval Giani [Tue, 10 Mar 2009 19:55:56 +0000 (12:55 -0700)]
kernel/user.c: fix a memory leak when freeing up non-init usernamespaces users

We were returning early in the sysfs directory cleanup function if the
user belonged to a non init usernamespace.  Due to this a lot of the
cleanup was not done and we were left with a leak.  Fix the leak.

Reported-by: Serge Hallyn <serue@linux.vnet.ibm.com>
Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Tested-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomtd: physmap: fix NULL pointer dereference in error path
Atsushi Nemoto [Tue, 10 Mar 2009 19:55:55 +0000 (12:55 -0700)]
mtd: physmap: fix NULL pointer dereference in error path

commit e480814f138cd5d78a8efe397756ba6b6518fdb6 ("[MTD] [MAPS] physmap:
fix wrong free and del_mtd_{partition,device}") introduces a NULL pointer
dereference in physmap_flash_remove when called from the error path in
physmap_flash_probe (if map_probe failed).

Call del_mtd_{partition,device} only if info->cmtd was not NULL.

Reported-by: pHilipp Zabel <philipp.zabel@gmail.com>
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agointel-agp: fix a panic with 1M of shared memory, no GTT entries
Lubomir Rintel [Tue, 10 Mar 2009 19:55:54 +0000 (12:55 -0700)]
intel-agp: fix a panic with 1M of shared memory, no GTT entries

When GTT size is equal to amount of video memory, the amount of GTT
entries is computed lower than zero, which is invalid and leads to
off-by-one error in intel_i915_configure()

Originally posted here:
http://bugzilla.kernel.org/show_bug.cgi?id=12539
http://bugzilla.redhat.com/show_bug.cgi?id=445592

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Cc: Lubomir Rintel <lkundrak@v3.sk>
Cc: Dave Airlie <airlied@linux.ie>
Reviewed-by: Eric Anholt <eric@anholt.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomtd_dataflash: fix probing of AT45DB321C chips.
Will Newton [Tue, 10 Mar 2009 19:55:53 +0000 (12:55 -0700)]
mtd_dataflash: fix probing of AT45DB321C chips.

Commit 771999b65f79264acde4b855e5d35696eca5e80c ("[MTD] DataFlash: bugfix,
binary page sizes now handled") broke support for probing AT45DB321C flash
chips.  These chips do not support the "page size" status bit, so if we
match the JEDEC id return early.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Will Newton <will.newton@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>