GitHub/LineageOS/android_kernel_motorola_exynos9610.git
16 years agoAMD IOMMU: use iommu_device_max_index, fix
Thomas Gleixner [Mon, 6 Oct 2008 12:15:24 +0000 (14:15 +0200)]
AMD IOMMU: use iommu_device_max_index, fix

include/linux/iommu-helper.h has no header guards, which breaks
sparc64 build. Add them.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: use iommu_device_max_index
FUJITA Tomonori [Sun, 28 Sep 2008 15:06:36 +0000 (00:06 +0900)]
AMD IOMMU: use iommu_device_max_index

AMD IOMMU can use iommu_device_max_index() instead of the homegrown
function.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: add PCI IDs for AMD Barcelona PCI devices
Ingo Molnar [Wed, 1 Oct 2008 06:59:40 +0000 (08:59 +0200)]
x86: add PCI IDs for AMD Barcelona PCI devices

Signed-off-by: Robert Richter <robert.richter@amd.com>
Cc: oprofile-list <oprofile-list@lists.sourceforge.net>
Cc: Barry Kasindorf <barry.kasindorf@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86/iommu: use __GFP_ZERO instead of memset for GART
Joerg Roedel [Thu, 25 Sep 2008 10:42:12 +0000 (12:42 +0200)]
x86/iommu: use __GFP_ZERO instead of memset for GART

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86/iommu: convert GART need_flush to bool
Joerg Roedel [Thu, 25 Sep 2008 10:13:54 +0000 (12:13 +0200)]
x86/iommu: convert GART need_flush to bool

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86/iommu: make GART driver checkpatch clean
Joerg Roedel [Thu, 25 Sep 2008 10:13:53 +0000 (12:13 +0200)]
x86/iommu: make GART driver checkpatch clean

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86 gart: remove unnecessary initialization
FUJITA Tomonori [Wed, 24 Sep 2008 13:41:10 +0000 (22:41 +0900)]
x86 gart: remove unnecessary initialization

There is no point to have such initialization in struct dma_mapping_ops.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: restore old GART alloc_coherent behavior
FUJITA Tomonori [Wed, 24 Sep 2008 11:48:37 +0000 (20:48 +0900)]
x86: restore old GART alloc_coherent behavior

Currently, GART alloc_coherent tries to allocate pages with GFP_DMA32
for a device having dma_masks > 24bit < 32bits. If GART gets an
address that a device can't access to, GART try to map the address to
a virtual I/O address that the device can access to.

But Andi pointed out, "The GART is somewhere in the 4GB range so you
cannot use it to map anything < 4GB. Also GART is pretty small."

http://lkml.org/lkml/2008/9/12/43

That is, it's possible that GART doesn't have virtual I/O address
space that a device can access to. The above behavior doesn't work for
a device having dma_masks > 24bit < 32bits.

This patch restores old GART alloc_coherent behavior (before the
alloc_coherent rewrite).

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agorevert "x86: make GART to respect device's dma_mask about virtual mappings"
FUJITA Tomonori [Wed, 24 Sep 2008 11:48:36 +0000 (20:48 +0900)]
revert "x86: make GART to respect device's dma_mask about virtual mappings"

This reverts:

commit bee44f294efd8417f5e68553778a6cc957af1547
Author: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date:   Fri Sep 12 19:42:35 2008 +0900

    x86: make GART to respect device's dma_mask about virtual mappings

I wrote the above commit to fix a GART alloc_coherent regression, that
can't handle a device having dma_masks > 24bit < 32bits, introduced by
the alloc_coherent rewrite:

  http://lkml.org/lkml/2008/8/12/200

After the alloc_coherent rewrite, GART alloc_coherent tried to
allocate pages with GFP_DMA32. If GART got an address that a device
can't access to, GART mapped the address to a virtual I/O address. But
GART mapping mechanism didn't take account of dma mask, so GART could
use a virtual I/O address that the device can't access to again.

Alan pointed out:

" This is indeed a specific problem found with things like older
  AACRAID where control blocks must be below 31bits and the GART
  is above 0x80000000. "

The above commit modified GART mapping mechanism to take care of dma
mask. But Andi pointed out, "The GART is somewhere in the 4GB range so
you cannot use it to map anything < 4GB. Also GART is pretty small."

http://lkml.org/lkml/2008/9/12/43

That means it's possible that GART doesn't have virtual I/O address
space that a device can access to. The above commit (to modify GART
mapping mechanism to take care of dma mask) can't fix the regression
reliably so let's avoid making GART more complicated.

We need a solution that always works for dma_masks > 24bit <
32bits. That's how GART worked before the alloc_coherent rewrite.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: export pci-nommu's alloc_coherent
FUJITA Tomonori [Wed, 24 Sep 2008 11:48:35 +0000 (20:48 +0900)]
x86: export pci-nommu's alloc_coherent

This patch exports nommu_alloc_coherent (renamed
dma_generic_alloc_coherent). GART needs this function.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoiommu: remove fullflush and nofullflush in IOMMU generic option
FUJITA Tomonori [Fri, 19 Sep 2008 16:23:30 +0000 (01:23 +0900)]
iommu: remove fullflush and nofullflush in IOMMU generic option

This patch against tip/x86/iommu virtually reverts
2842e5bf3115193f05dc9dac20f940e7abf44c1a. But just reverting the
commit breaks AMD IOMMU so this patch also includes some fixes.

The above commit adds new two options to x86 IOMMU generic kernel boot
options, fullflush and nofullflush. But such change that affects all
the IOMMUs needs more discussion (all IOMMU parties need the chance to
discuss it):

http://lkml.org/lkml/2008/9/19/106

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: remove set_bit_string()
FUJITA Tomonori [Mon, 22 Sep 2008 16:15:10 +0000 (01:15 +0900)]
x86: remove set_bit_string()

"export iommu_area_reserve helper funciton" patch converted all the
users of set_bit_string, GART, Calgary and AMD IOMMU drivers, to use
iommu_area_reserve helper function. Now we can remove unused
set_bit_string function.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoiommu: export iommu_area_reserve helper function
FUJITA Tomonori [Mon, 22 Sep 2008 13:35:07 +0000 (22:35 +0900)]
iommu: export iommu_area_reserve helper function

x86 has set_bit_string() that does the exact same thing that
set_bit_area() in lib/iommu-helper.c does.

This patch exports set_bit_area() in lib/iommu-helper.c as
iommu_area_reserve(), converts GART, Calgary, and AMD IOMMU to use it.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: use coherent_dma_mask in alloc_coherent
Joerg Roedel [Thu, 18 Sep 2008 13:54:23 +0000 (15:54 +0200)]
AMD IOMMU: use coherent_dma_mask in alloc_coherent

The alloc_coherent implementation for AMD IOMMU currently uses
*dev->dma_mask per default. This patch changes it to prefer
dev->coherent_dma_mask if it is set.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoadd AMD IOMMU tree to MAINTAINERS file
Joerg Roedel [Wed, 17 Sep 2008 15:22:45 +0000 (17:22 +0200)]
add AMD IOMMU tree to MAINTAINERS file

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: use cmd_buf_size when freeing the command buffer
Joerg Roedel [Wed, 17 Sep 2008 15:18:17 +0000 (17:18 +0200)]
AMD IOMMU: use cmd_buf_size when freeing the command buffer

The command buffer release function uses the CMD_BUF_SIZE macro for
get_order. Replace this with iommu->cmd_buf_size which is more reliable
about the actual size of the buffer.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: calculate IVHD size with a function
Joerg Roedel [Wed, 17 Sep 2008 15:14:27 +0000 (17:14 +0200)]
AMD IOMMU: calculate IVHD size with a function

The current calculation of the IVHD entry size is hard to read. So move
this code to a seperate function to make it more clear what this
calculation does.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: remove unnecessary cast to u64 in the init code
Joerg Roedel [Wed, 17 Sep 2008 14:45:59 +0000 (16:45 +0200)]
AMD IOMMU: remove unnecessary cast to u64 in the init code

The ctrl variable is only u32 and readl also returns a 32 bit value. So
the cast to u64 is pointless. Remove it with this patch.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: free domain bitmap with its allocation order
Joerg Roedel [Wed, 17 Sep 2008 10:19:58 +0000 (12:19 +0200)]
AMD IOMMU: free domain bitmap with its allocation order

The amd_iommu_pd_alloc_bitmap is allocated with a calculated order and
freed with order 1. This is not a bug since the calculated order always
evaluates to 1, but its unclean code. So replace the 1 with the
calculation in the release path.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: simplify dma_mask_to_pages
Joerg Roedel [Wed, 17 Sep 2008 10:17:00 +0000 (12:17 +0200)]
AMD IOMMU: simplify dma_mask_to_pages

The current calculation is very complicated. This patch replaces it with
a much simpler version.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: replace memset with __GFP_ZERO in alloc_coherent
Joerg Roedel [Thu, 11 Sep 2008 08:59:15 +0000 (10:59 +0200)]
AMD IOMMU: replace memset with __GFP_ZERO in alloc_coherent

Remove the memset and use __GFP_ZERO at allocation time instead.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: avoid unnecessary low zone allocation in alloc_coherent
FUJITA Tomonori [Wed, 10 Sep 2008 11:19:40 +0000 (20:19 +0900)]
AMD IOMMU: avoid unnecessary low zone allocation in alloc_coherent

x86's common alloc_coherent (dma_alloc_coherent in dma-mapping.h) sets
up the gfp flag according to the device dma_mask but AMD IOMMU doesn't
need it for devices that the IOMMU can do virtual mappings for. This
patch avoids unnecessary low zone allocation.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: some set_device_domain cleanups
Joerg Roedel [Thu, 11 Sep 2008 08:38:32 +0000 (10:38 +0200)]
AMD IOMMU: some set_device_domain cleanups

Remove some magic numbers and split the pte_root using standard
functions.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: don't assign preallocated protection domains to devices
Joerg Roedel [Thu, 11 Sep 2008 08:24:48 +0000 (10:24 +0200)]
AMD IOMMU: don't assign preallocated protection domains to devices

In isolation mode the protection domains for the devices are
preallocated and preassigned. This is bad if a device should be passed
to a virtualization guest because the IOMMU code does not know if it is
in use by a driver. This patch changes the code to assign the device to
the preallocated domain only if there are dma mapping requests for it.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: add dma_supported callback
Joerg Roedel [Tue, 9 Sep 2008 16:40:46 +0000 (18:40 +0200)]
AMD IOMMU: add dma_supported callback

This function determines if the AMD IOMMU implementation is responsible
for a given device. So the DMA layer can get this information from the
driver.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: allow IO page faults from devices
Joerg Roedel [Tue, 9 Sep 2008 15:55:28 +0000 (17:55 +0200)]
AMD IOMMU: allow IO page faults from devices

There is a bit in the device entry to suppress all IO page faults
generated by a device. This bit was set until now because there was no
event logging. Now that there is event logging this patch allows IO page
faults from devices to see them in the kernel log.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: enable event logging
Joerg Roedel [Tue, 9 Sep 2008 14:47:35 +0000 (16:47 +0200)]
AMD IOMMU: enable event logging

The code to log IOMMU events is in place now. So enable event logging
with this patch.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: add event handling code
Joerg Roedel [Tue, 9 Sep 2008 14:41:05 +0000 (16:41 +0200)]
AMD IOMMU: add event handling code

This patch adds code for polling and printing out events generated by
the AMD IOMMU.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: add MSI interrupt support
Joerg Roedel [Thu, 11 Sep 2008 14:51:41 +0000 (16:51 +0200)]
AMD IOMMU: add MSI interrupt support

The AMD IOMMU can generate interrupts for various reasons. This patch
adds the basic interrupt enabling infrastructure to the driver.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: save pci_dev instead of devid
Joerg Roedel [Mon, 8 Sep 2008 13:55:10 +0000 (15:55 +0200)]
AMD IOMMU: save pci_dev instead of devid

We need the pci_dev later anyways to enable MSI for the IOMMU hardware.
So remove the devid pointing to the BDF and replace it with the pci_dev
structure where the IOMMU is implemented.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: save pci segment from ACPI tables
Joerg Roedel [Mon, 8 Sep 2008 12:48:04 +0000 (14:48 +0200)]
AMD IOMMU: save pci segment from ACPI tables

This patch adds the pci_seg field to the amd_iommu structure and fills
it with the corresponding value from the ACPI table.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: add event buffer allocation
Joerg Roedel [Fri, 5 Sep 2008 12:29:07 +0000 (14:29 +0200)]
AMD IOMMU: add event buffer allocation

This patch adds the allocation of a event buffer for each AMD IOMMU in
the system. The hardware will log events like device page faults or
other errors to this buffer once this is enabled.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: align alloc_coherent addresses properly
Joerg Roedel [Thu, 4 Sep 2008 17:18:02 +0000 (19:18 +0200)]
AMD IOMMU: align alloc_coherent addresses properly

The API definition for dma_alloc_coherent states that the bus address
has to be aligned to the next power of 2 boundary greater than the
allocation size. This is violated by AMD IOMMU so far and this patch
fixes it.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: add branch hints to completion wait checks
Joerg Roedel [Thu, 4 Sep 2008 17:01:02 +0000 (19:01 +0200)]
AMD IOMMU: add branch hints to completion wait checks

This patch adds branch hints to the cecks if a completion_wait is
necessary. The completion_waits in the mapping paths are unlikly because
they will only happen on software implementations of AMD IOMMU which
don't exists today or with lazy IO/TLB flushing when the allocator wraps
around the address space. With lazy IO/TLB flushing the completion_wait
in the unmapping path is unlikely too.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: implement lazy IO/TLB flushing
Joerg Roedel [Thu, 4 Sep 2008 16:40:05 +0000 (18:40 +0200)]
AMD IOMMU: implement lazy IO/TLB flushing

The IO/TLB flushing on every unmaping operation is the most expensive
part in AMD IOMMU code and not strictly necessary. It is sufficient to
do the flush before any entries are reused. This is patch implements
lazy IO/TLB flushing which does exactly this.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: move GART TLB flushing options to generic code
Joerg Roedel [Thu, 18 Sep 2008 13:23:43 +0000 (15:23 +0200)]
x86: move GART TLB flushing options to generic code

The GART currently implements the iommu=[no]fullflush command line
parameters which influence its IO/TLB flushing strategy. This patch
makes these parameters generic so that they can be used by the AMD IOMMU
too.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: move TLB flushing to the map/unmap helper functions
Joerg Roedel [Thu, 4 Sep 2008 13:49:46 +0000 (15:49 +0200)]
AMD IOMMU: move TLB flushing to the map/unmap helper functions

This patch moves the invocation of the flushing functions to the
map/unmap helpers because its common code in all dma_ops relevant
mapping/unmapping code.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoAMD IOMMU: check for invalid device pointers
Joerg Roedel [Thu, 4 Sep 2008 13:04:26 +0000 (15:04 +0200)]
AMD IOMMU: check for invalid device pointers

Currently AMD IOMMU code triggers a BUG_ON if NULL is passed as the
device. This is inconsistent with other IOMMU implementations.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoconvert swiotlb to use dma_get_mask
FUJITA Tomonori [Thu, 18 Sep 2008 17:02:05 +0000 (02:02 +0900)]
convert swiotlb to use dma_get_mask

swiotlb can use dma_get_mask() instead of the homegrown function.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: tony.luck@intel.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: gart alloc_coherent does virtual mapppings only when necessary
FUJITA Tomonori [Thu, 11 Sep 2008 14:08:48 +0000 (23:08 +0900)]
x86: gart alloc_coherent does virtual mapppings only when necessary

gart alloc_coherent need to do virtual mapppings only when an
allocated buffer is not DMA-capable for a device.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: avoid unnecessary low zone allocation in Calgary's alloc_coherent
FUJITA Tomonori [Thu, 11 Sep 2008 14:08:47 +0000 (23:08 +0900)]
x86: avoid unnecessary low zone allocation in Calgary's alloc_coherent

x86's common alloc_coherent (dma_alloc_coherent in dma-mapping.h) sets
up the gfp flag according to the device dma_mask but Calgary doesn't
need it because of virtual mappings. This patch avoids unnecessary low
zone allocation.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agox86: make GART to respect device's dma_mask about virtual mappings
FUJITA Tomonori [Fri, 12 Sep 2008 10:42:35 +0000 (19:42 +0900)]
x86: make GART to respect device's dma_mask about virtual mappings

Currently, GART IOMMU ingores device's dma_mask when it does virtual
mappings. So it could give a device a virtual address that the device
can't access to.

This patch fixes the above problem.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoiommu: add dma_get_mask helper function
FUJITA Tomonori [Fri, 12 Sep 2008 10:42:34 +0000 (19:42 +0900)]
iommu: add dma_get_mask helper function

Several IOMMUs do the same thing to get the dma_mask of a device. This
adds a helper function to do the same thing to sweep them.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoiommu: add iommu_device_max_index IOMMU helper function
FUJITA Tomonori [Fri, 12 Sep 2008 10:42:33 +0000 (19:42 +0900)]
iommu: add iommu_device_max_index IOMMU helper function

This function helps IOMMUs to know the highest address that a device
can access to.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge branch 'linus' into x86/iommu
Ingo Molnar [Sun, 14 Sep 2008 12:07:00 +0000 (14:07 +0200)]
Merge branch 'linus' into x86/iommu

Conflicts:
lib/swiotlb.c

Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Sat, 13 Sep 2008 21:51:22 +0000 (14:51 -0700)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] Fix PCI_DMA_BUS_IS_PHYS for ARM
  [ARM] 5247/1: tosa: SW_EAR_IN support
  [ARM] 5246/1: tosa: add proper clock alias for tc6393xb clock
  [ARM] 5245/1: Fix warning about unused return value in drivers/pcmcia
  [ARM] OMAP: Fix MMC device data
  imx serial: fix rts handling for non imx1 based hardware
  imx serial: set RXD mux bit on i.MX27 and i.MX31
  i.MX serial: fix init failure
  pcm037: add rts/cts support for serial port

16 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Sat, 13 Sep 2008 21:48:14 +0000 (14:48 -0700)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [libata] LBA28/LBA48 off-by-one bug in ata.h
  sata_inic162x: enable LED blinking
  ata: duplicate variable sparse warning

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Sat, 13 Sep 2008 21:47:33 +0000 (14:47 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  PCI: re-add debug prints for unmodified BARs
  PCI: fix pciehp_free_irq()
  PCI Hotplug: fakephp: fix deadlock... again
  PCI: Fix printk warnings in setup-bus.c
  PCI: Fix printk warnings in probe.c
  PCI/iommu: blacklist DMAR on Intel G31/G33 chipsets

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Sat, 13 Sep 2008 21:46:57 +0000 (14:46 -0700)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  niu: panic on reset
  netlink: fix overrun in attribute iteration
  [Bluetooth] Fix regression from using default link policy
  ath9k: Assign seq# when mac80211 requests this

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Sat, 13 Sep 2008 21:45:42 +0000 (14:45 -0700)]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc: Fix user_regset 'n' field values.
  sparc64: Fix PCI error interrupt registry on PSYCHO.
  sparc32: Fix function signature of of_bus_sbus_get_flags().
  sparc64: Fix interrupt register calculations on Psycho and Sabre.

16 years agomemstick: fix MSProHG 8-bit interface mode support
Alex Dubov [Sat, 13 Sep 2008 09:33:26 +0000 (02:33 -0700)]
memstick: fix MSProHG 8-bit interface mode support

- 8-bit interface mode never worked properly.  The only adapter I have
  which supports the 8b mode (the Jmicron) had some problems with its
  clock wiring and they discovered it only now.  We also discovered that
  ProHG media is more sensitive to the ordering of initialization
  commands.

- Make the driver fall back to highest supported mode instead of always
  falling back to serial.  The driver will attempt the switch to 8b mode
  for any new MSPro card, but not all of them support it.  Previously,
  these new cards ended up in serial mode, which is not the best idea
  (they work fine with 4b, after all).

- Edit some macros for better conformance to Sony documentation

Signed-off-by: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agorescan_partitions(): make device capacity errors non-fatal
Andrew Morton [Sat, 13 Sep 2008 09:33:25 +0000 (02:33 -0700)]
rescan_partitions(): make device capacity errors non-fatal

Herton Krzesinski reports that the error-checking changes in
04ebd4aee52b06a2c38127d9208546e5b96f3a19 ("block/ioctl.c and
fs/partition/check.c: check value returned by add_partition") cause his
buggy USB camera to no longer mount.  "The camera is an Olympus X-840.
The original issue comes from the camera itself: its format program
creates a partition with an off by one error".

Buggy devices happen.  It is better for the kernel to warn and to proceed
with the mount.

Reported-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Cc: Abdel Benamrouche <draconux@gmail.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
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>
16 years agospi_s3c24xx: fix section warning
Ben Dooks [Sat, 13 Sep 2008 09:33:24 +0000 (02:33 -0700)]
spi_s3c24xx: fix section warning

Fix the section mismatch warning generated by the incorrect naming of
s3c24xx_spidrv which should be s3c24xx_spi_driver:

WARNING: drivers/spi/spi_s3c24xx.o(.data+0x4):
  Section mismatch in reference from the variable s3c24xx_spidrv
  to the (unknown reference) .exit.text:(unknown)

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoatmel_lcdfb: disable LCD and DMA engines when suspending
Haavard Skinnemoen [Sat, 13 Sep 2008 09:33:23 +0000 (02:33 -0700)]
atmel_lcdfb: disable LCD and DMA engines when suspending

When suspending the system with atmel_lcdfb enabled, I sometimes see
this:

atmel_lcdfb atmel_lcdfb.0: FIFO underflow 0x10

Which can be explained by the fact that we're not stopping the LCD
controller and its DMA engine when suspending, we're just gating the
clocks to them.

There's another potential issue which may be harder to trigger but
much more nasty: If we gate the clocks at _just_ the right moment,
e.g. when the DMA engine is doing a bus transaction, we may cause the
DMA engine to violate the system bus protocol and cause a lockup.

Avoid these issues by shutting down the LCD controller before entering
suspend (and restarting it when resuming). This prevents the underrun
from happening in the first place, and prevents whatever nastiness is
happening when the bus clock stops in the middle of a DMA transfer.

Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoia64: fix panic during `modprobe -r xpc'
Robin Holt [Sat, 13 Sep 2008 09:33:22 +0000 (02:33 -0700)]
ia64: fix panic during `modprobe -r xpc'

If you are on ia64 and you modprobe xpc then modprobe -r xpc, you
immediately get a panic.  xpc depends on xp which depends on gru for a
symbol.  That symbol is only used when we are running on UV hardware.

Currently, the GRU driver detects we are not on UV hardware and does no
initializing.  It does not do the same check when unloading.  As a result,
the gru driver attempts to tear down stuff that was not setup.

This is a simple two-line workaround to get us through this release.  Once
2.6.28 is opened, we need to rework the symbols that xp is depending on
from gru so the gru driver can properly fail to load when hardware is not
available.

Signed-off-by: Robin Holt <holt@sgi.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMAINTAINERS: fix USB VIDEO CLASS mail list address
Ming Lei [Sat, 13 Sep 2008 09:33:21 +0000 (02:33 -0700)]
MAINTAINERS: fix USB VIDEO CLASS mail list address

It should be linux-uvc-devel@lists.berlios.de.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoDocumentation/ABI: /sys/class/gpio
David Brownell [Sat, 13 Sep 2008 09:33:20 +0000 (02:33 -0700)]
Documentation/ABI: /sys/class/gpio

Provide summary ABI docs about the /sys/class/gpio files.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: mark the correct zone as full when scanning zonelists
Mel Gorman [Sat, 13 Sep 2008 09:33:19 +0000 (02:33 -0700)]
mm: mark the correct zone as full when scanning zonelists

The iterator for_each_zone_zonelist() uses a struct zoneref *z cursor when
scanning zonelists to keep track of where in the zonelist it is.  The
zoneref that is returned corresponds to the the next zone that is to be
scanned, not the current one.  It was intended to be treated as an opaque
list.

When the page allocator is scanning a zonelist, it marks elements in the
zonelist corresponding to zones that are temporarily full.  As the
zonelist is being updated, it uses the cursor here;

  if (NUMA_BUILD)
        zlc_mark_zone_full(zonelist, z);

This is intended to prevent rescanning in the near future but the zoneref
cursor does not correspond to the zone that has been found to be full.
This is an easy misunderstanding to make so this patch corrects the
problem by changing zoneref cursor to be the current zone being scanned
instead of the next one.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopxa2xx_spi: dma bugfixes
Ned Forrester [Sat, 13 Sep 2008 09:33:18 +0000 (02:33 -0700)]
pxa2xx_spi: dma bugfixes

Fixes two DMA bugs in the pxa2xx_spi driver.  The first bug is in all
versions of this driver; the second was introduced in the 2.6.20 kernel,
and prevents using the driver with chips like m25p16 flash (which can
issue large DMA reads).

 1. Zero length transfers are permitted for use to insert timing,
    but pxa2xx_spi.c will fail if this is requested in DMA mode.
    Fixed by using programmed I/O (PIO) mode for such transfers.

 2. Transfers larger than 8191 are not permitted in DMA mode.  A
    test for length rejects all large transfers regardless of DMA
    or PIO mode.  Worked around by rejecting only large transfers
    with DMA mapped buffers, and forcing all other transfers
    larger than 8191 to use PIO mode.  A rate limited warning is
    issued for DMA transfers forced to PIO mode.

This patch should apply to all kernels back to and including 2.6.20;
it was test patched against 2.6.20.  An additional patch would be
required for older kernels, but those versions are very buggy anyway.

Signed-off-by: Ned Forrester <nforrester@whoi.edu>
Cc: Vernon Sauder <vernoninhand@gmail.com>
Cc: Eric Miao <eric.y.miao@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopxa2xx_spi: chipselect bugfixes
Ned Forrester [Sat, 13 Sep 2008 09:33:17 +0000 (02:33 -0700)]
pxa2xx_spi: chipselect bugfixes

Fixes several chipselect bugs in the pxa2xx_spi driver.  These bugs are in
all versions of this driver and prevent using it with chips like m25p16
flash.

 1. The spi_transfer.cs_change flag is handled too early:
    before spi_transfer.delay_usecs applies, thus making the
    delay ineffective at holding chip select.

 2. spi_transfer.delay_usecs is ignored on the last transfer
    of a message (likewise not holding chipselect long enough).

 3. If spi_transfer.cs_change is set on the last transfer, the
    chip select is always disabled, instead of the intended
    meaning: optionally holding chip select enabled for the
    next message.

Those first three bugs were fixed with a relocation of delays
and chip select de-assertions.

 4. If a message has the cs_change flag set on the last transfer,
    and had the chip select stayed enabled as requested (see 3,
    above), it would not have been disabled if the next message is
    for a different chip.  Fixed by dropping chip select regardless
    of cs_change at end of a message, if there is no next message
    or if the next message is for a different chip.

This patch should apply to all kernels back to and including 2.6.20;
it was test patched against 2.6.20.  An additional patch would be
required for older kernels, but those versions are very buggy anyway.

Signed-off-by: Ned Forrester <nforrester@whoi.edu>
Cc: Vernon Sauder <vernoninhand@gmail.com>
Cc: Eric Miao <eric.y.miao@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi_mpc83xx: reject invalid transfer sizes
Peter Korsgaard [Sat, 13 Sep 2008 09:33:15 +0000 (02:33 -0700)]
spi_mpc83xx: reject invalid transfer sizes

Error out on transfer length != multiple of bytes per word with -EINVAL.
Fixes a buffer overrun crash if length < bytes per word.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Acked-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agospi_mpc83xx: fix clockrate calculation for low speed
Peter Korsgaard [Sat, 13 Sep 2008 09:33:14 +0000 (02:33 -0700)]
spi_mpc83xx: fix clockrate calculation for low speed

Commit a61f5345 (spi_mpc83xx clockrate fixes) broke clockrate calculation
for low speeds.  SPMODE_DIV16 should be set if the divider is higher than
64, not only if the divider gets clipped to 1024.

Furthermore, the clipping check was off by a factor 16 as well.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: ifdef Quicklists in /proc/meminfo
Hugh Dickins [Sat, 13 Sep 2008 09:33:13 +0000 (02:33 -0700)]
mm: ifdef Quicklists in /proc/meminfo

A "Quicklists:          0 kB" line has just started appearing in
/proc/meminfo, but most architectures (including x86) don't have
them configured, so #ifdef it, like the highmem lines.

And those architectures which do have quicklists configured are
using them for page tables: so let's place it next to PageTables.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Christoph Lameter <cl@linux-foundation.org>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agobfs: fix Lockdep warning
Eric Sesterhenn [Sat, 13 Sep 2008 09:33:12 +0000 (02:33 -0700)]
bfs: fix Lockdep warning

This fixes:

  =============================================
  [ INFO: possible recursive locking detected ]
  2.6.27-rc5-00283-g70bb089 #68
  ---------------------------------------------
  touch/6855 is trying to acquire lock:
   (&info->bfs_lock){--..}, at: [<c02262f5>] bfs_delete_inode+0x9e/0x18c

  but task is already holding lock:
   (&info->bfs_lock){--..}, at: [<c0226c00>] bfs_create+0x45/0x187

  other info that might help us debug this:
  2 locks held by touch/6855:
   #0:  (&type->i_mutex_dir_key#5){--..}, at: [<c018ad13>] do_filp_open+0x10b/0x62f
   #1:  (&info->bfs_lock){--..}, at: [<c0226c00>] bfs_create+0x45/0x187

  stack backtrace:
  Pid: 6855, comm: touch Not tainted 2.6.27-rc5-00283-g70bb089 #68
   [<c013e769>] validate_chain+0x458/0x9f4
   [<c013bece>] ? trace_hardirqs_off+0xb/0xd
   [<c013f36b>] __lock_acquire+0x666/0x6e0
   [<c013f440>] lock_acquire+0x5b/0x77
   [<c02262f5>] ? bfs_delete_inode+0x9e/0x18c
   [<c06aab74>] mutex_lock_nested+0xbc/0x234
   [<c02262f5>] ? bfs_delete_inode+0x9e/0x18c
   [<c02262f5>] ? bfs_delete_inode+0x9e/0x18c
   [<c02262f5>] bfs_delete_inode+0x9e/0x18c
   [<c0226257>] ? bfs_delete_inode+0x0/0x18c
   [<c01925e1>] generic_delete_inode+0x94/0xfe
   [<c019265d>] generic_drop_inode+0x12/0x12f
   [<c0191b7e>] iput+0x4b/0x4e
   [<c0226d1e>] bfs_create+0x163/0x187
   [<c0188b42>] vfs_create+0xa6/0x114
   [<c018adb5>] do_filp_open+0x1ad/0x62f
   [<c0107cdc>] ? native_sched_clock+0x82/0x96
   [<c06ac309>] ? _spin_unlock+0x27/0x3c
   [<c019379e>] ? alloc_fd+0xbf/0xc9
   [<c06ae2f4>] ? sub_preempt_count+0x9d/0xab
   [<c019379e>] ? alloc_fd+0xbf/0xc9
   [<c0180391>] do_sys_open+0x42/0xb8
   [<c041d564>] ? trace_hardirqs_on_thunk+0xc/0x10
   [<c0180449>] sys_open+0x1e/0x26
   [<c01038bd>] sysenter_do_call+0x12/0x31
   =======================

The problem is that we don't unlock the bfs->lock mutex before calling
iput (we do in the other cases).

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocoredump_filter: add description of bit 4
Hidehiro Kawai [Sat, 13 Sep 2008 09:33:10 +0000 (02:33 -0700)]
coredump_filter: add description of bit 4

There is no description of bit 4 of coredump_filter in the
documentation.  This patch adds it.

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocpuset: hotplug documentation fix
Li Zefan [Sat, 13 Sep 2008 09:33:09 +0000 (02:33 -0700)]
cpuset: hotplug documentation fix

If all the cpus in a cpuset are offlined, the tasks in it will be moved to
the nearest ancestor with non-empty cpus.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Paul Jackson <pj@sgi.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agocpuset: avoid changing cpuset's cpus when -errno returned
Li Zefan [Sat, 13 Sep 2008 09:33:08 +0000 (02:33 -0700)]
cpuset: avoid changing cpuset's cpus when -errno returned

After the patch:

commit 0b2f630a28d53b5a2082a5275bc3334b10373508
Author: Miao Xie <miaox@cn.fujitsu.com>
Date:   Fri Jul 25 01:47:21 2008 -0700

    cpusets: restructure the function update_cpumask() and update_nodemask()

It might happen that 'echo 0 > /cpuset/sub/cpus' returned failure but 'cpus'
has been changed, because cpus was changed before calling heap_init() which
may return -ENOMEM.

This patch restores the orginal behavior.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Paul Menage <menage@google.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoinclude/linux/ioport.h: add missing macro argument for devm_release_* family
Hiroshi DOYU [Sat, 13 Sep 2008 09:33:07 +0000 (02:33 -0700)]
include/linux/ioport.h: add missing macro argument for devm_release_* family

akpm: these have no callers at this time, but they shall soon, so let's
get them right.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Hiroshi DOYU <Hiroshi.DOYU@nokia.com>
Cc: Tony Lindgren <tony@atomide.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoproc: more debugging for "already registered" case
Alexey Dobriyan [Sat, 13 Sep 2008 09:33:06 +0000 (02:33 -0700)]
proc: more debugging for "already registered" case

Print parent directory name as well.

The aim is to catch non-creation of parent directory when proc_mkdir will
return NULL and all subsequent registrations go directly in /proc instead
of intended directory.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Fixed insane printk string while at it.  - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years ago[libata] LBA28/LBA48 off-by-one bug in ata.h
Taisuke Yamada [Sat, 13 Sep 2008 20:46:15 +0000 (16:46 -0400)]
[libata] LBA28/LBA48 off-by-one bug in ata.h

I recently bought 3 HGST P7K500-series 500GB SATA drives and
had trouble accessing the block right on the LBA28-LBA48 border.
Here's how it fails (same for all 3 drives):

  # dd if=/dev/sdc bs=512 count=1 skip=268435455 > /dev/null
  dd: reading `/dev/sdc': Input/output error
  0+0 records in
  0+0 records out
  0 bytes (0 B) copied, 0.288033 seconds, 0.0 kB/s
  # dmesg
  ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
  ata1.00: BMDMA stat 0x25
  ata1.00: cmd c8/00:08:f8:ff:ff/00:00:00:00:00/ef tag 0 dma 4096 in
  res 51/04:08:f8:ff:ff/00:00:00:00:00/ef Emask 0x1 (device error)
  ata1.00: status: { DRDY ERR }
  ata1.00: error: { ABRT }
  ata1.00: configured for UDMA/33
  ata1: EH complete
  ...

After some investigations, it turned out this seems to be caused
by misinterpretation of the ATA specification on LBA28 access.
Following part is the code in question:

  === include/linux/ata.h ===
  static inline int lba_28_ok(u64 block, u32 n_block)
  {
    /* check the ending block number */
    return ((block + n_block - 1) < ((u64)1 << 28)) && (n_block <= 256);
  }

HGST drive (sometimes) fails with LBA28 access of {block = 0xfffffff,
n_block = 1}, and this behavior seems to be comformant. Other drives,
including other HGST drives are not that strict, through.

>From the ATA specification:
(http://www.t13.org/Documents/UploadedDocuments/project/d1410r3b-ATA-ATAPI-6.pdf)

  8.15.29  Word (61:60): Total number of user addressable sectors
  This field contains a value that is one greater than the total number
  of user addressable sectors (see 6.2). The maximum value that shall
  be placed in this field is 0FFFFFFFh.

So the driver shouldn't use the value of 0xfffffff for LBA28 request
as this exceeds maximum user addressable sector. The logical maximum
value for LBA28 is 0xffffffe.

The obvious fix is to cut "- 1" part, and the patch attached just do
that. I've been using the patched kernel for about a month now, and
the same fix is also floating on the net for some time. So I believe
this fix works reliably.

Just FYI, many Windows/Intel platform users also seems to be struck
by this, and HGST has issued a note pointing to Intel ICH8/9 driver.

  "28-bit LBA command is being used to access LBAs 29-bits in length"
http://www.hitachigst.com/hddt/knowtree.nsf/cffe836ed7c12018862565b000530c74/b531b8bce8745fb78825740f00580e23

Also, *BSDs seems to have similar fix included sometime around ~2004,
through I have not checked out exact portion of the code.

Signed-off-by: Taisuke Yamada <tai@rakugaki.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agosata_inic162x: enable LED blinking
Bob Stewart [Thu, 11 Sep 2008 09:50:03 +0000 (11:50 +0200)]
sata_inic162x: enable LED blinking

Enable LED blinking.

Signed-off-by: Bob Stewart <bob@evoria.net>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years agoata: duplicate variable sparse warning
Stephen Hemminger [Mon, 8 Sep 2008 16:31:39 +0000 (09:31 -0700)]
ata: duplicate variable sparse warning

drivers/ata/ata_piix.c:1502:7: warning: symbol 'rc' shadows an earlier one

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
16 years ago[ARM] Fix PCI_DMA_BUS_IS_PHYS for ARM
Russell King [Sat, 13 Sep 2008 20:23:06 +0000 (21:23 +0100)]
[ARM] Fix PCI_DMA_BUS_IS_PHYS for ARM

PCI_DMA_BUS_IS_PHYS was defined to be zero, which meant we ignored
the DMA mask for IDE and SCSI transfers.  This is wrong - we have
no DMA translation hardware.  We want to obey DMA masks so that the
block layer performs bouncing itself.

Reported-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
16 years agoMerge branch 'for-rmk' of git://pasiphae.extern.pengutronix.de/git/imx/linux-2.6.git
Russell King [Sat, 13 Sep 2008 19:10:59 +0000 (20:10 +0100)]
Merge branch 'for-rmk' of git://pasiphae.extern.pengutronix.de/git/imx/linux-2.6.git

16 years ago[ARM] 5247/1: tosa: SW_EAR_IN support
Dmitry Baryshkov [Thu, 11 Sep 2008 00:28:51 +0000 (01:28 +0100)]
[ARM] 5247/1: tosa: SW_EAR_IN support

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Eric Miao <eric.miao@marvell.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
16 years ago[ARM] 5246/1: tosa: add proper clock alias for tc6393xb clock
Dmitry Baryshkov [Wed, 10 Sep 2008 09:30:37 +0000 (10:30 +0100)]
[ARM] 5246/1: tosa: add proper clock alias for tc6393xb clock

Add clock alias for clock that is used by tc6393xb device on tosa.
As that chip plays pretty major part in tosa life and is currently
disabled, this is 2.4.27 material.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
16 years ago[ARM] 5245/1: Fix warning about unused return value in drivers/pcmcia
Jürgen Schindele [Mon, 8 Sep 2008 21:06:29 +0000 (22:06 +0100)]
[ARM] 5245/1: Fix warning about unused return value in drivers/pcmcia

Fix warning when compiling "drivers/pcmcia/soc-common.c"
The return value of the function "device_create_file"
was not used / assigned.

Signed-off-by: Jrgen Schindele <linux@schindele.name>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
16 years agoPCI: re-add debug prints for unmodified BARs
Yinghai Lu [Tue, 9 Sep 2008 19:27:52 +0000 (12:27 -0700)]
PCI: re-add debug prints for unmodified BARs

Print out for device BAR values before the kernel tries to update them.
Also make related output use KERN_DEBUG.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
16 years agoniu: panic on reset
Santwona Behera [Fri, 12 Sep 2008 23:04:26 +0000 (16:04 -0700)]
niu: panic on reset

The reset_task function in the niu driver does not reset the tx and rx
buffers properly. This leads to panic on reset. This patch is a
modified implementation of the previously posted fix.

Signed-off-by: Santwona Behera <santwona.behera@sun.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agosparc: Fix user_regset 'n' field values.
David S. Miller [Fri, 12 Sep 2008 22:01:31 +0000 (15:01 -0700)]
sparc: Fix user_regset 'n' field values.

As noticed by Russell King, we were not setting this properly
to the number of entries, but rather the total size.

This results in the core dumping code allocating waayyyy too
much memory.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agosparc64: Fix PCI error interrupt registry on PSYCHO.
David S. Miller [Fri, 12 Sep 2008 22:13:15 +0000 (15:13 -0700)]
sparc64: Fix PCI error interrupt registry on PSYCHO.

We need to pass IRQF_SHARED, otherwise we get things like:

IRQ handler type mismatch for IRQ 33
current handler: PSYCHO_UE
Call Trace:
 [000000000048394c] request_irq+0xac/0x120
 [00000000007c5f6c] psycho_scan_bus+0x98/0x158
 [00000000007c2bc0] pcibios_init+0xdc/0x12c
 [0000000000426a5c] do_one_initcall+0x1c/0x160
 [00000000007c0180] kernel_init+0x9c/0xfc
 [0000000000427050] kernel_thread+0x30/0x60
 [00000000006ae1d0] rest_init+0x10/0x60

on e3500 and similar systems.

On a single board, the UE interrupts of two Psycho nodes
are funneled through the same interrupt, from of_debug=3
dump:

/pci@b,4000: direct translate 2ee --> 21
 ...
/pci@b,2000: direct translate 2ee --> 21

Decimal "33" mentioned above is the hex "21" mentioned here.

Thanks to Meelis Roos for dumps and testing.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/holtmann/bluet...
David S. Miller [Fri, 12 Sep 2008 02:11:50 +0000 (19:11 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/holtmann/bluetooth-2.6

16 years agonetlink: fix overrun in attribute iteration
Vegard Nossum [Fri, 12 Sep 2008 02:05:29 +0000 (19:05 -0700)]
netlink: fix overrun in attribute iteration

kmemcheck reported this:

  kmemcheck: Caught 16-bit read from uninitialized memory (f6c1ba30)
  0500110001508abf050010000500000002017300140000006f72672e66726565
   i i i i i i i i i i i i i u u u u u u u u u u u u u u u u u u u
                                   ^

  Pid: 3462, comm: wpa_supplicant Not tainted (2.6.27-rc3-00054-g6397ab9-dirty #13)
  EIP: 0060:[<c05de64a>] EFLAGS: 00010296 CPU: 0
  EIP is at nla_parse+0x5a/0xf0
  EAX: 00000008 EBX: fffffffd ECX: c06f16c0 EDX: 00000005
  ESI: 00000010 EDI: f6c1ba30 EBP: f6367c6c ESP: c0a11e88
   DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
  CR0: 8005003b CR2: f781cc84 CR3: 3632f000 CR4: 000006d0
  DR0: c0ead9bc DR1: 00000000 DR2: 00000000 DR3: 00000000
  DR6: ffff4ff0 DR7: 00000400
   [<c05d4b23>] rtnl_setlink+0x63/0x130
   [<c05d5f75>] rtnetlink_rcv_msg+0x165/0x200
   [<c05ddf66>] netlink_rcv_skb+0x76/0xa0
   [<c05d5dfe>] rtnetlink_rcv+0x1e/0x30
   [<c05dda21>] netlink_unicast+0x281/0x290
   [<c05ddbe9>] netlink_sendmsg+0x1b9/0x2b0
   [<c05beef2>] sock_sendmsg+0xd2/0x100
   [<c05bf945>] sys_sendto+0xa5/0xd0
   [<c05bf9a6>] sys_send+0x36/0x40
   [<c05c03d6>] sys_socketcall+0x1e6/0x2c0
   [<c020353b>] sysenter_do_call+0x12/0x3f
   [<ffffffff>] 0xffffffff

This is the line in nla_ok():

  /**
   * nla_ok - check if the netlink attribute fits into the remaining bytes
   * @nla: netlink attribute
   * @remaining: number of bytes remaining in attribute stream
   */
  static inline int nla_ok(const struct nlattr *nla, int remaining)
  {
          return remaining >= sizeof(*nla) &&
                 nla->nla_len >= sizeof(*nla) &&
                 nla->nla_len <= remaining;
  }

It turns out that remaining can become negative due to alignment in
nla_next(). But GCC promotes "remaining" to unsigned in the test
against sizeof(*nla) above. Therefore the test succeeds, and the
nla_for_each_attr() may access memory outside the received buffer.

A short example illustrating this point is here:

  #include <stdio.h>

  main(void)
  {
          printf("%d\n", -1 >= sizeof(int));
  }

...which prints "1".

This patch adds a cast in front of the sizeof so that GCC will make
a signed comparison and fix the illegal memory dereference. With the
patch applied, there is no kmemcheck report.

Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[Bluetooth] Fix regression from using default link policy
Marcel Holtmann [Fri, 12 Sep 2008 01:11:54 +0000 (03:11 +0200)]
[Bluetooth] Fix regression from using default link policy

To speed up the Simple Pairing connection setup, the support for the
default link policy has been enabled. This is in contrast to settings
the link policy on every connection setup. Using the default link policy
is the preferred way since there is no need to dynamically change it for
every connection.

For backward compatibility reason and to support old userspace the
HCISETLINKPOL ioctl has been switched over to using hci_request() to
issue the HCI command for setting the default link policy instead of
just storing it in the HCI device structure.

However the hci_request() can only be issued when the device is
brought up. If used on a device that is registered, but still down
it will timeout and fail. This is problematic since the command is
put on the TX queue and the Bluetooth core tries to submit it to
hardware that is not ready yet. The timeout for these requests is
10 seconds and this causes a significant regression when setting up
a new device.

The userspace can perfectly handle a failure of the HCISETLINKPOL
ioctl and will re-submit it later, but the 10 seconds delay causes
a problem. So in case hci_request() is called on a device that is
still down, just fail it with ENETDOWN to indicate what happens.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
16 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Thu, 11 Sep 2008 18:50:15 +0000 (11:50 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  sg: disable interrupts inside sg_copy_buffer

16 years agoMN10300: Change the fault handler to check in_atomic() not in_interrupt()
David Howells [Thu, 11 Sep 2008 16:18:56 +0000 (17:18 +0100)]
MN10300: Change the fault handler to check in_atomic() not in_interrupt()

Change the MN10300 fault handler to make it check in_atomic() rather than
in_interrupt() as commit 6edaf68a87d17570790fd55f0c451a29ec1d6703 did for other
architectures:

Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date:   Wed Dec 6 20:32:18 2006 -0800

[PATCH] mm: arch do_page_fault() vs in_atomic()

In light of the recent pagefault and filemap_copy_from_user work I've
gone through all the arch pagefault handlers to make sure the
inc_preempt_count() 'feature' works as expected.

Several sections of code (including the new filemap_copy_from_user)
rely on the fact that faults do not take locks under increased preempt
count.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agosg: disable interrupts inside sg_copy_buffer
FUJITA Tomonori [Thu, 11 Sep 2008 16:35:39 +0000 (18:35 +0200)]
sg: disable interrupts inside sg_copy_buffer

The callers of sg_copy_buffer must disable interrupts before calling
it (since it uses kmap_atomic). Some callers use it on
interrupt-disabled code but some need to take the trouble to disable
interrupts just for this. No wonder they forget about it and we hit a
bug like:

http://bugzilla.kernel.org/show_bug.cgi?id=11529

James said that it might be better to disable interrupts inside the
function rather than risk the callers getting it wrong.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Thu, 11 Sep 2008 15:42:55 +0000 (08:42 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: disable sysfs parts of the disk command filter

16 years agoMerge branch 'kvm-updates/2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 11 Sep 2008 15:42:14 +0000 (08:42 -0700)]
Merge branch 'kvm-updates/2.6.27' of git://git./linux/kernel/git/avi/kvm

* 'kvm-updates/2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm:
  KVM: VMX: Always return old for clear_flush_young() when using EPT
  KVM: SVM: fix guest global tlb flushes with NPT
  KVM: SVM: fix random segfaults with NPT enabled

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Thu, 11 Sep 2008 15:41:17 +0000 (08:41 -0700)]
Merge git://git./linux/kernel/git/jejb/scsi-rc-fixes-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
  [SCSI] fix check of PQ and PDT bits for WLUNs
  [SCSI] make scsi_check_sense HARDWARE_ERROR return ADD_TO_MLQUEUE on retry
  [SCSI] scsi_dh: make check_sense return ADD_TO_MLQUEUE
  [SCSI] zfcp: Remove duplicated unlikely() macros.
  [SCSI] zfcp: channel cannot be detached due to refcount imbalance
  [SCSI] zfcp: Fix reference counter for remote ports
  [SCSI] zfcp: Simplify ccw notify handler
  [SCSI] zfcp: Correctly query end flag in gpn_ft response
  [SCSI] zfcp: Fix request queue locking
  [SCSI] sd: select CRC_T10DIF only when necessary

16 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Thu, 11 Sep 2008 15:40:32 +0000 (08:40 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] prevent ia64 from invoking irq handlers on offline CPUs
  [IA64] arch/ia64/sn/pci/tioca_provider.c: introduce missing kfree
  [IA64] fix up bte.h
  [IA64] fix compile failure with non modular builds

16 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux...
Linus Torvalds [Thu, 11 Sep 2008 15:40:11 +0000 (08:40 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-udf-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-udf-2.6:
  udf: add llseek method
  udf: Fix error paths in udf_new_inode()
  udf: Fix lock inversion between iprune_mutex and alloc_mutex (v2)

16 years agoath9k: Assign seq# when mac80211 requests this
Jouni Malinen [Mon, 11 Aug 2008 11:01:50 +0000 (14:01 +0300)]
ath9k: Assign seq# when mac80211 requests this

Use TX control flag IEEE80211_TX_CTL_ASSIGN_SEQ as a request to update
the seq# for the frames. This will likely require some further cleanup
to get seq# correctly for Beacons vs. other frames and also potentially
for multiple BSSes. Anyway, this is better than ending up sending out
most frames with seq# 0.

(This is a backport of patch w/ same title already in net-next-2.6.
It is verified to fix http://bugzilla.kernel.org/show_bug.cgi?id=11394
and it should be acceptable for -rc due to the driver being new
in 2.6.27.)

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
16 years agoblock: disable sysfs parts of the disk command filter
Jens Axboe [Thu, 11 Sep 2008 12:20:23 +0000 (14:20 +0200)]
block: disable sysfs parts of the disk command filter

We still have life time issues with the sysfs command filter kobject,
so disable it for 2.6.27 release. We can revisit this and make it work
properly for 2.6.28, for 2.6.27 release it's too risky.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
16 years ago[ARM] OMAP: Fix MMC device data
Russell King [Tue, 9 Sep 2008 09:16:22 +0000 (10:16 +0100)]
[ARM] OMAP: Fix MMC device data

OMAPs MMC device data was passing the wrong structure via the platform
device.  Moreover, a missing function means that both sx1_defconfig
and omap_h2_1610_defconfig builds failed with

undefined reference to `omap_set_mmc_info'

errors.  Fix this by updating the MMC support from the omapzoom tree.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
16 years agoKVM: VMX: Always return old for clear_flush_young() when using EPT
Sheng Yang [Mon, 8 Sep 2008 07:12:30 +0000 (15:12 +0800)]
KVM: VMX: Always return old for clear_flush_young() when using EPT

As well as discard fake accessed bit and dirty bit of EPT.

Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: fix guest global tlb flushes with NPT
Joerg Roedel [Tue, 9 Sep 2008 17:11:51 +0000 (19:11 +0200)]
KVM: SVM: fix guest global tlb flushes with NPT

Accesses to CR4 are intercepted even with Nested Paging enabled. But the code
does not check if the guest wants to do a global TLB flush. So this flush gets
lost. This patch adds the check and the flush to svm_set_cr4.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: fix random segfaults with NPT enabled
Joerg Roedel [Wed, 27 Aug 2008 12:18:43 +0000 (14:18 +0200)]
KVM: SVM: fix random segfaults with NPT enabled

This patch introduces a guest TLB flush on every NPF exit in KVM. This fixes
random segfaults and #UD exceptions in the guest seen under some workloads
(e.g. long running compile workloads or tbench). A kernbench run with and
without that fix showed that it has a slowdown lower than 0.5%

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agosparc32: Fix function signature of of_bus_sbus_get_flags().
David S. Miller [Thu, 11 Sep 2008 06:38:51 +0000 (23:38 -0700)]
sparc32: Fix function signature of of_bus_sbus_get_flags().

This doesn't match the function pointer type it gets assigned
to.  Luckily, this was harmless.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Wed, 10 Sep 2008 21:16:53 +0000 (14:16 -0700)]
Merge git://git./linux/kernel/git/bart/ide-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6:
  add deprecated ide-scsi to feature-removal-schedule.txt
  ide: Fix pointer arithmetic in hpt3xx driver code (3rd try)