linux-edison.git
12 years agopowerpc/pseries: Update NUMA VDSO information when updating CPU maps
Jesse Larrew [Wed, 24 Apr 2013 06:03:48 +0000 (06:03 +0000)]
powerpc/pseries: Update NUMA VDSO information when updating CPU maps

The following patch adds vdso_getcpu_init(), which stores the NUMA node for
a cpu in SPRG3:

Commit 18ad51dd34 ("powerpc: Add VDSO version of getcpu") adds
vdso_getcpu_init(), which stores the NUMA node for a cpu in SPRG3.

This patch ensures that this information is also updated when the NUMA
affinity of a cpu changes.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Use stop machine to update cpu maps
Nathan Fontenot [Wed, 24 Apr 2013 06:02:13 +0000 (06:02 +0000)]
powerpc/pseries: Use stop machine to update cpu maps

The new PRRN firmware feature allows CPU and memory resources to be
transparently reassigned across NUMA boundaries. When this happens, the
kernel must update the node maps to reflect the new affinity information.

Although the NUMA maps can be protected by locking primitives during the
update itself, this is insufficient to prevent concurrent accesses to these
structures. Since cpumask_of_node() hands out a pointer to these
structures, they can still be modified outside of the lock. Furthermore,
tracking down each usage of these pointers and adding locks would be quite
invasive and difficult to maintain.

The approach used is to make a list of affected cpus and call stop_machine
to have the update routine run on each of the affected cpus allowing them
to update themselves. Each cpu finds itself in the list of cpus and makes
the appropriate updates. We need to have each cpu do this for themselves to
handle calls to vdso_getcpu_init() added in a subsequent patch.

Situations like these are best handled using stop_machine(). Since the NUMA
affinity updates are exceptionally rare events, this approach has the
benefit of not adding any overhead while accessing the NUMA maps during
normal operation.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Update CPU maps when device tree is updated
Jesse Larrew [Wed, 24 Apr 2013 06:00:35 +0000 (06:00 +0000)]
powerpc/pseries: Update CPU maps when device tree is updated

Platform events such as partition migration or the new PRRN firmware
feature can cause the NUMA characteristics of a CPU to change, and these
changes will be reflected in the device tree nodes for the affected
CPUs.

This patch registers a handler for Open Firmware device tree updates
and reconfigures the CPU and node maps whenever the associativity
changes. Currently, this is accomplished by marking the affected CPUs in
the cpu_associativity_changes_mask and allowing
arch_update_cpu_topology() to retrieve the new associativity information
using hcall_vphn().

Protecting the NUMA cpu maps from concurrent access during an update
operation will be addressed in a subsequent patch in this series.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Update numa.c to use updated firmware_has_feature()
Nathan Fontenot [Wed, 24 Apr 2013 05:58:23 +0000 (05:58 +0000)]
powerpc/pseries: Update numa.c to use updated firmware_has_feature()

Update the numa code to use the updated firmware_has_feature() when checking
for type 1 affinity.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Update firmware_has_feature() to check architecture vector 5 bits
Nathan Fontenot [Wed, 24 Apr 2013 05:57:18 +0000 (05:57 +0000)]
powerpc/pseries: Update firmware_has_feature() to check architecture vector 5 bits

The firmware_has_feature() function makes it easy to check for supported
features of the hypervisor. This patch extends the capability of
firmware_has_feature() to include checking for specified bits
in vector 5 of the architecture vector as reported in the device tree.

As part of this the #defines used for the architecture vector are re-defined
such that each option has the index into vector 5 and the feature bit encoded
into it. This makes checking for architecture bits when initiating data
for firmware_has_feature much easier.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Use ARRAY_SIZE to iterate over firmware_features_table array
Nathan Fontenot [Wed, 24 Apr 2013 05:55:08 +0000 (05:55 +0000)]
powerpc/pseries: Use ARRAY_SIZE to iterate over firmware_features_table array

When iterating over the entries in firmware_features_table we only need
to go over the actual number of entries in the array instead of declaring
it to be bigger and checking to make sure there is a valid entry in every
slot.

This patch removes the FIRMWARE_MAX_FEATURES #define and replaces the
array looping with the use of ARRAY_SIZE().

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Move architecture vector definitions to prom.h
Nathan Fontenot [Wed, 24 Apr 2013 05:53:10 +0000 (05:53 +0000)]
powerpc/pseries: Move architecture vector definitions to prom.h

As part of handling of PRRN events we need to check vector 5 of the
architecture vector bits reported in the device tree to ensure PRRN event
handling is enabled. To do this firmware_has_feature() is updated (in a
subsequent patch) to make this check vector 5 bits. To avoid having to
re-define bits in the architecture vector the bit definitions are moved
to prom.h.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Add PRRN RTAS event handler
Jesse Larrew [Wed, 24 Apr 2013 05:51:33 +0000 (05:51 +0000)]
powerpc/pseries: Add PRRN RTAS event handler

A PRRN event is signaled via the RTAS event-scan mechanism, which
returns a Hot Plug Event message "fixed part" indicating "Platform
Resource Reassignment". In response to the Hot Plug Event message,
we must call ibm,update-nodes to determine which resources were
reassigned and then ibm,update-properties to obtain the new affinity
information about those resources.

The PRRN event-scan RTAS message contains only the "fixed part" with
the "Type" field set to the value 160 and no Extended Event Log. The
four-byte Extended Event Log Length field is re-purposed (since no
Extended Event Log message is included) to pass the "scope" parameter
that causes the ibm,update-nodes to return the nodes affected by the
specific resource reassignment.

This patch adds a handler for RTAS events. The function
pseries_devicetree_update() (from mobility.c) is used to make the
ibm,update-nodes/ibm,update-properties RTAS calls. Updating the NUMA maps
(handled by a subsequent patch) will require significant processing,
so pseries_devicetree_update() is called from an asynchronous workqueue
to allow event processing to continue.

PRRN RTAS events on pseries systems are rare events that have to be
initiated from the HMC console for the system by an IBM tech. This allows
us to assume that these events are widely spaced. Additionally, all work
on the queue is flushed before handling any new work to ensure we only have
one event in flight being handled at a time.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Correct buffer parsing in update_dt_node()
Nathan Fontenot [Wed, 24 Apr 2013 05:49:36 +0000 (05:49 +0000)]
powerpc/pseries: Correct buffer parsing in update_dt_node()

Correct parsing of the buffer returned from ibm,update-properties. The first
element is a length and the path to the property which is slightly different
from the list of properties in the buffer so we need to specifically
handle this.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/pseries: Expose pseries devicetree_update()
Nathan Fontenot [Wed, 24 Apr 2013 05:47:11 +0000 (05:47 +0000)]
powerpc/pseries: Expose pseries devicetree_update()

Newer firmware on Power systems can transparently reassign platform resources
(CPU and Memory) in use. For instance, if a processor or memory unit is
predicted to fail, the platform may transparently move the processing to an
equivalent unused processor or the memory state to an equivalent unused
memory unit. However, reassigning resources across NUMA boundaries may alter
the performance of the partition. When such reassignment is necessary, the
Platform Resource Reassignment Notification (PRRN) option provides a
mechanism to inform the Linux kernel of changes to the NUMA affinity of
its platform resources.

When rtasd receives a PRRN event, it needs to make a series of RTAS
calls (ibm,update-nodes and ibm,update-properties) to retrieve the
updated device tree information. These calls are already handled in the
pseries_devicetree_update() routine used in partition migration.

This patch exposes pseries_devicetree_update() to make it accessible
to other pseries routines, this patch also updates pseries_devicetree_update()
to take a 32-bit scope parameter. The scope value, which was previously hard
coded to 1 for partition migration, is used for the RTAS calls
ibm,update-nodes/properties to update the device tree.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Fix hardware IRQs with MMU on exceptions when HV=0
Michael Neuling [Thu, 25 Apr 2013 15:30:57 +0000 (15:30 +0000)]
powerpc: Fix hardware IRQs with MMU on exceptions when HV=0

POWER8 allows us to take interrupts with the MMU on.  This gives us a
second set of vectors offset at 0x4000.

Unfortunately when coping these vectors we missed checking for MSR HV
for hardware interrupts (0x500).  This results in us trying to use
HSRR0/1 when HV=0, rather than SRR0/1 on HW IRQs

The below fixes this to check CPU_FTR_HVMODE when patching the code at
0x4500.

Also we remove the check for CPU_FTR_ARCH_206 since relocation on IRQs
are only available in arch 2.07 and beyond.

Thanks to benh for helping find this.

Signed-off-by: Michael Neuling <mikey@neuling.org>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/power8: Fix secondary CPUs hanging on boot for HV=0
Michael Neuling [Wed, 24 Apr 2013 21:00:37 +0000 (21:00 +0000)]
powerpc/power8: Fix secondary CPUs hanging on boot for HV=0

In __restore_cpu_power8 we determine if we are HV and if not, we return
before setting HV only resources.

Unfortunately we forgot to restore the link register from r11 before
returning.

This will happen on boot and with secondary CPUs not coming online.

This adds the missing link register restore.

Signed-off-by: Michael Neuling <mikey@neuling.org>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Add isync to copy_and_flush
Michael Neuling [Wed, 24 Apr 2013 00:30:09 +0000 (00:30 +0000)]
powerpc: Add isync to copy_and_flush

In __after_prom_start we copy the kernel down to zero in two calls to
copy_and_flush.  After the first call (copy from 0 to copy_to_here:)
we jump to the newly copied code soon after.

Unfortunately there's no isync between the copy of this code and the
jump to it.  Hence it's possible that stale instructions could still be
in the icache or pipeline before we branch to it.

We've seen this on real machines and it's results in no console output
after:
  calling quiesce...
  returning from prom_init

The below adds an isync to ensure that the copy and flushing has
completed before any branching to the new instructions occurs.

Signed-off-by: Michael Neuling <mikey@neuling.org>
CC: <stable@vger.kernel.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Add HWCAP2 aux entry
Michael Neuling [Wed, 17 Apr 2013 17:33:11 +0000 (17:33 +0000)]
powerpc: Add HWCAP2 aux entry

We are currently out of free bits in AT_HWCAP. With POWER8, we have
several hardware features that we need to advertise.

Tested on POWER and x86.

Signed-off-by: Michael Neuling <michael@neuling.org>
Signed-off-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agoMerge branch 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux...
Dave Airlie [Fri, 26 Apr 2013 05:42:02 +0000 (15:42 +1000)]
Merge branch 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6 into drm-next

"Nothing overly exciting here aside from calim's fermi/kepler vram
compression patches.  The rest is misc fixes I gathered from the list.

Most of the stuff from me is fixing issues that have come up from the
work on kepler PM, as well as a commit moving all the old-school
modesetting out of the way (no code changes here).  There's other
patches to go on top of that, but, it'll have to wait until I can rip
out the old PM code, it's a bit tangled."

* 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6: (38 commits)
  drm/nouveau/fifo: implement channel creation event generation
  drm/nouveau/core: allow non-maskable events
  drm/nouveau/timer: allow alarms to be cancelled
  drm/nouveau/device: tweak the device/subdev relationship a little
  drm/nouveau/device: enable proper constructor/destructor
  drm/nouveau/device: have engine object initialised before creation
  drm/nouveau/device: convert to engine, rather than subdev
  drm/nv50-/disp: use self as parent for subobjects
  drm/nv50-/fifo: use parent as self for subobjects
  drm/nv20-nv30/gr: use parent as self for subobjects
  drm/nvc0-/gr: use self as parent for subobjects
  drm/nv04-nv40/instmem: use self as parent for subobjects
  drm/nv04-nv40/vm: use self as parent for subobjects
  drm/nv50-/bar: use self as parent for subobjects
  drm/nv04-nv40/instmem: remove parent deref hack
  drm/nouveau/i2c: remove parent deref hack
  drm/nouveau/core: rebase object ref/use counts after ctor/init/fini events
  drm/nv50/disp: inform core when we're not creating a new context
  drm/nouveau/therm: send some messages to debug level
  drm/nve0/gr: add handling for a bunch of PGRAPH traps
  ...

12 years agodrm/nouveau/fifo: implement channel creation event generation
Ben Skeggs [Fri, 26 Apr 2013 01:44:26 +0000 (11:44 +1000)]
drm/nouveau/fifo: implement channel creation event generation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/core: allow non-maskable events
Ben Skeggs [Fri, 26 Apr 2013 02:03:10 +0000 (12:03 +1000)]
drm/nouveau/core: allow non-maskable events

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/timer: allow alarms to be cancelled
Ben Skeggs [Thu, 25 Apr 2013 23:17:22 +0000 (09:17 +1000)]
drm/nouveau/timer: allow alarms to be cancelled

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/device: tweak the device/subdev relationship a little
Ben Skeggs [Thu, 25 Apr 2013 01:43:54 +0000 (11:43 +1000)]
drm/nouveau/device: tweak the device/subdev relationship a little

Fixes not-in-use engines not having their reset() method called on
resume.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/device: enable proper constructor/destructor
Ben Skeggs [Thu, 25 Apr 2013 01:35:18 +0000 (11:35 +1000)]
drm/nouveau/device: enable proper constructor/destructor

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/device: have engine object initialised before creation
Ben Skeggs [Thu, 25 Apr 2013 07:56:03 +0000 (17:56 +1000)]
drm/nouveau/device: have engine object initialised before creation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/device: convert to engine, rather than subdev
Ben Skeggs [Thu, 25 Apr 2013 07:23:43 +0000 (17:23 +1000)]
drm/nouveau/device: convert to engine, rather than subdev

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv50-/disp: use self as parent for subobjects
Ben Skeggs [Wed, 24 Apr 2013 08:04:22 +0000 (18:04 +1000)]
drm/nv50-/disp: use self as parent for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv50-/fifo: use parent as self for subobjects
Ben Skeggs [Wed, 24 Apr 2013 08:02:35 +0000 (18:02 +1000)]
drm/nv50-/fifo: use parent as self for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv20-nv30/gr: use parent as self for subobjects
Ben Skeggs [Wed, 24 Apr 2013 07:55:28 +0000 (17:55 +1000)]
drm/nv20-nv30/gr: use parent as self for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0-/gr: use self as parent for subobjects
Ben Skeggs [Wed, 24 Apr 2013 07:51:20 +0000 (17:51 +1000)]
drm/nvc0-/gr: use self as parent for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv04-nv40/instmem: use self as parent for subobjects
Ben Skeggs [Wed, 24 Apr 2013 07:48:56 +0000 (17:48 +1000)]
drm/nv04-nv40/instmem: use self as parent for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv04-nv40/vm: use self as parent for subobjects
Ben Skeggs [Wed, 24 Apr 2013 07:46:54 +0000 (17:46 +1000)]
drm/nv04-nv40/vm: use self as parent for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv50-/bar: use self as parent for subobjects
Ben Skeggs [Wed, 24 Apr 2013 07:44:54 +0000 (17:44 +1000)]
drm/nv50-/bar: use self as parent for subobjects

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv04-nv40/instmem: remove parent deref hack
Ben Skeggs [Wed, 24 Apr 2013 07:38:38 +0000 (17:38 +1000)]
drm/nv04-nv40/instmem: remove parent deref hack

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/i2c: remove parent deref hack
Ben Skeggs [Wed, 24 Apr 2013 07:36:57 +0000 (17:36 +1000)]
drm/nouveau/i2c: remove parent deref hack

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/core: rebase object ref/use counts after ctor/init/fini events
Ben Skeggs [Wed, 24 Apr 2013 07:34:18 +0000 (17:34 +1000)]
drm/nouveau/core: rebase object ref/use counts after ctor/init/fini events

This is intended to support named (with a handle, etc) objects having
children that don't have an outside reference.

This will replace the various hacks around the place where subdev
objects have children, and have to manually drop the self-refs so
that they can be destroyed etc when all the outside refs have gone.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv50/disp: inform core when we're not creating a new context
Ben Skeggs [Thu, 25 Apr 2013 14:12:59 +0000 (00:12 +1000)]
drm/nv50/disp: inform core when we're not creating a new context

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/therm: send some messages to debug level
Ben Skeggs [Wed, 24 Apr 2013 03:14:02 +0000 (13:14 +1000)]
drm/nouveau/therm: send some messages to debug level

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nve0/gr: add handling for a bunch of PGRAPH traps
Christoph Bumiller [Wed, 27 Mar 2013 20:59:11 +0000 (21:59 +0100)]
drm/nve0/gr: add handling for a bunch of PGRAPH traps

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/mc: handle irq-related setup ourselves
Ben Skeggs [Fri, 22 Mar 2013 00:05:03 +0000 (10:05 +1000)]
drm/nouveau/mc: handle irq-related setup ourselves

We need to be able to process interrupts before the DRM code is able to
actually enable them, set it up ourselves.  Also, it's less convoluted
to *not* use the DRM wrappers it appears...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv04/disp: hide all the cruft away in its own little hole
Ben Skeggs [Thu, 21 Mar 2013 05:45:11 +0000 (15:45 +1000)]
drm/nv04/disp: hide all the cruft away in its own little hole

It'd be pretty awesome if someone would care enough to port this all
properly to a class interface, perhaps submitting a command stream to
the core via a sw object on PFIFO (emulating how EVO works basically,
and also what nvidia have done forever..)..

But, this seems unlikely given how old this hardware is now, so, lets
just hide it away.

There's a heap of other bits and pieces laying around that are still
tangled.  I'll (re)move them in pieces.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/bios: add missing newline on IO*_OR opcode debugging
Ben Skeggs [Fri, 19 Apr 2013 00:22:09 +0000 (10:22 +1000)]
drm/nouveau/bios: add missing newline on IO*_OR opcode debugging

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/bios: suppress some parser errors when dry-running scripts
Ben Skeggs [Mon, 15 Apr 2013 04:47:05 +0000 (14:47 +1000)]
drm/nouveau/bios: suppress some parser errors when dry-running scripts

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nve0: magic up some support for GF117
Ben Skeggs [Sat, 30 Mar 2013 11:56:26 +0000 (21:56 +1000)]
drm/nve0: magic up some support for GF117

Seen in the wild, don't have the hardware but this hacks things up to
treat it the same as GF119 for now.

Should be relatively safe, I'd be very surprised if anything major
changed outside of PGRAPH.  PGRAPH (3D etc) is disabled by default
however until it's confirmed working.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0-/ltcg: Fix build on 32-bit platforms (v2)
Roy Spliet [Sun, 14 Apr 2013 11:48:45 +0000 (13:48 +0200)]
drm/nvc0-/ltcg: Fix build on 32-bit platforms (v2)

v2: read, don't assume.. *puts on brown paper bag*

Signed-off-by: Roy Spliet <r.spliet@student.tudelft.nl>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/drm: adding support for backlight control of GT525M (NVC0)
Florian Scholz [Tue, 2 Apr 2013 21:16:38 +0000 (23:16 +0200)]
drm/nouveau/drm: adding support for backlight control of GT525M (NVC0)

This patch adds support for the backlight control of the NVIDIA GT
525M, which identifies itself as a member of the NVC0 family.

v2. Extended to handle Kepler too

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/drm: bump the driver version to 1.1.1 to report new features
Christoph Bumiller [Wed, 27 Mar 2013 21:25:52 +0000 (22:25 +0100)]
drm/nouveau/drm: bump the driver version to 1.1.1 to report new features

Signed-off-by: Christoph Bumiller <e0425955@student.tuwien.ac.at>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0/gr: add software methods to control some MP regs
Christoph Bumiller [Wed, 27 Mar 2013 21:16:55 +0000 (22:16 +0100)]
drm/nvc0/gr: add software methods to control some MP regs

Signed-off-by: Christoph Bumiller <e0425955@student.tuwien.ac.at>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0-: support NOUVEAU_GETPARAM_GRAPH_UNITS
Christoph Bumiller [Wed, 27 Mar 2013 21:16:54 +0000 (22:16 +0100)]
drm/nvc0-: support NOUVEAU_GETPARAM_GRAPH_UNITS

Signed-off-by: Christoph Bumiller <e0425955@student.tuwien.ac.at>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0-/ltcg: implement VRAM compression
Christoph Bumiller [Wed, 27 Mar 2013 21:16:53 +0000 (22:16 +0100)]
drm/nvc0-/ltcg: implement VRAM compression

Signed-off-by: Christoph Bumiller <e0425955@student.tuwien.ac.at>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/ttm: allow tiled memtype on system memory buffer objects
Ben Skeggs [Fri, 22 Mar 2013 02:12:17 +0000 (12:12 +1000)]
drm/nouveau/ttm: allow tiled memtype on system memory buffer objects

Compression not supported, and will be silently dropped.  Original G80
can't handle this either and requires LINEAR memtype, though it's still
possible to correctly texture and m2mf to/from these objects anyway.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nve0/grctx: initial attempt at unhardcoding yet more magic
Ben Skeggs [Wed, 20 Mar 2013 01:34:32 +0000 (11:34 +1000)]
drm/nve0/grctx: initial attempt at unhardcoding yet more magic

Not sure about the (gpc_nr == 1) condition, it's probably wrong but for
all the examples I've seen so far it matches what NVIDIA end up poking.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nv50/therm: implement temperature reading
Martin Peres [Sat, 23 Feb 2013 16:15:18 +0000 (17:15 +0100)]
drm/nv50/therm: implement temperature reading

Signed-off-by: Martin Peres <martin.peres@labri.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/therm: split the nv50 and nv84 code
Martin Peres [Sat, 23 Feb 2013 15:45:51 +0000 (16:45 +0100)]
drm/nouveau/therm: split the nv50 and nv84 code

This is needed because temperature management on nv50 can be enabled and it
looks about the same as nv40.

Signed-off-by: Martin Peres <martin.peres@labri.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/drm: fix crash in vram manager debug callback
Marcin Slusarz [Sat, 2 Mar 2013 19:00:31 +0000 (20:00 +0100)]
drm/nouveau/drm: fix crash in vram manager debug callback

It's probably impossible to hit it now on mainline kernel.
I only noticed it because one of my debugging patches uses it.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nvc0/fb: ignore readback page alloc failure to support userspace
Ben Skeggs [Wed, 16 Jan 2013 00:05:31 +0000 (10:05 +1000)]
drm/nvc0/fb: ignore readback page alloc failure to support userspace

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agodrm/nouveau/disp: fix uninitialised eq_done in error path
Ben Skeggs [Thu, 11 Apr 2013 00:12:48 +0000 (10:12 +1000)]
drm/nouveau/disp: fix uninitialised eq_done in error path

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
12 years agoALSA: usb - Avoid unnecessary sample rate changes on USB 2.0 clock sources
David Henningsson [Thu, 25 Apr 2013 12:32:24 +0000 (14:32 +0200)]
ALSA: usb - Avoid unnecessary sample rate changes on USB 2.0 clock sources

The Scarlett 2i2 seems to take almost 500 ms to set the sample rate,
even if the clock is currently set to that value. This patch speeds
up prepare of the device, by avoiding setting the clock to something
it already is.

Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agosound: oss/dmabuf: use dma_map_single
Arnd Bergmann [Thu, 25 Apr 2013 17:28:50 +0000 (19:28 +0200)]
sound: oss/dmabuf: use dma_map_single

The virt_to_bus/bus_to_virt functions have been deprecated
for as long as I can remember, and they are used in very
few remaining instances, usually in obscure ISA device
drivers. The OSS sound drivers are the only ones that are
still used on the ARM architecture, and only on some of
the earliest StrongARM machines.

The problem for converting the OSS subsystem to use
dma_map_single instead is that the caller of virt_to_bus
does not have a device pointer, since the subsystem has
never been ported to use the common device infrastructure.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agoALSA: ali5451: use mdelay instead of large udelay constants
Arnd Bergmann [Thu, 25 Apr 2013 17:28:49 +0000 (19:28 +0200)]
ALSA: ali5451: use mdelay instead of large udelay constants

ARM cannot handle udelay for more than 2 miliseconds, so we
should use mdelay instead for those.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agofbdev: fix check for fb_mmap's mmio availability
Tomi Valkeinen [Wed, 24 Apr 2013 05:35:20 +0000 (08:35 +0300)]
fbdev: fix check for fb_mmap's mmio availability

Commit fc9bbca8f650e5f738af8806317c0a041a48ae4a (vm: convert fb_mmap to
vm_iomap_memory() helper) made fbmem.c use vm_iomap_memory, but also
accidentally removed the check for mmio's availability.

Add the check back.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
12 years agofbdev: improve fb_mmap bounds checks
Tomi Valkeinen [Thu, 18 Apr 2013 05:39:47 +0000 (08:39 +0300)]
fbdev: improve fb_mmap bounds checks

Improve fb_mmap bounds checks in gbefb, smscufx, udlfb and vfb drivers to
prevent possible uint overflows.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Steve Glendinning <steve.glendinning@smsc.com>
Cc: Bernie Thompson <bernie@plugable.com>
12 years agofbdev/ps3fb: use vm_iomap_memory()
Tomi Valkeinen [Thu, 18 Apr 2013 04:52:42 +0000 (07:52 +0300)]
fbdev/ps3fb: use vm_iomap_memory()

Use vm_iomap_memory() instead of [io_]remap_pfn_range().
vm_iomap_memory() gives us much simpler API to map memory to userspace,
and reduces possibilities for bugs.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
12 years agoixgbe: add mac type to the version in ethtool_regs
Emil Tantilov [Fri, 19 Apr 2013 09:31:17 +0000 (09:31 +0000)]
ixgbe: add mac type to the version in ethtool_regs

This patch adds the mac type to the version in ethtool_regs.

This will make it easier to check the mac type when dumping registers with
ethtool. The drawback of this is that older versions of ethtool will only
be able to dump in hex format for 82599 and above  when used with the updated
driver.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbe: add support for disabling link at boot time on 82599
Emil Tantilov [Fri, 12 Apr 2013 08:36:47 +0000 (08:36 +0000)]
ixgbe: add support for disabling link at boot time on 82599

This patch adds support for disabling link during boot time. This
feature was requested by customers and is configurable through the EEPROM.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbe: cache AUTOC reads
Emil Tantilov [Fri, 12 Apr 2013 08:36:42 +0000 (08:36 +0000)]
ixgbe: cache AUTOC reads

This patch removes majority of the AUTOC register reads by using a cached
value instead.

The reason for this change is to avoid writing corrupted values to AUTOC
due to bad FW.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbe: fix register access during ethtool loopback test
Emil Tantilov [Fri, 12 Apr 2013 02:10:25 +0000 (02:10 +0000)]
ixgbe: fix register access during ethtool loopback test

This patch cleans up the logic in ixgbe_setup_loopback_test() to only access
registers applicable to the MAC type. AUTOC is only valid on MACs older than
X540. MACC is used for X540.

In addition it removes a read of AUTOC and uses the stored value to force the
link up.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbe: fix EICR write in ixgbe_msix_other
Jacob Keller [Sat, 2 Mar 2013 07:51:42 +0000 (07:51 +0000)]
ixgbe: fix EICR write in ixgbe_msix_other

Previously, the ixgbe_msix_other was writing the full 32bits of the set
interrupts, instead of only the ones which the ixgbe_msix_other is
handling. This resulted in a loss of performance when the X540's PPS feature is
enabled due to sometimes clearing queue interrupts which resulted in the driver
not getting the interrupt for cleaning the q_vector rings often enough. The fix
is to simply mask the lower 16bits off so that this handler does not write them
in the EICR, which causes them to remain high and be properly handled by the
clean_rings interrupt routine as normal.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Cc: stable <stable@vger.kernel.org>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: limit udelay for phy changes to 10000us
Andi Kleen [Mon, 22 Apr 2013 07:46:40 +0000 (07:46 +0000)]
igb: limit udelay for phy changes to 10000us

If you really want 100000us you should really use mdelay or so.

Found by the LTO kernel build

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoe1000e: panic caused by Rx traffic arriving while interface going down
Bruce Allan [Sat, 20 Apr 2013 05:37:29 +0000 (05:37 +0000)]
e1000e: panic caused by Rx traffic arriving while interface going down

An "unable to handle kernel paging request" panic can occur when receiving
traffic while the interface is going down.  Wait for NAPI to be done with
current context after disabling interrupts and then disable NAPI.

See https://bugzilla.vyatta.com/show_bug.cgi?id=8837.

Reported-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agof2fs: check nid == 0 in add_free_nid
Jaegeuk Kim [Thu, 25 Apr 2013 04:21:12 +0000 (13:21 +0900)]
f2fs: check nid == 0 in add_free_nid

It is more obvious that add_free_nid checks whether the free nid is zero or not.

Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
12 years agof2fs: add REQ_META about metadata requests for submit
Namjae Jeon [Thu, 25 Apr 2013 02:45:21 +0000 (11:45 +0900)]
f2fs: add REQ_META about metadata requests for submit

Adding REQ_META for all the metadata requests can help in improving the
FS performance, if the underlying device supports TAGGING.
So, when considering the submit_bio path for all the f2fs requests. We can
add REQ_META for all the META requests.
As a precursor to this change we considered the commit
4265900e0be653f5b78baf2816857ef57cf1332f 'mmc: MMC-4.5 Data Tag Support'

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
12 years agof2fs: give a chance to merge IOs by IO scheduler
Jaegeuk Kim [Wed, 24 Apr 2013 04:19:56 +0000 (13:19 +0900)]
f2fs: give a chance to merge IOs by IO scheduler

Previously, background GC submits many 4KB read requests to load victim blocks
and/or its (i)node blocks.

...
f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb61, blkaddr = 0x3b964ed
f2fs_gc : block_rq_complete: 8,16 R () 499854968 + 8 [0]
f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb6f, blkaddr = 0x3b964ee
f2fs_gc : block_rq_complete: 8,16 R () 499854976 + 8 [0]
f2fs_gc : f2fs_readpage: ino = 1, page_index = 0xb79, blkaddr = 0x3b964ef
f2fs_gc : block_rq_complete: 8,16 R () 499854984 + 8 [0]
...

However, by the fact that many IOs are sequential, we can give a chance to merge
the IOs by IO scheduler.
In order to do that, let's use blk_plug.

...
f2fs_gc : f2fs_iget: ino = 143
f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c6, blkaddr = 0x2e6ee
f2fs_gc : f2fs_iget: ino = 143
f2fs_gc : f2fs_readpage: ino = 143, page_index = 0x1c7, blkaddr = 0x2e6ef
<idle> : block_rq_complete: 8,16 R () 1519616 + 8 [0]
<idle> : block_rq_complete: 8,16 R () 1519848 + 8 [0]
<idle> : block_rq_complete: 8,16 R () 1520432 + 96 [0]
<idle> : block_rq_complete: 8,16 R () 1520536 + 104 [0]
<idle> : block_rq_complete: 8,16 R () 1521008 + 112 [0]
<idle> : block_rq_complete: 8,16 R () 1521440 + 152 [0]
<idle> : block_rq_complete: 8,16 R () 1521688 + 144 [0]
<idle> : block_rq_complete: 8,16 R () 1522128 + 192 [0]
<idle> : block_rq_complete: 8,16 R () 1523256 + 328 [0]
...

Note that this issue should be addressed in checkpoint, and some readahead
flows too.

Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
12 years agof2fs: avoid frequent background GC
Jaegeuk Kim [Wed, 24 Apr 2013 04:00:14 +0000 (13:00 +0900)]
f2fs: avoid frequent background GC

If there is no victim segments selected by background GC, let's wait
a little bit longer time to collect dirty segments.
By default, let's give 5 minutes.

Reviewed-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
12 years agoe1000e: fix numeric overflow in phc settime method
Richard Cochran [Tue, 23 Apr 2013 01:56:34 +0000 (01:56 +0000)]
e1000e: fix numeric overflow in phc settime method

The PTP Hardware Clock settime function in the e1000e driver
computes nanoseconds from a struct timespec. The code converts the
seconds field .tv_sec by multiplying it with NSEC_PER_SEC. However,
both operands are of type long, resulting in an unintended overflow.
The patch fixes the issue by using the helper function from time.h.

CC: stable <stable@vger.kernel.org>
Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agodrm/edid: Check both 60Hz and 59.94Hz when looking for a CEA mode
Ville Syrjälä [Wed, 24 Apr 2013 16:07:18 +0000 (19:07 +0300)]
drm/edid: Check both 60Hz and 59.94Hz when looking for a CEA mode

drm_match_cea_mode() should be able to match both the 60Hz version,
and the 59.94Hz version of modes.

We only store one pixel clock value per mode in edid_cea_modes, so the
other value must be calculated. Depending on the mode, edid_cea_modes
contains the pixel clock for either the 60Hz version or the 59.94Hz
version, so a bit of care is needed so that the calculation produces
the correct result.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=46800
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/edid: Populate vrefresh for CEA modes
Ville Syrjälä [Wed, 24 Apr 2013 16:07:17 +0000 (19:07 +0300)]
drm/edid: Populate vrefresh for CEA modes

Well have use for the vrefresh information of CEA modes later. Just
populate the information into the table to avoid having to calculate
it.

I'm too lazy to check if someone relies on newly allocated CEA
modes having 0 vrefresh, so just clear vrefresh back to 0 when
adding the mode to the connector's modelist.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm: Add drm_mode_equal_no_clocks()
Ville Syrjälä [Wed, 24 Apr 2013 16:07:16 +0000 (19:07 +0300)]
drm: Add drm_mode_equal_no_clocks()

drm_mode_equal_no_clocks() is like drm_mode_equal() except it doesn't
compare the clock or vrefresh values. drm_mode_equal() is now
implemented by first doing the clock checks, and then calling
drm_mode_equal_no_clocks().

v2: Add missing EXPORT_SYMBOL()

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm: Remove explicit vrefresh initialization from DRM_MODE()
Ville Syrjälä [Wed, 24 Apr 2013 16:07:15 +0000 (19:07 +0300)]
drm: Remove explicit vrefresh initialization from DRM_MODE()

No need to zero initialize .vrefresh in DRM_MODE() since it's using
desgignated initializers.

This will also avoid some duplicate initialization warnings later.

Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/shmobile: Fix race condition between page flip request and handler
Laurent Pinchart [Tue, 12 Mar 2013 14:38:43 +0000 (15:38 +0100)]
drm/shmobile: Fix race condition between page flip request and handler

The page flip handler stores the page flip event pointer and then calls
drm_vblank_get() to enable the vblank interrupt. Due to the vblank off
delay, the vblank interrupt can be enabled in the hardware at that
point, even if the vblank reference count is equal to 0. If a vblank
interrupt is triggered between storing the event pointer and calling
drm_vblank_get(), the page flip completion handler will process the
event and call drm_vblank_put() with a reference count equal to 0. This
will result in a BUG_ON.

Fix the race condition by calling drm_vblank_get() before storing the
event pointer.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm: export drm_vm_open_locked
Arnd Bergmann [Thu, 25 Apr 2013 17:28:52 +0000 (19:28 +0200)]
drm: export drm_vm_open_locked

The EXYNOS DRM driver uses drm_vm_open_locked in its mmap() function,
and it can be built as a loadable module, which currently fails.
This exports the symbol from the DRM core to avoid

ERROR: "drm_vm_open_locked" [drivers/gpu/drm/exynos/exynosdrm.ko] undefined!

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Inki Dae <inki.dae@samsung.com>
Cc: Joonyoung Shim <jy0922.shim@samsung.com>
Cc: Seung-Woo Kim <sw0312.kim@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/tilcdc: use only a single module device table
Arnd Bergmann [Tue, 23 Apr 2013 16:30:43 +0000 (18:30 +0200)]
drm/tilcdc: use only a single module device table

The tilcdc driver fails to be built as a module because of extraneous
MODULE_DEVICE_TABLE entries:

drivers/gpu/drm/tilcdc/tilcdc_slave.o:(.data+0x54): multiple definition of `__mod_of_device_table'
drivers/gpu/drm/tilcdc/tilcdc_tfp410.o:(.data+0x54): first defined here
drivers/gpu/drm/tilcdc/tilcdc_panel.o:(.data+0x54): multiple definition of `__mod_of_device_table'
drivers/gpu/drm/tilcdc/tilcdc_tfp410.o:(.data+0x54): first defined here
drivers/gpu/drm/tilcdc/tilcdc_drv.o:(.data+0x184): multiple definition of `__mod_of_device_table'
drivers/gpu/drm/tilcdc/tilcdc_tfp410.o:(.data+0x54): first defined here

Since the entire point of these entries is to make the module autoload
when one of the devices is present, it's enough to keep the one entry
for "ti,am33xx-tilcdc", which should always be there if any of the
others are.

Acked-by: Rob Clark <robdclark@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agoMerge branch 'gma500-fixes' of git://github.com/patjak/drm-gma500 into drm-next
Dave Airlie [Fri, 26 Apr 2013 00:16:10 +0000 (10:16 +1000)]
Merge branch 'gma500-fixes' of git://github.com/patjak/drm-gma500 into drm-next

Two fixes for gma500. First one from Anisse allows us to handle ASLE irqs even
when BIOS doesn't trigger a pipe event irq. The second one allows dual head
setups to have a big shared framebuffer.

* 'gma500-fixes' of git://github.com/patjak/drm-gma500:
  drm/gma500: Increase max resolution for mode setting
  drm/gma500: fix backlight hotkeys behaviour on netbooks

12 years agovfio: Convert container->group_lock to rwsem
Alex Williamson [Thu, 25 Apr 2013 22:12:38 +0000 (16:12 -0600)]
vfio: Convert container->group_lock to rwsem

All current users are writers, maintaining current mutual exclusion.
This lets us add read users next.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
12 years agoMerge tag 'efi-urgent' into x86/urgent
H. Peter Anvin [Thu, 25 Apr 2013 21:00:22 +0000 (14:00 -0700)]
Merge tag 'efi-urgent' into x86/urgent

 * The EFI variable anti-bricking algorithm merged in -rc8 broke booting
   on some Apple machines because they implement EFI spec 1.10, which
   doesn't provide a QueryVariableInfo() runtime function and the logic
   used to check for the existence of that function was insufficient.
   Fix from Josh Boyer.

 * The anti-bricking algorithm also introduced a compiler warning on
   32-bit. Fix from Borislav Petkov.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
12 years agoparisc: use spin_lock_irqsave/spin_unlock_irqrestore for PTE updates
John David Anglin [Tue, 23 Apr 2013 20:42:07 +0000 (22:42 +0200)]
parisc: use spin_lock_irqsave/spin_unlock_irqrestore for PTE updates

User applications running on SMP kernels have long suffered from instability
and random segmentation faults.  This patch improves the situation although
there is more work to be done.

One of the problems is the various routines in pgtable.h that update page table
entries use different locking mechanisms, or no lock at all (set_pte_at).  This
change modifies the routines to all use the same lock pa_dbit_lock.  This lock
is used for dirty bit updates in the interruption code. The patch also purges
the TLB entries associated with the PTE to ensure that inconsistent values are
not used after the page table entry is updated.  The UP and SMP code are now
identical.

The change also includes a minor update to the purge_tlb_entries function in
cache.c to improve its efficiency.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Helge Deller <deller@gmx.de>
12 years agoparisc: disable -mlong-calls compiler option for kernel modules
Helge Deller [Tue, 23 Apr 2013 19:29:03 +0000 (21:29 +0200)]
parisc: disable -mlong-calls compiler option for kernel modules

CONFIG_MLONGCALLS was introduced in commit
ec758f98328da3eb933a25dc7a2eed01ef44d849 to overcome linker issues when linking
huge linux kernels, e.g. with many modules linked in.

But in the kernel module loader there is no support yet for the new relocation
types, which is why modules built with -mlong-calls can't be loaded.
Furthermore, for modules long calls are not really necessary, since we already
use stub sections which resolve long distance calls.

So, let's just disable this compiler option when compiling kernel modules.

Signed-off-by: Helge Deller <deller@gmx.de>
12 years agoparisc: uaccess: fix compiler warnings caused by __put_user casting
Will Deacon [Mon, 22 Apr 2013 12:53:43 +0000 (12:53 +0000)]
parisc: uaccess: fix compiler warnings caused by __put_user casting

When targetting 32-bit processors, __put_user emits a pair of stw
instructions for the 8-byte case. If the type of __val is a pointer, the
marshalling code casts it to the wider integer type of u64, resulting
in the following compiler warnings:

  kernel/signal.c: In function 'copy_siginfo_to_user':
  kernel/signal.c:2752:11: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
  kernel/signal.c:2752:11: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
  [...]

This patch fixes the warnings by removing the marshalling code and using
the correct output modifiers in the __put_{user,kernel}_asm64 macros
so that GCC will allocate the right registers without the need to
extract the two words explicitly.

Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Helge Deller <deller@gmx.de>
12 years agoparisc: Change kunmap macro to static inline function
John David Anglin [Tue, 23 Apr 2013 00:23:50 +0000 (00:23 +0000)]
parisc: Change kunmap macro to static inline function

Change kunmap macro to static inline function to fix build error
compiling drivers/base/dma-buf.c.

Without the change, the following error can occur:

   CC      drivers/base/dma-buf.o
drivers/base/dma-buf.c: In function 'dma_buf_kunmap':
drivers/base/dma-buf.c:427:46:
error: macro "kunmap" passed 3 arguments, but takes just 1

I believe parisc is the only arch to implement kunmap using a macro.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Helge Deller <deller@gmx.de>
12 years agoparisc: Provide __ucmpdi2 to resolve undefined references in 32 bit builds.
John David Anglin [Sat, 20 Apr 2013 19:41:06 +0000 (19:41 +0000)]
parisc: Provide __ucmpdi2 to resolve undefined references in 32 bit builds.

The Debian experimental linux source package (3.8.5-1) build fails
with the following errors:
...
MODPOST 2016 modules
ERROR: "__ucmpdi2" [fs/btrfs/btrfs.ko] undefined!
ERROR: "__ucmpdi2" [drivers/md/dm-verity.ko] undefined!

The attached patch resolves this problem.  It is based on the s390
implementation of ucmpdi2.c.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Signed-off-by: Helge Deller <deller@gmx.de>
12 years agodrm/gma500: Increase max resolution for mode setting
Patrik Jakobsson [Thu, 25 Apr 2013 20:23:36 +0000 (22:23 +0200)]
drm/gma500: Increase max resolution for mode setting

By having a higher max resolution we can now set up a virtual
framebuffer that spans several monitors. 4096 should be ok since we're
gen 3 or higher and should be enough for most dual head setups.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
12 years agodrm/gma500: fix backlight hotkeys behaviour on netbooks
Anisse Astier [Wed, 24 Apr 2013 15:36:01 +0000 (17:36 +0200)]
drm/gma500: fix backlight hotkeys behaviour on netbooks

Backlight hotkeys weren't working before on certain cedartrail laptops.

The source of this problem is that the hotkeys' ASLE opregion interrupts
were simply ignored. Driver seemed to expect the interrupt to be
associated with a pipe, but it wasn't.

Accepting the ASLE interrupt without an associated pipe event flag fixes
the issue, the backlight code is called when needed, making the
brightness keys work properly.

[patrik: This patch affects irq handling on any netbook with opregion support]

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=833597
Reference: http://lists.freedesktop.org/archives/dri-devel/2012-July/025279.html
Cc: stable@kernel.org
Signed-off-by: Anisse Astier <anisse@astier.eu>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
12 years agoDocumentation/devicetree: make semantic of initrd-end more explicit
Uwe Kleine-König [Thu, 25 Apr 2013 12:27:57 +0000 (14:27 +0200)]
Documentation/devicetree: make semantic of initrd-end more explicit

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
12 years ago[media] s5c73m3: fix indentation of the help section in Kconfig
Gianluca Gennari [Thu, 25 Apr 2013 13:46:38 +0000 (10:46 -0300)]
[media] s5c73m3: fix indentation of the help section in Kconfig

The 'help' section of the Kconfig entry for this driver is missing
an extra alignment. That seems to violate what's stated at:
Documentation/kbuild/kconfig-language.txt

Even if it works, the better is to add 2 extra spaces there, as
this is the common practice and helps human reading of the file.

Also, the way it is, it breaks backport trees.

Signed-off-by: Gianluca Gennari <gennarone@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] cx25821-alsa: get rid of a __must_check warning
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:58:44 +0000 (15:58 -0300)]
[media] cx25821-alsa: get rid of a __must_check warning

The hole reason for __must_check is to not ignore an error.

However, a "ret" value is used at cx25821 just to avoid the
Kernel compilation to compain about it.

That, however, produces another warning (with W=1):

drivers/media/pci/cx25821/cx25821-alsa.c: In function 'cx25821_audio_fini':
drivers/media/pci/cx25821/cx25821-alsa.c:727:6: warning: variable 'ret' set but not used [-Wunused-but-set-variable]

With the current implementation of driver_for_each_device() and
cx25821_alsa_exit_callback(), there's actually just one
very unlikely condition where it will currently produce
an error: if driver_find() returns NULL.

Ok, there's not much that can be done, as it is on a driver's
function that returns void, but it can at least print some message
if the error happens.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] cx25821-video: declare cx25821_vidioc_s_std as static
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:49:33 +0000 (15:49 -0300)]
[media] cx25821-video: declare cx25821_vidioc_s_std as static

Fixes the following warning:

drivers/media/pci/cx25821/cx25821-video.c: At top level:
drivers/media/pci/cx25821/cx25821-video.c:766:5: warning: no previous prototype for 'cx25821_vidioc_s_std' [-Wmissing-prototypes]

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] cx25821-video: remove maxw from cx25821_vidioc_try_fmt_vid_cap
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:46:36 +0000 (15:46 -0300)]
[media] cx25821-video: remove maxw from cx25821_vidioc_try_fmt_vid_cap

After cx25821-video cleanup, this var is not used anymore:

drivers/media/pci/cx25821/cx25821-video.c: In function 'cx25821_vidioc_try_fmt_vid_cap':
drivers/media/pci/cx25821/cx25821-video.c:591:15: warning: variable 'maxw' set but not used [-Wunused-but-set-variable]

as the code now checks the max width as the default case for the
range check.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] r820t: Remove a warning for an unused value
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:40:21 +0000 (15:40 -0300)]
[media] r820t: Remove a warning for an unused value

Currently, the driver complains about the pre_detect var:

drivers/media/tuners/r820t.c: In function 'r820t_sysfreq_sel':
drivers/media/tuners/r820t.c:722:31: warning: variable 'pre_dect' set but not used [-Wunused-but-set-variable]

While rtl8232 code comments it, perhaps some other driver may use.
So, the better is to keep the code there, allowing to enable it
via r820t config data.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] dib0090: Fix a warning at dib0090_set_EFUSE
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:30:40 +0000 (15:30 -0300)]
[media] dib0090: Fix a warning at dib0090_set_EFUSE

The check if the values for c, h and n are within the range is
always true, as, if one of this values is out of range, the
previous "if" clauses will default to a value within the
range.

That fixes the following warning:

drivers/media/dvb-frontends/dib0090.c: In function 'dib0090_set_EFUSE':
drivers/media/dvb-frontends/dib0090.c:1545:5: warning: comparison is always true due to limited range of data type [-Wtype-limits]

and makes the code easier to read.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] dib8000: fix a warning
Mauro Carvalho Chehab [Thu, 25 Apr 2013 18:22:20 +0000 (15:22 -0300)]
[media] dib8000: fix a warning

drivers/media/dvb-frontends/dib8000.c: In function 'dib8000_wait_lock':
drivers/media/dvb-frontends/dib8000.c:3972:1: warning: 'value' may be used uninitialized in this function [-Wmaybe-uninitialized]
drivers/media/dvb-frontends/dib8000.c:2419:6: note: 'value' was declared here

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years ago[media] dib8000: Fix sub-channel range
Mauro Carvalho Chehab [Thu, 25 Apr 2013 14:15:54 +0000 (11:15 -0300)]
[media] dib8000: Fix sub-channel range

isdbt_sb_subchannel is unsigned with 8 bits. So, it will
never be -1. Instead, any value bigger than 13 is invalid.

As is, the current code generates the following warnings:

drivers/media/dvb-frontends/dib8000.c: In function 'dib8000_set_isdbt_common_channel':
drivers/media/dvb-frontends/dib8000.c:2358:3: warning: comparison is always true due to limited range of data type [-Wtype-limits]
drivers/media/dvb-frontends/dib8000.c: In function 'dib8000_tune':
drivers/media/dvb-frontends/dib8000.c:3107:8: warning: comparison is always false due to limited range of data type [-Wtype-limits]
drivers/media/dvb-frontends/dib8000.c:3153:9: warning: comparison is always false due to limited range of data type [-Wtype-limits]
drivers/media/dvb-frontends/dib8000.c:3160:5: warning: comparison is always false

It should also be noticed that ARIB STD-B31, item
"3.15.6.8 Number of segments" at TMCC table defines the
value 15 for unused segment, and 14 as reserved.

So, better to change the check to consider any value
bigger than 13 to mean that sub-channels should be
disabled, fixing the warning and doing the right thing
even if an invalid value is filled by userspace.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
12 years agocpufreq: pxa2xx: initialize variables
Arnd Bergmann [Thu, 25 Apr 2013 17:28:57 +0000 (19:28 +0200)]
cpufreq: pxa2xx: initialize variables

gcc-3.8 correctly found that the variables set by find_freq_tables()
are not initialized if this function is called on something other
than a pxa2xx or pxa3xx:

pxa2xx-cpufreq.c: In function 'pxa_verify_policy':
pxa2xx-cpufreq.c:272:6: warning: 'pxa_freqs_table' may be used uninitialized in this function [-Wmaybe-uninitialized]
pxa2xx-cpufreq.c: In function 'pxa_set_target':
pxa2xx-cpufreq.c:345:23: warning: 'pxa_freq_settings' may be used uninitialized in this function [-Wmaybe-uninitialized]

Rather than adding a bogus initialization that would let us
get a little further before crashing, add an explicit BUG().
We know that this code is designed to run on only these cpus,
so this will fix the build warning and give a more helpful
diagnostic if the code ever changes to run on other machines.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
12 years agoUSB: ehci-msm: USB_MSM_OTG needs USB_PHY
Arnd Bergmann [Thu, 25 Apr 2013 17:29:01 +0000 (19:29 +0200)]
USB: ehci-msm: USB_MSM_OTG needs USB_PHY

The Kconfig entry for USB_EHCI_MSM unconditionally selects USB_MSM_OTG,
which is now only visible when USB_PHY is also enabled.

This adds an appropriate dependency and enables USB_PHY in the msm
defconfig, avoiding the Kbuild warning:

warning: (USB_EHCI_MSM) selects USB_MSM_OTG which has unmet direct dependencies (USB_SUPPORT && USB_PHY && (USB || USB_GADGET) && ARCH_MSM)

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Felipe Balbi <balbi@ti.com>
Cc: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
12 years agoUSB: OHCI: avoid conflicting platform drivers
Arnd Bergmann [Thu, 25 Apr 2013 17:29:04 +0000 (19:29 +0200)]
USB: OHCI: avoid conflicting platform drivers

Like the EHCI driver, OHCI supports a large number of different platform
glue drivers by directly including them, which causes problems with
conflicting macro definitions in some cases. As more ARM architecture
specific back-ends are required to coexist in a single build, we should
split those out into separate drivers. Unfortunately, the infrastructure
for that is still under development, so to give us more time, this uses
a separate *_PLATFORM_DRIVER macro for each ARM specific OHCI backend,
just like we already do on PowerPC and some of the other ARM platforms.

In linux-3.10, only the SPEAr and CNS3xxx back-ends would actually conflict
without this patch, but over time we would get more of them, so this
is a way to avoid having to patch the driver every time it breaks. We
should still split out all back-ends into separate loadable modules,
but that work is only needed to improve code size and cleanliness after
this patch, not for correctness.

While we're here, this fixes the incorrectly sorted error path
for the OMAP1 and OMAP3 backends to ensure we always unregister
the exact set of drivers that were registered before erroring out.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>