review.tizen.org Git - kernel/kernel-mfld-blackbay.git/log

gfx: pvr: fix corrupted command trace for SGX transfer command

When the transfer command is called with the SGXMKIF_TQFLAGS_KEEPPENDING
source and destination sync counters were not zeroed in the
corresponding trace record, leading to a corrupted trace output.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Tested-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Add trace events for powermanagement

SGX may lockup because of incorrectly timed or missing power state
trasition for SGX or display controller.

To make future debugging simpler for these cases add command trace
entries for realeted power management changes.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Add command trace points to flip

To make easier to debug SGX lockups with page flip code we need command
trace to have page flip events. There already exists API to add trace
events for flip chain. We can use same API for drm mode page flips.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: overlay: Set CC_OUT bit in OCONFIG

Unfortunately the docs aren't exactly clear on when the bit should be set.
However all user space solutions I've seen set the bit, so let's follow suit.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: overlay: More thorough fix for SWIDTHSW issues

The current code can still misprogram SWIDTHSW if fb->offsets[] is
suitably misaligned. Simply limit SWIDTHSW to 2k pixels in such cases.

Change the SWIDTHSW alignment code to be more readable, and align the
width to 32 bytes (SWORD), instead of 64 bytes.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

MUST_REVERT: drm: psb: Added gralloc buffer support for video

Gralloc buffers are created and bound to
TTM buffers independent of the BCD module.

This patch has received many change requests from Pauli and Jani.
Jani Nacked it. However, Sean believes the requests are not critical.
Pierre said that we have to take this patch because it is urgent,
and improve it later. This is why it is marked as MUST_REVERT. Sean is
supposed to send the new version which should be taken instead of this
one, and this commit message should be removed.

Signed-off-by: Naeem M Afzal <naeem.m.afzal@intel.com>
Signed-off-by: Sean V Kelley <sean.v.kelley@intel.com>

MUST_REVERT: gfx: drm: explicitly authenticate for Android

Added as an interim measure to proper client authentication using
DRM_AUTH IOCTLs in user space. Applied to drm_fops.c:drm_open_helper
where there is already a similar hack. Shall be removed upon
userspace implementaiton.

Signed-off-by: Sean V Kelley <sean.v.kelley@intel.com>

gfx: drv: overlay: Fix NV12 chroma SWIDTHSW

NV12 chroma SWIDTHSW must be calculated in a somewhat bizarre way.

Also fix an off-by-one error for other pixel formats.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: remove the outdated README

The README is outdated, and this information is better shared outside of
the kernel tree anyway.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: remove unused uopt (user options)

Remove unused code for uopt, user options.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: pvr: remove unused code in intel-linux directories

The directories pvr/services4/{include,srvkm}/env/linux-intel are not used
for anything. Remove.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

Revert "drm: Protect drm/drm_os_linux.h inclusion with #ifdef"

This reverts commit "drm: Protect drm/drm_os_linux.h inclusion with #ifdef"
to minimize gfx related diff outside drivers/staging/mrst. The header file
<drm/drm_os_linux.h> is not supposed to be included directly anyway.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: remove redundant #include <drm/drm_os_linux.h>

The header file <drm/drm_os_linux.h> is not supposed to be included
directly, it does not have #ifdef protection for multiple inclusion, and
it's included through <drm/drmP.h> anyway. Remove the #include.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: remove obsolete .gitignore

The .gitignore is no longer relevant with Medfield only build.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: make tc35876x independent of tmd vid driver

The TC35876X driver used to be bolted on the TMD panel driver with some
ifs. Make it a more independent driver instead.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: remove redundant dpi_panel_on setting

dpi_panel_on is false at init time, no need to set it again.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

pvr: unifdef pvr power management code for improved readability

The PVR power management code is difficult to read due to excess #ifdef 1
and #ifdef 0. Remove them.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: allow powering down GL3 cache regardless of GL3 config

The commit "psb_video: remove OSPM_GL3_CACHE_ISLAND when not enabled" made
it impossible to power down the GL3 cache island when
CONFIG_MDFD_GL3=n. It's powered on at boot (likely by firmware) so allow
powering down regardless of config.

This fixes the following situation in /sys/kernel/debug/mid_pmu_states when
display is off:

pci 8086 0130 0000:00:02.0             pvrsrvkm:
                      |------->             GFX                   D0i3
                      |------->   Video Decoder                   D0i3
                      |------->   Video Encoder                   D0i3
                      |------->       GL3 Cache                   D0
                      |------->       Display A                   D0i3
                      |------->       Display B                   D0i3
                      |------->       Display C                   D0i3
                      |------->        MIPI-DSI                   D0i3

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: remove another redundant panel state variable gbdispstatus

The dev_priv->dpi_panel_on{,2} fields already track panel state. Use those
instead of gbdispstatus.

The old implementation could not have worked any better for DBI panels than
this one, but leave a comment to revisit this.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: refactor psb_runtime_{idle,suspend}

Reuse psb_runtime_idle for checking whether runtime suspend is possible in
psb_runtime_suspend, making the conditions the same for both. Do HDMI audio
suspend in suspend, not idle.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: tc35876x: remove redundant switching of dpi_panel_on state

The dev_priv->dpi_panel_on{,2} state is flipped accordingly, and on the
correct pipe, by the caller of mdfld_dsi_configure_{up,down}, so don't do
it again.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: tc35876x: make mdfld_dsi_configure_{up,down} more pipe aware

This does not fix everything about pipes in mdfld_dsi_configure_{up,down},
but is a step in the right direction.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: display: tc35876x: remove extra dsi_device_ready state variable

The driver is full of display state variables. Remove dsi_device_ready used
by the tc35876x, and use the existing dev_priv->dpi_panel_on instead.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Clear links when freeing head of list

Linked list is freed in psb_cleanup_pending_events is called just before
head of linked list is freed. All list entries needs to be initialized
to be empty lists to avoid accessing freed memory if logic changes in
future.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: pvr: Fix SGX failing to complete queued rendering

SGXCommandComplete wakeups SGX to render when blocking external
operation completes. If blocking operation completes in IRQ handler
driver only sets flag to wakeup hardware but doesn't schedule the work
to handle wakeup operation.

Issue: ANDROID-529
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Clean pending page flip events when device is closed

If userspace closes the device node with pending page flips driver keeps
a pointer to file private that is already freed. Following backtrace
shows the oops happening.

[ 643.828884] BUG: unable to handle kernel paging request at 6b6b6b6b
[ 643.835082] IP: [<c1224b6e>] __list_add+0x2a/0x5c
[ 643.888465] EIP is at __list_add+0x2a/0x5c
[ 643.943931] Call Trace:
[ 643.946376] [<c139c0d8>] psb_intel_flip_complete+0x71/0xcb
[ 643.951937] [<c139c1c8>] sync_callback+0x96/0x9f
[ 643.956631] [<c1371310>] PVRSRVCallbackOnSync+0x37/0xb7
[ 643.961934] [<c139c132>] ? psb_intel_flip_complete+0xcb/0xcb
[ 643.967673] [<c139c049>] psb_intel_crtc_page_flip+0xa1/0xbf
[ 643.973324] [<c1282f5d>] drm_mode_page_flip_ioctl+0x13f/0x190
[ 643.979151] [<c12778bf>] drm_ioctl+0x276/0x332

To cleanup pending events driver has to keep linked list of pending
flips in file private. When file is closed all events in pending flips
for the file are cleared.

To avoid race conditions dev->event_lock is used to protect pending
flips lists from parallel access.

Issue: KER-47
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
CC: Rahul Verman <rahul.verma@intel.com>
Acked-by: Sean V Kelley <sean.v.kelley@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Fix page flip lockup when requesting vblank fails

drm_vblank_get may fail if specified pipe is off. In that case we won't
get vblank even and can't call drm_vblank_put. But driver still has to
complete page flip to send required completion event to userspace.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Remove useless global variable

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: ttm: Fix clearing of highmem pages

ttm might allocate highmem pages for GPU usage. But page_address works
only for normal pages. clear_highpage can be used to clear both type of
pages.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reported-and-tested-by: Daniel Charles <daniel.charles@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Move NULL check outside spin lock

pending_flip->event is owned by single kernel thread at a time so it can
be checked outside of the spin lock. Spin lock is only protecting the
linked list where the event is added.

Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: pvr: Dereference a pointer after null check

Dereferencing a pointer may only happen after NULL check to avoid crash
if the pointer is NULL.

[pvr/services4/srvkm/common/buffer_manager.c:1152]: (error) Possible
null pointer dereference: psBMHeap - otherwise it is redundant to check
if psBMHeap is null at line 1156

Reported-by: Olev Kartau <olev.kartau@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: Dereference a pointer after null check

Dereferencing a pointer may only happen after NULL check to avoid crash
if the pointer is NULL.

[imgv/psb_fence.c:33]: (error) Possible null pointer dereference:
dev_priv - otherwise it is redundant to check if dev_priv is null at
line 36

Reported-by: Olev Kartau <olev.kartau@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drv: Remove unused variable

Remove unused stack variable which would crash -O0 optimised kernel.

[drv/psb_page_flip.c:61]: (error) Uninitialized variable: e

Reported-by: Olev Kartau <olev.kartau@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: add support for wrapping a pvr buffer as ttm buffer

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: refactor the placement allocator ioctls

Refactoring is done for better code reuse and to prepare for the next
patch.

Also while at it:
- get rid of unnecessary unlikely() annotations
- treat only negative return values as error instead of any non-zero
values

Other than the above no functional changes.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: add memory backend support for fixed pages

Needed by the upcoming patch adding support for wrapping a pvr buffers
as ttm buffer.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: pvr: add helper function to lookup a pvr buf by id

Needed by the upcoming patch adding support for wrapping a pvr buffer
as ttm buffer.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: refactor the checking of buffer placement

The driver checks whether a client process has enough permission to
set/clear TTM_PL_FLAG_NO_EVICT. For busy placements we don't need to do
this every time a related IOCTL is called, since those placements never
change and they don't include this flag. So remove the check for busy
placements.

An addition in a later patch will want to check the final flags value
besides the set/clear values, so add support for it now.

While at it, also normalize the cacheability flags in
ttm_pl_setstatus_ioctl similarly to what the buffer creation IOCTLs do.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: imgv: fix parameter checking for exec cmd IOCTL

As the comment in code suggests the IOCTL accepts only flags specifying
a GPU read or write operation, for all other flags we should return
-EINVAL. The call to psb_ttm_bo_check_placement that would seemingly
check this is actually a no-op in this case. Replace it with an actual
check. While at it, also clean up the preceeding check.

This is also needed by an upcoming patch getting rid of
psb_ttm_bo_check_placement.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drm: ttm: add support for non-swappable buffers

Needed by buffers that can move in GPU address space, but are not
swappable. An example is a wrapped PVR buffer, added by an upcoming
patch.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: drm: add missing header

Fix struct undefined error for struct drm_mode_modeinfo when drm_mode.h
is not included explicitly. Needed by an upcoming patch.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

gfx: pvr: fix uninitialized var bug on error path

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

psb_video: implement reset function by power up/down MSVDX

When MSVDX FW uploading is done by PUNIT, just after
MSVDX power up operation, In order to reset HW, driver need to
do one power off and power on, and wait for PUNIT re-upload FW
again.
This patch will enable driver to reset HW when there is a request.

Signed-off-by: Binglin Chen <binglin.chen@intel.com>
Signed-off-by: Daniel Charles <daniel.charles@intel.com>
Acked-by: Sean V Kelley <sean.v.kelley@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

psb_video: remove MSVDX firmware uploading from driver for PNW D0

For PNW C0 and older steppings, MSVDX(Video Decoder) firmware is
uploaded by driver. In PNW D0 silicon, MSVDX uploading is done by PUNIT
so video driver needs to remove firmware uploading functionality
and add adaption codes to work with PUNIT on firmware uploading,
verification, and MTX initialization.

Signed-off-by: Binglin Chen <binglin.chen@intel.com>
Signed-off-by: Daniel Charles <daniel.charles@intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Acked-by: Sean V Kelley <sean.v.kelley@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

psb_video: remove OSPM_GL3_CACHE_ISLAND when not enabled

When setting the island up for OSPM_GL3_CACHE_ISLAND the code was not
protected when the config is not enabled

Signed-off-by: Daniel Charles <daniel.charles@intel.com>
Reviewed-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Sean V Kelley <sean.v.kelley@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Fix BUG_ON() triggering in drm_vblank_put()

We're calling drm_vblank_get() too late. It must be done before the
xchg() to avoid psb_flip_complete() from calling drm_vblank_put() before
drm_vblank_get(). This could trigger a a BUG_ON() in drm_vblank_get().

Earlier the pvr sync callbacks were usually executed with interrupts
disabled, which made the problem extremely unlikely to happen.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Make sync_list and sync_lock static

drivers/staging/mrst/pvr/services4/srvkm/env/linux/ossync.c:30:1: warning: symbol 'sync_lock' was not declared. Should it be static?
drivers/staging/mrst/pvr/services4/srvkm/env/linux/ossync.c:31:1: warning: symbol 'sync_list' was not declared. Should it be static?

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: Remove an unused variable

drivers/staging/mrst/drv/psb_page_flip.c: In function 'psb_intel_crtc_page_flip':
drivers/staging/mrst/drv/psb_page_flip.c:193: warning: unused variable 'ret'

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: psb_gtt: Fix smatch warnings

drivers/staging/mrst/drv/psb_gtt.c:54 psb_gtt_takedown(4) warn: variable dereferenced before check 'pg' (see line 52) (smatch)
drivers/staging/mrst/drv/psb_gtt.c:343 psb_gtt_mm_init(10) warn: variable dereferenced before check 'pg' (see line 336) (smatch)

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: Silence a compiler warning

drivers/staging/mrst/drv/psb_page_flip.c: In function 'sync_callback':
drivers/staging/mrst/drv/psb_page_flip.c:102: warning: 'reg_offset' may be used uninitialized in this function

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Avoid double kmalloc()

Allow the user to embed the pending_sync structure inside its own
structure, and use container_of() in the callback.

This means we can now get by with one kmalloc() per flip, instead
of two.

As PVRSRVCallbackOnSync() can no longer fail, change the return
type to void.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Execute sync callbacks outside the spinlock

No need to hold the spinlock while executing the sync callbacks.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Use spin_lock_irq() in PVRSRVCallbackOnSync()

The code already calls kmalloc(..., GFP_KERNEL), so interrupts can't be
off anyway.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Make flags parameter to PVRSRVCallbackOnSync() unsigned

It's a bitmask, so it should be unsigned.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: ossync: Rewrite sync counter comparisons

Make the sync counter comparisons handle wraparound. Also reorganize
the code a bit to make it more readable.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: overlay: Use msleep(1) in ovl_wait()

Having ovl_wait() react immediately to hardware state changes is not
critical, so use msleep(1) instead of cpu_relax().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: overlay: Use double buffering for overlay registers and add suspend/resume hooks

The existing suspend/resume code that fiddles with the OVADD registers
simply does not work. Replace it with something sensible.

Allocate two pages for the registers and ping-pong between them. This
allows update_plane() to do its thing without waiting for the previous
update to complete. The previous update, if not completed, will simply
be overwritten by a new update.

If the hardware is powered down when update_plane() is called we can
simply skip programming OVADD, and instead wait for the reusme hook
to do it next time the hardware is woken up. It's also theoretically
possible that the hardware gets woken up just after we've pointed the
driver to the new register page, in which case we can simply skip
programming OVADD in update_plane(), since we know the resume hook
already did it for us.

There is a slight concern whether the hardware fetches the register
contents from memory in an atomic fashion wrt. the OVADD register
writes. That is, if the hardware can still be fetching the registers
from the old page after OVADD is pointed to the new page, there
might be trouble. So far I did not observe any ill effects. The
hardware really lacks any kind of sensible status + atomic flip
registers, so if it turns out there are problems we may need to
poll the pipe scanline counter to avoid updating OVADD too close
to the time when the hardware fetches the registers from memory.
But so far that doesn't seem necessary.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: overlay: Flush posted writes to the OVADD registers

We generally want the overlay update to happen without much delay, so
read the register back to make sure posted writes are flushed.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: overlay: Use jiffies based timeout in ovl_wait()

Wait at most 100 ms for overlay updates.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: overlay: Use set_memory_wc() instead of vmap()

x86 has set_memory_*() API so no need to vmap(). Also change from UC
to WC to make register updates, coeff tables in particular, more
efficient. There's a wmb() already in the code, and AFAICS that
should be enough to flush the WC buffers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: psb_gtt: Remove the rw semaphore

The read/write semaphore is only taken in read mode, so it
doesn't actually do anything meaningful. Remove it.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: Remove dead code

Nuke a big chunk if #if 0'ed code.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

staging: mrst: Fix error handling in psbfb_create

The two goto labels for unwinding on error were inverted. Swap them
around.

Fixes a smatch error:
drivers/staging/mrst/drv/psb_fb.c:421 psbfb_create(82) error: we previously assumed 'fb' could be null (see line 368) (smatch)

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

drm: Fix __user sparse warnings

Several pointers and casts were missing __user annotations.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>

drm: plane: mutex_unlock() was missing

Unlock the mode_config mutex if drm_plane_init() fails.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Install drm_fourcc.h

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Fix zorder handling while overlay updates are pending

Configuring the Z order may potentially need to touch both overlays
simultaneosly. So ovl_wait() needs to be performed for both overlays.
Do that before touching any state to allow set_plane_opts() to fail
in a clean fashion.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Need to wait for overlay in set_plane_opts()

Touching the overlay register memory while the overlay is fetching (or
primed to) the data. Do an ovl_wait() to make sure no updates are
pending.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add overlay Z order support

Allow the Z order of the overlay to be selected.

Note that there is something strange going on though. It seems changing
the Z order also affects destination color keying. When Z order is
configured one way destination color keying works, but when the Z order
is flipped destination color keying no longer is effective.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add overlay color keying and constant alpha

Both source and destination color keying can be enabled for the
overlays.

Also the overlay constant alpha blending factor can be changed.
Note that the overlay constant alpha isn't all that useful, though.
What you usually want is to control graphics alpha. That will need
to be handled through the CRTC somehow (a new ioctl possibly) until
such time when the scanout functionality can be ripped out of the
CRTC object. When that happens the graphics plane would be just another
drm plane.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add overlay CSC matrix and chroma siting settings

The used color space conversion matrix can be toggled between BT.601 and
BT.709 versions. Also the siting of chroma samples can now be selected.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add overlay color correction settings

Allow the overlay brightness, contrast, hue and saturation to be
adjusted. Also allows one to select the input range (JPEG vs. MPEG)
for the color space conversion.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add support for Medfield video overlays

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Use drm_framebuffer_check()

Call drm_framebuffer_check() to verify that the requested fb layout is
sane.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Add alignment argument to psb_gtt_map_pvr_memory()

The video overlay register block must be aligned on a 64K boundary.
Allow the caller of psb_gtt_map_pvr_memory() to specify the desired
alignment.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_chroma_phase_offsets() utility function

This function is is there to help driver writers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Add plane options ioctl

Add a new ioctl DRM_IOCTL_MODE_PLANE_OPTS which is used to configure
various settings for the plane.

I left out gamma correction thinking that it could be added using a
separate ioctl, since there's already a gamma ioctl for CRTCs.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_calc_{hscale, vscale}() utility functions

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add struct drm_region and assorted utility functions

struct drm_region represents a two dimensional region. The utility
functions are there to help driver writers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_framebuffer_check() utility function

This function performs a battery of sanity checks on the requested
framebuffer layout. Drivers can call it from their fb_create hook.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_format_{horz, vert}_chroma_subsampling() utility functions

These functions return the chroma subsampling factors for the specified
pixel format.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_format_plane_cpp() utility function

This function returns the bytes per pixel value based on the pixel
format and plane index.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add drm_format_num_planes() utility function

This function returns the number of planes used by a specific pixel
format.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Check that the requested pixel format is valid

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Handle duplicate FOURCCs

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Replace pitch with pitches[] in drm_framebuffer

Otherwise each driver would need to keep the information inside
their own framebuffer object structure. Also add offsets[]. BOs
on the other hand are driver specific, so those can be kept in
driver specific structures.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Check that the fb pixel format is supported by the plane

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Make 'formats' parameter to drm_plane_init() const

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Check crtc coordinates against integer overflows in setplane ioctl

Help drivers a little by guaranteeing that crtc_x+crtc_w and
crtc_y+crtc_h don't overflow.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Check source coordinates

Make sure the source coordinates stay within the buffer.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: fourcc: Use __u32 instead of u32

drm_fourcc.h can be included from user space so appropriate types.

Also fix some sparse errors while were at it.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: plane: Clear plane.crtc and plane.fb after disable_plane()

These are the only indication to user space that the plane was disabled.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Redefine pixel formats

Name the formats as DRM_FORMAT_X instead of DRM_FOURCC_X. Use consistent
names, especially for the RGB formats. Component order and byte order are
now strictly specified for each format.

The RGB format naming follows a convention where the components names
and sizes are listed from left to right, matching the order within a
single pixel from most significant bit to least significant bit.

The YUV format names vary more. For the 4:2:2 packed formats and 2
plane formats use the fourcc. For the three plane formats the
name includes the subsampling information using the standard
subsampling notation.

The fourccs for for all the RGB formats and some of the YUV formats
I invented myself. The idea was that looking at just the fourcc you
get some idea what the format is about without having to decode it
using some external reference.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: Add a missing ')'

The code happened to compile because the flag wasn't actually used yet.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: add an fb creation ioctl that takes a pixel format v5

To properly support the various plane formats supported by different
hardware, the kernel must know the pixel format of a framebuffer object.
So add a new ioctl taking a format argument corresponding to a fourcc
name from the new drm_fourcc.h header file.  Implement the fb creation
hooks in terms of the new mode_fb_cmd2 using helpers where the old
bpp/depth values are needed.

v2: create DRM specific fourcc header file for sharing with libdrm etc
v3: fix rebase failure and use DRM fourcc codes in intel_display.c and
    update commit message
v4: make fb_cmd2 handle field into an array for multi-object formats
    pull in Ville's fix for the memcpy in drm_plane_init
    apply Ville's cleanup to zero out fb_cmd2 arg in drm_mode_addfb
v5: add 'flags' field for interlaced support (from Ville)

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Reviewed-by: Rob Clark <rob.clark@linaro.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

drm: add plane support v3

Planes are a bit like half-CRTCs. They have a location and fb, but
don't drive outputs directly. Add support for handling them to the core
KMS code.

v2: fix ABI of get_plane - move format_type_ptr to the end
v3: add 'flags' field for interlaced support (from Ville)

Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Reviewed-by: Rob Clark <rob.clark@linaro.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

staging: mrst: Return ERR_PTR() from fb_create hook

In case of an error, the caller expects ERR_PTR() to be
used instead of returning NULL.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Pauli Nieminen <pauli.nieminen@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>