Commit Graph

159443 Commits

Author SHA1 Message Date
Zeng Zhaoming
afca8068ea ENGR00138121 Fix system hangs when arecord/aplay continuously
Sdma iapi start loading sdma script by write HSTART register as
memory. When instruction reorder and IRQ delay may let the next
synchronize operation wait forever.

We change it by using writel() to access sdma registers,
and introduce timeout to show this error.

HSTART and STOP_STAT contain bits that are reset by hardware.
So if we read-modify-write, we are in danger of setting a bit
after SDMA has cleared it.

The spec calls these registers "write-ones" register.  So the
ARM can write a 1 to any bit, but does not need to worry about
clearing any bits that were previously set.  SDMA hardware
keeps track of all bits that were set.

Signed-off-by: Zeng Zhaoming <b32542@freescale.com>
Signed-off-by: Alan Tull <alan.tull@freescale.com>
2011-02-12 03:17:18 -06:00
Matt Sealey
258c10961b build fix for mismerged ENGR00138516 2011-02-12 03:09:59 -06:00
Jason Chen
6d02e13e4b ENGR00137158 ipuv3: fix flash issue by first setting FG position
fix flash issue by first setting FG position.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit d6e7fa4bc91fcff2e6f4aff02998b22f8b482358)
2011-02-12 03:01:06 -06:00
Jason Chen
67dd799948 ENGR00137058 ipuv3:update DP SRM by MCU if FG disable for FG position setting
During FG position set, we used to update SRM by FSU all the time,
if FG is disabled, then after you set position then enable  FG,
it will come out NFB4ENF error for FG channel.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit 6651726077719c68714887202d309ec1b2971047)
2011-02-12 03:00:47 -06:00
Ranjani Vaidyanathan
eda3daf6fb ENGR00138516: MX5X: Initialize CPU working point correctly for DVFS-CORE
CPU working point needs to be initialized based on the current CPU
frequency in start_dvfs() and should correspond with the correct entry
in the cpu_wp_auto table.

If its not initialized correctly, DVFS-CORE will fail when PLL-relock
results in a frequency that is not the same as set in the cpu_wp_auto table.
So fix it by finding the entry that closely matches the CPU frequency.

Signed-off-by: Ranjani Vaidyanathan <ra5478@freescale.com>
2011-02-12 03:00:32 -06:00
Liu Ying
7aba289ff0 ENGR00119661 MXC IPUv3 FB:Fix arbitary bpp issue
We used to force the framebuffer bpp to be default bpp
set by the user in the kernel boot up command line.
This patch fixes this issue. This patch also
gets TVE framebuffer bpp value from IPUv3 framebuffer.

Signed-off-by: Liu Ying <Ying.Liu@freescale.com>
2011-02-12 02:53:54 -06:00
Jason Chen
dd3c31bc69 ENGR00138270-3 ipuv3 fb: refine pan display method
Refine pan display method.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit b00590b8f5e964a7192d8af863a5cd4d24844d0d)
2011-02-12 02:52:06 -06:00
Jason Chen
202aa76653 ENGR00138270-2 ipuv3: change ipu_check_buffer_busy to ipu_check_buffer_ready
Change ipu_check_buffer_busy to ipu_check_buffer_ready.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit 3e60177e176d78e41351c9d51bc4889762c44627)
2011-02-12 02:51:46 -06:00
Jason Chen
96f9b6eb92 ENGR00138270-1 header file: change ipu_check_buffer_busy func name.
Change ipu_check_buffer_busy to ipu_check_buffer_ready.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit 6a0f0935318eac8733f79964854d2c18fe012099)
2011-02-12 02:51:36 -06:00
Richard Zhao
a2f5eeeefe ENGR00138141 GPU: add a wait loop to check timestamp for yamato
When bus is busy, for example vpu is working too, the timestamp is
possiblly not yet refreshed to memory by yamato when we get
GSL_INTR_YDX_CP_RING_BUFFER. For most cases, it will hit on first
loop cycle. So it don't effect performance.

Signed-off-by: Richard Zhao <richard.zhao@freescale.com>
(cherry picked from commit e97cab9b93d3e3f125684577f2522404da77872e)
2011-02-12 02:50:55 -06:00
Jason Chen
482f5ee366 ENGR00137873 ipuv3 fb: can not set correct bpp for tve through cmdline
fb cmdline option bpp= should pass the default bpp to fb driver, for tve etc
case, it will set_var by itself, which var->bits_per_pixel = 0,
the mxcfb_check_var will then fix it to 16. It's not correct
when we need set bpp to other value.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit f8a69aa6bf401b5a682e88040cf7abbf53406e38)
2011-02-12 02:49:35 -06:00
Jason Chen
e8b36e7b04 ENGR00137980 spdif: enable spdif_clk src support
Add spdif_clk src support, and add clk src check.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit 97134c928a436448c182c3bdf6053b5186184bfa)
2011-02-12 02:48:57 -06:00
Liu Ying
e18adb3f1d ENGR00137783-1 MXC V4L2 capture:Add several ioctrl support
This patch supports VIDIOC_ENUM_FMT, VIDIOC_ENUM_FRAMESIZES
and VIDIOC_DBG_G_CHIP_IDENT ioctrls.

Signed-off-by: Liu Ying <b17645@freescale.com>
(cherry picked from commit 90fdf1cded2ce26ac6e8c81944a78a3f998ddaf2)
2011-02-12 02:48:06 -06:00
Liu Ying
d2ba6ce83b ENGR00137733 MXC V4L2 overlay:Fill black data to fb before preview
This patch fills black data to overlay fb before preview to get rid
of dirty data splashing the screen.

Signed-off-by: Liu Ying <b17645@freescale.com>
(cherry picked from commit 3dd91a35a55263252ab5d4bb63762713c138b567)
2011-02-12 02:47:38 -06:00
Liu Ying
a7553845cd ENGR00137624 MXC V4L2 capture:Don't enable mclk after resume
We used to enable mclk in resume function of mxc v4l2 capture
driver even if overlay or catpture is off when entering suspend
mode. This patch disables mclk after resume in this case.

Signed-off-by: Liu Ying <b17645@freescale.com>
2011-02-12 02:46:24 -06:00
Liu Ying
4100e61a89 ENGR00137557 MXC V4L2 overlay:Fix preview issue on DI1 BG fb
1) Add DISP3 BG-DI1 output support.
2) Remove ADC output support.

Signed-off-by: Liu Ying <b17645@freescale.com>
2011-02-12 02:42:59 -06:00
Liu Ying
3ac0117c12 ENGR00137517 MXC V4L2 capture:Improve streaming capture
1) Remove skip frame support.
2) Add dummy buffer support for CSI_PRP_ENC_MEM and
   MEM_ROT_ENC_MEM channel to get rid of lost frame
   error.
3) Clear buffer ready bit for one time whenever we fail
   to update buffer address.
4) Fix potential tearing issue by sync the working buf
   with EOF interrupt.

Signed-off-by: Liu Ying <b17645@freescale.com>
(cherry picked from commit f35a2d44710fe0f73516d52f67f7d063eef87097)
2011-02-12 02:41:35 -06:00
Liu Ying
91cb774f78 ENGR00137559 MXC V4L2 overlay:Don't clear FG position after preview
The FG window will go to the left-top corner of the display panel
after preview. So, remove the code to clear FG position when we
stop preview.

Signed-off-by: Liu Ying <b17645@freescale.com>
(cherry picked from commit 29a459e692e4dde9f3dee64aba176a357dfd0062)
2011-02-12 02:41:16 -06:00
Matt Sealey
8101a2a15d ENGR00137447-3 ipuv3: change video mode timming adjust method
Just change v_sync_start, v_sync_width and v_sync_end to fit ipu restricitions.

Signed-off-by: Jason Chen <b02280@freescale.com>
(cherry picked from commit 6e0251471e24a483d774a552b947653c73cbbcf9)

Conflicts:

	drivers/mxc/ipu3/ipu_disp.c
2011-02-12 02:39:47 -06:00
Matt Sealey
dbdaf49758 config(mx): re-modularize audio 2011-02-11 22:41:04 -06:00
Matt Sealey
de462c995e siihdmi: rename tx->enable_audio => tx->basic_audio, add audio property if our monitor supports basic_audio. Not hotplug aware! 2011-02-11 20:48:21 -06:00
Matt Sealey
52b63e7922 siihdmi: make snapping to 720p configurable ('seventwenty', enabled by default) and clean up teneighty setup. 2011-02-11 18:03:23 -06:00
Matt Sealey
481f0bd252 siihdmi: teneighty mode was picking non-1080p modes and not letting us past. Restrict it to 1920x1080 matches. 2011-02-11 17:53:29 -06:00
Matt Sealey
0a207e8028 siihdmi: make the mode culling debug more like the mode supported debug 2011-02-11 17:47:53 -06:00
Matt Sealey
63c62f4d05 siihdmi: clean up default/720p mode setting comments and code
also re-enable mode culling debug to watch this more closely
2011-02-11 15:51:38 -06:00
Matt Sealey
c23690839f siihdmi: improve debug for modesetting to show refresh rate 2011-02-11 15:43:37 -06:00
Matt Sealey
3f8f2e78df siihdmi: remove doublescanned modes (VESA or CEA type, we don't want them) 2011-02-11 15:36:54 -06:00
Matt Sealey
f3520cb2dd siihdmi: SPDIF audio support!!! 2011-02-11 15:36:05 -06:00
Matt Sealey
8e999abf24 siihdmi: revisit teneighty mode
* Turn it on
* Enable searching for all 3 1080p modes we can display (24, 25, 30Hz) in descending order of refresh rate
* Default video mode is now taken from the awesome CEA modes database
* Move EDID->ModeDB to before CEA extension parsing so we can do it ourselves, better
* Add SVD parsing to the CEA EDID helper

If your monitor supports a low field-rate 1080p mode, it'll work now :)
2011-02-11 15:31:35 -06:00
Matt Sealey
e5d7ec8f04 cea861: introduced clock doubled modes (not the same thing as traditional vesa doublescan but..) 2011-02-11 15:14:59 -06:00
Matt Sealey
c9db1540d1 siihdmi: add audio interface control register bit definitions 2011-02-11 14:56:45 -06:00
Matt Sealey
b1ebfeffe7 cea861: add Short Video Descriptor block to the definitions, so we can parse out SVDs
Also correct definition of vendor specific block data (maximum payload is 28 bytes beyond the registration identifier)
2011-02-11 14:55:26 -06:00
Matt Sealey
959360f210 mx: add IOMUX for SPDIF_OUT 2011-02-11 14:54:45 -06:00
Matt Sealey
ef17a617c2 siihdmi: re-enable HDMI sink support ONLY if audio is to be enabled 2011-02-11 09:35:07 -06:00
Matt Sealey
fbea4a92ce siihdmi: re-enable annoying pink line 2011-02-11 09:30:56 -06:00
Matt Sealey
1dbee6176e Revert "video: add fb_edid_add_monspecs for parsing extended edid information"
This reverts commit 4779967299.
2011-02-11 09:21:58 -06:00
Matt Sealey
9a41560735 Revert "fbdev: export fb_edid_add_monspecs() for modules, improve algorithm"
This reverts commit 39274f9d7c.
2011-02-11 09:21:43 -06:00
Matt Sealey
5f7d0e5283 Revert "fbdev: when parsing E-EDID blocks, also use SVD entries"
This reverts commit 08235ab815.

Conflicts:

	drivers/video/fbmon.c
	drivers/video/modedb.c
2011-02-11 09:21:16 -06:00
Matt Sealey
bfebf81e87 siihdmi: remove fb_edid_add_monspecs since the mainline support is terrible by design 2011-02-11 09:17:59 -06:00
Matt Sealey
21118749bb cea861: every VIC mode in the spec 2011-02-11 09:17:22 -06:00
Matt Sealey
a69c2a53c5 siihdmi: temporarily disable action on HDMI sink detection to solve the "annoying pink line" 2011-02-06 20:18:54 -06:00
Matt Sealey
7b0b570a3a siihdmi: update vmode register set to supply fractional refresh values (60.00, 74.25, etc.) 2011-02-06 20:10:14 -06:00
Matt Sealey
e944fc336a sihdmi: optimize match code for mode culling by not matching 3 times over 2011-02-06 20:09:28 -06:00
Steev Klimaszewski
767836d497 siihdmi: comment out fb_add_monspecs 2011-02-06 18:45:49 -06:00
Saleem Abdulrasool
504ec6f122 siihdmi: fix cea861 data block traversal 2011-02-06 01:23:14 -08:00
Matt Sealey
94c8dd6810 siihdmi: fix vmode register set refresh rate calculation to match common sense (74250KHz is >65535 so we can't store it in a u16) and specifications (clock/htotal*vtotal) 2011-02-04 17:01:11 -06:00
Matt Sealey
bb05469897 siihdmi: experimental 1080p24 support (kernel option "teneighty=1") 2011-02-04 16:56:47 -06:00
Matt Sealey
dbefa3daff siihdmi: report when HDMI OUI is found and we're using an HDMI sink 2011-02-04 16:48:01 -06:00
Matt Sealey
fc94e49598 siihdmi: add internal/hidden paging registers 2011-02-04 16:46:41 -06:00
Matt Sealey
f9c5dd2af7 siihdmi: add structure definitions for versioning the chip and features 2011-02-04 16:44:46 -06:00