2850 replies to this topic

[fosser2]

ejtagle,

Sorry to derail the current conversation. We're still having issues w/ HD video on our 3.1 kernel. All our avp code is the same as what scanno has on his git. I'm looking back through clocks and this is what you guys have active.

{ "uartd", "pll_p", 216000000, true },
{ "i2s1", "pll_a_out0", 0, false},
{ "i2s2", "pll_a_out0", 0, false},
{ "spdif_out", "pll_a_out0", 0, false},
{ "sdmmc1", "clk_m", 48000000, true },
{ "sdmmc2", "clk_m", 48000000, true },
{ "sdmmc4", "clk_m", 48000000, true },
{ "ndflash", "pll_p", 108000000, true },
{ "pwm", "clk_m", 12000000, false},
{ "usbd", "clk_m", 12000000, true }, /* fsl-tegra-udc , utmip-pad , tegra_ehci.0 , tegra_otg - we need this to be always on to always get hotplug events */
{ "usb2", "clk_m", 12000000, false}, /* tegra_ehci.1 - Really unused*/
{ "usb3", "clk_m", 12000000, true }, /* tegra_ehci.2 - we need this to be always on to always get hotplug events */
{ "i2c1", "clk_m", 800000, false}, /* tegra-i2c.0 */
{ "i2c2", "clk_m", 315789, false}, /* tegra-i2c.1 */
{ "i2c3", "clk_m", 800000, false}, /* tegra-i2c.2 */
{ "dvc", "clk_m", 2400000, false}, /* tegra-i2c.3 */

{ NULL, NULL, 0, 0},

This is what we are using right now:

{ "cdev1", NULL, 0, true },
{ "blink", "clk_32k", 32768, false},
{ "pll_p_out4", "pll_p", 24000000, true },
{ "pwm", "clk_32k", 32768, false},
{ "i2s1", "pll_a_out0", 0, false},
{ "i2s2", "pll_a_out0", 0, false},
{ "spdif_out", "pll_a_out0", 0, false},
{ NULL, NULL, 0, 0},

I'm wondering if I'm missing something important that has to do with the video decoding. Thank you for the help.

-fosser2

[ejtagle]

fosser2, on 30 October 2012 - 08:43 PM, said:

ejtagle,

Sorry to derail the current conversation. We're still having issues w/ HD video on our 3.1 kernel. All our avp code is the same as what scanno has on his git. I'm looking back through clocks and this is what you guys have active.

{ "uartd", "pll_p", 216000000, true },
{ "i2s1", "pll_a_out0", 0, false},
{ "i2s2", "pll_a_out0", 0, false},
{ "spdif_out", "pll_a_out0", 0, false},
{ "sdmmc1", "clk_m", 48000000, true },
{ "sdmmc2", "clk_m", 48000000, true },
{ "sdmmc4", "clk_m", 48000000, true },
{ "ndflash", "pll_p", 108000000, true },
{ "pwm", "clk_m", 12000000, false},
{ "usbd", "clk_m", 12000000, true }, /* fsl-tegra-udc , utmip-pad , tegra_ehci.0 , tegra_otg - we need this to be always on to always get hotplug events */
{ "usb2", "clk_m", 12000000, false}, /* tegra_ehci.1 - Really unused*/
{ "usb3", "clk_m", 12000000, true }, /* tegra_ehci.2 - we need this to be always on to always get hotplug events */
{ "i2c1", "clk_m", 800000, false}, /* tegra-i2c.0 */
{ "i2c2", "clk_m", 315789, false}, /* tegra-i2c.1 */
{ "i2c3", "clk_m", 800000, false}, /* tegra-i2c.2 */
{ "dvc", "clk_m", 2400000, false}, /* tegra-i2c.3 */

{ NULL, NULL, 0, 0},

This is what we are using right now:

{ "cdev1", NULL, 0, true },
{ "blink", "clk_32k", 32768, false},
{ "pll_p_out4", "pll_p", 24000000, true },
{ "pwm", "clk_32k", 32768, false},
{ "i2s1", "pll_a_out0", 0, false},
{ "i2s2", "pll_a_out0", 0, false},
{ "spdif_out", "pll_a_out0", 0, false},
{ NULL, NULL, 0, 0},

I'm wondering if I'm missing something important that has to do with the video decoding. Thank you for the help.

-fosser2

Clocks seems to be ok. Neither on our table, nor in yours, there is any reference to avp ... Be sure the problem is not memory related.. (gpu memory, i mean..)

[king0r]

king0r, on 28 October 2012 - 01:55 AM, said:

Hello,

I have also buyed an used MN10U. I immediately flashed it to latest JellyBean ROM available (JellyBean v6 Image).
But there is no option to overclock the Tegra CPU?

Is there any kernel out which supports OC the CPU? When not is it already in development?

Thank you!

Answers please ...

[Cass67]

king0r, on 31 October 2012 - 03:03 PM, said:

Answers please ...

Not that im aware of...

[fosser2]

ejtagle, on 31 October 2012 - 12:06 AM, said:

Clocks seems to be ok. Neither on our table, nor in yours, there is any reference to avp ... Be sure the problem is not memory related.. (gpu memory, i mean..)

Ejtagle,

Here is a pastebin of our HD decoder crashing. As far as I can tell our GPU memory has been alocated the same as the Vega's. Here is a snippit.

// SMBA1002 memory is 1xSZ_512M
#define SMBA1002_MEM_SIZE SZ_512M /* Total memory */
#define SMBA1002_MEM_BANKS 1

#define SMBA1002_GPU_MEM_SIZE SZ_128M /* Memory reserved for GPU */

#define SMBA1002_FB1_MEM_SIZE SZ_8M /* Memory reserved for Framebuffer 1: LCD */
#define SMBA1002_FB2_MEM_SIZE SZ_16M /* Memory reserved for Framebuffer 2: HDMI out */

#define DYNAMIC_GPU_MEM 1 /* use dynamic memory for GPU */

I also diffed our avp files and they are both stock 3.1 from NVIDIA. Do you have any other insite on what we're missing? Thank you in advance.

PS: here is a link to the git

Also, I'm looking for a recent copy of the shuttle def_config. Do you know where I can find that?

Edited by fosser2, 31 October 2012 - 08:40 PM.

[king0r]

Cass67, on 31 October 2012 - 04:17 PM, said:

Not that im aware of...

But the source code of the kernel is bublished. Why isn´t it possible to add OC functionality?

Edited by king0r, 31 October 2012 - 08:56 PM.

[Cass67]

king0r, on 31 October 2012 - 08:54 PM, said:

But the source code of the kernel is bublished. Why isn´t it possible to add OC functionality?

Of course its possible, just no one is interested in doing it or at least has been interested in doing it ... Until this point stability was key and way more important than a few extra mhz.. undervolting is probably much more interesting...

[ejtagle]

king0r, on 31 October 2012 - 08:54 PM, said:

But the source code of the kernel is bublished. Why isn´t it possible to add OC functionality?

Cass67, on 31 October 2012 - 10:36 PM, said:

Of course its possible, just no one is interested in doing it or at least has been interested in doing it ... Until this point stability was key and way more important than a few extra mhz.. undervolting is probably much more interesting...

The answer is easy: We still have some stability issues. There are tablets that show strange behaviours. We don't want to add another cause of troubles here. And, just to be honest, i consider adding overclocking to the tegra2 kernel a bad idea. There is no cpu temperature sensor on the P10AN01. You can easily fry your tablet... There is no provision for evacuating the extra heat generated by the overclocked chipset. Look on this same thread for an extra detailed explanation on why we chose not to implement OC as a default, why it should not be offered to people that are unaware of the implications of overclocking (read: If abused, you WILL BURN your tablet! ...) ..

Regards,
Eduardo

[ejtagle]

fosser2, on 31 October 2012 - 07:49 PM, said:

Ejtagle,

Here is a pastebin of our HD decoder crashing. As far as I can tell our GPU memory has been alocated the same as the Vega's. Here is a snippit.

// SMBA1002 memory is 1xSZ_512M
#define SMBA1002_MEM_SIZE SZ_512M /* Total memory */
#define SMBA1002_MEM_BANKS 1

#define SMBA1002_GPU_MEM_SIZE SZ_128M /* Memory reserved for GPU */

#define SMBA1002_FB1_MEM_SIZE SZ_8M /* Memory reserved for Framebuffer 1: LCD */
#define SMBA1002_FB2_MEM_SIZE SZ_16M /* Memory reserved for Framebuffer 2: HDMI out */

#define DYNAMIC_GPU_MEM 1 /* use dynamic memory for GPU */

I also diffed our avp files and they are both stock 3.1 from NVIDIA. Do you have any other insite on what we're missing? Thank you in advance.

PS: here is a link to the git

Also, I'm looking for a recent copy of the shuttle def_config. Do you know where I can find that?

Quote

<6>[  121.259746] avp_node_try_connect: got conn ack 'daaaaaaa+' (cf8c6d40 <-> 11aba8)
<3>[  121.268375] avp_svc: can't allocate for handle 0xcf8bb460 (-12)

Seems to be a GPU out of memory problem... You could use our latest def_config, or use the one i posted for the DVC10, that should give you a very good base to start with...

Edited by ejtagle, 01 November 2012 - 03:48 AM.

[MikhailM]

king0r, on 31 October 2012 - 03:03 PM, said:

Answers please ...

You have to choose - latest rom, or OC. If overclocking is what you want, you can use Corvus, which I tihnk overclocks to 1.5. HoneyIce Chimera I think overclocks to 1.2. But if you want Jelly Bean, you have to use VegaBean Beta 6 without overclocking. The reasons were explained by ejtagle and Cass67.

[DerArtem]

Hi,

I have added a custom
arm_pm_restart = betelgeuse_restart;

to my kernel in order to write a magic value to EEPROM to reboot to recovery.

Now the tablet is feezing some times on reboot. Have you had something like that? Any ideas where I should dig arround?

I have also sometimes issue with shutdown that the tablet is also freezing during shutdown. This only happends when the tablet was on LP0 standby before...

Here is the commit: https://github.com/D...426d8f1c2041fb9

Edited by DerArtem, 06 November 2012 - 08:19 PM.

[ejtagle]

DerArtem, on 06 November 2012 - 08:19 PM, said:

Hi,

I have added a custom
arm_pm_restart = betelgeuse_restart;

to my kernel in order to write a magic value to EEPROM to reboot to recovery.

Now the tablet is feezing some times on reboot. Have you had something like that? Any ideas where I should dig arround?

I have also sometimes issue with shutdown that the tablet is also freezing during shutdown. This only happends when the tablet was on LP0 standby before...

Here is the commit: https://github.com/D...426d8f1c2041fb9

Well, i have had such problems, but not on the p10an01 (as far as i can tell, has no eeprom...) ... The problems that i had were on the p10an10 (that DOES have the eeprom)... Not sure about LP0 problems... Did you manage to get a last_kmsg from the kernel when shutting down ? .. It could give a clue ...

[karthik050782]

Hi,

We are trying to build CM10 for samsung galaxyR-I9103. HW acceleration is not working, using Nvidia codec, it says "failed to allocate node" in OMXCodec.cpp. Tried with the changes of DerArtem(used for ICS), but it didn't worked, same error.
Also avp_init won't get performed in our kmsg.
Please help us.

If I posted wrongly, please forgive me! This is my first post in the modaco!!!

Edit: the ejtagle libraries helped, now Hw accelration is working!!! Thanks a million!


Thanks,
Karthik

Edited by karthik050782, 08 November 2012 - 08:22 PM.

[Scanno]

Seems google (JBQ) is uploading AOSP 4.2

[ejtagle]

Scanno, on 13 November 2012 - 07:18 PM, said:

Seems google (JBQ) is uploading AOSP 4.2

I think porting to 4.2 will be trivial... NVIDIA has not released a new kernel ... Neither new graphics libs... :S ... AOSP 4.2 seems to be just a bugfixed version of 4.1.2..

Let's see ... Syncing right now to 4.2...

Edited by ejtagle, 14 November 2012 - 12:19 AM.

[Scanno]

ejtagle, on 14 November 2012 - 12:10 AM, said:

I think porting to 4.2 will be trivial... NVIDIA has not released a new kernel ... Neither new graphics libs... :S ... AOSP 4.2 seems to be just a bugfixed version of 4.1.2..

Let's see ... Syncing right now to 4.2...

I hope that there it is just a bugfix version with multi-user support
Anyway... as soon as I have beta 7 out I will sync 4.2 also and see how far the build is going with the current device config and kernel.

Just a question... I see in log cat a lot of errors from the shuttle powerhal complaining about not able to write to /proc/sys/....
Am I right in the assumption that I have some permissions wrongly setup?

[ejtagle]

Scanno, on 14 November 2012 - 02:17 PM, said:

I hope that there it is just a bugfix version with multi-user support
Anyway... as soon as I have beta 7 out I will sync 4.2 also and see how far the build is going with the current device config and kernel.

Just a question... I see in log cat a lot of errors from the shuttle powerhal complaining about not able to write to /proc/sys/....
Am I right in the assumption that I have some permissions wrongly setup?

It's the default CPU power governor used... Android tries to fine tune it (should improve power management, as Android knows when to speed the CPU up or down. You will need the interactive governor as the default kernel governor for this to work..

[Scanno]

ejtagle, on 14 November 2012 - 06:21 PM, said:

It's the default CPU power governor used... Android tries to fine tune it (should improve power management, as Android knows when to speed the CPU up or down. You will need the interactive governor as the default kernel governor for this to work..

The interactive governor IS the default:
CONFIG_CPU_FREQ_DEFAULT_GOV_INTERACTIVE=y

and i get:

E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load: No such file or directory
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/boost_factor: Permission denied
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/input_boost: No such file or directory
I/SystemServer( 341): Power Manager

and

E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/input_boost: No such file or directory
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/boost_factor: Permission denied

This is the directory structure i see:

root@android:/sys/devices/system/cpu/cpufreq/interactive # ll
-rw-r--r-- root	 root		 4096 2012-11-14 20:38 boost_factor
-rw-r--r-- root	 root		 4096 2012-11-14 20:37 go_maxspeed_load
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 io_is_busy
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 max_boost
-rw-rw---- system system	 4096 2012-11-14 20:37 min_sample_time
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 sustain_load
-rw-rw---- system system	 4096 2012-11-14 20:37 timer_rate
root@android:/sys/devices/system/cpu/cpufreq/interactive #

init.rc contains this:

chown system system /sys/devices/system/cpu/cpufreq/interactive/timer_rate
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/timer_rate
chown system system /sys/devices/system/cpu/cpufreq/interactive/min_sample_time
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/min_sample_time
chown system system /sys/devices/system/cpu/cpufreq/interactive/hispeed_freq
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/hispeed_freq
chown system system /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load
chown system system /sys/devices/system/cpu/cpufreq/interactive/above_hispeed_delay
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/above_hispeed_delay
chown system system /sys/devices/system/cpu/cpufreq/interactive/boost
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/boost
chown system system /sys/devices/system/cpu/cpufreq/interactive/boostpulse
chown system system /sys/devices/system/cpu/cpufreq/interactive/input_boost
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/input_boost

So i am a bit lost at the moment why this is.
It seems that the kernel generates other files then Android expects... This is why the permissions are not set correctly.

Edited by Scanno, 14 November 2012 - 09:22 PM.

[ejtagle]

Scanno, on 14 November 2012 - 07:35 PM, said:

The interactive governor IS the default:
CONFIG_CPU_FREQ_DEFAULT_GOV_INTERACTIVE=y

and i get:

E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load: No such file or directory
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/boost_factor: Permission denied
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/input_boost: No such file or directory
I/SystemServer( 341): Power Manager

and

E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/input_boost: No such file or directory
E/Shuttle PowerHAL( 341): Error opening /sys/devices/system/cpu/cpufreq/interactive/boost_factor: Permission denied

This is the directory structure i see:

root@android:/sys/devices/system/cpu/cpufreq/interactive # ll
-rw-r--r-- root	 root		 4096 2012-11-14 20:38 boost_factor
-rw-r--r-- root	 root		 4096 2012-11-14 20:37 go_maxspeed_load
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 io_is_busy
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 max_boost
-rw-rw---- system system	 4096 2012-11-14 20:37 min_sample_time
-rw-r--r-- root	 root		 4096 2012-11-14 20:48 sustain_load
-rw-rw---- system system	 4096 2012-11-14 20:37 timer_rate
root@android:/sys/devices/system/cpu/cpufreq/interactive #

init.rc contains this:

chown system system /sys/devices/system/cpu/cpufreq/interactive/timer_rate
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/timer_rate
chown system system /sys/devices/system/cpu/cpufreq/interactive/min_sample_time
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/min_sample_time
chown system system /sys/devices/system/cpu/cpufreq/interactive/hispeed_freq
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/hispeed_freq
chown system system /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/go_hispeed_load
chown system system /sys/devices/system/cpu/cpufreq/interactive/above_hispeed_delay
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/above_hispeed_delay
chown system system /sys/devices/system/cpu/cpufreq/interactive/boost
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/boost
chown system system /sys/devices/system/cpu/cpufreq/interactive/boostpulse
chown system system /sys/devices/system/cpu/cpufreq/interactive/input_boost
chmod 0660 /sys/devices/system/cpu/cpufreq/interactive/input_boost

So i am a bit lost at the moment why this is.
It seems that the kernel generates other files then Android expects... This is why the permissions are not set correctly.

Perhaps Google has modified the Interactive governor of the Grouper kernel to export those entries... That could be the most probable cause... ... As i don't like modified kernels, i guess i will modify the power.c file to support the default interactive governor...

[Scanno]

ejtagle, on 15 November 2012 - 01:32 AM, said:

Perhaps Google has modified the Interactive governor of the Grouper kernel to export those entries... That could be the most probable cause... ... As i don't like modified kernels, i guess i will modify the power.c file to support the default interactive governor...

From what I see the power.c just writes values in the (hard coded) filenames. If I know the mapping between the current filenames and the what they should be I can modify power.c myself. Offcourse  I should also modify the ramdisk to chown/chmod those files.