Jump to content

Advent Vega kernel source code now available!


Guest PaulOBrien

Recommended Posts

Guest the_corvus

Regarding the camera, it is an USB standard camera. This could differ significantly from the Acer... One of the tegra2 SoC peripherals allows the designer to connect a digital camera, and allows to drive it directly, without using USB ... and, that peripheral does also implement autofocus and video processing. But P10AN01 does not use it. Instead, they use an standard USB camera that does not require firmware at all. Once the linux kernel recognizes it, it is exported as an standard Video 4 linux. So, though it should be easy to make it work, perhaps the Acer people have written an *.so library that is trying to access the (unused) Tegra2 SoC camera interface peripheral, and trying to load firmware for it... Perhaps, we can get a *.so library that only uses standard V4L interfaces ... In such case, it should work..

Agree.

I take out all the tegra soc camera stuff from kernel, and i can start camera and stop it, i see a video0 device, etc. So camera seems to be working. We only need a libcamera that use V4L devices.

About BT. I condigure all needed files, but it doesnt run. It gives me a timeout, so i search for the device and i think no device is switched on. It must be ttyHS2, but in /sys/class/rfkill (where i get a video0 device in video4linux) i dont get any device in bluetooth folder. Of course i start device using power_on, and in dmesg i get BT adapter enabled.

About backlight... is working now correctly, we can swith off and on the screen, but dmesg goes crazy with error logs. (failed to read points)

Corvus.

Link to comment
Share on other sites

Guest ejtagle

With this lines commented, i get the same errors...

In nvec.0 all files have 0 as value.

Corvus.

Are you still getting the same error (nvec nvec.0: NvEC not found) ?

Regarding BT, i also agree.. It should be in /dev/ttyHS2 ...Seems something is still missing.. I'll try to check it...

If you could post dmesg logs, it would help with nvec problems ;)

Edited by ejtagle
Link to comment
Share on other sites

Guest Cass67

Also i attached my patched libhardware_legacy.so, with this Android loads the driver an configure wpa_supplicant.

The path for module and atheros blobs is with this: /system/lib/hw/wlan/

rebel1

libhardware_legacy.so.zip

Rebel1,

Do you happen to have the test 15 newbe5 libhw legacy so around? Its the TF build with no sound ... if so could you patch that the same to support ar6k ? Mine always fail, or if not can you tell me how to do it correctly ;) ...

i managed to get sound on theTF build so hopefully with this in place we can proceed where newbe5 does not have to fight with icona builds no more.. was a missing lib that was up... /system/lib/hw/tegra_alsa.tegra.so, I took one from an icona build i found.... the build now makes the familiar crackle on boot as well as other sounds :)

Cheers

EDIT : - im having trouble attaching the file ... ping me if you need it ...

Edited by Cass67
Link to comment
Share on other sites

Guest Cass67

About BT. I condigure all needed files, but it doesnt run. It gives me a timeout, so i search for the device and i think no device is switched on. It must be ttyHS2, but in /sys/class/rfkill (where i get a video0 device in video4linux) i dont get any device in bluetooth folder. Of course i start device using power_on, and in dmesg i get BT adapter enabled.

Agreed, BT is a bit borked, i could not make it work either ... seems to be ok but in the end hciattach fails with timeout...

Link to comment
Share on other sites

Guest ejtagle

Agreed, BT is a bit borked, i could not make it work either ... seems to be ok but in the end hciattach fails with timeout...

Well, i have investigated a bit the BT issue...It is nearly guesswork, as there is no information available onthe bluetooth module used. Or they want to sell more than 10kunits to give the datasheet... But, nowadays i am suspecting that there is no separate bluetooth module. WLAN and bluetooth are integrated in just one module. (http://www.artimar.com.br/uploads/MMT_wireless_100615.pdf), look for LBEE19QMBC.. It is said to be compatible with the MW2024 module. That module has *again, i am trying to do an educated guess here!) 4 independant power supplies, and 2 other ones called VDD_HOSTIF_B (VDD for Host interface for bluetooth), and VDD_HOSTIF_WLAN) (vdd for host interface of WLAN).

I don'r know if it is making sense to you, but for me, taking into account tbat tbe Bluetooth power control driver i wrote for linux ONLY powers up the VDD_HOSTIF_B) , this could mean that there are still some power supplies to enable (the same that are enabled ONLY if WLAN is active).

So i would ask if someone could try to enable bluetooth, while the WLAN is enabled... I will try it myself, but you willhave to wait a couple of days *currently unable to try it)

Well, thanks a lot in advance,

Eduardo

Link to comment
Share on other sites

Guest the_corvus

Well, i have investigated a bit the BT issue...It is nearly guesswork, as there is no information available onthe bluetooth module used. Or they want to sell more than 10kunits to give the datasheet... But, nowadays i am suspecting that there is no separate bluetooth module. WLAN and bluetooth are integrated in just one module. (http://www.artimar.c...less_100615.pdf), look for LBEE19QMBC.. It is said to be compatible with the MW2024 module. That module has *again, i am trying to do an educated guess here!) 4 independant power supplies, and 2 other ones called VDD_HOSTIF_B (VDD for Host interface for bluetooth), and VDD_HOSTIF_WLAN) (vdd for host interface of WLAN).

I don'r know if it is making sense to you, but for me, taking into account tbat tbe Bluetooth power control driver i wrote for linux ONLY powers up the VDD_HOSTIF_B) , this could mean that there are still some power supplies to enable (the same that are enabled ONLY if WLAN is active).

So i would ask if someone could try to enable bluetooth, while the WLAN is enabled... I will try it myself, but you willhave to wait a couple of days *currently unable to try it)

Well, thanks a lot in advance,

Eduardo

Sorry... sometimes i forgot i spend too much time with vega and think everyone knows his internals.

Yes, i can confirm that wlan and BT module are the same chip, they are "connected". If i remember correclty, in stock rom, when bt is activated, wlan is activated too, but is not show in status bar.

I will test to switch on BT with wlan switched on. Will report...

About NVEC, i will post dmesg, with and without the lines you post earlier, commented.

Compile Time :P

Corvus.

Link to comment
Share on other sites

Guest the_corvus

Sorry... sometimes i forgot i spend too much time with vega and think everyone knows his internals.

Yes, i can confirm that wlan and BT module are the same chip, they are "connected". If i remember correclty, in stock rom, when bt is activated, wlan is activated too, but is not show in status bar.

I will test to switch on BT with wlan switched on. Will report...

About NVEC, i will post dmesg, with and without the lines you post earlier, commented.

Compile Time :P

Corvus.

Eduardo...

We have a problem with reboot. Tablet shutdown correctly, but reboot only switch off the screen, but tablet remains on.. :(

Can you see if can fix the problem?

Corvus.

Link to comment
Share on other sites

Guest a_appleby

Rebel1,

Do you happen to have the test 15 newbe5 libhw legacy so around? Its the TF build with no sound ... if so could you patch that the same to support ar6k ? Mine always fail, or if not can you tell me how to do it correctly ;) ...

i managed to get sound on theTF build so hopefully with this in place we can proceed where newbe5 does not have to fight with icona builds no more.. was a missing lib that was up... /system/lib/hw/tegra_alsa.tegra.so, I took one from an icona build i found.... the build now makes the familiar crackle on boot as well as other sounds :)

Cheers

EDIT : - im having trouble attaching the file ... ping me if you need it ...

Do you mean test 13? The test13 has a tegra_alsa.ventana.so and so does test22 from what I recall.

Link to comment
Share on other sites

Guest Cass67

Do you mean test 13? The test13 has a tegra_alsa.ventana.so and so does test22 from what I recall.

Im not sure of the status of sound in test 13, i was working from test 15 and it required tegra_alsa.tegra.so, the ventana file was not required and im not sure it was even in 15..

Link to comment
Share on other sites

Guest a_appleby

Im not sure of the status of sound in test 13, i was working from test 15 and it required tegra_alsa.tegra.so, the ventana file was not required and im not sure it was even in 15..

Can you share test build 15 ? Thanks.

Link to comment
Share on other sites

Guest ejtagle

Eduardo...

We have a problem with reboot. Tablet shutdown correctly, but reboot only switch off the screen, but tablet remains on.. :(

Can you see if can fix the problem?

Corvus.

Of course... System reboot (also) depends on NvEC drivers... With tegra2, there are 3 CPUs in fact... The first one is an ARMv7, and that is the one that starts execution. That CPU then enables one of the CortexA9 cpus (main cpu) and starts execution on that cpu, and disables itself...Then, everything happens with that cortex a9 (from it, the other CPU core (another a9) is enabled based on needs... To do a full system reset, we need an external means to trigger a reset, and that external means is the NvEC :(

So, all failures are related... Even, total shutodown depends on NvEC .. ;) --- So, i guess once we get NvEC running as it should, everything will work as expected (battery and reboots :D )

Link to comment
Share on other sites

Guest the_corvus

Of course... System reboot (also) depends on NvEC drivers... With tegra2, there are 3 CPUs in fact... The first one is an ARMv7, and that is the one that starts execution. That CPU then enables one of the CortexA9 cpus (main cpu) and starts execution on that cpu, and disables itself...Then, everything happens with that cortex a9 (from it, the other CPU core (another a9) is enabled based on needs... To do a full system reset, we need an external means to trigger a reset, and that external means is the NvEC :(

So, all failures are related... Even, total shutodown depends on NvEC .. ;) --- So, i guess once we get NvEC running as it should, everything will work as expected (battery and reboots :D )

This is great. I really hate nvec :D

I want to make some snapshot of the status of our work, what is working and what not:

1) HW acceleration. Works, flash included.

2) Wifi, works.

3) BT, works at linux level (hci0 up) but dont detected by android.

4) SDcard, works, but fixing some bugs

5) Camera, works at linux level, but not detected by android. Maybe the hardest part...

6) Rotation, doesnt work, but no one try to fix it, so maybe an easy task

7) Reboot, will work :P (then we will have access to CWM)

8) GSM and GPS may work, rebel made some test and seems to work.

We are choosing the best image for start with (Transformer, Iconia, Xoom, etc.)

Corvus.

Link to comment
Share on other sites

Guest Cass67

Can you share test build 15 ? Thanks.

what you have is probably as good as t15.. id not bother to be honest as its quite broken .. there will be a release of a much newer and very much improved build shortly .. its up to newbe5 when though .. its his baby ... wont be too long.

Link to comment
Share on other sites

Guest a_appleby

what you have is probably as good as t15.. id not bother to be honest as its quite broken .. there will be a release of a much newer and very much improved build shortly .. its up to newbe5 when though .. its his baby ... wont be too long.

It's not ok at all to just sit tight and wait. I will release the ROM I have built based on everything done so far soon.

It's enough that google doesn't release the HC sources. I will do all I can to share as much as possible.

We're supposed to stick it to the man and work together to get what Google & their buddies don't want us to. Sharing is the key and I will do my part.

Link to comment
Share on other sites

Guest newbe5

It's not ok at all to just sit tight and wait. I will release the ROM I have built based on everything done so far soon.

It's enough that google doesn't release the HC sources. I will do all I can to share as much as possible.

We're supposed to stick it to the man and work together to get what Google & their buddies don't want us to. Sharing is the key and I will do my part.

Yea, you were really forthcoming instantly when I asked for help with the init... It'll be coming soon, you can drop me a PM if you want and I can give you a link for development purposes. But you have to understand that releasing something this volatile in to the wild is going to open up a whole can of worms as people who don't know what they are doing flash it and then can't get back. We hope to release the current working image asap, but I'm hardly going to instantly release an update into the wild for every single change that is made, that's not unfair at all, it would add a ton of extra work...

Also, we are hardly sitting tight and waiting, we are working our asses off. Can you please show me all the links you've posted to things you have fixed and we will compare?

If you want some help, ask, don't get pissy.

newbe5

Edited by newbe5
Link to comment
Share on other sites

Guest Paul K

It's not ok at all to just sit tight and wait. I will release the ROM I have built based on everything done so far soon.

It's enough that google doesn't release the HC sources. I will do all I can to share as much as possible.

We're supposed to stick it to the man and work together to get what Google & their buddies don't want us to. Sharing is the key and I will do my part.

Give the guy a break, he has been and is working hard to get the best Vegacomb he can. If newbe5 decides not to realise we should wait until he is ready ........

sorry for the spam but could not just sit and ignore it.....

Link to comment
Share on other sites

Guest a_appleby

Yea, you were really forthcoming instantly when I asked for help with the init... It'll be coming soon, you can drop me a PM if you want and I can give you a link for development purposes. But you have to understand that releasing something this volatile in to the wild is going to open up a whole can of worms as people who don't know what they are doing flash it and then can't get back. We hope to release the current working image asap, but I'm hardly going to instantly release an update into the wild for every single change that is made, that's not unfair at all, it would add a ton of extra work...

Also, we are hardly sitting tight and waiting, we are working our asses off. Can you please show me all the links you've posted to things you have fixed and we will compare?

If you want some help, ask, don't get pissy.

newbe5

I'm sorry if that's what you understood from what I said, I should have mentioned what it was about. I was talking about that test 15 build, hoping it would be a bit more stable than 13 and could run more apps than 22.

As for all the rest, you will have it all tomorrow.

Give the guy a break, he has been and is working hard to get the best Vegacomb he can. If newbe5 decides not to realise we should wait until he is ready ........

sorry for the spam but could not just sit and ignore it.....

That was not meant for newbe5.

P.S.: I did send you a PM asking you to do just that, perhaps I should've mentioned more clearly I wanted the ROM for testing and development.

update:

test22 leaks memory after using the wireless for 30 minutes or more. Trying to find out what is doing this.

Edited by a_appleby
Link to comment
Share on other sites

Guest Cass67

Yeah it was meant for me... Point is t15 was bad and there was no point in putting it out as it was not usable and given you were using t13 it really was pointless.. I would have expected a bit less throwing toys out the pram and a bit more patience as what you want was coming ASAP.. t38+ is much nicer and worth the day or so wait.... what's the problem with that? Also its not my place to share other peoples projects, I'm helping out, nothing more...

Link to comment
Share on other sites

Guest Paul K

Yeah it was meant for me... Point is t15 was bad and there was no point in putting it out as it was not usable and given you were using t13 it really was pointless.. I would have expected a bit less throwing toys out the pram and a bit more patience as what you want was coming ASAP.. t38+ is much nicer and worth the day or so wait.... what's the problem with that? Also its not my place to share other peoples projects, I'm helping out, nothing more...

I will keep quiet now then ........ keep up the good work guys ........ SPAM over

Link to comment
Share on other sites

Guest the_corvus

Ej, i post you some logs.

We are having problems with memory.

This is /proc/meminfo once booted:

http://pastebin.com/EPkQjaEp

This is playing a youtube video with youtube app (it works):

http://pastebin.com/FB4Dzx74

And this is playing a megavideo video in browser (it doesn't work, touchscreen dont work, but adb do):

http://pastebin.com/ZgJHiM1d

In logcat:

http://pastebin.com/GuxrRLd7

in dmesg:

http://pastebin.com/KGUsmkxM

In other tests with other kernels (not this one) rebel modify kernel so GPU has 128M and not 64M (SHUTTLE_GPU_MEM_SIZE). We do test too with 96M.

With 128M we dont have any problem with flash playing, but free mem is so little for some memory hungry apps.

We need to search for some balance with flash working ok and get all the free mem we can...

Any help is welcome...

Corvus.

Link to comment
Share on other sites

Guest a_appleby

Ej, i post you some logs.

We are having problems with memory.

This is /proc/meminfo once booted:

http://pastebin.com/EPkQjaEp

This is playing a youtube video with youtube app (it works):

http://pastebin.com/FB4Dzx74

And this is playing a megavideo video in browser (it doesn't work, touchscreen dont work, but adb do):

http://pastebin.com/ZgJHiM1d

In logcat:

http://pastebin.com/GuxrRLd7

in dmesg:

http://pastebin.com/KGUsmkxM

In other tests with other kernels (not this one) rebel modify kernel so GPU has 128M and not 64M (SHUTTLE_GPU_MEM_SIZE). We do test too with 96M.

With 128M we dont have any problem with flash playing, but free mem is so little for some memory hungry apps.

We need to search for some balance with flash working ok and get all the free mem we can...

Any help is welcome...

Corvus.

Was playing back flash video any better with 96MB of RAM for the GPU than with 64MB? What was happening when it was set to 64MB?

Edited by a_appleby
Link to comment
Share on other sites

Guest newbe5

Was playing back flash video any better with 96MB of RAM for the GPU than with 64MB? What was happening when it was set to 64MB?

96mb works better, but need to do more tests to see if it is ENOUGH, because with extended use the buffers may hit cap again.

So, next question, how easy would it be to add a 25% compcache to this kernel? I'd prefer to avoid using a swap partition if possible, but compcache may give us just about enough to get the job done?

newbe5

Link to comment
Share on other sites

Guest a_appleby

96mb works better, but need to do more tests to see if it is ENOUGH, because with extended use the buffers may hit cap again.

So, next question, how easy would it be to add a 25% compcache to this kernel? I'd prefer to avoid using a swap partition if possible, but compcache may give us just about enough to get the job done?

newbe5

It helps when you have regular apps which allocate memory for themselves, but it doesn't help much when we have to deal with things which can't be compressed. A video buffer can't be compressed further, APKs don't compress further either.

I will give it a shot and provide the results and the boot.img.

Link to comment
Share on other sites

Guest newbe5

It helps when you have regular apps which allocate memory for themselves, but it doesn't help much when we have to deal with things which can't be compressed. A video buffer can't be compressed further, APKs don't compress further either.

I will give it a shot and provide the results and the boot.img.

Not bothered about compressing video memory, but if we are using 96mb vmem, having a compcache on the rest of the system might give us a bit of a boost, since we will be losing valuable system resources. If this helps, even if it's only 20%, it might let us avoid using a swap partition, which I don't think would be a great answer.

newbe5

Link to comment
Share on other sites

Guest a_appleby

Not bothered about compressing video memory, but if we are using 96mb vmem, having a compcache on the rest of the system might give us a bit of a boost, since we will be losing valuable system resources. If this helps, even if it's only 20%, it might let us avoid using a swap partition, which I don't think would be a great answer.

newbe5

I just checked it out myself and it looks like there are some memory leaks somewhere. I still have to investigate it.

On a fresh boot with just one megavideo video played in Opera, free memory in "running apps" went down to 80-90MB. It was somewhat higher (went to 150, dropped to 90 again in one minute) after flushing Opera which was taking 62MB of RAM.

I think something is terribly wrong or leaky as almost nothing was cached and I found no memory hogs in ps afterwards!

update: memtotal is at 235836 instead of 365964 for me, something is wrong with the splitting of memory for the GPU, it might be just the kernel I built

Edited by a_appleby
Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.