TDA4VM: EGL: ERROR: eglInitialize() failed !!!

Hello,

The SDK demos are tested so that they work out-of-the-box. This seems to be an issue with the state of the driver or it's support.

What changes have you made to the filesystem, if any?

And is your current display that is connected using the R5 rendering? Or do you see the weston homescreen?

Please also run "rgx_compute_test" as a sanity check that the GPU driver is working.

Regards,

Erick

Hi, Erick, 

Erick Narvaez said:

What changes have you made to the filesystem, if any?

I'm not sure is there anything different. Could you help list the dependencies required for running GPU and libEGL on the 0805 SDK? I'll check if there's anything missing or if the version is consistent.

Erick Narvaez said:

And is your current display that is connected using the R5 rendering? Or do you see the weston homescreen?

I use the DP connected monitor directly and I can see display like this:

Erick Narvaez said:

Please also run "rgx_compute_test" as a sanity check that the GPU driver is working.

Here is the running result, please help check. 

root@optimus:/data# rgx_compute_test
------------------ RGX compute test -----------------
----------------------- Start -----------------------
Call PVRSRVConnectionCreateDevice with a valid argument:
Connecting to first (0) default pvr device
 OK
Create dev var context:
 OK
Looking up General heap handle
 OK
Getting event object
 OK
Creating robustness buffer
 OK
Mapping robustness buffer
 OK
Creating Compute Context
 OK
Creating Buffer
Creating DWord for CDM Event Object
 OK
 OK
Create PDS Heap
 OK
Create USC Heap
 OK
Reset event object value
Creating NOP instruction
Creating Data Segment
Creating Code Segment
Write Kernel 0
Creating Fence Data Segment
Creating Code Segment
Write Fence Kernel
Write Terminate
Call services to kick CDM
 OK
Poll for CDM event object data
Event object value: 0xa1b2c3d4
 OK
 OK
Destroy Compute Context
 OK

Total time: 1ms
Disconnect from services:
 OK
------------------------ End ------------------------

Thank you. 

Hi,

muqun.yun said:

I'm not sure is there anything different. Could you help list the dependencies required for running GPU and libEGL on the 0805 SDK? I'll check if there's anything missing or if the version is consistent.

It looks like you have the dependencies for the GPU driver to be running, so this is fine.

muqun.yun said:

I use the DP connected monitor directly and I can see display like this:

This means your display is working (R5 based).

The last step is to see why the GPU won't render to the framebuffer. Did you build your filesystem with Yocto? The default SDK will work out of the box, but if the configuration or GPU driver used is different, it can cause issues. Especially if it's pulling some of the latest changes we made to the driver.

I'll check SDK 8.5 Vision Apps on my side to see if the same error is present. Meanwhile, can you also check one of the other GPU applications? Like this one:

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/08_06_00_12/exports/docs/vision_apps/docs/user_guide/group_apps_basic_demos_app_arm_mosaic.html

To see if you get the same error.

Thanks,

Erick

Hi, Eirck,

Erick Narvaez said:

Did you build your filesystem with Yocto?

Yes, I use a Yocto project to build filesystem. 

Erick Narvaez said:

but if the configuration or GPU driver used is different, it can cause issues.

Is the configuration you said here refer to the Linux kernel configuration? 

Does “GPU driver” refer to ti-img-rogue-driver? I have already integrated the ti-img-rogue-driver_1.15.6133109.bb and ti-img-rogue-umlibs_1.15.6133109.bb in my Yocto project and I did not make any changes. 

Erick Narvaez said:

To see if you get the same error.

Yes, there is the same error: 

root@optimus:/data# ./vx_app_arm_opengl_mosaic.out
Test data path is NULL. Defaulting to current folder
APP: Init ... !!!
MEM: Init ... !!!
MEM: Initialized DMA HEAP (fd=4) !!!
MEM: Init ... Done !!!
IPC: Init ... !!!
IPC: Init ... Done !!!
REMOTE_SERVICE: Init ... !!!
REMOTE_SERVICE: Init ... Done !!!
 10209.441509 s: GTC Frequency = 200 MHz
APP: Init ... Done !!!
 10209.441582 s:  VX_ZONE_INIT:Enabled
 10209.441593 s:  VX_ZONE_ERROR:Enabled
 10209.441600 s:  VX_ZONE_WARNING:Enabled
 10209.444302 s:  VX_ZONE_INIT:[tivxInitLocal:145] Initialization Done !!!
 10209.444580 s:  VX_ZONE_INIT:[tivxHostInitLocal:93] Initialization Done for HOST !!!
libEGL warning: MESA-LOADER: failed to open swrast: /usr/lib/dri/swrast_dri.so: cannot open shared object file: No such file or directory (search paths /usr/lib/dri)

EGL: eglInitialize() returned 0
EGL: after eglInitialize() eglError (0x3001)
EGL: ERROR: eglInitialize() failed !!!
 10209.477582 s:  VX_ZONE_ERROR:[tivxOpenglMosaicProcessEglInfo:352] OpenGL Mosaic: ERROR: Couldn't open openGL window!
Segmentation fault (core dumped)

Thank you,

 Muqun

Muqun,

Are you trying to use the SDK 8.5 configuration file for the Yocto build? I'm see a discrepancy in the GPU driver version that we used in the SDK (1.15) and the one reported in your system. Can you send the dmesg logs? I think it is loading 1.13, which might be causing some issues.

muqun.yun said:

Does “GPU driver” refer to ti-img-rogue-driver? I have already integrated the ti-img-rogue-driver_1.15.6133109.bb and ti-img-rogue-umlibs_1.15.6133109.bb in my Yocto project and I did not make any changes. 

Regards,

Erick

Hi, Erick,

Thanks for finding this problem. Now I have corrected the problem, but the error still exists.

The dmesg log shows that it is loading 1.15: 

root@optimus:/# dmesg |grep PVR
[    7.788448] PVR_K:  207: Read BVNC 22.104.208.318 from HW device registers
[    8.084593] PVR_K:  207: RGX Device registered with BVNC 22.104.208.318
[  109.758524] PVR_K:  1777: RGX Firmware image 'rgx.fw.22.104.208.318.64k' loaded
[  109.772137] PVR_K:  1777: Shader binary image 'rgx.sh.22.104.208.318' loaded
root@optimus:/# dmesg |grep gpu
[    7.726795] pvrsrvkm 4e20000000.gpu: More than one power domain for this GPU, gpu driver manages power domains
[    8.178182] [drm] Initialized pvr 1.15.6133109 20170530 for 4e20000000.gpu on minor 0

I run rgx_compute_test again: 

root@optimus:/# rgx_compute_test
------------------ RGX compute test -----------------
----------------------- Start -----------------------
Call PVRSRVConnectionCreateDevice with a valid argument:
Connecting to first (0) default pvr device
 OK
Create dev var context:
 OK
Looking up General heap handle
 OK
Getting event object
 OK
Creating robustness buffer
 OK
Mapping robustness buffer
 OK
Creating Compute Context
 OK
Creating Buffer
Creating DWord for CDM Event Object
 OK
 OK
Create PDS Heap
 OK
Create USC Heap
 OK
Reset event object value
Creating NOP instruction
Creating Data Segment
Creating Code Segment
Write Kernel 0
Creating Fence Data Segment
Creating Code Segment
Write Fence Kernel
Write Terminate
Call services to kick CDM
 OK
Poll for CDM event object data
Event object value: 0xa1b2c3d4
 OK
 OK
Destroy Compute Context
 OK

Total time: 1ms
Disconnect from services:
 OK
------------------------ End ------------------------

Then I find that this two logs

[ 109.758524] PVR_K: 1777: RGX Firmware image 'rgx.fw.22.104.208.318.64k' loaded
[ 109.772137] PVR_K: 1777: Shader binary image 'rgx.sh.22.104.208.318' loaded

appear after I run rgx_compute_test, will this have any impact?

Then I run vx_app_arm_opengl_mosaic.out again, however, the error still exists. 

Thank you,

Muqun

Muqun,

muqun.yun said:

Then I find that this two logs

[ 109.758524] PVR_K: 1777: RGX Firmware image 'rgx.fw.22.104.208.318.64k' loaded
[ 109.772137] PVR_K: 1777: Shader binary image 'rgx.sh.22.104.208.318' loaded

appear after I run rgx_compute_test, will this have any impact?

This is OK, actually the GPU driver will only load the firmware until an application is run. The message coming through dmesg might take a bit longer since rgx_compute_test runs quickly.

muqun.yun said:

Then I run vx_app_arm_opengl_mosaic.out again, however, the error still exists. 

Ok. Perhaps there is still something missing in the Yocto build. The default SDK image runs fine. Let me check if there are configuration options in Yocto that could be triggering the wrong Graphics driver options.

Regards,

Erick

Hi, Erick, 

I found the problem. There were some problems when building the Yocto project, and some .so files were not linked correctly, and pkgconfig is not packaged into the filesystem .

Now after I manually add these files, the srv camera demo can run normally.

Next I will go and see where there are problems with our Yocto project.

Thanks a lot. 

Best Regards,

Muqun