i.MX6DL Android gpu-viv driver update

[waynekuo]

Hi all,

Our customer is facing a issue while use i.MX6DL + Android 6.0.1 Kernel 3.14 for their product.

While the app running, sometimes the LVDS will hand and show up the following logs (more details please see the attachment):
   [71484.964912] [galcore]: GPU[0] hang, automatic recovery.
   [71484.970228] [galcore]: recovery done

I check the source code, the log is from ../drivers/mxc/gpu-viv/ driver.
I found a community thread has the similar issue, seems update Kernel may solve the issue.

GPU hang 

However, customer is not allowed to upgrade bsp or Kernel right now.
Do NXP have gpu-viv driver update or patch for Kernel 3.14 to solve the issue?

Any suggestion?

Best Regards,
Wayne Kuo

Original Attachment has been moved to: log_0104_app_crash_x3.txt.zip

[Bio_TICFSL]

Hi wayne,

your issue is due to: galcore: clk_get vg clock failed, disable vg!

There was a problem with the i.MX6 v2.0 that required that  GPu2D and VG Core Clock and Axi Clock to be separated,  you may had a previous version of the device and it is having the problem

https://git.congatec.com/arm/qmx6_kernel/commit/765af27537bbbdea4e280823e6eeecbd73f2c604

regards

[waynekuo · Accepted solution]

Hi Bio,

Thanks for your suggestion, customer has solved this issue.
We think this issue cause by customer themselves.

We probe the VDD_PU_CAP voltage on customer'sboard.
We saw the VDD_PU_CAP voltage was cyclical High-to-Low, and which on SabreSD board is stable High.
We tried to set the PMU_REG_CORE register but no luck, seems this register was always been overwrite.
After customer's S/W optimization engineer did some modify and release another Kernel version, the PMU_REG_CORE register can be set to a fix value and VDD_PU_CAP voltage is stable, this iuuse is solved.
We think this issue is cause by unstable VDD_PU_CAP voltage.

Best Regards,
Wayne Kuo