Warning in boot log about vc4/vc4_hdmi_phy.c

[tompreston]

Is this the right place for my bug report?
Possibly, it's to do with the HDMI component of the vc4 driver. I don't know if this should live here or in github.com/raspberrypi/firmware.

Describe the bug
On Raspberry Pi 4 with dtoverlay=vc4-kms-v3d or dtoverlay=vc4-kms-v3d-pi4, I see a warning at boot time relating to the vc4 HDMI component.

[    6.055586] WARNING: CPU: 1 PID: 135 at drivers/gpu/drm/vc4/vc4_hdmi_phy.c:414 vc5_hdmi_phy_init+0x7b4/0x2078 [vc4]

To reproduce
Connect rpi4 to HDMI display. Build and boot rpi-5.4.y on Raspberry Pi 4. Check dmesg.

Expected behaviour
There should be no warning.

Actual behaviour
A warning appears in the boot log

System

• Which model of Raspberry Pi? Raspberry Pi 4
• Which OS and version (cat /etc/rpi-issue)? Custom OS, but I believe this is a kernel issue.
• Which firmware version (vcgencmd version)? I'm not sure how to check this without vcgencmd
• Which kernel version (uname -a)? Linux localhost 5.4.35-v8 #1 SMP PREEMPT Fri May 1 17:22:32 UTC 2020 aarch64 GNU/Linux

Logs
Full dmesg https://pastebin.com/4NYaaEffper

Additional context
The warning comes from the VC4_SET_FIELD macro, which checks that the register value is contained within the register mask. It looks like the warning is because phy_get_cp_current returns an 8-bit value, but the mask is only 6 bits. Either way the mask is applied so the 0xc8 turns into 0x08.

I need to find out what the register is supposed to be set to and what the effect of setting it to 0x08 is, but I don't know where the datasheet for the BCM2711 HDMI registers is.

Please can someone point me to the datasheet containing the registers for the BCM2711 HDMI peripheral.

Thanks!

[6by9]

I thought @popcornmix had already committed a fix for this. The correct value is 0x18.

[popcornmix]

It was cherry picked by you into #3535 but that PR got delayed.
I'll commit it directly.

[6by9]

Thanks @popcornmix - I'm glad someone remembers these things!

[tompreston]

@6by9 You mean the return value from phy_get_cp_current() is 0x18? That makes more sense, but I still don't understand why. Presumably 0x18 and 0x1c are (electrical) current values which need to be set depending on the vco frequency? 🤷 😄

Out of curiosity, is the datasheet publicly available?

Thanks!

[6by9]

You mean the return value from phy_get_cp_current() is 0x18? That makes more sense, but I still don't understand why.

They are values provided by Broadcom.
There was a typo in transferring the value from the Broadcom code to the firmware, and that was replicated in the KMS driver. It's now fixed.

Out of curiosity, is the datasheet publicly available?

No.

[tompreston]

I'm just testing this, then I'll close the issue.

[tompreston]

I can confirm the warning goes away with 71d5db1. Thanks!