On Fri, 26 Jul 2019 13:35:26 +0000 Matthias Brugger mbrugger@suse.com wrote:

On 26/07/2019 14:54, Andre Przywara wrote:

...

Before we use random pointers that bcm2835_set_video_params() is supposed to populate, we should check its return value, which could indicate that the function bailed out early. This happend when there is no monitor connected.

That fixes a horrible crash when just booting with a serial console.

Can you test this patch? https://patchwork.ozlabs.org/patch/1130837/

It should fix your issue.

Yes, of course it does. Sorry that I missed that email, just replied to that as well.

By the way which version of the FW do you use?

Some older one, I guess. I downloaded some image on the day the board was announced.

Btw: I run this with ATF as armstub8.bin, which gives me PSCI, errata fixes and the ability to tweak the DT: https://github.com/Andre-ARM/arm-trusted-firmware/commits/rpi4 That works with the kernel or U-Boot as kernel8.img. The upstreaming discussion is happening on the official Gerrit (yikes!) server: https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/1629

Cheers, Andre

...

Signed-off-by: Andre Przywara andre.przywara@arm.com

Hi,

I saw this on the RPi4 (with Andrei's v2 series), but it looks like this would happen on the other boards as well.

Cheers, Andre.

drivers/video/bcm2835.c | 2 ++ 1 file changed, 2 insertions(+)

diff --git a/drivers/video/bcm2835.c b/drivers/video/bcm2835.c index bc41090aed..4188a94cba 100644 --- a/drivers/video/bcm2835.c +++ b/drivers/video/bcm2835.c @@ -26,6 +26,8 @@ static int bcm2835_video_probe(struct udevice *dev) ret = bcm2835_set_video_params(&w, &h, 32, BCM2835_MBOX_PIXEL_ORDER_RGB, BCM2835_MBOX_ALPHA_MODE_IGNORED, &fb_base, &fb_size, &pitch);

• if (ret)

• return ret;
  
  

  debug("bcm2835: Final resolution is %d x %d\n", w, h);