Bug #80
openScreen garbled in Seabios, does not boot, Kernel panic (X60)
0%
Description
I tried updating from Coreboot-4.4 to current HEAD on my Thinkpad X60. The build went fine, but afterwards:
- screen was partly disorted / garbled (big green artifacts in the bottom part)
- first attempt to boot resulted in kernel panic
- could get machine to boot only after rebooting and then it booted normally
Downgraded now to 4.4, and things are working okay again.
Files
Updated by Daniel Kulesz over 7 years ago
I had the same issue with the official 4.5 release. Then I switched off the following option:
Display => Keep VESA framebuffer
Since then, everything is back to normal. However, in 4.4 it used to work with the option enabled (just double-checked), so I would consider this a regression.
Updated by Arthur Heymans over 7 years ago
Could you see if commit 9c5fc62f9 "nb/i945/gma.c: use IS_ENABLED instead of #if, #endif" was in your git tree that you flashed? This breaks native graphic init, but is fixed in https://review.coreboot.org/#/c/17075/ .
Updated by Daniel Kulesz over 7 years ago
Sorry, didn't see your comment. Since the commit is merged, I will try with the current master and report back.
Updated by Daniel Kulesz over 7 years ago
- File IMG_20161118_161249.jpg IMG_20161118_161249.jpg added
I retried in master as of cd2afc0df034670a83479aded514b22b99124cf5. Here is what changed:
- again, first attempt to boot after flashing resulted in kernel panic (blinking scroll lock led)
- the issue that the device does not boot up at the first time got better. Lockup during boot happens only occasionally but is not easily reproducible.
- however, the device sometimes does not init correctly (empty screen, backlight on)
- when using the "Keep VESA framebuffer" option, the screen is still garbled during bootup, and also when booting into text-based Payloads such as meminfo or memtest86+. Memtest86+ itself does not report any memory issues (when running coreboot without the VESA option since no output is visible when activated).
I am attaching pictures for some of the symptoms.
Updated by Daniel Kulesz over 7 years ago
- File IMG_20161118_132934.jpg IMG_20161118_132934.jpg added
- File IMG_20161118_142310.jpg IMG_20161118_142310.jpg added
Updated by Daniel Kulesz over 7 years ago
- File IMG_20161118_161220.jpg IMG_20161118_161220.jpg added
- File IMG_20161118_161249.jpg IMG_20161118_161249.jpg added
Updated by Arthur Heymans over 7 years ago
I don't seem to have these issues on my x60. Linux can be unhappy if gpu was not initiated properly.
Could you include a cblog?
Would also be nice if you could bisect this issue, if you say this is a regression.
Updated by Daniel Kulesz about 7 years ago
I tried on current master, and there the issue is still present. I am trying to recover from this issue - until then, I will not be able to provide cblog.
Updated by Daniel Kulesz about 7 years ago
- File cbfs.txt cbfs.txt added
- File config.short.txt config.short.txt added
- File config.txt config.txt added
- File coreboot_console.txt coreboot_console.txt added
- File coreboot_timestamps.txt coreboot_timestamps.txt added
- File kernel_log.txt kernel_log.txt added
- File payload_config.txt payload_config.txt added
- File revision.txt revision.txt added
- File rom_checksum.txt rom_checksum.txt added
Here are the logs.
Updated by Daniel Kulesz about 7 years ago
One more thing I noticed in the VESA "broken" Mode:
If I hit "Esc" in Seabios after cold boot and choose the boot device, the system will always hang. If I do not hit Esc but let it continue booting, the Linux kernel will succeed in initializing the graphics most of the time. Some timing issue?