Hi there Rogier,
Why no tint? ;-) tint patch adds a checksum verification for the archive that will be downloaded later from Free Software Foundation during the compilation, so no security risk there. Perhaps the only downside of this tint tetris is that currently it's not compressed while being added to CBFS, so occupies about 90KB if I remember correctly, while that could be much less. Although there is a lot of free space at CBFS (>3 MB out of 4MB) while simply building a coreboot, if you'd like to add a lot of floppy-based OS later - even as few as 90KB might be an issue.
It's strange that you got a variable degree of this USB issue at different laptops, it should've been the same regardless of your dGPU type. Seems to be a Qubes-only problem, a bit random, and might be possible to fix if you'd read a note here - http://dangerousprototypes.com/docs/Lenovo_G505S_hacking#TODO_list . Here's a copy-paste :
" 2) Both XHCI options in Coreboot menu should be disabled (unless you'd like to add the XHCI firmware) or the left side ports won't work at all. These options are already disabled by default and all the ports are functioning as USB 2.0.
awokd tells:
"You may have to use irqpoll in sys-usb kernel options with Qubes OS. USB interrupts don't seem to be routing correctly in Coreboot, and the only way to use them in Qubes is with irqpoll in the kernel options." "
Perhaps awokd knows more about this issue, you could try to e-mail him directly.
P.S. Hope you don't mind that I CC a coreboot mailing list, this troubleshooting knowledge may be useful to someone else so should be public
On Tue, May 14, 2019 at 7:48 AM Rogier en Ingrid rogieringrid@gmail.com wrote:
I got a few issues - mainly with Qubes 4.0.1. Not included the tint patch, also changed the .config to not include tint. For the rest, included all your nice patches.
After installing Qubes 4.0.1, initialization of the OS, i noticed some errors when starting (will include them later today in 2nd mail) and mainly some weird USB port behavior: the laptop with dgpu R5-M230: none of the USB ports work when running Qubes (though all work at boot). the laptop with dgpu HD-8570M: usb 2.0 port does not work, both usb 3 on the left work. Only again when running Qubes, not when booting a live Ubuntu for example.
Mike Banon:
It's strange that you got a variable degree of this USB issue at different laptops, it should've been the same regardless of your dGPU type. Seems to be a Qubes-only problem, a bit random, and might be possible to fix if you'd read a note here - http://dangerousprototypes.com/docs/Lenovo_G505S_hacking#TODO_list . Here's a copy-paste :
" 2) Both XHCI options in Coreboot menu should be disabled (unless you'd like to add the XHCI firmware) or the left side ports won't work at all. These options are already disabled by default and all the ports are functioning as USB 2.0.
awokd tells:
"You may have to use irqpoll in sys-usb kernel options with Qubes OS. USB interrupts don't seem to be routing correctly in Coreboot, and the only way to use them in Qubes is with irqpoll in the kernel options." "
Perhaps awokd knows more about this issue, you could try to e-mail him directly.
Not much more than I wrote there. When I look at log messages, I see the VM trying to send interrupt #s that don't exist. Not sure how to troubleshoot further (am open to suggestions!), but irqpoll worked well enough. Maybe different GPUs have different interrupt tables, and the one supplied doesn't fit all? They each have slightly different revisions of the G505s board.
-
awokd -
Mike Banon