From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from felloff.net ([199.191.58.38]) by pp; Sat Dec 6 06:19:23 EST 2014 Message-ID: List-ID: <9front.9front.org> X-Glyph: ➈ X-Bullshit: SSL realtime-scale service Date: Sat, 6 Dec 2014 12:19:04 +0100 From: cinap_lenrek@felloff.net To: 9front@9front.org Subject: Re: [9front] Booting EFI (or work-arounds to install 9front on GPT partition)? In-Reply-To: <5454299.WklWF47Z7a@krypton> MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit we currently have no gpt support in fdisk. but mischief is working on it. from the description, it seems the efi loader has some problems. i assume the "some sort of shell" refers to the boot console (see 9boot(8)). in one case, it seemed to be unable to locate the kernel file from the boot media. the bootloader; as a last action; prints "boot" before jumping to the kernel. if that just hangs, it means the kernel fails *before* it is able to attach to the GOP framebuffer. possible causes (speculation): - bad memory map (see *e820= boot parameter) - efi doesnt provide GOP (see *bootscreen= boot parameter) - maybe kernel image got corrupted (in memory?) due to wrong assumtptions - you might have tried to load an older kernel that has no *bootscreen= support - acpi? what we can do to get some evidence: hit any key on boot to get to the bootloader console (until you see the ">" prompt). then read what boot parameters where established by the loader: *e820=, *bootscreen=, and *acpi= you can use the "show" command to print the boot parameter environment. you can try serial console by specifying console=0 before booting the kernel. that way, we might see more early output of the kernel. if thats not an option, then all we can do commenting out code before bootscreeninit() in the kenrel to figure out where it fails. the efi loader has pxe boot support. so you might be able todo the debugging this way without having to copy stuff to usb or burn cdroms. keep in mind that the efi stuff is very fresh and has only been tested with thinkpads and qemu/vmware and vbox so far. it would be good to know what kind of machine this is. -- cinap