Package: linux-2.6 Severity: normal In the course of other testing, I need to boot systems using GRUB2 from a USB drive. To support systems that can't find their boot drive if the drive numbering changes, I use GRUB2's drivemap command like this:
set root=(hd1) drivemap (hd0) (hd1) chainloader +1 boot This results in GRUB2 hooking the e820 interrupt to reserve memory for its hook of the BIOS disk interrupt. However, note that this problem *also* occurs if GRUB2 hooks e820 and does not run drivemap (part of the testing involves allocating persistent memory for other purposes). This works fine with most OSes. It works with the Linux 2.6.26 kernel shipped in Lenny (specifically, linux-image-2.6.26-2-amd64 version 2.6.26-22lenny1). However, if I try it with 2.6.32, as shipped in unstable or backports, it fails to boot. From what I can see of the boot messages, and from the rare occasions when it manages to give me a shell in the initramfs with a working keyboard, it looks like Linux only detects 64MB of memory, goes nuts with the OOM killer on processes essential to booting (udev, for instance), and usually ends by panicing because it can't find something to kill. (Not surprising when booting a 16-way or 24-way SMP system with only 64MB of RAM.) I've manually verified the e820 map provided by the GRUB2-hooked interrupt (after booting a kernel that *does* boot), and it looks perfectly correct; thus, I don't think the problem lies in GRUB2, but in the kernel. Given the nature of the problem, it proves fairly difficult to extract the boot messages, but I can try to gather more information if you have a suggestion for what to look for. I may be able to bisect this in the future. - Josh Triplett -- System Information: Debian Release: 5.0.4 APT prefers stable APT policy: (500, 'stable') Architecture: amd64 (x86_64) Kernel: Linux 2.6.32-bpo.3-amd64 (SMP w/16 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/bash -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20100607031431.3006.88542.report...@scratch
