Thank you for your kind words. And I really agree with you in that kvm should better report error conditions. So here, I attach more information which might or might not be a hint.
Since I could get any help from vger, so I try to find the bug with gdb by myself. Because I do not know much about the kernel, I just traced the kvm executable. The segmentation fault was occured in kvm-all.c:1073 (kvm_vm_ioctl()), and its caller was kvm-all.c:215 (kvm_init_vcpu()). At that time, kvm_state is already null, so that s (==kvm_state) is delivered as null to the callee. The grand-caller was cpus.c:792 (qemu_kvm_cpu_thread_fn()). As you know, many functions in qemu-kvm are just based on kernel API's, so I had to stop tracing then. Anyway, we may add some guard code around those lines to avoid segmentation faults, even if it is not a really good solution. I really appreciate you. Have a good weekend ! Regards, Hae-woo Park. 2011/11/6 Michael Tokarev <[email protected]>: > On 05.11.2011 13:21, Hae-woo Park wrote: >> I use Gigabyte GA-MA78GM-US2H main-board, >> and I solved this problem by upgrading BIOS to version F9D (beta). >> I think the previous version (maybe F9C, but I'm not sure now) has some >> problem. > > For fun, it was the same motherboard where I did many experiments > with kvm some time ago. But I upgraded BIOS right away, because > of some unrelated thing (I think it was something to do with > automatic CPU fan speed control, but I don't remember anymore), > so I had no issues like this there. > > It's sad I don't have this board anymore - it'd be a nice experiment > to find out what's wrong with it + kvm. Kvm should better report > error conditions, especially in such obscure situations, and > definitely should not crash like you observed. -- The last part, > crash after problem enabling kvm in kernel - is still a problem. > >> You may close this bug. > > No, I think it should stay for now, due to second part of it which > is really a bug which is still present - it should not crash. > >> I really thank you, and sorry for bothering you. > > It is very difficult to know where the problem is. If kvm had better > error reporting, maybe this whole thing didn't exist and you didn't > have all this issues which you had. > > Thank you for your patience! > > /mjt > -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
