Hide Forgot
Description of problem: It takes qemu-kvm about 40 seconds to start booting the kernel if the -initrd specifies a initramfs of 150 MB. Once the kernel starts booting, it boots within 4 seconds, all the way through systemd to the console. qemu-kvm seems to be doing something strange. Version-Release number of selected component (if applicable): qemu-system-x86-0.14.0-7.fc15.x86_64 kernel 2.6.38.8-32.fc15.x86_64 How reproducible: Always Steps to Reproduce: 1. sudo yum -y --releasever=15 --installroot `pwd`/root install bash systemd 2. cd root; ln -s /sbin/init 3. find . | sudo cpio -o -H newc | gzip > ../initramfs.img 4. qemu-kvm -m 4096M -nographic -net none -kernel root/boot/vmlinuz-2.6.38.8-35.fc15.x86_64 -append "console=ttyS0 selinux=0 clocksource=hpet ipv6.disable=1 plymouth.enable=0 rd.plymouth=0" -initrd initramfs.img Actual results: qemu-kvm takes about 40 seconds before the kernel starts booting. Expected results: qemu-kvm should boot fast Additional info: Hardware is a Mac Pro with 2* Intel Xeon E5462 CPUs and 20 GB RAM.
This package has changed ownership in the Fedora Package Database. Reassigning to the new owner of this component.
Changing to rawhide, since this basically affects all QEMU versions in Fedora & we don't want it auto-closed when F15 goes EOL.
*** Bug 618720 has been marked as a duplicate of this bug. ***
Rawhide bugs just tend to linger, so moving to F17. I thought I saw commits go by about fixing this with qemu 1.1, but can't find them now. Here's older info from Rich's report: The root cause is possibly: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=7972995b0c346de76 See discussion here: http://lists.gnu.org/archive/html/qemu-devel/2010-08/threads.html#00133
I'm fairly sure this is going to be a problem in qemu forever since upstream devs don't want to change the way that the initrd is loaded to make it more efficient. In libguestfs we sidestepped the issue by not using an initrd to load the appliance.
I'm quite certain that more recent qemu has improved on this, so closing. And as Rich mentioned even if there are still issues, the substantial speedups were rejected in the past.