I'm using genkernel to build 2.6.6-rc3-mm1, currently I'm running gentoo-sources-2.4.22-r5. Genkernel freezes after saying "cleaning up and compressing initrd" and my computer goes into ~100% System CPU usage, being used by umount. Here's the end of genkernel's log: INSTALL drivers/ieee1394/video1394.ko INSTALL drivers/pcmcia/yenta_socket.ko if [ -r System.map ]; then /sbin/depmod -ae -F System.map 2.6.6-rc3-mm1; fi * Copying config for successful build to /etc/kernels/kernel-config-x86-2.6.6-rc3-mm1 * Dietlibc: Found bincache at /usr/share/genkernel/pkg/x86/dietlibc-0.24-x86.tar.bz2 * Busybox: Found bincache at /usr/share/genkernel/pkg/x86/busybox-1.00-pre7-losetup-crypto-alpha-x86.bz2 * module-init-tools: Found bincache at /usr/share/genkernel/pkg/x86/insmod-x86-static-2.6.bz2 * devfsd: Found bincache at /usr/share/genkernel/pkg/x86/devfsd-1.3.25-dietlibc-kernel25-x86.bz2 and /usr/share/genkernel/pkg/x86/devfsd-conf-0.24-x86.bz2 * initrd: creating base system * initrd: copying modules * initrd: copying auxilary files * initrd: calculating initrd size * initrd: calculated size 1306 + 100k slop for fs overhead * initrd: real size 1406 * initrd: creating loopback filesystem 1406+0 records in 1406+0 records out * initrd: moving initrd fs to loopback * initrd: cleaning up and compressing initrd This has happened twice in a row, the second time was after rebooting, I assume it would happen every time. Here are the genkernel arguments I used: genkernel all --menuconfig --bootsplash --no-clean --no-mrproper --install --kerneldir=/usr/src/linux-2.6.6-rc3-mm1 Reproducible: Always Steps to Reproduce: 1. 2. 3.
This doesn't happen every time, only sometimes (maybe every other time). This is with the latest genkernel, version 3.0.2a.
Could you try building a kernel without genkernel? Let's narrow down a genkernel bug or a kernel bug.
This is a kernel bug - the system hangs when unmounting a loop...
Does this happen on the 2.6.7-r2 kernel? Or just with a -mm kernel?
Please retest with different/newer kernels as already requested, and reopen if this is still an issue.