On 03/11/2013 03:01 PM, Marcus Schäfer wrote:
Hi Dirk, Hi Alex
I've debugged it further. the issue is that apparently kiwi does some kpartx -a magic, and then later reads through /dev/mapper/* to find out about the mappings (that it just created itself). For reasons that I don't understand, the /dev inside the build chroot does not get any files inside /dev/mapper/*. With the help from Alex we potentially found a fix: it seems the default of kpartx for "sync mode" changed between 12.2 and 12.3. changing the kpartx -a/-d calls to -sa/-sd (for enabling sync mode) might fix the issue (not fully tested yet, but a quick try looked promising).
Marcus, any opinion about this patch? Thanks much for the patch makes a lot sense. I also stumbled over that problem and could solve it by waiting for the udev queue to finish. looks like on 12.3 the creation of the devices is done via udev ?
Yup :). That fails for chroot mode though. It also fails for qemu-user - probably because the ARM libudev and the x86_64 udev daemon on the host don't understand each other.
The sync mode seems to be fine in addition to that
With sync mode, kpartx reverts to its old behavior where it creates the device nodes itself.
patch is added to master, will cherry pick to 12.3 as well
Awesome - thanks a lot! Alex -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org