Description of problem: I am trying to do my database testing with raw devices. I defined devices as follows . . . <disk type='block' device='disk'> <source dev='/dev/perfvg/system_001'/> <target dev='vdz' bus='virtio'/> </disk> <disk type='block' device='disk'> <source dev='/dev/perfvg/temp_0_0'/> <target dev='vdaa' bus='virtio'/> </disk> <disk type='block' device='disk'> <source dev='/dev/perfvg/undo_sys_ts.dbf'/> <target dev='vdab' bus='virtio'/> </disk> <disk type='block' device='disk'> <source dev='/dev/perfvg/ware_0_0'/> <target dev='vdac' bus='virtio'/> </disk> . . . But in my guest I see devices only upto vdz. The vdaa, vdab and vdac devices were not created. After exchanging emails, it was narrowed down to libvirt code. src/util.c::virDiskNameToIndex() int idx = 0; ... while (*ptr) { idx = idx * 26; if (!c_islower(*ptr)) return -1; idx += *ptr - 'a'; <-- looks broken here ptr++; } Version-Release number of selected component (if applicable): How reproducible: Readily reproducible Steps to Reproduce: 1. Create KVM guest with raw devices exceeding vdz (eg vdaa, .. 2. Start guest and devices are created only upto vdz 3. Actual results: Expected results: Additional info:
Hum, there is a serious problem with this bug report, it states RhEL-5.3, but KVM is not supported on 5.3 and the libvirt for 5.3 really is not ready for KVM. I cannot infer from that bug report which version of libvirt is being used, and in any case KVM support for libvirt in 5.3 is not fixable with a simple device patch. Can you clarify the versions used ? If the fix is needed upstream, then the mailing-list is a better place to carry this request. Daniel
Note that the fix was commited upstream: https://www.redhat.com/archives/libvir-list/2008-October/msg00270.html but I'm not sure we can do this for RHEL-5.3 proper, virDiskNameToIndex doesn't exist in 0.3.3, only if we rebase in RHEL will this need to be applied. Daniel
This was simply mis-filed, should be oVirt.