Re: [libvirt] [PATCH v2 2/3] qemu: Handle device mapper targets properly

On 03/26/2018 05:17 PM, Peter Krempa wrote: > On Mon, Mar 26, 2018 at 16:43:02 +0200, Michal Privoznik wrote: >> https://bugzilla.redhat.com/show_bug.cgi?id=1557769 >> >> Problem with device mapper targets is that there can be several >> other devices 'hidden' behind them. For instance, /dev/dm-1 can >> consist of /dev/sda, /dev/sdb and /dev/sdc. Therefore, when >> setting up devices CGroup and namespaces we have to take this >> into account. >> >> Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com&gt; >> ---

>> + for (i = 0; i < nmaj; i++) { >> + if (virAsprintf(&devPath, "/dev/block/%u:%u", maj[i], min[i]) < 0) >> + goto cleanup; >> + >> + if (qemuDomainCreateDevice(devPath, data, false) < 0) >> + goto cleanup; > > So now that I see this new version, this part starts looking suspicious > to me. Since this did not care much that the path changed, is it really > necessary to create the /dev/ entries in the container? > > Looks like even device mapper is returning them as the node > specificator, so I'd presume it really does not matter if they are > present. > > More specifically we can't really reverse engineer from the major:minor > numbers which actual path the user used so it should not really be > necessary for it to be present in the container. Yes, looks like I was too eager trying to fix this bug. I've rebuilt libvirt without qemu_domain.c change (so only CGroup code was modified) and the bug still did not reproduce. So I guess namespace changes are not necessary after all. I'll drop them.