On Thu, 2007-07-05 at 11:28 +0400, Vasily Tarasov wrote: > On Wed, 2007-07-04 at 11:51 -0400, Erez Zadok wrote: > > Do you mean the branch-management ioctls? If that's the only difference, we > > can work around it easily (wrapper scripts, for example). Or did you write > > programs that use the ioctl(2) syscall directly? > > > > The branch-management ioctls in 1.0 were really racy; the remount-based > > branch-management is not just race free, but also the "approved" way that > > the kernel folks prefer. > > > > Unionfs 2.0 has so much new stuff, not to mention fixes; perhaps you'd > > consider using 2.0 for kernels 2.6.18 and higher, and special case 2.6.9? > > > > Or maybe I'll now have to see if we can backport the 2.6.18 code to 2.6.9? > > That's going to be an interesting adventure. :-) > > > > Erez. > > Hello, > > I meant the branch-management ioctls. And at the moment I'm in the > process of creating script for (re-)mounting unionfs 2.0 without these > iotctl on 2.6.20 kernel ;) If it'll be ok, we probably will use unionfs > 2.0 on 2.6.18 too. > > However I don't think that the races during branch-management ioctls are > severe problem in case of LiveCD. We only add branches once during boot: > no other actions ;) > > I tell you why we need 2.6.9. OpenVZ community supports now three > kernels: > 2.6.9-rhel4-based > 2.6.18-based > 2.6.18-rhel5-based > 2.6.20-based. > And some people prefer to use 2.6.9-rhel4-based kernels on the > rhel4/centos4 nodes, because they consider the system to be more > "solid". > > Thank you for your help I'll inform you about the progress. ;) > > Vasily.
Hello, I just want to inform you that we successfully use unionfs 2.0 with 2.6.20 kernel on our LiveCD. Thank you for the help, Vasily. _______________________________________________ unionfs mailing list: http://unionfs.filesystems.org/ [email protected] http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs
