On Thu, 2009-06-04 at 10:11 +0200, Thorsten Leemhuis wrote: > On 04.06.2009 09:29, Hans Kristian Rosbach wrote: > > I was unsure whether to post this feature request on the bugzilla > > or here, and since I found no feature request queue; here it is. > > Not much time for a long answer right now, but here is a short one: It > maybe impossible to do what you want (not sure, haven't looked that > close at the patch), but even if not it's likely a bad idea, as > dm-raid45 will be needed quite early during boot which leads to all > sorts of interesting problems and corner cases. > > Maybe asking the Kernel-Developers in Fedora to integrate raid45 might > be a good idea and the first step. Normally they are quite reluctant to > include out-of-tree code, but the code is maintained by a RH developer > and the feature afaics is in RHEL, so it might be a bit different in > this case.
If it is not possible, then that is fine. But I am not thinking about using the raid as a boot/system drive. The problem we have at the moment is that there is no way to get at the data at all from fedora. The 6 disks in raid5 were used as a data store and for backups, but at the moment we'd have to reinstall windows on that computer in order to get access to the data at all. Having a 3rd party kernel module that could be loaded dynamically or even manually would make it possible to access the data in order to back it up and migrate to running md-raid instead. If the module would allow anything more it would be great, but even just read-only functionality would be awesome for those trapped by this problem. There are also hpt18xx highpoint (open source and closed source) modules that can be built given just the kernel headers, but unfortunately those are so out-of-date that they do not compile against newer kernels. So atleast for us, that solution is out. Sincerly Hans K. Rosbach
