Colin Watson wrote: > On Fri, Jan 09, 2009 at 12:41:21PM +0000, Colin Watson wrote: >> Here's a version of my patch merged with yours. Although I have not yet >> been able to test them extensively, the block count patches are >> particularly important; without them, parted will detect the wrong >> geometry for a >2TB ext4 filesystem. > > And of course it turns out to be rather hard to test these, because > ext3's maximum file size is 2TB and we need 2^32 4KB blocks (i.e. 16TB) > before needing to use the high longword of the block count, so it's > rather hard to create this situation with the sparse loopback file > approach that would work reasonably in parted's test suite. I can't even > create an ext4 filesystem and put a loopback file in that (not that that > would be viable in parted's test suite) because ext4's maximum file size > is 16TB. Any ideas? >
How about using a sparse block device instead of a loopback device backed by a sparse file? The principal is the same but it uses device-mapper's COW target to provide the sparsness rather than the file system sparse file support. You just need to set up a relatively small writable backing store and a very large device that uses the device-mapper zero target (a block version of /dev/zero). You then snapshot the zero device using the backing store and snapshot target and this gives a writable block device that is the same size as the zero device used. I've used this heavily for testing file systems much larger than 16T - obviously, it won't work if you need to fill the fs with lots of data but should cover the same set of situations as a loopback device can. This approach was used during the development of ext4 and worked quite well there. Heinze Mauelshagen has a pair of scripts that automate the setup/remove steps: http://www.bullopensource.org/ext4/files/sparse_remove http://www.bullopensource.org/ext4/files/sparse_create http://www.bullopensource.org/ext4/files/sparse.txt Regards, Bryn. _______________________________________________ parted-devel mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/parted-devel
