On Fri 08-04-11 19:36:56, Jean Delvare wrote:
On Thursday 07 April 2011 08:57:12 pm Jan Kara wrote:
Hello,
we are figuring out upstream how to eventually get rid of maintaing three code bases of ext2, ext3, and ext4 in parallel. These days ext4 kernel driver is able to handle both ext2 and ext3 filesystems in a backward compatible manner - i.e., you can mount, access, modify the filesystem with ext4 driver and then still be able to use it with the old ext2 driver. So I'd think that we could try changing config of our kernel in Factory so that ext4 driver is used for ext2 filesystem and see whether something breaks or not.
Do people have any opinion on this?
Wouldn't it be a safest path to replace the use cases of ext2 with ext_3_? It's way more mature than ext4, and presumably people have already been using ext3 to mount ext2 partitions for years.
There's a problem with ext3 replacing ext2 as ext3 adds journaling which adds visible performance overhead. I suppose people who had not migrated from ext2 to ext3 by now either care about the performance difference or just don't care about anything all :). In both cases ext4 should be
Or are using flash cards :-). Oh yeah, that's a good point (for other readers, some flash cards tend to have "interesting" properties invalidating assumptions any sane journalling filesystem has about the behavior of underlying device. Thus crashing in
On Fri 15-04-11 18:29:49, Pavel Machek wrote: the middle of write is going to corrupt filesystem on these regardless of journalling). But ext4 in nojournal mode should be fine in this case either :) Honza -- Jan Kara <jack@suse.cz> SUSE Labs, CR -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org