On Mon, 24 Mar 2008 09:46:50 -0400 Theodore Tso <tytso@xxxxxxx> wrote: > This turns out to be a kludge whose short comings show other problems. > The real problem is that most of the libext2fs isn't BLOCK_UNINIT > aware. So for example, if debugfs is used to write a file into the > filesystem, and the block group doesn't have an initialized bitmap, > the Wrong Thing will happen. More to the point, if you use mke2fs to > a 1k blocksize filesystem, and the journal is bigger than 16 megs, (or > with a 4k blocksize filesystem, if the journal is bigger than 512 > megs), you could easily end up allocating the journal into a block > group with BG_BLOCK_UNINIT. Oops. > > This wasn't that much of a big deal since up until now lazy_bg was > only used for debugging really big filesystems, and not much else. It > was a quick hack for debugging purposes only. But given that > uninititalized blockgroups are intended for more general use, we have > to make sure all of these corner cases are handled correctly. > > Just looking at it quickly, it seems like the right thing to do is > split setup_lazy_bg() into two parts. The first part sets > EXT2_BG_BLOCK_UNINIT for all block groups, and then we modify the > block allocation functions in lib/ext2fs to clear the BLOCK_UNINIT > flag --- and then later on, we update the bg_free_blocks_count and > s_free_blocks_count for the lazy_bg case. It seems that libext2fs is BLOCK_UNINIT aware since the only time we would need to update the BLOCK_UNINIT flag is if we update the group descriptors bg_free_blocks_count. In ext2fs_block_alloc_stats(), the EXT2_BG_BLOCK_UNINIT flag is unset when this happens. Having said that, I think there is a bug in this code if we do not use GDT_CSUM since setup_lazy_bg() would set bg_free_block_count to zero and this routine does nothing to set the right number of free blocks. If GDT_CSUM is not used, s_free_blocks_count would also need to be updated as well. Other than fixing ext2fs_block_alloc_stats(), I don't see anywhere else in libext2fs where we would need to add awareness to unset the BLOCK_UNINIT flag. -JRS -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
