On Aug 20, 2008 16:58 -0700, Mingming Cao wrote: > > Also, if you are doing this by changing the aops on the inode, isn't > > it possible that a large write starts outside the EXT4_MIN_FREE_BLOCKS > > boundary and then still runs out of space without changing the aops? > > > Instead it is maybe better to do the check at the start of > > ext4_da_write_begin() and if it fails then call the non-delalloc > > write_begin from there? > > Yeah that's better. > > But I realize a problem. Actually now I think we can't fall back to > nondelalloc mode if the inode has any dirty pages in the page cache, as > those pages need delalloc aops ->ext4_da_writepages() to handle delayed > allocation writeout.. That is only if you are changing the aops, which I already don't think is a good idea. Instead, as I suggest there should be a check in the ext4_da_write_begin() that makes the decision to delay the allocation or to actually do it, based on the current free blocks. Instead of doing this check repeatedly (if we think it is expensive, maybe it isn't), it might make sense to set a superblock flag that indicates "no more delalloc, too close to ENOSPC" and this is cleared when blocks are freed beyond the low watermark. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. -- 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
