On Thu, Mar 21, 2019 at 03:30:35PM -0700, Randy Dunlap wrote: > I briefly scanned but did not look terribly closely: > > Does this patch series ignore ext3 filesystems that are being handled > by the ext4fs code? Like all ext2/3/4 features, new functionality is gated by a feature bit in the compat, rocompat, or incompat bitmasks. A file system implementation (e.g., in the Linux kernel, NetBSD/FreeBSD, Grub, et. al) which sees a feature bit it doesn't recognize in an rocompat bitmask will only allow the file system to be mounted read-only. If it sees a bit in the incompat bitmask it doesn't recognize, it won't allow the file system to be mounted at all. The kernel doesn't care if it finds a bit it doesn't recognize in the compat bitmask; however, e2fsck will not touch a file system that has a compat bit which it doesn't know how to deal with. Strictly speaking, there is no such thing as an "ext3 file system". There are just file systems with a set of feature bits set. And the set of features supported by NetBSD or the Grub installer may not fall neatly into the "ext2", "ext3", or "ext4" buckets. If by "ext3 file system" you mean a file system with those featire bits set by mke2fs -t ext3 by default (e.g., not overridden by the -O option, or by the system administrator editing /etc/mke2fs.conf), then yes, this patch series will not result in any substantive change for "ext3 file systems", since it by default "mke2fs -t ext3" or "mkfs.ext3" will not enable the feature flag that turns on case-insensitive support. For more information, see the ext4 and mke2fs.conf manual pages. Cheers, - Ted
