The "too small to fail" memory-allocation rule

Posted Dec 24, 2014 5:13 UTC (Wed) by zblaxell (subscriber, #26385)
In reply to: The "too small to fail" memory-allocation rule by flussence
Parent article: The "too small to fail" memory-allocation rule

> This isn't the first time I've seen memory management problems and XFS mentioned in the same article. Is that just a side effect of more eyes on the code compared to other FSes, or is there something more to it?

If you're comparing XFS to vfat or ext3, remember that those are really simple filesystems. Almost everything is in fixed-size arrays or bitmaps, and disk blocks have a 1:1 relationship with page cache. Not much interesting for MM to do.

Comparing XFS to reiserfs, NTFS, ZFS: these filesystems have similar complexity, but they're not developed any more, or their best implementations are not in the Linux kernel proper, so their developers don't have newsworthy interactions with MM developers.

Comparing XFS to btrfs: btrfs has MM bugs. but it also has lots of other kinds of bugs, so the MM bugs don't stand out as much. ;)

If you count up lines of code and commits in the last two years, XFS is one of the most actively developed filesystems in Linux, second only to btrfs among non-network filesystems by those crude metrics.