On Sun, 10 Dec 2023 07:56:36 +0000 Stephan =?ISO-8859- 1?Q?Verb=FCcheln?= <verbuech...@posteo.de> wrote: > > - Is it safe to assume that other filesystems (like BTRFS) are not > affected? > - Does this cause filesystem corruption or only file corruption? > - Does this affect metadata or only file contents? > - Is there a way to detect corrupted files? > - If metadata is not affected, is it safe to assume that all files with > a modification date older than the update are fine? > - Does it help to shut down services (such as Apache) or the whole > machine until the fix is available?
The bug appears to be triggered when an ->end_io handler returns a non- zero value to iomap after a direct IO write. It looks like the ext4 handler is the only one that returns non-zero in kernel 6.1.64, so for now one can assume that only ext4 filesystems are affected. The bug corrupts file data during a direct write operation, so I would also assume that files last modified before 6.1.64 was installed will not be corrupted. As far as I can tell, the corruption only affects file data (not metadata) but perhaps someone with more kernel experience than me can confirm. Best regards, Dale Richards
