Bug 219132 - Redundant "re-mounted ro" message
Summary: Redundant "re-mounted ro" message
Status: NEW
Alias: None
Product: File System
Classification: Unclassified
Component: ext4 (show other bugs)
Hardware: All Linux
: P3 low
Assignee: fs_ext4@kernel-bugs.osdl.org
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2024-08-06 18:20 UTC by Artem S. Tashkinov
Modified: 2024-08-07 13:41 UTC (History)
1 user (show)

See Also:
Kernel Version:
Subsystem:
Regression: No
Bisected commit-id:


Attachments

Description Artem S. Tashkinov 2024-08-06 18:20:06 UTC
I'm not a fan of ext4 logging this:

EXT4-fs (nvme0n1p3): re-mounted UUID ro. Quota mode: none.

when the filesystem was already mounted RO.

It's confusing and may lead the user to believe the FS was mounted RW previously.

I'd like the ext4 driver to ignore or not log RO remount requests when the FS is already mounted RO.
Comment 1 Artem S. Tashkinov 2024-08-06 19:14:09 UTC
I guess hiding this message might break someone's workflow, so let's do it better:

EXT4-fs (nvme0n1p3): Ignoring the RO remount request, there's nothing to do.

or something like that.
Comment 2 Theodore Tso 2024-08-07 03:33:16 UTC
The suggestion in #1 isn't necessarily correct, though, because we could be remounting to change some other superblock options.  For example:

mount -o remount,errors=continue /dev/vdc

Fundamentally, whenever runs "mount -o remount ...", we issue the "EXT4-fs (DEVICE): re-mounted ..." message.   The fact that we print the ro/rw and quota mode is completely arbitrary and more due to historical reasons than anything else.   For example, the fact that we print the quota mode is pretty much useless in this day and era and there are plenty of other mount option/state that would probably be much more useful.

So the fact that we print the ro/rw state doesn't imply that it has changed.  For example, if the file system is mounted read/write and we change the errors= mode from continue to remount-ro, etc., we will print:

EXT4-fs (nvme0n1p3): re-mounted UUID rw. Quota mode: none.

... and it doesn't mean that we changed the rw/ro mode from ro to r/w.   

Should we change the behavior to something else?   Perhaps, although to be honest it's not the highest priority thing for me.   I could see dropping the quota mode and only printing the message when the r/o state changes.  Or maybe we display any mount option that changes (which would be a lot more work).  Is it worth the effort?  Meh.....
.
Comment 3 Artem S. Tashkinov 2024-08-07 09:53:16 UTC
Log messages normally convey what has actually been done.

The current messaging implies the kernel has indeed remounted something which is not the case. That's my concern.
Comment 4 Theodore Tso 2024-08-07 13:41:45 UTC
I hear you, and I agree that this is a reasonable interpretation of the log message, even if it's never been the case.   If you'd like to submit a patch which makes the change where it only logs when the ro/rw state changes, and drops the quota mode, I'd certainly accept it.

Note You need to log in before you can comment on or make changes to this bug.