Details

Reviewers

des
olce

Commits

rG404fbe6847b4: Fix getmntpoint(3) to operate as it is documented in its manual page.
rG45197d29ced2: Fix getmntpoint(3) to operate as it is documented in its manual page.
rG99bf680a8499: Fix getmntpoint(3) to operate as it is documented in its manual page.

Summary

Dag-Erling Smørgrav <des@FreeBSD.org> reoprts:
While working on quot(8), I noticed that getmntpoint(3) does not actually accept a device name “with or without /dev/ prepended to it” as the manual page and code comments claim.

This patch corrects the problem by prepending /dev/ to names that do not begin with a '/'.

Since there are now two places in the code that want to prepend /dev/ the code to do this is broken out into a separate function.

Test Plan

Note that this fix cannot be tested with the df(1) command as it does not use getmntpoint(3).

This program can be used to test getmntpoit(3): {F132597285}

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Event Timeline

mckusick created this revision.Oct 18 2025, 7:11 AM

Herald added a reviewer: olce. · View Herald TranscriptOct 18 2025, 7:11 AM

Herald added a subscriber: imp. · View Herald Transcript

mckusick requested review of this revision.Oct 18 2025, 7:11 AM

des requested changes to this revision.Oct 18 2025, 7:36 AM

des added inline comments.

lib/libutil/mntopts.c
151	Please use a caller-provided buffer instead
179	Missing a null check here

This revision now requires changes to proceed.Oct 18 2025, 7:36 AM

Respond to Dag-Erling's comments.

Responding to Dag-Erling's comments.

lib/libutil/mntopts.c
151	It takes more parameters, more lines of code, and generally makes the code harder to follow. That said, I have made the change.
179	stat(NULL, &statbuf) returns -1 (EBADF) which will have the correct response here. It seems pointless to add an explicit test for NULL.

Add comment on why it is not necessary to check for NULL return from prependdevtopath().

des added inline comments.Oct 20 2025, 5:43 PM

lib/libutil/mntopts.c
151	agreed, but making `getmntpoint()` thread-unsafe is worse, imo
179	`EBADF` isn't a very useful `errno` value for our caller, though. I think it would be better to restore the `errno` value from the original `stat(2)` call if `prependdevtopath()` fails.
208	this `else` is unnecessary

Get rid of unneeded else statement.

Respond to latest comments.

lib/libutil/mntopts.c
179	We make no claims about what systems calls we have done or what error they have returned. We simply return a pointer to a statfs structure or NULL. We do several systems calls to implement this routine all of which could fail for various reasons, No caller of this routine should make any use or interpretation of errno.