Thank you for the feedback! On 10/16/26 23:35, Mouse wrote: > I'd be inclined to replace the second part with "or is root" (I can't > recall how that's spelled these days), not just drop it.
On 10/17/26 02:35, Michael van Elst wrote: > So a valid implementation could just ignore the flag (which could be done in the compat layer, if it isn't already). Sounds good on both counts, I will go ahead and attempt a patch that validates privileges as discussed (based on UID or is root), and otherwise ignore the flag. Thanks again. Aiman On 10/16/26 23:00, Aiman Najjar wrote: > Hello tech-kern, > > I'm new to NetBSD and I am interested in contributing to compat_linux. > > I've had a chance to look at the work Henrique did for the LTP as part > of GSoC, I am interested in working on some of the failing test cases. > > In particular, `open02` test case seems like a good candidate for a > first patch. The failing test case passes `O_NOATIME` mode flag to open, > and expects it to fail with `EPERM` error. > > My question: what is the correct way to address this, do we: 1) add > native support for `O_NOATIME`, which allows users to flag (or hint) > that they don't want the file's access time to be updated (useful for > backup solutions processing large amount of files), or 2) do we emulate > it, somehow, at `compat_linux` layer (e.g. only support this behavior > when the module is installed) or 3) Do we only check for user privileges > (which would make the test pass) but do not actually implement the behavior? > > Additionally, re. privilege check: Linux checks the privilege by > ensuring process effective UID matches the file owner UID, or if the > process has `CAP_FOWNER` [*]. On NetBSD, I presume we'd only do the > first check? > > Also, stumbled upon a similar discussion in FreeBSD forums: > https://forums.freebsd.org/threads/o_noatime-option-in-open.92523/ > > Thanks, > Aiman > > [*] > https://manpages.debian.org/trixie/manpages-dev/openat.2.en.html#O_NOATIME
