Re: zfs question about atime update

[Takashi YAMAMOTO <yamt9999%gmail.com@localhost>]

On Tue, Mar 17, 2026 at 10:45 PM Jason Thorpe <thorpej%me.com@localhost> wrote:
>
>
>
> > On Mar 17, 2026, at 5:36 AM, Takashi YAMAMOTO <yamt9999%gmail.com@localhost> wrote:
> >
> > hi,
> >
> > our zfs has a logic to flush atime updates in zfs_netbsd_sync:
> >>      /*
> >>       * On NetBSD, we need to push out atime updates.  Solaris does
> >>       * this during VOP_INACTIVE, but that does not work well with the
> >>       * BSD VFS, so we do it in batch here.
> >>       */
> >
> > does anyone here remember why it "doesn't work well with the BSD VFS"?
> >
> > i'm running a kernel with a local patch to put it back to VOP_INACTIVE.
> > i haven't noticed anything wrong with it so far.
>
> Indeed, seems like a fine place to do it.  In fact, I’d go so far as to say it should be done even lazier than INACTIVE; do it in RECLAIM.

well, actually, the current code does it in VFS_SYNC _and_ VOP_RECLAIM.

i prefer to do it in INACTIVE as it's what the upstream zfs does.
less divergence from the upstream is better as we will need to update
the codebase sooner or later.

>
> -- thorpej
>