Re: [PATCH RFC 7/7] xfs: error tag to force zeroing on debug kernels

[Brian Foster <bfoster@redhat.com>]

On Tue, Jun 10, 2025 at 06:30:29AM -0700, Christoph Hellwig wrote:
> On Tue, Jun 10, 2025 at 08:26:45AM -0400, Brian Foster wrote:
> > Well that is kind of the question.. ;) My preference was to either add
> > something to fstests to enable select errortags by default on every
> > mount (or do the same in-kernel via XFS_DEBUG[_ERRTAGS] or some such)
> > over just creating a one-off test that runs fsx or whatever with this
> > error tag turned on. [1].
> > 
> > That said, I wouldn't be opposed to just doing both if folks prefer
> > that. It just bugs me to add yet another test that only runs a specific
> > fsx test when we get much more coverage by running the full suite of
> > tests. IOW, whenever somebody is testing a kernel that would actually
> > run a custom test (XFS_DEBUG plus specific errortag support), we could
> > in theory be running the whole suite with the same errortag turned on
> > (albeit perhaps at a lesser frequency than a custom test would use). So
> > from that perspective I'm not sure it makes a whole lot of sense to do
> > both.
> > 
> > So any thoughts from anyone on a custom test vs. enabling errortag
> > defaults (via fstests or kernel) vs. some combination of both?
> 
> I definitively like a targeted test to exercise it.  If you want
> additional knows to turn on error tags that's probably fine if it
> works out.  I'm worried about adding more flags to xfstests because
> it makes it really hard to figure out what runs are need for good
> test coverage.
> 
> 

Yeah, an fstests variable would add yet another configuration to test,
which maybe defeats the point. But we could still turn on certain tags
by default in the kernel. For example, see the couple of open coded
get_random_u32_below() callsites in XFS where we already effectively do
this for XFS_DEBUG, they just aren't implemented as proper errortags.

I think the main thing that would need to change is to not xfs_warn() on
those knobs when they are enabled by default. I think there are a few
different ways that could possibly be done, ideally so we go back to
default/warn behavior when userspace makes an explicit errortag change,
but I'd have to play around with it a little bit. Hm?

Anyways, given the fstests config matrix concern I'm inclined to at
least give something like that a try first and then fall back to a
custom test if that fails or is objectionable for some other reason..

Brian