Re: [PATCH 3/3] vfs: return -EOVERFLOW in generic_remap_checks() when overflow check fails

[Christoph Hellwig <hch@infradead.org>]

On Fri, Sep 20, 2024 at 07:37:27AM -0700, Darrick J. Wong wrote:
> On Fri, Sep 20, 2024 at 07:19:28AM -0700, Christoph Hellwig wrote:
> > On Fri, Sep 20, 2024 at 08:30:22PM +0800, Julian Sun wrote:
> > > Keep it consistent with the handling of the same check within
> > > generic_copy_file_checks().
> > > Also, returning -EOVERFLOW in this case is more appropriate.
> > 
> > Maybe:
> > 
> > Keep the errno value consistent with the equivalent check in
> > generic_copy_file_checks() that returns -EOVERFLOW, which feels like the
> > more appropriate value to return compared to the overly generic -EINVAL.
> 
> The manpage for clone/dedupe/exchange don't say anything about
> EOVERFLOW, but they do have this to say about EINVAL:
> 
> EINVAL
> The  filesystem  does  not  support  reflinking the ranges of the given
> files.

Which isn't exactly the integer overflow case described here :)

> Does this errno code change cause any regressions in fstests?

Given our rather sparse test coverage of it I doubt it, but it
would be great to have that confirmed by the submitter.

While we're talking about that - a simple exerciser for the overflow
condition for xfstests would be very useful to have.