Re: Linux Xen PV CPA W^X violation false-positives

[Jason Andryuk <jandryuk@xxxxxxxxx>]

On Thu, Mar 28, 2024 at 9:00 AM Jürgen Groß <jgross@xxxxxxxx> wrote:
>
> Hi Jason,
>
> On 28.03.24 02:24, Jason Andryuk wrote:
> > On Wed, Mar 27, 2024 at 7:46 AM Jürgen Groß <jgross@xxxxxxxx> wrote:
> >>
> >> On 24.01.24 17:54, Jason Andryuk wrote:
> >>> +
> >>> +                     return new;
> >>> +             }
> >>> +     }
> >>> +
> >>>        end = start + npg * PAGE_SIZE - 1;
> >>>        WARN_ONCE(1, "CPA detected W^X violation: %016llx -> %016llx 
> >>> range: 0x%016lx - 0x%016lx PFN %lx\n",
> >>>                  (unsigned long long)pgprot_val(old),
> >>
> >> Jason, do you want to send a V2 with your Signed-off, or would you like me 
> >> to
> >> try upstreaming the patch?
> >
> > Hi Jürgen,
> >
> > Yes, please upstream your approach.  I wasn't sure how to deal with
> > it, so it was more of a bug report.
>
> The final solution was a bit more complicated, as there are some
> corner cases to be considered. OTOH it is now complete by looking
> at all used translation entries.
>
> Are you able to test the attached patch? I don't see the original
> issue and can only verify the patch doesn't cause any regression.

I'm no longer involved with OpenXT, but I reached out to some of the
developers.  Hopefully they try this out and respond here.

The backtrace in this thread is from BPF - I don't know how that was
triggered.  The other case I saw was in dom0.  That looked like it was
from freeing a module's (nouveau) .init section.  I don't seem to be
able to reproduce that on a non-OpenXT box.

Thanks,
Jason