Re: [PATCH v33 10/21] mm: Introduce vm_ops->may_mprotect()

From: Jarkko Sakkinen
Date: Thu Jun 25 2020 - 18:40:19 EST


On Thu, Jun 25, 2020 at 11:06:46AM -0700, Sean Christopherson wrote:
> On Thu, Jun 25, 2020 at 06:30:50PM +0100, Matthew Wilcox wrote:
> > On Thu, Jun 25, 2020 at 07:14:16PM +0200, Borislav Petkov wrote:
> > > On Thu, Jun 18, 2020 at 01:08:32AM +0300, Jarkko Sakkinen wrote:
> > > > diff --git a/mm/mprotect.c b/mm/mprotect.c
> > > > index ce8b8a5eacbb..f7731dc13ff0 100644
> > > > --- a/mm/mprotect.c
> > > > +++ b/mm/mprotect.c
> > > > @@ -603,13 +603,21 @@ static int do_mprotect_pkey(unsigned long start, size_t len,
> > > > goto out;
> > > > }
> > > >
> > > > + tmp = vma->vm_end;
> > > > + if (tmp > end)
> > > > + tmp = end;
> > > > +
> > > > + if (vma->vm_ops && vma->vm_ops->may_mprotect) {
> > > > + error = vma->vm_ops->may_mprotect(vma, nstart, tmp,
> > > > + prot);
> > > > + if (error)
> > > > + goto out;
> > > > + }
> > > > +
> > > > error = security_file_mprotect(vma, reqprot, prot);
> > > > if (error)
> > > > goto out;
> > > >
> >
> > I think the right way to do this is:
> >
> > error = security_file_mprotect(vma, reqprot, prot);
> > if (error)
> > goto out;
> >
> > tmp = vma->vm_end;
> > if (tmp > end)
> > tmp = end;
> > + if (vma->vm_ops->mprotect)
> > + error = vma->vm_ops->mprotect(vma, &prev, nstart, tmp,
> > + newflags);
> > + else
> > + error = mprotect_fixup(vma, &prev, nstart, tmp,
> > + newflags);
> > - error = mprotect_fixup(vma, &prev, nstart, tmp, newflags);
> > if (error)
> > goto out;
> >
> > and then the vma owner can do whatever it needs to before calling
> > mprotect_fixup(), which is already not static.
>
> I'm certainly not opposed to a straight ->mprotect() hook. ->may_protect()
> came about because I/we thought it would be less objectionable to allow the
> vma owner to apply additional restrictions as opposed to a wholesale
> replacement.

Can you send fixes to associated patches to linux-sgx (i.e. this and
driver)?

> > (how did we get to v33 with this kind of problem still in the patch set?)
>
> Because no one from the mm world has looked at it. Which is completely
> understandable because it's a giant patch set and the first 25 or so versions
> were spent sorting out fundamental architectural/design issue (there have
> been a _lot_ of speed bumps), e.g. the need for hooking mprotect() didn't
> even come about until v21.

Actually also because we did not have an explicit linux-mm CC.

/Jarkko