On Mon, Oct 24, 2022 at 10:14:03AM -0500, Haitao Huang wrote:
> Hi Jarkko,
>
> On Sun, 23 Oct 2022 16:23:30 -0500, Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> wrote:
> ...
> > > + /* Here: vm_start <= pos < end <= vm_end */
> > > + while (pos < end) {
> > > + if (xa_load(&encl->page_array, PFN_DOWN(pos)))
> > > + continue;
> > > + if (signal_pending(current)) {
> > > + if (pos == start)
> > > + ret = -ERESTARTSYS;
> > > + else
> > > + ret = -EINTR;
> > > + goto unlock;
> > > + }
> > > + ret = sgx_encl_eaug_page(vma, encl, pos);
> >
> > You should instead just do the export in the previous commit,
> > and convert the return values here. Less pollution to the
> > existing code base.
> >
>
> Thanks for the feedback. Just to clarify, you prefer following:
>
> 1. do export only in separate patch, no return type and value change
> 2. change return type and value in this patch.
>
> Note the sgx_fadvise function needs to stop EAUGing and return in case of
> -EBUSY. So sgx_encl_eaug_page has to return error code -EBUSY separately,
> instead of lumping -EBUSY with VM_FAULT_NOPAGE which was fine for page fault
> handler.
I guess the problem is that any of this is not explained in the commit
message, i.e. should already answer to this question. My preference is
argumented solution.
BR, Jarkko
