On Fri, 2017-12-08 at 07:31 -0800, Christopherson, Sean J wrote:
> Jarkko Sakkinen <jarkko.sakkinen@xxxxxxxxxxxxxxx> wrote:
> > On Thu, Dec 07, 2017 at 02:46:39PM +0000, Christopherson, Sean J wrote:
> > > > + for (i = 0; i < 2; i++) {
> > > > + va_page = list_first_entry(&encl->va_pages,
> > > > + struct sgx_va_page, list);
> > > > + va_offset = sgx_alloc_va_slot(va_page);
> > > > + if (va_offset < PAGE_SIZE)
> > > > + break;
> > > > +
> > > > + list_move_tail(&va_page->list, &encl->va_pages);
> > > > + }
> > >
> > > This is broken, there is no guarantee that the next VA page will have
> > > a free slot. You have to walk over all VA pages to guarantee a slot
> > > is found, e.g. this caused EWB and ELDU errors.
> >
> > I did run some extensive stress tests on this and did not experience any
> > issues. Full VA pages are always put to the end. Please point me to the
> > test where this breaks so that I can fix the issue if it persists.
>
> Three VA pages in the enclave: A, B and C. Evict all pages in the
> enclave, i.e. consume all slots in A, B and C. The list can be in
> any order at this point, but for the sake of argument let's say the
> order is C->A->B, i.e. C was originally the last VA page in the list.
> Fault in page X, whose VA is in B. Evict X. This code looks at C
> and A, and finds no available slot, but continues with VA page A and
> a va_offset of PAGE_SIZE.
So it looks like you avoid the described case by moving B to the head of
the list in sgx_eldu. The bug I am seeing is still straightforward to
theorize:
1. Three VA pages. List = A->B->C
2. Fill A and B, use one entry in C. List = C->B->A
3. ELDU, freeing a slot in B. List = B->C->A
4. EWB, consuming the last slot in B. List = B->C->A
5. ELDU, freeing a slot in A. List = A->B->C
6. EWB, consuming the last slot in A. List = A->B->C
7. ELDU, but both A and B are full
8. Explode
