Re: [PATCH] x86: sgx: Don't track poisoned pages for reclaiming

[Huang, Kai]

On 12/02/2025 12:31 pm, Dave Hansen wrote:
> On 2/11/25 13:18, Huang, Kai wrote:
> > > > This requires low-level SGX implementation knowledge to fully
> > > > understand. Both what "ETRACK, EBLOCK and EWB" are in the first place,
> > > > how they are involved in reclaim and also why EREMOVE doesn't lead to
> > > > the same fate.
> > >
> > > Does it? [I'll dig up Intel SDM to check this]
> > I just did. 🙂
> >
> > It seems EREMOVE only reads and updates the EPCM entry for the target
> > EPC page but won't actually access that EPC page.
>
> Actually, now that I think about it even more, why would ETRACK or
> EBLOCK access the page itself? They seem superficially like they'd be
> metadata-only too.

Looking at the pseudo code, AFAICT EBLOCK doesn't touch the EPC page 
either, but ETRACK actually reads SECS (ETRACK must take SECS page as 
input):

    (* All processors must have completed the previous tracking cycle*)
    IF ( (DS:RCX).TRACKING ≠ 0) )
	......

Here the DS:RCX is the SECS page.

I think this also is consistent with what Andrew said:

  "I haven't seen a crash in either of these (always in EWB), ..."

because a poisoned EPC page being regular enclave page has much higher 
possibility.  In fact, ETRACK only takes SECS page but I think the 
chance that the kernel code can still reach ETRACK while SECS page is 
poisoned should be just 0.