On Fri, 2023-07-21 at 00:32 +0000, Huang, Kai wrote:
On Wed, 2023-07-19 at 08:53 -0500, Haitao Huang wrote:
Hi Dave and Kai On Tue, 18 Jul 2023 19:21:54 -0500, Dave Hansen dave.hansen@intel.com wrote:
On 7/18/23 17:14, Huang, Kai wrote:
Also perhaps the patch title is too vague. Adding more information doesn't hurt I think, e.g., mentioning it is a fix for NULL pointer dereference in the EAUG flow.
Yeah, let's say something like:
x86/sgx: Resolve SECS reclaim vs. page fault race
The patch is not to resolve SECS vs #PF race though the race is a necessary condition to cause the NULL pointer. The same condition does not cause NULL pointer in the ELDU path of #PF, only in EAUG path of #PF.
And the issue really is the NULL pointer not checked and fix was to reuse the same code to reload SECS in ELDU code path for EAUG code path
How about this:
x86/sgx: Reload reclaimed SECS for EAUG on #PF
or
x86/sgx: Fix a NULL pointer to SECS used for EAUG on #PF
Perhaps you can add "EAUG" part to what Dave suggested?
x86/sgx: Resolves SECS reclaim vs. page fault race on EAUG
(assuming Dave is fine with this :-))
Btw, do you have a real call trace? If you have, I think you can add that to the changelog too because that catches people's eye immediately.