Hi Andrew, On 09/12/2022 22:18, Andrew Cooper wrote:
On 09/12/2022 18:51, Julien Grall wrote: That said, moving the creation side of things doesn't change the teardown requirements. When I get time to respin the fault_ttl series, Gitlab CI will be able to demonstrate the bug I keep on telling you is still there, the fix for which is taking the patch I already wrote for you.
Is it just a matter of rebasing the series? If so, I am happy to give a try to see what comes up...
There is no correct way to move the final loop out of complete_domain_destroy(), even if in the general case you can make it more preemptible by moving the allocation later.
Hmmm... I don't think the call to p2m_teardown() in complete_domain_destroy() will be necessary once we move the P2M allocation out of arch_domain_create().
This is because there will be no use of the P2M after domain_relinquish_resources().
Cheers, -- Julien Grall