On Tue, Apr 06, 2021 at 08:09:26AM +0200, Paolo Bonzini wrote:
On 05/04/21 10:54, Greg Kroah-Hartman wrote:
From: Ben Gardon bgardon@google.com
[ Upstream commit 9a77daacc87dee9fd63e31243f21894132ed8407 ]
To prepare for handling page faults in parallel, change the TDP MMU page fault handler to use atomic operations to set SPTEs so that changes are not lost if multiple threads attempt to modify the same SPTE.
Reviewed-by: Peter Feiner pfeiner@google.com Signed-off-by: Ben Gardon bgardon@google.com
Whoa no, you have included basically a whole new feature, except for the final patch that actually enables the feature. The whole new MMU
Right, we would usually grab dependencies rather than modifying the patch. It means we diverge less with upstream, and custom backports tend to be buggier than just grabbing dependencies.
is still not meant to be used in production and development is still happening as of 5.13.
Unrelated to this disucssion, but how are folks supposed to know which feature can and which feature can't be used in production? If it's a released kernel, in theory anyone can pick up 5.12 and use it in production.
Were all these patches (82-97) included just to enable patch 98 ("KVM: x86/mmu: Ensure TLBs are flushed for TDP MMU during NX zapping")? Same for 105-120 in 5.11.
Yup. Is there anything wrong with those patches?