On Thu, May 26, 2022 at 04:49:09PM +0800, Lai Jiangshan wrote: > On Sat, May 21, 2022 at 9:16 PM Lai Jiangshan <jiangshanlai@xxxxxxxxx> wrote: > > > > From: Lai Jiangshan <jiangshan.ljs@xxxxxxxxxxxx> > > > > Current code uses mmu->pae_root, mmu->pml4_root, and mmu->pml5_root to > > setup special roots. The initialization code is complex and the roots > > are not associated with struct kvm_mmu_page which causes the code more > > complex. > > > > So add new local shadow pages to simplify it. > > > > The local shadow pages are associated with struct kvm_mmu_page and > > VCPU-local. > > > > The local shadow pages are created and freed when the roots are > > changed (or one-off) which can be optimized but not in the patchset > > since the re-creating is light way (in normal case only the struct > > kvm_mmu_page needs to be re-allocated and sp->spt doens't, because > > it is likely to be mmu->pae_root) > > > > The patchset also fixes a possible bug described in: > > https://lore.kernel.org/lkml/20220415103414.86555-1-jiangshanlai@xxxxxxxxx/ > > as patch1. > > > > Ping and please ignore patch1 and patch9. It would not cause any conflict > without patch1 and patch9 if both are ignored together. > > The fix is wrong (see new discussion in the above link). So the possible > correct fix will not have any conflict with this patchset of one-off > local shadow page. I don't want to add extra stuff in this patchset > anymore. Yeah I agree with splitting this fix out to a separate patchset, and ordered after this cleanup so it can be done in one patch. When you get around to it, can you also implement a kvm-unit-test to demonstrate the bug? It would be good to have a regression test. > > Thanks > Lai
