------- Comment on attachment From [email protected] 2018-05-09 01:55 EDT-------
I notice that kvm_radix_test_clear_dirty() is not taking the kvm->mmu_lock, unlike all the other code that manipulates the partition-scoped page tables for the guest. Although I haven't been able to construct a scenario where this could cause a dirty bit to get lost, I think it is worth trying this patch, which makes kvm_radix_test_clear_dirty() take the kvm->mmu_lock and therefore serialize with the other code that manipulates ptes. This would not necessarily be the final patch but it should be good enough for testing. The patch is against the Ubuntu bionic 4.15.0-21.22 kernel. The other thing that could be worth trying is to test with a current upstream kernel. There have been changes in the partition-scoped page table management since 4.15, and it would be worth knowing whether the bug still exists upstream. ** Attachment added: "Patch to take mmu_lock in kvm_radix_test_clear_dirty" https://bugs.launchpad.net/bugs/1768115/+attachment/5136564/+files/dirty-take-lock.patch -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1768115 Title: ISST-LTE:KVM:Ubuntu1804:BostonLC:boslcp3g1: Migration guest running with IO stress crashed@security_file_permission+0xf4/0x160. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu-power-systems/+bug/1768115/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
