mbox series

[v2,0/4] KVM: x86/mmu: Clean up indirect_shadow_pages usage

Message ID 20240203002343.383056-1-seanjc@google.com
Headers
Series KVM: x86/mmu: Clean up indirect_shadow_pages usage |

Message

Sean Christopherson Feb. 3, 2024, 12:23 a.m. UTC 
  Resurrect a 6 month old patch from Mingwei, add a few cleanps on top, and
fix a largely theoretical race between emulating writes and write-protecting
shadow pages.  At least, I'm pretty sure there's a race.  Memory ordering
isn't exactly my strong suit.

v2:
 - Drop the unnecessary READ_ONCE(). [Jim]
 - Cleanup more old crud in reexecute_instruction().
 - Fix the aforementioned race.

v1: https://lore.kernel.org/all/20230605004334.1930091-1-mizhang@google.com

Mingwei Zhang (1):
  KVM: x86/mmu: Don't acquire mmu_lock when using indirect_shadow_pages
    as a heuristic

Sean Christopherson (3):
  KVM: x86: Drop dedicated logic for direct MMUs in
    reexecute_instruction()
  KVM: x86: Drop superfluous check on direct MMU vs. WRITE_PF_TO_SP flag
  KVM: x86/mmu: Fix a *very* theoretical race in kvm_mmu_track_write()

 arch/x86/kvm/mmu/mmu.c | 19 ++++++++++++++++---
 arch/x86/kvm/x86.c     | 35 ++++++++++++++---------------------
 2 files changed, 30 insertions(+), 24 deletions(-)


base-commit: 60eedcfceda9db46f1b333e5e1aa9359793f04fb

Comments

Sean Christopherson Feb. 23, 2024, 1:35 a.m. UTC | #1 
On Fri, 02 Feb 2024 16:23:39 -0800, Sean Christopherson wrote:
> Resurrect a 6 month old patch from Mingwei, add a few cleanps on top, and
> fix a largely theoretical race between emulating writes and write-protecting
> shadow pages.  At least, I'm pretty sure there's a race.  Memory ordering
> isn't exactly my strong suit.
> 
> v2:
>  - Drop the unnecessary READ_ONCE(). [Jim]
>  - Cleanup more old crud in reexecute_instruction().
>  - Fix the aforementioned race.
> 
> [...]

Applied 1-3 to kvm-x86 mmu.  I will likely apply patch 4 for 6.9 as well, but
assuming it doesn't get attention "soon", I'll apply it dead last in kvm-x86/mmu
so that it's easy to discard if my paranoia is unfounded.

[1/4] KVM: x86/mmu: Don't acquire mmu_lock when using indirect_shadow_pages as a heuristic
      https://github.com/kvm-x86/linux/commit/474b99ed703b
[2/4] KVM: x86: Drop dedicated logic for direct MMUs in reexecute_instruction()
      https://github.com/kvm-x86/linux/commit/515c18a64e70
[3/4] KVM: x86: Drop superfluous check on direct MMU vs. WRITE_PF_TO_SP flag
      https://github.com/kvm-x86/linux/commit/dfeef3d3f310
[4/4] KVM: x86/mmu: Fix a *very* theoretical race in kvm_mmu_track_write()
      (not applied, yet)

--
https://github.com/kvm-x86/linux/tree/next