Message-ID: <20250113020925.18789-1-yan.y.zhao@intel.com>
Date: Mon, 13 Jan 2025 10:09:25 +0800
From: Yan Zhao <yan.y.zhao@...el.com>
To: pbonzini@...hat.com,
	seanjc@...gle.com,
	kvm@...r.kernel.org
Cc: linux-kernel@...r.kernel.org,
	rick.p.edgecombe@...el.com,
	kai.huang@...el.com,
	adrian.hunter@...el.com,
	reinette.chatre@...el.com,
	xiaoyao.li@...el.com,
	tony.lindgren@...el.com,
	binbin.wu@...ux.intel.com,
	dmatlack@...gle.com,
	isaku.yamahata@...el.com,
	isaku.yamahata@...il.com,
	Yan Zhao <yan.y.zhao@...el.com>
Subject: [PATCH 0/7] KVM: TDX SEPT SEAMCALL retry

This series aims to provide a clean solution to avoid the blind retries in
the previous hack [1] in "TDX MMU Part 2," following the initial
discussions to [2], further discussions in the RFC, and the PUCK [3].

A full analysis of the lock status for each SEAMCALL relevant to KVM is
available at [4].

This series categorizes the SEPT-related SEAMCALLs (used for page
installation and uninstallation) into three groups:

Group 1: tdh_mem_page_add().
       - Invoked only during TD build time.
       - Proposal: Return -EBUSY on TDX_OPERAND_BUSY.
       - Patch 1.

Group 2: tdh_mem_sept_add(), tdh_mem_page_aug().
       - Invoked for TD runtime page installation.
       - Proposal: Retry locally in the TDX EPT violation handler for
         RET_PF_RETRY.
       - Patches 2-3.

Group 3: tdh_mem_range_block(), tdh_mem_track(), tdh_mem_page_remove().
       - Invoked for page uninstallation, with KVM mmu_lock held for write.
       - Proposal: Kick off vCPUs and no vCPU entry on TDX_OPERAND_BUSY.
       - Patch 4.

Patches 5/6/7 are fixup patches:
Patch 5: Return -EBUSY instead of -EAGAIN when tdh_mem_sept_add() is busy.
Patch 6: Remove the retry loop for tdh_phymem_page_wbinvd_hkid().
Patch 7: Warn on force_immediate_exit in tdx_vcpu_run().

Code base: kvm-coco-queue 2f30b837bf7b.
Applies to the tail since the dependence on
commit 8e801e55ba8f ("KVM: TDX: Handle EPT violation/misconfig exit"),

Thanks
Yan

RFC --> v1:
- Split patch 1 in RFC into patches 1,2,3,5, and add new fixup patches 6/7.
- Add contention analysis of tdh_mem_page_add() in patch 1 log.
- Provide justification in patch 2 log and add checks for RET_PF_CONTINUE.
- Use "a per-VM flag wait_for_sept_zap + KVM_REQ_OUTSIDE_GUEST_MODE"
  instead of a arch-specific request to prevent vCPUs from TD entry in patch 4
  (Sean).

RFC: https://lore.kernel.org/all/20241121115139.26338-1-yan.y.zhao@intel.com
[1] https://lore.kernel.org/all/20241112073909.22326-1-yan.y.zhao@intel.com
[2] https://lore.kernel.org/kvm/20240904030751.117579-10-rick.p.edgecombe@intel.com/
[3] https://drive.google.com/drive/folders/1k0qOarKuZXpzRsKDtVeC5Lpl9-amJ6AJ?resourcekey=0-l9uVpVEBC34Uar1ReaqisQ
[4] https://lore.kernel.org/kvm/ZuP5eNXFCljzRgWo@yzhao56-desk.sh.intel.com

Yan Zhao (7):
  KVM: TDX: Return -EBUSY when tdh_mem_page_add() encounters
    TDX_OPERAND_BUSY
  KVM: x86/mmu: Return RET_PF* instead of 1 in kvm_mmu_page_fault()
  KVM: TDX: Retry locally in TDX EPT violation handler on RET_PF_RETRY
  KVM: TDX: Kick off vCPUs when SEAMCALL is busy during TD page removal
  fixup! KVM: TDX: Implement hooks to propagate changes of TDP MMU
    mirror page table
  fixup! KVM: TDX: Implement hooks to propagate changes of TDP MMU
    mirror page table
  fixup! KVM: TDX: Implement TDX vcpu enter/exit path

 arch/x86/kvm/mmu/mmu.c          |  10 ++-
 arch/x86/kvm/mmu/mmu_internal.h |  12 ++-
 arch/x86/kvm/vmx/tdx.c          | 135 ++++++++++++++++++++++++++------
 arch/x86/kvm/vmx/tdx.h          |   7 ++
 4 files changed, 134 insertions(+), 30 deletions(-)

-- 
2.43.2

Powered by blists - more mailing lists