[PATCH v2 0/8] KVM: x86/mmu: Use separate namespaces gPTEs and SPTEs

[PATCH v2 0/8] KVM: x86/mmu: Use separate namespaces gPTEs and SPTEs

[Sean Christopherson]

This is based on kvm/queue + "Remove unused PT32_DIR_BASE_ADDR_MASK from mmu.c",
so I think it should apply cleanly on top of what you already grabbed.

Clean up the x86 MMU code to use separate namespaces for guest PTEs and
shadow PTEs.  While there is most definitely overlap, especially in
attribute bits, the rules for walking gPTEs and for generating/walking
SPTEs differ in subtle ways, e.g. see commit fc9bf2e087ef ("KVM: x86/mmu:
Do not apply HPA (memory encryption) mask to GPAs").  The paging32 macros
in particular should never be used outside of paging_tmpl.h.

Separating gPTEs from SPTEs actually provides for nice cleanups (see the
diffstat) as KVM has ended up with a fair bit of copy+paste code that can
be deduplicated once KVM isn't trying to use PT64_* defines for both gPTEs
and SPTEs.

This is a spiritual successor to patches 4-7 of the series[*] that added
the aformentenioned commit.

v2:
  - Don't move is_cpuid_PSE36(). [Lai Jiangshan]
  - Change author for patch 1 (yeah, I was lazy). [Lai Jiangshan]
  - Fix a random typo in patch 7's changelog.
  - Rebase to play nice with PT32_DIR_BASE_ADDR_MASK.

v1: https://lore.kernel.org/all/20220613225723.2734132-2-seanjc@google.com

[*] https://lore.kernel.org/all/20210623230552.4027702-1-seanjc@google.com

Lai Jiangshan (1):
  KVM: x86/mmu: Drop unused CMPXCHG macro from paging_tmpl.h

Sean Christopherson (7):
  KVM: VMX: Refactor 32-bit PSE PT creation to avoid using MMU macro
  KVM: x86/mmu: Bury 32-bit PSE paging helpers in paging_tmpl.h
  KVM: x86/mmu: Dedup macros for computing various page table masks
  KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs
  KVM: x86/mmu: Use common macros to compute 32/64-bit paging masks
  KVM: x86/mmu: Truncate paging32's PT_BASE_ADDR_MASK to 32 bits
  KVM: x86/mmu: Use common logic for computing the 32/64-bit base PA
    mask

 arch/x86/kvm/mmu.h              | 10 ------
 arch/x86/kvm/mmu/mmu.c          | 57 +++++++++------------------------
 arch/x86/kvm/mmu/mmu_internal.h | 17 ++++++++++
 arch/x86/kvm/mmu/paging.h       | 14 --------
 arch/x86/kvm/mmu/paging_tmpl.h  | 55 ++++++++++++++++---------------
 arch/x86/kvm/mmu/spte.c         |  2 +-
 arch/x86/kvm/mmu/spte.h         | 28 +++++++---------
 arch/x86/kvm/mmu/tdp_iter.c     |  6 ++--
 arch/x86/kvm/mmu/tdp_mmu.c      |  6 ++--
 arch/x86/kvm/vmx/vmx.c          |  2 +-
 10 files changed, 82 insertions(+), 115 deletions(-)
 delete mode 100644 arch/x86/kvm/mmu/paging.h


base-commit: ccf1c220033d8d6fe50d8d11daa3dec5640f8c4d
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 1/8] KVM: x86/mmu: Drop unused CMPXCHG macro from paging_tmpl.h

[Sean Christopherson]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Drop the CMPXCHG macro from paging_tmpl.h, it's no longer used now that
KVM uses a common uaccess helper to do 8-byte CMPXCHG.

Fixes: f122dfe44768 ("KVM: x86: Use __try_cmpxchg_user() to update guest PTE A/D bits")
Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
[sean: drop only CMPXCHG, update changelog accordingly]
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/paging_tmpl.h | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index fe35d8fd3276..f595c4b8657f 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -34,7 +34,6 @@
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) true
 	#ifdef CONFIG_X86_64
 	#define PT_MAX_FULL_LEVELS PT64_ROOT_MAX_LEVEL
-	#define CMPXCHG "cmpxchgq"
 	#else
 	#define PT_MAX_FULL_LEVELS 2
 	#endif
@@ -51,7 +50,6 @@
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
 	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) true
-	#define CMPXCHG "cmpxchgl"
 #elif PTTYPE == PTTYPE_EPT
 	#define pt_element_t u64
 	#define guest_walker guest_walkerEPT
@@ -64,9 +62,6 @@
 	#define PT_GUEST_DIRTY_SHIFT 9
 	#define PT_GUEST_ACCESSED_SHIFT 8
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) (!(mmu)->cpu_role.base.ad_disabled)
-	#ifdef CONFIG_X86_64
-	#define CMPXCHG "cmpxchgq"
-	#endif
 	#define PT_MAX_FULL_LEVELS PT64_ROOT_MAX_LEVEL
 #else
 	#error Invalid PTTYPE value
@@ -1100,7 +1095,6 @@ static int FNAME(sync_page)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
 #undef PT_MAX_FULL_LEVELS
 #undef gpte_to_gfn
 #undef gpte_to_gfn_lvl
-#undef CMPXCHG
 #undef PT_GUEST_ACCESSED_MASK
 #undef PT_GUEST_DIRTY_MASK
 #undef PT_GUEST_DIRTY_SHIFT
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 2/8] KVM: VMX: Refactor 32-bit PSE PT creation to avoid using MMU macro

[Sean Christopherson]

Compute the number of PTEs to be filled for the 32-bit PSE page tables
using the page size and the size of each entry.  While using the MMU's
PT32_ENT_PER_PAGE macro is arguably better in isolation, removing VMX's
usage will allow a future namespacing cleanup to move the guest page
table macros into paging_tmpl.h, out of the reach of code that isn't
directly related to shadow paging.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/vmx/vmx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 5e14e4c40007..b774f8c1b952 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -3704,7 +3704,7 @@ static int init_rmode_identity_map(struct kvm *kvm)
 	}
 
 	/* Set up identity-mapping pagetable for EPT in real mode */
-	for (i = 0; i < PT32_ENT_PER_PAGE; i++) {
+	for (i = 0; i < (PAGE_SIZE / sizeof(tmp)); i++) {
 		tmp = (i << 22) + (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER |
 			_PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_PSE);
 		if (__copy_to_user(uaddr + i * sizeof(tmp), &tmp, sizeof(tmp))) {
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 3/8] KVM: x86/mmu: Bury 32-bit PSE paging helpers in paging_tmpl.h

[Sean Christopherson]

Move a handful of one-off macros and helpers for 32-bit PSE paging into
paging_tmpl.h and hide them behind "PTTYPE == 32".  Under no circumstance
should anything but 32-bit shadow paging care about PSE paging.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu.h             |  5 -----
 arch/x86/kvm/mmu/mmu.c         |  7 -------
 arch/x86/kvm/mmu/paging_tmpl.h | 18 +++++++++++++++++-
 3 files changed, 17 insertions(+), 13 deletions(-)

diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index f8192864b496..d1021e34ac15 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -34,11 +34,6 @@
 #define PT_DIR_PAT_SHIFT 12
 #define PT_DIR_PAT_MASK (1ULL << PT_DIR_PAT_SHIFT)
 
-#define PT32_DIR_PSE36_SIZE 4
-#define PT32_DIR_PSE36_SHIFT 13
-#define PT32_DIR_PSE36_MASK \
-	(((1ULL << PT32_DIR_PSE36_SIZE) - 1) << PT32_DIR_PSE36_SHIFT)
-
 #define PT64_ROOT_5LEVEL 5
 #define PT64_ROOT_4LEVEL 4
 #define PT32_ROOT_LEVEL 2
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index f168693695bd..73497da1a99b 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -324,13 +324,6 @@ static int is_cpuid_PSE36(void)
 	return 1;
 }
 
-static gfn_t pse36_gfn_delta(u32 gpte)
-{
-	int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;
-
-	return (gpte & PT32_DIR_PSE36_MASK) << shift;
-}
-
 #ifdef CONFIG_X86_64
 static void __set_spte(u64 *sptep, u64 spte)
 {
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index f595c4b8657f..55fd35b1b227 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -50,6 +50,11 @@
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
 	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) true
+
+	#define PT32_DIR_PSE36_SIZE 4
+	#define PT32_DIR_PSE36_SHIFT 13
+	#define PT32_DIR_PSE36_MASK \
+		(((1ULL << PT32_DIR_PSE36_SIZE) - 1) << PT32_DIR_PSE36_SHIFT)
 #elif PTTYPE == PTTYPE_EPT
 	#define pt_element_t u64
 	#define guest_walker guest_walkerEPT
@@ -92,6 +97,15 @@ struct guest_walker {
 	struct x86_exception fault;
 };
 
+#if PTTYPE == 32
+static inline gfn_t pse36_gfn_delta(u32 gpte)
+{
+	int shift = 32 - PT32_DIR_PSE36_SHIFT - PAGE_SHIFT;
+
+	return (gpte & PT32_DIR_PSE36_MASK) << shift;
+}
+#endif
+
 static gfn_t gpte_to_gfn_lvl(pt_element_t gpte, int lvl)
 {
 	return (gpte & PT_LVL_ADDR_MASK(lvl)) >> PAGE_SHIFT;
@@ -416,8 +430,10 @@ static int FNAME(walk_addr_generic)(struct guest_walker *walker,
 	gfn = gpte_to_gfn_lvl(pte, walker->level);
 	gfn += (addr & PT_LVL_OFFSET_MASK(walker->level)) >> PAGE_SHIFT;
 
-	if (PTTYPE == 32 && walker->level > PG_LEVEL_4K && is_cpuid_PSE36())
+#if PTTYPE == 32
+	if (walker->level > PG_LEVEL_4K && is_cpuid_PSE36())
 		gfn += pse36_gfn_delta(pte);
+#endif
 
 	real_gpa = kvm_translate_gpa(vcpu, mmu, gfn_to_gpa(gfn), access, &walker->fault);
 	if (real_gpa == UNMAPPED_GVA)
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 4/8] KVM: x86/mmu: Dedup macros for computing various page table masks

[Sean Christopherson]

Provide common helper macros to generate various masks, shifts, etc...
for 32-bit vs. 64-bit page tables.  Only the inputs differ, the actual
calculations are identical.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu.h              |  4 ++--
 arch/x86/kvm/mmu/mmu.c          | 14 +++++---------
 arch/x86/kvm/mmu/mmu_internal.h | 14 ++++++++++++++
 arch/x86/kvm/mmu/paging.h       |  9 +++++----
 arch/x86/kvm/mmu/spte.h         |  7 +++----
 5 files changed, 29 insertions(+), 19 deletions(-)

diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index d1021e34ac15..6efe6bd7fb6e 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -7,9 +7,9 @@
 #include "cpuid.h"
 
 #define PT64_PT_BITS 9
-#define PT64_ENT_PER_PAGE (1 << PT64_PT_BITS)
+#define PT64_ENT_PER_PAGE __PT_ENT_PER_PAGE(PT64_PT_BITS)
 #define PT32_PT_BITS 10
-#define PT32_ENT_PER_PAGE (1 << PT32_PT_BITS)
+#define PT32_ENT_PER_PAGE __PT_ENT_PER_PAGE(PT32_PT_BITS)
 
 #define PT_WRITABLE_SHIFT 1
 #define PT_USER_SHIFT 2
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 73497da1a99b..b3edff05a53a 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -113,21 +113,17 @@ module_param(dbg, bool, 0644);
 
 #define PT32_LEVEL_BITS 10
 
-#define PT32_LEVEL_SHIFT(level) \
-		(PAGE_SHIFT + (level - 1) * PT32_LEVEL_BITS)
+#define PT32_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT32_LEVEL_BITS)
 
 #define PT32_LVL_OFFSET_MASK(level) \
-	(PT32_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
-						* PT32_LEVEL_BITS))) - 1))
-
-#define PT32_INDEX(address, level)\
-	(((address) >> PT32_LEVEL_SHIFT(level)) & ((1 << PT32_LEVEL_BITS) - 1))
+	__PT_LVL_OFFSET_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
 
+#define PT32_INDEX(address, level) __PT_INDEX(address, level, PT32_LEVEL_BITS)
 
 #define PT32_BASE_ADDR_MASK PAGE_MASK
+
 #define PT32_LVL_ADDR_MASK(level) \
-	(PAGE_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
-					    * PT32_LEVEL_BITS))) - 1))
+	__PT_LVL_ADDR_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
 
 #include <trace/events/kvm.h>
 
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index bd2a26897b97..5e1e3c8f8aaa 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -20,6 +20,20 @@ extern bool dbg;
 #define MMU_WARN_ON(x) do { } while (0)
 #endif
 
+/* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
+#define __PT_LEVEL_SHIFT(level, bits_per_level)	\
+	(PAGE_SHIFT + ((level) - 1) * (bits_per_level))
+#define __PT_INDEX(address, level, bits_per_level) \
+	(((address) >> __PT_LEVEL_SHIFT(level, bits_per_level)) & ((1 << (bits_per_level)) - 1))
+
+#define __PT_LVL_ADDR_MASK(base_addr_mask, level, bits_per_level) \
+	((base_addr_mask) & ~((1ULL << (PAGE_SHIFT + (((level) - 1) * (bits_per_level)))) - 1))
+
+#define __PT_LVL_OFFSET_MASK(base_addr_mask, level, bits_per_level) \
+	((base_addr_mask) & ((1ULL << (PAGE_SHIFT + (((level) - 1) * (bits_per_level)))) - 1))
+
+#define __PT_ENT_PER_PAGE(bits_per_level)  (1 << (bits_per_level))
+
 /*
  * Unlike regular MMU roots, PAE "roots", a.k.a. PDPTEs/PDPTRs, have a PRESENT
  * bit, and thus are guaranteed to be non-zero when valid.  And, when a guest
diff --git a/arch/x86/kvm/mmu/paging.h b/arch/x86/kvm/mmu/paging.h
index de8ab323bb70..23f3f64b8092 100644
--- a/arch/x86/kvm/mmu/paging.h
+++ b/arch/x86/kvm/mmu/paging.h
@@ -4,11 +4,12 @@
 #define __KVM_X86_PAGING_H
 
 #define GUEST_PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
+
 #define PT64_LVL_ADDR_MASK(level) \
-	(GUEST_PT64_BASE_ADDR_MASK & ~((1ULL << (PAGE_SHIFT + (((level) - 1) \
-						* PT64_LEVEL_BITS))) - 1))
+	__PT_LVL_ADDR_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
+
 #define PT64_LVL_OFFSET_MASK(level) \
-	(GUEST_PT64_BASE_ADDR_MASK & ((1ULL << (PAGE_SHIFT + (((level) - 1) \
-						* PT64_LEVEL_BITS))) - 1))
+	__PT_LVL_OFFSET_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
+
 #endif /* __KVM_X86_PAGING_H */
 
diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
index 0127bb6e3c7d..d5a8183b7232 100644
--- a/arch/x86/kvm/mmu/spte.h
+++ b/arch/x86/kvm/mmu/spte.h
@@ -55,11 +55,10 @@ static_assert(SPTE_TDP_AD_ENABLED_MASK == 0);
 
 #define PT64_LEVEL_BITS 9
 
-#define PT64_LEVEL_SHIFT(level) \
-		(PAGE_SHIFT + (level - 1) * PT64_LEVEL_BITS)
+#define PT64_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT64_LEVEL_BITS)
+
+#define PT64_INDEX(address, level) __PT_INDEX(address, level, PT64_LEVEL_BITS)
 
-#define PT64_INDEX(address, level)\
-	(((address) >> PT64_LEVEL_SHIFT(level)) & ((1 << PT64_LEVEL_BITS) - 1))
 #define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
 
 /*
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 5/8] KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs

[Sean Christopherson]

Separate the macros for KVM's shadow PTEs (SPTE) from guest 64-bit PTEs
(PT64).  SPTE and PT64 are _mostly_ the same, but the few differences are
quite critical, e.g. *_BASE_ADDR_MASK must differentiate between host and
guest physical address spaces, and SPTE_PERM_MASK (was PT64_PERM_MASK) is
very much specific to SPTEs.

Opportunistically (and temporarily) move most guest macros into paging.h
to clearly associate them with shadow paging, and to ensure that they're
not used as of this commit.  A future patch will eliminate them entirely.

Sadly, PT32_LEVEL_BITS is left behind in mmu_internal.h because it's
needed for the quadrant calculation in kvm_mmu_get_page().  The quadrant
calculation is hot enough (when using shadow paging with 32-bit guests)
that adding a per-context helper is undesirable, and burying the
computation in paging_tmpl.h with a forward declaration isn't exactly an
improvement.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu.h              |  5 ----
 arch/x86/kvm/mmu/mmu.c          | 44 +++++++++++----------------------
 arch/x86/kvm/mmu/mmu_internal.h |  3 +++
 arch/x86/kvm/mmu/paging.h       | 17 +++++++++++++
 arch/x86/kvm/mmu/paging_tmpl.h  |  4 +--
 arch/x86/kvm/mmu/spte.c         |  2 +-
 arch/x86/kvm/mmu/spte.h         | 27 +++++++++-----------
 arch/x86/kvm/mmu/tdp_iter.c     |  6 ++---
 arch/x86/kvm/mmu/tdp_mmu.c      |  6 ++---
 9 files changed, 56 insertions(+), 58 deletions(-)

diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index 6efe6bd7fb6e..a99acec925eb 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -6,11 +6,6 @@
 #include "kvm_cache_regs.h"
 #include "cpuid.h"
 
-#define PT64_PT_BITS 9
-#define PT64_ENT_PER_PAGE __PT_ENT_PER_PAGE(PT64_PT_BITS)
-#define PT32_PT_BITS 10
-#define PT32_ENT_PER_PAGE __PT_ENT_PER_PAGE(PT32_PT_BITS)
-
 #define PT_WRITABLE_SHIFT 1
 #define PT_USER_SHIFT 2
 
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index b3edff05a53a..81f2e58dc85b 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -111,20 +111,6 @@ module_param(dbg, bool, 0644);
 
 #define PTE_PREFETCH_NUM		8
 
-#define PT32_LEVEL_BITS 10
-
-#define PT32_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT32_LEVEL_BITS)
-
-#define PT32_LVL_OFFSET_MASK(level) \
-	__PT_LVL_OFFSET_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
-
-#define PT32_INDEX(address, level) __PT_INDEX(address, level, PT32_LEVEL_BITS)
-
-#define PT32_BASE_ADDR_MASK PAGE_MASK
-
-#define PT32_LVL_ADDR_MASK(level) \
-	__PT_LVL_ADDR_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
-
 #include <trace/events/kvm.h>
 
 /* make pte_list_desc fit well in cache lines */
@@ -704,7 +690,7 @@ static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
 	if (!sp->role.direct)
 		return sp->gfns[index];
 
-	return sp->gfn + (index << ((sp->role.level - 1) * PT64_LEVEL_BITS));
+	return sp->gfn + (index << ((sp->role.level - 1) * SPTE_LEVEL_BITS));
 }
 
 static void kvm_mmu_page_set_gfn(struct kvm_mmu_page *sp, int index, gfn_t gfn)
@@ -1776,7 +1762,7 @@ static int __mmu_unsync_walk(struct kvm_mmu_page *sp,
 			continue;
 		}
 
-		child = to_shadow_page(ent & PT64_BASE_ADDR_MASK);
+		child = to_shadow_page(ent & SPTE_BASE_ADDR_MASK);
 
 		if (child->unsync_children) {
 			if (mmu_pages_add(pvec, child, i))
@@ -2027,8 +2013,8 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
 	role.direct = direct;
 	role.access = access;
 	if (role.has_4_byte_gpte) {
-		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
-		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
+		quadrant = gaddr >> (PAGE_SHIFT + (SPTE_LEVEL_BITS * level));
+		quadrant &= (1 << ((PT32_LEVEL_BITS - SPTE_LEVEL_BITS) * level)) - 1;
 		role.quadrant = quadrant;
 	}
 	if (level <= vcpu->arch.mmu->cpu_role.base.level)
@@ -2132,7 +2118,7 @@ static void shadow_walk_init_using_root(struct kvm_shadow_walk_iterator *iterato
 
 		iterator->shadow_addr
 			= vcpu->arch.mmu->pae_root[(addr >> 30) & 3];
-		iterator->shadow_addr &= PT64_BASE_ADDR_MASK;
+		iterator->shadow_addr &= SPTE_BASE_ADDR_MASK;
 		--iterator->level;
 		if (!iterator->shadow_addr)
 			iterator->level = 0;
@@ -2151,7 +2137,7 @@ static bool shadow_walk_okay(struct kvm_shadow_walk_iterator *iterator)
 	if (iterator->level < PG_LEVEL_4K)
 		return false;
 
-	iterator->index = SHADOW_PT_INDEX(iterator->addr, iterator->level);
+	iterator->index = SPTE_INDEX(iterator->addr, iterator->level);
 	iterator->sptep	= ((u64 *)__va(iterator->shadow_addr)) + iterator->index;
 	return true;
 }
@@ -2164,7 +2150,7 @@ static void __shadow_walk_next(struct kvm_shadow_walk_iterator *iterator,
 		return;
 	}
 
-	iterator->shadow_addr = spte & PT64_BASE_ADDR_MASK;
+	iterator->shadow_addr = spte & SPTE_BASE_ADDR_MASK;
 	--iterator->level;
 }
 
@@ -2203,7 +2189,7 @@ static void validate_direct_spte(struct kvm_vcpu *vcpu, u64 *sptep,
 		 * so we should update the spte at this point to get
 		 * a new sp with the correct access.
 		 */
-		child = to_shadow_page(*sptep & PT64_BASE_ADDR_MASK);
+		child = to_shadow_page(*sptep & SPTE_BASE_ADDR_MASK);
 		if (child->role.access == direct_access)
 			return;
 
@@ -2224,7 +2210,7 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
 		if (is_last_spte(pte, sp->role.level)) {
 			drop_spte(kvm, spte);
 		} else {
-			child = to_shadow_page(pte & PT64_BASE_ADDR_MASK);
+			child = to_shadow_page(pte & SPTE_BASE_ADDR_MASK);
 			drop_parent_pte(child, spte);
 
 			/*
@@ -2250,7 +2236,7 @@ static int kvm_mmu_page_unlink_children(struct kvm *kvm,
 	int zapped = 0;
 	unsigned i;
 
-	for (i = 0; i < PT64_ENT_PER_PAGE; ++i)
+	for (i = 0; i < SPTE_ENT_PER_PAGE; ++i)
 		zapped += mmu_page_zap_pte(kvm, sp, sp->spt + i, invalid_list);
 
 	return zapped;
@@ -2663,7 +2649,7 @@ static int mmu_set_spte(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot,
 			struct kvm_mmu_page *child;
 			u64 pte = *sptep;
 
-			child = to_shadow_page(pte & PT64_BASE_ADDR_MASK);
+			child = to_shadow_page(pte & SPTE_BASE_ADDR_MASK);
 			drop_parent_pte(child, sptep);
 			flush = true;
 		} else if (pfn != spte_to_pfn(*sptep)) {
@@ -3252,7 +3238,7 @@ static void mmu_free_root_page(struct kvm *kvm, hpa_t *root_hpa,
 	if (!VALID_PAGE(*root_hpa))
 		return;
 
-	sp = to_shadow_page(*root_hpa & PT64_BASE_ADDR_MASK);
+	sp = to_shadow_page(*root_hpa & SPTE_BASE_ADDR_MASK);
 	if (WARN_ON(!sp))
 		return;
 
@@ -3724,7 +3710,7 @@ void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu)
 		hpa_t root = vcpu->arch.mmu->pae_root[i];
 
 		if (IS_VALID_PAE_ROOT(root)) {
-			root &= PT64_BASE_ADDR_MASK;
+			root &= SPTE_BASE_ADDR_MASK;
 			sp = to_shadow_page(root);
 			mmu_sync_children(vcpu, sp, true);
 		}
@@ -5186,11 +5172,11 @@ static bool need_remote_flush(u64 old, u64 new)
 		return false;
 	if (!is_shadow_present_pte(new))
 		return true;
-	if ((old ^ new) & PT64_BASE_ADDR_MASK)
+	if ((old ^ new) & SPTE_BASE_ADDR_MASK)
 		return true;
 	old ^= shadow_nx_mask;
 	new ^= shadow_nx_mask;
-	return (old & ~new & PT64_PERM_MASK) != 0;
+	return (old & ~new & SPTE_PERM_MASK) != 0;
 }
 
 static u64 mmu_pte_write_fetch_gpte(struct kvm_vcpu *vcpu, gpa_t *gpa,
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index 5e1e3c8f8aaa..cb9d4d358335 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -20,6 +20,9 @@ extern bool dbg;
 #define MMU_WARN_ON(x) do { } while (0)
 #endif
 
+/* The number of bits for 32-bit PTEs is to needed compute the quandrant. */
+#define PT32_LEVEL_BITS 10
+
 /* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
 #define __PT_LEVEL_SHIFT(level, bits_per_level)	\
 	(PAGE_SHIFT + ((level) - 1) * (bits_per_level))
diff --git a/arch/x86/kvm/mmu/paging.h b/arch/x86/kvm/mmu/paging.h
index 23f3f64b8092..6a63727cc7e8 100644
--- a/arch/x86/kvm/mmu/paging.h
+++ b/arch/x86/kvm/mmu/paging.h
@@ -5,11 +5,28 @@
 
 #define GUEST_PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
 
+#define PT64_LEVEL_BITS 9
+
+#define PT64_INDEX(address, level) __PT_INDEX(address, level, PT64_LEVEL_BITS)
+
 #define PT64_LVL_ADDR_MASK(level) \
 	__PT_LVL_ADDR_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
 
 #define PT64_LVL_OFFSET_MASK(level) \
 	__PT_LVL_OFFSET_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
 
+
+#define PT32_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT32_LEVEL_BITS)
+
+#define PT32_LVL_OFFSET_MASK(level) \
+	__PT_LVL_OFFSET_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
+
+#define PT32_INDEX(address, level) __PT_INDEX(address, level, PT32_LEVEL_BITS)
+
+#define PT32_BASE_ADDR_MASK PAGE_MASK
+
+#define PT32_LVL_ADDR_MASK(level) \
+	__PT_LVL_ADDR_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
+
 #endif /* __KVM_X86_PAGING_H */
 
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 55fd35b1b227..d68cc7a5ef81 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -899,7 +899,7 @@ static gpa_t FNAME(get_level1_sp_gpa)(struct kvm_mmu_page *sp)
 	WARN_ON(sp->role.level != PG_LEVEL_4K);
 
 	if (PTTYPE == 32)
-		offset = sp->role.quadrant << PT64_LEVEL_BITS;
+		offset = sp->role.quadrant << SPTE_LEVEL_BITS;
 
 	return gfn_to_gpa(sp->gfn) + offset * sizeof(pt_element_t);
 }
@@ -1034,7 +1034,7 @@ static int FNAME(sync_page)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
 
 	first_pte_gpa = FNAME(get_level1_sp_gpa)(sp);
 
-	for (i = 0; i < PT64_ENT_PER_PAGE; i++) {
+	for (i = 0; i < SPTE_ENT_PER_PAGE; i++) {
 		u64 *sptep, spte;
 		struct kvm_memory_slot *slot;
 		unsigned pte_access;
diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
index cda1851ec155..242e4828d7df 100644
--- a/arch/x86/kvm/mmu/spte.c
+++ b/arch/x86/kvm/mmu/spte.c
@@ -301,7 +301,7 @@ u64 kvm_mmu_changed_pte_notifier_make_spte(u64 old_spte, kvm_pfn_t new_pfn)
 {
 	u64 new_spte;
 
-	new_spte = old_spte & ~PT64_BASE_ADDR_MASK;
+	new_spte = old_spte & ~SPTE_BASE_ADDR_MASK;
 	new_spte |= (u64)new_pfn << PAGE_SHIFT;
 
 	new_spte &= ~PT_WRITABLE_MASK;
diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
index d5a8183b7232..121c5eaaec77 100644
--- a/arch/x86/kvm/mmu/spte.h
+++ b/arch/x86/kvm/mmu/spte.h
@@ -36,12 +36,12 @@ extern bool __read_mostly enable_mmio_caching;
 static_assert(SPTE_TDP_AD_ENABLED_MASK == 0);
 
 #ifdef CONFIG_DYNAMIC_PHYSICAL_MASK
-#define PT64_BASE_ADDR_MASK (physical_mask & ~(u64)(PAGE_SIZE-1))
+#define SPTE_BASE_ADDR_MASK (physical_mask & ~(u64)(PAGE_SIZE-1))
 #else
-#define PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
+#define SPTE_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
 #endif
 
-#define PT64_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | shadow_user_mask \
+#define SPTE_PERM_MASK (PT_PRESENT_MASK | PT_WRITABLE_MASK | shadow_user_mask \
 			| shadow_x_mask | shadow_nx_mask | shadow_me_mask)
 
 #define ACC_EXEC_MASK    1
@@ -50,16 +50,13 @@ static_assert(SPTE_TDP_AD_ENABLED_MASK == 0);
 #define ACC_ALL          (ACC_EXEC_MASK | ACC_WRITE_MASK | ACC_USER_MASK)
 
 /* The mask for the R/X bits in EPT PTEs */
-#define PT64_EPT_READABLE_MASK			0x1ull
-#define PT64_EPT_EXECUTABLE_MASK		0x4ull
+#define SPTE_EPT_READABLE_MASK			0x1ull
+#define SPTE_EPT_EXECUTABLE_MASK		0x4ull
 
-#define PT64_LEVEL_BITS 9
-
-#define PT64_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT64_LEVEL_BITS)
-
-#define PT64_INDEX(address, level) __PT_INDEX(address, level, PT64_LEVEL_BITS)
-
-#define SHADOW_PT_INDEX(addr, level) PT64_INDEX(addr, level)
+#define SPTE_LEVEL_BITS			9
+#define SPTE_LEVEL_SHIFT(level)		__PT_LEVEL_SHIFT(level, SPTE_LEVEL_BITS)
+#define SPTE_INDEX(address, level)	__PT_INDEX(address, level, SPTE_LEVEL_BITS)
+#define SPTE_ENT_PER_PAGE		__PT_ENT_PER_PAGE(SPTE_LEVEL_BITS)
 
 /*
  * The mask/shift to use for saving the original R/X bits when marking the PTE
@@ -68,8 +65,8 @@ static_assert(SPTE_TDP_AD_ENABLED_MASK == 0);
  * restored only when a write is attempted to the page.  This mask obviously
  * must not overlap the A/D type mask.
  */
-#define SHADOW_ACC_TRACK_SAVED_BITS_MASK (PT64_EPT_READABLE_MASK | \
-					  PT64_EPT_EXECUTABLE_MASK)
+#define SHADOW_ACC_TRACK_SAVED_BITS_MASK (SPTE_EPT_READABLE_MASK | \
+					  SPTE_EPT_EXECUTABLE_MASK)
 #define SHADOW_ACC_TRACK_SAVED_BITS_SHIFT 54
 #define SHADOW_ACC_TRACK_SAVED_MASK	(SHADOW_ACC_TRACK_SAVED_BITS_MASK << \
 					 SHADOW_ACC_TRACK_SAVED_BITS_SHIFT)
@@ -281,7 +278,7 @@ static inline bool is_executable_pte(u64 spte)
 
 static inline kvm_pfn_t spte_to_pfn(u64 pte)
 {
-	return (pte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
+	return (pte & SPTE_BASE_ADDR_MASK) >> PAGE_SHIFT;
 }
 
 static inline bool is_accessed_spte(u64 spte)
diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index ee4802d7b36c..9c65a64a56d9 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -11,7 +11,7 @@
 static void tdp_iter_refresh_sptep(struct tdp_iter *iter)
 {
 	iter->sptep = iter->pt_path[iter->level - 1] +
-		SHADOW_PT_INDEX(iter->gfn << PAGE_SHIFT, iter->level);
+		SPTE_INDEX(iter->gfn << PAGE_SHIFT, iter->level);
 	iter->old_spte = kvm_tdp_mmu_read_spte(iter->sptep);
 }
 
@@ -116,8 +116,8 @@ static bool try_step_side(struct tdp_iter *iter)
 	 * Check if the iterator is already at the end of the current page
 	 * table.
 	 */
-	if (SHADOW_PT_INDEX(iter->gfn << PAGE_SHIFT, iter->level) ==
-            (PT64_ENT_PER_PAGE - 1))
+	if (SPTE_INDEX(iter->gfn << PAGE_SHIFT, iter->level) ==
+	    (SPTE_ENT_PER_PAGE - 1))
 		return false;
 
 	iter->gfn += KVM_PAGES_PER_HPAGE(iter->level);
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 7b9265d67131..26cb9fed2f18 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -425,7 +425,7 @@ static void handle_removed_pt(struct kvm *kvm, tdp_ptep_t pt, bool shared)
 
 	tdp_mmu_unlink_sp(kvm, sp, shared);
 
-	for (i = 0; i < PT64_ENT_PER_PAGE; i++) {
+	for (i = 0; i < SPTE_ENT_PER_PAGE; i++) {
 		tdp_ptep_t sptep = pt + i;
 		gfn_t gfn = base_gfn + i * KVM_PAGES_PER_HPAGE(level);
 		u64 old_spte;
@@ -1487,7 +1487,7 @@ static int tdp_mmu_split_huge_page(struct kvm *kvm, struct tdp_iter *iter,
 	 * No need for atomics when writing to sp->spt since the page table has
 	 * not been linked in yet and thus is not reachable from any other CPU.
 	 */
-	for (i = 0; i < PT64_ENT_PER_PAGE; i++)
+	for (i = 0; i < SPTE_ENT_PER_PAGE; i++)
 		sp->spt[i] = make_huge_page_split_spte(huge_spte, level, i);
 
 	/*
@@ -1507,7 +1507,7 @@ static int tdp_mmu_split_huge_page(struct kvm *kvm, struct tdp_iter *iter,
 	 * are overwriting from the page stats. But we have to manually update
 	 * the page stats with the new present child pages.
 	 */
-	kvm_update_page_stats(kvm, level - 1, PT64_ENT_PER_PAGE);
+	kvm_update_page_stats(kvm, level - 1, SPTE_ENT_PER_PAGE);
 
 out:
 	trace_kvm_mmu_split_huge_page(iter->gfn, huge_spte, level, ret);
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 6/8] KVM: x86/mmu: Use common macros to compute 32/64-bit paging masks

[Sean Christopherson]

Dedup the code for generating (most of) the per-type PT_* masks in
paging_tmpl.h.  The relevant macros only vary based on the number of bits
per level, and that smidge of info is already provided in a common form
as PT_LEVEL_BITS.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/paging.h      | 23 -----------------------
 arch/x86/kvm/mmu/paging_tmpl.h | 25 +++++++++++--------------
 2 files changed, 11 insertions(+), 37 deletions(-)

diff --git a/arch/x86/kvm/mmu/paging.h b/arch/x86/kvm/mmu/paging.h
index 6a63727cc7e8..9de4976b2d46 100644
--- a/arch/x86/kvm/mmu/paging.h
+++ b/arch/x86/kvm/mmu/paging.h
@@ -5,28 +5,5 @@
 
 #define GUEST_PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
 
-#define PT64_LEVEL_BITS 9
-
-#define PT64_INDEX(address, level) __PT_INDEX(address, level, PT64_LEVEL_BITS)
-
-#define PT64_LVL_ADDR_MASK(level) \
-	__PT_LVL_ADDR_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
-
-#define PT64_LVL_OFFSET_MASK(level) \
-	__PT_LVL_OFFSET_MASK(GUEST_PT64_BASE_ADDR_MASK, level, PT64_LEVEL_BITS)
-
-
-#define PT32_LEVEL_SHIFT(level) __PT_LEVEL_SHIFT(level, PT32_LEVEL_BITS)
-
-#define PT32_LVL_OFFSET_MASK(level) \
-	__PT_LVL_OFFSET_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
-
-#define PT32_INDEX(address, level) __PT_INDEX(address, level, PT32_LEVEL_BITS)
-
-#define PT32_BASE_ADDR_MASK PAGE_MASK
-
-#define PT32_LVL_ADDR_MASK(level) \
-	__PT_LVL_ADDR_MASK(PT32_BASE_ADDR_MASK, level, PT32_LEVEL_BITS)
-
 #endif /* __KVM_X86_PAGING_H */
 
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index d68cc7a5ef81..4fcde3a18f5f 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -16,8 +16,9 @@
  */
 
 /*
- * We need the mmu code to access both 32-bit and 64-bit guest ptes,
- * so the code in this file is compiled twice, once per pte size.
+ * The MMU needs to be able to access/walk 32-bit and 64-bit guest page tables,
+ * as well as guest EPT tables, so the code in this file is compiled thrice,
+ * once per guest PTE type.  The per-type defines are #undef'd at the end.
  */
 
 #if PTTYPE == 64
@@ -25,10 +26,7 @@
 	#define guest_walker guest_walker64
 	#define FNAME(name) paging##64_##name
 	#define PT_BASE_ADDR_MASK GUEST_PT64_BASE_ADDR_MASK
-	#define PT_LVL_ADDR_MASK(lvl) PT64_LVL_ADDR_MASK(lvl)
-	#define PT_LVL_OFFSET_MASK(lvl) PT64_LVL_OFFSET_MASK(lvl)
-	#define PT_INDEX(addr, level) PT64_INDEX(addr, level)
-	#define PT_LEVEL_BITS PT64_LEVEL_BITS
+	#define PT_LEVEL_BITS 9
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
 	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) true
@@ -41,10 +39,7 @@
 	#define pt_element_t u32
 	#define guest_walker guest_walker32
 	#define FNAME(name) paging##32_##name
-	#define PT_BASE_ADDR_MASK PT32_BASE_ADDR_MASK
-	#define PT_LVL_ADDR_MASK(lvl) PT32_LVL_ADDR_MASK(lvl)
-	#define PT_LVL_OFFSET_MASK(lvl) PT32_LVL_OFFSET_MASK(lvl)
-	#define PT_INDEX(addr, level) PT32_INDEX(addr, level)
+	#define PT_BASE_ADDR_MASK PAGE_MASK
 	#define PT_LEVEL_BITS PT32_LEVEL_BITS
 	#define PT_MAX_FULL_LEVELS 2
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
@@ -60,10 +55,7 @@
 	#define guest_walker guest_walkerEPT
 	#define FNAME(name) ept_##name
 	#define PT_BASE_ADDR_MASK GUEST_PT64_BASE_ADDR_MASK
-	#define PT_LVL_ADDR_MASK(lvl) PT64_LVL_ADDR_MASK(lvl)
-	#define PT_LVL_OFFSET_MASK(lvl) PT64_LVL_OFFSET_MASK(lvl)
-	#define PT_INDEX(addr, level) PT64_INDEX(addr, level)
-	#define PT_LEVEL_BITS PT64_LEVEL_BITS
+	#define PT_LEVEL_BITS 9
 	#define PT_GUEST_DIRTY_SHIFT 9
 	#define PT_GUEST_ACCESSED_SHIFT 8
 	#define PT_HAVE_ACCESSED_DIRTY(mmu) (!(mmu)->cpu_role.base.ad_disabled)
@@ -72,6 +64,11 @@
 	#error Invalid PTTYPE value
 #endif
 
+/* Common logic, but per-type values.  These also need to be undefined. */
+#define PT_LVL_ADDR_MASK(lvl)	__PT_LVL_ADDR_MASK(PT_BASE_ADDR_MASK, lvl, PT_LEVEL_BITS)
+#define PT_LVL_OFFSET_MASK(lvl)	__PT_LVL_OFFSET_MASK(PT_BASE_ADDR_MASK, lvl, PT_LEVEL_BITS)
+#define PT_INDEX(addr, lvl)	__PT_INDEX(addr, lvl, PT_LEVEL_BITS)
+
 #define PT_GUEST_DIRTY_MASK    (1 << PT_GUEST_DIRTY_SHIFT)
 #define PT_GUEST_ACCESSED_MASK (1 << PT_GUEST_ACCESSED_SHIFT)
 
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 7/8] KVM: x86/mmu: Truncate paging32's PT_BASE_ADDR_MASK to 32 bits

[Sean Christopherson]

Truncate paging32's PT_BASE_ADDR_MASK to a pt_element_t, i.e. to 32 bits.
Ignoring PSE huge pages, the mask is only used in conjunction with gPTEs,
which are 32 bits, and so the address is limited to bits 31:12.

PSE huge pages encoded PA bits 39:32 in PTE bits 20:13, i.e. need custom
logic to handle their funky encoding regardless of PT_BASE_ADDR_MASK.

Note, PT_LVL_OFFSET_MASK is somewhat confusing in that it computes the
offset of the _gfn_, not of the gpa, i.e. not having bits 63:32 set in
PT_BASE_ADDR_MASK is again correct.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/paging_tmpl.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 4fcde3a18f5f..3ed7ba4730b4 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -39,7 +39,7 @@
 	#define pt_element_t u32
 	#define guest_walker guest_walker32
 	#define FNAME(name) paging##32_##name
-	#define PT_BASE_ADDR_MASK PAGE_MASK
+	#define PT_BASE_ADDR_MASK ((pt_element_t)PAGE_MASK)
 	#define PT_LEVEL_BITS PT32_LEVEL_BITS
 	#define PT_MAX_FULL_LEVELS 2
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
-- 
2.36.1.476.g0c4daa206d-goog

[PATCH v2 8/8] KVM: x86/mmu: Use common logic for computing the 32/64-bit base PA mask

[Sean Christopherson]

Use common logic for computing PT_BASE_ADDR_MASK for 32-bit, 64-bit, and
EPT paging.  Both PAGE_MASK and the new-common logic are supsersets of
what is actually needed for 32-bit paging.  PAGE_MASK sets bits 63:12 and
the former GUEST_PT64_BASE_ADDR_MASK sets bits 51:12, so regardless of
which value is used, the result will always be bits 31:12.

No functional change intended.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c         | 2 --
 arch/x86/kvm/mmu/paging.h      | 9 ---------
 arch/x86/kvm/mmu/paging_tmpl.h | 4 +---
 3 files changed, 1 insertion(+), 14 deletions(-)
 delete mode 100644 arch/x86/kvm/mmu/paging.h

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 81f2e58dc85b..a67aac727dc2 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -53,8 +53,6 @@
 #include <asm/kvm_page_track.h>
 #include "trace.h"
 
-#include "paging.h"
-
 extern bool itlb_multihit_kvm_mitigation;
 
 int __read_mostly nx_huge_pages = -1;
diff --git a/arch/x86/kvm/mmu/paging.h b/arch/x86/kvm/mmu/paging.h
deleted file mode 100644
index 9de4976b2d46..000000000000
--- a/arch/x86/kvm/mmu/paging.h
+++ /dev/null
@@ -1,9 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/* Shadow paging constants/helpers that don't need to be #undef'd. */
-#ifndef __KVM_X86_PAGING_H
-#define __KVM_X86_PAGING_H
-
-#define GUEST_PT64_BASE_ADDR_MASK (((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1))
-
-#endif /* __KVM_X86_PAGING_H */
-
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 3ed7ba4730b4..6c29aef4092b 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -25,7 +25,6 @@
 	#define pt_element_t u64
 	#define guest_walker guest_walker64
 	#define FNAME(name) paging##64_##name
-	#define PT_BASE_ADDR_MASK GUEST_PT64_BASE_ADDR_MASK
 	#define PT_LEVEL_BITS 9
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
 	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT
@@ -39,7 +38,6 @@
 	#define pt_element_t u32
 	#define guest_walker guest_walker32
 	#define FNAME(name) paging##32_##name
-	#define PT_BASE_ADDR_MASK ((pt_element_t)PAGE_MASK)
 	#define PT_LEVEL_BITS PT32_LEVEL_BITS
 	#define PT_MAX_FULL_LEVELS 2
 	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
@@ -54,7 +52,6 @@
 	#define pt_element_t u64
 	#define guest_walker guest_walkerEPT
 	#define FNAME(name) ept_##name
-	#define PT_BASE_ADDR_MASK GUEST_PT64_BASE_ADDR_MASK
 	#define PT_LEVEL_BITS 9
 	#define PT_GUEST_DIRTY_SHIFT 9
 	#define PT_GUEST_ACCESSED_SHIFT 8
@@ -65,6 +62,7 @@
 #endif
 
 /* Common logic, but per-type values.  These also need to be undefined. */
+#define PT_BASE_ADDR_MASK	((pt_element_t)(((1ULL << 52) - 1) & ~(u64)(PAGE_SIZE-1)))
 #define PT_LVL_ADDR_MASK(lvl)	__PT_LVL_ADDR_MASK(PT_BASE_ADDR_MASK, lvl, PT_LEVEL_BITS)
 #define PT_LVL_OFFSET_MASK(lvl)	__PT_LVL_OFFSET_MASK(PT_BASE_ADDR_MASK, lvl, PT_LEVEL_BITS)
 #define PT_INDEX(addr, lvl)	__PT_INDEX(addr, lvl, PT_LEVEL_BITS)
-- 
2.36.1.476.g0c4daa206d-goog

Re: [PATCH v2 5/8] KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs

[Paolo Bonzini]

On 6/15/22 01:33, Sean Christopherson wrote:
> @@ -2027,8 +2013,8 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>   	role.direct = direct;
>   	role.access = access;
>   	if (role.has_4_byte_gpte) {
> -		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
> -		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
> +		quadrant = gaddr >> (PAGE_SHIFT + (SPTE_LEVEL_BITS * level));
> +		quadrant &= (1 << ((PT32_LEVEL_BITS - SPTE_LEVEL_BITS) * level)) - 1;
>   		role.quadrant = quadrant;

That's just a fancy 1, though, and this is just

	/*
	 * Isolate the bits of the address that have already been used by the
	 * 8-byte shadow page table structures, but not yet in the 4-byte guest
	 * page tables.  For example, a 4-byte PDE consumes bits 31:22 and an
	 * 8-byte PDE consumes bits 29:21, so bits 31:30 go in the hash
	 * key.  The hash table look up up ensures that each sPTE points to
	 * the page for the correct portion of the guest page table structure.
	 */
	quadrant = gaddr >> (PAGE_SHIFT + (SPTE_LEVEL_BITS * level));
	quadrant &= (1 << level) - 1;

(Not the best comment, understood).

Paolo

Re: [PATCH v2 5/8] KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs

[Paolo Bonzini]

On 6/15/22 01:33, Sean Christopherson wrote:
> Separate the macros for KVM's shadow PTEs (SPTE) from guest 64-bit PTEs
> (PT64).  SPTE and PT64 are _mostly_ the same, but the few differences are
> quite critical, e.g. *_BASE_ADDR_MASK must differentiate between host and
> guest physical address spaces, and SPTE_PERM_MASK (was PT64_PERM_MASK) is
> very much specific to SPTEs.
> 
> Opportunistically (and temporarily) move most guest macros into paging.h
> to clearly associate them with shadow paging, and to ensure that they're
> not used as of this commit.  A future patch will eliminate them entirely.
> 
> Sadly, PT32_LEVEL_BITS is left behind in mmu_internal.h because it's
> needed for the quadrant calculation in kvm_mmu_get_page().  The quadrant
> calculation is hot enough (when using shadow paging with 32-bit guests)
> that adding a per-context helper is undesirable, and burying the
> computation in paging_tmpl.h with a forward declaration isn't exactly an
> improvement.
> 
> Signed-off-by: Sean Christopherson <seanjc@google.com>

A better try:

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 54b3e39d07b3..cd561b49cc84 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -2011,8 +2011,21 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
  	role.direct = direct;
  	role.access = access;
  	if (role.has_4_byte_gpte) {
+		/*
+		 * The "quadrant" value corresponds to those bits of the address
+		 * that have already been used by the 8-byte shadow page table
+		 * lookup, but not yet in the 4-byte guest page tables.  Having
+		 * the quadrant as part of the role ensures that each upper sPTE
+		 * points to the the correct portion of the guest page table
+		 * structure.
+		 *
+		 * For example, a 4-byte PDE consumes bits 31:22 and an 8-byte PDE
+		 * consumes bits 29:21.  Each guest PD must be expanded into four
+		 * shadow PDs, one for each value of bits 31:30, and the PDPEs
+		 * will use the four quadrants in round-robin fashion.
+		 */
  		quadrant = gaddr >> (PAGE_SHIFT + (SPTE_LEVEL_BITS * level));
-		quadrant &= (1 << ((PT32_LEVEL_BITS - SPTE_LEVEL_BITS) * level)) - 1;
+		quadrant &= (1 << level) - 1;
  		role.quadrant = quadrant;
  	}
  	if (level <= vcpu->arch.mmu->cpu_role.base.level)
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index cb9d4d358335..5e1e3c8f8aaa 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -20,9 +20,6 @@ extern bool dbg;
  #define MMU_WARN_ON(x) do { } while (0)
  #endif
  
-/* The number of bits for 32-bit PTEs is to needed compute the quandrant. */
-#define PT32_LEVEL_BITS 10
-
  /* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
  #define __PT_LEVEL_SHIFT(level, bits_per_level)	\
  	(PAGE_SHIFT + ((level) - 1) * (bits_per_level))
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 6c29aef4092b..e4655056e651 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -38,7 +38,7 @@
  	#define pt_element_t u32
  	#define guest_walker guest_walker32
  	#define FNAME(name) paging##32_##name
-	#define PT_LEVEL_BITS PT32_LEVEL_BITS
+	#define PT_LEVEL_BITS 10
  	#define PT_MAX_FULL_LEVELS 2
  	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
  	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT

Paolo

Re: [PATCH v2 5/8] KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs

[Sean Christopherson]

On Wed, Jun 15, 2022, Paolo Bonzini wrote:
> On 6/15/22 01:33, Sean Christopherson wrote:
> > Separate the macros for KVM's shadow PTEs (SPTE) from guest 64-bit PTEs
> > (PT64).  SPTE and PT64 are _mostly_ the same, but the few differences are
> > quite critical, e.g. *_BASE_ADDR_MASK must differentiate between host and
> > guest physical address spaces, and SPTE_PERM_MASK (was PT64_PERM_MASK) is
> > very much specific to SPTEs.
> > 
> > Opportunistically (and temporarily) move most guest macros into paging.h
> > to clearly associate them with shadow paging, and to ensure that they're
> > not used as of this commit.  A future patch will eliminate them entirely.
> > 
> > Sadly, PT32_LEVEL_BITS is left behind in mmu_internal.h because it's
> > needed for the quadrant calculation in kvm_mmu_get_page().  The quadrant
> > calculation is hot enough (when using shadow paging with 32-bit guests)
> > that adding a per-context helper is undesirable, and burying the
> > computation in paging_tmpl.h with a forward declaration isn't exactly an
> > improvement.
> > 
> > Signed-off-by: Sean Christopherson <seanjc@google.com>
> 
> A better try:

Very nice!  It's obvious once someone else writes the code.  :-)

> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 54b3e39d07b3..cd561b49cc84 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -2011,8 +2011,21 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>  	role.direct = direct;
>  	role.access = access;
>  	if (role.has_4_byte_gpte) {
> +		/*
> +		 * The "quadrant" value corresponds to those bits of the address
> +		 * that have already been used by the 8-byte shadow page table
> +		 * lookup, but not yet in the 4-byte guest page tables.  Having
> +		 * the quadrant as part of the role ensures that each upper sPTE
> +		 * points to the the correct portion of the guest page table
> +		 * structure.
> +		 *
> +		 * For example, a 4-byte PDE consumes bits 31:22 and an 8-byte PDE
> +		 * consumes bits 29:21.  Each guest PD must be expanded into four
> +		 * shadow PDs, one for each value of bits 31:30, and the PDPEs
> +		 * will use the four quadrants in round-robin fashion.

It's not round-robin, that would imply KVM rotates through each quadrant on its
own.  FWIW, I like David's comment from his patch that simplifies this mess in a
similar way.

https://lore.kernel.org/all/20220516232138.1783324-5-dmatlack@google.com
> +		 */
>  		quadrant = gaddr >> (PAGE_SHIFT + (SPTE_LEVEL_BITS * level));
> -		quadrant &= (1 << ((PT32_LEVEL_BITS - SPTE_LEVEL_BITS) * level)) - 1;
> +		quadrant &= (1 << level) - 1;
>  		role.quadrant = quadrant;
>  	}
>  	if (level <= vcpu->arch.mmu->cpu_role.base.level)
> diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
> index cb9d4d358335..5e1e3c8f8aaa 100644
> --- a/arch/x86/kvm/mmu/mmu_internal.h
> +++ b/arch/x86/kvm/mmu/mmu_internal.h
> @@ -20,9 +20,6 @@ extern bool dbg;
>  #define MMU_WARN_ON(x) do { } while (0)
>  #endif
> -/* The number of bits for 32-bit PTEs is to needed compute the quandrant. */

Heh, and it gets rid of my typo.

> -#define PT32_LEVEL_BITS 10
> -
>  /* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
>  #define __PT_LEVEL_SHIFT(level, bits_per_level)	\
>  	(PAGE_SHIFT + ((level) - 1) * (bits_per_level))
> diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
> index 6c29aef4092b..e4655056e651 100644
> --- a/arch/x86/kvm/mmu/paging_tmpl.h
> +++ b/arch/x86/kvm/mmu/paging_tmpl.h
> @@ -38,7 +38,7 @@
>  	#define pt_element_t u32
>  	#define guest_walker guest_walker32
>  	#define FNAME(name) paging##32_##name
> -	#define PT_LEVEL_BITS PT32_LEVEL_BITS
> +	#define PT_LEVEL_BITS 10
>  	#define PT_MAX_FULL_LEVELS 2
>  	#define PT_GUEST_DIRTY_SHIFT PT_DIRTY_SHIFT
>  	#define PT_GUEST_ACCESSED_SHIFT PT_ACCESSED_SHIFT
> 
> Paolo
> 

Re: [PATCH v2 5/8] KVM: x86/mmu: Use separate namespaces for guest PTEs and shadow PTEs

[Paolo Bonzini]

On 6/15/22 16:26, Sean Christopherson wrote:
>> +		 * For example, a 4-byte PDE consumes bits 31:22 and an 8-byte PDE
>> +		 * consumes bits 29:21.  Each guest PD must be expanded into four
>> +		 * shadow PDs, one for each value of bits 31:30, and the PDPEs
>> +		 * will use the four quadrants in round-robin fashion.
> It's not round-robin, that would imply KVM rotates through each quadrant on its
> own.  FWIW, I like David's comment from his patch that simplifies this mess in a
> similar way.
> 
> https://lore.kernel.org/all/20220516232138.1783324-5-dmatlack@google.com

Yeah, by round-robin I meant that the 512 entries will look like q=0 q=1 
q=2 q=3 q=0 q=1 etc.

I'll incorporate David's comment, minus the last paragraph.

Paolo