[PATCH v2 0/9] KVM: x86/mmu: Clean up MMU_DEBUG and BUG/WARN usage

[PATCH v2 0/9] KVM: x86/mmu: Clean up MMU_DEBUG and BUG/WARN usage

[Sean Christopherson]

This series consist of three loosely related miniseries:

 1. Remove the noisy prints buried behind MMU_DEBUG, and replace MMU_DEBUG
    with a KVM_PROVE_MMU Kconfig.

 2. Use WARN_ON_ONCE() for all runtime WARNs, i.e. avoid spamming the
    kernel log if something goes awry in the MMU.

 3. Demote BUG() usage in the shadow MMU to KVM_BUG_ON() when the kernel
    is built with CONFIG_BUG_ON_DATA_CORRUPTION=n.


v2:
 - Collect a review. [Mingwei]
 - Call out the more obvious path to FNAME(walk_addr_generic) in patch 6. [David]
 - Use BUILD_BUG_ON_INVALID() for the stub. [David]

v1: https://lore.kernel.org/all/20230511235917.639770-1-seanjc@google.com

Mingwei Zhang (1):
  KVM: x86/mmu: Plumb "struct kvm" all the way to pte_list_remove()

Sean Christopherson (8):
  KVM: x86/mmu: Delete rmap_printk() and all its usage
  KVM: x86/mmu: Delete the "dbg" module param
  KVM: x86/mmu: Rename MMU_WARN_ON() to KVM_MMU_WARN_ON()
  KVM: x86/mmu: Convert "runtime" WARN_ON() assertions to WARN_ON_ONCE()
  KVM: x86/mmu: Bug the VM if a vCPU ends up in long mode without PAE
    enabled
  KVM: x86/mmu: Replace MMU_DEBUG with proper KVM_PROVE_MMU Kconfig
  KVM: x86/mmu: Use BUILD_BUG_ON_INVALID() for KVM_MMU_WARN_ON() stub
  KVM: x86/mmu: BUG() in rmap helpers iff
    CONFIG_BUG_ON_DATA_CORRUPTION=y

 arch/x86/kvm/Kconfig            |  13 ++++
 arch/x86/kvm/mmu/mmu.c          | 127 ++++++++++++++------------------
 arch/x86/kvm/mmu/mmu_internal.h |  12 +--
 arch/x86/kvm/mmu/page_track.c   |  16 ++--
 arch/x86/kvm/mmu/paging_tmpl.h  |  16 +++-
 arch/x86/kvm/mmu/spte.c         |   4 +-
 arch/x86/kvm/mmu/spte.h         |   8 +-
 arch/x86/kvm/mmu/tdp_iter.c     |   4 +-
 arch/x86/kvm/mmu/tdp_mmu.c      |  28 +++----
 include/linux/kvm_host.h        |  19 +++++
 10 files changed, 134 insertions(+), 113 deletions(-)


base-commit: 294ebdda1560d3eb515da88b805ba7f88eb28d21
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 1/9] KVM: x86/mmu: Delete rmap_printk() and all its usage

[Sean Christopherson]

Delete rmap_printk() so that MMU_WARN_ON() and MMU_DEBUG can be morphed
into something that can be regularly enabled for debug kernels.  The
information provided by rmap_printk() isn't all that useful now that the
rmap and unsync code is mature, as the prints are simultaneously too
verbose (_lots_ of message) and yet not verbose enough to be helpful for
debug (most instances print just the SPTE pointer/value, which is rarely
sufficient to root cause anything but trivial bugs).

Alternatively, rmap_printk() could be reworked to into tracepoints, but
it's not clear there is a real need as rmap bugs rarely escape initial
development, and when bugs do escape to production, they are often edge
cases and/or reside in code that isn't directly related to the rmaps.
In other words, the problems with rmap_printk() being unhelpful also apply
to tracepoints.  And deleting rmap_printk() doesn't preclude adding
tracepoints in the future.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c          | 12 ------------
 arch/x86/kvm/mmu/mmu_internal.h |  2 --
 2 files changed, 14 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index bc24d430db6e..8e36e07719bf 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -941,10 +941,8 @@ static int pte_list_add(struct kvm_mmu_memory_cache *cache, u64 *spte,
 	int count = 0;
 
 	if (!rmap_head->val) {
-		rmap_printk("%p %llx 0->1\n", spte, *spte);
 		rmap_head->val = (unsigned long)spte;
 	} else if (!(rmap_head->val & 1)) {
-		rmap_printk("%p %llx 1->many\n", spte, *spte);
 		desc = kvm_mmu_memory_cache_alloc(cache);
 		desc->sptes[0] = (u64 *)rmap_head->val;
 		desc->sptes[1] = spte;
@@ -953,7 +951,6 @@ static int pte_list_add(struct kvm_mmu_memory_cache *cache, u64 *spte,
 		rmap_head->val = (unsigned long)desc | 1;
 		++count;
 	} else {
-		rmap_printk("%p %llx many->many\n", spte, *spte);
 		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
 		count = desc->tail_count + desc->spte_count;
 
@@ -1018,14 +1015,12 @@ static void pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
 		pr_err("%s: %p 0->BUG\n", __func__, spte);
 		BUG();
 	} else if (!(rmap_head->val & 1)) {
-		rmap_printk("%p 1->0\n", spte);
 		if ((u64 *)rmap_head->val != spte) {
 			pr_err("%s:  %p 1->BUG\n", __func__, spte);
 			BUG();
 		}
 		rmap_head->val = 0;
 	} else {
-		rmap_printk("%p many->many\n", spte);
 		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
 		while (desc) {
 			for (i = 0; i < desc->spte_count; ++i) {
@@ -1241,8 +1236,6 @@ static bool spte_write_protect(u64 *sptep, bool pt_protect)
 	    !(pt_protect && is_mmu_writable_spte(spte)))
 		return false;
 
-	rmap_printk("spte %p %llx\n", sptep, *sptep);
-
 	if (pt_protect)
 		spte &= ~shadow_mmu_writable_mask;
 	spte = spte & ~PT_WRITABLE_MASK;
@@ -1267,8 +1260,6 @@ static bool spte_clear_dirty(u64 *sptep)
 {
 	u64 spte = *sptep;
 
-	rmap_printk("spte %p %llx\n", sptep, *sptep);
-
 	MMU_WARN_ON(!spte_ad_enabled(spte));
 	spte &= ~shadow_dirty_mask;
 	return mmu_spte_update(sptep, spte);
@@ -1480,9 +1471,6 @@ static bool kvm_set_pte_rmap(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
 
 restart:
 	for_each_rmap_spte(rmap_head, &iter, sptep) {
-		rmap_printk("spte %p %llx gfn %llx (%d)\n",
-			    sptep, *sptep, gfn, level);
-
 		need_flush = true;
 
 		if (pte_write(pte)) {
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index 4f1e4b327f40..9c9dd9340c63 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -11,10 +11,8 @@
 #ifdef MMU_DEBUG
 extern bool dbg;
 
-#define rmap_printk(fmt, args...) do { if (dbg) printk("%s: " fmt, __func__, ## args); } while (0)
 #define MMU_WARN_ON(x) WARN_ON(x)
 #else
-#define rmap_printk(x...) do { } while (0)
 #define MMU_WARN_ON(x) do { } while (0)
 #endif
 
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 2/9] KVM: x86/mmu: Delete the "dbg" module param

[Sean Christopherson]

Delete KVM's "dbg" module param now that its usage in KVM is gone (it
used to guard pgprintk() and rmap_printk()).

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c          | 5 -----
 arch/x86/kvm/mmu/mmu_internal.h | 2 --
 2 files changed, 7 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 8e36e07719bf..b16092d71d3f 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -115,11 +115,6 @@ static int max_huge_page_level __read_mostly;
 static int tdp_root_level __read_mostly;
 static int max_tdp_level __read_mostly;
 
-#ifdef MMU_DEBUG
-bool dbg = 0;
-module_param(dbg, bool, 0644);
-#endif
-
 #define PTE_PREFETCH_NUM		8
 
 #include <trace/events/kvm.h>
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index 9c9dd9340c63..9ea80e4d463c 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -9,8 +9,6 @@
 #undef MMU_DEBUG
 
 #ifdef MMU_DEBUG
-extern bool dbg;
-
 #define MMU_WARN_ON(x) WARN_ON(x)
 #else
 #define MMU_WARN_ON(x) do { } while (0)
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 3/9] KVM: x86/mmu: Rename MMU_WARN_ON() to KVM_MMU_WARN_ON()

[Sean Christopherson]

Rename MMU_WARN_ON() to make it super obvious that the assertions are
all about KVM's MMU, not the primary MMU.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c          | 4 ++--
 arch/x86/kvm/mmu/mmu_internal.h | 4 ++--
 arch/x86/kvm/mmu/spte.h         | 8 ++++----
 arch/x86/kvm/mmu/tdp_mmu.c      | 8 ++++----
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index b16092d71d3f..c87539dd1ac0 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -1255,7 +1255,7 @@ static bool spte_clear_dirty(u64 *sptep)
 {
 	u64 spte = *sptep;
 
-	MMU_WARN_ON(!spte_ad_enabled(spte));
+	KVM_MMU_WARN_ON(!spte_ad_enabled(spte));
 	spte &= ~shadow_dirty_mask;
 	return mmu_spte_update(sptep, spte);
 }
@@ -1735,7 +1735,7 @@ static void kvm_unaccount_mmu_page(struct kvm *kvm, struct kvm_mmu_page *sp)
 
 static void kvm_mmu_free_shadow_page(struct kvm_mmu_page *sp)
 {
-	MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
+	KVM_MMU_WARN_ON(!is_empty_shadow_page(sp->spt));
 	hlist_del(&sp->hash_link);
 	list_del(&sp->link);
 	free_page((unsigned long)sp->spt);
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index 9ea80e4d463c..bb1649669bc9 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -9,9 +9,9 @@
 #undef MMU_DEBUG
 
 #ifdef MMU_DEBUG
-#define MMU_WARN_ON(x) WARN_ON(x)
+#define KVM_MMU_WARN_ON(x) WARN_ON(x)
 #else
-#define MMU_WARN_ON(x) do { } while (0)
+#define KVM_MMU_WARN_ON(x) do { } while (0)
 #endif
 
 /* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
index 1279db2eab44..83e6614f3720 100644
--- a/arch/x86/kvm/mmu/spte.h
+++ b/arch/x86/kvm/mmu/spte.h
@@ -265,13 +265,13 @@ static inline bool sp_ad_disabled(struct kvm_mmu_page *sp)
 
 static inline bool spte_ad_enabled(u64 spte)
 {
-	MMU_WARN_ON(!is_shadow_present_pte(spte));
+	KVM_MMU_WARN_ON(!is_shadow_present_pte(spte));
 	return (spte & SPTE_TDP_AD_MASK) != SPTE_TDP_AD_DISABLED;
 }
 
 static inline bool spte_ad_need_write_protect(u64 spte)
 {
-	MMU_WARN_ON(!is_shadow_present_pte(spte));
+	KVM_MMU_WARN_ON(!is_shadow_present_pte(spte));
 	/*
 	 * This is benign for non-TDP SPTEs as SPTE_TDP_AD_ENABLED is '0',
 	 * and non-TDP SPTEs will never set these bits.  Optimize for 64-bit
@@ -282,13 +282,13 @@ static inline bool spte_ad_need_write_protect(u64 spte)
 
 static inline u64 spte_shadow_accessed_mask(u64 spte)
 {
-	MMU_WARN_ON(!is_shadow_present_pte(spte));
+	KVM_MMU_WARN_ON(!is_shadow_present_pte(spte));
 	return spte_ad_enabled(spte) ? shadow_accessed_mask : 0;
 }
 
 static inline u64 spte_shadow_dirty_mask(u64 spte)
 {
-	MMU_WARN_ON(!is_shadow_present_pte(spte));
+	KVM_MMU_WARN_ON(!is_shadow_present_pte(spte));
 	return spte_ad_enabled(spte) ? shadow_dirty_mask : 0;
 }
 
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 512163d52194..f881de40f9ef 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -1548,8 +1548,8 @@ static bool clear_dirty_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
 		if (!is_shadow_present_pte(iter.old_spte))
 			continue;
 
-		MMU_WARN_ON(kvm_ad_enabled() &&
-			    spte_ad_need_write_protect(iter.old_spte));
+		KVM_MMU_WARN_ON(kvm_ad_enabled() &&
+				spte_ad_need_write_protect(iter.old_spte));
 
 		if (!(iter.old_spte & dbit))
 			continue;
@@ -1607,8 +1607,8 @@ static void clear_dirty_pt_masked(struct kvm *kvm, struct kvm_mmu_page *root,
 		if (!mask)
 			break;
 
-		MMU_WARN_ON(kvm_ad_enabled() &&
-			    spte_ad_need_write_protect(iter.old_spte));
+		KVM_MMU_WARN_ON(kvm_ad_enabled() &&
+				spte_ad_need_write_protect(iter.old_spte));
 
 		if (iter.level > PG_LEVEL_4K ||
 		    !(mask & (1UL << (iter.gfn - gfn))))
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 4/9] KVM: x86/mmu: Convert "runtime" WARN_ON() assertions to WARN_ON_ONCE()

[Sean Christopherson]

Convert all "runtime" assertions, i.e. assertions that can be triggered
while running vCPUs, from WARN_ON() to WARN_ON_ONCE().  Every WARN in the
MMU that is tied to running vCPUs, i.e. not contained to loading and
initializing KVM, is likely to fire _a lot_ when it does trigger.  E.g. if
KVM ends up with a bug that causes a root to be invalidated before the
page fault handler is invoked, pretty much _every_ page fault VM-Exit
triggers the WARN.

If a WARN is triggered frequently, the resulting spam usually causes a lot
of damage of its own, e.g. consumes resources to log the WARN and pollutes
the kernel log, often to the point where other useful information can be
lost.  In many case, the damage caused by the spam is actually worse than
the bug itself, e.g. KVM can almost always recover from an unexpectedly
invalid root.

On the flip side, warning every time is rarely helpful for debug and
triage, i.e. a single splat is usually sufficient to point a debugger in
the right direction, and automated testing, e.g. syzkaller, typically runs
with warn_on_panic=1, i.e. will never get past the first WARN anyways.

Lastly, when an assertions fails multiple times, the stack traces in KVM
are almost always identical, i.e. the full splat only needs to be captured
once.  And _if_ there is value in captruing information about the failed
assert, a ratelimited printk() is sufficient and less likely to rack up a
large amount of collateral damage.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c          | 48 ++++++++++++++++-----------------
 arch/x86/kvm/mmu/mmu_internal.h |  2 +-
 arch/x86/kvm/mmu/page_track.c   | 16 +++++------
 arch/x86/kvm/mmu/paging_tmpl.h  |  4 +--
 arch/x86/kvm/mmu/spte.c         |  4 +--
 arch/x86/kvm/mmu/tdp_iter.c     |  4 +--
 arch/x86/kvm/mmu/tdp_mmu.c      | 20 +++++++-------
 7 files changed, 49 insertions(+), 49 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index c87539dd1ac0..eb6af9c4cf14 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -485,7 +485,7 @@ static u64 __get_spte_lockless(u64 *sptep)
  */
 static void mmu_spte_set(u64 *sptep, u64 new_spte)
 {
-	WARN_ON(is_shadow_present_pte(*sptep));
+	WARN_ON_ONCE(is_shadow_present_pte(*sptep));
 	__set_spte(sptep, new_spte);
 }
 
@@ -497,7 +497,7 @@ static u64 mmu_spte_update_no_track(u64 *sptep, u64 new_spte)
 {
 	u64 old_spte = *sptep;
 
-	WARN_ON(!is_shadow_present_pte(new_spte));
+	WARN_ON_ONCE(!is_shadow_present_pte(new_spte));
 	check_spte_writable_invariants(new_spte);
 
 	if (!is_shadow_present_pte(old_spte)) {
@@ -510,7 +510,7 @@ static u64 mmu_spte_update_no_track(u64 *sptep, u64 new_spte)
 	else
 		old_spte = __update_clear_spte_slow(sptep, new_spte);
 
-	WARN_ON(spte_to_pfn(old_spte) != spte_to_pfn(new_spte));
+	WARN_ON_ONCE(spte_to_pfn(old_spte) != spte_to_pfn(new_spte));
 
 	return old_spte;
 }
@@ -592,7 +592,7 @@ static u64 mmu_spte_clear_track_bits(struct kvm *kvm, u64 *sptep)
 	 * by a refcounted page, the refcount is elevated.
 	 */
 	page = kvm_pfn_to_refcounted_page(pfn);
-	WARN_ON(page && !page_count(page));
+	WARN_ON_ONCE(page && !page_count(page));
 
 	if (is_accessed_spte(old_spte))
 		kvm_set_pfn_accessed(pfn);
@@ -807,7 +807,7 @@ static void update_gfn_disallow_lpage_count(const struct kvm_memory_slot *slot,
 	for (i = PG_LEVEL_2M; i <= KVM_MAX_HUGEPAGE_LEVEL; ++i) {
 		linfo = lpage_info_slot(gfn, slot, i);
 		linfo->disallow_lpage += count;
-		WARN_ON(linfo->disallow_lpage < 0);
+		WARN_ON_ONCE(linfo->disallow_lpage < 0);
 	}
 }
 
@@ -1202,7 +1202,7 @@ static void drop_large_spte(struct kvm *kvm, u64 *sptep, bool flush)
 	struct kvm_mmu_page *sp;
 
 	sp = sptep_to_sp(sptep);
-	WARN_ON(sp->role.level == PG_LEVEL_4K);
+	WARN_ON_ONCE(sp->role.level == PG_LEVEL_4K);
 
 	drop_spte(kvm, sptep);
 
@@ -1461,7 +1461,7 @@ static bool kvm_set_pte_rmap(struct kvm *kvm, struct kvm_rmap_head *rmap_head,
 	u64 new_spte;
 	kvm_pfn_t new_pfn;
 
-	WARN_ON(pte_huge(pte));
+	WARN_ON_ONCE(pte_huge(pte));
 	new_pfn = pte_pfn(pte);
 
 restart:
@@ -1823,7 +1823,7 @@ static int mmu_pages_add(struct kvm_mmu_pages *pvec, struct kvm_mmu_page *sp,
 static inline void clear_unsync_child_bit(struct kvm_mmu_page *sp, int idx)
 {
 	--sp->unsync_children;
-	WARN_ON((int)sp->unsync_children < 0);
+	WARN_ON_ONCE((int)sp->unsync_children < 0);
 	__clear_bit(idx, sp->unsync_child_bitmap);
 }
 
@@ -1881,7 +1881,7 @@ static int mmu_unsync_walk(struct kvm_mmu_page *sp,
 
 static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp)
 {
-	WARN_ON(!sp->unsync);
+	WARN_ON_ONCE(!sp->unsync);
 	trace_kvm_mmu_sync_page(sp);
 	sp->unsync = 0;
 	--kvm->stat.mmu_unsync;
@@ -2056,11 +2056,11 @@ static int mmu_pages_first(struct kvm_mmu_pages *pvec,
 	if (pvec->nr == 0)
 		return 0;
 
-	WARN_ON(pvec->page[0].idx != INVALID_INDEX);
+	WARN_ON_ONCE(pvec->page[0].idx != INVALID_INDEX);
 
 	sp = pvec->page[0].sp;
 	level = sp->role.level;
-	WARN_ON(level == PG_LEVEL_4K);
+	WARN_ON_ONCE(level == PG_LEVEL_4K);
 
 	parents->parent[level-2] = sp;
 
@@ -2082,7 +2082,7 @@ static void mmu_pages_clear_parents(struct mmu_page_path *parents)
 		if (!sp)
 			return;
 
-		WARN_ON(idx == INVALID_INDEX);
+		WARN_ON_ONCE(idx == INVALID_INDEX);
 		clear_unsync_child_bit(sp, idx);
 		level++;
 	} while (!sp->unsync_children);
@@ -2203,7 +2203,7 @@ static struct kvm_mmu_page *kvm_mmu_find_shadow_page(struct kvm *kvm,
 			if (ret < 0)
 				break;
 
-			WARN_ON(!list_empty(&invalid_list));
+			WARN_ON_ONCE(!list_empty(&invalid_list));
 			if (ret > 0)
 				kvm_flush_remote_tlbs(kvm);
 		}
@@ -2658,7 +2658,7 @@ static void kvm_mmu_commit_zap_page(struct kvm *kvm,
 	kvm_flush_remote_tlbs(kvm);
 
 	list_for_each_entry_safe(sp, nsp, invalid_list, link) {
-		WARN_ON(!sp->role.invalid || sp->root_count);
+		WARN_ON_ONCE(!sp->role.invalid || sp->root_count);
 		kvm_mmu_free_shadow_page(sp);
 	}
 }
@@ -2853,7 +2853,7 @@ int mmu_try_to_unsync_pages(struct kvm *kvm, const struct kvm_memory_slot *slot,
 				continue;
 		}
 
-		WARN_ON(sp->role.level != PG_LEVEL_4K);
+		WARN_ON_ONCE(sp->role.level != PG_LEVEL_4K);
 		kvm_unsync_page(kvm, sp);
 	}
 	if (locked)
@@ -3006,7 +3006,7 @@ static void __direct_pte_prefetch(struct kvm_vcpu *vcpu,
 	u64 *spte, *start = NULL;
 	int i;
 
-	WARN_ON(!sp->role.direct);
+	WARN_ON_ONCE(!sp->role.direct);
 
 	i = spte_index(sptep) & ~(PTE_PREFETCH_NUM - 1);
 	spte = sp->spt + i;
@@ -3552,7 +3552,7 @@ static void mmu_free_root_page(struct kvm *kvm, hpa_t *root_hpa,
 	 * SPTE to ensure any non-PA bits are dropped.
 	 */
 	sp = spte_to_child_sp(*root_hpa);
-	if (WARN_ON(!sp))
+	if (WARN_ON_ONCE(!sp))
 		return;
 
 	if (is_tdp_mmu_page(sp))
@@ -4167,7 +4167,7 @@ static int handle_mmio_page_fault(struct kvm_vcpu *vcpu, u64 addr, bool direct)
 		return RET_PF_EMULATE;
 
 	reserved = get_mmio_spte(vcpu, addr, &spte);
-	if (WARN_ON(reserved))
+	if (WARN_ON_ONCE(reserved))
 		return -EINVAL;
 
 	if (is_mmio_spte(spte)) {
@@ -5502,9 +5502,9 @@ void kvm_mmu_unload(struct kvm_vcpu *vcpu)
 	struct kvm *kvm = vcpu->kvm;
 
 	kvm_mmu_free_roots(kvm, &vcpu->arch.root_mmu, KVM_MMU_ROOTS_ALL);
-	WARN_ON(VALID_PAGE(vcpu->arch.root_mmu.root.hpa));
+	WARN_ON_ONCE(VALID_PAGE(vcpu->arch.root_mmu.root.hpa));
 	kvm_mmu_free_roots(kvm, &vcpu->arch.guest_mmu, KVM_MMU_ROOTS_ALL);
-	WARN_ON(VALID_PAGE(vcpu->arch.guest_mmu.root.hpa));
+	WARN_ON_ONCE(VALID_PAGE(vcpu->arch.guest_mmu.root.hpa));
 	vcpu_clear_mmio_info(vcpu, MMIO_GVA_ANY);
 }
 
@@ -5708,7 +5708,7 @@ int noinline kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 err
 	int r, emulation_type = EMULTYPE_PF;
 	bool direct = vcpu->arch.mmu->root_role.direct;
 
-	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root.hpa)))
+	if (WARN_ON_ONCE(!VALID_PAGE(vcpu->arch.mmu->root.hpa)))
 		return RET_PF_RETRY;
 
 	r = RET_PF_INVALID;
@@ -6065,7 +6065,7 @@ static void kvm_zap_obsolete_pages(struct kvm *kvm)
 		 * pages.  Skip the bogus page, otherwise we'll get stuck in an
 		 * infinite loop if the page gets put back on the list (again).
 		 */
-		if (WARN_ON(sp->role.invalid))
+		if (WARN_ON_ONCE(sp->role.invalid))
 			continue;
 
 		/*
@@ -6707,7 +6707,7 @@ void kvm_mmu_zap_all(struct kvm *kvm)
 	write_lock(&kvm->mmu_lock);
 restart:
 	list_for_each_entry_safe(sp, node, &kvm->arch.active_mmu_pages, link) {
-		if (WARN_ON(sp->role.invalid))
+		if (WARN_ON_ONCE(sp->role.invalid))
 			continue;
 		if (__kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list, &ign))
 			goto restart;
@@ -6725,7 +6725,7 @@ void kvm_mmu_zap_all(struct kvm *kvm)
 
 void kvm_mmu_invalidate_mmio_sptes(struct kvm *kvm, u64 gen)
 {
-	WARN_ON(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
+	WARN_ON_ONCE(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS);
 
 	gen &= MMIO_SPTE_GEN_MASK;
 
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index bb1649669bc9..cfe925fefa68 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -9,7 +9,7 @@
 #undef MMU_DEBUG
 
 #ifdef MMU_DEBUG
-#define KVM_MMU_WARN_ON(x) WARN_ON(x)
+#define KVM_MMU_WARN_ON(x) WARN_ON_ONCE(x)
 #else
 #define KVM_MMU_WARN_ON(x) do { } while (0)
 #endif
diff --git a/arch/x86/kvm/mmu/page_track.c b/arch/x86/kvm/mmu/page_track.c
index 0a2ac438d647..fd16918b3a7a 100644
--- a/arch/x86/kvm/mmu/page_track.c
+++ b/arch/x86/kvm/mmu/page_track.c
@@ -94,7 +94,7 @@ static void update_gfn_track(struct kvm_memory_slot *slot, gfn_t gfn,
 
 	val = slot->arch.gfn_track[mode][index];
 
-	if (WARN_ON(val + count < 0 || val + count > USHRT_MAX))
+	if (WARN_ON_ONCE(val + count < 0 || val + count > USHRT_MAX))
 		return;
 
 	slot->arch.gfn_track[mode][index] += count;
@@ -117,11 +117,11 @@ void kvm_slot_page_track_add_page(struct kvm *kvm,
 				  enum kvm_page_track_mode mode)
 {
 
-	if (WARN_ON(!page_track_mode_is_valid(mode)))
+	if (WARN_ON_ONCE(!page_track_mode_is_valid(mode)))
 		return;
 
-	if (WARN_ON(mode == KVM_PAGE_TRACK_WRITE &&
-		    !kvm_page_track_write_tracking_enabled(kvm)))
+	if (WARN_ON_ONCE(mode == KVM_PAGE_TRACK_WRITE &&
+			 !kvm_page_track_write_tracking_enabled(kvm)))
 		return;
 
 	update_gfn_track(slot, gfn, mode, 1);
@@ -155,11 +155,11 @@ void kvm_slot_page_track_remove_page(struct kvm *kvm,
 				     struct kvm_memory_slot *slot, gfn_t gfn,
 				     enum kvm_page_track_mode mode)
 {
-	if (WARN_ON(!page_track_mode_is_valid(mode)))
+	if (WARN_ON_ONCE(!page_track_mode_is_valid(mode)))
 		return;
 
-	if (WARN_ON(mode == KVM_PAGE_TRACK_WRITE &&
-		    !kvm_page_track_write_tracking_enabled(kvm)))
+	if (WARN_ON_ONCE(mode == KVM_PAGE_TRACK_WRITE &&
+			 !kvm_page_track_write_tracking_enabled(kvm)))
 		return;
 
 	update_gfn_track(slot, gfn, mode, -1);
@@ -181,7 +181,7 @@ bool kvm_slot_page_track_is_active(struct kvm *kvm,
 {
 	int index;
 
-	if (WARN_ON(!page_track_mode_is_valid(mode)))
+	if (WARN_ON_ONCE(!page_track_mode_is_valid(mode)))
 		return false;
 
 	if (!slot)
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 7a97f769a7cb..a3fc7c1a7f8d 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -633,7 +633,7 @@ static int FNAME(fetch)(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault,
 	if (FNAME(gpte_changed)(vcpu, gw, top_level))
 		goto out_gpte_changed;
 
-	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root.hpa)))
+	if (WARN_ON_ONCE(!VALID_PAGE(vcpu->arch.mmu->root.hpa)))
 		goto out_gpte_changed;
 
 	for_each_shadow_entry(vcpu, fault->addr, it) {
@@ -830,7 +830,7 @@ static gpa_t FNAME(get_level1_sp_gpa)(struct kvm_mmu_page *sp)
 {
 	int offset = 0;
 
-	WARN_ON(sp->role.level != PG_LEVEL_4K);
+	WARN_ON_ONCE(sp->role.level != PG_LEVEL_4K);
 
 	if (PTTYPE == 32)
 		offset = sp->role.quadrant << SPTE_LEVEL_BITS;
diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
index 438a86bda9f3..4a599130e9c9 100644
--- a/arch/x86/kvm/mmu/spte.c
+++ b/arch/x86/kvm/mmu/spte.c
@@ -61,7 +61,7 @@ static u64 generation_mmio_spte_mask(u64 gen)
 {
 	u64 mask;
 
-	WARN_ON(gen & ~MMIO_SPTE_GEN_MASK);
+	WARN_ON_ONCE(gen & ~MMIO_SPTE_GEN_MASK);
 
 	mask = (gen << MMIO_SPTE_GEN_LOW_SHIFT) & MMIO_SPTE_GEN_LOW_MASK;
 	mask |= (gen << MMIO_SPTE_GEN_HIGH_SHIFT) & MMIO_SPTE_GEN_HIGH_MASK;
@@ -240,7 +240,7 @@ bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
 
 	if ((spte & PT_WRITABLE_MASK) && kvm_slot_dirty_track_enabled(slot)) {
 		/* Enforced by kvm_mmu_hugepage_adjust. */
-		WARN_ON(level > PG_LEVEL_4K);
+		WARN_ON_ONCE(level > PG_LEVEL_4K);
 		mark_page_dirty_in_slot(vcpu->kvm, slot, gfn);
 	}
 
diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index d2eb0d4f8710..5bb09f8d9fc6 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -41,8 +41,8 @@ void tdp_iter_start(struct tdp_iter *iter, struct kvm_mmu_page *root,
 {
 	int root_level = root->role.level;
 
-	WARN_ON(root_level < 1);
-	WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
+	WARN_ON_ONCE(root_level < 1);
+	WARN_ON_ONCE(root_level > PT64_ROOT_MAX_LEVEL);
 
 	iter->next_last_level_gfn = next_last_level_gfn;
 	iter->root_level = root_level;
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index f881de40f9ef..b2068c47f78c 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -475,9 +475,9 @@ static void handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
 	bool is_leaf = is_present && is_last_spte(new_spte, level);
 	bool pfn_changed = spte_to_pfn(old_spte) != spte_to_pfn(new_spte);
 
-	WARN_ON(level > PT64_ROOT_MAX_LEVEL);
-	WARN_ON(level < PG_LEVEL_4K);
-	WARN_ON(gfn & (KVM_PAGES_PER_HPAGE(level) - 1));
+	WARN_ON_ONCE(level > PT64_ROOT_MAX_LEVEL);
+	WARN_ON_ONCE(level < PG_LEVEL_4K);
+	WARN_ON_ONCE(gfn & (KVM_PAGES_PER_HPAGE(level) - 1));
 
 	/*
 	 * If this warning were to trigger it would indicate that there was a
@@ -522,9 +522,9 @@ static void handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
 		 * impact the guest since both the former and current SPTEs
 		 * are nonpresent.
 		 */
-		if (WARN_ON(!is_mmio_spte(old_spte) &&
-			    !is_mmio_spte(new_spte) &&
-			    !is_removed_spte(new_spte)))
+		if (WARN_ON_ONCE(!is_mmio_spte(old_spte) &&
+				 !is_mmio_spte(new_spte) &&
+				 !is_removed_spte(new_spte)))
 			pr_err("Unexpected SPTE change! Nonpresent SPTEs\n"
 			       "should not be replaced with another,\n"
 			       "different nonpresent SPTE, unless one or both\n"
@@ -661,7 +661,7 @@ static u64 tdp_mmu_set_spte(struct kvm *kvm, int as_id, tdp_ptep_t sptep,
 	 * should be used. If operating under the MMU lock in write mode, the
 	 * use of the removed SPTE should not be necessary.
 	 */
-	WARN_ON(is_removed_spte(old_spte) || is_removed_spte(new_spte));
+	WARN_ON_ONCE(is_removed_spte(old_spte) || is_removed_spte(new_spte));
 
 	old_spte = kvm_tdp_mmu_write_spte(sptep, old_spte, new_spte, level);
 
@@ -709,7 +709,7 @@ static inline bool __must_check tdp_mmu_iter_cond_resched(struct kvm *kvm,
 							  struct tdp_iter *iter,
 							  bool flush, bool shared)
 {
-	WARN_ON(iter->yielded);
+	WARN_ON_ONCE(iter->yielded);
 
 	/* Ensure forward progress has been made before yielding. */
 	if (iter->next_last_level_gfn == iter->yielded_gfn)
@@ -728,7 +728,7 @@ static inline bool __must_check tdp_mmu_iter_cond_resched(struct kvm *kvm,
 
 		rcu_read_lock();
 
-		WARN_ON(iter->gfn > iter->next_last_level_gfn);
+		WARN_ON_ONCE(iter->gfn > iter->next_last_level_gfn);
 
 		iter->yielded = true;
 	}
@@ -1241,7 +1241,7 @@ static bool set_spte_gfn(struct kvm *kvm, struct tdp_iter *iter,
 	u64 new_spte;
 
 	/* Huge pages aren't expected to be modified without first being zapped. */
-	WARN_ON(pte_huge(range->pte) || range->start + 1 != range->end);
+	WARN_ON_ONCE(pte_huge(range->pte) || range->start + 1 != range->end);
 
 	if (iter->level != PG_LEVEL_4K ||
 	    !is_shadow_present_pte(iter->old_spte))
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 5/9] KVM: x86/mmu: Bug the VM if a vCPU ends up in long mode without PAE enabled

[Sean Christopherson]

Promote the ASSERT(), which is quite dead code in KVM, into a KVM_BUG_ON()
for KVM's sanity check that CR4.PAE=1 if the vCPU is in long mode when
performing a walk of guest page tables.  The sanity is quite cheap since
neither EFER nor CR4.PAE requires a VMREAD, especially relative to the
cost of walking the guest page tables.

More importantly, the sanity check would have prevented the true badness
fixed by commit 112e66017bff ("KVM: nVMX: add missing consistency checks
for CR0 and CR4").  The missed consistency check resulted in some versions
of KVM corrupting the on-stack guest_walker structure due to KVM thinking
there are 4/5 levels of page tables, but wiring up the MMU hooks to point
at the paging32 implementation, which only allocates space for two levels
of page tables in "struct guest_walker32".

Queue a page fault for injection if the assertion fails, as both callers,
FNAME(gva_to_gpa) and FNAME(walk_addr_generic), assume that walker.fault
contains sane info on a walk failure.  E.g. not populating the fault info
could result in KVM consuming and/or exposing uninitialized stack data
before the vCPU is kicked out to userspace, which doesn't happen until
KVM checks for KVM_REQ_VM_DEAD on the next enter.

Move the check below the initialization of "pte_access" so that the
aforementioned to-be-injected page fault doesn't consume uninitialized
stack data.  The information _shouldn't_ reach the guest or userspace,
but there's zero downside to being paranoid in this case.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/paging_tmpl.h | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index a3fc7c1a7f8d..f8d358226ac6 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -338,7 +338,6 @@ static int FNAME(walk_addr_generic)(struct guest_walker *walker,
 	}
 #endif
 	walker->max_level = walker->level;
-	ASSERT(!(is_long_mode(vcpu) && !is_pae(vcpu)));
 
 	/*
 	 * FIXME: on Intel processors, loads of the PDPTE registers for PAE paging
@@ -348,6 +347,17 @@ static int FNAME(walk_addr_generic)(struct guest_walker *walker,
 	nested_access = (have_ad ? PFERR_WRITE_MASK : 0) | PFERR_USER_MASK;
 
 	pte_access = ~0;
+
+	/*
+	 * Queue a page fault for injection if this assertion fails, as callers
+	 * assume that walker.fault contains sane info on a walk failure.  I.e.
+	 * avoid making the situation worse by inducing even worse badness
+	 * between when the assertion fails and when KVM kicks the vCPU out to
+	 * userspace (because the VM is bugged).
+	 */
+	if (KVM_BUG_ON(is_long_mode(vcpu) && !is_pae(vcpu), vcpu->kvm))
+		goto error;
+
 	++walker->level;
 
 	do {
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 6/9] KVM: x86/mmu: Replace MMU_DEBUG with proper KVM_PROVE_MMU Kconfig

[Sean Christopherson]

Replace MMU_DEBUG, which requires manually modifying KVM to enable the
macro, with a proper Kconfig, KVM_PROVE_MMU.  Now that pgprintk() and
rmap_printk() are gone, i.e. the macro guards only KVM_MMU_WARN_ON() and
won't flood the kernel logs, enabling the option for debug kernels is both
desirable and feasible.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/Kconfig            | 13 +++++++++++++
 arch/x86/kvm/mmu/mmu.c          |  4 ++--
 arch/x86/kvm/mmu/mmu_internal.h |  4 +---
 3 files changed, 16 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig
index 89ca7f4c1464..4e5a282cc518 100644
--- a/arch/x86/kvm/Kconfig
+++ b/arch/x86/kvm/Kconfig
@@ -138,6 +138,19 @@ config KVM_XEN
 
 	  If in doubt, say "N".
 
+config KVM_PROVE_MMU
+	bool "Prove KVM MMU correctness"
+	depends on DEBUG_KERNEL
+	depends on KVM
+	depends on EXPERT
+	help
+	  Enables runtime assertions in KVM's MMU that are too costly to enable
+	  in anything remotely resembling a production environment, e.g. this
+	  gates code that verifies a to-be-freed page table doesn't have any
+	  present SPTEs.
+
+	  If in doubt, say "N".
+
 config KVM_EXTERNAL_WRITE_TRACKING
 	bool
 
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index eb6af9c4cf14..933e48a73a9a 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -1693,7 +1693,7 @@ bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
 	return young;
 }
 
-#ifdef MMU_DEBUG
+#ifdef CONFIG_KVM_PROVE_MMU
 static int is_empty_shadow_page(u64 *spt)
 {
 	u64 *pos;
@@ -1707,7 +1707,7 @@ static int is_empty_shadow_page(u64 *spt)
 		}
 	return 1;
 }
-#endif
+#endif /* CONFIG_KVM_PROVE_MMU */
 
 /*
  * This value is the sum of all of the kvm instances's
diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index cfe925fefa68..40e74db6a7d5 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -6,9 +6,7 @@
 #include <linux/kvm_host.h>
 #include <asm/kvm_host.h>
 
-#undef MMU_DEBUG
-
-#ifdef MMU_DEBUG
+#ifdef CONFIG_KVM_PROVE_MMU
 #define KVM_MMU_WARN_ON(x) WARN_ON_ONCE(x)
 #else
 #define KVM_MMU_WARN_ON(x) do { } while (0)
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 7/9] KVM: x86/mmu: Use BUILD_BUG_ON_INVALID() for KVM_MMU_WARN_ON() stub

[Sean Christopherson]

Use BUILD_BUG_ON_INVALID() instead of an empty do-while loop to stub out
KVM_MMU_WARN_ON() when CONFIG_KVM_PROVE_MMU=n, that way _some_ build
issues with the usage of KVM_MMU_WARN_ON() will be dected even if the
kernel is using the stubs, e.g. basic syntax errors will be detected.

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

diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index 40e74db6a7d5..f1ef670058e5 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -9,7 +9,7 @@
 #ifdef CONFIG_KVM_PROVE_MMU
 #define KVM_MMU_WARN_ON(x) WARN_ON_ONCE(x)
 #else
-#define KVM_MMU_WARN_ON(x) do { } while (0)
+#define KVM_MMU_WARN_ON(x) BUILD_BUG_ON_INVALID(x)
 #endif
 
 /* Page table builder macros common to shadow (host) PTEs and guest PTEs. */
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 8/9] KVM: x86/mmu: Plumb "struct kvm" all the way to pte_list_remove()

[Sean Christopherson]

From: Mingwei Zhang <mizhang@google.com>

Plumb "struct kvm" all the way to pte_list_remove() to allow the usage of
KVM_BUG() and/or KVM_BUG_ON().  This will allow killing only the offending
VM instead of doing BUG() if the kernel is built with
CONFIG_BUG_ON_DATA_CORRUPTION=n, i.e. does NOT want to BUG() if KVM's data
structures (rmaps) appear to be corrupted.

Signed-off-by: Mingwei Zhang <mizhang@google.com>
[sean: tweak changelog]
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c | 33 ++++++++++++++++++---------------
 1 file changed, 18 insertions(+), 15 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 933e48a73a9a..b6cc261d7748 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -965,7 +965,8 @@ static int pte_list_add(struct kvm_mmu_memory_cache *cache, u64 *spte,
 	return count;
 }
 
-static void pte_list_desc_remove_entry(struct kvm_rmap_head *rmap_head,
+static void pte_list_desc_remove_entry(struct kvm *kvm,
+				       struct kvm_rmap_head *rmap_head,
 				       struct pte_list_desc *desc, int i)
 {
 	struct pte_list_desc *head_desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
@@ -1001,7 +1002,8 @@ static void pte_list_desc_remove_entry(struct kvm_rmap_head *rmap_head,
 	mmu_free_pte_list_desc(head_desc);
 }
 
-static void pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
+static void pte_list_remove(struct kvm *kvm, u64 *spte,
+			    struct kvm_rmap_head *rmap_head)
 {
 	struct pte_list_desc *desc;
 	int i;
@@ -1020,7 +1022,8 @@ static void pte_list_remove(u64 *spte, struct kvm_rmap_head *rmap_head)
 		while (desc) {
 			for (i = 0; i < desc->spte_count; ++i) {
 				if (desc->sptes[i] == spte) {
-					pte_list_desc_remove_entry(rmap_head, desc, i);
+					pte_list_desc_remove_entry(kvm, rmap_head,
+								   desc, i);
 					return;
 				}
 			}
@@ -1035,7 +1038,7 @@ static void kvm_zap_one_rmap_spte(struct kvm *kvm,
 				  struct kvm_rmap_head *rmap_head, u64 *sptep)
 {
 	mmu_spte_clear_track_bits(kvm, sptep);
-	pte_list_remove(sptep, rmap_head);
+	pte_list_remove(kvm, sptep, rmap_head);
 }
 
 /* Return true if at least one SPTE was zapped, false otherwise */
@@ -1110,7 +1113,7 @@ static void rmap_remove(struct kvm *kvm, u64 *spte)
 	slot = __gfn_to_memslot(slots, gfn);
 	rmap_head = gfn_to_rmap(gfn, sp->role.level, slot);
 
-	pte_list_remove(spte, rmap_head);
+	pte_list_remove(kvm, spte, rmap_head);
 }
 
 /*
@@ -1758,16 +1761,16 @@ static void mmu_page_add_parent_pte(struct kvm_mmu_memory_cache *cache,
 	pte_list_add(cache, parent_pte, &sp->parent_ptes);
 }
 
-static void mmu_page_remove_parent_pte(struct kvm_mmu_page *sp,
+static void mmu_page_remove_parent_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
 				       u64 *parent_pte)
 {
-	pte_list_remove(parent_pte, &sp->parent_ptes);
+	pte_list_remove(kvm, parent_pte, &sp->parent_ptes);
 }
 
-static void drop_parent_pte(struct kvm_mmu_page *sp,
+static void drop_parent_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
 			    u64 *parent_pte)
 {
-	mmu_page_remove_parent_pte(sp, parent_pte);
+	mmu_page_remove_parent_pte(kvm, sp, parent_pte);
 	mmu_spte_clear_no_track(parent_pte);
 }
 
@@ -2482,7 +2485,7 @@ static void validate_direct_spte(struct kvm_vcpu *vcpu, u64 *sptep,
 		if (child->role.access == direct_access)
 			return;
 
-		drop_parent_pte(child, sptep);
+		drop_parent_pte(vcpu->kvm, child, sptep);
 		kvm_flush_remote_tlbs_sptep(vcpu->kvm, sptep);
 	}
 }
@@ -2500,7 +2503,7 @@ static int mmu_page_zap_pte(struct kvm *kvm, struct kvm_mmu_page *sp,
 			drop_spte(kvm, spte);
 		} else {
 			child = spte_to_child_sp(pte);
-			drop_parent_pte(child, spte);
+			drop_parent_pte(kvm, child, spte);
 
 			/*
 			 * Recursively zap nested TDP SPs, parentless SPs are
@@ -2531,13 +2534,13 @@ static int kvm_mmu_page_unlink_children(struct kvm *kvm,
 	return zapped;
 }
 
-static void kvm_mmu_unlink_parents(struct kvm_mmu_page *sp)
+static void kvm_mmu_unlink_parents(struct kvm *kvm, struct kvm_mmu_page *sp)
 {
 	u64 *sptep;
 	struct rmap_iterator iter;
 
 	while ((sptep = rmap_get_first(&sp->parent_ptes, &iter)))
-		drop_parent_pte(sp, sptep);
+		drop_parent_pte(kvm, sp, sptep);
 }
 
 static int mmu_zap_unsync_children(struct kvm *kvm,
@@ -2576,7 +2579,7 @@ static bool __kvm_mmu_prepare_zap_page(struct kvm *kvm,
 	++kvm->stat.mmu_shadow_zapped;
 	*nr_zapped = mmu_zap_unsync_children(kvm, sp, invalid_list);
 	*nr_zapped += kvm_mmu_page_unlink_children(kvm, sp, invalid_list);
-	kvm_mmu_unlink_parents(sp);
+	kvm_mmu_unlink_parents(kvm, sp);
 
 	/* Zapping children means active_mmu_pages has become unstable. */
 	list_unstable = *nr_zapped;
@@ -2934,7 +2937,7 @@ static int mmu_set_spte(struct kvm_vcpu *vcpu, struct kvm_memory_slot *slot,
 			u64 pte = *sptep;
 
 			child = spte_to_child_sp(pte);
-			drop_parent_pte(child, sptep);
+			drop_parent_pte(vcpu->kvm, child, sptep);
 			flush = true;
 		} else if (pfn != spte_to_pfn(*sptep)) {
 			drop_spte(vcpu->kvm, sptep);
-- 
2.41.0.487.g6d72f3e995-goog

[PATCH v2 9/9] KVM: x86/mmu: BUG() in rmap helpers iff CONFIG_BUG_ON_DATA_CORRUPTION=y

[Sean Christopherson]

Introduce KVM_BUG_ON_DATA_CORRUPTION() and use it in the low-level rmap
helpers to convert the existing BUG()s to WARN_ON_ONCE() when the kernel
is built with CONFIG_BUG_ON_DATA_CORRUPTION=n, i.e. does NOT want to BUG()
on corruption of host kernel data structures.  Environments that don't
have infrastructure to automatically capture crash dumps, i.e. aren't
likely to enable CONFIG_BUG_ON_DATA_CORRUPTION=y, are typically better
served overall by WARN-and-continue behavior (for the kernel, the VM is
dead regardless), as a BUG() while holding mmu_lock all but guarantees
the _best_ case scenario is a panic().

Make the BUG()s conditional instead of removing/replacing them entirely as
there's a non-zero chance (though by no means a guarantee) that the damage
isn't contained to the target VM, e.g. if no rmap is found for a SPTE then
KVM may be double-zapping the SPTE, i.e. has already freed the memory the
SPTE pointed at and thus KVM is reading/writing memory that KVM no longer
owns.

Link: https://lore.kernel.org/all/20221129191237.31447-1-mizhang@google.com
Suggested-by: Mingwei Zhang <mizhang@google.com>
Cc: David Matlack <dmatlack@google.com>
Cc: Jim Mattson <jmattson@google.com>
Reviewed-by: Mingwei Zhang <mizhang@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/mmu/mmu.c   | 21 ++++++++++-----------
 include/linux/kvm_host.h | 19 +++++++++++++++++++
 2 files changed, 29 insertions(+), 11 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index b6cc261d7748..69f65f7b6158 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -977,7 +977,7 @@ static void pte_list_desc_remove_entry(struct kvm *kvm,
 	 * when adding an entry and the previous head is full, and heads are
 	 * removed (this flow) when they become empty.
 	 */
-	BUG_ON(j < 0);
+	KVM_BUG_ON_DATA_CORRUPTION(j < 0, kvm);
 
 	/*
 	 * Replace the to-be-freed SPTE with the last valid entry from the head
@@ -1008,14 +1008,13 @@ static void pte_list_remove(struct kvm *kvm, u64 *spte,
 	struct pte_list_desc *desc;
 	int i;
 
-	if (!rmap_head->val) {
-		pr_err("%s: %p 0->BUG\n", __func__, spte);
-		BUG();
-	} else if (!(rmap_head->val & 1)) {
-		if ((u64 *)rmap_head->val != spte) {
-			pr_err("%s:  %p 1->BUG\n", __func__, spte);
-			BUG();
-		}
+	if (KVM_BUG_ON_DATA_CORRUPTION(!rmap_head->val, kvm))
+		return;
+
+	if (!(rmap_head->val & 1)) {
+		if (KVM_BUG_ON_DATA_CORRUPTION((u64 *)rmap_head->val != spte, kvm))
+			return;
+
 		rmap_head->val = 0;
 	} else {
 		desc = (struct pte_list_desc *)(rmap_head->val & ~1ul);
@@ -1029,8 +1028,8 @@ static void pte_list_remove(struct kvm *kvm, u64 *spte,
 			}
 			desc = desc->more;
 		}
-		pr_err("%s: %p many->many\n", __func__, spte);
-		BUG();
+
+		KVM_BUG_ON_DATA_CORRUPTION(true, kvm);
 	}
 }
 
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 9d3ac7720da9..cb86108c624d 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -865,6 +865,25 @@ static inline void kvm_vm_bugged(struct kvm *kvm)
 	unlikely(__ret);					\
 })
 
+/*
+ * Note, "data corruption" refers to corruption of host kernel data structures,
+ * not guest data.  Guest data corruption, suspected or confirmed, that is tied
+ * and contained to a single VM should *never* BUG() and potentially panic the
+ * host, i.e. use this variant of KVM_BUG() if and only if a KVM data structure
+ * is corrupted and that corruption can have a cascading effect to other parts
+ * of the hosts and/or to other VMs.
+ */
+#define KVM_BUG_ON_DATA_CORRUPTION(cond, kvm)			\
+({								\
+	bool __ret = !!(cond);					\
+								\
+	if (IS_ENABLED(CONFIG_BUG_ON_DATA_CORRUPTION))		\
+		BUG_ON(__ret);					\
+	else if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged))	\
+		kvm_vm_bugged(kvm);				\
+	unlikely(__ret);					\
+})
+
 static inline void kvm_vcpu_srcu_read_lock(struct kvm_vcpu *vcpu)
 {
 #ifdef CONFIG_PROVE_RCU
-- 
2.41.0.487.g6d72f3e995-goog