[PATCH v2 0/4] Fix RCU warnings in TDP MMU

[PATCH v2 0/4] Fix RCU warnings in TDP MMU

[Ben Gardon]

The Linux Test Robot found a few RCU warnings in the TDP MMU:
https://www.spinics.net/lists/kernel/msg3845500.html
https://www.spinics.net/lists/kernel/msg3845521.html

Fix these warnings and cleanup a hack in tdp_mmu_iter_cond_resched.

Tested by compiling as suggested in the test robot report and confirmed
that the warnings go away with this series applied. Also ran
kvm-unit-tests on an Intel Skylake machine with the TDP MMU enabled and
confirmed that the series introduced no new failures.

Ben Gardon (4):
  KVM: x86/mmu: Fix RCU usage in handle_removed_tdp_mmu_page
  KVM: x86/mmu: Fix RCU usage when atomically zapping SPTEs
  KVM: x86/mmu: Factor out tdp_iter_return_to_root
  KVM: x86/mmu: Store the address space ID in the TDP iterator

 arch/x86/kvm/mmu/mmu_internal.h |  5 +++++
 arch/x86/kvm/mmu/tdp_iter.c     | 30 +++++++++++++++----------
 arch/x86/kvm/mmu/tdp_iter.h     |  4 +++-
 arch/x86/kvm/mmu/tdp_mmu.c      | 40 +++++++++++++--------------------
 4 files changed, 41 insertions(+), 38 deletions(-)

-- 
2.31.0.rc2.261.g7f71774620-goog

[PATCH v2 1/4] KVM: x86/mmu: Fix RCU usage in handle_removed_tdp_mmu_page

[Ben Gardon]

The pt passed into handle_removed_tdp_mmu_page does not need RCU
protection, as it is not at any risk of being freed by another thread at
that point. However, the implicit cast from tdp_sptep_t to u64 * dropped
the __rcu annotation without a proper rcu_derefrence. Fix this by
passing the pt as a tdp_ptep_t and then rcu_dereferencing it in
the function.

Suggested-by: Sean Christopherson <seanjc@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Ben Gardon <bgardon@google.com>
---
 arch/x86/kvm/mmu/tdp_mmu.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index d78915019b08..db2936cca4bf 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -301,11 +301,16 @@ static void tdp_mmu_unlink_page(struct kvm *kvm, struct kvm_mmu_page *sp,
  *
  * Given a page table that has been removed from the TDP paging structure,
  * iterates through the page table to clear SPTEs and free child page tables.
+ *
+ * Note that pt is passed in as a tdp_ptep_t, but it does not need RCU
+ * protection. Since this thread removed it from the paging structure,
+ * this thread will be responsible for ensuring the page is freed. Hence the
+ * early rcu_dereferences in the function.
  */
-static void handle_removed_tdp_mmu_page(struct kvm *kvm, u64 *pt,
+static void handle_removed_tdp_mmu_page(struct kvm *kvm, tdp_ptep_t pt,
 					bool shared)
 {
-	struct kvm_mmu_page *sp = sptep_to_sp(pt);
+	struct kvm_mmu_page *sp = sptep_to_sp(rcu_dereference(pt));
 	int level = sp->role.level;
 	gfn_t base_gfn = sp->gfn;
 	u64 old_child_spte;
@@ -318,7 +323,7 @@ static void handle_removed_tdp_mmu_page(struct kvm *kvm, u64 *pt,
 	tdp_mmu_unlink_page(kvm, sp, shared);
 
 	for (i = 0; i < PT64_ENT_PER_PAGE; i++) {
-		sptep = pt + i;
+		sptep = rcu_dereference(pt) + i;
 		gfn = base_gfn + (i * KVM_PAGES_PER_HPAGE(level - 1));
 
 		if (shared) {
-- 
2.31.0.rc2.261.g7f71774620-goog

[PATCH v2 2/4] KVM: x86/mmu: Fix RCU usage when atomically zapping SPTEs

[Ben Gardon]

Fix a missing rcu_dereference in tdp_mmu_zap_spte_atomic.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Ben Gardon <bgardon@google.com>
---
 arch/x86/kvm/mmu/tdp_mmu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index db2936cca4bf..946da74e069c 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -543,7 +543,7 @@ static inline bool tdp_mmu_zap_spte_atomic(struct kvm *kvm,
 	 * here since the SPTE is going from non-present
 	 * to non-present.
 	 */
-	WRITE_ONCE(*iter->sptep, 0);
+	WRITE_ONCE(*rcu_dereference(iter->sptep), 0);
 
 	return true;
 }
-- 
2.31.0.rc2.261.g7f71774620-goog

Re: [PATCH v2 2/4] KVM: x86/mmu: Fix RCU usage when atomically zapping SPTEs

[Sean Christopherson]

On Mon, Mar 15, 2021, Ben Gardon wrote:
> Fix a missing rcu_dereference in tdp_mmu_zap_spte_atomic.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---

Reviewed-by: Sean Christopherson <seanjc@google.com>

[PATCH v2 3/4] KVM: x86/mmu: Factor out tdp_iter_return_to_root

[Ben Gardon]

In tdp_mmu_iter_cond_resched there is a call to tdp_iter_start which
causes the iterator to continue its walk over the paging structure from
the root. This is needed after a yield as paging structure could have
been freed in the interim.

The tdp_iter_start call is not very clear and something of a hack. It
requires exposing tdp_iter fields not used elsewhere in tdp_mmu.c and
the effect is not obvious from the function name. Factor a more aptly
named function out of tdp_iter_start and call it from
tdp_mmu_iter_cond_resched and tdp_iter_start.

No functional change intended.

Signed-off-by: Ben Gardon <bgardon@google.com>
---
 arch/x86/kvm/mmu/tdp_iter.c | 24 +++++++++++++++++-------
 arch/x86/kvm/mmu/tdp_iter.h |  1 +
 arch/x86/kvm/mmu/tdp_mmu.c  |  4 +---
 3 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index e5f148106e20..f7f94ea65243 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -20,6 +20,21 @@ static gfn_t round_gfn_for_level(gfn_t gfn, int level)
 	return gfn & -KVM_PAGES_PER_HPAGE(level);
 }
 
+/*
+ * Return the TDP iterator to the root PT and allow it to continue its
+ * traversal over the paging structure from there.
+ */
+void tdp_iter_restart(struct tdp_iter *iter)
+{
+	iter->yielded_gfn = iter->next_last_level_gfn;
+	iter->level = iter->root_level;
+
+	iter->gfn = round_gfn_for_level(iter->next_last_level_gfn, iter->level);
+	tdp_iter_refresh_sptep(iter);
+
+	iter->valid = true;
+}
+
 /*
  * Sets a TDP iterator to walk a pre-order traversal of the paging structure
  * rooted at root_pt, starting with the walk to translate next_last_level_gfn.
@@ -31,16 +46,11 @@ void tdp_iter_start(struct tdp_iter *iter, u64 *root_pt, int root_level,
 	WARN_ON(root_level > PT64_ROOT_MAX_LEVEL);
 
 	iter->next_last_level_gfn = next_last_level_gfn;
-	iter->yielded_gfn = iter->next_last_level_gfn;
 	iter->root_level = root_level;
 	iter->min_level = min_level;
-	iter->level = root_level;
-	iter->pt_path[iter->level - 1] = (tdp_ptep_t)root_pt;
+	iter->pt_path[iter->root_level - 1] = (tdp_ptep_t)root_pt;
 
-	iter->gfn = round_gfn_for_level(iter->next_last_level_gfn, iter->level);
-	tdp_iter_refresh_sptep(iter);
-
-	iter->valid = true;
+	tdp_iter_restart(iter);
 }
 
 /*
diff --git a/arch/x86/kvm/mmu/tdp_iter.h b/arch/x86/kvm/mmu/tdp_iter.h
index 4cc177d75c4a..8eb424d17c91 100644
--- a/arch/x86/kvm/mmu/tdp_iter.h
+++ b/arch/x86/kvm/mmu/tdp_iter.h
@@ -63,5 +63,6 @@ void tdp_iter_start(struct tdp_iter *iter, u64 *root_pt, int root_level,
 		    int min_level, gfn_t next_last_level_gfn);
 void tdp_iter_next(struct tdp_iter *iter);
 tdp_ptep_t tdp_iter_root_pt(struct tdp_iter *iter);
+void tdp_iter_restart(struct tdp_iter *iter);
 
 #endif /* __KVM_X86_MMU_TDP_ITER_H */
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 946da74e069c..38b6b6936171 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -664,9 +664,7 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
 
 		WARN_ON(iter->gfn > iter->next_last_level_gfn);
 
-		tdp_iter_start(iter, iter->pt_path[iter->root_level - 1],
-			       iter->root_level, iter->min_level,
-			       iter->next_last_level_gfn);
+		tdp_iter_restart(iter);
 
 		return true;
 	}
-- 
2.31.0.rc2.261.g7f71774620-goog

Re: [PATCH v2 3/4] KVM: x86/mmu: Factor out tdp_iter_return_to_root

[Sean Christopherson]

On Mon, Mar 15, 2021, Ben Gardon wrote:
> In tdp_mmu_iter_cond_resched there is a call to tdp_iter_start which
> causes the iterator to continue its walk over the paging structure from
> the root. This is needed after a yield as paging structure could have
> been freed in the interim.
> 
> The tdp_iter_start call is not very clear and something of a hack. It
> requires exposing tdp_iter fields not used elsewhere in tdp_mmu.c and
> the effect is not obvious from the function name. Factor a more aptly
> named function out of tdp_iter_start and call it from
> tdp_mmu_iter_cond_resched and tdp_iter_start.
> 
> No functional change intended.
> 
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---

Very nice, sooo much easier to read.

Reviewed-by: Sean Christopherson <seanjc@google.com>

[PATCH v2 4/4] KVM: x86/mmu: Store the address space ID in the TDP iterator

[Ben Gardon]

Store the address space ID in the TDP iterator so that it can be
retrieved without having to bounce through the root shadow page.  This
streamlines the code and fixes a Sparse warning about not properly using
rcu_dereference() when grabbing the ID from the root on the fly.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Ben Gardon <bgardon@google.com>
---
 arch/x86/kvm/mmu/mmu_internal.h |  5 +++++
 arch/x86/kvm/mmu/tdp_iter.c     |  6 +-----
 arch/x86/kvm/mmu/tdp_iter.h     |  3 ++-
 arch/x86/kvm/mmu/tdp_mmu.c      | 23 +++++------------------
 4 files changed, 13 insertions(+), 24 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu_internal.h b/arch/x86/kvm/mmu/mmu_internal.h
index ec4fc28b325a..1f6f98c76bdf 100644
--- a/arch/x86/kvm/mmu/mmu_internal.h
+++ b/arch/x86/kvm/mmu/mmu_internal.h
@@ -78,6 +78,11 @@ static inline struct kvm_mmu_page *sptep_to_sp(u64 *sptep)
 	return to_shadow_page(__pa(sptep));
 }
 
+static inline int kvm_mmu_page_as_id(struct kvm_mmu_page *sp)
+{
+	return sp->role.smm ? 1 : 0;
+}
+
 static inline bool kvm_vcpu_ad_need_write_protect(struct kvm_vcpu *vcpu)
 {
 	/*
diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index f7f94ea65243..b3ed302c1a35 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -49,6 +49,7 @@ void tdp_iter_start(struct tdp_iter *iter, u64 *root_pt, int root_level,
 	iter->root_level = root_level;
 	iter->min_level = min_level;
 	iter->pt_path[iter->root_level - 1] = (tdp_ptep_t)root_pt;
+	iter->as_id = kvm_mmu_page_as_id(sptep_to_sp(root_pt));
 
 	tdp_iter_restart(iter);
 }
@@ -169,8 +170,3 @@ void tdp_iter_next(struct tdp_iter *iter)
 	iter->valid = false;
 }
 
-tdp_ptep_t tdp_iter_root_pt(struct tdp_iter *iter)
-{
-	return iter->pt_path[iter->root_level - 1];
-}
-
diff --git a/arch/x86/kvm/mmu/tdp_iter.h b/arch/x86/kvm/mmu/tdp_iter.h
index 8eb424d17c91..b1748b988d3a 100644
--- a/arch/x86/kvm/mmu/tdp_iter.h
+++ b/arch/x86/kvm/mmu/tdp_iter.h
@@ -36,6 +36,8 @@ struct tdp_iter {
 	int min_level;
 	/* The iterator's current level within the paging structure */
 	int level;
+	/* The address space ID, i.e. SMM vs. regular. */
+	int as_id;
 	/* A snapshot of the value at sptep */
 	u64 old_spte;
 	/*
@@ -62,7 +64,6 @@ tdp_ptep_t spte_to_child_pt(u64 pte, int level);
 void tdp_iter_start(struct tdp_iter *iter, u64 *root_pt, int root_level,
 		    int min_level, gfn_t next_last_level_gfn);
 void tdp_iter_next(struct tdp_iter *iter);
-tdp_ptep_t tdp_iter_root_pt(struct tdp_iter *iter);
 void tdp_iter_restart(struct tdp_iter *iter);
 
 #endif /* __KVM_X86_MMU_TDP_ITER_H */
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 38b6b6936171..462b1f71c77f 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -203,11 +203,6 @@ static void handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
 				u64 old_spte, u64 new_spte, int level,
 				bool shared);
 
-static int kvm_mmu_page_as_id(struct kvm_mmu_page *sp)
-{
-	return sp->role.smm ? 1 : 0;
-}
-
 static void handle_changed_spte_acc_track(u64 old_spte, u64 new_spte, int level)
 {
 	bool pfn_changed = spte_to_pfn(old_spte) != spte_to_pfn(new_spte);
@@ -497,10 +492,6 @@ static inline bool tdp_mmu_set_spte_atomic(struct kvm *kvm,
 					   struct tdp_iter *iter,
 					   u64 new_spte)
 {
-	u64 *root_pt = tdp_iter_root_pt(iter);
-	struct kvm_mmu_page *root = sptep_to_sp(root_pt);
-	int as_id = kvm_mmu_page_as_id(root);
-
 	lockdep_assert_held_read(&kvm->mmu_lock);
 
 	/*
@@ -514,8 +505,8 @@ static inline bool tdp_mmu_set_spte_atomic(struct kvm *kvm,
 		      new_spte) != iter->old_spte)
 		return false;
 
-	handle_changed_spte(kvm, as_id, iter->gfn, iter->old_spte, new_spte,
-			    iter->level, true);
+	handle_changed_spte(kvm, iter->as_id, iter->gfn, iter->old_spte,
+			    new_spte, iter->level, true);
 
 	return true;
 }
@@ -569,10 +560,6 @@ static inline void __tdp_mmu_set_spte(struct kvm *kvm, struct tdp_iter *iter,
 				      u64 new_spte, bool record_acc_track,
 				      bool record_dirty_log)
 {
-	tdp_ptep_t root_pt = tdp_iter_root_pt(iter);
-	struct kvm_mmu_page *root = sptep_to_sp(root_pt);
-	int as_id = kvm_mmu_page_as_id(root);
-
 	lockdep_assert_held_write(&kvm->mmu_lock);
 
 	/*
@@ -586,13 +573,13 @@ static inline void __tdp_mmu_set_spte(struct kvm *kvm, struct tdp_iter *iter,
 
 	WRITE_ONCE(*rcu_dereference(iter->sptep), new_spte);
 
-	__handle_changed_spte(kvm, as_id, iter->gfn, iter->old_spte, new_spte,
-			      iter->level, false);
+	__handle_changed_spte(kvm, iter->as_id, iter->gfn, iter->old_spte,
+			      new_spte, iter->level, false);
 	if (record_acc_track)
 		handle_changed_spte_acc_track(iter->old_spte, new_spte,
 					      iter->level);
 	if (record_dirty_log)
-		handle_changed_spte_dirty_log(kvm, as_id, iter->gfn,
+		handle_changed_spte_dirty_log(kvm, iter->as_id, iter->gfn,
 					      iter->old_spte, new_spte,
 					      iter->level);
 }
-- 
2.31.0.rc2.261.g7f71774620-goog

Re: [PATCH v2 4/4] KVM: x86/mmu: Store the address space ID in the TDP iterator

[Sean Christopherson]

On Mon, Mar 15, 2021, Ben Gardon wrote:

Missing "From: Sean Christopherson <seanjc@google.com>", i.e. the commit in your
local tree needs "git commit --amend --author="Sean Christopherson <seanjc@google.com>".
Alternatively, you could just erase my SOB ;-)

> Store the address space ID in the TDP iterator so that it can be
> retrieved without having to bounce through the root shadow page.  This
> streamlines the code and fixes a Sparse warning about not properly using
> rcu_dereference() when grabbing the ID from the root on the fly.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Sean Christopherson <seanjc@google.com>
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---