1 files changed, 61 insertions, 17 deletions

diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h
index 48a706f641f2..68da67df304d 100644
--- a/arch/x86/include/asm/qspinlock.h
+++ b/arch/x86/include/asm/qspinlock.h
@@ -1,9 +1,38 @@
+/* SPDX-License-Identifier: GPL-2.0 */
 #ifndef _ASM_X86_QSPINLOCK_H
 #define _ASM_X86_QSPINLOCK_H
 
+#include <linux/jump_label.h>
 #include <asm/cpufeature.h>
 #include <asm-generic/qspinlock_types.h>
 #include <asm/paravirt.h>
+#include <asm/rmwcc.h>
+
+#define _Q_PENDING_LOOPS	(1 << 9)
+
+#define queued_fetch_set_pending_acquire queued_fetch_set_pending_acquire
+static __always_inline u32 queued_fetch_set_pending_acquire(struct qspinlock *lock)
+{
+	u32 val;
+
+	/*
+	 * We can't use GEN_BINARY_RMWcc() inside an if() stmt because asm goto
+	 * and CONFIG_PROFILE_ALL_BRANCHES=y results in a label inside a
+	 * statement expression, which GCC doesn't like.
+	 */
+	val = GEN_BINARY_RMWcc(LOCK_PREFIX "btsl", lock->val.counter, c,
+			       "I", _Q_PENDING_OFFSET) * _Q_PENDING_VAL;
+	val |= atomic_read(&lock->val) & ~_Q_PENDING_MASK;
+
+	return val;
+}
+
+#ifdef CONFIG_PARAVIRT_SPINLOCKS
+extern void native_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
+extern void __pv_init_lock_hash(void);
+extern void __pv_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
+extern void __raw_callee_save___pv_queued_spin_unlock(struct qspinlock *lock);
+extern bool nopvspin;
 
 #define	queued_spin_unlock queued_spin_unlock
 /**
@@ -14,15 +43,9 @@
  */
 static inline void native_queued_spin_unlock(struct qspinlock *lock)
 {
-	smp_store_release((u8 *)lock, 0);
+	smp_store_release(&lock->locked, 0);
 }
 
-#ifdef CONFIG_PARAVIRT_SPINLOCKS
-extern void native_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
-extern void __pv_init_lock_hash(void);
-extern void __pv_queued_spin_lock_slowpath(struct qspinlock *lock, u32 val);
-extern void __raw_callee_save___pv_queued_spin_unlock(struct qspinlock *lock);
-
 static inline void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
 {
 	pv_queued_spin_lock_slowpath(lock, val);
@@ -30,6 +53,7 @@ static inline void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
 
 static inline void queued_spin_unlock(struct qspinlock *lock)
 {
+	kcsan_release();
 	pv_queued_spin_unlock(lock);
 }
 
@@ -38,18 +62,34 @@ static inline bool vcpu_is_preempted(long cpu)
 {
 	return pv_vcpu_is_preempted(cpu);
 }
-#else
-static inline void queued_spin_unlock(struct qspinlock *lock)
-{
-	native_queued_spin_unlock(lock);
-}
 #endif
 
 #ifdef CONFIG_PARAVIRT
+/*
+ * virt_spin_lock_key - disables by default the virt_spin_lock() hijack.
+ *
+ * Native (and PV wanting native due to vCPU pinning) should keep this key
+ * disabled. Native does not touch the key.
+ *
+ * When in a guest then native_pv_lock_init() enables the key first and
+ * KVM/XEN might conditionally disable it later in the boot process again.
+ */
+DECLARE_STATIC_KEY_FALSE(virt_spin_lock_key);
+
+/*
+ * Shortcut for the queued_spin_lock_slowpath() function that allows
+ * virt to hijack it.
+ *
+ * Returns:
+ *   true - lock has been negotiated, all done;
+ *   false - queued_spin_lock_slowpath() will do its thing.
+ */
 #define virt_spin_lock virt_spin_lock
 static inline bool virt_spin_lock(struct qspinlock *lock)
 {
-	if (!static_cpu_has(X86_FEATURE_HYPERVISOR))
+	int val;
+
+	if (!static_branch_likely(&virt_spin_lock_key))
 		return false;
 
 	/*
@@ -58,13 +98,17 @@ static inline bool virt_spin_lock(struct qspinlock *lock)
 	 * horrible lock 'holder' preemption issues.
 	 */
 
-	do {
-		while (atomic_read(&lock->val) != 0)
-			cpu_relax();
-	} while (atomic_cmpxchg(&lock->val, 0, _Q_LOCKED_VAL) != 0);
+ __retry:
+	val = atomic_read(&lock->val);
+
+	if (val || !atomic_try_cmpxchg(&lock->val, &val, _Q_LOCKED_VAL)) {
+		cpu_relax();
+		goto __retry;
+	}
 
 	return true;
 }
+
 #endif /* CONFIG_PARAVIRT */
 
 #include <asm-generic/qspinlock.h>