Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull s390 updates from Martin Schwidefsky:
 "The first part of the s390 updates for 4.14:

   - Add machine type 0x3906 for IBM z14

   - Add IBM z14 TLB flushing improvements for KVM guests

   - Exploit the TOD clock epoch extension to provide a continuous TOD
     clock afer 2042/09/17

   - Add NIAI spinlock hints for IBM z14

   - Rework the vmcp driver and use CMA for the respone buffer of z/VM
     CP commands

   - Drop some s390 specific asm headers and use the generic version

   - Add block discard for DASD-FBA devices under z/VM

   - Add average request times to DASD statistics

   - A few of those constify patches which seem to be in vogue right now

   - Cleanup and bug fixes"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (50 commits)
  s390/mm: avoid empty zero pages for KVM guests to avoid postcopy hangs
  s390/dasd: Add discard support for FBA devices
  s390/zcrypt: make CPRBX const
  s390/uaccess: avoid mvcos jump label
  s390/mm: use generic mm_hooks
  s390/facilities: fix typo
  s390/vmcp: simplify vmcp_response_free()
  s390/topology: Remove the unused parent_node() macro
  s390/dasd: Change unsigned long long to unsigned long
  s390/smp: convert cpuhp_setup_state() return code to zero on success
  s390: fix 'novx' early parameter handling
  s390/dasd: add average request times to dasd statistics
  s390/scm: use common completion path
  s390/pci: log changes to uid checking
  s390/vmcp: simplify vmcp_ioctl()
  s390/vmcp: return -ENOTTY for unknown ioctl commands
  s390/vmcp: split vmcp header file and move to uapi
  s390/vmcp: make use of contiguous memory allocator
  s390/cpcmd,vmcp: avoid GFP_DMA allocations
  s390/vmcp: fix uaccess check and avoid undefined behavior
  ...

1 files changed, 51 insertions, 36 deletions

diff --git a/arch/s390/lib/spinlock.c b/arch/s390/lib/spinlock.c
index ffb15bd4c593..b12663d653d8 100644
--- a/arch/s390/lib/spinlock.c
+++ b/arch/s390/lib/spinlock.c
@@ -32,42 +32,63 @@ static int __init spin_retry_setup(char *str)
 }
 __setup("spin_retry=", spin_retry_setup);
 
+static inline int arch_load_niai4(int *lock)
+{
+	int owner;
+
+	asm volatile(
+#ifdef CONFIG_HAVE_MARCH_ZEC12_FEATURES
+		"	.long	0xb2fa0040\n"	/* NIAI 4 */
+#endif
+		"	l	%0,%1\n"
+		: "=d" (owner) : "Q" (*lock) : "memory");
+       return owner;
+}
+
+static inline int arch_cmpxchg_niai8(int *lock, int old, int new)
+{
+	int expected = old;
+
+	asm volatile(
+#ifdef CONFIG_HAVE_MARCH_ZEC12_FEATURES
+		"	.long	0xb2fa0080\n"	/* NIAI 8 */
+#endif
+		"	cs	%0,%3,%1\n"
+		: "=d" (old), "=Q" (*lock)
+		: "0" (old), "d" (new), "Q" (*lock)
+		: "cc", "memory");
+	return expected == old;
+}
+
 void arch_spin_lock_wait(arch_spinlock_t *lp)
 {
 	int cpu = SPINLOCK_LOCKVAL;
-	int owner, count, first_diag;
+	int owner, count;
+
+	/* Pass the virtual CPU to the lock holder if it is not running */
+	owner = arch_load_niai4(&lp->lock);
+	if (owner && arch_vcpu_is_preempted(~owner))
+		smp_yield_cpu(~owner);
 
-	first_diag = 1;
+	count = spin_retry;
 	while (1) {
-		owner = ACCESS_ONCE(lp->lock);
+		owner = arch_load_niai4(&lp->lock);
 		/* Try to get the lock if it is free. */
 		if (!owner) {
-			if (__atomic_cmpxchg_bool(&lp->lock, 0, cpu))
+			if (arch_cmpxchg_niai8(&lp->lock, 0, cpu))
 				return;
 			continue;
 		}
-		/* First iteration: check if the lock owner is running. */
-		if (first_diag && arch_vcpu_is_preempted(~owner)) {
-			smp_yield_cpu(~owner);
-			first_diag = 0;
+		if (count-- >= 0)
 			continue;
-		}
-		/* Loop for a while on the lock value. */
 		count = spin_retry;
-		do {
-			owner = ACCESS_ONCE(lp->lock);
-		} while (owner && count-- > 0);
-		if (!owner)
-			continue;
 		/*
 		 * For multiple layers of hypervisors, e.g. z/VM + LPAR
 		 * yield the CPU unconditionally. For LPAR rely on the
 		 * sense running status.
 		 */
-		if (!MACHINE_IS_LPAR || arch_vcpu_is_preempted(~owner)) {
+		if (!MACHINE_IS_LPAR || arch_vcpu_is_preempted(~owner))
 			smp_yield_cpu(~owner);
-			first_diag = 0;
-		}
 	}
 }
 EXPORT_SYMBOL(arch_spin_lock_wait);
@@ -75,42 +96,36 @@ EXPORT_SYMBOL(arch_spin_lock_wait);
 void arch_spin_lock_wait_flags(arch_spinlock_t *lp, unsigned long flags)
 {
 	int cpu = SPINLOCK_LOCKVAL;
-	int owner, count, first_diag;
+	int owner, count;
 
 	local_irq_restore(flags);
-	first_diag = 1;
+
+	/* Pass the virtual CPU to the lock holder if it is not running */
+	owner = arch_load_niai4(&lp->lock);
+	if (owner && arch_vcpu_is_preempted(~owner))
+		smp_yield_cpu(~owner);
+
+	count = spin_retry;
 	while (1) {
-		owner = ACCESS_ONCE(lp->lock);
+		owner = arch_load_niai4(&lp->lock);
 		/* Try to get the lock if it is free. */
 		if (!owner) {
 			local_irq_disable();
-			if (__atomic_cmpxchg_bool(&lp->lock, 0, cpu))
+			if (arch_cmpxchg_niai8(&lp->lock, 0, cpu))
 				return;
 			local_irq_restore(flags);
 			continue;
 		}
-		/* Check if the lock owner is running. */
-		if (first_diag && arch_vcpu_is_preempted(~owner)) {
-			smp_yield_cpu(~owner);
-			first_diag = 0;
+		if (count-- >= 0)
 			continue;
-		}
-		/* Loop for a while on the lock value. */
 		count = spin_retry;
-		do {
-			owner = ACCESS_ONCE(lp->lock);
-		} while (owner && count-- > 0);
-		if (!owner)
-			continue;
 		/*
 		 * For multiple layers of hypervisors, e.g. z/VM + LPAR
 		 * yield the CPU unconditionally. For LPAR rely on the
 		 * sense running status.
 		 */
-		if (!MACHINE_IS_LPAR || arch_vcpu_is_preempted(~owner)) {
+		if (!MACHINE_IS_LPAR || arch_vcpu_is_preempted(~owner))
 			smp_yield_cpu(~owner);
-			first_diag = 0;
-		}
 	}
 }
 EXPORT_SYMBOL(arch_spin_lock_wait_flags);