14 files changed, 475 insertions, 150 deletions

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 6dab9c1be508..5b39918b7042 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -117,6 +117,7 @@ config S390
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_DEBUG_PAGEALLOC
 	select ARCH_SUPPORTS_HUGETLBFS
+	select ARCH_SUPPORTS_INT128 if CC_HAS_INT128 && CC_IS_CLANG
 	select ARCH_SUPPORTS_NUMA_BALANCING
 	select ARCH_SUPPORTS_PER_VMA_LOCK
 	select ARCH_USE_BUILTIN_BSWAP
diff --git a/arch/s390/crypto/paes_s390.c b/arch/s390/crypto/paes_s390.c
index 29dc827e0fe8..d29a9d908797 100644
--- a/arch/s390/crypto/paes_s390.c
+++ b/arch/s390/crypto/paes_s390.c
@@ -5,7 +5,7 @@
  * s390 implementation of the AES Cipher Algorithm with protected keys.
  *
  * s390 Version:
- *   Copyright IBM Corp. 2017,2020
+ *   Copyright IBM Corp. 2017, 2023
  *   Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>
  *		Harald Freudenberger <freude@de.ibm.com>
  */
@@ -132,7 +132,8 @@ static inline int __paes_keyblob2pkey(struct key_blob *kb,
 		if (i > 0 && ret == -EAGAIN && in_task())
 			if (msleep_interruptible(1000))
 				return -EINTR;
-		ret = pkey_keyblob2pkey(kb->key, kb->keylen, pk);
+		ret = pkey_keyblob2pkey(kb->key, kb->keylen,
+					pk->protkey, &pk->len, &pk->type);
 		if (ret == 0)
 			break;
 	}
@@ -145,6 +146,7 @@ static inline int __paes_convert_key(struct s390_paes_ctx *ctx)
 	int ret;
 	struct pkey_protkey pkey;
 
+	pkey.len = sizeof(pkey.protkey);
 	ret = __paes_keyblob2pkey(&ctx->kb, &pkey);
 	if (ret)
 		return ret;
@@ -414,6 +416,9 @@ static inline int __xts_paes_convert_key(struct s390_pxts_ctx *ctx)
 {
 	struct pkey_protkey pkey0, pkey1;
 
+	pkey0.len = sizeof(pkey0.protkey);
+	pkey1.len = sizeof(pkey1.protkey);
+
 	if (__paes_keyblob2pkey(&ctx->kb[0], &pkey0) ||
 	    __paes_keyblob2pkey(&ctx->kb[1], &pkey1))
 		return -EINVAL;
diff --git a/arch/s390/include/asm/asm-prototypes.h b/arch/s390/include/asm/asm-prototypes.h
index c37eb921bfbf..a873e873e1ee 100644
--- a/arch/s390/include/asm/asm-prototypes.h
+++ b/arch/s390/include/asm/asm-prototypes.h
@@ -6,4 +6,8 @@
 #include <asm/fpu/api.h>
 #include <asm-generic/asm-prototypes.h>
 
+__int128_t __ashlti3(__int128_t a, int b);
+__int128_t __ashrti3(__int128_t a, int b);
+__int128_t __lshrti3(__int128_t a, int b);
+
 #endif /* _ASM_S390_PROTOTYPES_H */
diff --git a/arch/s390/include/asm/cpacf.h b/arch/s390/include/asm/cpacf.h
index 646b12981f20..b378e2b57ad8 100644
--- a/arch/s390/include/asm/cpacf.h
+++ b/arch/s390/include/asm/cpacf.h
@@ -2,7 +2,7 @@
 /*
  * CP Assist for Cryptographic Functions (CPACF)
  *
- * Copyright IBM Corp. 2003, 2017
+ * Copyright IBM Corp. 2003, 2023
  * Author(s): Thomas Spatzier
  *	      Jan Glauber
  *	      Harald Freudenberger (freude@de.ibm.com)
@@ -132,6 +132,11 @@
 #define CPACF_PCKMO_ENC_AES_128_KEY	0x12
 #define CPACF_PCKMO_ENC_AES_192_KEY	0x13
 #define CPACF_PCKMO_ENC_AES_256_KEY	0x14
+#define CPACF_PCKMO_ENC_ECC_P256_KEY	0x20
+#define CPACF_PCKMO_ENC_ECC_P384_KEY	0x21
+#define CPACF_PCKMO_ENC_ECC_P521_KEY	0x22
+#define CPACF_PCKMO_ENC_ECC_ED25519_KEY	0x28
+#define CPACF_PCKMO_ENC_ECC_ED448_KEY	0x29
 
 /*
  * Function codes for the PRNO (PERFORM RANDOM NUMBER OPERATION)
diff --git a/arch/s390/include/asm/os_info.h b/arch/s390/include/asm/os_info.h
index 0d1c74a7a650..a4d2e103f116 100644
--- a/arch/s390/include/asm/os_info.h
+++ b/arch/s390/include/asm/os_info.h
@@ -16,6 +16,9 @@
 
 #define OS_INFO_VMCOREINFO	0
 #define OS_INFO_REIPL_BLOCK	1
+#define OS_INFO_FLAGS_ENTRY	2
+
+#define OS_INFO_FLAG_REIPL_CLEAR	(1UL << 0)
 
 struct os_info_entry {
 	u64	addr;
@@ -30,8 +33,8 @@ struct os_info {
 	u16	version_minor;
 	u64	crashkernel_addr;
 	u64	crashkernel_size;
-	struct os_info_entry entry[2];
-	u8	reserved[4024];
+	struct os_info_entry entry[3];
+	u8	reserved[4004];
 } __packed;
 
 void os_info_init(void);
diff --git a/arch/s390/include/asm/pkey.h b/arch/s390/include/asm/pkey.h
index dd3d20c332ac..47d80a7451a6 100644
--- a/arch/s390/include/asm/pkey.h
+++ b/arch/s390/include/asm/pkey.h
@@ -2,7 +2,7 @@
 /*
  * Kernelspace interface to the pkey device driver
  *
- * Copyright IBM Corp. 2016,2019
+ * Copyright IBM Corp. 2016, 2023
  *
  * Author: Harald Freudenberger <freude@de.ibm.com>
  *
@@ -23,6 +23,6 @@
  * @return 0 on success, negative errno value on failure
  */
 int pkey_keyblob2pkey(const u8 *key, u32 keylen,
-		      struct pkey_protkey *protkey);
+		      u8 *protkey, u32 *protkeylen, u32 *protkeytype);
 
 #endif /* _KAPI_PKEY_H */
diff --git a/arch/s390/include/uapi/asm/pkey.h b/arch/s390/include/uapi/asm/pkey.h
index 924b876f992c..f7bae1c63bd6 100644
--- a/arch/s390/include/uapi/asm/pkey.h
+++ b/arch/s390/include/uapi/asm/pkey.h
@@ -2,7 +2,7 @@
 /*
  * Userspace interface to the pkey device driver
  *
- * Copyright IBM Corp. 2017, 2019
+ * Copyright IBM Corp. 2017, 2023
  *
  * Author: Harald Freudenberger <freude@de.ibm.com>
  *
@@ -32,10 +32,15 @@
 #define MINKEYBLOBSIZE	SECKEYBLOBSIZE
 
 /* defines for the type field within the pkey_protkey struct */
-#define PKEY_KEYTYPE_AES_128		      1
-#define PKEY_KEYTYPE_AES_192		      2
-#define PKEY_KEYTYPE_AES_256		      3
-#define PKEY_KEYTYPE_ECC		      4
+#define PKEY_KEYTYPE_AES_128		1
+#define PKEY_KEYTYPE_AES_192		2
+#define PKEY_KEYTYPE_AES_256		3
+#define PKEY_KEYTYPE_ECC		4
+#define PKEY_KEYTYPE_ECC_P256		5
+#define PKEY_KEYTYPE_ECC_P384		6
+#define PKEY_KEYTYPE_ECC_P521		7
+#define PKEY_KEYTYPE_ECC_ED25519	8
+#define PKEY_KEYTYPE_ECC_ED448		9
 
 /* the newer ioctls use a pkey_key_type enum for type information */
 enum pkey_key_type {
diff --git a/arch/s390/kernel/ipl.c b/arch/s390/kernel/ipl.c
index f44f70de9661..85a00d97a314 100644
--- a/arch/s390/kernel/ipl.c
+++ b/arch/s390/kernel/ipl.c
@@ -176,6 +176,8 @@ static bool reipl_fcp_clear;
 static bool reipl_ccw_clear;
 static bool reipl_eckd_clear;
 
+static unsigned long os_info_flags;
+
 static inline int __diag308(unsigned long subcode, unsigned long addr)
 {
 	union register_pair r1;
@@ -1938,6 +1940,20 @@ static void dump_reipl_run(struct shutdown_trigger *trigger)
 	struct lowcore *abs_lc;
 	unsigned int csum;
 
+	/*
+	 * Set REIPL_CLEAR flag in os_info flags entry indicating
+	 * 'clear' sysfs attribute has been set on the panicked system
+	 * for specified reipl type.
+	 * Always set for IPL_TYPE_NSS and IPL_TYPE_UNKNOWN.
+	 */
+	if ((reipl_type == IPL_TYPE_CCW && reipl_ccw_clear) ||
+	    (reipl_type == IPL_TYPE_ECKD && reipl_eckd_clear) ||
+	    (reipl_type == IPL_TYPE_FCP && reipl_fcp_clear) ||
+	    (reipl_type == IPL_TYPE_NVME && reipl_nvme_clear) ||
+	    reipl_type == IPL_TYPE_NSS ||
+	    reipl_type == IPL_TYPE_UNKNOWN)
+		os_info_flags |= OS_INFO_FLAG_REIPL_CLEAR;
+	os_info_entry_add(OS_INFO_FLAGS_ENTRY, &os_info_flags, sizeof(os_info_flags));
 	csum = (__force unsigned int)
 	       csum_partial(reipl_block_actual, reipl_block_actual->hdr.len, 0);
 	abs_lc = get_abs_lowcore();
diff --git a/arch/s390/kernel/module.c b/arch/s390/kernel/module.c
index f1b35dcdf3eb..42215f9404af 100644
--- a/arch/s390/kernel/module.c
+++ b/arch/s390/kernel/module.c
@@ -352,7 +352,8 @@ static int apply_rela(Elf_Rela *rela, Elf_Addr base, Elf_Sym *symtab,
 			rc = apply_rela_bits(loc, val, 0, 64, 0, write);
 		else if (r_type == R_390_GOTENT ||
 			 r_type == R_390_GOTPLTENT) {
-			val += (Elf_Addr) me->mem[MOD_TEXT].base - loc;
+			val += (Elf_Addr)me->mem[MOD_TEXT].base +
+				me->arch.got_offset - loc;
 			rc = apply_rela_bits(loc, val, 1, 32, 1, write);
 		}
 		break;
diff --git a/arch/s390/kernel/perf_cpum_cf.c b/arch/s390/kernel/perf_cpum_cf.c
index cf1b6e8a708d..90679143534b 100644
--- a/arch/s390/kernel/perf_cpum_cf.c
+++ b/arch/s390/kernel/perf_cpum_cf.c
@@ -76,6 +76,7 @@ static inline int ctr_stcctm(enum cpumf_ctr_set set, u64 range, u64 *dest)
 }
 
 struct cpu_cf_events {
+	refcount_t refcnt;		/* Reference count */
 	atomic_t		ctr_set[CPUMF_CTR_SET_MAX];
 	u64			state;		/* For perf_event_open SVC */
 	u64			dev_state;	/* For /dev/hwctr */
@@ -88,9 +89,6 @@ struct cpu_cf_events {
 	unsigned int sets;		/* # Counter set saved in memory */
 };
 
-/* Per-CPU event structure for the counter facility */
-static DEFINE_PER_CPU(struct cpu_cf_events, cpu_cf_events);
-
 static unsigned int cfdiag_cpu_speed;	/* CPU speed for CF_DIAG trailer */
 static debug_info_t *cf_dbg;
 
@@ -103,6 +101,221 @@ static debug_info_t *cf_dbg;
  */
 static struct cpumf_ctr_info	cpumf_ctr_info;
 
+struct cpu_cf_ptr {
+	struct cpu_cf_events *cpucf;
+};
+
+static struct cpu_cf_root {		/* Anchor to per CPU data */
+	refcount_t refcnt;		/* Overall active events */
+	struct cpu_cf_ptr __percpu *cfptr;
+} cpu_cf_root;
+
+/*
+ * Serialize event initialization and event removal. Both are called from
+ * user space in task context with perf_event_open() and close()
+ * system calls.
+ *
+ * This mutex serializes functions cpum_cf_alloc_cpu() called at event
+ * initialization via cpumf_pmu_event_init() and function cpum_cf_free_cpu()
+ * called at event removal via call back function hw_perf_event_destroy()
+ * when the event is deleted. They are serialized to enforce correct
+ * bookkeeping of pointer and reference counts anchored by
+ * struct cpu_cf_root and the access to cpu_cf_root::refcnt and the
+ * per CPU pointers stored in cpu_cf_root::cfptr.
+ */
+static DEFINE_MUTEX(pmc_reserve_mutex);
+
+/*
+ * Get pointer to per-cpu structure.
+ *
+ * Function get_cpu_cfhw() is called from
+ * - cfset_copy_all(): This function is protected by cpus_read_lock(), so
+ *   CPU hot plug remove can not happen. Event removal requires a close()
+ *   first.
+ *
+ * Function this_cpu_cfhw() is called from perf common code functions:
+ * - pmu_{en|dis}able(), pmu_{add|del}()and pmu_{start|stop}():
+ *   All functions execute with interrupts disabled on that particular CPU.
+ * - cfset_ioctl_{on|off}, cfset_cpu_read(): see comment cfset_copy_all().
+ *
+ * Therefore it is safe to access the CPU specific pointer to the event.
+ */
+static struct cpu_cf_events *get_cpu_cfhw(int cpu)
+{
+	struct cpu_cf_ptr __percpu *p = cpu_cf_root.cfptr;
+
+	if (p) {
+		struct cpu_cf_ptr *q = per_cpu_ptr(p, cpu);
+
+		return q->cpucf;
+	}
+	return NULL;
+}
+
+static struct cpu_cf_events *this_cpu_cfhw(void)
+{
+	return get_cpu_cfhw(smp_processor_id());
+}
+
+/* Disable counter sets on dedicated CPU */
+static void cpum_cf_reset_cpu(void *flags)
+{
+	lcctl(0);
+}
+
+/* Free per CPU data when the last event is removed. */
+static void cpum_cf_free_root(void)
+{
+	if (!refcount_dec_and_test(&cpu_cf_root.refcnt))
+		return;
+	free_percpu(cpu_cf_root.cfptr);
+	cpu_cf_root.cfptr = NULL;
+	irq_subclass_unregister(IRQ_SUBCLASS_MEASUREMENT_ALERT);
+	on_each_cpu(cpum_cf_reset_cpu, NULL, 1);
+	debug_sprintf_event(cf_dbg, 4, "%s2 root.refcnt %u cfptr %px\n",
+			    __func__, refcount_read(&cpu_cf_root.refcnt),
+			    cpu_cf_root.cfptr);
+}
+
+/*
+ * On initialization of first event also allocate per CPU data dynamically.
+ * Start with an array of pointers, the array size is the maximum number of
+ * CPUs possible, which might be larger than the number of CPUs currently
+ * online.
+ */
+static int cpum_cf_alloc_root(void)
+{
+	int rc = 0;
+
+	if (refcount_inc_not_zero(&cpu_cf_root.refcnt))
+		return rc;
+
+	/* The memory is already zeroed. */
+	cpu_cf_root.cfptr = alloc_percpu(struct cpu_cf_ptr);
+	if (cpu_cf_root.cfptr) {
+		refcount_set(&cpu_cf_root.refcnt, 1);
+		on_each_cpu(cpum_cf_reset_cpu, NULL, 1);
+		irq_subclass_register(IRQ_SUBCLASS_MEASUREMENT_ALERT);
+	} else {
+		rc = -ENOMEM;
+	}
+
+	return rc;
+}
+
+/* Free CPU counter data structure for a PMU */
+static void cpum_cf_free_cpu(int cpu)
+{
+	struct cpu_cf_events *cpuhw;
+	struct cpu_cf_ptr *p;
+
+	mutex_lock(&pmc_reserve_mutex);
+	/*
+	 * When invoked via CPU hotplug handler, there might be no events
+	 * installed or that particular CPU might not have an
+	 * event installed. This anchor pointer can be NULL!
+	 */
+	if (!cpu_cf_root.cfptr)
+		goto out;
+	p = per_cpu_ptr(cpu_cf_root.cfptr, cpu);
+	cpuhw = p->cpucf;
+	/*
+	 * Might be zero when called from CPU hotplug handler and no event
+	 * installed on that CPU, but on different CPUs.
+	 */
+	if (!cpuhw)
+		goto out;
+
+	if (refcount_dec_and_test(&cpuhw->refcnt)) {
+		kfree(cpuhw);
+		p->cpucf = NULL;
+	}
+	cpum_cf_free_root();
+out:
+	mutex_unlock(&pmc_reserve_mutex);
+}
+
+/* Allocate CPU counter data structure for a PMU. Called under mutex lock. */
+static int cpum_cf_alloc_cpu(int cpu)
+{
+	struct cpu_cf_events *cpuhw;
+	struct cpu_cf_ptr *p;
+	int rc;
+
+	mutex_lock(&pmc_reserve_mutex);
+	rc = cpum_cf_alloc_root();
+	if (rc)
+		goto unlock;
+	p = per_cpu_ptr(cpu_cf_root.cfptr, cpu);
+	cpuhw = p->cpucf;
+
+	if (!cpuhw) {
+		cpuhw = kzalloc(sizeof(*cpuhw), GFP_KERNEL);
+		if (cpuhw) {
+			p->cpucf = cpuhw;
+			refcount_set(&cpuhw->refcnt, 1);
+		} else {
+			rc = -ENOMEM;
+		}
+	} else {
+		refcount_inc(&cpuhw->refcnt);
+	}
+	if (rc) {
+		/*
+		 * Error in allocation of event, decrement anchor. Since
+		 * cpu_cf_event in not created, its destroy() function is not
+		 * invoked. Adjust the reference counter for the anchor.
+		 */
+		cpum_cf_free_root();
+	}
+unlock:
+	mutex_unlock(&pmc_reserve_mutex);
+	return rc;
+}
+
+/*
+ * Create/delete per CPU data structures for /dev/hwctr interface and events
+ * created by perf_event_open().
+ * If cpu is -1, track task on all available CPUs. This requires
+ * allocation of hardware data structures for all CPUs. This setup handles
+ * perf_event_open() with task context and /dev/hwctr interface.
+ * If cpu is non-zero install event on this CPU only. This setup handles
+ * perf_event_open() with CPU context.
+ */
+static int cpum_cf_alloc(int cpu)
+{
+	cpumask_var_t mask;
+	int rc;
+
+	if (cpu == -1) {
+		if (!zalloc_cpumask_var(&mask, GFP_KERNEL))
+			return -ENOMEM;
+		for_each_online_cpu(cpu) {
+			rc = cpum_cf_alloc_cpu(cpu);
+			if (rc) {
+				for_each_cpu(cpu, mask)
+					cpum_cf_free_cpu(cpu);
+				break;
+			}
+			cpumask_set_cpu(cpu, mask);
+		}
+		free_cpumask_var(mask);
+	} else {
+		rc = cpum_cf_alloc_cpu(cpu);
+	}
+	return rc;
+}
+
+static void cpum_cf_free(int cpu)
+{
+	if (cpu == -1) {
+		for_each_online_cpu(cpu)
+			cpum_cf_free_cpu(cpu);
+	} else {
+		cpum_cf_free_cpu(cpu);
+	}
+}
+
 #define	CF_DIAG_CTRSET_DEF		0xfeef	/* Counter set header mark */
 						/* interval in seconds */
 
@@ -451,10 +664,10 @@ static int validate_ctr_version(const u64 config, enum cpumf_ctr_set set)
  */
 static void cpumf_pmu_enable(struct pmu *pmu)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	int err;
 
-	if (cpuhw->flags & PMU_F_ENABLED)
+	if (!cpuhw || (cpuhw->flags & PMU_F_ENABLED))
 		return;
 
 	err = lcctl(cpuhw->state | cpuhw->dev_state);
@@ -471,11 +684,11 @@ static void cpumf_pmu_enable(struct pmu *pmu)
  */
 static void cpumf_pmu_disable(struct pmu *pmu)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
-	int err;
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	u64 inactive;
+	int err;
 
-	if (!(cpuhw->flags & PMU_F_ENABLED))
+	if (!cpuhw || !(cpuhw->flags & PMU_F_ENABLED))
 		return;
 
 	inactive = cpuhw->state & ~((1 << CPUMF_LCCTL_ENABLE_SHIFT) - 1);
@@ -487,58 +700,10 @@ static void cpumf_pmu_disable(struct pmu *pmu)
 		cpuhw->flags &= ~PMU_F_ENABLED;
 }
 
-#define PMC_INIT      0UL
-#define PMC_RELEASE   1UL
-
-static void cpum_cf_setup_cpu(void *flags)
-{
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
-
-	switch ((unsigned long)flags) {
-	case PMC_INIT:
-		cpuhw->flags |= PMU_F_RESERVED;
-		break;
-
-	case PMC_RELEASE:
-		cpuhw->flags &= ~PMU_F_RESERVED;
-		break;
-	}
-
-	/* Disable CPU counter sets */
-	lcctl(0);
-	debug_sprintf_event(cf_dbg, 5, "%s flags %#x flags %#x state %#llx\n",
-			    __func__, *(int *)flags, cpuhw->flags,
-			    cpuhw->state);
-}
-
-/* Initialize the CPU-measurement counter facility */
-static int __kernel_cpumcf_begin(void)
-{
-	on_each_cpu(cpum_cf_setup_cpu, (void *)PMC_INIT, 1);
-	irq_subclass_register(IRQ_SUBCLASS_MEASUREMENT_ALERT);
-
-	return 0;
-}
-
-/* Release the CPU-measurement counter facility */
-static void __kernel_cpumcf_end(void)
-{
-	on_each_cpu(cpum_cf_setup_cpu, (void *)PMC_RELEASE, 1);
-	irq_subclass_unregister(IRQ_SUBCLASS_MEASUREMENT_ALERT);
-}
-
-/* Number of perf events counting hardware events */
-static atomic_t num_events = ATOMIC_INIT(0);
-/* Used to avoid races in calling reserve/release_cpumf_hardware */
-static DEFINE_MUTEX(pmc_reserve_mutex);
-
 /* Release the PMU if event is the last perf event */
 static void hw_perf_event_destroy(struct perf_event *event)
 {
-	mutex_lock(&pmc_reserve_mutex);
-	if (atomic_dec_return(&num_events) == 0)
-		__kernel_cpumcf_end();
-	mutex_unlock(&pmc_reserve_mutex);
+	cpum_cf_free(event->cpu);
 }
 
 /* CPUMF <-> perf event mappings for kernel+userspace (basic set) */
@@ -562,14 +727,6 @@ static const int cpumf_generic_events_user[] = {
 	[PERF_COUNT_HW_BUS_CYCLES]	    = -1,
 };
 
-static void cpumf_hw_inuse(void)
-{
-	mutex_lock(&pmc_reserve_mutex);
-	if (atomic_inc_return(&num_events) == 1)
-		__kernel_cpumcf_begin();
-	mutex_unlock(&pmc_reserve_mutex);
-}
-
 static int is_userspace_event(u64 ev)
 {
 	return cpumf_generic_events_user[PERF_COUNT_HW_CPU_CYCLES] == ev ||
@@ -653,7 +810,8 @@ static int __hw_perf_event_init(struct perf_event *event, unsigned int type)
 	}
 
 	/* Initialize for using the CPU-measurement counter facility */
-	cpumf_hw_inuse();
+	if (cpum_cf_alloc(event->cpu))
+		return -ENOMEM;
 	event->destroy = hw_perf_event_destroy;
 
 	/*
@@ -756,7 +914,7 @@ static void cpumf_pmu_read(struct perf_event *event)
 
 static void cpumf_pmu_start(struct perf_event *event, int flags)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	struct hw_perf_event *hwc = &event->hw;
 	int i;
 
@@ -830,7 +988,7 @@ static int cfdiag_push_sample(struct perf_event *event,
 
 static void cpumf_pmu_stop(struct perf_event *event, int flags)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	struct hw_perf_event *hwc = &event->hw;
 	int i;
 
@@ -857,8 +1015,7 @@ static void cpumf_pmu_stop(struct perf_event *event, int flags)
 						      false);
 			if (cfdiag_diffctr(cpuhw, event->hw.config_base))
 				cfdiag_push_sample(event, cpuhw);
-		} else if (cpuhw->flags & PMU_F_RESERVED) {
-			/* Only update when PMU not hotplugged off */
+		} else {
 			hw_perf_event_update(event);
 		}
 		hwc->state |= PERF_HES_UPTODATE;
@@ -867,7 +1024,7 @@ static void cpumf_pmu_stop(struct perf_event *event, int flags)
 
 static int cpumf_pmu_add(struct perf_event *event, int flags)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 
 	ctr_set_enable(&cpuhw->state, event->hw.config_base);
 	event->hw.state = PERF_HES_UPTODATE | PERF_HES_STOPPED;
@@ -880,7 +1037,7 @@ static int cpumf_pmu_add(struct perf_event *event, int flags)
 
 static void cpumf_pmu_del(struct perf_event *event, int flags)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	int i;
 
 	cpumf_pmu_stop(event, PERF_EF_UPDATE);
@@ -912,29 +1069,83 @@ static struct pmu cpumf_pmu = {
 	.read	      = cpumf_pmu_read,
 };
 
-static int cpum_cf_setup(unsigned int cpu, unsigned long flags)
-{
-	local_irq_disable();
-	cpum_cf_setup_cpu((void *)flags);
-	local_irq_enable();
-	return 0;
-}
+static struct cfset_session {		/* CPUs and counter set bit mask */
+	struct list_head head;		/* Head of list of active processes */
+} cfset_session = {
+	.head = LIST_HEAD_INIT(cfset_session.head)
+};
+
+static refcount_t cfset_opencnt = REFCOUNT_INIT(0);	/* Access count */
+/*
+ * Synchronize access to device /dev/hwc. This mutex protects against
+ * concurrent access to functions cfset_open() and cfset_release().
+ * Same for CPU hotplug add and remove events triggering
+ * cpum_cf_online_cpu() and cpum_cf_offline_cpu().
+ * It also serializes concurrent device ioctl access from multiple
+ * processes accessing /dev/hwc.
+ *
+ * The mutex protects concurrent access to the /dev/hwctr session management
+ * struct cfset_session and reference counting variable cfset_opencnt.
+ */
+static DEFINE_MUTEX(cfset_ctrset_mutex);
 
+/*
+ * CPU hotplug handles only /dev/hwctr device.
+ * For perf_event_open() the CPU hotplug handling is done on kernel common
+ * code:
+ * - CPU add: Nothing is done since a file descriptor can not be created
+ *   and returned to the user.
+ * - CPU delete: Handled by common code via pmu_disable(), pmu_stop() and
+ *   pmu_delete(). The event itself is removed when the file descriptor is
+ *   closed.
+ */
 static int cfset_online_cpu(unsigned int cpu);
+
 static int cpum_cf_online_cpu(unsigned int cpu)
 {
-	debug_sprintf_event(cf_dbg, 4, "%s cpu %d in_irq %ld\n", __func__,
-			    cpu, in_interrupt());
-	cpum_cf_setup(cpu, PMC_INIT);
-	return cfset_online_cpu(cpu);
+	int rc = 0;
+
+	debug_sprintf_event(cf_dbg, 4, "%s cpu %d root.refcnt %d "
+			    "opencnt %d\n", __func__, cpu,
+			    refcount_read(&cpu_cf_root.refcnt),
+			    refcount_read(&cfset_opencnt));
+	/*
+	 * Ignore notification for perf_event_open().
+	 * Handle only /dev/hwctr device sessions.
+	 */
+	mutex_lock(&cfset_ctrset_mutex);
+	if (refcount_read(&cfset_opencnt)) {
+		rc = cpum_cf_alloc_cpu(cpu);
+		if (!rc)
+			cfset_online_cpu(cpu);
+	}
+	mutex_unlock(&cfset_ctrset_mutex);
+	return rc;
 }
 
 static int cfset_offline_cpu(unsigned int cpu);
+
 static int cpum_cf_offline_cpu(unsigned int cpu)
 {
-	debug_sprintf_event(cf_dbg, 4, "%s cpu %d\n", __func__, cpu);
-	cfset_offline_cpu(cpu);
-	return cpum_cf_setup(cpu, PMC_RELEASE);
+	debug_sprintf_event(cf_dbg, 4, "%s cpu %d root.refcnt %d opencnt %d\n",
+			    __func__, cpu, refcount_read(&cpu_cf_root.refcnt),
+			    refcount_read(&cfset_opencnt));
+	/*
+	 * During task exit processing of grouped perf events triggered by CPU
+	 * hotplug processing, pmu_disable() is called as part of perf context
+	 * removal process. Therefore do not trigger event removal now for
+	 * perf_event_open() created events. Perf common code triggers event
+	 * destruction when the event file descriptor is closed.
+	 *
+	 * Handle only /dev/hwctr device sessions.
+	 */
+	mutex_lock(&cfset_ctrset_mutex);
+	if (refcount_read(&cfset_opencnt)) {
+		cfset_offline_cpu(cpu);
+		cpum_cf_free_cpu(cpu);
+	}
+	mutex_unlock(&cfset_ctrset_mutex);
+	return 0;
 }
 
 /* Return true if store counter set multiple instruction is available */
@@ -953,13 +1164,13 @@ static void cpumf_measurement_alert(struct ext_code ext_code,
 		return;
 
 	inc_irq_stat(IRQEXT_CMC);
-	cpuhw = this_cpu_ptr(&cpu_cf_events);
 
 	/*
 	 * Measurement alerts are shared and might happen when the PMU
 	 * is not reserved.  Ignore these alerts in this case.
 	 */
-	if (!(cpuhw->flags & PMU_F_RESERVED))
+	cpuhw = this_cpu_cfhw();
+	if (!cpuhw)
 		return;
 
 	/* counter authorization change alert */
@@ -1039,19 +1250,11 @@ out1:
  * counter set via normal file operations.
  */
 
-static atomic_t cfset_opencnt = ATOMIC_INIT(0);		/* Access count */
-static DEFINE_MUTEX(cfset_ctrset_mutex);/* Synchronize access to hardware */
 struct cfset_call_on_cpu_parm {		/* Parm struct for smp_call_on_cpu */
 	unsigned int sets;		/* Counter set bit mask */
 	atomic_t cpus_ack;		/* # CPUs successfully executed func */
 };
 
-static struct cfset_session {		/* CPUs and counter set bit mask */
-	struct list_head head;		/* Head of list of active processes */
-} cfset_session = {
-	.head = LIST_HEAD_INIT(cfset_session.head)
-};
-
 struct cfset_request {			/* CPUs and counter set bit mask */
 	unsigned long ctrset;		/* Bit mask of counter set to read */
 	cpumask_t mask;			/* CPU mask to read from */
@@ -1113,11 +1316,11 @@ static void cfset_session_add(struct cfset_request *p)
 /* Stop all counter sets via ioctl interface */
 static void cfset_ioctl_off(void *parm)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	struct cfset_call_on_cpu_parm *p = parm;
 	int rc;
 
-	/* Check if any counter set used by /dev/hwc */
+	/* Check if any counter set used by /dev/hwctr */
 	for (rc = CPUMF_CTR_SET_BASIC; rc < CPUMF_CTR_SET_MAX; ++rc)
 		if ((p->sets & cpumf_ctr_ctl[rc])) {
 			if (!atomic_dec_return(&cpuhw->ctr_set[rc])) {
@@ -1141,7 +1344,7 @@ static void cfset_ioctl_off(void *parm)
 /* Start counter sets on particular CPU */
 static void cfset_ioctl_on(void *parm)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	struct cfset_call_on_cpu_parm *p = parm;
 	int rc;
 
@@ -1163,7 +1366,7 @@ static void cfset_ioctl_on(void *parm)
 
 static void cfset_release_cpu(void *p)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	int rc;
 
 	debug_sprintf_event(cf_dbg, 4, "%s state %#llx dev_state %#llx\n",
@@ -1203,27 +1406,41 @@ static int cfset_release(struct inode *inode, struct file *file)
 		kfree(file->private_data);
 		file->private_data = NULL;
 	}
-	if (!atomic_dec_return(&cfset_opencnt))
+	if (refcount_dec_and_test(&cfset_opencnt)) {	/* Last close */
 		on_each_cpu(cfset_release_cpu, NULL, 1);
+		cpum_cf_free(-1);
+	}
 	mutex_unlock(&cfset_ctrset_mutex);
-
-	hw_perf_event_destroy(NULL);
 	return 0;
 }
 
+/*
+ * Open via /dev/hwctr device. Allocate all per CPU resources on the first
+ * open of the device. The last close releases all per CPU resources.
+ * Parallel perf_event_open system calls also use per CPU resources.
+ * These invocations are handled via reference counting on the per CPU data
+ * structures.
+ */
 static int cfset_open(struct inode *inode, struct file *file)
 {
-	if (!capable(CAP_SYS_ADMIN))
+	int rc = 0;
+
+	if (!perfmon_capable())
 		return -EPERM;
+	file->private_data = NULL;
+
 	mutex_lock(&cfset_ctrset_mutex);
-	if (atomic_inc_return(&cfset_opencnt) == 1)
-		cfset_session_init();
+	if (!refcount_inc_not_zero(&cfset_opencnt)) {	/* First open */
+		rc = cpum_cf_alloc(-1);
+		if (!rc) {
+			cfset_session_init();
+			refcount_set(&cfset_opencnt, 1);
+		}
+	}
 	mutex_unlock(&cfset_ctrset_mutex);
 
-	cpumf_hw_inuse();
-	file->private_data = NULL;
 	/* nonseekable_open() never fails */
-	return nonseekable_open(inode, file);
+	return rc ?: nonseekable_open(inode, file);
 }
 
 static int cfset_all_start(struct cfset_request *req)
@@ -1280,7 +1497,7 @@ static int cfset_all_copy(unsigned long arg, cpumask_t *mask)
 	ctrset_read = (struct s390_ctrset_read __user *)arg;
 	uptr = ctrset_read->data;
 	for_each_cpu(cpu, mask) {
-		struct cpu_cf_events *cpuhw = per_cpu_ptr(&cpu_cf_events, cpu);
+		struct cpu_cf_events *cpuhw = get_cpu_cfhw(cpu);
 		struct s390_ctrset_cpudata __user *ctrset_cpudata;
 
 		ctrset_cpudata = uptr;
@@ -1324,7 +1541,7 @@ static size_t cfset_cpuset_read(struct s390_ctrset_setdata *p, int ctrset,
 /* Read all counter sets. */
 static void cfset_cpu_read(void *parm)
 {
-	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cpu_cf_events *cpuhw = this_cpu_cfhw();
 	struct cfset_call_on_cpu_parm *p = parm;
 	int set, set_size;
 	size_t space;
@@ -1348,9 +1565,9 @@ static void cfset_cpu_read(void *parm)
 			cpuhw->used += space;
 			cpuhw->sets += 1;
 		}
+		debug_sprintf_event(cf_dbg, 4, "%s sets %d used %zd\n", __func__,
+				    cpuhw->sets, cpuhw->used);
 	}
-	debug_sprintf_event(cf_dbg, 4, "%s sets %d used %zd\n", __func__,
-			    cpuhw->sets, cpuhw->used);
 }
 
 static int cfset_all_read(unsigned long arg, struct cfset_request *req)
@@ -1502,6 +1719,7 @@ static struct miscdevice cfset_dev = {
 	.name	= S390_HWCTR_DEVICE,
 	.minor	= MISC_DYNAMIC_MINOR,
 	.fops	= &cfset_fops,
+	.mode	= 0666,
 };
 
 /* Hotplug add of a CPU. Scan through all active processes and add
@@ -1512,7 +1730,6 @@ static int cfset_online_cpu(unsigned int cpu)
 	struct cfset_call_on_cpu_parm p;
 	struct cfset_request *rp;
 
-	mutex_lock(&cfset_ctrset_mutex);
 	if (!list_empty(&cfset_session.head)) {
 		list_for_each_entry(rp, &cfset_session.head, node) {
 			p.sets = rp->ctrset;
@@ -1520,19 +1737,18 @@ static int cfset_online_cpu(unsigned int cpu)
 			cpumask_set_cpu(cpu, &rp->mask);
 		}
 	}
-	mutex_unlock(&cfset_ctrset_mutex);
 	return 0;
 }
 
 /* Hotplug remove of a CPU. Scan through all active processes and clear
  * that CPU from the list of CPUs supplied with ioctl(..., START, ...).
+ * Adjust reference counts.
  */
 static int cfset_offline_cpu(unsigned int cpu)
 {
 	struct cfset_call_on_cpu_parm p;
 	struct cfset_request *rp;
 
-	mutex_lock(&cfset_ctrset_mutex);
 	if (!list_empty(&cfset_session.head)) {
 		list_for_each_entry(rp, &cfset_session.head, node) {
 			p.sets = rp->ctrset;
@@ -1540,7 +1756,6 @@ static int cfset_offline_cpu(unsigned int cpu)
 			cpumask_clear_cpu(cpu, &rp->mask);
 		}
 	}
-	mutex_unlock(&cfset_ctrset_mutex);
 	return 0;
 }
 
@@ -1618,7 +1833,8 @@ static int cfdiag_event_init(struct perf_event *event)
 	}
 
 	/* Initialize for using the CPU-measurement counter facility */
-	cpumf_hw_inuse();
+	if (cpum_cf_alloc(event->cpu))
+		return -ENOMEM;
 	event->destroy = hw_perf_event_destroy;
 
 	err = cfdiag_event_init2(event);
diff --git a/arch/s390/kernel/perf_pai_crypto.c b/arch/s390/kernel/perf_pai_crypto.c
index a7b339c4fd7c..fe7d1774ded1 100644
--- a/arch/s390/kernel/perf_pai_crypto.c
+++ b/arch/s390/kernel/perf_pai_crypto.c
@@ -36,7 +36,7 @@ struct paicrypt_map {
 	unsigned long *page;		/* Page for CPU to store counters */
 	struct pai_userdata *save;	/* Page to store no-zero counters */
 	unsigned int active_events;	/* # of PAI crypto users */
-	unsigned int refcnt;		/* Reference count mapped buffers */
+	refcount_t refcnt;		/* Reference count mapped buffers */
 	enum paievt_mode mode;		/* Type of event */
 	struct perf_event *event;	/* Perf event for sampling */
 };
@@ -57,10 +57,11 @@ static void paicrypt_event_destroy(struct perf_event *event)
 	static_branch_dec(&pai_key);
 	mutex_lock(&pai_reserve_mutex);
 	debug_sprintf_event(cfm_dbg, 5, "%s event %#llx cpu %d users %d"
-			    " mode %d refcnt %d\n", __func__,
+			    " mode %d refcnt %u\n", __func__,
 			    event->attr.config, event->cpu,
-			    cpump->active_events, cpump->mode, cpump->refcnt);
-	if (!--cpump->refcnt) {
+			    cpump->active_events, cpump->mode,
+			    refcount_read(&cpump->refcnt));
+	if (refcount_dec_and_test(&cpump->refcnt)) {
 		debug_sprintf_event(cfm_dbg, 4, "%s page %#lx save %p\n",
 				    __func__, (unsigned long)cpump->page,
 				    cpump->save);
@@ -149,8 +150,10 @@ static int paicrypt_busy(struct perf_event_attr *a, struct paicrypt_map *cpump)
 	/* Allocate memory for counter page and counter extraction.
 	 * Only the first counting event has to allocate a page.
 	 */
-	if (cpump->page)
+	if (cpump->page) {
+		refcount_inc(&cpump->refcnt);
 		goto unlock;
+	}
 
 	rc = -ENOMEM;
 	cpump->page = (unsigned long *)get_zeroed_page(GFP_KERNEL);
@@ -164,18 +167,18 @@ static int paicrypt_busy(struct perf_event_attr *a, struct paicrypt_map *cpump)
 		goto unlock;
 	}
 	rc = 0;
+	refcount_set(&cpump->refcnt, 1);
 
 unlock:
 	/* If rc is non-zero, do not set mode and reference count */
 	if (!rc) {
-		cpump->refcnt++;
 		cpump->mode = a->sample_period ? PAI_MODE_SAMPLING
 					       : PAI_MODE_COUNTING;
 	}
 	debug_sprintf_event(cfm_dbg, 5, "%s sample_period %#llx users %d"
-			    " mode %d refcnt %d page %#lx save %p rc %d\n",
+			    " mode %d refcnt %u page %#lx save %p rc %d\n",
 			    __func__, a->sample_period, cpump->active_events,
-			    cpump->mode, cpump->refcnt,
+			    cpump->mode, refcount_read(&cpump->refcnt),
 			    (unsigned long)cpump->page, cpump->save, rc);
 	mutex_unlock(&pai_reserve_mutex);
 	return rc;
diff --git a/arch/s390/kernel/perf_pai_ext.c b/arch/s390/kernel/perf_pai_ext.c
index fcea307d7529..3b4f384f77f7 100644
--- a/arch/s390/kernel/perf_pai_ext.c
+++ b/arch/s390/kernel/perf_pai_ext.c
@@ -50,7 +50,7 @@ struct paiext_map {
 	struct pai_userdata *save;	/* Area to store non-zero counters */
 	enum paievt_mode mode;		/* Type of event */
 	unsigned int active_events;	/* # of PAI Extension users */
-	unsigned int refcnt;
+	refcount_t refcnt;
 	struct perf_event *event;	/* Perf event for sampling */
 	struct paiext_cb *paiext_cb;	/* PAI extension control block area */
 };
@@ -60,14 +60,14 @@ struct paiext_mapptr {
 };
 
 static struct paiext_root {		/* Anchor to per CPU data */
-	int refcnt;			/* Overall active events */
+	refcount_t refcnt;		/* Overall active events */
 	struct paiext_mapptr __percpu *mapptr;
 } paiext_root;
 
 /* Free per CPU data when the last event is removed. */
 static void paiext_root_free(void)
 {
-	if (!--paiext_root.refcnt) {
+	if (refcount_dec_and_test(&paiext_root.refcnt)) {
 		free_percpu(paiext_root.mapptr);
 		paiext_root.mapptr = NULL;
 	}
@@ -80,7 +80,7 @@ static void paiext_root_free(void)
  */
 static int paiext_root_alloc(void)
 {
-	if (++paiext_root.refcnt == 1) {
+	if (!refcount_inc_not_zero(&paiext_root.refcnt)) {
 		/* The memory is already zeroed. */
 		paiext_root.mapptr = alloc_percpu(struct paiext_mapptr);
 		if (!paiext_root.mapptr) {
@@ -91,6 +91,7 @@ static int paiext_root_alloc(void)
 			 */
 			return -ENOMEM;
 		}
+		refcount_set(&paiext_root.refcnt, 1);
 	}
 	return 0;
 }
@@ -122,7 +123,7 @@ static void paiext_event_destroy(struct perf_event *event)
 
 	mutex_lock(&paiext_reserve_mutex);
 	cpump->event = NULL;
-	if (!--cpump->refcnt)		/* Last reference gone */
+	if (refcount_dec_and_test(&cpump->refcnt))	/* Last reference gone */
 		paiext_free(mp);
 	paiext_root_free();
 	mutex_unlock(&paiext_reserve_mutex);
@@ -163,7 +164,7 @@ static int paiext_alloc(struct perf_event_attr *a, struct perf_event *event)
 		rc = -ENOMEM;
 		cpump = kzalloc(sizeof(*cpump), GFP_KERNEL);
 		if (!cpump)
-			goto unlock;
+			goto undo;
 
 		/* Allocate memory for counter area and counter extraction.
 		 * These are
@@ -183,8 +184,9 @@ static int paiext_alloc(struct perf_event_attr *a, struct perf_event *event)
 					     GFP_KERNEL);
 		if (!cpump->save || !cpump->area || !cpump->paiext_cb) {
 			paiext_free(mp);
-			goto unlock;
+			goto undo;
 		}
+		refcount_set(&cpump->refcnt, 1);
 		cpump->mode = a->sample_period ? PAI_MODE_SAMPLING
 					       : PAI_MODE_COUNTING;
 	} else {
@@ -195,15 +197,15 @@ static int paiext_alloc(struct perf_event_attr *a, struct perf_event *event)
 		if (cpump->mode == PAI_MODE_SAMPLING ||
 		    (cpump->mode == PAI_MODE_COUNTING && a->sample_period)) {
 			rc = -EBUSY;
-			goto unlock;
+			goto undo;
 		}
+		refcount_inc(&cpump->refcnt);
 	}
 
 	rc = 0;
 	cpump->event = event;
-	++cpump->refcnt;
 
-unlock:
+undo:
 	if (rc) {
 		/* Error in allocation of event, decrement anchor. Since
 		 * the event in not created, its destroy() function is never
@@ -211,6 +213,7 @@ unlock:
 		 */
 		paiext_root_free();
 	}
+unlock:
 	mutex_unlock(&paiext_reserve_mutex);
 	/* If rc is non-zero, no increment of counter/sampler was done. */
 	return rc;
diff --git a/arch/s390/lib/Makefile b/arch/s390/lib/Makefile
index 580d2e3265cb..7c50eca85ca4 100644
--- a/arch/s390/lib/Makefile
+++ b/arch/s390/lib/Makefile
@@ -3,7 +3,7 @@
 # Makefile for s390-specific library files..
 #
 
-lib-y += delay.o string.o uaccess.o find.o spinlock.o
+lib-y += delay.o string.o uaccess.o find.o spinlock.o tishift.o
 obj-y += mem.o xor.o
 lib-$(CONFIG_KPROBES) += probes.o
 lib-$(CONFIG_UPROBES) += probes.o
diff --git a/arch/s390/lib/tishift.S b/arch/s390/lib/tishift.S
new file mode 100644
index 000000000000..de33cf02cfd2
--- /dev/null
+++ b/arch/s390/lib/tishift.S
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#include <linux/linkage.h>
+#include <asm/nospec-insn.h>
+#include <asm/export.h>
+
+	.section .noinstr.text, "ax"
+
+	GEN_BR_THUNK %r14
+
+SYM_FUNC_START(__ashlti3)
+	lmg	%r0,%r1,0(%r3)
+	cije	%r4,0,1f
+	lhi	%r3,64
+	sr	%r3,%r4
+	jnh	0f
+	srlg	%r3,%r1,0(%r3)
+	sllg	%r0,%r0,0(%r4)
+	sllg	%r1,%r1,0(%r4)
+	ogr	%r0,%r3
+	j	1f
+0:	sllg	%r0,%r1,-64(%r4)
+	lghi	%r1,0
+1:	stmg	%r0,%r1,0(%r2)
+	BR_EX	%r14
+SYM_FUNC_END(__ashlti3)
+EXPORT_SYMBOL(__ashlti3)
+
+SYM_FUNC_START(__ashrti3)
+	lmg	%r0,%r1,0(%r3)
+	cije	%r4,0,1f
+	lhi	%r3,64
+	sr	%r3,%r4
+	jnh	0f
+	sllg	%r3,%r0,0(%r3)
+	srlg	%r1,%r1,0(%r4)
+	srag	%r0,%r0,0(%r4)
+	ogr	%r1,%r3
+	j	1f
+0:	srag	%r1,%r0,-64(%r4)
+	srag	%r0,%r0,63
+1:	stmg	%r0,%r1,0(%r2)
+	BR_EX	%r14
+SYM_FUNC_END(__ashrti3)
+EXPORT_SYMBOL(__ashrti3)
+
+SYM_FUNC_START(__lshrti3)
+	lmg	%r0,%r1,0(%r3)
+	cije	%r4,0,1f
+	lhi	%r3,64
+	sr	%r3,%r4
+	jnh	0f
+	sllg	%r3,%r0,0(%r3)
+	srlg	%r1,%r1,0(%r4)
+	srlg	%r0,%r0,0(%r4)
+	ogr	%r1,%r3
+	j	1f
+0:	srlg	%r1,%r0,-64(%r4)
+	lghi	%r0,0
+1:	stmg	%r0,%r1,0(%r2)
+	BR_EX	%r14
+SYM_FUNC_END(__lshrti3)
+EXPORT_SYMBOL(__lshrti3)