12 files changed, 660 insertions, 81 deletions

diff --git a/arch/s390/include/asm/irq_work.h b/arch/s390/include/asm/irq_work.h
new file mode 100644
index 000000000000..603783766d0a
--- /dev/null
+++ b/arch/s390/include/asm/irq_work.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_S390_IRQ_WORK_H
+#define _ASM_S390_IRQ_WORK_H
+
+static inline bool arch_irq_work_has_interrupt(void)
+{
+	return true;
+}
+
+void arch_irq_work_raise(void);
+
+#endif /* _ASM_S390_IRQ_WORK_H */
diff --git a/arch/s390/include/asm/pgalloc.h b/arch/s390/include/asm/pgalloc.h
index d1297d6bbdcf..6b187cd72251 100644
--- a/arch/s390/include/asm/pgalloc.h
+++ b/arch/s390/include/asm/pgalloc.h
@@ -135,7 +135,7 @@ static inline void pmd_populate(struct mm_struct *mm,
 #define pmd_populate_kernel(mm, pmd, pte) pmd_populate(mm, pmd, pte)
 
 #define pmd_pgtable(pmd) \
-	(pgtable_t)(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE)
+	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
 
 /*
  * page table entry allocation/free routines.
diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index 794746a32806..29c7ecd5ad1d 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -1219,8 +1219,8 @@ static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 
-#define p4d_deref(pud) (p4d_val(pud) & _REGION_ENTRY_ORIGIN)
-#define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
+#define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
+#define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
 
 static inline unsigned long pmd_deref(pmd_t pmd)
 {
@@ -1229,12 +1229,12 @@ static inline unsigned long pmd_deref(pmd_t pmd)
 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
 	if (pmd_large(pmd))
 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
-	return pmd_val(pmd) & origin_mask;
+	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
 }
 
 static inline unsigned long pmd_pfn(pmd_t pmd)
 {
-	return pmd_deref(pmd) >> PAGE_SHIFT;
+	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
 }
 
 static inline unsigned long pud_deref(pud_t pud)
@@ -1244,12 +1244,12 @@ static inline unsigned long pud_deref(pud_t pud)
 	origin_mask = _REGION_ENTRY_ORIGIN;
 	if (pud_large(pud))
 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
-	return pud_val(pud) & origin_mask;
+	return (unsigned long)__va(pud_val(pud) & origin_mask);
 }
 
 static inline unsigned long pud_pfn(pud_t pud)
 {
-	return pud_deref(pud) >> PAGE_SHIFT;
+	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
 }
 
 /*
@@ -1329,7 +1329,7 @@ static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
 }
 #define gup_fast_permitted gup_fast_permitted
 
-#define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
+#define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
 #define pte_page(x) pfn_to_page(pte_pfn(x))
 
@@ -1636,7 +1636,7 @@ static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
 }
 #define pmdp_collapse_flush pmdp_collapse_flush
 
-#define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
+#define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
 
 static inline int pmd_trans_huge(pmd_t pmd)
diff --git a/arch/s390/include/uapi/asm/perf_cpum_cf_diag.h b/arch/s390/include/uapi/asm/perf_cpum_cf_diag.h
new file mode 100644
index 000000000000..3d8284b95f87
--- /dev/null
+++ b/arch/s390/include/uapi/asm/perf_cpum_cf_diag.h
@@ -0,0 +1,51 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/*
+ * Copyright IBM Corp. 2021
+ * Interface implementation for communication with the CPU Measurement
+ * counter facility device driver.
+ *
+ * Author(s): Thomas Richter <tmricht@linux.ibm.com>
+ *
+ * Define for ioctl() commands to communicate with the CPU Measurement
+ * counter facility device driver.
+ */
+
+#ifndef _PERF_CPUM_CF_DIAG_H
+#define _PERF_CPUM_CF_DIAG_H
+
+#include <linux/ioctl.h>
+#include <linux/types.h>
+
+#define S390_HWCTR_DEVICE		"hwctr"
+#define S390_HWCTR_START_VERSION	1
+
+struct s390_ctrset_start {		/* Set CPUs to operate on */
+	__u64 version;			/* Version of interface */
+	__u64 data_bytes;		/* # of bytes required */
+	__u64 cpumask_len;		/* Length of CPU mask in bytes */
+	__u64 *cpumask;			/* Pointer to CPU mask */
+	__u64 counter_sets;		/* Bit mask of counter sets to get */
+};
+
+struct s390_ctrset_setdata {		/* Counter set data */
+	__u32 set;			/* Counter set number */
+	__u32 no_cnts;			/* # of counters stored in cv[] */
+	__u64 cv[0];			/* Counter values (variable length) */
+};
+
+struct s390_ctrset_cpudata {		/* Counter set data per CPU */
+	__u32 cpu_nr;			/* CPU number */
+	__u32 no_sets;			/* # of counters sets in data[] */
+	struct s390_ctrset_setdata data[0];
+};
+
+struct s390_ctrset_read {		/* Structure to get all ctr sets */
+	__u64 no_cpus;			/* Total # of CPUs data taken from */
+	struct s390_ctrset_cpudata data[0];
+};
+
+#define S390_HWCTR_MAGIC	'C'	/* Random magic # for ioctls */
+#define	S390_HWCTR_START	_IOWR(S390_HWCTR_MAGIC, 1, struct s390_ctrset_start)
+#define	S390_HWCTR_STOP		_IO(S390_HWCTR_MAGIC, 2)
+#define	S390_HWCTR_READ		_IOWR(S390_HWCTR_MAGIC, 3, struct s390_ctrset_read)
+#endif
diff --git a/arch/s390/kernel/perf_cpum_cf_diag.c b/arch/s390/kernel/perf_cpum_cf_diag.c
index b5c86fb70d63..db4877bbb9aa 100644
--- a/arch/s390/kernel/perf_cpum_cf_diag.c
+++ b/arch/s390/kernel/perf_cpum_cf_diag.c
@@ -2,7 +2,7 @@
 /*
  * Performance event support for s390x - CPU-measurement Counter Sets
  *
- *  Copyright IBM Corp. 2019
+ *  Copyright IBM Corp. 2019, 2021
  *  Author(s): Hendrik Brueckner <brueckner@linux.ibm.com>
  *	       Thomas Richer <tmricht@linux.ibm.com>
  */
@@ -17,6 +17,8 @@
 #include <linux/export.h>
 #include <linux/slab.h>
 #include <linux/processor.h>
+#include <linux/miscdevice.h>
+#include <linux/mutex.h>
 
 #include <asm/ctl_reg.h>
 #include <asm/irq.h>
@@ -24,15 +26,20 @@
 #include <asm/timex.h>
 #include <asm/debug.h>
 
-#define	CF_DIAG_CTRSET_DEF		0xfeef	/* Counter set header mark */
+#include <asm/perf_cpum_cf_diag.h>
 
+#define	CF_DIAG_CTRSET_DEF		0xfeef	/* Counter set header mark */
+#define CF_DIAG_MIN_INTERVAL		60	/* Minimum counter set read */
+						/* interval in seconds */
+static unsigned long cf_diag_interval = CF_DIAG_MIN_INTERVAL;
 static unsigned int cf_diag_cpu_speed;
 static debug_info_t *cf_diag_dbg;
 
-struct cf_diag_csd {		/* Counter set data per CPU */
+struct cf_diag_csd {			/* Counter set data per CPU */
 	size_t used;			/* Bytes used in data/start */
 	unsigned char start[PAGE_SIZE];	/* Counter set at event start */
 	unsigned char data[PAGE_SIZE];	/* Counter set at event delete */
+	unsigned int sets;		/* # Counter set saved in data */
 };
 static DEFINE_PER_CPU(struct cf_diag_csd, cf_diag_csd);
 
@@ -178,18 +185,35 @@ static void cf_diag_disable(struct pmu *pmu)
 
 /* Number of perf events counting hardware events */
 static atomic_t cf_diag_events = ATOMIC_INIT(0);
+/* Used to avoid races in calling reserve/release_cpumf_hardware */
+static DEFINE_MUTEX(cf_diag_reserve_mutex);
 
 /* Release the PMU if event is the last perf event */
 static void cf_diag_perf_event_destroy(struct perf_event *event)
 {
 	debug_sprintf_event(cf_diag_dbg, 5,
 			    "%s event %p cpu %d cf_diag_events %d\n",
-			    __func__, event, event->cpu,
+			    __func__, event, smp_processor_id(),
 			    atomic_read(&cf_diag_events));
 	if (atomic_dec_return(&cf_diag_events) == 0)
 		__kernel_cpumcf_end();
 }
 
+static int get_authctrsets(void)
+{
+	struct cpu_cf_events *cpuhw;
+	unsigned long auth = 0;
+	enum cpumf_ctr_set i;
+
+	cpuhw = &get_cpu_var(cpu_cf_events);
+	for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i) {
+		if (cpuhw->info.auth_ctl & cpumf_ctr_ctl[i])
+			auth |= cpumf_ctr_ctl[i];
+	}
+	put_cpu_var(cpu_cf_events);
+	return auth;
+}
+
 /* Setup the event. Test for authorized counter sets and only include counter
  * sets which are authorized at the time of the setup. Including unauthorized
  * counter sets result in specification exception (and panic).
@@ -197,15 +221,12 @@ static void cf_diag_perf_event_destroy(struct perf_event *event)
 static int __hw_perf_event_init(struct perf_event *event)
 {
 	struct perf_event_attr *attr = &event->attr;
-	struct cpu_cf_events *cpuhw;
-	enum cpumf_ctr_set i;
 	int err = 0;
 
 	debug_sprintf_event(cf_diag_dbg, 5, "%s event %p cpu %d\n", __func__,
 			    event, event->cpu);
 
 	event->hw.config = attr->config;
-	event->hw.config_base = 0;
 
 	/* Add all authorized counter sets to config_base. The
 	 * the hardware init function is either called per-cpu or just once
@@ -215,11 +236,7 @@ static int __hw_perf_event_init(struct perf_event *event)
 	 * Checking the authorization on any CPU is fine as the hardware
 	 * applies the same authorization settings to all CPUs.
 	 */
-	cpuhw = &get_cpu_var(cpu_cf_events);
-	for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i)
-		if (cpuhw->info.auth_ctl & cpumf_ctr_ctl[i])
-			event->hw.config_base |= cpumf_ctr_ctl[i];
-	put_cpu_var(cpu_cf_events);
+	event->hw.config_base = get_authctrsets();
 
 	/* No authorized counter sets, nothing to count/sample */
 	if (!event->hw.config_base) {
@@ -237,6 +254,25 @@ out:
 	return err;
 }
 
+/* Return 0 if the CPU-measurement counter facility is currently free
+ * and an error otherwise.
+ */
+static int cf_diag_perf_event_inuse(void)
+{
+	int err = 0;
+
+	if (!atomic_inc_not_zero(&cf_diag_events)) {
+		mutex_lock(&cf_diag_reserve_mutex);
+		if (atomic_read(&cf_diag_events) == 0 &&
+		    __kernel_cpumcf_begin())
+			err = -EBUSY;
+		else
+			err = atomic_inc_return(&cf_diag_events);
+		mutex_unlock(&cf_diag_reserve_mutex);
+	}
+	return err;
+}
+
 static int cf_diag_event_init(struct perf_event *event)
 {
 	struct perf_event_attr *attr = &event->attr;
@@ -264,13 +300,9 @@ static int cf_diag_event_init(struct perf_event *event)
 	}
 
 	/* Initialize for using the CPU-measurement counter facility */
-	if (atomic_inc_return(&cf_diag_events) == 1) {
-		if (__kernel_cpumcf_begin()) {
-			atomic_dec(&cf_diag_events);
-			err = -EBUSY;
-			goto out;
-		}
-	}
+	err = cf_diag_perf_event_inuse();
+	if (err < 0)
+		goto out;
 	event->destroy = cf_diag_perf_event_destroy;
 
 	err = __hw_perf_event_init(event);
@@ -599,6 +631,8 @@ static void cf_diag_del(struct perf_event *event, int flags)
 	cpuhw->flags &= ~PMU_F_IN_USE;
 }
 
+/* Default counter set events and format attribute groups */
+
 CPUMF_EVENT_ATTR(CF_DIAG, CF_DIAG, PERF_EVENT_CPUM_CF_DIAG);
 
 static struct attribute *cf_diag_events_attr[] = {
@@ -663,6 +697,452 @@ static void cf_diag_get_cpu_speed(void)
 	}
 }
 
+/* Code to create device and file I/O operations */
+static atomic_t ctrset_opencnt = ATOMIC_INIT(0);	/* Excl. access */
+
+static int cf_diag_open(struct inode *inode, struct file *file)
+{
+	int err = 0;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+	if (atomic_xchg(&ctrset_opencnt, 1))
+		return -EBUSY;
+
+	/* Avoid concurrent access with perf_event_open() system call */
+	mutex_lock(&cf_diag_reserve_mutex);
+	if (atomic_read(&cf_diag_events) || __kernel_cpumcf_begin())
+		err = -EBUSY;
+	mutex_unlock(&cf_diag_reserve_mutex);
+	if (err) {
+		atomic_set(&ctrset_opencnt, 0);
+		return err;
+	}
+	file->private_data = NULL;
+	debug_sprintf_event(cf_diag_dbg, 2, "%s\n", __func__);
+	/* nonseekable_open() never fails */
+	return nonseekable_open(inode, file);
+}
+
+/* Variables for ioctl() interface support */
+static DEFINE_MUTEX(cf_diag_ctrset_mutex);
+static struct cf_diag_ctrset {
+	unsigned long ctrset;		/* Bit mask of counter set to read */
+	cpumask_t mask;			/* CPU mask to read from */
+	time64_t lastread;		/* Epoch counter set last read */
+} cf_diag_ctrset;
+
+static void cf_diag_ctrset_clear(void)
+{
+	cpumask_clear(&cf_diag_ctrset.mask);
+	cf_diag_ctrset.ctrset = 0;
+}
+
+static void cf_diag_release_cpu(void *p)
+{
+	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+
+	debug_sprintf_event(cf_diag_dbg, 3, "%s cpu %d\n", __func__,
+			    smp_processor_id());
+	lcctl(0);		/* Reset counter sets */
+	cpuhw->state = 0;	/* Save state in CPU hardware state */
+}
+
+/* Release function is also called when application gets terminated without
+ * doing a proper ioctl(..., S390_HWCTR_STOP, ...) command.
+ * Since only one application is allowed to open the device, simple stop all
+ * CPU counter sets.
+ */
+static int cf_diag_release(struct inode *inode, struct file *file)
+{
+	on_each_cpu(cf_diag_release_cpu, NULL, 1);
+	cf_diag_ctrset_clear();
+	atomic_set(&ctrset_opencnt, 0);
+	__kernel_cpumcf_end();
+	debug_sprintf_event(cf_diag_dbg, 2, "%s\n", __func__);
+	return 0;
+}
+
+struct cf_diag_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 int cf_diag_all_copy(unsigned long arg, cpumask_t *mask)
+{
+	struct s390_ctrset_read __user *ctrset_read;
+	unsigned int cpu, cpus, rc;
+	void __user *uptr;
+
+	ctrset_read = (struct s390_ctrset_read __user *)arg;
+	uptr = ctrset_read->data;
+	for_each_cpu(cpu, mask) {
+		struct cf_diag_csd *csd = per_cpu_ptr(&cf_diag_csd, cpu);
+		struct s390_ctrset_cpudata __user *ctrset_cpudata;
+
+		ctrset_cpudata = uptr;
+		debug_sprintf_event(cf_diag_dbg, 5, "%s cpu %d used %zd\n",
+				    __func__, cpu, csd->used);
+		rc  = put_user(cpu, &ctrset_cpudata->cpu_nr);
+		rc |= put_user(csd->sets, &ctrset_cpudata->no_sets);
+		rc |= copy_to_user(ctrset_cpudata->data, csd->data, csd->used);
+		if (rc)
+			return -EFAULT;
+		uptr += sizeof(struct s390_ctrset_cpudata) + csd->used;
+		cond_resched();
+	}
+	cpus = cpumask_weight(mask);
+	if (put_user(cpus, &ctrset_read->no_cpus))
+		return -EFAULT;
+	debug_sprintf_event(cf_diag_dbg, 5, "%s copied %ld\n",
+			    __func__, uptr - (void __user *)ctrset_read->data);
+	return 0;
+}
+
+static size_t cf_diag_cpuset_read(struct s390_ctrset_setdata *p, int ctrset,
+				  int ctrset_size, size_t room)
+{
+	size_t need = 0;
+	int rc = -1;
+
+	need = sizeof(*p) + sizeof(u64) * ctrset_size;
+	debug_sprintf_event(cf_diag_dbg, 5,
+			    "%s room %zd need %zd set %#x set_size %d\n",
+			    __func__, room, need, ctrset, ctrset_size);
+	if (need <= room) {
+		p->set = cpumf_ctr_ctl[ctrset];
+		p->no_cnts = ctrset_size;
+		rc = ctr_stcctm(ctrset, ctrset_size, (u64 *)p->cv);
+		if (rc == 3)		/* Nothing stored */
+			need = 0;
+	}
+	debug_sprintf_event(cf_diag_dbg, 5, "%s need %zd rc %d\n", __func__,
+			    need, rc);
+	return need;
+}
+
+/* Read all counter sets. Since the perf_event_open() system call with
+ * event cpum_cf_diag/.../ is blocked when this interface is active, reuse
+ * the perf_event_open() data buffer to store the counter sets.
+ */
+static void cf_diag_cpu_read(void *parm)
+{
+	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cf_diag_csd *csd = this_cpu_ptr(&cf_diag_csd);
+	struct cf_diag_call_on_cpu_parm *p = parm;
+	int set, set_size;
+	size_t space;
+
+	debug_sprintf_event(cf_diag_dbg, 5,
+			    "%s new %#x flags %#x state %#llx\n",
+			    __func__, p->sets, cpuhw->flags,
+			    cpuhw->state);
+	/* No data saved yet */
+	csd->used = 0;
+	csd->sets = 0;
+	memset(csd->data, 0, sizeof(csd->data));
+
+	/* Scan the counter sets */
+	for (set = CPUMF_CTR_SET_BASIC; set < CPUMF_CTR_SET_MAX; ++set) {
+		struct s390_ctrset_setdata *sp = (void *)csd->data + csd->used;
+
+		if (!(p->sets & cpumf_ctr_ctl[set]))
+			continue;	/* Counter set not in list */
+		set_size = cf_diag_ctrset_size(set, &cpuhw->info);
+		space = sizeof(csd->data) - csd->used;
+		space = cf_diag_cpuset_read(sp, set, set_size, space);
+		if (space) {
+			csd->used += space;
+			csd->sets += 1;
+		}
+		debug_sprintf_event(cf_diag_dbg, 5, "%s sp %px space %zd\n",
+				    __func__, sp, space);
+	}
+	debug_sprintf_event(cf_diag_dbg, 5, "%s sets %d used %zd\n", __func__,
+			    csd->sets, csd->used);
+}
+
+static int cf_diag_all_read(unsigned long arg)
+{
+	struct cf_diag_call_on_cpu_parm p;
+	cpumask_var_t mask;
+	time64_t now;
+	int rc = 0;
+
+	debug_sprintf_event(cf_diag_dbg, 5, "%s\n", __func__);
+	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+		return -ENOMEM;
+	now = ktime_get_seconds();
+	if (cf_diag_ctrset.lastread + cf_diag_interval > now) {
+		debug_sprintf_event(cf_diag_dbg, 5, "%s now %lld "
+				    " lastread %lld\n", __func__, now,
+				    cf_diag_ctrset.lastread);
+		rc = -EAGAIN;
+		goto out;
+	} else {
+		cf_diag_ctrset.lastread = now;
+	}
+	p.sets = cf_diag_ctrset.ctrset;
+	cpumask_and(mask, &cf_diag_ctrset.mask, cpu_online_mask);
+	on_each_cpu_mask(mask, cf_diag_cpu_read, &p, 1);
+	rc = cf_diag_all_copy(arg, mask);
+out:
+	free_cpumask_var(mask);
+	debug_sprintf_event(cf_diag_dbg, 5, "%s rc %d\n", __func__, rc);
+	return rc;
+}
+
+/* Stop all counter sets via ioctl interface */
+static void cf_diag_ioctl_off(void *parm)
+{
+	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cf_diag_call_on_cpu_parm *p = parm;
+	int rc;
+
+	debug_sprintf_event(cf_diag_dbg, 5,
+			    "%s new %#x flags %#x state %#llx\n",
+			    __func__, p->sets, cpuhw->flags,
+			    cpuhw->state);
+
+	ctr_set_multiple_disable(&cpuhw->state, p->sets);
+	ctr_set_multiple_stop(&cpuhw->state, p->sets);
+	rc = lcctl(cpuhw->state);		/* Stop counter sets */
+	if (!cpuhw->state)
+		cpuhw->flags &= ~PMU_F_IN_USE;
+	debug_sprintf_event(cf_diag_dbg, 5,
+			    "%s rc %d flags %#x state %#llx\n", __func__,
+			     rc, cpuhw->flags, cpuhw->state);
+}
+
+/* Start counter sets on particular CPU */
+static void cf_diag_ioctl_on(void *parm)
+{
+	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	struct cf_diag_call_on_cpu_parm *p = parm;
+	int rc;
+
+	debug_sprintf_event(cf_diag_dbg, 5,
+			    "%s new %#x flags %#x state %#llx\n",
+			    __func__, p->sets, cpuhw->flags,
+			    cpuhw->state);
+
+	if (!(cpuhw->flags & PMU_F_IN_USE))
+		cpuhw->state = 0;
+	cpuhw->flags |= PMU_F_IN_USE;
+	rc = lcctl(cpuhw->state);		/* Reset unused counter sets */
+	ctr_set_multiple_enable(&cpuhw->state, p->sets);
+	ctr_set_multiple_start(&cpuhw->state, p->sets);
+	rc |= lcctl(cpuhw->state);		/* Start counter sets */
+	if (!rc)
+		atomic_inc(&p->cpus_ack);
+	debug_sprintf_event(cf_diag_dbg, 5, "%s rc %d state %#llx\n",
+			    __func__, rc, cpuhw->state);
+}
+
+static int cf_diag_all_stop(void)
+{
+	struct cf_diag_call_on_cpu_parm p = {
+		.sets = cf_diag_ctrset.ctrset,
+	};
+	cpumask_var_t mask;
+
+	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+		return -ENOMEM;
+	cpumask_and(mask, &cf_diag_ctrset.mask, cpu_online_mask);
+	on_each_cpu_mask(mask, cf_diag_ioctl_off, &p, 1);
+	free_cpumask_var(mask);
+	return 0;
+}
+
+static int cf_diag_all_start(void)
+{
+	struct cf_diag_call_on_cpu_parm p = {
+		.sets = cf_diag_ctrset.ctrset,
+		.cpus_ack = ATOMIC_INIT(0),
+	};
+	cpumask_var_t mask;
+	int rc = 0;
+
+	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+		return -ENOMEM;
+	cpumask_and(mask, &cf_diag_ctrset.mask, cpu_online_mask);
+	on_each_cpu_mask(mask, cf_diag_ioctl_on, &p, 1);
+	if (atomic_read(&p.cpus_ack) != cpumask_weight(mask)) {
+		on_each_cpu_mask(mask, cf_diag_ioctl_off, &p, 1);
+		rc = -EIO;
+	}
+	free_cpumask_var(mask);
+	return rc;
+}
+
+/* Return the maximum required space for all possible CPUs in case one
+ * CPU will be onlined during the START, READ, STOP cycles.
+ * To find out the size of the counter sets, any one CPU will do. They
+ * all have the same counter sets.
+ */
+static size_t cf_diag_needspace(unsigned int sets)
+{
+	struct cpu_cf_events *cpuhw = this_cpu_ptr(&cpu_cf_events);
+	size_t bytes = 0;
+	int i;
+
+	for (i = CPUMF_CTR_SET_BASIC; i < CPUMF_CTR_SET_MAX; ++i) {
+		if (!(sets & cpumf_ctr_ctl[i]))
+			continue;
+		bytes += cf_diag_ctrset_size(i, &cpuhw->info) * sizeof(u64) +
+			 sizeof(((struct s390_ctrset_setdata *)0)->set) +
+			 sizeof(((struct s390_ctrset_setdata *)0)->no_cnts);
+	}
+	bytes = sizeof(((struct s390_ctrset_read *)0)->no_cpus) + nr_cpu_ids *
+		(bytes + sizeof(((struct s390_ctrset_cpudata *)0)->cpu_nr) +
+		     sizeof(((struct s390_ctrset_cpudata *)0)->no_sets));
+	debug_sprintf_event(cf_diag_dbg, 5, "%s bytes %ld\n", __func__,
+			    bytes);
+	return bytes;
+}
+
+static long cf_diag_ioctl_read(unsigned long arg)
+{
+	struct s390_ctrset_read read;
+	int ret = 0;
+
+	debug_sprintf_event(cf_diag_dbg, 5, "%s\n", __func__);
+	if (copy_from_user(&read, (char __user *)arg, sizeof(read)))
+		return -EFAULT;
+	ret = cf_diag_all_read(arg);
+	debug_sprintf_event(cf_diag_dbg, 5, "%s ret %d\n", __func__, ret);
+	return ret;
+}
+
+static long cf_diag_ioctl_stop(void)
+{
+	int ret;
+
+	debug_sprintf_event(cf_diag_dbg, 5, "%s\n", __func__);
+	ret = cf_diag_all_stop();
+	cf_diag_ctrset_clear();
+	debug_sprintf_event(cf_diag_dbg, 5, "%s ret %d\n", __func__, ret);
+	return ret;
+}
+
+static long cf_diag_ioctl_start(unsigned long arg)
+{
+	struct s390_ctrset_start __user *ustart;
+	struct s390_ctrset_start start;
+	void __user *umask;
+	unsigned int len;
+	int ret = 0;
+	size_t need;
+
+	if (cf_diag_ctrset.ctrset)
+		return -EBUSY;
+	ustart = (struct s390_ctrset_start __user *)arg;
+	if (copy_from_user(&start, ustart, sizeof(start)))
+		return -EFAULT;
+	if (start.version != S390_HWCTR_START_VERSION)
+		return -EINVAL;
+	if (start.counter_sets & ~(cpumf_ctr_ctl[CPUMF_CTR_SET_BASIC] |
+				   cpumf_ctr_ctl[CPUMF_CTR_SET_USER] |
+				   cpumf_ctr_ctl[CPUMF_CTR_SET_CRYPTO] |
+				   cpumf_ctr_ctl[CPUMF_CTR_SET_EXT] |
+				   cpumf_ctr_ctl[CPUMF_CTR_SET_MT_DIAG]))
+		return -EINVAL;		/* Invalid counter set */
+	if (!start.counter_sets)
+		return -EINVAL;		/* No counter set at all? */
+	cpumask_clear(&cf_diag_ctrset.mask);
+	len = min_t(u64, start.cpumask_len, cpumask_size());
+	umask = (void __user *)start.cpumask;
+	if (copy_from_user(&cf_diag_ctrset.mask, umask, len))
+		return -EFAULT;
+	if (cpumask_empty(&cf_diag_ctrset.mask))
+		return -EINVAL;
+	need = cf_diag_needspace(start.counter_sets);
+	if (put_user(need, &ustart->data_bytes))
+		ret = -EFAULT;
+	if (ret)
+		goto out;
+	cf_diag_ctrset.ctrset = start.counter_sets;
+	ret = cf_diag_all_start();
+out:
+	if (ret)
+		cf_diag_ctrset_clear();
+	debug_sprintf_event(cf_diag_dbg, 2, "%s sets %#lx need %ld ret %d\n",
+			    __func__, cf_diag_ctrset.ctrset, need, ret);
+	return ret;
+}
+
+static long cf_diag_ioctl(struct file *file, unsigned int cmd,
+			  unsigned long arg)
+{
+	int ret;
+
+	debug_sprintf_event(cf_diag_dbg, 2, "%s cmd %#x arg %lx\n", __func__,
+			    cmd, arg);
+	get_online_cpus();
+	mutex_lock(&cf_diag_ctrset_mutex);
+	switch (cmd) {
+	case S390_HWCTR_START:
+		ret = cf_diag_ioctl_start(arg);
+		break;
+	case S390_HWCTR_STOP:
+		ret = cf_diag_ioctl_stop();
+		break;
+	case S390_HWCTR_READ:
+		ret = cf_diag_ioctl_read(arg);
+		break;
+	default:
+		ret = -ENOTTY;
+		break;
+	}
+	mutex_unlock(&cf_diag_ctrset_mutex);
+	put_online_cpus();
+	debug_sprintf_event(cf_diag_dbg, 2, "%s ret %d\n", __func__, ret);
+	return ret;
+}
+
+static const struct file_operations cf_diag_fops = {
+	.owner = THIS_MODULE,
+	.open = cf_diag_open,
+	.release = cf_diag_release,
+	.unlocked_ioctl	= cf_diag_ioctl,
+	.compat_ioctl = cf_diag_ioctl,
+	.llseek = no_llseek
+};
+
+static struct miscdevice cf_diag_dev = {
+	.name	= S390_HWCTR_DEVICE,
+	.minor	= MISC_DYNAMIC_MINOR,
+	.fops	= &cf_diag_fops,
+};
+
+static int cf_diag_online_cpu(unsigned int cpu)
+{
+	struct cf_diag_call_on_cpu_parm p;
+
+	mutex_lock(&cf_diag_ctrset_mutex);
+	if (!cf_diag_ctrset.ctrset)
+		goto out;
+	p.sets = cf_diag_ctrset.ctrset;
+	cf_diag_ioctl_on(&p);
+out:
+	mutex_unlock(&cf_diag_ctrset_mutex);
+	return 0;
+}
+
+static int cf_diag_offline_cpu(unsigned int cpu)
+{
+	struct cf_diag_call_on_cpu_parm p;
+
+	mutex_lock(&cf_diag_ctrset_mutex);
+	if (!cf_diag_ctrset.ctrset)
+		goto out;
+	p.sets = cf_diag_ctrset.ctrset;
+	cf_diag_ioctl_off(&p);
+out:
+	mutex_unlock(&cf_diag_ctrset_mutex);
+	return 0;
+}
+
 /* Initialize the counter set PMU to generate complete counter set data as
  * event raw data. This relies on the CPU Measurement Counter Facility device
  * already being loaded and initialized.
@@ -685,21 +1165,43 @@ static int __init cf_diag_init(void)
 		return -ENOMEM;
 	}
 
+	rc = misc_register(&cf_diag_dev);
+	if (rc) {
+		pr_err("Registration of /dev/" S390_HWCTR_DEVICE
+		       "failed rc=%d\n", rc);
+		goto out;
+	}
+
 	/* Setup s390dbf facility */
 	cf_diag_dbg = debug_register(KMSG_COMPONENT, 2, 1, 128);
 	if (!cf_diag_dbg) {
 		pr_err("Registration of s390dbf(cpum_cf_diag) failed\n");
-		return -ENOMEM;
+		rc = -ENOMEM;
+		goto out_dbf;
 	}
 	debug_register_view(cf_diag_dbg, &debug_sprintf_view);
 
 	rc = perf_pmu_register(&cf_diag, "cpum_cf_diag", -1);
 	if (rc) {
-		debug_unregister_view(cf_diag_dbg, &debug_sprintf_view);
-		debug_unregister(cf_diag_dbg);
 		pr_err("Registration of PMU(cpum_cf_diag) failed with rc=%i\n",
 		       rc);
+		goto out_perf;
 	}
+	rc = cpuhp_setup_state_nocalls(CPUHP_AP_PERF_S390_CFD_ONLINE,
+				       "perf/s390/cfd:online",
+				       cf_diag_online_cpu, cf_diag_offline_cpu);
+	if (!rc)
+		goto out;
+
+	pr_err("Registration of CPUHP_AP_PERF_S390_CFD_ONLINE failed rc=%i\n",
+	       rc);
+	perf_pmu_unregister(&cf_diag);
+out_perf:
+	debug_unregister_view(cf_diag_dbg, &debug_sprintf_view);
+	debug_unregister(cf_diag_dbg);
+out_dbf:
+	misc_deregister(&cf_diag_dev);
+out:
 	return rc;
 }
-arch_initcall(cf_diag_init);
+device_initcall(cf_diag_init);
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index e299892440b6..58c8afa3da65 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -30,6 +30,7 @@
 #include <linux/delay.h>
 #include <linux/interrupt.h>
 #include <linux/irqflags.h>
+#include <linux/irq_work.h>
 #include <linux/cpu.h>
 #include <linux/slab.h>
 #include <linux/sched/hotplug.h>
@@ -62,6 +63,7 @@ enum {
 	ec_call_function_single,
 	ec_stop_cpu,
 	ec_mcck_pending,
+	ec_irq_work,
 };
 
 enum {
@@ -434,10 +436,12 @@ void notrace smp_yield_cpu(int cpu)
  */
 void notrace smp_emergency_stop(void)
 {
-	cpumask_t cpumask;
+	static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED;
+	static cpumask_t cpumask;
 	u64 end;
 	int cpu;
 
+	arch_spin_lock(&lock);
 	cpumask_copy(&cpumask, cpu_online_mask);
 	cpumask_clear_cpu(smp_processor_id(), &cpumask);
 
@@ -458,6 +462,7 @@ void notrace smp_emergency_stop(void)
 			break;
 		cpu_relax();
 	}
+	arch_spin_unlock(&lock);
 }
 NOKPROBE_SYMBOL(smp_emergency_stop);
 
@@ -505,6 +510,8 @@ static void smp_handle_ext_call(void)
 		generic_smp_call_function_single_interrupt();
 	if (test_bit(ec_mcck_pending, &bits))
 		__s390_handle_mcck();
+	if (test_bit(ec_irq_work, &bits))
+		irq_work_run();
 }
 
 static void do_ext_call_interrupt(struct ext_code ext_code,
@@ -537,6 +544,13 @@ void smp_send_reschedule(int cpu)
 	pcpu_ec_call(pcpu_devices + cpu, ec_schedule);
 }
 
+#ifdef CONFIG_IRQ_WORK
+void arch_irq_work_raise(void)
+{
+	pcpu_ec_call(pcpu_devices + smp_processor_id(), ec_irq_work);
+}
+#endif
+
 /*
  * parameter area for the set/clear control bit callbacks
  */
@@ -775,11 +789,13 @@ static int smp_add_core(struct sclp_core_entry *core, cpumask_t *avail,
 static int __smp_rescan_cpus(struct sclp_core_info *info, bool early)
 {
 	struct sclp_core_entry *core;
-	cpumask_t avail;
+	static cpumask_t avail;
 	bool configured;
 	u16 core_id;
 	int nr, i;
 
+	get_online_cpus();
+	mutex_lock(&smp_cpu_state_mutex);
 	nr = 0;
 	cpumask_xor(&avail, cpu_possible_mask, cpu_present_mask);
 	/*
@@ -800,6 +816,8 @@ static int __smp_rescan_cpus(struct sclp_core_info *info, bool early)
 		configured = i < info->configured;
 		nr += smp_add_core(&info->core[i], &avail, configured, early);
 	}
+	mutex_unlock(&smp_cpu_state_mutex);
+	put_online_cpus();
 	return nr;
 }
 
@@ -847,9 +865,7 @@ void __init smp_detect_cpus(void)
 	pr_info("%d configured CPUs, %d standby CPUs\n", c_cpus, s_cpus);
 
 	/* Add CPUs present at boot */
-	get_online_cpus();
 	__smp_rescan_cpus(info, true);
-	put_online_cpus();
 	memblock_free_early((unsigned long)info, sizeof(*info));
 }
 
@@ -1178,11 +1194,7 @@ int __ref smp_rescan_cpus(void)
 	if (!info)
 		return -ENOMEM;
 	smp_get_core_info(info, 0);
-	get_online_cpus();
-	mutex_lock(&smp_cpu_state_mutex);
 	nr = __smp_rescan_cpus(info, false);
-	mutex_unlock(&smp_cpu_state_mutex);
-	put_online_cpus();
 	kfree(info);
 	if (nr)
 		topology_schedule_update();
diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c
index ca47141a5be9..e7ce447651b9 100644
--- a/arch/s390/kernel/topology.c
+++ b/arch/s390/kernel/topology.c
@@ -62,16 +62,16 @@ static struct mask_info drawer_info;
 struct cpu_topology_s390 cpu_topology[NR_CPUS];
 EXPORT_SYMBOL_GPL(cpu_topology);
 
-static cpumask_t cpu_group_map(struct mask_info *info, unsigned int cpu)
+static void cpu_group_map(cpumask_t *dst, struct mask_info *info, unsigned int cpu)
 {
-	cpumask_t mask;
+	static cpumask_t mask;
 
 	cpumask_copy(&mask, cpumask_of(cpu));
 	switch (topology_mode) {
 	case TOPOLOGY_MODE_HW:
 		while (info) {
 			if (cpumask_test_cpu(cpu, &info->mask)) {
-				mask = info->mask;
+				cpumask_copy(&mask, &info->mask);
 				break;
 			}
 			info = info->next;
@@ -89,23 +89,24 @@ static cpumask_t cpu_group_map(struct mask_info *info, unsigned int cpu)
 		break;
 	}
 	cpumask_and(&mask, &mask, cpu_online_mask);
-	return mask;
+	cpumask_copy(dst, &mask);
 }
 
-static cpumask_t cpu_thread_map(unsigned int cpu)
+static void cpu_thread_map(cpumask_t *dst, unsigned int cpu)
 {
-	cpumask_t mask;
+	static cpumask_t mask;
 	int i;
 
 	cpumask_copy(&mask, cpumask_of(cpu));
 	if (topology_mode != TOPOLOGY_MODE_HW)
-		return mask;
+		goto out;
 	cpu -= cpu % (smp_cpu_mtid + 1);
 	for (i = 0; i <= smp_cpu_mtid; i++)
 		if (cpu_present(cpu + i))
 			cpumask_set_cpu(cpu + i, &mask);
 	cpumask_and(&mask, &mask, cpu_online_mask);
-	return mask;
+out:
+	cpumask_copy(dst, &mask);
 }
 
 #define TOPOLOGY_CORE_BITS	64
@@ -250,10 +251,10 @@ void update_cpu_masks(void)
 
 	for_each_possible_cpu(cpu) {
 		topo = &cpu_topology[cpu];
-		topo->thread_mask = cpu_thread_map(cpu);
-		topo->core_mask = cpu_group_map(&socket_info, cpu);
-		topo->book_mask = cpu_group_map(&book_info, cpu);
-		topo->drawer_mask = cpu_group_map(&drawer_info, cpu);
+		cpu_thread_map(&topo->thread_mask, cpu);
+		cpu_group_map(&topo->core_mask, &socket_info, cpu);
+		cpu_group_map(&topo->book_mask, &book_info, cpu);
+		cpu_group_map(&topo->drawer_mask, &drawer_info, cpu);
 		topo->booted_cores = 0;
 		if (topology_mode != TOPOLOGY_MODE_HW) {
 			id = topology_mode == TOPOLOGY_MODE_PACKAGE ? 0 : cpu;
diff --git a/arch/s390/mm/pgalloc.c b/arch/s390/mm/pgalloc.c
index 4e87c819ddea..781965f7210e 100644
--- a/arch/s390/mm/pgalloc.c
+++ b/arch/s390/mm/pgalloc.c
@@ -58,7 +58,7 @@ unsigned long *crst_table_alloc(struct mm_struct *mm)
 	if (!page)
 		return NULL;
 	arch_set_page_dat(page, 2);
-	return (unsigned long *) page_to_phys(page);
+	return (unsigned long *) page_to_virt(page);
 }
 
 void crst_table_free(struct mm_struct *mm, unsigned long *table)
@@ -161,7 +161,7 @@ struct page *page_table_alloc_pgste(struct mm_struct *mm)
 
 	page = alloc_page(GFP_KERNEL);
 	if (page) {
-		table = (u64 *)page_to_phys(page);
+		table = (u64 *)page_to_virt(page);
 		memset64(table, _PAGE_INVALID, PTRS_PER_PTE);
 		memset64(table + PTRS_PER_PTE, 0, PTRS_PER_PTE);
 	}
@@ -194,7 +194,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 			mask = atomic_read(&page->_refcount) >> 24;
 			mask = (mask | (mask >> 4)) & 3;
 			if (mask != 3) {
-				table = (unsigned long *) page_to_phys(page);
+				table = (unsigned long *) page_to_virt(page);
 				bit = mask & 1;		/* =1 -> second 2K */
 				if (bit)
 					table += PTRS_PER_PTE;
@@ -217,7 +217,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 	}
 	arch_set_page_dat(page, 0);
 	/* Initialize page table */
-	table = (unsigned long *) page_to_phys(page);
+	table = (unsigned long *) page_to_virt(page);
 	if (mm_alloc_pgste(mm)) {
 		/* Return 4K page table with PGSTEs */
 		atomic_xor_bits(&page->_refcount, 3 << 24);
@@ -239,10 +239,10 @@ void page_table_free(struct mm_struct *mm, unsigned long *table)
 	struct page *page;
 	unsigned int bit, mask;
 
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
+	page = virt_to_page(table);
 	if (!mm_alloc_pgste(mm)) {
 		/* Free 2K page table fragment of a 4K page */
-		bit = (__pa(table) & ~PAGE_MASK)/(PTRS_PER_PTE*sizeof(pte_t));
+		bit = ((unsigned long) table & ~PAGE_MASK)/(PTRS_PER_PTE*sizeof(pte_t));
 		spin_lock_bh(&mm->context.lock);
 		mask = atomic_xor_bits(&page->_refcount, 1U << (bit + 24));
 		mask >>= 24;
@@ -269,14 +269,14 @@ void page_table_free_rcu(struct mmu_gather *tlb, unsigned long *table,
 	unsigned int bit, mask;
 
 	mm = tlb->mm;
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
+	page = virt_to_page(table);
 	if (mm_alloc_pgste(mm)) {
 		gmap_unlink(mm, table, vmaddr);
-		table = (unsigned long *) (__pa(table) | 3);
+		table = (unsigned long *) ((unsigned long)table | 3);
 		tlb_remove_table(tlb, table);
 		return;
 	}
-	bit = (__pa(table) & ~PAGE_MASK) / (PTRS_PER_PTE*sizeof(pte_t));
+	bit = ((unsigned long) table & ~PAGE_MASK) / (PTRS_PER_PTE*sizeof(pte_t));
 	spin_lock_bh(&mm->context.lock);
 	mask = atomic_xor_bits(&page->_refcount, 0x11U << (bit + 24));
 	mask >>= 24;
@@ -285,7 +285,7 @@ void page_table_free_rcu(struct mmu_gather *tlb, unsigned long *table,
 	else
 		list_del(&page->lru);
 	spin_unlock_bh(&mm->context.lock);
-	table = (unsigned long *) (__pa(table) | (1U << bit));
+	table = (unsigned long *) ((unsigned long) table | (1U << bit));
 	tlb_remove_table(tlb, table);
 }
 
@@ -293,7 +293,7 @@ void __tlb_remove_table(void *_table)
 {
 	unsigned int mask = (unsigned long) _table & 3;
 	void *table = (void *)((unsigned long) _table ^ mask);
-	struct page *page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
+	struct page *page = virt_to_page(table);
 
 	switch (mask) {
 	case 0:		/* pmd, pud, or p4d */
diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c
index 82dbf9450105..96897fab89dc 100644
--- a/arch/s390/mm/vmem.c
+++ b/arch/s390/mm/vmem.c
@@ -27,14 +27,14 @@ static void __ref *vmem_alloc_pages(unsigned int order)
 
 	if (slab_is_available())
 		return (void *)__get_free_pages(GFP_KERNEL, order);
-	return (void *) memblock_phys_alloc(size, size);
+	return memblock_alloc(size, size);
 }
 
 static void vmem_free_pages(unsigned long addr, int order)
 {
 	/* We don't expect boot memory to be removed ever. */
 	if (!slab_is_available() ||
-	    WARN_ON_ONCE(PageReserved(phys_to_page(addr))))
+	    WARN_ON_ONCE(PageReserved(virt_to_page(addr))))
 		return;
 	free_pages(addr, order);
 }
@@ -57,7 +57,7 @@ pte_t __ref *vmem_pte_alloc(void)
 	if (slab_is_available())
 		pte = (pte_t *) page_table_alloc(&init_mm);
 	else
-		pte = (pte_t *) memblock_phys_alloc(size, size);
+		pte = (pte_t *) memblock_alloc(size, size);
 	if (!pte)
 		return NULL;
 	memset64((u64 *)pte, _PAGE_INVALID, PTRS_PER_PTE);
@@ -85,7 +85,7 @@ static void vmemmap_flush_unused_sub_pmd(void)
 {
 	if (!unused_sub_pmd_start)
 		return;
-	memset(__va(unused_sub_pmd_start), PAGE_UNUSED,
+	memset((void *)unused_sub_pmd_start, PAGE_UNUSED,
 	       ALIGN(unused_sub_pmd_start, PMD_SIZE) - unused_sub_pmd_start);
 	unused_sub_pmd_start = 0;
 }
@@ -98,7 +98,7 @@ static void vmemmap_mark_sub_pmd_used(unsigned long start, unsigned long end)
 	 * getting removed (just in case the memmap never gets initialized,
 	 * e.g., because the memory block never gets onlined).
 	 */
-	memset(__va(start), 0, sizeof(struct page));
+	memset((void *)start, 0, sizeof(struct page));
 }
 
 static void vmemmap_use_sub_pmd(unsigned long start, unsigned long end)
@@ -119,7 +119,7 @@ static void vmemmap_use_sub_pmd(unsigned long start, unsigned long end)
 
 static void vmemmap_use_new_sub_pmd(unsigned long start, unsigned long end)
 {
-	void *page = __va(ALIGN_DOWN(start, PMD_SIZE));
+	unsigned long page = ALIGN_DOWN(start, PMD_SIZE);
 
 	vmemmap_flush_unused_sub_pmd();
 
@@ -128,7 +128,7 @@ static void vmemmap_use_new_sub_pmd(unsigned long start, unsigned long end)
 
 	/* Mark the unused parts of the new memmap page PAGE_UNUSED. */
 	if (!IS_ALIGNED(start, PMD_SIZE))
-		memset(page, PAGE_UNUSED, start - __pa(page));
+		memset((void *)page, PAGE_UNUSED, start - page);
 	/*
 	 * We want to avoid memset(PAGE_UNUSED) when populating the vmemmap of
 	 * consecutive sections. Remember for the last added PMD the last
@@ -141,11 +141,11 @@ static void vmemmap_use_new_sub_pmd(unsigned long start, unsigned long end)
 /* Returns true if the PMD is completely unused and can be freed. */
 static bool vmemmap_unuse_sub_pmd(unsigned long start, unsigned long end)
 {
-	void *page = __va(ALIGN_DOWN(start, PMD_SIZE));
+	unsigned long page = ALIGN_DOWN(start, PMD_SIZE);
 
 	vmemmap_flush_unused_sub_pmd();
-	memset(__va(start), PAGE_UNUSED, end - start);
-	return !memchr_inv(page, PAGE_UNUSED, PMD_SIZE);
+	memset((void *)start, PAGE_UNUSED, end - start);
+	return !memchr_inv((void *)page, PAGE_UNUSED, PMD_SIZE);
 }
 
 /* __ref: we'll only call vmemmap_alloc_block() via vmemmap_populate() */
@@ -166,7 +166,7 @@ static int __ref modify_pte_table(pmd_t *pmd, unsigned long addr,
 			if (pte_none(*pte))
 				continue;
 			if (!direct)
-				vmem_free_pages(pfn_to_phys(pte_pfn(*pte)), 0);
+				vmem_free_pages((unsigned long) pfn_to_virt(pte_pfn(*pte)), 0);
 			pte_clear(&init_mm, addr, pte);
 		} else if (pte_none(*pte)) {
 			if (!direct) {
@@ -176,7 +176,7 @@ static int __ref modify_pte_table(pmd_t *pmd, unsigned long addr,
 					goto out;
 				pte_val(*pte) = __pa(new_page) | prot;
 			} else {
-				pte_val(*pte) = addr | prot;
+				pte_val(*pte) = __pa(addr) | prot;
 			}
 		} else {
 			continue;
@@ -201,7 +201,7 @@ static void try_free_pte_table(pmd_t *pmd, unsigned long start)
 		if (!pte_none(*pte))
 			return;
 	}
-	vmem_pte_free(__va(pmd_deref(*pmd)));
+	vmem_pte_free((unsigned long *) pmd_deref(*pmd));
 	pmd_clear(pmd);
 }
 
@@ -242,7 +242,7 @@ static int __ref modify_pmd_table(pud_t *pud, unsigned long addr,
 			    IS_ALIGNED(next, PMD_SIZE) &&
 			    MACHINE_HAS_EDAT1 && addr && direct &&
 			    !debug_pagealloc_enabled()) {
-				pmd_val(*pmd) = addr | prot;
+				pmd_val(*pmd) = __pa(addr) | prot;
 				pages++;
 				continue;
 			} else if (!direct && MACHINE_HAS_EDAT1) {
@@ -338,7 +338,7 @@ static int modify_pud_table(p4d_t *p4d, unsigned long addr, unsigned long end,
 			    IS_ALIGNED(next, PUD_SIZE) &&
 			    MACHINE_HAS_EDAT2 && addr && direct &&
 			    !debug_pagealloc_enabled()) {
-				pud_val(*pud) = addr | prot;
+				pud_val(*pud) = __pa(addr) | prot;
 				pages++;
 				continue;
 			}
diff --git a/arch/s390/tools/opcodes.txt b/arch/s390/tools/opcodes.txt
index 46d8ed96cf06..0e207c46e8da 100644
--- a/arch/s390/tools/opcodes.txt
+++ b/arch/s390/tools/opcodes.txt
@@ -597,7 +597,7 @@ b9b3	cu42	RRE_RR
 b9bd	trtre	RRF_U0RR
 b9be	srstu	RRE_RR
 b9bf	trte	RRF_U0RR
-b9c0	selhhhr	RRF_RURR
+b9c0	selfhr	RRF_RURR
 b9c8	ahhhr	RRF_R0RR2
 b9c9	shhhr	RRF_R0RR2
 b9ca	alhhhr	RRF_R0RR2
diff --git a/drivers/s390/virtio/virtio_ccw.c b/drivers/s390/virtio/virtio_ccw.c
index 5730572b52cd..54e686dca6de 100644
--- a/drivers/s390/virtio/virtio_ccw.c
+++ b/drivers/s390/virtio/virtio_ccw.c
@@ -117,7 +117,7 @@ struct virtio_rev_info {
 };
 
 /* the highest virtio-ccw revision we support */
-#define VIRTIO_CCW_REV_MAX 1
+#define VIRTIO_CCW_REV_MAX 2
 
 struct virtio_ccw_vq_info {
 	struct virtqueue *vq;
@@ -952,7 +952,7 @@ static u8 virtio_ccw_get_status(struct virtio_device *vdev)
 	u8 old_status = vcdev->dma_area->status;
 	struct ccw1 *ccw;
 
-	if (vcdev->revision < 1)
+	if (vcdev->revision < 2)
 		return vcdev->dma_area->status;
 
 	ccw = ccw_device_dma_zalloc(vcdev->cdev, sizeof(*ccw));
diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index ee09a39627d6..9129ba231423 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -168,6 +168,7 @@ enum cpuhp_state {
 	CPUHP_AP_PERF_X86_CQM_ONLINE,
 	CPUHP_AP_PERF_X86_CSTATE_ONLINE,
 	CPUHP_AP_PERF_S390_CF_ONLINE,
+	CPUHP_AP_PERF_S390_CFD_ONLINE,
 	CPUHP_AP_PERF_S390_SF_ONLINE,
 	CPUHP_AP_PERF_ARM_CCI_ONLINE,
 	CPUHP_AP_PERF_ARM_CCN_ONLINE,