s390: add support for runtime instrumentation

Allow user-space threads to use runtime instrumentation (RI). To enable RI
for a thread there is a new s390 specific system call, sys_s390_runtime_instr,
that takes as parameter a realtime signal number. If the RI facility is
available the system call sets up a control block for the calling thread with
the appropriate permissions for the thread to modify the control block.

The user-space thread can then use the store and modify RI instructions to
alter the control block and start/stop the instrumentation via RION/RIOFF.

If the user specified program buffer runs full RI triggers an external
interrupt. The external interrupt is translated to a real-time signal that
is delivered to the thread that enabled RI on that CPU. The number of
the real-time signal is the number specified in the RI system call. So,
user-space can select any available real-time signal number in case the
application itself uses real-time signals for other purposes.

The kernel saves the RI control blocks on task switch only if the running
thread was enabled for RI. Therefore, the performance impact on task switch
should be negligible if RI is not used.

RI is only enabled for user-space mode and is disabled for the supervisor
state.

Reviewed-by: Heiko Carstens <[e-mail]>
Signed-off-by: Jan Glauber <[e-mail]>
Signed-off-by: Martin Schwidefsky <[e-mail]>

arch/s390/include/asm/cpu_mf.h

···
 #define CPU_MF_INT_SF_LSDA	(1 << 22)	/* loss of sample data alert */
 #define CPU_MF_INT_CF_CACA	(1 <<  7)	/* counter auth. change alert */
 #define CPU_MF_INT_CF_LCDA	(1 <<  6)	/* loss of counter data alert */
+#define CPU_MF_INT_RI_HALTED	(1 <<  5)	/* run-time instr. halted */
+#define CPU_MF_INT_RI_BUF_FULL	(1 <<  4)	/* run-time instr. program
+						   buffer full */
 
 #define CPU_MF_INT_CF_MASK	(CPU_MF_INT_CF_CACA|CPU_MF_INT_CF_LCDA)
 #define CPU_MF_INT_SF_MASK	(CPU_MF_INT_SF_IAE|CPU_MF_INT_SF_ISE|	\
 				 CPU_MF_INT_SF_PRA|CPU_MF_INT_SF_SACA|	\
 				 CPU_MF_INT_SF_LSDA)
+#define CPU_MF_INT_RI_MASK	(CPU_MF_INT_RI_HALTED|CPU_MF_INT_RI_BUF_FULL)
 
 /* CPU measurement facility support */
 static inline int cpum_cf_avail(void)

arch/s390/include/asm/irq.h

···
 	EXTINT_IUC,
 	EXTINT_CMS,
 	EXTINT_CMC,
+	EXTINT_CMR,
 	IOINT_CIO,
 	IOINT_QAI,
 	IOINT_DAS,

arch/s390/include/asm/processor.h

···
 #include <asm/page.h>
 #include <asm/ptrace.h>
 #include <asm/setup.h>
+#include <asm/runtime_instr.h>
 
 /*
  * Default implementation of macro that returns current
···
         /* pfault_wait is used to block the process on a pfault event */
 	unsigned long pfault_wait;
 	struct list_head list;
+	/* cpu runtime instrumentation */
+	struct runtime_instr_cb *ri_cb;
+	int ri_signum;
 };
 
 typedef struct thread_struct thread_struct;

arch/s390/include/asm/ptrace.h

···
 #define PSW_MASK_ASC		0x0000C000UL
 #define PSW_MASK_CC		0x00003000UL
 #define PSW_MASK_PM		0x00000F00UL
+#define PSW_MASK_RI		0x00000000UL
 #define PSW_MASK_EA		0x00000000UL
 #define PSW_MASK_BA		0x00000000UL
 
···
 #define PSW_MASK_ASC		0x0000C00000000000UL
 #define PSW_MASK_CC		0x0000300000000000UL
 #define PSW_MASK_PM		0x00000F0000000000UL
+#define PSW_MASK_RI		0x0000008000000000UL
 #define PSW_MASK_EA		0x0000000100000000UL
 #define PSW_MASK_BA		0x0000000080000000UL
 
-#define PSW_MASK_USER		0x00003F0180000000UL
+#define PSW_MASK_USER		0x00003F8180000000UL
 
 #define PSW_ADDR_AMODE		0x0000000000000000UL
 #define PSW_ADDR_INSN		0xFFFFFFFFFFFFFFFFUL

arch/s390/include/asm/runtime_instr.h

···
+#ifndef _RUNTIME_INSTR_H
+#define _RUNTIME_INSTR_H
+
+#define S390_RUNTIME_INSTR_START	0x1
+#define S390_RUNTIME_INSTR_STOP		0x2
+
+struct runtime_instr_cb {
+	__u64 buf_current;
+	__u64 buf_origin;
+	__u64 buf_limit;
+
+	__u32 valid		: 1;
+	__u32 pstate		: 1;
+	__u32 pstate_set_buf	: 1;
+	__u32 home_space	: 1;
+	__u32 altered		: 1;
+	__u32			: 3;
+	__u32 pstate_sample	: 1;
+	__u32 sstate_sample	: 1;
+	__u32 pstate_collect	: 1;
+	__u32 sstate_collect	: 1;
+	__u32			: 1;
+	__u32 halted_int	: 1;
+	__u32 int_requested	: 1;
+	__u32 buffer_full_int	: 1;
+	__u32 key		: 4;
+	__u32			: 9;
+	__u32 rgs		: 3;
+
+	__u32 mode		: 4;
+	__u32 next		: 1;
+	__u32 mae		: 1;
+	__u32			: 2;
+	__u32 call_type_br	: 1;
+	__u32 return_type_br	: 1;
+	__u32 other_type_br	: 1;
+	__u32 bc_other_type	: 1;
+	__u32 emit		: 1;
+	__u32 tx_abort		: 1;
+	__u32			: 2;
+	__u32 bp_xn		: 1;
+	__u32 bp_xt		: 1;
+	__u32 bp_ti		: 1;
+	__u32 bp_ni		: 1;
+	__u32 suppr_y		: 1;
+	__u32 suppr_z		: 1;
+
+	__u32 dc_miss_extra	: 1;
+	__u32 lat_lev_ignore	: 1;
+	__u32 ic_lat_lev	: 4;
+	__u32 dc_lat_lev	: 4;
+
+	__u64 reserved1;
+	__u64 scaling_factor;
+	__u64 rsic;
+	__u64 reserved2;
+} __packed __aligned(8);
+
+extern struct runtime_instr_cb runtime_instr_empty_cb;
+
+static inline void load_runtime_instr_cb(struct runtime_instr_cb *cb)
+{
+	asm volatile(".insn	rsy,0xeb0000000060,0,0,%0"	/* LRIC */
+		: : "Q" (*cb));
+}
+
+static inline void store_runtime_instr_cb(struct runtime_instr_cb *cb)
+{
+	asm volatile(".insn	rsy,0xeb0000000061,0,0,%0"	/* STRIC */
+		: "=Q" (*cb) : : "cc");
+}
+
+static inline void save_ri_cb(struct runtime_instr_cb *cb_prev)
+{
+#ifdef CONFIG_64BIT
+	if (cb_prev)
+		store_runtime_instr_cb(cb_prev);
+#endif
+}
+
+static inline void restore_ri_cb(struct runtime_instr_cb *cb_next,
+				 struct runtime_instr_cb *cb_prev)
+{
+#ifdef CONFIG_64BIT
+	if (cb_next)
+		load_runtime_instr_cb(cb_next);
+	else if (cb_prev)
+		load_runtime_instr_cb(&runtime_instr_empty_cb);
+#endif
+}
+
+#ifdef CONFIG_64BIT
+extern void exit_thread_runtime_instr(void);
+#else
+static inline void exit_thread_runtime_instr(void) { }
+#endif
+
+#endif /* _RUNTIME_INSTR_H */

arch/s390/include/asm/switch_to.h

···
 	if (prev->mm) {							\
 		save_fp_regs(&prev->thread.fp_regs);			\
 		save_access_regs(&prev->thread.acrs[0]);		\
+		save_ri_cb(prev->thread.ri_cb);				\
 	}								\
 	if (next->mm) {							\
 		restore_fp_regs(&next->thread.fp_regs);			\
 		restore_access_regs(&next->thread.acrs[0]);		\
+		restore_ri_cb(next->thread.ri_cb, prev->thread.ri_cb);	\
 		update_per_regs(next);					\
 	}								\
 	prev = __switch_to(prev,next);					\

arch/s390/include/asm/unistd.h

···
 #define __NR_setns		339
 #define __NR_process_vm_readv	340
 #define __NR_process_vm_writev	341
-#define NR_syscalls 342
+#define __NR_s390_runtime_instr 342
+#define NR_syscalls 343
 
 /* 
  * There are some system calls that are not present on 64 bit, some

arch/s390/kernel/Makefile

···
 obj-$(CONFIG_FTRACE_SYSCALLS)  += ftrace.o
 obj-$(CONFIG_CRASH_DUMP)	+= crash_dump.o
 obj-$(CONFIG_PERF_EVENTS)	+= perf_event.o perf_cpum_cf.o
-obj-$(CONFIG_64BIT)		+= cache.o
+obj-$(CONFIG_64BIT)		+= runtime_instr.o cache.o
 
 # Kexec part
 S390_KEXEC_OBJS := machine_kexec.o

arch/s390/kernel/compat_wrapper.S

···
 	llgf	%r0,164(%r15)		# unsigned long
 	stg	%r0,160(%r15)
 	jg	compat_sys_process_vm_writev
+
+ENTRY(sys_s390_runtime_instr_wrapper)
+	lgfr	%r2,%r2			# int
+	lgfr	%r3,%r3			# int
+	jg	sys_s390_runtime_instr

arch/s390/kernel/dis.c

···
 	LONG_INSN_POPCNT,
 	LONG_INSN_RISBHG,
 	LONG_INSN_RISBLG,
+	LONG_INSN_RINEXT,
+	LONG_INSN_RIEMIT,
 };
 
 static char *long_insn_name[] = {
···
 	[LONG_INSN_POPCNT] = "popcnt",
 	[LONG_INSN_RISBHG] = "risbhg",
 	[LONG_INSN_RISBLG] = "risblk",
+	[LONG_INSN_RINEXT] = "rinext",
+	[LONG_INSN_RIEMIT] = "riemit",
 };
 
 static struct insn opcode[] = {
···
 	{ "ahi", 0x0a, INSTR_RI_RI },
 	{ "mhi", 0x0c, INSTR_RI_RI },
 	{ "chi", 0x0e, INSTR_RI_RI },
+	{ "", 0, INSTR_INVALID }
+};
+
+static struct insn opcode_aa[] = {
+#ifdef CONFIG_64BIT
+	{ { 0, LONG_INSN_RINEXT }, 0x00, INSTR_RI_RI },
+	{ "rion", 0x01, INSTR_RI_RI },
+	{ "tric", 0x02, INSTR_RI_RI },
+	{ "rioff", 0x03, INSTR_RI_RI },
+	{ { 0, LONG_INSN_RIEMIT }, 0x04, INSTR_RI_RI },
+#endif
 	{ "", 0, INSTR_INVALID }
 };
 
···
 	{ "cliy", 0x55, INSTR_SIY_URD },
 	{ "oiy", 0x56, INSTR_SIY_URD },
 	{ "xiy", 0x57, INSTR_SIY_URD },
+	{ "lric", 0x60, INSTR_RSY_RDRM },
+	{ "stric", 0x61, INSTR_RSY_RDRM },
+	{ "mric", 0x62, INSTR_RSY_RDRM },
 	{ "icmh", 0x80, INSTR_RSE_RURD },
 	{ "icmh", 0x80, INSTR_RSY_RURD },
 	{ "icmy", 0x81, INSTR_RSY_RURD },
···
 		break;
 	case 0xa7:
 		table = opcode_a7;
+		break;
+	case 0xaa:
+		table = opcode_aa;
 		break;
 	case 0xb2:
 		table = opcode_b2;

arch/s390/kernel/irq.c

···
 	{.name = "IUC", .desc = "[EXT] IUCV" },
 	{.name = "CMS", .desc = "[EXT] CPU-Measurement: Sampling" },
 	{.name = "CMC", .desc = "[EXT] CPU-Measurement: Counter" },
+	{.name = "CMR", .desc = "[EXT] CPU-Measurement: RI" },
 	{.name = "CIO", .desc = "[I/O] Common I/O Layer Interrupt" },
 	{.name = "QAI", .desc = "[I/O] QDIO Adapter Interrupt" },
 	{.name = "DAS", .desc = "[I/O] DASD" },

arch/s390/kernel/process.c

···
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/switch_to.h>
+#include <asm/runtime_instr.h>
 #include "entry.h"
 
 asmlinkage void ret_from_fork(void) asm ("ret_from_fork");
···
  */
 void exit_thread(void)
 {
+	exit_thread_runtime_instr();
 }
 
 void flush_thread(void)
···
 
 	/* Save access registers to new thread structure. */
 	save_access_regs(&p->thread.acrs[0]);
+
+	/* Don't copy runtime instrumentation info */
+	p->thread.ri_cb = NULL;
+	p->thread.ri_signum = 0;
+	frame->childregs.psw.mask &= ~PSW_MASK_RI;
 
 #ifndef CONFIG_64BIT
 	/*

arch/s390/kernel/runtime_instr.c

···
+/*
+ * Copyright IBM Corp. 2012
+ * Author(s): Jan Glauber <[e-mail]>
+ */
+
+#include <linux/kernel.h>
+#include <linux/syscalls.h>
+#include <linux/signal.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/kernel_stat.h>
+#include <asm/runtime_instr.h>
+#include <asm/cpu_mf.h>
+#include <asm/irq.h>
+
+/* empty control block to disable RI by loading it */
+struct runtime_instr_cb runtime_instr_empty_cb;
+
+static int runtime_instr_avail(void)
+{
+	return test_facility(64);
+}
+
+static void disable_runtime_instr(void)
+{
+	struct pt_regs *regs = task_pt_regs(current);
+
+	load_runtime_instr_cb(&runtime_instr_empty_cb);
+
+	/*
+	 * Make sure the RI bit is deleted from the PSW. If the user did not
+	 * switch off RI before the system call the process will get a
+	 * specification exception otherwise.
+	 */
+	regs->psw.mask &= ~PSW_MASK_RI;
+}
+
+static void init_runtime_instr_cb(struct runtime_instr_cb *cb)
+{
+	cb->buf_limit = 0xfff;
+	if (addressing_mode == HOME_SPACE_MODE)
+		cb->home_space = 1;
+	cb->int_requested = 1;
+	cb->pstate = 1;
+	cb->pstate_set_buf = 1;
+	cb->pstate_sample = 1;
+	cb->pstate_collect = 1;
+	cb->key = PAGE_DEFAULT_KEY;
+	cb->valid = 1;
+}
+
+void exit_thread_runtime_instr(void)
+{
+	struct task_struct *task = current;
+
+	if (!task->thread.ri_cb)
+		return;
+	disable_runtime_instr();
+	kfree(task->thread.ri_cb);
+	task->thread.ri_signum = 0;
+	task->thread.ri_cb = NULL;
+}
+
+static void runtime_instr_int_handler(struct ext_code ext_code,
+				unsigned int param32, unsigned long param64)
+{
+	struct siginfo info;
+
+	if (!(param32 & CPU_MF_INT_RI_MASK))
+		return;
+
+	kstat_cpu(smp_processor_id()).irqs[EXTINT_CMR]++;
+
+	if (!current->thread.ri_cb)
+		return;
+	if (current->thread.ri_signum < SIGRTMIN ||
+	    current->thread.ri_signum > SIGRTMAX) {
+		WARN_ON_ONCE(1);
+		return;
+	}
+
+	memset(&info, 0, sizeof(info));
+	info.si_signo = current->thread.ri_signum;
+	info.si_code = SI_QUEUE;
+	if (param32 & CPU_MF_INT_RI_BUF_FULL)
+		info.si_int = ENOBUFS;
+	else if (param32 & CPU_MF_INT_RI_HALTED)
+		info.si_int = ECANCELED;
+	else
+		return; /* unknown reason */
+
+	send_sig_info(current->thread.ri_signum, &info, current);
+}
+
+SYSCALL_DEFINE2(s390_runtime_instr, int, command, int, signum)
+{
+	struct runtime_instr_cb *cb;
+
+	if (!runtime_instr_avail())
+		return -EOPNOTSUPP;
+
+	if (command == S390_RUNTIME_INSTR_STOP) {
+		preempt_disable();
+		exit_thread_runtime_instr();
+		preempt_enable();
+		return 0;
+	}
+
+	if (command != S390_RUNTIME_INSTR_START ||
+	    (signum < SIGRTMIN || signum > SIGRTMAX))
+		return -EINVAL;
+
+	if (!current->thread.ri_cb) {
+		cb = kzalloc(sizeof(*cb), GFP_KERNEL);
+		if (!cb)
+			return -ENOMEM;
+	} else {
+		cb = current->thread.ri_cb;
+		memset(cb, 0, sizeof(*cb));
+	}
+
+	init_runtime_instr_cb(cb);
+	current->thread.ri_signum = signum;
+
+	/* now load the control block to make it available */
+	preempt_disable();
+	current->thread.ri_cb = cb;
+	load_runtime_instr_cb(cb);
+	preempt_enable();
+	return 0;
+}
+
+static int __init runtime_instr_init(void)
+{
+	int rc;
+
+	if (!runtime_instr_avail())
+		return 0;
+
+	measurement_alert_subclass_register();
+	rc = register_external_interrupt(0x1407, runtime_instr_int_handler);
+	if (rc)
+		measurement_alert_subclass_unregister();
+	else
+		pr_info("Runtime instrumentation facility initialized\n");
+	return rc;
+}
+device_initcall(runtime_instr_init);

arch/s390/kernel/syscalls.S

···
 SYSCALL(sys_setns,sys_setns,sys_setns_wrapper)
 SYSCALL(sys_process_vm_readv,sys_process_vm_readv,compat_sys_process_vm_readv_wrapper) /* 340 */
 SYSCALL(sys_process_vm_writev,sys_process_vm_writev,compat_sys_process_vm_writev_wrapper)
+SYSCALL(sys_ni_syscall,sys_s390_runtime_instr,sys_s390_runtime_instr_wrapper)