Implement syscall clock_getcpuclockid2, so we can get a clock id

for process, thread or others we want to support.
Use the syscall to implement POSIX API clock_getcpuclock and
pthread_getcpuclockid.

PR:	168417

15 files changed, 225 insertions, 17 deletions

diff --git a/sys/compat/freebsd32/freebsd32_syscall.h b/sys/compat/freebsd32/freebsd32_syscall.h
index b1684a1..240f1ca 100644
--- a/sys/compat/freebsd32/freebsd32_syscall.h
+++ b/sys/compat/freebsd32/freebsd32_syscall.h
@@ -210,6 +210,7 @@
 #define	FREEBSD32_SYS_ffclock_getcounter	241
 #define	FREEBSD32_SYS_ffclock_setestimate	242
 #define	FREEBSD32_SYS_ffclock_getestimate	243
+#define	FREEBSD32_SYS_clock_getcpuclockid2	247
 #define	FREEBSD32_SYS_minherit	250
 #define	FREEBSD32_SYS_rfork	251
 #define	FREEBSD32_SYS_openbsd_poll	252
diff --git a/sys/compat/freebsd32/freebsd32_syscalls.c b/sys/compat/freebsd32/freebsd32_syscalls.c
index 18ade01..75e4d22 100644
--- a/sys/compat/freebsd32/freebsd32_syscalls.c
+++ b/sys/compat/freebsd32/freebsd32_syscalls.c
@@ -257,7 +257,7 @@ const char *freebsd32_syscallnames[] = {
 	"#244",			/* 244 = nosys */
 	"#245",			/* 245 = nosys */
 	"#246",			/* 246 = nosys */
-	"#247",			/* 247 = nosys */
+	"clock_getcpuclockid2",			/* 247 = clock_getcpuclockid2 */
 	"#248",			/* 248 = ntp_gettime */
 	"#249",			/* 249 = nosys */
 	"minherit",			/* 250 = minherit */
diff --git a/sys/compat/freebsd32/freebsd32_sysent.c b/sys/compat/freebsd32/freebsd32_sysent.c
index f657cc6..334c520 100644
--- a/sys/compat/freebsd32/freebsd32_sysent.c
+++ b/sys/compat/freebsd32/freebsd32_sysent.c
@@ -294,7 +294,7 @@ struct sysent freebsd32_sysent[] = {
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 244 = nosys */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 245 = nosys */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 246 = nosys */
-	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 247 = nosys */
+	{ AS(clock_getcpuclockid2_args), (sy_call_t *)sys_clock_getcpuclockid2, AUE_NULL, NULL, 0, 0, 0, SY_THR_STATIC },	/* 247 = clock_getcpuclockid2 */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 248 = ntp_gettime */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 249 = nosys */
 	{ AS(minherit_args), (sy_call_t *)sys_minherit, AUE_MINHERIT, NULL, 0, 0, 0, SY_THR_STATIC },	/* 250 = minherit */
diff --git a/sys/compat/freebsd32/freebsd32_systrace_args.c b/sys/compat/freebsd32/freebsd32_systrace_args.c
index c7a43d0..e5588d0 100644
--- a/sys/compat/freebsd32/freebsd32_systrace_args.c
+++ b/sys/compat/freebsd32/freebsd32_systrace_args.c
@@ -1224,6 +1224,15 @@ systrace_args(int sysnum, void *params, uint64_t *uarg, int *n_args)
 		*n_args = 1;
 		break;
 	}
+	/* clock_getcpuclockid2 */
+	case 247: {
+		struct clock_getcpuclockid2_args *p = params;
+		iarg[0] = p->id; /* id_t */
+		iarg[1] = p->which; /* int */
+		uarg[2] = (intptr_t) p->clock_id; /* clockid_t * */
+		*n_args = 3;
+		break;
+	}
 	/* minherit */
 	case 250: {
 		struct minherit_args *p = params;
@@ -4991,6 +5000,22 @@ systrace_entry_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
 			break;
 		};
 		break;
+	/* clock_getcpuclockid2 */
+	case 247:
+		switch(ndx) {
+		case 0:
+			p = "id_t";
+			break;
+		case 1:
+			p = "int";
+			break;
+		case 2:
+			p = "clockid_t *";
+			break;
+		default:
+			break;
+		};
+		break;
 	/* minherit */
 	case 250:
 		switch(ndx) {
@@ -8890,6 +8915,11 @@ systrace_return_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
 		if (ndx == 0 || ndx == 1)
 			p = "int";
 		break;
+	/* clock_getcpuclockid2 */
+	case 247:
+		if (ndx == 0 || ndx == 1)
+			p = "int";
+		break;
 	/* minherit */
 	case 250:
 		if (ndx == 0 || ndx == 1)
diff --git a/sys/compat/freebsd32/syscalls.master b/sys/compat/freebsd32/syscalls.master
index 97f09d0..478558a 100644
--- a/sys/compat/freebsd32/syscalls.master
+++ b/sys/compat/freebsd32/syscalls.master
@@ -457,7 +457,8 @@
 244	AUE_NULL	UNIMPL	nosys
 245	AUE_NULL	UNIMPL	nosys
 246	AUE_NULL	UNIMPL	nosys
-247	AUE_NULL	UNIMPL	nosys
+247	AUE_NULL	NOPROTO	{ int clock_getcpuclockid2(id_t id,\
+					int which, clockid_t *clock_id); }
 248	AUE_NULL	UNIMPL	ntp_gettime
 249	AUE_NULL	UNIMPL	nosys
 ; syscall numbers initially used in OpenBSD
diff --git a/sys/kern/init_sysent.c b/sys/kern/init_sysent.c
index 4a4328a..721143d 100644
--- a/sys/kern/init_sysent.c
+++ b/sys/kern/init_sysent.c
@@ -281,7 +281,7 @@ struct sysent sysent[] = {
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 244 = nosys */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 245 = nosys */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 246 = nosys */
-	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 247 = nosys */
+	{ AS(clock_getcpuclockid2_args), (sy_call_t *)sys_clock_getcpuclockid2, AUE_NULL, NULL, 0, 0, 0, SY_THR_STATIC },	/* 247 = clock_getcpuclockid2 */
 	{ AS(ntp_gettime_args), (sy_call_t *)sys_ntp_gettime, AUE_NULL, NULL, 0, 0, SYF_CAPENABLED, SY_THR_STATIC },	/* 248 = ntp_gettime */
 	{ 0, (sy_call_t *)nosys, AUE_NULL, NULL, 0, 0, 0, SY_THR_ABSENT },			/* 249 = nosys */
 	{ AS(minherit_args), (sy_call_t *)sys_minherit, AUE_MINHERIT, NULL, 0, 0, SYF_CAPENABLED, SY_THR_STATIC },	/* 250 = minherit */
diff --git a/sys/kern/kern_time.c b/sys/kern/kern_time.c
index 4ba6117..c0e7831 100644
--- a/sys/kern/kern_time.c
+++ b/sys/kern/kern_time.c
@@ -58,6 +58,12 @@ __FBSDID("$FreeBSD$");
 #include <vm/vm_extern.h>
 
 #define MAX_CLOCKS 	(CLOCK_MONOTONIC+1)
+#define CPUCLOCK_BIT		0x80000000
+#define CPUCLOCK_PROCESS_BIT	0x40000000
+#define CPUCLOCK_ID_MASK	(~(CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT))
+#define MAKE_THREAD_CPUCLOCK(tid)	(CPUCLOCK_BIT|(tid))
+#define MAKE_PROCESS_CPUCLOCK(pid)	\
+	(CPUCLOCK_BIT|CPUCLOCK_PROCESS_BIT|(pid))
 
 static struct kclock	posix_clocks[MAX_CLOCKS];
 static uma_zone_t	itimer_zone = NULL;
@@ -165,6 +171,52 @@ settime(struct thread *td, struct timeval *tv)
 }
 
 #ifndef _SYS_SYSPROTO_H_
+struct clock_getcpuclockid2_args {
+	id_t id;
+	int which,
+	clockid_t *clock_id;
+};
+#endif
+/* ARGSUSED */
+int
+sys_clock_getcpuclockid2(struct thread *td, struct clock_getcpuclockid2_args *uap)
+{
+	clockid_t clk_id;
+	struct proc *p;
+	pid_t pid;
+	lwpid_t tid;
+	int error;
+
+	switch(uap->which) {
+	case CPUCLOCK_WHICH_PID:
+		if (uap->id != 0) {
+			p = pfind(uap->id);
+			if (p == NULL)
+				return (ESRCH);
+			error = p_cansee(td, p);
+			PROC_UNLOCK(p);
+			if (error)
+				return (error);
+			pid = uap->id;
+		} else {
+			pid = td->td_proc->p_pid;
+		}
+		clk_id = MAKE_PROCESS_CPUCLOCK(pid);
+		break;
+	case CPUCLOCK_WHICH_TID:
+		if (uap->id == 0)
+			tid = td->td_tid;
+		else
+			tid = uap->id;
+		clk_id = MAKE_THREAD_CPUCLOCK(tid);
+		break;
+	default:
+		return (EINVAL);
+	}
+	return (copyout(&clk_id, uap->clock_id, sizeof(clockid_t)));
+}
+
+#ifndef _SYS_SYSPROTO_H_
 struct clock_gettime_args {
 	clockid_t clock_id;
 	struct	timespec *tp;
@@ -184,12 +236,85 @@ sys_clock_gettime(struct thread *td, struct clock_gettime_args *uap)
 	return (error);
 }
 
+static inline void 
+cputick2timespec(uint64_t runtime, struct timespec *ats)
+{
+	runtime = cputick2usec(runtime);
+	ats->tv_sec = runtime / 1000000;
+	ats->tv_nsec = runtime % 1000000 * 1000;
+}
+
+static void
+get_thread_cputime(struct thread *targettd, struct timespec *ats)
+{
+	uint64_t runtime, curtime, switchtime;
+
+	if (targettd == NULL) { /* current thread */
+		critical_enter();
+		switchtime = PCPU_GET(switchtime);
+		curtime = cpu_ticks();
+		runtime = curthread->td_runtime;
+		critical_exit();
+		runtime += curtime - switchtime;
+	} else {
+		thread_lock(targettd);
+		runtime = targettd->td_runtime;
+		thread_unlock(targettd);
+	}
+	cputick2timespec(runtime, ats);
+}
+
+static void
+get_process_cputime(struct proc *targetp, struct timespec *ats)
+{
+	uint64_t runtime;
+	struct rusage ru;
+
+	PROC_SLOCK(targetp);
+	rufetch(targetp, &ru);
+	runtime = targetp->p_rux.rux_runtime;
+	PROC_SUNLOCK(targetp);
+	cputick2timespec(runtime, ats);
+}
+
+static int
+get_cputime(struct thread *td, clockid_t clock_id, struct timespec *ats)
+{
+	struct proc *p, *p2;
+	struct thread *td2;
+	lwpid_t tid;
+	pid_t pid;
+	int error;
+
+	p = td->td_proc;
+	if ((clock_id & CPUCLOCK_PROCESS_BIT) == 0) {
+		tid = clock_id & CPUCLOCK_ID_MASK;
+		td2 = tdfind(tid, p->p_pid);
+		if (td2 == NULL)
+			return (EINVAL);
+		get_thread_cputime(td2, ats);
+		PROC_UNLOCK(td2->td_proc);
+	} else {
+		pid = clock_id & CPUCLOCK_ID_MASK;
+		p2 = pfind(pid);
+		if (p2 == NULL)
+			return (EINVAL);
+		error = p_cansee(td, p2);
+		if (error) {
+			PROC_UNLOCK(p2);
+			return (EINVAL);
+		}
+		get_process_cputime(p2, ats);
+		PROC_UNLOCK(p2);
+	}
+	return (0);
+}
+
 int
 kern_clock_gettime(struct thread *td, clockid_t clock_id, struct timespec *ats)
 {
 	struct timeval sys, user;
 	struct proc *p;
-	uint64_t runtime, curtime, switchtime;
 
 	p = td->td_proc;
 	switch (clock_id) {
@@ -232,17 +357,17 @@ kern_clock_gettime(struct thread *td, clockid_t clock_id, struct timespec *ats)
 		ats->tv_nsec = 0;
 		break;
 	case CLOCK_THREAD_CPUTIME_ID:
-		critical_enter();
-		switchtime = PCPU_GET(switchtime);
-		curtime = cpu_ticks();
-		runtime = td->td_runtime;
-		critical_exit();
-		runtime = cputick2usec(runtime + curtime - switchtime);
-		ats->tv_sec = runtime / 1000000;
-		ats->tv_nsec = runtime % 1000000 * 1000;
+		get_thread_cputime(NULL, ats);
+		break;
+	case CLOCK_PROCESS_CPUTIME_ID:
+		PROC_LOCK(p);
+		get_process_cputime(p, ats);
+		PROC_UNLOCK(p);
 		break;
 	default:
-		return (EINVAL);
+		if ((int)clock_id >= 0)
+			return (EINVAL);
+		return (get_cputime(td, clock_id, ats));
 	}
 	return (0);
 }
@@ -336,12 +461,16 @@ kern_clock_getres(struct thread *td, clockid_t clock_id, struct timespec *ts)
 		ts->tv_nsec = 0;
 		break;
 	case CLOCK_THREAD_CPUTIME_ID:
+	case CLOCK_PROCESS_CPUTIME_ID:
+	cputime:
 		/* sync with cputick2usec */
 		ts->tv_nsec = 1000000 / cpu_tickrate();
 		if (ts->tv_nsec == 0)
 			ts->tv_nsec = 1000;
 		break;
 	default:
+		if ((int)clock_id < 0)
+			goto cputime;
 		return (EINVAL);
 	}
 	return (0);
diff --git a/sys/kern/syscalls.c b/sys/kern/syscalls.c
index a2ea5c8..34829a6 100644
--- a/sys/kern/syscalls.c
+++ b/sys/kern/syscalls.c
@@ -254,7 +254,7 @@ const char *syscallnames[] = {
 	"#244",			/* 244 = nosys */
 	"#245",			/* 245 = nosys */
 	"#246",			/* 246 = nosys */
-	"#247",			/* 247 = nosys */
+	"clock_getcpuclockid2",			/* 247 = clock_getcpuclockid2 */
 	"ntp_gettime",			/* 248 = ntp_gettime */
 	"#249",			/* 249 = nosys */
 	"minherit",			/* 250 = minherit */
diff --git a/sys/kern/syscalls.master b/sys/kern/syscalls.master
index f62dad7..5cb90f8 100644
--- a/sys/kern/syscalls.master
+++ b/sys/kern/syscalls.master
@@ -464,7 +464,8 @@
 244	AUE_NULL	UNIMPL	nosys
 245	AUE_NULL	UNIMPL	nosys
 246	AUE_NULL	UNIMPL	nosys
-247	AUE_NULL	UNIMPL	nosys
+247	AUE_NULL	STD	{ int clock_getcpuclockid2(id_t id,\
+				    int which, clockid_t *clock_id); }
 248	AUE_NULL	STD	{ int ntp_gettime(struct ntptimeval *ntvp); }
 249	AUE_NULL	UNIMPL	nosys
 ; syscall numbers initially used in OpenBSD
diff --git a/sys/kern/systrace_args.c b/sys/kern/systrace_args.c
index 522954b..4c57237 100644
--- a/sys/kern/systrace_args.c
+++ b/sys/kern/systrace_args.c
@@ -1358,6 +1358,15 @@ systrace_args(int sysnum, void *params, uint64_t *uarg, int *n_args)
 		*n_args = 1;
 		break;
 	}
+	/* clock_getcpuclockid2 */
+	case 247: {
+		struct clock_getcpuclockid2_args *p = params;
+		iarg[0] = p->id; /* id_t */
+		iarg[1] = p->which; /* int */
+		uarg[2] = (intptr_t) p->clock_id; /* clockid_t * */
+		*n_args = 3;
+		break;
+	}
 	/* ntp_gettime */
 	case 248: {
 		struct ntp_gettime_args *p = params;
@@ -5432,6 +5441,22 @@ systrace_entry_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
 			break;
 		};
 		break;
+	/* clock_getcpuclockid2 */
+	case 247:
+		switch(ndx) {
+		case 0:
+			p = "id_t";
+			break;
+		case 1:
+			p = "int";
+			break;
+		case 2:
+			p = "clockid_t *";
+			break;
+		default:
+			break;
+		};
+		break;
 	/* ntp_gettime */
 	case 248:
 		switch(ndx) {
@@ -9464,6 +9489,11 @@ systrace_return_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
 		if (ndx == 0 || ndx == 1)
 			p = "int";
 		break;
+	/* clock_getcpuclockid2 */
+	case 247:
+		if (ndx == 0 || ndx == 1)
+			p = "int";
+		break;
 	/* ntp_gettime */
 	case 248:
 		if (ndx == 0 || ndx == 1)
diff --git a/sys/sys/syscall.h b/sys/sys/syscall.h
index 321b70a..5a9e7cf 100644
--- a/sys/sys/syscall.h
+++ b/sys/sys/syscall.h
@@ -219,6 +219,7 @@
 #define	SYS_ffclock_getcounter	241
 #define	SYS_ffclock_setestimate	242
 #define	SYS_ffclock_getestimate	243
+#define	SYS_clock_getcpuclockid2	247
 #define	SYS_ntp_gettime	248
 #define	SYS_minherit	250
 #define	SYS_rfork	251
diff --git a/sys/sys/syscall.mk b/sys/sys/syscall.mk
index 7d9e06c..fe1332d 100644
--- a/sys/sys/syscall.mk
+++ b/sys/sys/syscall.mk
@@ -171,6 +171,7 @@ MIASM =  \
 	ffclock_getcounter.o \
 	ffclock_setestimate.o \
 	ffclock_getestimate.o \
+	clock_getcpuclockid2.o \
 	ntp_gettime.o \
 	minherit.o \
 	rfork.o \
diff --git a/sys/sys/sysproto.h b/sys/sys/sysproto.h
index d0163c6..ca17ea1 100644
--- a/sys/sys/sysproto.h
+++ b/sys/sys/sysproto.h
@@ -736,6 +736,11 @@ struct ffclock_setestimate_args {
 struct ffclock_getestimate_args {
 	char cest_l_[PADL_(struct ffclock_estimate *)]; struct ffclock_estimate * cest; char cest_r_[PADR_(struct ffclock_estimate *)];
 };
+struct clock_getcpuclockid2_args {
+	char id_l_[PADL_(id_t)]; id_t id; char id_r_[PADR_(id_t)];
+	char which_l_[PADL_(int)]; int which; char which_r_[PADR_(int)];
+	char clock_id_l_[PADL_(clockid_t *)]; clockid_t * clock_id; char clock_id_r_[PADR_(clockid_t *)];
+};
 struct ntp_gettime_args {
 	char ntvp_l_[PADL_(struct ntptimeval *)]; struct ntptimeval * ntvp; char ntvp_r_[PADR_(struct ntptimeval *)];
 };
@@ -1907,6 +1912,7 @@ int	sys_nanosleep(struct thread *, struct nanosleep_args *);
 int	sys_ffclock_getcounter(struct thread *, struct ffclock_getcounter_args *);
 int	sys_ffclock_setestimate(struct thread *, struct ffclock_setestimate_args *);
 int	sys_ffclock_getestimate(struct thread *, struct ffclock_getestimate_args *);
+int	sys_clock_getcpuclockid2(struct thread *, struct clock_getcpuclockid2_args *);
 int	sys_ntp_gettime(struct thread *, struct ntp_gettime_args *);
 int	sys_minherit(struct thread *, struct minherit_args *);
 int	sys_rfork(struct thread *, struct rfork_args *);
@@ -2597,6 +2603,7 @@ int	freebsd7_shmctl(struct thread *, struct freebsd7_shmctl_args *);
 #define	SYS_AUE_ffclock_getcounter	AUE_NULL
 #define	SYS_AUE_ffclock_setestimate	AUE_NULL
 #define	SYS_AUE_ffclock_getestimate	AUE_NULL
+#define	SYS_AUE_clock_getcpuclockid2	AUE_NULL
 #define	SYS_AUE_ntp_gettime	AUE_NULL
 #define	SYS_AUE_minherit	AUE_MINHERIT
 #define	SYS_AUE_rfork	AUE_RFORK
diff --git a/sys/sys/time.h b/sys/sys/time.h
index 82bbd97..2890d1f 100644
--- a/sys/sys/time.h
+++ b/sys/sys/time.h
@@ -266,6 +266,7 @@ struct clockinfo {
 #define CLOCK_MONOTONIC_FAST	12	/* FreeBSD-specific. */
 #define CLOCK_SECOND	13		/* FreeBSD-specific. */
 #define CLOCK_THREAD_CPUTIME_ID	14
+#define CLOCK_PROCESS_CPUTIME_ID	15
 #endif
 
 #ifndef TIMER_ABSTIME
@@ -273,6 +274,11 @@ struct clockinfo {
 #define TIMER_ABSTIME	0x1	/* absolute timer */
 #endif
 
+#if __BSD_VISIBLE
+#define	CPUCLOCK_WHICH_PID	0
+#define	CPUCLOCK_WHICH_TID	1
+#endif
+
 #ifdef _KERNEL
 
 /*
@@ -344,6 +350,7 @@ int	utimes(const char *, const struct timeval *);
 
 #if __BSD_VISIBLE
 int	adjtime(const struct timeval *, struct timeval *);
+int	clock_getcpuclockid2(id_t, int, clockid_t *);
 int	futimes(int, const struct timeval *);
 int	futimesat(int, const char *, const struct timeval [2]);
 int	lutimes(const char *, const struct timeval *);
diff --git a/sys/sys/unistd.h b/sys/sys/unistd.h
index c558e8b..d6d92ab 100644
--- a/sys/sys/unistd.h
+++ b/sys/sys/unistd.h
@@ -53,7 +53,7 @@
 #define	_POSIX_ASYNCHRONOUS_IO		0
 #define	_POSIX_CHOWN_RESTRICTED		1
 #define	_POSIX_CLOCK_SELECTION		(-1)
-#define	_POSIX_CPUTIME			(-1)
+#define	_POSIX_CPUTIME			200112L
 #define	_POSIX_FSYNC			200112L
 #define	_POSIX_IPV6			0
 #define	_POSIX_JOB_CONTROL		1