thread: add td_wantedlock

This enables obtaining lock information threads are actively waiting for
while sampling. Without the change one would only see a bunch of calls
to lock_delay(), where the stacktrace often does not reveal what the
lock might be.

Note this is not the same as lock profiling, which only produces data
for cases which wait for locks.

struct thread already has a td_lockname field, but I did not use it
because it has different semantics -- denotes when the thread is off
cpu. At the same time it could not be converted to hold a lock_object
pointer because non-curthread access would no longer be guaranteed to be
safe -- by the time it reads the pointer the lock might have been taken,
released and the object containing it freed.

Sample usage with dtrace:
rm /tmp/out.kern_stacks ; dtrace -x stackframes=100 -n 'profile-997 { @[curthread->td_wantedlock != NULL ? stringof(curthread->td_wantedlock->lo_name) : stringof("\n"), stack()] = count(); }' -o /tmp/out.kern_stacks

This also facilitates addition of lock information to traces produced by
hwpmc.

Note: spinlocks are not supported at the moment.

Sponsored by:	Rubicon Communications, LLC ("Netgate")

sys/kern/kern_mutex.c

@@ -574,6 +574,8 @@ __mtx_lock_sleep(volatile uintptr_t *c, uintptr_t v)
 		    "_mtx_lock_sleep: %s contested (lock=%p) at %s:%d",
 		    m->lock_object.lo_name, (void *)m->mtx_lock, file, line);
 
+	THREAD_CONTENDS_ON_LOCK(&m->lock_object);
+
 	for (;;) {
 		if (v == MTX_UNOWNED) {
 			if (_mtx_obtain_lock_fetch(m, &v, tid))
@@ -670,6 +672,7 @@ retry_turnstile:
 #endif
 		v = MTX_READ_VALUE(m);
 	}
+	THREAD_CONTENTION_DONE(&m->lock_object);
 #if defined(KDTRACE_HOOKS) || defined(LOCK_PROFILING)
 	if (__predict_true(!doing_lockprof))
 		return;

sys/kern/kern_rwlock.c

@@ -482,6 +482,8 @@ __rw_rlock_hard(struct rwlock *rw, struct thread *td, uintptr_t v
 	lock_profile_obtain_lock_failed(&rw->lock_object, false,
 	    &contested, &waittime);
 
+	THREAD_CONTENDS_ON_LOCK(&rw->lock_object);
+
 	for (;;) {
 		if (__rw_rlock_try(rw, td, &v, false LOCK_FILE_LINE_ARG))
 			break;
@@ -628,6 +630,7 @@ retry_ts:
 			    __func__, rw);
 		v = RW_READ_VALUE(rw);
 	}
+	THREAD_CONTENTION_DONE(&rw->lock_object);
 #if defined(KDTRACE_HOOKS) || defined(LOCK_PROFILING)
 	if (__predict_true(!doing_lockprof))
 		return;
@@ -976,6 +979,8 @@ __rw_wlock_hard(volatile uintptr_t *c, uintptr_t v LOCK_FILE_LINE_ARG_DEF)
 	lock_profile_obtain_lock_failed(&rw->lock_object, false,
 	    &contested, &waittime);
 
+	THREAD_CONTENDS_ON_LOCK(&rw->lock_object);
+
 	for (;;) {
 		if (v == RW_UNLOCKED) {
 			if (_rw_write_lock_fetch(rw, &v, tid))
@@ -1161,6 +1166,7 @@ retry_ts:
 #endif
 		v = RW_READ_VALUE(rw);
 	}
+	THREAD_CONTENTION_DONE(&rw->lock_object);
 	if (__predict_true(!extra_work))
 		return;
 #ifdef ADAPTIVE_RWLOCKS

sys/kern/kern_sx.c

@@ -653,6 +653,8 @@ _sx_xlock_hard(struct sx *sx, uintptr_t x, int opts LOCK_FILE_LINE_ARG_DEF)
 	GIANT_SAVE(extra_work);
 #endif
 
+	THREAD_CONTENDS_ON_LOCK(&sx->lock_object);
+
 	for (;;) {
 		if (x == SX_LOCK_UNLOCKED) {
 			if (atomic_fcmpset_acq_ptr(&sx->sx_lock, &x, tid))
@@ -872,6 +874,7 @@ retry_sleepq:
 			    __func__, sx);
 		x = SX_READ_VALUE(sx);
 	}
+	THREAD_CONTENTION_DONE(&sx->lock_object);
 	if (__predict_true(!extra_work))
 		return (error);
 #ifdef ADAPTIVE_SX
@@ -1074,6 +1077,8 @@ _sx_slock_hard(struct sx *sx, int opts, uintptr_t x LOCK_FILE_LINE_ARG_DEF)
 	GIANT_SAVE(extra_work);
 #endif
 
+	THREAD_CONTENDS_ON_LOCK(&sx->lock_object);
+
 	/*
 	 * As with rwlocks, we don't make any attempt to try to block
 	 * shared locks once there is an exclusive waiter.
@@ -1224,6 +1229,7 @@ retry_sleepq:
 			    __func__, sx);
 		x = SX_READ_VALUE(sx);
 	}
+	THREAD_CONTENTION_DONE(&sx->lock_object);
 #if defined(KDTRACE_HOOKS) || defined(LOCK_PROFILING)
 	if (__predict_true(!extra_work))
 		return (error);

sys/kern/kern_thread.c

@@ -87,7 +87,7 @@ _Static_assert(offsetof(struct thread, td_flags) == 0x108,
     "struct thread KBI td_flags");
 _Static_assert(offsetof(struct thread, td_pflags) == 0x114,
     "struct thread KBI td_pflags");
-_Static_assert(offsetof(struct thread, td_frame) == 0x4b0,
+_Static_assert(offsetof(struct thread, td_frame) == 0x4b8,
     "struct thread KBI td_frame");
 _Static_assert(offsetof(struct thread, td_emuldata) == 0x6c0,
     "struct thread KBI td_emuldata");
@@ -107,9 +107,9 @@ _Static_assert(offsetof(struct thread, td_flags) == 0x9c,
     "struct thread KBI td_flags");
 _Static_assert(offsetof(struct thread, td_pflags) == 0xa8,
     "struct thread KBI td_pflags");
-_Static_assert(offsetof(struct thread, td_frame) == 0x314,
+_Static_assert(offsetof(struct thread, td_frame) == 0x318,
     "struct thread KBI td_frame");
-_Static_assert(offsetof(struct thread, td_emuldata) == 0x358,
+_Static_assert(offsetof(struct thread, td_emuldata) == 0x35c,
     "struct thread KBI td_emuldata");
 _Static_assert(offsetof(struct proc, p_flag) == 0x6c,
     "struct proc KBI p_flag");

sys/sys/proc.h

@@ -277,6 +277,7 @@ struct thread {
 	int		td_rw_rlocks;	/* (k) Count of rwlock read locks. */
 	int		td_sx_slocks;	/* (k) Count of sx shared locks. */
 	int		td_lk_slocks;	/* (k) Count of lockmgr shared locks. */
+	struct lock_object *td_wantedlock; /* (k) Lock we are contending on */
 	struct turnstile *td_blocked;	/* (t) Lock thread is blocked on. */
 	const char	*td_lockname;	/* (t) Name of lock blocked on. */
 	LIST_HEAD(, turnstile) td_contested;	/* (q) Contested locks. */
@@ -1074,6 +1075,16 @@ extern pid_t pid_max;
 
 #define	THREAD_CAN_SLEEP()		((curthread)->td_no_sleeping == 0)
 
+#define	THREAD_CONTENDS_ON_LOCK(lo)		do {			\
+	MPASS(curthread->td_wantedlock == NULL);			\
+	curthread->td_wantedlock = lo;					\
+} while (0)
+
+#define	THREAD_CONTENTION_DONE(lo)		do {			\
+	MPASS(curthread->td_wantedlock == lo);				\
+	curthread->td_wantedlock = NULL;				\
+} while (0)
+
 #define	PIDHASH(pid)	(&pidhashtbl[(pid) & pidhash])
 #define	PIDHASHLOCK(pid) (&pidhashtbl_lock[((pid) & pidhashlock)])
 extern LIST_HEAD(pidhashhead, proc) *pidhashtbl;