Suppress tdata creation in reentrancy

This change suppresses tdata initialization and prof sample threshold
update in interrupting malloc calls.  Interrupting calls have no need
for tdata.  Delaying tdata creation aligns better with our lazy tdata
creation principle, and it also helps us gain control back from
interrupting calls more quickly and reduces any risk of delegating
tdata creation to an interrupting call.

include/jemalloc/internal/prof_inlines_b.h

@@ -22,6 +22,7 @@ prof_tdata_get(tsd_t *tsd, bool create) {
 
 	tdata = tsd_prof_tdata_get(tsd);
 	if (create) {
+		assert(tsd_reentrancy_level_get(tsd) == 0);
 		if (unlikely(tdata == NULL)) {
 			if (tsd_nominal(tsd)) {
 				tdata = prof_tdata_init(tsd);
@@ -109,7 +110,11 @@ prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
 		return true;
 	}
 
-	bool booted = tsd_prof_tdata_get(tsd);
+	if (tsd_reentrancy_level_get(tsd) > 0) {
+		return true;
+	}
+
+	bool booted = prof_tdata_get(tsd, false);
 	tdata = prof_tdata_get(tsd, true);
 	if (unlikely((uintptr_t)tdata <= (uintptr_t)PROF_TDATA_STATE_MAX)) {
 		tdata = NULL;
@@ -132,9 +137,6 @@ prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
 		return true;
 	}
 
-	if (tsd_reentrancy_level_get(tsd) > 0) {
-		return true;
-	}
 	/* Compute new sample threshold. */
 	if (update) {
 		prof_sample_threshold_update(tdata);

src/prof.c

@@ -127,10 +127,15 @@ prof_tctx_should_destroy(tsdn_t *tsdn, prof_tctx_t *tctx) {
 
 void
 prof_alloc_rollback(tsd_t *tsd, prof_tctx_t *tctx, bool updated) {
-	prof_tdata_t *tdata;
-
 	cassert(config_prof);
 
+	if (tsd_reentrancy_level_get(tsd) > 0) {
+		assert((uintptr_t)tctx == (uintptr_t)1U);
+		return;
+	}
+
+	prof_tdata_t *tdata;
+
 	if (updated) {
 		/*
 		 * Compute a new sample threshold.  This isn't very important in
@@ -810,6 +815,8 @@ prof_active_set(tsdn_t *tsdn, bool active) {
 
 const char *
 prof_thread_name_get(tsd_t *tsd) {
+	assert(tsd_reentrancy_level_get(tsd) == 0);
+
 	prof_tdata_t *tdata;
 
 	tdata = prof_tdata_get(tsd, true);
@@ -821,6 +828,8 @@ prof_thread_name_get(tsd_t *tsd) {
 
 int
 prof_thread_name_set(tsd_t *tsd, const char *thread_name) {
+	assert(tsd_reentrancy_level_get(tsd) == 0);
+
 	prof_tdata_t *tdata;
 	unsigned i;
 	char *s;
@@ -859,6 +868,8 @@ prof_thread_name_set(tsd_t *tsd, const char *thread_name) {
 
 bool
 prof_thread_active_get(tsd_t *tsd) {
+	assert(tsd_reentrancy_level_get(tsd) == 0);
+
 	prof_tdata_t *tdata;
 
 	tdata = prof_tdata_get(tsd, true);
@@ -870,6 +881,8 @@ prof_thread_active_get(tsd_t *tsd) {
 
 bool
 prof_thread_active_set(tsd_t *tsd, bool active) {
+	assert(tsd_reentrancy_level_get(tsd) == 0);
+
 	prof_tdata_t *tdata;
 
 	tdata = prof_tdata_get(tsd, true);

src/prof_data.c

@@ -1199,6 +1199,8 @@ prof_bt_keycomp(const void *k1, const void *k2) {
 prof_tdata_t *
 prof_tdata_init_impl(tsd_t *tsd, uint64_t thr_uid, uint64_t thr_discrim,
     char *thread_name, bool active) {
+	assert(tsd_reentrancy_level_get(tsd) == 0);
+
 	prof_tdata_t *tdata;
 
 	cassert(config_prof);