Add basic reentrancy-checking support, and allow arena_new to reenter.

This checks whether or not we're reentrant using thread-local data, and, if we
are, moves certain internal allocations to use arena 0 (which should be properly
initialized after bootstrapping).

The immediate thing this allows is spinning up threads in arena_new, which will
enable spinning up background threads there.

include/jemalloc/internal/hooks.h

@@ -1,8 +1,8 @@
 #ifndef JEMALLOC_INTERNAL_HOOKS_H
 #define JEMALLOC_INTERNAL_HOOKS_H
 
-extern void (*hooks_arena_new_hook)();
-extern void (*hooks_libc_hook)();
+extern JEMALLOC_EXPORT void (*hooks_arena_new_hook)();
+extern JEMALLOC_EXPORT void (*hooks_libc_hook)();
 
 #define JEMALLOC_HOOK(fn, hook) ((void)(hook != NULL && (hook(), 0)), fn)
 

include/jemalloc/internal/jemalloc_internal.h.in

@@ -1013,6 +1013,11 @@ arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
 		return arena;
 	}
 
+	/* During reentrancy, arena 0 is the safest bet. */
+	if (*tsd_reentrancy_levelp_get(tsd) > 1) {
+		return arena_get(tsd_tsdn(tsd), 0, true);
+	}
+
 	ret = internal ? tsd_iarena_get(tsd) : tsd_arena_get(tsd);
 	if (unlikely(ret == NULL)) {
 		ret = arena_choose_hard(tsd, internal);
@@ -1193,7 +1198,9 @@ idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_internal,
 	if (config_stats && is_internal) {
 		arena_internal_sub(iaalloc(tsdn, ptr), isalloc(tsdn, ptr));
 	}
-
+	if (!is_internal && *tsd_reentrancy_levelp_get(tsdn_tsd(tsdn)) != 0) {
+		tcache = NULL;
+	}
 	arena_dalloc(tsdn, ptr, tcache, slow_path);
 }
 

include/jemalloc/internal/private_symbols.txt

@@ -232,6 +232,7 @@ hash_rotl_64
 hash_x64_128
 hash_x86_128
 hash_x86_32
+hooks_arena_new_hook
 hooks_libc_hook
 iaalloc
 ialloc
@@ -537,6 +538,9 @@ tsd_init_head
 tsd_narenas_tdata_get
 tsd_narenas_tdata_set
 tsd_narenas_tdatap_get
+tsd_reentrancy_level_get
+tsd_reentrancy_level_set
+tsd_reentrancy_levelp_get
 tsd_wrapper_get
 tsd_wrapper_set
 tsd_nominal

include/jemalloc/internal/tsd_structs.h

@@ -65,7 +65,8 @@ struct tsd_init_head_s {
     O(witnesses,		witness_list_t,	no,	no,	yes)	\
     O(rtree_leaf_elm_witnesses,	rtree_leaf_elm_witness_tsd_t,		\
 						no,	no,	no)	\
-    O(witness_fork,		bool,		yes,	no,	no)
+    O(witness_fork,		bool,		yes,	no,	no)	\
+    O(reentrancy_level,		int,		no,	no,	no)
 
 #define TSD_INITIALIZER {						\
     tsd_state_uninitialized,						\
@@ -82,7 +83,8 @@ struct tsd_init_head_s {
     TCACHE_ZERO_INITIALIZER,						\
     ql_head_initializer(witnesses),					\
     RTREE_ELM_WITNESS_TSD_INITIALIZER,					\
-    false								\
+    false,								\
+    0									\
 }
 
 struct tsd_s {