Fix fork(2)-related deadlocks.

Add a library constructor for jemalloc that initializes the allocator.
This fixes a race that could occur if threads were created by the main
thread prior to any memory allocation, followed by fork(2), and then
memory allocation in the child process.

Fix the prefork/postfork functions to acquire/release the ctl, prof, and
rtree mutexes.  This fixes various fork() child process deadlocks, but
one possible deadlock remains (intentionally) unaddressed: prof
backtracing can acquire runtime library mutexes, so deadlock is still
possible if heap profiling is enabled during fork().  This deadlock is
known to be a real issue in at least the case of libgcc-based
backtracing.

Reported by tfengjun.

include/jemalloc/internal/chunk.h

@@ -45,6 +45,9 @@ extern size_t		arena_maxclass; /* Max size class for arenas. */
 void	*chunk_alloc(size_t size, size_t alignment, bool base, bool *zero);
 void	chunk_dealloc(void *chunk, size_t size, bool unmap);
 bool	chunk_boot(void);
+void	chunk_prefork(void);
+void	chunk_postfork_parent(void);
+void	chunk_postfork_child(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/

include/jemalloc/internal/ctl.h

@@ -75,6 +75,9 @@ int	ctl_nametomib(const char *name, size_t *mibp, size_t *miblenp);
 int	ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
     void *newp, size_t newlen);
 bool	ctl_boot(void);
+void	ctl_prefork(void);
+void	ctl_postfork_parent(void);
+void	ctl_postfork_child(void);
 
 #define	xmallctl(name, oldp, oldlenp, newp, newlen) do {		\
 	if (je_mallctl(name, oldp, oldlenp, newp, newlen)		\

include/jemalloc/internal/private_namespace.h

@@ -59,6 +59,7 @@
 #define	arenas_lock JEMALLOC_N(arenas_lock)
 #define	arenas_lrun_i_index JEMALLOC_N(arenas_lrun_i_index)
 #define	arenas_tls JEMALLOC_N(arenas_tls)
+#define	arenas_tsd JEMALLOC_N(arenas_tsd)
 #define	arenas_tsd_boot JEMALLOC_N(arenas_tsd_boot)
 #define	arenas_tsd_cleanup_wrapper JEMALLOC_N(arenas_tsd_cleanup_wrapper)
 #define	arenas_tsd_get JEMALLOC_N(arenas_tsd_get)
@@ -104,6 +105,9 @@
 #define	chunk_dss_prefork JEMALLOC_N(chunk_dss_prefork)
 #define	chunk_in_dss JEMALLOC_N(chunk_in_dss)
 #define	chunk_npages JEMALLOC_N(chunk_npages)
+#define	chunk_postfork_child JEMALLOC_N(chunk_postfork_child)
+#define	chunk_postfork_parent JEMALLOC_N(chunk_postfork_parent)
+#define	chunk_prefork JEMALLOC_N(chunk_prefork)
 #define	chunks_mtx JEMALLOC_N(chunks_mtx)
 #define	chunks_rtree JEMALLOC_N(chunks_rtree)
 #define	chunksize JEMALLOC_N(chunksize)
@@ -129,6 +133,9 @@
 #define	ctl_bymib JEMALLOC_N(ctl_bymib)
 #define	ctl_byname JEMALLOC_N(ctl_byname)
 #define	ctl_nametomib JEMALLOC_N(ctl_nametomib)
+#define	ctl_postfork_child JEMALLOC_N(ctl_postfork_child)
+#define	ctl_postfork_parent JEMALLOC_N(ctl_postfork_parent)
+#define	ctl_prefork JEMALLOC_N(ctl_prefork)
 #define	extent_tree_ad_first JEMALLOC_N(extent_tree_ad_first)
 #define	extent_tree_ad_insert JEMALLOC_N(extent_tree_ad_insert)
 #define	extent_tree_ad_iter JEMALLOC_N(extent_tree_ad_iter)
@@ -161,6 +168,7 @@
 #define	extent_tree_szad_reverse_iter_recurse JEMALLOC_N(extent_tree_szad_reverse_iter_recurse)
 #define	extent_tree_szad_reverse_iter_start JEMALLOC_N(extent_tree_szad_reverse_iter_start)
 #define	extent_tree_szad_search JEMALLOC_N(extent_tree_szad_search)
+#define	get_errno JEMALLOC_N(get_errno)
 #define	hash JEMALLOC_N(hash)
 #define	huge_allocated JEMALLOC_N(huge_allocated)
 #define	huge_boot JEMALLOC_N(huge_boot)
@@ -254,6 +262,9 @@
 #define	prof_lookup JEMALLOC_N(prof_lookup)
 #define	prof_malloc JEMALLOC_N(prof_malloc)
 #define	prof_mdump JEMALLOC_N(prof_mdump)
+#define	prof_postfork_child JEMALLOC_N(prof_postfork_child)
+#define	prof_postfork_parent JEMALLOC_N(prof_postfork_parent)
+#define	prof_prefork JEMALLOC_N(prof_prefork)
 #define	prof_promote JEMALLOC_N(prof_promote)
 #define	prof_realloc JEMALLOC_N(prof_realloc)
 #define	prof_sample_accum_update JEMALLOC_N(prof_sample_accum_update)
@@ -264,6 +275,7 @@
 #define	prof_tdata_init JEMALLOC_N(prof_tdata_init)
 #define	prof_tdata_initialized JEMALLOC_N(prof_tdata_initialized)
 #define	prof_tdata_tls JEMALLOC_N(prof_tdata_tls)
+#define	prof_tdata_tsd JEMALLOC_N(prof_tdata_tsd)
 #define	prof_tdata_tsd_boot JEMALLOC_N(prof_tdata_tsd_boot)
 #define	prof_tdata_tsd_cleanup_wrapper JEMALLOC_N(prof_tdata_tsd_cleanup_wrapper)
 #define	prof_tdata_tsd_get JEMALLOC_N(prof_tdata_tsd_get)
@@ -278,9 +290,13 @@
 #define	rtree_get JEMALLOC_N(rtree_get)
 #define	rtree_get_locked JEMALLOC_N(rtree_get_locked)
 #define	rtree_new JEMALLOC_N(rtree_new)
+#define	rtree_postfork_child JEMALLOC_N(rtree_postfork_child)
+#define	rtree_postfork_parent JEMALLOC_N(rtree_postfork_parent)
+#define	rtree_prefork JEMALLOC_N(rtree_prefork)
 #define	rtree_set JEMALLOC_N(rtree_set)
 #define	s2u JEMALLOC_N(s2u)
 #define	sa2u JEMALLOC_N(sa2u)
+#define	set_errno JEMALLOC_N(set_errno)
 #define	stats_arenas_i_bins_j_index JEMALLOC_N(stats_arenas_i_bins_j_index)
 #define	stats_arenas_i_index JEMALLOC_N(stats_arenas_i_index)
 #define	stats_arenas_i_lruns_j_index JEMALLOC_N(stats_arenas_i_lruns_j_index)
@@ -311,6 +327,7 @@
 #define	tcache_enabled_initialized JEMALLOC_N(tcache_enabled_initialized)
 #define	tcache_enabled_set JEMALLOC_N(tcache_enabled_set)
 #define	tcache_enabled_tls JEMALLOC_N(tcache_enabled_tls)
+#define	tcache_enabled_tsd JEMALLOC_N(tcache_enabled_tsd)
 #define	tcache_enabled_tsd_boot JEMALLOC_N(tcache_enabled_tsd_boot)
 #define	tcache_enabled_tsd_cleanup_wrapper JEMALLOC_N(tcache_enabled_tsd_cleanup_wrapper)
 #define	tcache_enabled_tsd_get JEMALLOC_N(tcache_enabled_tsd_get)
@@ -325,6 +342,7 @@
 #define	tcache_stats_merge JEMALLOC_N(tcache_stats_merge)
 #define	tcache_thread_cleanup JEMALLOC_N(tcache_thread_cleanup)
 #define	tcache_tls JEMALLOC_N(tcache_tls)
+#define	tcache_tsd JEMALLOC_N(tcache_tsd)
 #define	tcache_tsd_boot JEMALLOC_N(tcache_tsd_boot)
 #define	tcache_tsd_cleanup_wrapper JEMALLOC_N(tcache_tsd_cleanup_wrapper)
 #define	tcache_tsd_get JEMALLOC_N(tcache_tsd_get)
@@ -332,6 +350,7 @@
 #define	thread_allocated_booted JEMALLOC_N(thread_allocated_booted)
 #define	thread_allocated_initialized JEMALLOC_N(thread_allocated_initialized)
 #define	thread_allocated_tls JEMALLOC_N(thread_allocated_tls)
+#define	thread_allocated_tsd JEMALLOC_N(thread_allocated_tsd)
 #define	thread_allocated_tsd_boot JEMALLOC_N(thread_allocated_tsd_boot)
 #define	thread_allocated_tsd_cleanup_wrapper JEMALLOC_N(thread_allocated_tsd_cleanup_wrapper)
 #define	thread_allocated_tsd_get JEMALLOC_N(thread_allocated_tsd_get)

include/jemalloc/internal/prof.h

@@ -223,6 +223,9 @@ void	prof_tdata_cleanup(void *arg);
 void	prof_boot0(void);
 void	prof_boot1(void);
 bool	prof_boot2(void);
+void	prof_prefork(void);
+void	prof_postfork_parent(void);
+void	prof_postfork_child(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/

include/jemalloc/internal/rtree.h

@@ -36,6 +36,9 @@ struct rtree_s {
 #ifdef JEMALLOC_H_EXTERNS
 
 rtree_t	*rtree_new(unsigned bits);
+void	rtree_prefork(rtree_t *rtree);
+void	rtree_postfork_parent(rtree_t *rtree);
+void	rtree_postfork_child(rtree_t *rtree);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/