Merge branch 'egil/ensure-erl_crash.dump/OTP-10422' into maint

* egil/ensure-erl_crash.dump/OTP-10422:
  test: Relax timeouts for heart_SUITE
  erts: Fix crash dump write to port hack
  erts: Fix lock check assertion
  doc: Document ERL_CRASH_DUMP_SECONDS behaviour
  test: Add test for heart restart on crash
  test: Add test for heart restart on crash
  erts: Change ERL_CRASH_DUMP_SECONDS behaviour
  test: Refactor away ?line macros in heart_SUITE
  erts: Search for heart in ports that are alive
  heart: Refactor heart debugging
  erts, heart: Ensure erl_crash.dump is written

10 files changed, 146 insertions, 33 deletions

diff --git a/erts/emulator/beam/atom.names b/erts/emulator/beam/atom.names
index 106fad030b..afcbd732df 100644
--- a/erts/emulator/beam/atom.names
+++ b/erts/emulator/beam/atom.names
@@ -252,6 +252,7 @@ atom heap_block_size
 atom heap_size
 atom heap_sizes
 atom heap_type
+atom heart_port
 atom heir
 atom hidden
 atom hide
diff --git a/erts/emulator/beam/break.c b/erts/emulator/beam/break.c
index 93aa2fb8d0..cf66f4e6b6 100644
--- a/erts/emulator/beam/break.c
+++ b/erts/emulator/beam/break.c
@@ -650,10 +650,13 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     ErtsThrPrgrData tpd_buf; /* in case we aren't a managed thread... */
 #endif
     int fd;
+    size_t envsz;
     time_t now;
+    char env[21]; /* enough to hold any 64-bit integer */
     size_t dumpnamebufsize = MAXPATHLEN;
     char dumpnamebuf[MAXPATHLEN];
     char* dumpname;
+    int secs;
 
     if (ERTS_SOMEONE_IS_CRASH_DUMPING)
 	return;
@@ -676,9 +679,41 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     erts_writing_erl_crash_dump = 1;
 #endif
 
-    erts_sys_prepare_crash_dump();
+    envsz = sizeof(env);
+    /* ERL_CRASH_DUMP_SECONDS not set
+     * same as ERL_CRASH_DUMP_SECONDS = 0
+     * - do not write dump
+     * - do not set an alarm
+     * - break immediately
+     *
+     * ERL_CRASH_DUMP_SECONDS = 0
+     * - do not write dump
+     * - do not set an alarm
+     * - break immediately
+     *
+     * ERL_CRASH_DUMP_SECONDS < 0
+     * - do not set alarm
+     * - write dump until done
+     *
+     * ERL_CRASH_DUMP_SECONDS = S (and S positive)
+     * - Don't dump file forever
+     * - set alarm (set in sys)
+     * - write dump until alarm or file is written completely
+     */
+	
+    if (erts_sys_getenv__("ERL_CRASH_DUMP_SECONDS", env, &envsz) != 0) {
+	return; /* break immediately */
+    } else {
+	secs = atoi(env);
+    }
+
+    if (secs == 0) {
+	return;
+    }
+
+    erts_sys_prepare_crash_dump(secs);
 
-    if (erts_sys_getenv_raw("ERL_CRASH_DUMP",&dumpnamebuf[0],&dumpnamebufsize) != 0)
+    if (erts_sys_getenv__("ERL_CRASH_DUMP",&dumpnamebuf[0],&dumpnamebufsize) != 0)
 	dumpname = "erl_crash.dump";
     else
 	dumpname = &dumpnamebuf[0];
diff --git a/erts/emulator/beam/erl_lock_check.c b/erts/emulator/beam/erl_lock_check.c
index b545ec07c0..1f388c1796 100644
--- a/erts/emulator/beam/erl_lock_check.c
+++ b/erts/emulator/beam/erl_lock_check.c
@@ -443,7 +443,7 @@ print_lock2(char *prefix, Sint16 id, Wterm extra, Uint16 flags, char *suffix)
 		     "%s'%s:%p%s'%s%s",
 		     prefix,
 		     lname,
-		     boxed_val(extra),
+		     _unchecked_boxed_val(extra),
 		     lock_type(flags),
 		     rw_op_str(flags),
 		     suffix);
diff --git a/erts/emulator/beam/erl_process_lock.c b/erts/emulator/beam/erl_process_lock.c
index f28ef6a6d7..f7900317cc 100644
--- a/erts/emulator/beam/erl_process_lock.c
+++ b/erts/emulator/beam/erl_process_lock.c
@@ -1349,7 +1349,7 @@ erts_proc_lc_chk_no_proc_locks(char *file, int line)
 		  lc_id.proc_lock_msgq,
 		  lc_id.proc_lock_status};
     erts_lc_have_lock_ids(resv, ids, 4);
-    if (resv[0] || resv[1] || resv[2] || resv[3]) {
+    if (!ERTS_IS_CRASH_DUMPING && (resv[0] || resv[1] || resv[2] || resv[3])) {
 	erts_lc_fail("%s:%d: Thread has process locks locked when expected "
 		     "not to have any process locks locked",
 		     file, line);
diff --git a/erts/emulator/beam/global.h b/erts/emulator/beam/global.h
index dbf95f5bd7..2c20e3da3b 100755
--- a/erts/emulator/beam/global.h
+++ b/erts/emulator/beam/global.h
@@ -983,6 +983,9 @@ Uint erts_port_ioq_size(Port *pp);
 void erts_stale_drv_select(Eterm, ErlDrvEvent, int, int);
 void erts_port_cleanup(Port *);
 void erts_fire_port_monitor(Port *prt, Eterm ref);
+
+Port *erts_get_heart_port(void);
+
 #ifdef ERTS_SMP
 void erts_smp_xports_unlock(Port *);
 #endif
diff --git a/erts/emulator/beam/index.c b/erts/emulator/beam/index.c
index a4a3007f93..ad4672c3de 100644
--- a/erts/emulator/beam/index.c
+++ b/erts/emulator/beam/index.c
@@ -82,7 +82,8 @@ index_put(IndexTable* t, void* tmpl)
     if (ix >= t->size) {
 	Uint sz;
 	if (ix >= t->limit) {
-	    erl_exit(1, "no more index entries in %s (max=%d)\n",
+	    /* A core dump is unnecessary */
+	    erl_exit(ERTS_DUMP_EXIT, "no more index entries in %s (max=%d)\n",
 		     t->htable.name, t->limit);
 	}
 	sz = INDEX_PAGE_SIZE*sizeof(IndexSlot*);
diff --git a/erts/emulator/beam/io.c b/erts/emulator/beam/io.c
index 35b194f927..609fe9f5fb 100644
--- a/erts/emulator/beam/io.c
+++ b/erts/emulator/beam/io.c
@@ -1132,7 +1132,7 @@ int erts_write_to_port(Eterm caller_id, Port *p, Eterm list)
     Uint size;
     int fpe_was_unmasked;
     
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(p));
+    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(p) || ERTS_IS_CRASH_DUMPING);
     ERTS_SMP_CHK_NO_PROC_LOCKS;
 
     p->caller = caller_id;
@@ -5235,3 +5235,27 @@ erl_drv_getenv(char *key, char *value, size_t *value_size)
 {
     return erts_sys_getenv_raw(key, value, value_size);
 }
+
+/* get heart_port
+ * used by erl_crash_dump
+ * - uses the fact that heart_port is registered when starting heart
+ */
+
+Port *erts_get_heart_port() {
+
+    Port* port;
+    Uint  ix;
+
+    for(ix = 0; ix < erts_max_ports; ix++) {
+	port = &erts_port[ix];
+	/* only examine undead or alive ports */
+	if (port->status & ERTS_PORT_SFLGS_DEAD)
+	    continue;
+	/* immediate atom compare */
+	if (port->reg && port->reg->name == am_heart_port) {
+	    return port;
+	}
+    }
+
+    return NULL;
+}
diff --git a/erts/emulator/beam/sys.h b/erts/emulator/beam/sys.h
index 2406c52f14..9dd8341520 100644
--- a/erts/emulator/beam/sys.h
+++ b/erts/emulator/beam/sys.h
@@ -652,7 +652,7 @@ void erts_sys_schedule_interrupt_timed(int set, erts_short_time_t msec);
 void erts_sys_main_thread(void);
 #endif
 
-extern void erts_sys_prepare_crash_dump(void);
+extern void erts_sys_prepare_crash_dump(int secs);
 extern void erts_sys_pre_init(void);
 extern void erl_sys_init(void);
 extern void erl_sys_args(int *argc, char **argv);
diff --git a/erts/emulator/sys/unix/sys.c b/erts/emulator/sys/unix/sys.c
index c1fa00b4ea..c485a4eece 100644
--- a/erts/emulator/sys/unix/sys.c
+++ b/erts/emulator/sys/unix/sys.c
@@ -58,7 +58,6 @@
 #define __DARWIN__ 1
 #endif
 
-
 #ifdef USE_THREADS
 #include "erl_threads.h"
 #endif
@@ -71,7 +70,6 @@ static erts_smp_rwmtx_t environ_rwmtx;
 #define MAX_VSIZE 16		/* Max number of entries allowed in an I/O
 				 * vector sock_sendv().
 				 */
-
 /*
  * Don't need global.h, but bif_table.h (included by bif.h),
  * won't compile otherwise
@@ -123,6 +121,15 @@ struct ErtsSysReportExit_ {
 #endif
 };
 
+/* This data is shared by these drivers - initialized by spawn_init() */
+static struct driver_data {
+    int port_num, ofd, packet_bytes;
+    ErtsSysReportExit *report_exit;
+    int pid;
+    int alive;
+    int status;
+} *driver_data;			/* indexed by fd */
+
 static ErtsSysReportExit *report_exit_list;
 #if CHLDWTHR && !defined(ERTS_SMP)
 static ErtsSysReportExit *report_exit_transit_list;
@@ -680,17 +687,40 @@ static RETSIGTYPE break_handler(int sig)
 #endif /* 0 */
 
 static ERTS_INLINE void
-prepare_crash_dump(void)
+prepare_crash_dump(int secs)
 {
+#define NUFBUF (3)
     int i, max;
     char env[21]; /* enough to hold any 64-bit integer */
     size_t envsz;
+    DeclareTmpHeapNoproc(heap,NUFBUF);
+    Port *heart_port;
+    Eterm *hp = heap;
+    Eterm list = NIL;
+    int heart_fd[2] = {-1,-1};
+
+    UseTmpHeapNoproc(NUFBUF);
 
     if (ERTS_PREPARED_CRASH_DUMP)
 	return; /* We have already been called */
 
+    heart_port = erts_get_heart_port();
+    if (heart_port) {
+	/* hearts input fd
+	 * We "know" drv_data is the in_fd since the port is started with read|write
+	 */
+	heart_fd[0] = (int)heart_port->drv_data;
+	heart_fd[1] = (int)driver_data[heart_fd[0]].ofd;
+
+	list = CONS(hp, make_small(8), list); hp += 2;
+
+	/* send to heart port, CMD = 8, i.e. prepare crash dump =o */
+	erts_write_to_port(ERTS_INVALID_PID, heart_port, list);
+    }
+
     /* Make sure we unregister at epmd (unknown fd) and get at least
        one free filedescriptor (for erl_crash.dump) */
+
     max = max_files;
     if (max < 1024)
 	max = 1024;
@@ -704,11 +734,15 @@ prepare_crash_dump(void)
 	if (i == async_fd[0] || i == async_fd[1])
 	    continue;
 #endif
+	/* We don't want to close our heart yet ... */
+	if (i == heart_fd[0] || i == heart_fd[1])
+	    continue;
+
 	close(i);
     }
 
     envsz = sizeof(env);
-    i = erts_sys_getenv_raw("ERL_CRASH_DUMP_NICE", env, &envsz);
+    i = erts_sys_getenv__("ERL_CRASH_DUMP_NICE", env, &envsz);
     if (i >= 0) {
 	int nice_val;
 	nice_val = i != 0 ? 0 : atoi(env);
@@ -717,21 +751,21 @@ prepare_crash_dump(void)
 	}
 	erts_silence_warn_unused_result(nice(nice_val));
     }
-    
-    envsz = sizeof(env);
-    i = erts_sys_getenv_raw("ERL_CRASH_DUMP_SECONDS", env, &envsz);
-    if (i >= 0) {
-	unsigned sec;
-	sec = (unsigned) i != 0 ? 0 : atoi(env);
-	alarm(sec);
-    }
 
+    /* Positive secs means an alarm must be set
+     * 0 or negative means no alarm
+     */
+    if (secs > 0) {
+	alarm((unsigned int)secs);
+    }
+    UnUseTmpHeapNoproc(NUFBUF);
+#undef NUFBUF
 }
 
 void
-erts_sys_prepare_crash_dump(void)
+erts_sys_prepare_crash_dump(int secs)
 {
-    prepare_crash_dump();
+    prepare_crash_dump(secs);
 }
 
 static ERTS_INLINE void
@@ -773,7 +807,7 @@ sigusr1_exit(void)
       is hung somewhere, so it won't be able to poll any flag we set here.
       */
     ERTS_SET_GOT_SIGUSR1;
-    prepare_crash_dump();
+    prepare_crash_dump((int)0);
     erl_exit(1, "Received SIGUSR1\n");
 }
 
@@ -1021,15 +1055,6 @@ void fini_getenv_state(GETENV_STATE *state)
 
 #define ERTS_SYS_READ_BUF_SZ (64*1024)
 
-/* This data is shared by these drivers - initialized by spawn_init() */
-static struct driver_data {
-    int port_num, ofd, packet_bytes;
-    ErtsSysReportExit *report_exit;
-    int pid;
-    int alive;
-    int status;
-} *driver_data;			/* indexed by fd */
-
 /* Driver interfaces */
 static ErlDrvData spawn_start(ErlDrvPort, char*, SysDriverOpts*);
 static ErlDrvData fd_start(ErlDrvPort, char*, SysDriverOpts*);
@@ -2417,6 +2442,15 @@ erts_sys_getenv_raw(char *key, char *value, size_t *size) {
     return erts_sys_getenv(key, value, size);
 }
 
+/*
+ * erts_sys_getenv
+ * returns:
+ *  -1, if environment key is not set with a value
+ *   0, if environment key is set and value fits into buffer size
+ *   1, if environment key is set but does not fit into buffer size
+ *      size is set with the needed buffer size value
+ */
+
 int
 erts_sys_getenv(char *key, char *value, size_t *size)
 {
diff --git a/erts/emulator/sys/win32/sys.c b/erts/emulator/sys/win32/sys.c
index c4e748ed3c..50f6219374 100755
--- a/erts/emulator/sys/win32/sys.c
+++ b/erts/emulator/sys/win32/sys.c
@@ -256,10 +256,25 @@ void erl_sys_args(int* argc, char** argv)
 }
 
 void
-erts_sys_prepare_crash_dump(void)
+erts_sys_prepare_crash_dump(int secs)
 {
+    Port *heart_port;
+    Eterm heap[3];
+    Eterm *hp = heap;
+    Eterm list = NIL;
+
+    heart_port = erts_get_heart_port();
+
+    if (heart_port) {
+
+	list = CONS(hp, make_small(8), list); hp += 2;
+
+	/* send to heart port, CMD = 8, i.e. prepare crash dump =o */
+	erts_write_to_port(NIL, heart_port, list);
+    }
+
     /* Windows - free file descriptors are hopefully available */
-    return;
+    /* Alarm not used on windows */
 }
 
 static void