diff options
author | Lukas Larsson <[email protected]> | 2014-12-19 10:53:27 +0100 |
---|---|---|
committer | Lukas Larsson <[email protected]> | 2014-12-19 10:53:27 +0100 |
commit | 3e3ef7b16fa03a7b9f4e5376387750a69957719f (patch) | |
tree | 40d74a3d6ccd8a4faa5c7194979584fa67517ea1 /erts | |
parent | a4d72d2cae3a9bd3f0dccdb663f70d54a45fe10f (diff) | |
parent | bd4b900c09308a6c3ab41e3a2070cb90410ee01d (diff) | |
download | otp-3e3ef7b16fa03a7b9f4e5376387750a69957719f.tar.gz otp-3e3ef7b16fa03a7b9f4e5376387750a69957719f.tar.bz2 otp-3e3ef7b16fa03a7b9f4e5376387750a69957719f.zip |
Merge branch 'lukas/erts/crashdump_improvements/OTP-12377'
* lukas/erts/crashdump_improvements/OTP-12377:
erts: Make main thread safe from pipe closed event
erts: Improve crash dumps
erts: Rename sys_sigset to sys_signal
erts: Introduce thread suspend functions
erts: Remove usage of QUANTIFY signal
erts: Add support for thread names
ets: Increase data available in crash dumps and ets:info
erts: Start compilation of beam_emu earlier
Diffstat (limited to 'erts')
28 files changed, 1035 insertions, 240 deletions
diff --git a/erts/aclocal.m4 b/erts/aclocal.m4 index 63f5a9c840..264a1726db 100644 --- a/erts/aclocal.m4 +++ b/erts/aclocal.m4 @@ -1349,7 +1349,56 @@ case "$THR_LIB_NAME" in AC_MSG_RESULT([$linux_futex]) test $linux_futex = yes && AC_DEFINE(ETHR_HAVE_LINUX_FUTEX, 1, [Define if you have a linux futex implementation.]) - fi + pthread_setname=no + AC_MSG_CHECKING([for pthread_setname_np]) + old_CFLAGS=$CFLAGS + CFLAGS="$CFLAGS -Werror" + AC_TRY_LINK([#define __USE_GNU + #include <pthread.h>], + [pthread_setname_np(pthread_self(), "name");], + pthread_setname=linux) + AC_TRY_LINK([#define __USE_GNU + #include <pthread.h>], + [pthread_set_name_np(pthread_self(), "name");], + pthread_setname=bsd) + AC_TRY_LINK([#define _DARWIN_C_SOURCE + #include <pthread.h>], + [pthread_setname_np("name");], + pthread_setname=darwin) + AC_MSG_RESULT([$pthread_setname]) + case $pthread_setname in + linux) AC_DEFINE(ETHR_HAVE_PTHREAD_SETNAME_NP_2, 1, + [Define if you have linux style pthread_setname_np]);; + bsd) AC_DEFINE(ETHR_HAVE_PTHREAD_SET_NAME_NP_2, 1, + [Define if you have bsd style pthread_set_name_np]);; + darwin) AC_DEFINE(ETHR_HAVE_PTHREAD_SETNAME_NP_1, 1, + [Define if you have darwin style pthread_setname_np]);; + *) ;; + esac + + pthread_getname=no + AC_MSG_CHECKING([for pthread_getname_np]) + AC_TRY_LINK([#define __USE_GNU + #define _DARWIN_C_SOURCE + #include <pthread.h>], + [char buff[256]; pthread_getname_np(pthread_self(), buff, 256);], + pthread_getname=normal) + AC_TRY_LINK([#define __USE_GNU + #define _DARWIN_C_SOURCE + #include <pthread.h>], + [char buff[256]; pthread_getname_np(pthread_self(), buff);], + pthread_getname=ibm) + AC_MSG_RESULT([$pthread_getname]) + case $pthread_getname in + linux) AC_DEFINE(ETHR_HAVE_PTHREAD_GETNAME_NP_3, 1, + [Define if you have linux style pthread_getname_np]);; + ibm) AC_DEFINE(ETHR_HAVE_PTHREAD_GETNAME_NP_2, 1, + [Define if you have ibm style pthread_getname_np]);; + *) ;; + esac + CFLAGS=$old_CFLAGS + + fi ## test "x$THR_LIB_NAME" = "xpthread" AC_CHECK_SIZEOF(int) AC_CHECK_SIZEOF(long) diff --git a/erts/doc/src/crash_dump.xml b/erts/doc/src/crash_dump.xml index 2b5fc877c3..8291bf38b7 100644 --- a/erts/doc/src/crash_dump.xml +++ b/erts/doc/src/crash_dump.xml @@ -55,10 +55,12 @@ emulator or the operating system can be reconfigured to avoid the crash, which is why interpreting the crash dump correctly is important.</p> + <p>On systems that support OS signals, it is also possible to stop + the runtime system and generate a crash dump by sending the SIGUSR1.</p> <p>The erlang crash dump is a readable text file, but it might not be very easy to read. Using the Crashdump Viewer tool in the <c><![CDATA[observer]]></c> application will simplify the task. This is an - HTML based tool for browsing Erlang crash dumps.</p> + wx-widget based tool for browsing Erlang crash dumps.</p> <section> <marker id="general_info"></marker> @@ -66,8 +68,9 @@ <p>The first part of the dump shows the creation time for the dump, a slogan indicating the reason for the dump, the system version, of the node from which the dump originates, the compile time of - the emulator running the originating node and the number of - atoms in the atom table. + the emulator running the originating node, the number of + atoms in the atom table and the runtime system thread that caused + the crash dump to happen. </p> <section> @@ -170,6 +173,60 @@ </section> <section> + <marker id="scheduler"></marker> + <title>Scheduler information</title> + <p>Under the tag <em>=scheduler</em> information about the current state + and statistics of the schedulers in the runtime system is displayed. + On OSs that do allow instant suspension of other threads, the data within + this section will reflect what the runtime system looks like at the moment + when the crash happens.</p> + <p>The following fields can exist for a process:</p> + <taglist> + <tag><em>=scheduler:id</em></tag> + <item>Header, states the scheduler identifier.</item> + <tag><em>Scheduler Sleep Info Flags</em></tag> + <item>If empty the scheduler was doing some work. + If not empty the scheduler is either in some state of sleep, + or suspended. This entry is only present in a SMP enabled emulator</item> + <tag><em>Scheduler Sleep Info Aux Work</em></tag> + <item>If not empty, a scheduler internal auxiliary work is scheduled + to be done.</item> + <tag><em>Current Port</em></tag> + <item>The port identifier of the port that is currently being + executed by the scheduler.</item> + <tag><em>Current Process</em></tag> + <item>The process identifier of the process that is currently being + executed by the scheduler. If there is such a process this entry is + followed by the <em>State</em>,<em>Internal State</em>, + <em>Program Counter</em>, <em>CP</em> of that same process. See + <seealso marker="#processes">Process Information</seealso> for a + description what the different entries mean. Keep in mind that + this is a snapshot of what the entries are exactly when the crash + dump is starting to be generated. Therefore they will most likely + be different (and more telling) then the entries for the same + processes found in the <em>=proc</em> section. If there is no currently + running process, only the <em>Current Process</em> entry will be printed. + </item> + <tag><em>Current Process Limited Stack Trace</em></tag> + <item>This entry only shows up if there is a current process. It is very + similar to <seealso marker="#proc_data"><em>=proc_stack</em></seealso>, + except that only the function frames are printed (i.e. the stack variables + are omited). It is also limited to only print the top and bottom part + of the stack. If the stack is small (less that 512 slots) then the + entire stack will be printed. If not, an entry stating + <code>skipping ## slots</code> will be printed where ## is + replaced by the number of slots that has been skipped.</item> + <tag><em>Run Queue</em></tag> + <item>Displays statistics about how many processes and ports + of different priorities are scheduled on this scheduler.</item> + <tag><em>** crashed **</em></tag> + <item>This entry is normally not printed. It signifies that getting + the rest of the information about this scheduler failed for some reason. + </item> + </taglist> + </section> + + <section> <marker id="memory"></marker> <title>Memory information</title> <p>Under the tag <em>=memory</em> you will find information similar @@ -314,6 +371,9 @@ <item>The number of live argument registers. The argument registers, if any are live, will follow. These may contain the arguments of the function if they are not yet moved to the stack.</item> + <item><em>Internal State</em></item> + <item>A more detailed internal represantation of the state of + this process.</item> </taglist> <p>See also the section about <seealso marker="#proc_data">process data</seealso>.</p> </section> @@ -339,18 +399,38 @@ <tag><em>Name</em></tag> <item>The name of the table, regardless of whether it is a <c><![CDATA[named_table]]></c> or not.</item> - <tag><em>Buckets</em></tag> + <tag><em>Hash table, Buckets</em></tag> <item>This occurs if the table is a hash table, i.e. if it is not an <c><![CDATA[ordered_set]]></c>.</item> + <tag><em>Hash table, Chain Length</em></tag> + <item>Only applicable for hash tables. Contains statistics about the + hash table, such as the max, min and avg chain length. Having a max much + larger than the avg, and a std dev much larger that + the expected std dev is a sign that the hashing of the terms is + behaving badly for some reason.</item> <tag><em>Ordered set (AVL tree), Elements</em></tag> <item>This occurs only if the table is an <c><![CDATA[ordered_set]]></c>. (The number of elements is the same as the number of objects in the table.)</item> + <tag><em>Fixed</em></tag> + <item>If the table is fixed using ets:safe_fixtable or some internal + mechanism.</item> <tag><em>Objects</em></tag> <item>The number of objects in the table</item> <tag><em>Words</em></tag> <item>The number of words (usually 4 bytes/word) allocated to data in the table.</item> + <tag><em>Type</em></tag> + <item>The type of the table, i.e. <c>set</c>, <c>bag</c>, + <c>dublicate_bag</c> or <c>ordered_set</c>.</item> + <tag><em>Compressed</em></tag> + <item>If this table was compressed.</item> + <tag><em>Protection</em></tag> + <item>The protection of this table.</item> + <tag><em>Write Concurrency</em></tag> + <item>If write_concurrency was enabled for this table.</item> + <tag><em>Read Concurrency</em></tag> + <item>If read_concurrency was enabled for this table.</item> </taglist> </section> diff --git a/erts/emulator/Makefile.in b/erts/emulator/Makefile.in index 53fc7bd713..e4cd566285 100644 --- a/erts/emulator/Makefile.in +++ b/erts/emulator/Makefile.in @@ -934,7 +934,7 @@ ifdef HIPE_ENABLED EXTRA_BASE_OBJS += $(HIPE_OBJS) endif -BASE_OBJS = $(RUN_OBJS) $(EMU_OBJS) $(OS_OBJS) $(EXTRA_BASE_OBJS) +BASE_OBJS = $(EMU_OBJS) $(RUN_OBJS) $(OS_OBJS) $(EXTRA_BASE_OBJS) before_DTrace_OBJS = $(BASE_OBJS) $(DRV_OBJS) diff --git a/erts/emulator/beam/break.c b/erts/emulator/beam/break.c index 5aee85174f..41765dad12 100644 --- a/erts/emulator/beam/break.c +++ b/erts/emulator/beam/break.c @@ -210,25 +210,12 @@ print_process_info(int to, void *to_arg, Process *p) erts_print(to, to_arg, "State: "); state = erts_smp_atomic32_read_acqb(&p->state); - if (state & ERTS_PSFLG_FREE) - erts_print(to, to_arg, "Non Existing\n"); /* Should never happen */ - else if (state & ERTS_PSFLG_EXITING) - erts_print(to, to_arg, "Exiting\n"); - else if (state & ERTS_PSFLG_GC) { - garbing = 1; - running = 1; - erts_print(to, to_arg, "Garbing\n"); - } - else if (state & ERTS_PSFLG_SUSPENDED) - erts_print(to, to_arg, "Suspended\n"); - else if (state & ERTS_PSFLG_RUNNING) { - running = 1; - erts_print(to, to_arg, "Running\n"); - } - else if (state & ERTS_PSFLG_ACTIVE) - erts_print(to, to_arg, "Scheduled\n"); - else - erts_print(to, to_arg, "Waiting\n"); + erts_dump_process_state(to, to_arg, state); + if (state & ERTS_PSFLG_GC) { + garbing = 1; + running = 1; + } else if (state & ERTS_PSFLG_RUNNING) + running = 1; /* * If the process is registered as a global process, display the @@ -352,6 +339,10 @@ print_process_info(int to, void *to_arg, Process *p) #endif erts_stack_dump(to, to_arg, p); } + + /* Display all states */ + erts_print(to, to_arg, "Internal State: "); + erts_dump_extended_process_state(to, to_arg, state); } static void @@ -671,6 +662,7 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args) { #ifdef ERTS_SMP ErtsThrPrgrData tpd_buf; /* in case we aren't a managed thread... */ + int bc; #endif int fd; size_t envsz; @@ -681,27 +673,39 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args) char* dumpname; int secs; int env_erl_crash_dump_seconds_set = 1; + int i; if (ERTS_SOMEONE_IS_CRASH_DUMPING) return; #ifdef ERTS_SMP + /* Order all managed threads to block, this has to be done + first to guarantee that this is the only thread to generate + crash dump. */ + bc = erts_thr_progress_fatal_error_block(&tpd_buf); + +#ifdef ERTS_THR_HAVE_SIG_FUNCS /* - * Wait for all managed threads to block. If all threads haven't blocked - * after a minute, we go anyway and hope for the best... - * - * We do not release system again. We expect an exit() or abort() after - * dump has been written. + * We suspend all scheduler threads so that we can dump some + * data about the currently running processes and scheduler data. + * We have to be very very careful when doing this as the schedulers + * could be anywhere. */ - erts_thr_progress_fatal_error_block(60000, &tpd_buf); - /* Either worked or not... */ + for (i = 0; i < erts_no_schedulers; i++) { + erts_tid_t tid = ERTS_SCHEDULER_IX(i)->tid; + if (!erts_equal_tids(tid,erts_thr_self())) + sys_thr_suspend(tid); + } + +#endif /* Allow us to pass certain places without locking... */ erts_smp_atomic32_set_mb(&erts_writing_erl_crash_dump, 1); erts_smp_tsd_set(erts_is_crash_dumping_key, (void *) 1); -#else + +#else /* !ERTS_SMP */ erts_writing_erl_crash_dump = 1; -#endif +#endif /* ERTS_SMP */ envsz = sizeof(env); /* ERL_CRASH_DUMP_SECONDS not set @@ -758,9 +762,8 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args) erts_fprintf(stderr,"\nCrash dump is being written to: %s...", dumpname); fd = open(dumpname,O_WRONLY | O_CREAT | O_TRUNC,0640); - if (fd < 0) + if (fd < 0) return; /* Can't create the crash dump, skip it */ - time(&now); erts_fdprintf(fd, "=erl_crash_dump:0.3\n%s", ctime(&now)); @@ -774,9 +777,74 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args) erts_fdprintf(fd, "System version: "); erts_print_system_version(fd, NULL, NULL); erts_fdprintf(fd, "%s\n", "Compiled: " ERLANG_COMPILE_DATE); + erts_fdprintf(fd, "Taints: "); erts_print_nif_taints(fd, NULL); erts_fdprintf(fd, "Atoms: %d\n", atom_table_size()); + +#ifdef USE_THREADS + /* We want to note which thread it was that called erl_exit */ + if (erts_get_scheduler_data()) { + erts_fdprintf(fd, "Calling Thread: scheduler:%d\n", + erts_get_scheduler_data()->no); + } else { + if (!erts_thr_getname(erts_thr_self(), dumpnamebuf, MAXPATHLEN)) + erts_fdprintf(fd, "Calling Thread: %s\n", dumpnamebuf); + else + erts_fdprintf(fd, "Calling Thread: %p\n", erts_thr_self()); + } +#else + erts_fdprintf(fd, "Calling Thread: scheduler:1\n"); +#endif + +#if defined(ERTS_HAVE_TRY_CATCH) + + /* + * erts_print_scheduler_info is not guaranteed to be safe to call + * here for all schedulers as we may have suspended a scheduler + * in the middle of updating the STACK_TOP and STACK_START + * variables and thus when scanning the stack we could get + * segmentation faults. We protect against this very unlikely + * scenario by using the ERTS_SYS_TRY_CATCH. + */ + for (i = 0; i < erts_no_schedulers; i++) { + ERTS_SYS_TRY_CATCH( + erts_print_scheduler_info(fd, NULL, ERTS_SCHEDULER_IX(i)), + erts_fdprintf(fd, "** crashed **\n")); + } +#endif + +#ifdef ERTS_SMP + +#if defined(ERTS_THR_HAVE_SIG_FUNCS) + + /* We resume all schedulers so that we are in a known safe state + when we write the rest of the crash dump */ + for (i = 0; i < erts_no_schedulers; i++) { + erts_tid_t tid = ERTS_SCHEDULER_IX(i)->tid; + if (!erts_equal_tids(tid,erts_thr_self())) + sys_thr_resume(tid); + } +#endif + + /* + * Wait for all managed threads to block. If all threads haven't blocked + * after a minute, we go anyway and hope for the best... + * + * We do not release system again. We expect an exit() or abort() after + * dump has been written. + */ + erts_thr_progress_fatal_error_wait(60000); + /* Either worked or not... */ +#endif + +#ifndef ERTS_HAVE_TRY_CATCH + /* This is safe to call here, as all schedulers are blocked */ + for (i = 0; i < erts_no_schedulers; i++) { + erts_print_scheduler_info(fd, NULL, ERTS_SCHEDULER_IX(i)); + } +#endif + info(fd, NULL); /* General system info */ if (erts_ptab_initialized(&erts_proc)) process_info(fd, NULL); /* Info about each process and port */ diff --git a/erts/emulator/beam/erl_async.c b/erts/emulator/beam/erl_async.c index decae6b2ca..bc06d41720 100644 --- a/erts/emulator/beam/erl_async.c +++ b/erts/emulator/beam/erl_async.c @@ -176,7 +176,7 @@ erts_init_async(void) ErtsThrQInit_t qinit = ERTS_THR_Q_INIT_DEFAULT; #endif erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER; - char *ptr; + char *ptr, thr_name[16]; size_t tot_size = 0; int i; @@ -227,23 +227,16 @@ erts_init_async(void) thr_opts.suggested_stack_size = erts_async_thread_suggested_stack_size; -#ifdef ETHR_HAVE_THREAD_NAMES - thr_opts.name = malloc(sizeof(char)*(strlen("async_XXXX")+1)); -#endif + thr_opts.name = thr_name; for (i = 0; i < erts_async_max_threads; i++) { ErtsAsyncQ *aq = async_q(i); -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(thr_opts.name, "async_%d", i+1); -#endif + erts_snprintf(thr_opts.name, 16, "async_%d", i+1); erts_thr_create(&aq->thr_id, async_main, (void*) aq, &thr_opts); } -#ifdef ETHR_HAVE_THREAD_NAMES - free(thr_opts.name); -#endif /* Wait for async threads to initialize... */ erts_mtx_lock(&async->init.data.mtx); diff --git a/erts/emulator/beam/erl_bif_info.c b/erts/emulator/beam/erl_bif_info.c index d2ee5e4224..d750e34be3 100644 --- a/erts/emulator/beam/erl_bif_info.c +++ b/erts/emulator/beam/erl_bif_info.c @@ -3900,6 +3900,13 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2) } } } + else if (ERTS_IS_ATOM_STR("broken_halt", BIF_ARG_1)) { + /* Ugly ugly code used by bif_SUITE:erlang_halt/1 */ +#if defined(ERTS_HAVE_TRY_CATCH) + erts_get_scheduler_data()->run_queue = NULL; +#endif + erl_exit(ERTS_DUMP_EXIT, "%T", BIF_ARG_2); + } } BIF_ERROR(BIF_P, BADARG); diff --git a/erts/emulator/beam/erl_db.c b/erts/emulator/beam/erl_db.c index fba1e32230..4806befd99 100644 --- a/erts/emulator/beam/erl_db.c +++ b/erts/emulator/beam/erl_db.c @@ -2668,7 +2668,9 @@ BIF_RETTYPE ets_match_object_3(BIF_ALIST_3) BIF_RETTYPE ets_info_1(BIF_ALIST_1) { static Eterm fields[] = {am_protection, am_keypos, am_type, am_named_table, - am_node, am_size, am_name, am_heir, am_owner, am_memory, am_compressed}; + am_node, am_size, am_name, am_heir, am_owner, am_memory, am_compressed, + am_write_concurrency, + am_read_concurrency}; Eterm results[sizeof(fields)/sizeof(Eterm)]; DbTable* tb; Eterm res; @@ -3695,6 +3697,10 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What) ret = am_protected; else if (tb->common.status & DB_PUBLIC) ret = am_public; + } else if (What == am_write_concurrency) { + ret = tb->common.status & DB_FINE_LOCKED ? am_true : am_false; + } else if (What == am_read_concurrency) { + ret = tb->common.status & DB_FREQ_READ ? am_true : am_false; } else if (What == am_name) { ret = tb->common.the_name; } else if (What == am_keypos) { @@ -3777,7 +3783,7 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What) avg, std_dev_real, std_dev_exp, make_small(stats.min_chain_len), make_small(stats.max_chain_len), - make_small(db_kept_items_hash(&tb->hash))); + make_small(stats.kept_items)); } else { ret = am_false; @@ -3799,6 +3805,11 @@ static void print_table(int to, void *to_arg, int show, DbTable* tb) + sizeof(Uint) - 1) / sizeof(Uint))); + erts_print(to, to_arg, "Type: %T\n", table_info(NULL, tb, am_type)); + erts_print(to, to_arg, "Protection: %T\n", table_info(NULL, tb, am_protection)); + erts_print(to, to_arg, "Compressed: %T\n", table_info(NULL, tb, am_compressed)); + erts_print(to, to_arg, "Write Concurrency: %T\n", table_info(NULL, tb, am_write_concurrency)); + erts_print(to, to_arg, "Read Concurrency: %T\n", table_info(NULL, tb, am_read_concurrency)); } void db_info(int to, void *to_arg, int show) /* Called by break handler */ diff --git a/erts/emulator/beam/erl_db_hash.c b/erts/emulator/beam/erl_db_hash.c index 6e8d892561..c2157457a0 100644 --- a/erts/emulator/beam/erl_db_hash.c +++ b/erts/emulator/beam/erl_db_hash.c @@ -648,25 +648,6 @@ restart: /* ToDo: Maybe try grow/shrink the table as well */ } -/* Only used by tests -*/ -Uint db_kept_items_hash(DbTableHash *tb) -{ - Uint kept_items = 0; - Uint ix = 0; - erts_smp_rwmtx_t* lck = RLOCK_HASH(tb,ix); - HashDbTerm* b; - do { - for (b = BUCKET(tb, ix); b != NULL; b = b->next) { - if (b->hvalue == INVALID_HASH) { - ++kept_items; - } - } - ix = next_slot(tb, ix, &lck); - }while (ix); - return kept_items; -} - int db_create_hash(Process *p, DbTable *tbl) { DbTableHash *tb = &tbl->hash; @@ -2161,10 +2142,38 @@ int db_mark_all_deleted_hash(DbTable *tbl) static void db_print_hash(int to, void *to_arg, int show, DbTable *tbl) { DbTableHash *tb = &tbl->hash; + DbHashStats stats; int i; erts_print(to, to_arg, "Buckets: %d\n", NACTIVE(tb)); - + +#ifdef ERTS_SMP + i = tbl->common.is_thread_safe; + /* If crash dumping we set table to thread safe in order to + avoid taking any locks */ + if (ERTS_IS_CRASH_DUMPING) + tbl->common.is_thread_safe = 1; + + db_calc_stats_hash(&tbl->hash, &stats); + + tbl->common.is_thread_safe = i; +#else + db_calc_stats_hash(&tbl->hash, &stats); +#endif + + erts_print(to, to_arg, "Chain Length Avg: %f\n", stats.avg_chain_len); + erts_print(to, to_arg, "Chain Length Max: %d\n", stats.max_chain_len); + erts_print(to, to_arg, "Chain Length Min: %d\n", stats.min_chain_len); + erts_print(to, to_arg, "Chain Length Std Dev: %f\n", + stats.std_dev_chain_len); + erts_print(to, to_arg, "Chain Length Expected Std Dev: %f\n", + stats.std_dev_expected); + + if (IS_FIXED(tb)) + erts_print(to, to_arg, "Fixed: %d\n", stats.kept_items); + else + erts_print(to, to_arg, "Fixed: false\n"); + if (show) { for (i = 0; i < NACTIVE(tb); i++) { HashDbTerm* list = BUCKET(tb,i); @@ -2890,6 +2899,7 @@ void db_calc_stats_hash(DbTableHash* tb, DbHashStats* stats) erts_smp_rwmtx_t* lck; int sum = 0; int sq_sum = 0; + int kept_items = 0; int ix; int len; @@ -2901,6 +2911,8 @@ void db_calc_stats_hash(DbTableHash* tb, DbHashStats* stats) len = 0; for (b = BUCKET(tb,ix); b!=NULL; b=b->next) { len++; + if (b->hvalue == INVALID_HASH) + ++kept_items; } sum += len; sq_sum += len*len; @@ -2912,7 +2924,8 @@ void db_calc_stats_hash(DbTableHash* tb, DbHashStats* stats) stats->std_dev_chain_len = sqrt((sq_sum - stats->avg_chain_len*sum) / NACTIVE(tb)); /* Expected standard deviation from a good uniform hash function, ie binomial distribution (not taking the linear hashing into acount) */ - stats->std_dev_expected = sqrt(stats->avg_chain_len * (1 - 1.0/NACTIVE(tb))); + stats->std_dev_expected = sqrt(stats->avg_chain_len * (1 - 1.0/NACTIVE(tb))); + stats->kept_items = kept_items; } #ifdef HARDDEBUG diff --git a/erts/emulator/beam/erl_db_hash.h b/erts/emulator/beam/erl_db_hash.h index e68081a5b1..f12cd363b0 100644 --- a/erts/emulator/beam/erl_db_hash.h +++ b/erts/emulator/beam/erl_db_hash.h @@ -104,6 +104,7 @@ typedef struct { float std_dev_expected; int max_chain_len; int min_chain_len; + int kept_items; }DbHashStats; void db_calc_stats_hash(DbTableHash* tb, DbHashStats*); diff --git a/erts/emulator/beam/erl_drv_thread.c b/erts/emulator/beam/erl_drv_thread.c index 147249f751..31b05d22af 100644 --- a/erts/emulator/beam/erl_drv_thread.c +++ b/erts/emulator/beam/erl_drv_thread.c @@ -601,17 +601,14 @@ erl_drv_thread_create(char *name, #ifdef USE_THREADS int res; struct ErlDrvTid_ *dtid; - ethr_thr_opts ethr_opts; + ethr_thr_opts ethr_opts = ETHR_THR_OPTS_DEFAULT_INITER; ethr_thr_opts *use_opts; - ethr_thr_opts def_ethr_opts = ETHR_THR_OPTS_DEFAULT_INITER; if (!opts) use_opts = NULL; else { - sys_memcpy((void *) ðr_opts, - (void *) &def_ethr_opts, - sizeof(ethr_thr_opts)); ethr_opts.suggested_stack_size = opts->suggested_stack_size; + ethr_opts.name = name; use_opts = ðr_opts; } diff --git a/erts/emulator/beam/erl_process.c b/erts/emulator/beam/erl_process.c index 07daabefc7..ba09ee57c2 100644 --- a/erts/emulator/beam/erl_process.c +++ b/erts/emulator/beam/erl_process.c @@ -457,8 +457,7 @@ do { \ static void exec_misc_ops(ErtsRunQueue *); static void print_function_from_pc(int to, void *to_arg, BeamInstr* x); -static int stack_element_dump(int to, void *to_arg, Process* p, Eterm* sp, - int yreg); +static int stack_element_dump(int to, void *to_arg, Eterm* sp, int yreg); static void aux_work_timeout(void *unused); static void aux_work_timeout_early_init(int no_schedulers); @@ -7834,23 +7833,17 @@ erts_start_schedulers(void) Uint actual; Uint wanted = erts_no_schedulers; Uint wanted_no_schedulers = erts_no_schedulers; + char name[16]; ethr_thr_opts opts = ETHR_THR_OPTS_DEFAULT_INITER; opts.detached = 1; -#ifdef ETHR_HAVE_THREAD_NAMES - opts.name = malloc(80); - if (!opts.name) { - ERTS_INTERNAL_ERROR("malloc failed to allocate memory!"); - } -#endif + opts.name = name; #ifdef ERTS_SMP if (erts_runq_supervision_interval) { opts.suggested_stack_size = 16; -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(opts.name, "runq_supervisor"); -#endif + erts_snprintf(opts.name, 16, "runq_supervisor"); erts_atomic_init_nob(&runq_supervisor_sleeping, 0); if (0 != ethr_event_init(&runq_supervision_event)) erl_exit(1, "Failed to create run-queue supervision event\n"); @@ -7877,9 +7870,7 @@ erts_start_schedulers(void) ASSERT(actual == esdp->no - 1); -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(opts.name, "scheduler_%d", actual + 1); -#endif + erts_snprintf(opts.name, 16, "%lu_scheduler", actual + 1); #ifdef __OSE__ /* This should be done in the bind strategy */ @@ -7901,18 +7892,14 @@ erts_start_schedulers(void) int ix; for (ix = 0; ix < erts_no_dirty_cpu_schedulers; ix++) { ErtsSchedulerData *esdp = ERTS_DIRTY_CPU_SCHEDULER_IX(ix); -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(opts.name,"dirty_cpu_scheduler_%d", ix + 1); -#endif + erts_snprintf(opts.name, 16, "%d_dirty_cpu_scheduler", ix + 1); res = ethr_thr_create(&esdp->tid,sched_dirty_cpu_thread_func,(void*)esdp,&opts); if (res != 0) erl_exit(1, "Failed to create dirty cpu scheduler thread %d\n", ix); } for (ix = 0; ix < erts_no_dirty_io_schedulers; ix++) { ErtsSchedulerData *esdp = ERTS_DIRTY_IO_SCHEDULER_IX(ix); -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(opts.name,"dirty_io_scheduler_%d", ix + 1); -#endif + erts_snprintf(opts.name, 16, "%d_dirty_io_scheduler", ix + 1); res = ethr_thr_create(&esdp->tid,sched_dirty_io_thread_func,(void*)esdp,&opts); if (res != 0) erl_exit(1, "Failed to create dirty io scheduler thread %d\n", ix); @@ -7923,9 +7910,7 @@ erts_start_schedulers(void) ERTS_THR_MEMORY_BARRIER; -#ifdef ETHR_HAVE_THREAD_NAMES - sprintf(opts.name, "aux"); -#endif + erts_snprintf(opts.name, 16, "aux"); #ifdef __OSE__ opts.coreNo = 0; @@ -7951,9 +7936,6 @@ erts_start_schedulers(void) erts_send_error_to_logger_nogl(dsbufp); } -#ifdef ETHR_HAVE_THREAD_NAMES - free(opts.name); -#endif } #endif /* ERTS_SMP */ @@ -12154,7 +12136,7 @@ erts_stack_dump(int to, void *to_arg, Process *p) } erts_program_counter_info(to, to_arg, p); for (sp = p->stop; sp < STACK_START(p); sp++) { - yreg = stack_element_dump(to, to_arg, p, sp, yreg); + yreg = stack_element_dump(to, to_arg, sp, yreg); } } @@ -12211,7 +12193,7 @@ print_function_from_pc(int to, void *to_arg, BeamInstr* x) } static int -stack_element_dump(int to, void *to_arg, Process* p, Eterm* sp, int yreg) +stack_element_dump(int to, void *to_arg, Eterm* sp, int yreg) { Eterm x = *sp; @@ -12240,6 +12222,214 @@ stack_element_dump(int to, void *to_arg, Process* p, Eterm* sp, int yreg) } /* + * Print scheduler information + */ +void +erts_print_scheduler_info(int to, void *to_arg, ErtsSchedulerData *esdp) { + int i; + erts_aint32_t flg; + Process *p; + + erts_print(to, to_arg, "=scheduler:%u\n", esdp->no); + +#ifdef ERTS_SMP + flg = erts_smp_atomic32_read_dirty(&esdp->ssi->flags); + erts_print(to, to_arg, "Scheduler Sleep Info Flags: "); + for (i = 0; i < ERTS_SSI_FLGS_MAX && flg; i++) { + erts_aint32_t chk = (1 << i); + if (flg & chk) { + switch (chk) { + case ERTS_SSI_FLG_SLEEPING: + erts_print(to, to_arg, "SLEEPING"); break; + case ERTS_SSI_FLG_POLL_SLEEPING: + erts_print(to, to_arg, "POLL_SLEEPING"); break; + case ERTS_SSI_FLG_TSE_SLEEPING: + erts_print(to, to_arg, "TSE_SLEEPING"); break; + case ERTS_SSI_FLG_WAITING: + erts_print(to, to_arg, "WAITING"); break; + case ERTS_SSI_FLG_SUSPENDED: + erts_print(to, to_arg, "SUSPENDED"); break; + default: + erts_print(to, to_arg, "UNKNOWN(%d)", flg); break; + } + if (flg > chk) + erts_print(to, to_arg, " | "); + flg -= chk; + } + } + erts_print(to, to_arg, "\n"); +#endif + + flg = erts_atomic32_read_dirty(&esdp->ssi->aux_work); + erts_print(to, to_arg, "Scheduler Sleep Info Aux Work: "); + for (i = 0; i < ERTS_SSI_AUX_WORK_MAX && flg; i++) { + erts_aint32_t chk = (1 << i); + if (flg & chk) { + switch (chk) { + case ERTS_SSI_AUX_WORK_DELAYED_AW_WAKEUP: + erts_print(to, to_arg, "DELAYED_AW_WAKEUP"); break; + case ERTS_SSI_AUX_WORK_DD: + erts_print(to, to_arg, "DELAYED_DEALLOC"); break; + case ERTS_SSI_AUX_WORK_DD_THR_PRGR: + erts_print(to, to_arg, "DELAYED_DEALLOC_THR_PRGR"); break; + case ERTS_SSI_AUX_WORK_FIX_ALLOC_DEALLOC: + erts_print(to, to_arg, "FIX_ALLOC_DEALLOC"); break; + case ERTS_SSI_AUX_WORK_FIX_ALLOC_LOWER_LIM: + erts_print(to, to_arg, "FIX_ALLOC_LOWER_LIM"); break; + case ERTS_SSI_AUX_WORK_THR_PRGR_LATER_OP: + erts_print(to, to_arg, "THR_PRGR_LATER_OP"); break; + case ERTS_SSI_AUX_WORK_ASYNC_READY: + erts_print(to, to_arg, "ASYNC_READY"); break; + case ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN: + erts_print(to, to_arg, "ASYNC_READY_CLEAN"); break; + case ERTS_SSI_AUX_WORK_MISC_THR_PRGR: + erts_print(to, to_arg, "MISC_THR_PRGR"); break; + case ERTS_SSI_AUX_WORK_MISC: + erts_print(to, to_arg, "MISC"); break; + case ERTS_SSI_AUX_WORK_CHECK_CHILDREN: + erts_print(to, to_arg, "CHECK_CHILDREN"); break; + case ERTS_SSI_AUX_WORK_SET_TMO: + erts_print(to, to_arg, "SET_TMO"); break; + case ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK: + erts_print(to, to_arg, "MSEG_CACHE_CHECK"); break; + case ERTS_SSI_AUX_WORK_REAP_PORTS: + erts_print(to, to_arg, "REAP_PORTS"); break; + default: + erts_print(to, to_arg, "UNKNOWN(%d)", flg); break; + } + if (flg > chk) + erts_print(to, to_arg, " | "); + flg -= chk; + } + } + erts_print(to, to_arg, "\n"); + + erts_print(to, to_arg, "Current Port: "); + if (esdp->current_port) + erts_print(to, to_arg, "%T", esdp->current_port->common.id); + erts_print(to, to_arg, "\n"); + + p = esdp->current_process; + erts_print(to, to_arg, "Current Process: "); + if (esdp->current_process && !(ERTS_TRACE_FLAGS(p) & F_SENSITIVE)) { + flg = erts_smp_atomic32_read_dirty(&p->state); + erts_print(to, to_arg, "%T\n", p->common.id); + + erts_print(to, to_arg, "Current Process State: "); + erts_dump_process_state(to, to_arg, flg); + + erts_print(to, to_arg, "Current Process Internal State: "); + erts_dump_extended_process_state(to, to_arg, flg); + + erts_print(to, to_arg, "Current Process Program counter: %p (", p->i); + print_function_from_pc(to, to_arg, p->i); + erts_print(to, to_arg, ")\n"); + erts_print(to, to_arg, "Current Process CP: %p (", p->cp); + print_function_from_pc(to, to_arg, p->cp); + erts_print(to, to_arg, ")\n"); + + /* Getting this stacktrace can segfault if we are very very + unlucky if called while a process is being garbage collected. + Therefore we only call this on other schedulers if we either + have protection against segfaults, or we know that the process + is not garbage collecting. It *should* always be safe to call + on a process owned by us, even if it is currently being garbage + collected. + */ + erts_print(to, to_arg, "Current Process Limited Stack Trace:\n"); + erts_limited_stack_trace(to, to_arg, p); + } else + erts_print(to, to_arg, "\n"); + + for (i = 0; i < ERTS_NO_PROC_PRIO_LEVELS; i++) { + erts_print(to, to_arg, "Run Queue "); + switch (i) { + case PRIORITY_MAX: + erts_print(to, to_arg, "Max "); + break; + case PRIORITY_HIGH: + erts_print(to, to_arg, "High "); + break; + case PRIORITY_NORMAL: + erts_print(to, to_arg, "Normal "); + break; + case PRIORITY_LOW: + erts_print(to, to_arg, "Low "); + break; + default: + erts_print(to, to_arg, "Unknown "); + break; + } + erts_print(to, to_arg, "Length: %d\n", + erts_smp_atomic32_read_dirty(&esdp->run_queue->procs.prio_info[i].len)); + } + erts_print(to, to_arg, "Run Queue Port Length: %d\n", + erts_smp_atomic32_read_dirty(&esdp->run_queue->ports.info.len)); + + flg = erts_smp_atomic32_read_dirty(&esdp->run_queue->flags); + erts_print(to, to_arg, "Run Queue Flags: "); + for (i = 0; i < ERTS_RUNQ_FLG_MAX && flg; i++) { + erts_aint32_t chk = (1 << i); + if (flg & chk) { + switch (chk) { + case (1 << PRIORITY_MAX): + erts_print(to, to_arg, "NONEMPTY_MAX"); break; + case (1 << PRIORITY_HIGH): + erts_print(to, to_arg, "NONEMPTY_HIGH"); break; + case (1 << PRIORITY_NORMAL): + erts_print(to, to_arg, "NONEMPTY_NORMAL"); break; + case (1 << PRIORITY_LOW): + erts_print(to, to_arg, "NONEMPTY_LOW"); break; + case (1 << (PRIORITY_MAX + ERTS_RUNQ_FLGS_EMIGRATE_SHFT)): + erts_print(to, to_arg, "EMIGRATE_MAX"); break; + case (1 << (PRIORITY_HIGH + ERTS_RUNQ_FLGS_EMIGRATE_SHFT)): + erts_print(to, to_arg, "EMIGRATE_HIGH"); break; + case (1 << (PRIORITY_NORMAL + ERTS_RUNQ_FLGS_EMIGRATE_SHFT)): + erts_print(to, to_arg, "EMIGRATE_NORMAL"); break; + case (1 << (PRIORITY_LOW + ERTS_RUNQ_FLGS_EMIGRATE_SHFT)): + erts_print(to, to_arg, "EMIGRATE_LOW"); break; + case (1 << (PRIORITY_MAX + ERTS_RUNQ_FLGS_IMMIGRATE_SHFT)): + erts_print(to, to_arg, "IMMIGRATE_MAX"); break; + case (1 << (PRIORITY_HIGH + ERTS_RUNQ_FLGS_IMMIGRATE_SHFT)): + erts_print(to, to_arg, "IMMIGRATE_HIGH"); break; + case (1 << (PRIORITY_NORMAL + ERTS_RUNQ_FLGS_IMMIGRATE_SHFT)): + erts_print(to, to_arg, "IMMIGRATE_NORMAL"); break; + case (1 << (PRIORITY_LOW + ERTS_RUNQ_FLGS_IMMIGRATE_SHFT)): + erts_print(to, to_arg, "IMMIGRATE_LOW"); break; + case (1 << (PRIORITY_MAX + ERTS_RUNQ_FLGS_EVACUATE_SHFT)): + erts_print(to, to_arg, "EVACUATE_MAX"); break; + case (1 << (PRIORITY_HIGH + ERTS_RUNQ_FLGS_EVACUATE_SHFT)): + erts_print(to, to_arg, "EVACUATE_HIGH"); break; + case (1 << (PRIORITY_NORMAL + ERTS_RUNQ_FLGS_EVACUATE_SHFT)): + erts_print(to, to_arg, "EVACUATE_NORMAL"); break; + case (1 << (PRIORITY_LOW + ERTS_RUNQ_FLGS_EVACUATE_SHFT)): + erts_print(to, to_arg, "EVACUATE_LOW"); break; + case ERTS_RUNQ_FLG_OUT_OF_WORK: + erts_print(to, to_arg, "OUT_OF_WORK"); break; + case ERTS_RUNQ_FLG_HALFTIME_OUT_OF_WORK: + erts_print(to, to_arg, "HALFTIME_OUT_OF_WORK"); break; + case ERTS_RUNQ_FLG_SUSPENDED: + erts_print(to, to_arg, "SUSPENDED"); break; + case ERTS_RUNQ_FLG_CHK_CPU_BIND: + erts_print(to, to_arg, "CHK_CPU_BIND"); break; + case ERTS_RUNQ_FLG_INACTIVE: + erts_print(to, to_arg, "INACTIVE"); break; + case ERTS_RUNQ_FLG_NONEMPTY: + erts_print(to, to_arg, "NONEMPTY"); break; + case ERTS_RUNQ_FLG_PROTECTED: + erts_print(to, to_arg, "PROTECTED"); break; + default: + erts_print(to, to_arg, "UNKNOWN(%d)", flg); break; + } + if (flg > chk) + erts_print(to, to_arg, " | "); + flg -= chk; + } + } + erts_print(to, to_arg, "\n"); +} + +/* * A nice system halt closing all open port goes as follows: * 1) This function schedules the aux work ERTS_SSI_AUX_WORK_REAP_PORTS * on all schedulers, then schedules itself out. diff --git a/erts/emulator/beam/erl_process.h b/erts/emulator/beam/erl_process.h index 55009ad904..d12ac792af 100644 --- a/erts/emulator/beam/erl_process.h +++ b/erts/emulator/beam/erl_process.h @@ -170,6 +170,8 @@ extern int erts_sched_thread_suggested_stack_size; #define ERTS_RUNQ_FLG_PROTECTED \ (((Uint32) 1) << (ERTS_RUNQ_FLG_BASE2 + 6)) +#define ERTS_RUNQ_FLG_MAX (ERTS_RUNQ_FLG_BASE2 + 7) + #define ERTS_RUNQ_FLGS_MIGRATION_QMASKS \ (ERTS_RUNQ_FLGS_EMIGRATE_QMASK \ | ERTS_RUNQ_FLGS_IMMIGRATE_QMASK \ @@ -252,6 +254,8 @@ typedef enum { #define ERTS_SSI_FLG_WAITING (((erts_aint32_t) 1) << 3) #define ERTS_SSI_FLG_SUSPENDED (((erts_aint32_t) 1) << 4) +#define ERTS_SSI_FLGS_MAX 5 + #define ERTS_SSI_FLGS_SLEEP_TYPE \ (ERTS_SSI_FLG_TSE_SLEEPING|ERTS_SSI_FLG_POLL_SLEEPING) @@ -283,6 +287,8 @@ typedef enum { #define ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK (((erts_aint32_t) 1) << 12) #define ERTS_SSI_AUX_WORK_REAP_PORTS (((erts_aint32_t) 1) << 13) +#define ERTS_SSI_AUX_WORK_MAX 14 + typedef struct ErtsSchedulerSleepInfo_ ErtsSchedulerSleepInfo; #ifdef ERTS_DIRTY_SCHEDULERS @@ -1075,6 +1081,9 @@ void erts_check_for_holes(Process* p); #define ERTS_PSFLG_DIRTY_IO_PROC ERTS_PSFLG_BIT(19) #define ERTS_PSFLG_DIRTY_CPU_PROC_IN_Q ERTS_PSFLG_BIT(20) #define ERTS_PSFLG_DIRTY_IO_PROC_IN_Q ERTS_PSFLG_BIT(21) +#define ERTS_PSFLG_MAX (ERTS_PSFLGS_ZERO_BIT_OFFSET + 22) +#else +#define ERTS_PSFLG_MAX (ERTS_PSFLGS_ZERO_BIT_OFFSET + 18) #endif #define ERTS_PSFLGS_IN_PRQ_MASK (ERTS_PSFLG_IN_PRQ_MAX \ @@ -1610,7 +1619,11 @@ void erts_cleanup_empty_process(Process* p); void erts_debug_verify_clean_empty_process(Process* p); #endif void erts_stack_dump(int to, void *to_arg, Process *); +void erts_limited_stack_trace(int to, void *to_arg, Process *); void erts_program_counter_info(int to, void *to_arg, Process *); +void erts_print_scheduler_info(int to, void *to_arg, ErtsSchedulerData *esdp); +void erts_dump_extended_process_state(int to, void *to_arg, erts_aint32_t psflg); +void erts_dump_process_state(int to, void *to_arg, erts_aint32_t psflg); Eterm erts_get_process_priority(Process *p); Eterm erts_set_process_priority(Process *p, Eterm prio); diff --git a/erts/emulator/beam/erl_process_dump.c b/erts/emulator/beam/erl_process_dump.c index 2f3cf23b00..36bb6b2f0e 100644 --- a/erts/emulator/beam/erl_process_dump.c +++ b/erts/emulator/beam/erl_process_dump.c @@ -43,8 +43,9 @@ static void dump_process_info(int to, void *to_arg, Process *p); static void dump_element(int to, void *to_arg, Eterm x); static void dump_dist_ext(int to, void *to_arg, ErtsDistExternal *edep); static void dump_element_nl(int to, void *to_arg, Eterm x); -static int stack_element_dump(int to, void *to_arg, Process* p, Eterm* sp, +static int stack_element_dump(int to, void *to_arg, Eterm* sp, int yreg); +static void stack_trace_dump(int to, void *to_arg, Eterm* sp); static void print_function_from_pc(int to, void *to_arg, BeamInstr* x); static void heap_dump(int to, void *to_arg, Eterm x); static void dump_binaries(int to, void *to_arg, Binary* root); @@ -148,7 +149,7 @@ dump_process_info(int to, void *to_arg, Process *p) if ((ERTS_TRACE_FLAGS(p) & F_SENSITIVE) == 0) { erts_print(to, to_arg, "=proc_stack:%T\n", p->common.id); for (sp = p->stop; sp < STACK_START(p); sp++) { - yreg = stack_element_dump(to, to_arg, p, sp, yreg); + yreg = stack_element_dump(to, to_arg, sp, yreg); } erts_print(to, to_arg, "=proc_heap:%T\n", p->common.id); @@ -243,9 +244,65 @@ dump_element_nl(int to, void *to_arg, Eterm x) erts_putc(to, to_arg, '\n'); } +static void +stack_trace_dump(int to, void *to_arg, Eterm *sp) { + Eterm x = *sp; + if (is_CP(x)) { + erts_print(to, to_arg, "%p:", sp); + erts_print(to, to_arg, "SReturn addr 0x%X (", cp_val(x)); + print_function_from_pc(to, to_arg, cp_val(x)); + erts_print(to, to_arg, ")\n"); + } +} + +void +erts_limited_stack_trace(int to, void *to_arg, Process *p) +{ + Eterm* sp; + + + if (ERTS_TRACE_FLAGS(p) & F_SENSITIVE) { + return; + } + + if (STACK_START(p) < STACK_TOP(p)) { + return; + } + + if ((STACK_START(p) - STACK_TOP(p)) < 512) { + if (erts_sys_is_area_readable((char*)STACK_TOP(p), + (char*)STACK_START(p))) + for (sp = STACK_TOP(p); sp < STACK_START(p); sp++) + stack_trace_dump(to, to_arg, sp); + else + erts_print(to, to_arg, "Could not read from stack memory: %p - %p\n", + STACK_TOP(p), STACK_START(p)); + } else { + sp = STACK_TOP(p); + if (erts_sys_is_area_readable((char*)STACK_TOP(p), + (char*)(STACK_TOP(p) + 25))) + for (; sp < (STACK_TOP(p) + 256); sp++) + stack_trace_dump(to, to_arg, sp); + else + erts_print(to, to_arg, "Could not read from stack memory: %p - %p\n", + STACK_TOP(p), STACK_TOP(p) + 256); + + erts_print(to, to_arg, "%p: skipping %d frames\n", + sp, STACK_START(p) - STACK_TOP(p) - 512); + + if (erts_sys_is_area_readable((char*)(STACK_START(p) - 256), + (char*)STACK_START(p))) + for (sp = STACK_START(p) - 256; sp < STACK_START(p); sp++) + stack_trace_dump(to, to_arg, sp); + else + erts_print(to, to_arg, "Could not read from stack memory: %p - %p\n", + STACK_START(p) - 256, STACK_START(p)); + } + +} static int -stack_element_dump(int to, void *to_arg, Process* p, Eterm* sp, int yreg) +stack_element_dump(int to, void *to_arg, Eterm* sp, int yreg) { Eterm x = *sp; @@ -508,3 +565,114 @@ dump_externally(int to, void *to_arg, Eterm term) erts_print(to, to_arg, "%02X", *s++); } } + +void erts_dump_process_state(int to, void *to_arg, erts_aint32_t psflg) { + if (psflg & ERTS_PSFLG_FREE) + erts_print(to, to_arg, "Non Existing\n"); /* Should never happen */ + else if (psflg & ERTS_PSFLG_EXITING) + erts_print(to, to_arg, "Exiting\n"); + else if (psflg & ERTS_PSFLG_GC) { + erts_print(to, to_arg, "Garbing\n"); + } + else if (psflg & ERTS_PSFLG_SUSPENDED) + erts_print(to, to_arg, "Suspended\n"); + else if (psflg & ERTS_PSFLG_RUNNING) { + erts_print(to, to_arg, "Running\n"); + } + else if (psflg & ERTS_PSFLG_ACTIVE) + erts_print(to, to_arg, "Scheduled\n"); + else + erts_print(to, to_arg, "Waiting\n"); +} + +void +erts_dump_extended_process_state(int to, void *to_arg, erts_aint32_t psflg) { + + int i; + + switch (ERTS_PSFLGS_GET_ACT_PRIO(psflg)) { + case PRIORITY_MAX: erts_print(to, to_arg, "ACT_PRIO_MAX | "); break; + case PRIORITY_HIGH: erts_print(to, to_arg, "ACT_PRIO_HIGH | "); break; + case PRIORITY_NORMAL: erts_print(to, to_arg, "ACT_PRIO_NORMAL | "); break; + case PRIORITY_LOW: erts_print(to, to_arg, "ACT_PRIO_LOW | "); break; + } + switch (ERTS_PSFLGS_GET_USR_PRIO(psflg)) { + case PRIORITY_MAX: erts_print(to, to_arg, "USR_PRIO_MAX | "); break; + case PRIORITY_HIGH: erts_print(to, to_arg, "USR_PRIO_HIGH | "); break; + case PRIORITY_NORMAL: erts_print(to, to_arg, "USR_PRIO_NORMAL | "); break; + case PRIORITY_LOW: erts_print(to, to_arg, "USR_PRIO_LOW | "); break; + } + switch (ERTS_PSFLGS_GET_PRQ_PRIO(psflg)) { + case PRIORITY_MAX: erts_print(to, to_arg, "PRQ_PRIO_MAX"); break; + case PRIORITY_HIGH: erts_print(to, to_arg, "PRQ_PRIO_HIGH"); break; + case PRIORITY_NORMAL: erts_print(to, to_arg, "PRQ_PRIO_NORMAL"); break; + case PRIORITY_LOW: erts_print(to, to_arg, "PRQ_PRIO_LOW"); break; + } + + psflg &= ~(ERTS_PSFLGS_ACT_PRIO_MASK | + ERTS_PSFLGS_USR_PRIO_MASK | + ERTS_PSFLGS_PRQ_PRIO_MASK); + + if (psflg) + erts_print(to, to_arg, " | "); + + for (i = 0; i < ERTS_PSFLG_MAX && psflg; i++) { + erts_aint32_t chk = (1 << i); + if (psflg & chk) { + switch (chk) { + case ERTS_PSFLG_IN_PRQ_MAX: + erts_print(to, to_arg, "IN_PRQ_MAX"); break; + case ERTS_PSFLG_IN_PRQ_HIGH: + erts_print(to, to_arg, "IN_PRQ_HIGH"); break; + case ERTS_PSFLG_IN_PRQ_NORMAL: + erts_print(to, to_arg, "IN_PRQ_NORMAL"); break; + case ERTS_PSFLG_IN_PRQ_LOW: + erts_print(to, to_arg, "IN_PRQ_LOW"); break; + case ERTS_PSFLG_FREE: + erts_print(to, to_arg, "FREE"); break; + case ERTS_PSFLG_EXITING: + erts_print(to, to_arg, "EXITING"); break; + case ERTS_PSFLG_PENDING_EXIT: + erts_print(to, to_arg, "PENDING_EXIT"); break; + case ERTS_PSFLG_ACTIVE: + erts_print(to, to_arg, "ACTIVE"); break; + case ERTS_PSFLG_IN_RUNQ: + erts_print(to, to_arg, "IN_RUNQ"); break; + case ERTS_PSFLG_RUNNING: + erts_print(to, to_arg, "RUNNING"); break; + case ERTS_PSFLG_SUSPENDED: + erts_print(to, to_arg, "SUSPENDED"); break; + case ERTS_PSFLG_GC: + erts_print(to, to_arg, "GC"); break; + case ERTS_PSFLG_BOUND: + erts_print(to, to_arg, "BOUND"); break; + case ERTS_PSFLG_TRAP_EXIT: + erts_print(to, to_arg, "TRAP_EXIT"); break; + case ERTS_PSFLG_ACTIVE_SYS: + erts_print(to, to_arg, "ACTIVE_SYS"); break; + case ERTS_PSFLG_RUNNING_SYS: + erts_print(to, to_arg, "RUNNING_SYS"); break; + case ERTS_PSFLG_PROXY: + erts_print(to, to_arg, "PROXY"); break; + case ERTS_PSFLG_DELAYED_SYS: + erts_print(to, to_arg, "DELAYED_SYS"); break; +#ifdef ERTS_DIRTY_SCHEDULERS + case ERTS_PSFLG_DIRTY_CPU_PROC: + erts_print(to, to_arg, "DIRTY_CPU_PROC"); break; + case ERTS_PSFLG_DIRTY_IO_PROC: + erts_print(to, to_arg, "DIRTY_IO_PROC"); break; + case ERTS_PSFLG_DIRTY_CPU_PROC_IN_Q: + erts_print(to, to_arg, "DIRTY_CPU_PROC_IN_Q"); break; + case ERTS_PSFLG_DIRTY_IO_PROC_IN_Q: + erts_print(to, to_arg, "DIRTY_IO_PROC_IN_Q"); break; +#endif + default: + erts_print(to, to_arg, "UNKNOWN(%d)", chk); break; + } + if (psflg > chk) + erts_print(to, to_arg, " | "); + psflg -= chk; + } + } + erts_print(to, to_arg, "\n"); +} diff --git a/erts/emulator/beam/erl_thr_progress.c b/erts/emulator/beam/erl_thr_progress.c index 664c479eb6..c2365c5cf7 100644 --- a/erts/emulator/beam/erl_thr_progress.c +++ b/erts/emulator/beam/erl_thr_progress.c @@ -1335,25 +1335,10 @@ erts_thr_progress_block(void) thr_progress_block(tmp_thr_prgr_data(NULL), 1); } -void -erts_thr_progress_fatal_error_block(SWord timeout, - ErtsThrPrgrData *tmp_tpd_bufp) +int +erts_thr_progress_fatal_error_block(ErtsThrPrgrData *tmp_tpd_bufp) { ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL); - erts_aint32_t bc; - SWord time_left = timeout; - SysTimeval to; - - /* - * Counting poll intervals may give us a too long timeout - * if cpu is busy. If we got tolerant time of day we use it - * to prevent this. - */ - if (!erts_disable_tolerant_timeofday) { - erts_get_timeval(&to); - to.tv_sec += timeout / 1000; - to.tv_sec += timeout % 1000; - } if (!tpd) { /* @@ -1366,9 +1351,26 @@ erts_thr_progress_fatal_error_block(SWord timeout, init_tmp_thr_prgr_data(tpd); } - bc = thr_progress_block(tpd, 0); - if (bc == 0) - return; /* Succefully blocked all managed threads */ + /* Returns number of threads that have not yes been blocked */ + return thr_progress_block(tpd, 0); +} + +void +erts_thr_progress_fatal_error_wait(SWord timeout) { + erts_aint32_t bc; + SWord time_left = timeout; + SysTimeval to; + + /* + * Counting poll intervals may give us a too long timeout + * if cpu is busy. If we got tolerant time of day we use it + * to prevent this. + */ + if (!erts_disable_tolerant_timeofday) { + erts_get_timeval(&to); + to.tv_sec += timeout / 1000; + to.tv_sec += timeout % 1000; + } while (1) { if (erts_milli_sleep(ERTS_THR_PRGR_FTL_ERR_BLCK_POLL_INTERVAL) == 0) diff --git a/erts/emulator/beam/erl_thr_progress.h b/erts/emulator/beam/erl_thr_progress.h index 03ddbd467c..cf11c4e114 100644 --- a/erts/emulator/beam/erl_thr_progress.h +++ b/erts/emulator/beam/erl_thr_progress.h @@ -83,8 +83,8 @@ typedef struct { ErtsThrPrgrLeaderState leader_state; } ErtsThrPrgrData; -void erts_thr_progress_fatal_error_block(SWord timeout, - ErtsThrPrgrData *tmp_tpd_bufp); +int erts_thr_progress_fatal_error_block(ErtsThrPrgrData *tmp_tpd_bufp); +void erts_thr_progress_fatal_error_wait(SWord timeout); #endif /* ERTS_SMP */ diff --git a/erts/emulator/beam/erl_threads.h b/erts/emulator/beam/erl_threads.h index 7214f3ea33..1fd800d524 100644 --- a/erts/emulator/beam/erl_threads.h +++ b/erts/emulator/beam/erl_threads.h @@ -476,6 +476,7 @@ ERTS_GLB_INLINE void erts_thr_detach(erts_tid_t tid); ERTS_GLB_INLINE void erts_thr_exit(void *res); ERTS_GLB_INLINE void erts_thr_install_exit_handler(void (*exit_handler)(void)); ERTS_GLB_INLINE erts_tid_t erts_thr_self(void); +ERTS_GLB_INLINE int erts_thr_getname(erts_tid_t tid, char *buf, size_t len); ERTS_GLB_INLINE int erts_equal_tids(erts_tid_t x, erts_tid_t y); ERTS_GLB_INLINE void erts_mtx_init_x(erts_mtx_t *mtx, char *name, Eterm extra, int enable_lcnt); @@ -656,11 +657,17 @@ ERTS_GLB_INLINE void erts_thr_set_main_status(int, int); ERTS_GLB_INLINE int erts_thr_get_main_status(void); ERTS_GLB_INLINE void erts_thr_yield(void); + #ifdef ETHR_HAVE_ETHR_SIG_FUNCS #define ERTS_THR_HAVE_SIG_FUNCS 1 ERTS_GLB_INLINE void erts_thr_sigmask(int how, const sigset_t *set, sigset_t *oset); ERTS_GLB_INLINE void erts_thr_sigwait(const sigset_t *set, int *sig); + +#ifdef USE_THREADS +ERTS_GLB_INLINE void erts_thr_kill(erts_tid_t tid, int sig); +#endif + #endif /* #ifdef HAVE_ETHR_SIG_FUNCS */ #ifdef USE_THREADS @@ -2129,6 +2136,16 @@ erts_thr_self(void) #endif } +ERTS_GLB_INLINE int +erts_thr_getname(erts_tid_t tid, char *buf, size_t len) +{ +#ifdef USE_THREADS + return ethr_getname(tid, buf, len); +#else + return -1; +#endif +} + ERTS_GLB_INLINE int erts_equal_tids(erts_tid_t x, erts_tid_t y) @@ -3517,6 +3534,15 @@ ERTS_GLB_INLINE void erts_thr_yield(void) #ifdef ETHR_HAVE_ETHR_SIG_FUNCS ERTS_GLB_INLINE void +erts_thr_kill(erts_tid_t tid, int sig) { +#ifdef USE_THREADS + int res = ethr_kill((ethr_tid)tid, sig); + if (res) + erts_thr_fatal_error(res, "killing thread"); +#endif +} + +ERTS_GLB_INLINE void erts_thr_sigmask(int how, const sigset_t *set, sigset_t *oset) { #ifdef USE_THREADS diff --git a/erts/emulator/beam/erl_trace.c b/erts/emulator/beam/erl_trace.c index eba07667ad..2f9969b0e7 100644 --- a/erts/emulator/beam/erl_trace.c +++ b/erts/emulator/beam/erl_trace.c @@ -3492,16 +3492,13 @@ init_sys_msg_dispatcher(void) thr_opts.coreNo = 0; #endif thr_opts.detached = 1; + thr_opts.name = "sys_msg_dispatcher"; init_smq_element_alloc(); sys_message_queue = NULL; sys_message_queue_end = NULL; erts_smp_cnd_init(&smq_cnd); erts_smp_mtx_init(&smq_mtx, "sys_msg_q"); -#ifdef ETHR_HAVE_THREAD_NAMES - thr_opts.name = "sys_msg_dispatcher"; -#endif - erts_smp_thr_create(&sys_msg_dispatcher_tid, sys_msg_dispatcher_func, NULL, diff --git a/erts/emulator/beam/sys.h b/erts/emulator/beam/sys.h index 7661bf2dc7..828f5b427a 100644 --- a/erts/emulator/beam/sys.h +++ b/erts/emulator/beam/sys.h @@ -772,6 +772,8 @@ typedef struct { } ErtsCheckIoDebugInfo; int erts_check_io_debug(ErtsCheckIoDebugInfo *ip); +int erts_sys_is_area_readable(char *start, char *stop); + /* xxxP */ #define SYS_DEFAULT_FLOAT_DECIMALS 20 void init_sys_float(void); @@ -800,6 +802,11 @@ int erts_sys_unsetenv(char *key); char *erts_read_env(char *key); void erts_free_read_env(void *value); +#if defined(ERTS_THR_HAVE_SIG_FUNCS) && !defined(ETHR_UNUSABLE_SIGUSRX) +extern void sys_thr_resume(erts_tid_t tid); +extern void sys_thr_suspend(erts_tid_t tid); +#endif + /* utils.c */ /* Options to sys_alloc_opt */ diff --git a/erts/emulator/drivers/unix/ttsl_drv.c b/erts/emulator/drivers/unix/ttsl_drv.c index be2fee1f25..a5960716f2 100644 --- a/erts/emulator/drivers/unix/ttsl_drv.c +++ b/erts/emulator/drivers/unix/ttsl_drv.c @@ -338,8 +338,8 @@ static ErlDrvData ttysl_start(ErlDrvPort port, char* buf) } #endif DEBUGLOG(("utf8_mode is %s\n",(utf8_mode) ? "on" : "off")); - sys_sigset(SIGCONT, cont); - sys_sigset(SIGWINCH, winch); + sys_signal(SIGCONT, cont); + sys_signal(SIGWINCH, winch); driver_select(port, (ErlDrvEvent)(UWord)ttysl_fd, ERL_DRV_READ|ERL_DRV_USE, 1); ttysl_port = port; @@ -423,8 +423,8 @@ static void ttysl_stop(ErlDrvData ttysl_data) tty_reset(ttysl_fd); driver_select(ttysl_port, (ErlDrvEvent)(UWord)ttysl_fd, ERL_DRV_WRITE|ERL_DRV_READ|ERL_DRV_USE, 0); - sys_sigset(SIGCONT, SIG_DFL); - sys_sigset(SIGWINCH, SIG_DFL); + sys_signal(SIGCONT, SIG_DFL); + sys_signal(SIGWINCH, SIG_DFL); } ttysl_port = (ErlDrvPort)-1; ttysl_fd = -1; @@ -1458,11 +1458,11 @@ static RETSIGTYPE suspend(int sig) exit(1); } - sys_sigset(sig, SIG_DFL); /* Set signal handler to default */ + sys_signal(sig, SIG_DFL); /* Set signal handler to default */ sys_sigrelease(sig); /* Allow 'sig' to come through */ kill(getpid(), sig); /* Send ourselves the signal */ sys_sigblock(sig); /* Reset to old mask */ - sys_sigset(sig, suspend); /* Reset signal handler */ + sys_signal(sig, suspend); /* Reset signal handler */ if (tty_set(ttysl_fd) < 0) { fprintf(stderr,"Can't set tty raw \n"); diff --git a/erts/emulator/sys/unix/erl_unix_sys.h b/erts/emulator/sys/unix/erl_unix_sys.h index b6bca5c0f4..f0050db114 100644 --- a/erts/emulator/sys/unix/erl_unix_sys.h +++ b/erts/emulator/sys/unix/erl_unix_sys.h @@ -45,7 +45,7 @@ #include <fcntl.h> #include "erl_errno.h" #include <signal.h> - +#include <setjmp.h> #if HAVE_SYS_SOCKETIO_H # include <sys/socketio.h> @@ -211,13 +211,8 @@ int sys_stop_hrvtime(void); #define SYS_CLOCK_RESOLUTION 1 /* These are defined in sys.c */ -#if defined(SIG_SIGSET) /* Old SysV */ -RETSIGTYPE (*sys_sigset())(); -#elif defined(SIG_SIGNAL) /* Old BSD */ -RETSIGTYPE (*sys_sigset())(); -#else -RETSIGTYPE (*sys_sigset(int, RETSIGTYPE (*func)(int)))(int); -#endif +typedef void (*SIGFUNC)(int); +extern SIGFUNC sys_signal(int, SIGFUNC); extern void sys_sigrelease(int); extern void sys_sigblock(int); extern void sys_stop_cat(void); @@ -354,4 +349,28 @@ extern int exit_async(void); #define ERTS_EXIT_AFTER_DUMP _exit +#if !defined(__APPLE__) && !defined(__MACH__) +/* Some OS X versions do not allow (ab)using signal handlers like this */ +#define ERTS_HAVE_TRY_CATCH 1 + +/* We try to simulate a try catch in C with the help of signal handlers. + * Only use this as a very last resort, as it is not very portable and + * quite unstable. It is also not thread safe, so make sure that only + * one thread can call this at a time! + */ +extern void erts_sys_sigsegv_handler(int); +extern jmp_buf erts_sys_sigsegv_jmp; +#define ERTS_SYS_TRY_CATCH(EXPR,CATCH) \ + do { \ + SIGFUNC prev_handler = sys_signal(SIGSEGV, \ + erts_sys_sigsegv_handler); \ + if (!setjmp(erts_sys_sigsegv_jmp)) { \ + EXPR; \ + } else { \ + CATCH; \ + } \ + sys_signal(SIGSEGV,prev_handler); \ + } while(0) +#endif + #endif /* #ifndef _ERL_UNIX_SYS_H */ diff --git a/erts/emulator/sys/unix/sys.c b/erts/emulator/sys/unix/sys.c index 47acd132ca..6248651882 100644 --- a/erts/emulator/sys/unix/sys.c +++ b/erts/emulator/sys/unix/sys.c @@ -224,10 +224,15 @@ static erts_smp_atomic_t sys_misc_mem_sz; #if defined(ERTS_SMP) static void smp_sig_notify(char c); static int sig_notify_fds[2] = {-1, -1}; + +static int sig_suspend_fds[2] = {-1, -1}; +#define ERTS_SYS_SUSPEND_SIGNAL SIGUSR2 #elif defined(USE_THREADS) static int async_fd[2]; #endif +jmp_buf erts_sys_sigsegv_jmp; + #if CHLDWTHR || defined(ERTS_SMP) erts_mtx_t chld_stat_mtx; #endif @@ -654,39 +659,7 @@ erl_sys_init(void) /* signal handling */ -#ifdef SIG_SIGSET /* Old SysV */ -RETSIGTYPE (*sys_sigset(sig, func))() -int sig; -RETSIGTYPE (*func)(); -{ - return(sigset(sig, func)); -} -void sys_sigblock(int sig) -{ - sighold(sig); -} -void sys_sigrelease(int sig) -{ - sigrelse(sig); -} -#else /* !SIG_SIGSET */ -#ifdef SIG_SIGNAL /* Old BSD */ -RETSIGTYPE (*sys_sigset(sig, func))(int, int) -int sig; -RETSIGTYPE (*func)(); -{ - return(signal(sig, func)); -} -sys_sigblock(int sig) -{ - sigblock(sig); -} -sys_sigrelease(int sig) -{ - sigsetmask(sigblock(0) & ~sigmask(sig)); -} -#else /* !SIG_SIGNAL */ /* The True Way - POSIX!:-) */ -RETSIGTYPE (*sys_sigset(int sig, RETSIGTYPE (*func)(int)))(int) +SIGFUNC sys_signal(int sig, SIGFUNC func) { struct sigaction act, oact; @@ -719,23 +692,35 @@ void sys_sigrelease(int sig) sigaddset(&mask, sig); sigprocmask(SIG_UNBLOCK, &mask, (sigset_t *)NULL); } -#endif /* !SIG_SIGNAL */ -#endif /* !SIG_SIGSET */ -#if (0) /* not used? -- gordon */ -static void (*break_func)(); -static RETSIGTYPE break_handler(int sig) -{ -#ifdef QNX - /* Turn off SIGCHLD during break processing */ - sys_sigblock(SIGCHLD); -#endif - (*break_func)(); -#ifdef QNX - sys_sigrelease(SIGCHLD); -#endif +void erts_sys_sigsegv_handler(int signo) { + if (signo == SIGSEGV) { + longjmp(erts_sys_sigsegv_jmp, 1); + } +} + +/* + * Function returns 1 if we can read from all values in between + * start and stop. + */ +int +erts_sys_is_area_readable(char *start, char *stop) { + int fds[2]; + if (!pipe(fds)) { + /* We let write try to figure out if the pointers are readable */ + int res = write(fds[1], start, (char*)stop - (char*)start); + if (res == -1) { + close(fds[0]); + close(fds[1]); + return 0; + } + close(fds[0]); + close(fds[1]); + return 1; + } + return 0; + } -#endif /* 0 */ static ERTS_INLINE int prepare_crash_dump(int secs) @@ -797,6 +782,9 @@ prepare_crash_dump(int secs) /* We don't want to close the signal notification pipe... */ if (i == sig_notify_fds[0] || i == sig_notify_fds[1]) continue; + /* We don't want to close the signal syspend pipe... */ + if (i == sig_suspend_fds[0] || i == sig_suspend_fds[1]) + continue; #elif defined(USE_THREADS) /* We don't want to close the async notification pipe... */ if (i == async_fd[0] || i == async_fd[1]) @@ -885,9 +873,23 @@ sigusr1_exit(void) #ifdef ETHR_UNUSABLE_SIGUSRX #warning "Unusable SIGUSR1 & SIGUSR2. Disabling use of these signals" -#endif -#ifndef ETHR_UNUSABLE_SIGUSRX +#else + +#ifdef ERTS_SMP +void +sys_thr_suspend(erts_tid_t tid) { + erts_thr_kill(tid, ERTS_SYS_SUSPEND_SIGNAL); +} + +void +sys_thr_resume(erts_tid_t tid) { + int i = 0, res; + do { + res = write(sig_suspend_fds[1],&i,sizeof(i)); + } while (res < 0 && errno == EAGAIN); +} +#endif #if (defined(SIG_SIGSET) || defined(SIG_SIGNAL)) static RETSIGTYPE user_signal1(void) @@ -902,20 +904,20 @@ static RETSIGTYPE user_signal1(int signum) #endif } -#ifdef QUANTIFY +#ifdef ERTS_SMP #if (defined(SIG_SIGSET) || defined(SIG_SIGNAL)) -static RETSIGTYPE user_signal2(void) +static RETSIGTYPE suspend_signal(void) #else -static RETSIGTYPE user_signal2(int signum) +static RETSIGTYPE suspend_signal(int signum) #endif { -#ifdef ERTS_SMP - smp_sig_notify('2'); -#else - quantify_save_data(); -#endif + int res; + int buf[1]; + do { + res = read(sig_suspend_fds[0], buf, sizeof(int)); + } while (res < 0 && errno == EINTR); } -#endif +#endif /* #ifdef ERTS_SMP */ #endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */ @@ -940,9 +942,9 @@ static RETSIGTYPE do_quit(int signum) /* Disable break */ void erts_set_ignore_break(void) { - sys_sigset(SIGINT, SIG_IGN); - sys_sigset(SIGQUIT, SIG_IGN); - sys_sigset(SIGTSTP, SIG_IGN); + sys_signal(SIGINT, SIG_IGN); + sys_signal(SIGQUIT, SIG_IGN); + sys_signal(SIGTSTP, SIG_IGN); } /* Don't use ctrl-c for break handler but let it be @@ -965,14 +967,14 @@ void erts_replace_intr(void) { void init_break_handler(void) { - sys_sigset(SIGINT, request_break); + sys_signal(SIGINT, request_break); #ifndef ETHR_UNUSABLE_SIGUSRX - sys_sigset(SIGUSR1, user_signal1); -#ifdef QUANTIFY - sys_sigset(SIGUSR2, user_signal2); -#endif + sys_signal(SIGUSR1, user_signal1); +#ifdef ERTS_SMP + sys_signal(ERTS_SYS_SUSPEND_SIGNAL, suspend_signal); +#endif /* #ifdef ERTS_SMP */ #endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */ - sys_sigset(SIGQUIT, do_quit); + sys_signal(SIGQUIT, do_quit); } int sys_max_files(void) @@ -989,8 +991,13 @@ static void block_signals(void) sys_sigblock(SIGINT); #ifndef ETHR_UNUSABLE_SIGUSRX sys_sigblock(SIGUSR1); +#endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */ +#endif /* #ifndef ERTS_SMP */ + +#if defined(ERTS_SMP) && !defined(ETHR_UNUSABLE_SIGUSRX) + sys_sigblock(ERTS_SYS_SUSPEND_SIGNAL); #endif -#endif + } static void unblock_signals(void) @@ -1004,8 +1011,14 @@ static void unblock_signals(void) #ifndef ETHR_UNUSABLE_SIGUSRX sys_sigrelease(SIGUSR1); #endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */ +#endif /* #ifndef ERTS_SMP */ + +#if defined(ERTS_SMP) && !defined(ETHR_UNUSABLE_SIGUSRX) + sys_sigrelease(ERTS_SYS_SUSPEND_SIGNAL); #endif + } + /************************** Time stuff **************************/ #ifdef HAVE_GETHRTIME #ifdef GETHRTIME_WITH_CLOCK_GETTIME @@ -1335,9 +1348,10 @@ static int spawn_init() thr_opts.detached = 0; thr_opts.suggested_stack_size = 0; /* Smallest possible */ + thr_opts.name = "child_waiter"; #endif - sys_sigset(SIGPIPE, SIG_IGN); /* Ignore - we'll handle the write failure */ + sys_signal(SIGPIPE, SIG_IGN); /* Ignore - we'll handle the write failure */ driver_data = (struct driver_data *) erts_alloc(ERTS_ALC_T_DRV_TAB, max_files * sizeof(struct driver_data)); erts_smp_atomic_add_nob(&sys_misc_mem_sz, @@ -1350,7 +1364,7 @@ static int spawn_init() sys_sigblock(SIGCHLD); #endif - sys_sigset(SIGCHLD, onchld); /* Reap children */ + sys_signal(SIGCHLD, onchld); /* Reap children */ #if CHLDWTHR erts_thr_create(&child_waiter_tid, child_waiter, NULL, &thr_opts); @@ -3210,13 +3224,6 @@ signal_dispatcher_thread_func(void *unused) case '1': /* SIGUSR1 */ sigusr1_exit(); break; -#ifdef QUANTIFY - case '2': /* SIGUSR2 */ - quantify_save_data(); /* Might take a substantial amount of - time, but this is a test/debug - build */ - break; -#endif default: erl_exit(ERTS_ABORT_EXIT, "signal-dispatcher thread received unknown " @@ -3234,6 +3241,7 @@ init_smp_sig_notify(void) { erts_smp_thr_opts_t thr_opts = ERTS_SMP_THR_OPTS_DEFAULT_INITER; thr_opts.detached = 1; + thr_opts.name = "sys_sig_dispatcher"; if (pipe(sig_notify_fds) < 0) { erl_exit(ERTS_ABORT_EXIT, @@ -3248,6 +3256,17 @@ init_smp_sig_notify(void) NULL, &thr_opts); } + +static void +init_smp_sig_suspend(void) { + if (pipe(sig_suspend_fds) < 0) { + erl_exit(ERTS_ABORT_EXIT, + "Failed to create sig_suspend pipe: %s (%d)\n", + erl_errno_id(errno), + errno); + } +} + #ifdef __DARWIN__ int erts_darwin_main_thread_pipe[2]; @@ -3275,9 +3294,11 @@ erts_sys_main_thread(void) #endif smp_sig_notify(0); /* Notify initialized */ - while (1) { - /* Wait for a signal to arrive... */ + + /* Wait for a signal to arrive... */ + #ifdef __DARWIN__ + while (1) { /* * The wx driver needs to be able to steal the main thread for Cocoa to * work properly. @@ -3292,12 +3313,24 @@ erts_sys_main_thread(void) void* (*func)(void*); void* arg; void *resp; - read(erts_darwin_main_thread_pipe[0],&func,sizeof(void* (*)(void*))); - read(erts_darwin_main_thread_pipe[0],&arg, sizeof(void*)); + res = read(erts_darwin_main_thread_pipe[0],&func,sizeof(void* (*)(void*))); + if (res != sizeof(void* (*)(void*))) + break; + res = read(erts_darwin_main_thread_pipe[0],&arg,sizeof(void*)); + if (res != sizeof(void*)) + break; resp = (*func)(arg); write(erts_darwin_main_thread_result_pipe[1],&resp,sizeof(void *)); } -#else + + if (res == -1 && errno != EINTR) + break; + } + /* Something broke with the main thread pipe, so we ignore it for now. + Most probably erts has closed this pipe and is about to exit. */ +#endif /* #ifdef __DARWIN__ */ + + while (1) { #ifdef DEBUG int res = #else @@ -3306,7 +3339,6 @@ erts_sys_main_thread(void) select(0, NULL, NULL, NULL, NULL); ASSERT(res < 0); ASSERT(errno == EINTR); -#endif } } @@ -3398,6 +3430,7 @@ erl_sys_args(int* argc, char** argv) #ifdef ERTS_SMP init_smp_sig_notify(); + init_smp_sig_suspend(); #endif /* Handled arguments have been marked with NULL. Slide arguments diff --git a/erts/emulator/sys/unix/sys_float.c b/erts/emulator/sys/unix/sys_float.c index cafeab547e..2ffa649767 100644 --- a/erts/emulator/sys/unix/sys_float.c +++ b/erts/emulator/sys/unix/sys_float.c @@ -32,7 +32,7 @@ void erts_sys_init_float(void) { # ifdef SIGFPE - sys_sigset(SIGFPE, SIG_IGN); /* Ignore so we can test for NaN and Inf */ + sys_signal(SIGFPE, SIG_IGN); /* Ignore so we can test for NaN and Inf */ # endif } @@ -667,7 +667,7 @@ static void fpe_sig_handler(int sig) static void erts_thread_catch_fp_exceptions(void) { - sys_sigset(SIGFPE, fpe_sig_handler); + sys_signal(SIGFPE, fpe_sig_handler); unmask_fpe(); } diff --git a/erts/emulator/sys/win32/erl_win_sys.h b/erts/emulator/sys/win32/erl_win_sys.h index 838f0c61eb..fde32c8684 100644 --- a/erts/emulator/sys/win32/erl_win_sys.h +++ b/erts/emulator/sys/win32/erl_win_sys.h @@ -236,4 +236,16 @@ typedef long ssize_t; int init_async(int); int exit_async(void); #endif + +#define ERTS_HAVE_TRY_CATCH 1 + +#define ERTS_SYS_TRY_CATCH(EXPR,CATCH) \ + __try { \ + EXPR; \ + } \ + __except(GetExceptionCode() == EXCEPTION_ACCESS_VIOLATION ? EXCEPTION_EXECUTE_HANDLER : EXCEPTION_CONTINUE_SEARCH) \ + { \ + CATCH; \ + } + #endif diff --git a/erts/emulator/sys/win32/sys.c b/erts/emulator/sys/win32/sys.c index 164ef95629..5d51659b4e 100644 --- a/erts/emulator/sys/win32/sys.c +++ b/erts/emulator/sys/win32/sys.c @@ -247,6 +247,27 @@ void erl_sys_args(int* argc, char** argv) #endif } +/* + * Function returns 1 if we can read from all values in between + * start and stop. + */ +int +erts_sys_is_area_readable(char *start, char *stop) { + volatile char tmp; + __try + { + while(start < stop) { + tmp = *start; + start++; + } + } + __except(EXCEPTION_EXECUTE_HANDLER) + { + return 0; + } + return 1; +} + int erts_sys_prepare_crash_dump(int secs) { Port *heart_port; diff --git a/erts/emulator/test/bif_SUITE.erl b/erts/emulator/test/bif_SUITE.erl index fbc229bc53..fc9bdae0a0 100644 --- a/erts/emulator/test/bif_SUITE.erl +++ b/erts/emulator/test/bif_SUITE.erl @@ -20,6 +20,7 @@ -module(bif_SUITE). -include_lib("test_server/include/test_server.hrl"). +-include_lib("kernel/include/file.hrl"). -export([all/0, suite/0,groups/0,init_per_suite/1, end_per_suite/1, init_per_group/2,end_per_group/2, @@ -681,8 +682,38 @@ erlang_halt(Config) when is_list(Config) -> {badrpc,nodedown} = rpc:call(N2, erlang, halt, [0]), {ok,N3} = slave:start(H, halt_node3), {badrpc,nodedown} = rpc:call(N3, erlang, halt, [0,[]]), - ok. + % This test triggers a segfault when dumping a crash dump + % to make sure that we can handle it properly. + {ok,N4} = slave:start(H, halt_node4), + CrashDump = filename:join(proplists:get_value(priv_dir,Config), + "segfault_erl_crash.dump"), + true = rpc:call(N4, os, putenv, ["ERL_CRASH_DUMP",CrashDump]), + false = rpc:call(N4, erts_debug, set_internal_state, + [available_internal_state, true]), + {badrpc,nodedown} = rpc:call(N4, erts_debug, set_internal_state, + [broken_halt, "Validate correct crash dump"]), + ok = wait_until_stable_size(CrashDump,-1), + {ok, Bin} = file:read_file(CrashDump), + case {string:str(binary_to_list(Bin),"\n=end\n"), + string:str(binary_to_list(Bin),"\r\n=end\r\n")} of + {0,0} -> ct:fail("Could not find end marker in crash dump"); + _ -> ok + end. + +wait_until_stable_size(_File,-10) -> + {error,enoent}; +wait_until_stable_size(File,PrevSz) -> + timer:sleep(250), + case file:read_file_info(File) of + {error,enoent} -> + wait_until_stable_size(File,PrevSz-1); + {ok,#file_info{size = PrevSz }} when PrevSz /= -1 -> + io:format("Crashdump file size was: ~p (~s)~n",[PrevSz,File]), + ok; + {ok,#file_info{size = NewSz }} -> + wait_until_stable_size(File,NewSz) + end. %% Helpers diff --git a/erts/include/internal/ethread.h b/erts/include/internal/ethread.h index ad5d05704c..0d9a4a4305 100644 --- a/erts/include/internal/ethread.h +++ b/erts/include/internal/ethread.h @@ -214,8 +214,6 @@ typedef OSPPDKEY ethr_tsd_key; /* Out own RW mutexes are probably faster, but use OSEs mutexes */ #define ETHR_USE_OWN_RWMTX_IMPL__ -#define ETHR_HAVE_THREAD_NAMES - #else /* No supported thread lib found */ #ifdef ETHR_NO_SUPP_THR_LIB_NOT_FATAL @@ -501,21 +499,18 @@ typedef struct { typedef struct { int detached; /* boolean (default false) */ int suggested_stack_size; /* kilo words (default sys dependent) */ + char *name; /* max 14 char long (default no-name) */ #ifdef ETHR_OSE_THREADS - char *name; U32 coreNo; #endif } ethr_thr_opts; #if defined(ETHR_OSE_THREADS) -/* Default ethr name is big as we want to be able to sprint stuff in there */ -#define ETHR_THR_OPTS_DEFAULT_INITER \ - {0, -1, "ethread", 0} +#define ETHR_THR_OPTS_DEFAULT_INITER {0, -1, NULL, 0} #else -#define ETHR_THR_OPTS_DEFAULT_INITER {0, -1} +#define ETHR_THR_OPTS_DEFAULT_INITER {0, -1, NULL} #endif - #if !defined(ETHR_TRY_INLINE_FUNCS) || defined(ETHR_AUX_IMPL__) # define ETHR_NEED_SPINLOCK_PROTOTYPES__ # define ETHR_NEED_RWSPINLOCK_PROTOTYPES__ @@ -529,6 +524,8 @@ int ethr_thr_join(ethr_tid, void **); int ethr_thr_detach(ethr_tid); void ethr_thr_exit(void *); ethr_tid ethr_self(void); +int ethr_getname(ethr_tid, char *, size_t); +void ethr_setname(char *); int ethr_equal_tids(ethr_tid, ethr_tid); int ethr_tsd_key_create(ethr_tsd_key *,char *); @@ -540,6 +537,7 @@ void *ethr_tsd_get(ethr_tsd_key); #include <signal.h> int ethr_sigmask(int how, const sigset_t *set, sigset_t *oset); int ethr_sigwait(const sigset_t *set, int *sig); +int ethr_kill(const ethr_tid tid, const int sig); #endif void ethr_compiler_barrier(void); diff --git a/erts/lib_src/pthread/ethread.c b/erts/lib_src/pthread/ethread.c index 79784c5b84..7cf38580c5 100644 --- a/erts/lib_src/pthread/ethread.c +++ b/erts/lib_src/pthread/ethread.c @@ -42,6 +42,7 @@ #include <sys/types.h> #include <unistd.h> #include <signal.h> +#include <string.h> #include <limits.h> @@ -77,6 +78,8 @@ typedef struct { void *(*thr_func)(void *); void *arg; void *prep_func_res; + char *name; + char name_buff[16]; } ethr_thr_wrap_data__; static void *thr_wrapper(void *vtwd) @@ -98,6 +101,8 @@ static void *thr_wrapper(void *vtwd) tsep = twd->tse; /* We aren't allowed to follow twd after result has been set! */ + if (twd->name) + ethr_setname(twd->name); ethr_atomic32_set(&twd->result, result); @@ -315,6 +320,12 @@ ethr_thr_create(ethr_tid *tid, void * (*func)(void *), void *arg, twd.thr_func = func; twd.arg = arg; + if (opts && opts->name) { + snprintf(twd.name_buff, 16, "%s", opts->name); + twd.name = twd.name_buff; + } else + twd.name = NULL; + res = pthread_attr_init(&attr); if (res != 0) return res; @@ -445,6 +456,30 @@ ethr_self(void) } int +ethr_getname(ethr_tid tid, char *buf, size_t len) +{ +#if defined(ETHR_HAVE_PTHREAD_GETNAME_NP_3) + return pthread_getname_np((pthread_t) tid, buf, len); +#elif defined(ETHR_HAVE_PTHREAD_GETNAME_NP_2) + return pthread_getname_np((pthread_t) tid, buf); +#else + return ENOSYS; +#endif +} + +void +ethr_setname(char *name) +{ +#if defined(ETHR_HAVE_PTHREAD_SETNAME_NP_2) + pthread_setname_np(ethr_self(), name); +#elif defined(ETHR_HAVE_PTHREAD_SET_NAME_NP_2) + pthread_set_name_np(ethr_self(), name); +#elif defined(ETHR_HAVE_PTHREAD_SETNAME_NP_1) + pthread_setname_np(name); +#endif +} + +int ethr_equal_tids(ethr_tid tid1, ethr_tid tid2) { return pthread_equal((pthread_t) tid1, (pthread_t) tid2); @@ -565,6 +600,17 @@ int ethr_sigwait(const sigset_t *set, int *sig) return 0; } +int ethr_kill(const ethr_tid tid, const int sig) +{ +#if ETHR_XCHK + if (ethr_not_inited__) { + ETHR_ASSERT(0); + return EACCES; + } +#endif + return pthread_kill((const pthread_t)tid, sig); +} + #endif /* #if ETHR_HAVE_ETHR_SIG_FUNCS */ ETHR_IMPL_NORETURN__ diff --git a/erts/lib_src/win/ethread.c b/erts/lib_src/win/ethread.c index 14d0b6deff..fe5d4a327f 100644 --- a/erts/lib_src/win/ethread.c +++ b/erts/lib_src/win/ethread.c @@ -508,6 +508,19 @@ ethr_self(void) return *tid; } +/* getname and setname are not available on windows */ +int +ethr_getname(ethr_tid tid, char *buf, size_t len) +{ + return ENOSYS; +} + +void +ethr_setname(char *name) +{ + return; +} + int ethr_equal_tids(ethr_tid tid1, ethr_tid tid2) { |