1 /*
2 * Copyright (c) 2003, 2025, Oracle and/or its affiliates. All rights reserved.
3 * Copyright (c) 2017, 2024 SAP SE. All rights reserved.
4 * Copyright (c) 2023, 2025, Red Hat, Inc. and/or its affiliates.
5 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 *
7 * This code is free software; you can redistribute it and/or modify it
8 * under the terms of the GNU General Public License version 2 only, as
9 * published by the Free Software Foundation.
10 *
11 * This code is distributed in the hope that it will be useful, but WITHOUT
12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 * version 2 for more details (a copy is included in the LICENSE file that
15 * accompanied this code).
16 *
17 * You should have received a copy of the GNU General Public License version
18 * 2 along with this work; if not, write to the Free Software Foundation,
19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 *
21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22 * or visit www.oracle.com if you need additional information or have any
23 * questions.
24 *
25 */
26
27 #include "cds/aotMetaspace.hpp"
28 #include "code/codeCache.hpp"
29 #include "compiler/compilationFailureInfo.hpp"
30 #include "compiler/compilationMemoryStatistic.hpp"
31 #include "compiler/compileBroker.hpp"
32 #include "compiler/disassembler.hpp"
33 #include "gc/shared/gcConfig.hpp"
34 #include "gc/shared/gcLogPrecious.hpp"
35 #include "jvm.h"
36 #include "logging/logConfiguration.hpp"
37 #include "memory/allocation.hpp"
38 #include "memory/metaspace.hpp"
39 #include "memory/metaspaceUtils.hpp"
40 #include "memory/resourceArea.inline.hpp"
41 #include "memory/universe.hpp"
42 #include "nmt/memTracker.hpp"
43 #include "oops/compressedOops.hpp"
44 #include "prims/whitebox.hpp"
45 #include "runtime/arguments.hpp"
46 #include "runtime/atomicAccess.hpp"
47 #include "runtime/flags/jvmFlag.hpp"
48 #include "runtime/frame.inline.hpp"
49 #include "runtime/init.hpp"
50 #include "runtime/javaThread.inline.hpp"
51 #include "runtime/os.inline.hpp"
52 #include "runtime/osThread.hpp"
53 #include "runtime/safefetch.hpp"
54 #include "runtime/safepointMechanism.hpp"
55 #include "runtime/stackFrameStream.inline.hpp"
56 #include "runtime/stackOverflow.hpp"
57 #include "runtime/threads.hpp"
58 #include "runtime/threadSMR.hpp"
59 #include "runtime/trimNativeHeap.hpp"
60 #include "runtime/vm_version.hpp"
61 #include "runtime/vmOperations.hpp"
62 #include "runtime/vmThread.hpp"
63 #include "sanitizers/address.hpp"
64 #include "sanitizers/ub.hpp"
65 #include "utilities/debug.hpp"
66 #include "utilities/decoder.hpp"
67 #include "utilities/defaultStream.hpp"
68 #include "utilities/events.hpp"
69 #include "utilities/globalDefinitions.hpp"
70 #include "utilities/macros.hpp"
71 #include "utilities/nativeStackPrinter.hpp"
72 #include "utilities/ostream.hpp"
73 #include "utilities/vmError.hpp"
74 #if INCLUDE_JFR
75 #include "jfr/jfr.hpp"
76 #endif
77 #if INCLUDE_JVMCI
78 #include "jvmci/jvmci.hpp"
79 #endif
80
81 #ifndef PRODUCT
82 #include <signal.h>
83 #endif // PRODUCT
84
85 bool VMError::coredump_status;
86 char VMError::coredump_message[O_BUFLEN];
87 int VMError::_current_step;
88 const char* VMError::_current_step_info;
89 volatile jlong VMError::_reporting_start_time = -1;
90 volatile bool VMError::_reporting_did_timeout = false;
91 volatile jlong VMError::_step_start_time = -1;
92 volatile bool VMError::_step_did_timeout = false;
93 volatile bool VMError::_step_did_succeed = false;
94 volatile intptr_t VMError::_first_error_tid = -1;
95 int VMError::_id;
96 const char* VMError::_message;
97 char VMError::_detail_msg[1024];
98 Thread* VMError::_thread;
99 address VMError::_pc;
100 const void* VMError::_siginfo;
101 const void* VMError::_context;
102 bool VMError::_print_stack_from_frame_used = false;
103 const char* VMError::_filename;
104 int VMError::_lineno;
105 size_t VMError::_size;
106 const size_t VMError::_reattempt_required_stack_headroom = 64 * K;
107 const intptr_t VMError::segfault_address = pd_segfault_address;
108 Thread* volatile VMError::_handshake_timed_out_thread = nullptr;
109 Thread* volatile VMError::_safepoint_timed_out_thread = nullptr;
110
111 // List of environment variables that should be reported in error log file.
112 static const char* env_list[] = {
113 // All platforms
114 "JAVA_HOME", "JAVA_TOOL_OPTIONS", "_JAVA_OPTIONS", "CLASSPATH",
115 "JDK_AOT_VM_OPTIONS",
116 "JAVA_OPTS", "PATH", "USERNAME",
117
118 "XDG_CACHE_HOME", "XDG_CONFIG_HOME", "FC_LANG", "FONTCONFIG_USE_MMAP",
119
120 // Env variables that are defined on Linux/BSD
121 "LD_LIBRARY_PATH", "LD_PRELOAD", "SHELL", "DISPLAY", "WAYLAND_DISPLAY",
122 "HOSTTYPE", "OSTYPE", "ARCH", "MACHTYPE",
123 "LANG", "LC_ALL", "LC_CTYPE", "LC_NUMERIC", "LC_TIME",
124 "TERM", "TMPDIR", "TZ",
125
126 // defined on AIX
127 "LIBPATH", "LDR_PRELOAD", "LDR_PRELOAD64",
128
129 // defined on Linux/AIX/BSD
130 "_JAVA_SR_SIGNUM",
131
132 // defined on Darwin
133 "DYLD_LIBRARY_PATH", "DYLD_FALLBACK_LIBRARY_PATH",
134 "DYLD_FRAMEWORK_PATH", "DYLD_FALLBACK_FRAMEWORK_PATH",
135 "DYLD_INSERT_LIBRARIES",
136
137 // defined on Windows
138 "OS", "PROCESSOR_IDENTIFIER", "_ALT_JAVA_HOME_DIR", "TMP", "TEMP",
139
140 nullptr // End marker.
141 };
142
143 // A simple parser for lists of commands such as -XX:OnError and -XX:OnOutOfMemoryError
144 // Command list (ptr) is expected to be a sequence of commands delineated by semicolons and/or newlines.
145 // Usage:
146 // ptr = OnError;
147 // while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr) != nullptr)
148 // ... ...
149 static char* next_OnError_command(char* buf, int buflen, const char** ptr) {
150 if (ptr == nullptr || *ptr == nullptr) return nullptr;
151
152 const char* cmd = *ptr;
153
154 // skip leading blanks, ';' or newlines
155 while (*cmd == ' ' || *cmd == ';' || *cmd == '\n') cmd++;
156
157 if (*cmd == '\0') return nullptr;
158
159 const char * cmdend = cmd;
160 while (*cmdend != '\0' && *cmdend != ';' && *cmdend != '\n') cmdend++;
161
162 Arguments::copy_expand_pid(cmd, cmdend - cmd, buf, buflen);
163
164 *ptr = (*cmdend == '\0' ? cmdend : cmdend + 1);
165 return buf;
166 }
167
168 static void print_bug_submit_message(outputStream *out, Thread *thread) {
169 if (out == nullptr) return;
170 const char *url = Arguments::java_vendor_url_bug();
171 if (url == nullptr || *url == '\0')
172 url = JDK_Version::runtime_vendor_vm_bug_url();
173 if (url != nullptr && *url != '\0') {
174 out->print_raw_cr("# If you would like to submit a bug report, please visit:");
175 out->print_raw ("# ");
176 out->print_raw_cr(url);
177 }
178 // If the crash is in native code, encourage user to submit a bug to the
179 // provider of that code.
180 if (thread && thread->is_Java_thread() &&
181 !thread->is_hidden_from_external_view()) {
182 if (JavaThread::cast(thread)->thread_state() == _thread_in_native) {
183 out->print_cr("# The crash happened outside the Java Virtual Machine in native code.\n# See problematic frame for where to report the bug.");
184 }
185 }
186 out->print_raw_cr("#");
187 }
188
189 static bool stack_has_headroom(size_t headroom) {
190 size_t stack_size = 0;
191 address stack_base = nullptr;
192 os::current_stack_base_and_size(&stack_base, &stack_size);
193
194 const size_t guard_size = StackOverflow::stack_guard_zone_size();
195 const size_t unguarded_stack_size = stack_size - guard_size;
196
197 if (unguarded_stack_size < headroom) {
198 return false;
199 }
200
201 const address unguarded_stack_end = stack_base - unguarded_stack_size;
202 const address stack_pointer = os::current_stack_pointer();
203
204 return stack_pointer >= unguarded_stack_end + headroom;
205 }
206
207 #ifdef ASSERT
208 PRAGMA_DIAG_PUSH
209 PRAGMA_INFINITE_RECURSION_IGNORED
210 void VMError::reattempt_test_hit_stack_limit(outputStream* st) {
211 if (stack_has_headroom(_reattempt_required_stack_headroom)) {
212 // Use all but (_reattempt_required_stack_headroom - K) unguarded stack space.
213 size_t stack_size = 0;
214 address stack_base = nullptr;
215 os::current_stack_base_and_size(&stack_base, &stack_size);
216
217 const size_t guard_size = StackOverflow::stack_guard_zone_size();
218 const address stack_pointer = os::current_stack_pointer();
219
220 const size_t unguarded_stack_size = stack_size - guard_size;
221 const address unguarded_stack_end = stack_base - unguarded_stack_size;
222 const size_t available_headroom = stack_pointer - unguarded_stack_end;
223 const size_t allocation_size = available_headroom - _reattempt_required_stack_headroom + K;
224
225 st->print_cr("Current Stack Pointer: " PTR_FORMAT " alloca %zu"
226 " of %zu bytes available unguarded stack space",
227 p2i(stack_pointer), allocation_size, available_headroom);
228
229 // Allocate byte blob on the stack. Make pointer volatile to avoid having
230 // the compiler removing later reads.
231 volatile char* stack_buffer = static_cast<char*>(alloca(allocation_size));
232 // Initialize the last byte.
233 stack_buffer[allocation_size - 1] = '\0';
234 // Recursive call should hit the stack limit.
235 reattempt_test_hit_stack_limit(st);
236 // Perform a volatile read of the last byte to avoid having the complier
237 // remove the allocation.
238 static_cast<void>(stack_buffer[allocation_size - 1] == '\0');
239 }
240 controlled_crash(14);
241 }
242 PRAGMA_DIAG_POP
243 #endif // ASSERT
244
245 bool VMError::can_reattempt_step(const char* &stop_reason) {
246 if (!stack_has_headroom(_reattempt_required_stack_headroom)) {
247 stop_reason = "Stack headroom limit reached";
248 return false;
249 }
250
251 if (_step_did_timeout) {
252 stop_reason = "Step time limit reached";
253 return false;
254 }
255
256 return true;
257 }
258
259 void VMError::record_coredump_status(const char* message, bool status) {
260 coredump_status = status;
261 strncpy(coredump_message, message, sizeof(coredump_message));
262 coredump_message[sizeof(coredump_message)-1] = 0;
263 }
264
265 // Return a string to describe the error
266 char* VMError::error_string(char* buf, int buflen) {
267 char signame_buf[64];
268 const char *signame = os::exception_name(_id, signame_buf, sizeof(signame_buf));
269
270 if (signame) {
271 jio_snprintf(buf, buflen,
272 "%s (0x%x) at pc=" PTR_FORMAT ", pid=%d, tid=%zu",
273 signame, _id, p2i(_pc),
274 os::current_process_id(), os::current_thread_id());
275 } else if (_filename != nullptr && _lineno > 0) {
276 // skip directory names
277 int n = jio_snprintf(buf, buflen,
278 "Internal Error at %s:%d, pid=%d, tid=%zu",
279 get_filename_only(), _lineno,
280 os::current_process_id(), os::current_thread_id());
281 if (n >= 0 && n < buflen && _message) {
282 if (strlen(_detail_msg) > 0) {
283 jio_snprintf(buf + n, buflen - n, "%s%s: %s",
284 os::line_separator(), _message, _detail_msg);
285 } else {
286 jio_snprintf(buf + n, buflen - n, "%sError: %s",
287 os::line_separator(), _message);
288 }
289 }
290 } else {
291 jio_snprintf(buf, buflen,
292 "Internal Error (0x%x), pid=%d, tid=%zu",
293 _id, os::current_process_id(), os::current_thread_id());
294 }
295
296 return buf;
297 }
298
299 void VMError::print_stack_trace(outputStream* st, JavaThread* jt,
300 char* buf, int buflen, bool verbose) {
301 #ifdef ZERO
302 if (jt->zero_stack()->sp() && jt->top_zero_frame()) {
303 // StackFrameStream uses the frame anchor, which may not have
304 // been set up. This can be done at any time in Zero, however,
305 // so if it hasn't been set up then we just set it up now and
306 // clear it again when we're done.
307 bool has_last_Java_frame = jt->has_last_Java_frame();
308 if (!has_last_Java_frame)
309 jt->set_last_Java_frame();
310 st->print("Java frames:");
311 st->cr();
312
313 // Print the frames
314 StackFrameStream sfs(jt, true /* update */, true /* process_frames */);
315 for(int i = 0; !sfs.is_done(); sfs.next(), i++) {
316 sfs.current()->zero_print_on_error(i, st, buf, buflen);
317 st->cr();
318 }
319
320 // Reset the frame anchor if necessary
321 if (!has_last_Java_frame)
322 jt->reset_last_Java_frame();
323 }
324 #else
325 if (jt->has_last_Java_frame()) {
326 st->print_cr("Java frames: (J=compiled Java code, j=interpreted, Vv=VM code)");
327 for (StackFrameStream sfs(jt, true /* update */, true /* process_frames */); !sfs.is_done(); sfs.next()) {
328 sfs.current()->print_on_error(st, buf, buflen, verbose);
329 st->cr();
330 }
331 }
332 #endif // ZERO
333 }
334
335 const char* VMError::get_filename_only() {
336 char separator = os::file_separator()[0];
337 const char* p = strrchr(_filename, separator);
338 return p ? p + 1 : _filename;
339 }
340
341 /**
342 * Adds `value` to `list` iff it's not already present and there is sufficient
343 * capacity (i.e. length(list) < `list_capacity`). The length of the list
344 * is the index of the first nullptr entry or `list_capacity` if there are
345 * no nullptr entries.
346 *
347 * @ return true if the value was added, false otherwise
348 */
349 static bool add_if_absent(address value, address* list, int list_capacity) {
350 for (int i = 0; i < list_capacity; i++) {
351 if (list[i] == value) {
352 return false;
353 }
354 if (list[i] == nullptr) {
355 list[i] = value;
356 if (i + 1 < list_capacity) {
357 list[i + 1] = nullptr;
358 }
359 return true;
360 }
361 }
362 return false;
363 }
364
365 /**
366 * Prints the VM generated code unit, if any, containing `pc` if it has not already
367 * been printed. If the code unit is an InterpreterCodelet or StubCodeDesc, it is
368 * only printed if `is_crash_pc` is true.
369 *
370 * @param printed array of code units that have already been printed (delimited by nullptr entry)
371 * @param printed_capacity the capacity of `printed`
372 * @return true if the code unit was printed, false otherwise
373 */
374 static bool print_code(outputStream* st, Thread* thread, address pc, bool is_crash_pc,
375 address* printed, int printed_capacity) {
376 if (Interpreter::contains(pc)) {
377 if (is_crash_pc) {
378 // The interpreter CodeBlob is very large so try to print the codelet instead.
379 InterpreterCodelet* codelet = Interpreter::codelet_containing(pc);
380 if (codelet != nullptr) {
381 if (add_if_absent((address) codelet, printed, printed_capacity)) {
382 codelet->print_on(st);
383 Disassembler::decode(codelet->code_begin(), codelet->code_end(), st);
384 return true;
385 }
386 }
387 }
388 } else {
389 StubCodeDesc* desc = StubCodeDesc::desc_for(pc);
390 if (desc != nullptr) {
391 if (is_crash_pc) {
392 if (add_if_absent((address) desc, printed, printed_capacity)) {
393 desc->print_on(st);
394 Disassembler::decode(desc->begin(), desc->end(), st);
395 return true;
396 }
397 }
398 } else if (thread != nullptr) {
399 CodeBlob* cb = CodeCache::find_blob(pc);
400 if (cb != nullptr && add_if_absent((address) cb, printed, printed_capacity)) {
401 // Disassembling nmethod will incur resource memory allocation,
402 // only do so when thread is valid.
403 ResourceMark rm(thread);
404 Disassembler::decode(cb, st);
405 st->cr();
406 return true;
407 }
408 }
409 }
410 return false;
411 }
412
413 // Like above, but only try to figure out a short name. Return nullptr if not found.
414 static const char* find_code_name(address pc) {
415 if (Interpreter::contains(pc)) {
416 InterpreterCodelet* codelet = Interpreter::codelet_containing(pc);
417 if (codelet != nullptr) {
418 return codelet->description();
419 }
420 } else {
421 StubCodeDesc* desc = StubCodeDesc::desc_for(pc);
422 if (desc != nullptr) {
423 return desc->name();
424 } else {
425 CodeBlob* cb = CodeCache::find_blob(pc);
426 if (cb != nullptr) {
427 return cb->name();
428 }
429 }
430 }
431 return nullptr;
432 }
433
434 static void print_oom_reasons(outputStream* st) {
435 st->print_cr("# Possible reasons:");
436 st->print_cr("# The system is out of physical RAM or swap space");
437 #ifdef LINUX
438 st->print_cr("# This process has exceeded the maximum number of memory mappings (check below");
439 st->print_cr("# for `/proc/sys/vm/max_map_count` and `Total number of mappings`)");
440 #endif
441 if (UseCompressedOops) {
442 st->print_cr("# This process is running with CompressedOops enabled, and the Java Heap may be blocking the growth of the native heap");
443 }
444 if (LogBytesPerWord == 2) {
445 st->print_cr("# In 32 bit mode, the process size limit was hit");
446 }
447 st->print_cr("# Possible solutions:");
448 st->print_cr("# Reduce memory load on the system");
449 st->print_cr("# Increase physical memory or swap space");
450 st->print_cr("# Check if swap backing store is full");
451 if (LogBytesPerWord == 2) {
452 st->print_cr("# Use 64 bit Java on a 64 bit OS");
453 }
454 st->print_cr("# Decrease Java heap size (-Xmx/-Xms)");
455 st->print_cr("# Decrease number of Java threads");
456 st->print_cr("# Decrease Java thread stack sizes (-Xss)");
457 st->print_cr("# Set larger code cache with -XX:ReservedCodeCacheSize=");
458 if (UseCompressedOops) {
459 switch (CompressedOops::mode()) {
460 case CompressedOops::UnscaledNarrowOop:
461 st->print_cr("# JVM is running with Unscaled Compressed Oops mode in which the Java heap is");
462 st->print_cr("# placed in the first 4GB address space. The Java Heap base address is the");
463 st->print_cr("# maximum limit for the native heap growth. Please use -XX:HeapBaseMinAddress");
464 st->print_cr("# to set the Java Heap base and to place the Java Heap above 4GB virtual address.");
465 break;
466 case CompressedOops::ZeroBasedNarrowOop:
467 st->print_cr("# JVM is running with Zero Based Compressed Oops mode in which the Java heap is");
468 st->print_cr("# placed in the first 32GB address space. The Java Heap base address is the");
469 st->print_cr("# maximum limit for the native heap growth. Please use -XX:HeapBaseMinAddress");
470 st->print_cr("# to set the Java Heap base and to place the Java Heap above 32GB virtual address.");
471 break;
472 default:
473 break;
474 }
475 }
476 st->print_cr("# This output file may be truncated or incomplete.");
477 }
478
479 static void print_stack_location(outputStream* st, const void* context, int& continuation) {
480 const int number_of_stack_slots = 8;
481
482 int i = continuation;
483 // Update continuation with next index before fetching frame
484 continuation = i + 1;
485 const frame fr = os::fetch_frame_from_context(context);
486 while (i < number_of_stack_slots) {
487 // Update continuation with next index before printing location
488 continuation = i + 1;
489 // decode stack contents if possible
490 const intptr_t *sp = fr.sp();
491 const intptr_t *slot = sp + i;
492 if (!is_aligned(slot, sizeof(intptr_t))) {
493 st->print_cr("Misaligned sp: " PTR_FORMAT, p2i(sp));
494 break;
495 } else if (os::is_readable_pointer(slot)) {
496 st->print("stack at sp + %d slots: ", i);
497 os::print_location(st, *(slot));
498 } else {
499 st->print_cr("unreadable stack slot at sp + %d", i);
500 }
501 ++i;
502 }
503 }
504
505 static void report_vm_version(outputStream* st, char* buf, int buflen) {
506 // VM version
507 st->print_cr("#");
508 JDK_Version::current().to_string(buf, buflen);
509 const char* runtime_name = JDK_Version::runtime_name() != nullptr ?
510 JDK_Version::runtime_name() : "";
511 const char* runtime_version = JDK_Version::runtime_version() != nullptr ?
512 JDK_Version::runtime_version() : "";
513 const char* vendor_version = JDK_Version::runtime_vendor_version() != nullptr ?
514 JDK_Version::runtime_vendor_version() : "";
515 const char* jdk_debug_level = VM_Version::printable_jdk_debug_level() != nullptr ?
516 VM_Version::printable_jdk_debug_level() : "";
517
518 st->print_cr("# JRE version: %s%s%s (%s) (%sbuild %s)", runtime_name,
519 (*vendor_version != '\0') ? " " : "", vendor_version,
520 buf, jdk_debug_level, runtime_version);
521
522 // This is the long version with some default settings added
523 st->print_cr("# Java VM: %s%s%s (%s%s, %s%s%s%s%s%s, %s, %s)",
524 VM_Version::vm_name(),
525 (*vendor_version != '\0') ? " " : "", vendor_version,
526 jdk_debug_level,
527 VM_Version::vm_release(),
528 VM_Version::vm_info_string(),
529 TieredCompilation ? ", tiered" : "",
530 #if INCLUDE_JVMCI
531 EnableJVMCI ? ", jvmci" : "",
532 UseJVMCICompiler ? ", jvmci compiler" : "",
533 #else
534 "", "",
535 #endif
536 UseCompressedOops ? ", compressed oops" : "",
537 UseCompactObjectHeaders ? ", compact obj headers"
538 : (UseCompressedClassPointers ? ", compressed class ptrs" : ""),
539 GCConfig::hs_err_name(),
540 VM_Version::vm_platform_string()
541 );
542 }
543
544 // Returns true if at least one thread reported a fatal error and fatal error handling is in process.
545 bool VMError::is_error_reported() {
546 return _first_error_tid != -1;
547 }
548
549 // Returns true if the current thread reported a fatal error.
550 bool VMError::is_error_reported_in_current_thread() {
551 return _first_error_tid == os::current_thread_id();
552 }
553
554 // Helper, return current timestamp for timeout handling.
555 jlong VMError::get_current_timestamp() {
556 return os::javaTimeNanos();
557 }
558 // Factor to translate the timestamp to seconds.
559 #define TIMESTAMP_TO_SECONDS_FACTOR (1000 * 1000 * 1000)
560
561 void VMError::record_reporting_start_time() {
562 const jlong now = get_current_timestamp();
563 AtomicAccess::store(&_reporting_start_time, now);
564 }
565
566 jlong VMError::get_reporting_start_time() {
567 return AtomicAccess::load(&_reporting_start_time);
568 }
569
570 void VMError::record_step_start_time() {
571 const jlong now = get_current_timestamp();
572 AtomicAccess::store(&_step_start_time, now);
573 }
574
575 jlong VMError::get_step_start_time() {
576 return AtomicAccess::load(&_step_start_time);
577 }
578
579 void VMError::clear_step_start_time() {
580 return AtomicAccess::store(&_step_start_time, (jlong)0);
581 }
582
583 // This is the main function to report a fatal error. Only one thread can
584 // call this function, so we don't need to worry about MT-safety. But it's
585 // possible that the error handler itself may crash or die on an internal
586 // error, for example, when the stack/heap is badly damaged. We must be
587 // able to handle recursive errors that happen inside error handler.
588 //
589 // Error reporting is done in several steps. If a crash or internal error
590 // occurred when reporting an error, the nested signal/exception handler
591 // can skip steps that are already (or partially) done. Error reporting will
592 // continue from the next step. This allows us to retrieve and print
593 // information that may be unsafe to get after a fatal error. If it happens,
594 // you may find nested report_and_die() frames when you look at the stack
595 // in a debugger.
596 //
597 // In general, a hang in error handler is much worse than a crash or internal
598 // error, as it's harder to recover from a hang. Deadlock can happen if we
599 // try to grab a lock that is already owned by current thread, or if the
600 // owner is blocked forever (e.g. in os::infinite_sleep()). If possible, the
601 // error handler and all the functions it called should avoid grabbing any
602 // lock. An important thing to notice is that memory allocation needs a lock.
603 //
604 // We should avoid using large stack allocated buffers. Many errors happen
605 // when stack space is already low. Making things even worse is that there
606 // could be nested report_and_die() calls on stack (see above). Only one
607 // thread can report error, so large buffers are statically allocated in data
608 // segment.
609 void VMError::report(outputStream* st, bool _verbose) {
610 // Used by reattempt step logic
611 static int continuation = 0;
612 const char* stop_reattempt_reason = nullptr;
613 # define BEGIN \
614 if (_current_step == 0) { \
615 _step_did_succeed = false; \
616 _current_step = __LINE__; \
617 {
618 // [Begin logic]
619
620 # define STEP_IF(s, cond) \
621 } \
622 _step_did_succeed = true; \
623 } \
624 if (_current_step < __LINE__) { \
625 _step_did_succeed = false; \
626 _current_step = __LINE__; \
627 _current_step_info = s; \
628 if ((cond)) { \
629 record_step_start_time(); \
630 _step_did_timeout = false;
631 // [Step logic]
632
633 # define STEP(s) STEP_IF(s, true)
634
635 # define REATTEMPT_STEP_IF(s, cond) \
636 } \
637 _step_did_succeed = true; \
638 } \
639 if (_current_step < __LINE__ && !_step_did_succeed) { \
640 _current_step = __LINE__; \
641 _current_step_info = s; \
642 const bool cond_value = (cond); \
643 if (cond_value && !can_reattempt_step( \
644 stop_reattempt_reason)) { \
645 st->print_cr("[stop reattempt (%s) reason: %s]", \
646 _current_step_info, \
647 stop_reattempt_reason); \
648 } else if (cond_value) {
649 // [Continue Step logic]
650
651 # define END \
652 } \
653 _step_did_succeed = true; \
654 clear_step_start_time(); \
655 }
656
657 // don't allocate large buffer on stack
658 static char buf[O_BUFLEN];
659
660 // Native stack trace may get stuck. We try to handle the last pc if it
661 // belongs to VM generated code.
662 address lastpc = nullptr;
663
664 BEGIN
665 if (MemTracker::enabled() &&
666 NmtVirtualMemory_lock != nullptr &&
667 NmtVirtualMemory_lock->owned_by_self()) {
668 // Manually unlock to avoid reentrancy due to mallocs in detailed mode.
669 NmtVirtualMemory_lock->unlock();
670 }
671
672 STEP("printing fatal error message")
673 st->print_cr("#");
674 if (should_report_bug(_id)) {
675 st->print_cr("# A fatal error has been detected by the Java Runtime Environment:");
676 } else {
677 st->print_cr("# There is insufficient memory for the Java "
678 "Runtime Environment to continue.");
679 }
680
681 // avoid the cache update for malloc/mmap errors
682 if (should_report_bug(_id)) {
683 os::prepare_native_symbols();
684 }
685
686 #ifdef ASSERT
687 // Error handler self tests
688 // Meaning of codes passed through in the tests.
689 #define TEST_SECONDARY_CRASH 14
690 #define TEST_REATTEMPT_SECONDARY_CRASH 15
691 #define TEST_RESOURCE_MARK_CRASH 2
692
693 // test secondary error handling. Test it twice, to test that resetting
694 // error handler after a secondary crash works.
695 STEP_IF("test secondary crash 1", _verbose && TestCrashInErrorHandler == TEST_SECONDARY_CRASH)
696 st->print_cr("Will crash now (TestCrashInErrorHandler=%u)...",
697 TestCrashInErrorHandler);
698 controlled_crash(TestCrashInErrorHandler);
699
700 STEP_IF("test secondary crash 2", _verbose && TestCrashInErrorHandler == TEST_SECONDARY_CRASH)
701 st->print_cr("Will crash now (TestCrashInErrorHandler=%u)...",
702 TestCrashInErrorHandler);
703 controlled_crash(TestCrashInErrorHandler);
704
705 // See corresponding test in test/runtime/ErrorHandling/ReattemptErrorTest.java
706 STEP_IF("test reattempt secondary crash",
707 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
708 st->print_cr("Will crash now (TestCrashInErrorHandler=%u)...",
709 TestCrashInErrorHandler);
710 controlled_crash(14);
711
712 REATTEMPT_STEP_IF("test reattempt secondary crash, attempt 2",
713 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
714 st->print_cr("test reattempt secondary crash. attempt 2");
715
716 REATTEMPT_STEP_IF("test reattempt secondary crash, attempt 3",
717 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
718 st->print_cr("test reattempt secondary crash. attempt 3");
719
720 STEP_IF("test reattempt timeout",
721 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
722 st->print_cr("test reattempt timeout");
723 os::infinite_sleep();
724
725 REATTEMPT_STEP_IF("test reattempt timeout, attempt 2",
726 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
727 st->print_cr("test reattempt timeout, attempt 2");
728
729 STEP_IF("test reattempt stack headroom",
730 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
731 st->print_cr("test reattempt stack headroom");
732 reattempt_test_hit_stack_limit(st);
733
734 REATTEMPT_STEP_IF("test reattempt stack headroom, attempt 2",
735 _verbose && TestCrashInErrorHandler == TEST_REATTEMPT_SECONDARY_CRASH)
736 st->print_cr("test reattempt stack headroom, attempt 2");
737
738 STEP_IF("test missing ResourceMark does not crash",
739 _verbose && TestCrashInErrorHandler == TEST_RESOURCE_MARK_CRASH)
740 stringStream message;
741 message.print("This is a message with no ResourceMark");
742 tty->print_cr("%s", message.as_string());
743
744 // TestUnresponsiveErrorHandler: We want to test both step timeouts and global timeout.
745 // Step to global timeout ratio is 4:1, so in order to be absolutely sure we hit the
746 // global timeout, let's execute the timeout step five times.
747 // See corresponding test in test/runtime/ErrorHandling/TimeoutInErrorHandlingTest.java
748 STEP_IF("setup for test unresponsive error reporting step",
749 _verbose && TestUnresponsiveErrorHandler)
750 // We record reporting_start_time for this test here because we
751 // care about the time spent executing TIMEOUT_TEST_STEP and not
752 // about the time it took us to get here.
753 tty->print_cr("Recording reporting_start_time for TestUnresponsiveErrorHandler.");
754 record_reporting_start_time();
755
756 #define TIMEOUT_TEST_STEP STEP_IF("test unresponsive error reporting step", \
757 _verbose && TestUnresponsiveErrorHandler) \
758 os::infinite_sleep();
759 TIMEOUT_TEST_STEP
760 TIMEOUT_TEST_STEP
761 TIMEOUT_TEST_STEP
762 TIMEOUT_TEST_STEP
763 TIMEOUT_TEST_STEP
764
765 STEP_IF("test safefetch in error handler", _verbose && TestSafeFetchInErrorHandler)
766 // test whether it is safe to use SafeFetch32 in Crash Handler. Test twice
767 // to test that resetting the signal handler works correctly.
768 st->print_cr("Will test SafeFetch...");
769 int* const invalid_pointer = (int*)segfault_address;
770 const int x = 0x76543210;
771 int i1 = SafeFetch32(invalid_pointer, x);
772 int i2 = SafeFetch32(invalid_pointer, x);
773 if (i1 == x && i2 == x) {
774 st->print_cr("SafeFetch OK."); // Correctly deflected and returned default pattern
775 } else {
776 st->print_cr("??");
777 }
778 #endif // ASSERT
779
780 STEP("printing type of error")
781 switch(static_cast<unsigned int>(_id)) {
782 case OOM_MALLOC_ERROR:
783 case OOM_MMAP_ERROR:
784 case OOM_MPROTECT_ERROR:
785 if (_size) {
786 st->print("# Native memory allocation ");
787 st->print((_id == (int)OOM_MALLOC_ERROR) ? "(malloc) failed to allocate " :
788 (_id == (int)OOM_MMAP_ERROR) ? "(mmap) failed to map " :
789 "(mprotect) failed to protect ");
790 jio_snprintf(buf, sizeof(buf), "%zu", _size);
791 st->print("%s", buf);
792 st->print(" bytes.");
793 if (strlen(_detail_msg) > 0) {
794 st->print(" Error detail: ");
795 st->print("%s", _detail_msg);
796 }
797 st->cr();
798 } else {
799 if (strlen(_detail_msg) > 0) {
800 st->print("# ");
801 st->print_cr("%s", _detail_msg);
802 }
803 }
804 // In error file give some solutions
805 if (_verbose) {
806 print_oom_reasons(st);
807 } else {
808 return; // that's enough for the screen
809 }
810 break;
811 case INTERNAL_ERROR:
812 default:
813 break;
814 }
815
816 STEP("printing exception/signal name")
817 st->print_cr("#");
818 st->print("# ");
819 // Is it an OS exception/signal?
820 if (os::exception_name(_id, buf, sizeof(buf))) {
821 st->print("%s", buf);
822 st->print(" (0x%x)", _id); // signal number
823 st->print(" at pc=" PTR_FORMAT, p2i(_pc));
824 if (_siginfo != nullptr && os::signal_sent_by_kill(_siginfo)) {
825 if (get_handshake_timed_out_thread() == _thread) {
826 st->print(" (sent by handshake timeout handler)");
827 } else if (get_safepoint_timed_out_thread() == _thread) {
828 st->print(" (sent by safepoint timeout handler)");
829 } else {
830 st->print(" (sent by kill)");
831 }
832 }
833 } else {
834 if (should_report_bug(_id)) {
835 st->print("Internal Error");
836 } else {
837 st->print("Out of Memory Error");
838 }
839 if (_filename != nullptr && _lineno > 0) {
840 #ifdef PRODUCT
841 // In product mode chop off pathname
842 const char *file = get_filename_only();
843 #else
844 const char *file = _filename;
845 #endif
846 st->print(" (%s:%d)", file, _lineno);
847 } else {
848 st->print(" (0x%x)", _id);
849 }
850 }
851
852 STEP("printing current thread and pid")
853 // process id, thread id
854 st->print(", pid=%d", os::current_process_id());
855 st->print(", tid=%zu", os::current_thread_id());
856 st->cr();
857
858 STEP_IF("printing error message", should_report_bug(_id)) // already printed the message.
859 // error message
860 if (strlen(_detail_msg) > 0) {
861 st->print_cr("# %s: %s", _message ? _message : "Error", _detail_msg);
862 } else if (_message) {
863 st->print_cr("# Error: %s", _message);
864 }
865
866 STEP("printing Java version string")
867 report_vm_version(st, buf, sizeof(buf));
868
869 STEP_IF("printing problematic frame", _context != nullptr)
870 // Print current frame if we have a context (i.e. it's a crash)
871 st->print_cr("# Problematic frame:");
872 st->print("# ");
873 frame fr = os::fetch_frame_from_context(_context);
874 fr.print_on_error(st, buf, sizeof(buf));
875 st->cr();
876 st->print_cr("#");
877
878 STEP("printing core file information")
879 st->print("# ");
880 if (CreateCoredumpOnCrash) {
881 if (coredump_status) {
882 st->print("Core dump will be written. Default location: %s", coredump_message);
883 } else {
884 st->print("No core dump will be written. %s", coredump_message);
885 }
886 } else {
887 st->print("CreateCoredumpOnCrash turned off, no core file dumped");
888 }
889 st->cr();
890 st->print_cr("#");
891
892 JFR_ONLY(STEP("printing jfr information"))
893 JFR_ONLY(Jfr::on_vm_error_report(st);)
894
895 STEP_IF("printing bug submit message", should_submit_bug_report(_id) && _verbose)
896 print_bug_submit_message(st, _thread);
897
898 STEP_IF("printing summary", _verbose)
899 st->cr();
900 st->print_cr("--------------- S U M M A R Y ------------");
901 st->cr();
902
903 STEP_IF("printing VM option summary", _verbose)
904 // VM options
905 Arguments::print_summary_on(st);
906 st->cr();
907
908 STEP_IF("printing summary machine and OS info", _verbose)
909 os::print_summary_info(st, buf, sizeof(buf));
910
911 STEP_IF("printing date and time", _verbose)
912 os::print_date_and_time(st, buf, sizeof(buf));
913
914 #ifdef ADDRESS_SANITIZER
915 STEP_IF("printing ASAN error information", _verbose && Asan::had_error())
916 st->cr();
917 st->print_cr("------------------ A S A N ----------------");
918 st->cr();
919 Asan::report(st);
920 st->cr();
921 #endif // ADDRESS_SANITIZER
922
923 STEP_IF("printing thread", _verbose)
924 st->cr();
925 st->print_cr("--------------- T H R E A D ---------------");
926 st->cr();
927
928 STEP_IF("printing current thread", _verbose)
929 // current thread
930 if (_thread) {
931 st->print("Current thread (" PTR_FORMAT "): ", p2i(_thread));
932 _thread->print_on_error(st, buf, sizeof(buf));
933 st->cr();
934 } else {
935 st->print_cr("Current thread is native thread");
936 }
937 st->cr();
938
939 STEP_IF("printing current compile task",
940 _verbose && _thread != nullptr && _thread->is_Compiler_thread())
941 CompilerThread* t = (CompilerThread*)_thread;
942 if (t->task()) {
943 st->cr();
944 st->print_cr("Current CompileTask:");
945 t->task()->print_line_on_error(st, buf, sizeof(buf));
946 st->cr();
947 }
948
949 STEP_IF("printing stack bounds", _verbose)
950 st->print("Stack: ");
951
952 address stack_top;
953 size_t stack_size;
954
955 if (_thread) {
956 stack_top = _thread->stack_base();
957 stack_size = _thread->stack_size();
958 } else {
959 os::current_stack_base_and_size(&stack_top, &stack_size);
960 }
961
962 address stack_bottom = stack_top - stack_size;
963 st->print("[" PTR_FORMAT "," PTR_FORMAT "]", p2i(stack_bottom), p2i(stack_top));
964
965 frame fr = _context ? os::fetch_frame_from_context(_context)
966 : os::current_frame();
967
968 if (fr.sp()) {
969 st->print(", sp=" PTR_FORMAT, p2i(fr.sp()));
970 size_t free_stack_size = pointer_delta(fr.sp(), stack_bottom, 1024);
971 st->print(", free space=%zuk", free_stack_size);
972 }
973
974 st->cr();
975
976 STEP_IF("printing native stack (with source info)", _verbose)
977
978 NativeStackPrinter nsp(_thread, _context, _filename != nullptr ? get_filename_only() : nullptr, _lineno);
979 if (nsp.print_stack(st, buf, sizeof(buf), lastpc,
980 true /*print_source_info */, -1 /* max stack */)) {
981 // We have printed the native stack in platform-specific code
982 // Windows/x64 needs special handling.
983 // Stack walking may get stuck. Try to find the calling code.
984 if (lastpc != nullptr) {
985 const char* name = find_code_name(lastpc);
986 if (name != nullptr) {
987 st->print_cr("The last pc belongs to %s (printed below).", name);
988 }
989 }
990 } else {
991 _print_stack_from_frame_used = true; // frame-based native stack walk done
992 }
993
994 REATTEMPT_STEP_IF("retry printing native stack (no source info)", _verbose)
995 st->cr();
996 st->print_cr("Retrying call stack printing without source information...");
997 NativeStackPrinter nsp(_thread, _context, get_filename_only(), _lineno);
998 nsp.print_stack_from_frame(st, buf, sizeof(buf),
999 false /*print_source_info */, -1 /* max stack */);
1000 _print_stack_from_frame_used = true;
1001
1002 STEP_IF("printing Java stack", _verbose && _thread && _thread->is_Java_thread())
1003 if (_verbose && _thread && _thread->is_Java_thread()) {
1004 print_stack_trace(st, JavaThread::cast(_thread), buf, sizeof(buf));
1005 }
1006
1007 STEP_IF("printing target Java thread stack",
1008 _verbose && _thread != nullptr && (_thread->is_Named_thread()))
1009 // printing Java thread stack trace if it is involved in GC crash
1010 Thread* thread = ((NamedThread *)_thread)->processed_thread();
1011 if (thread != nullptr && thread->is_Java_thread()) {
1012 JavaThread* jt = JavaThread::cast(thread);
1013 st->print_cr("JavaThread " PTR_FORMAT " (nid = %d) was being processed", p2i(jt), jt->osthread()->thread_id());
1014 print_stack_trace(st, jt, buf, sizeof(buf), true);
1015 }
1016
1017 STEP_IF("printing siginfo", _verbose && _siginfo != nullptr)
1018 // signal no, signal code, address that caused the fault
1019 st->cr();
1020 os::print_siginfo(st, _siginfo);
1021 st->cr();
1022
1023 STEP_IF("CDS archive access warning", _verbose && _siginfo != nullptr)
1024 // Print an explicit hint if we crashed on access to the CDS archive.
1025 check_failing_cds_access(st, _siginfo);
1026 st->cr();
1027
1028 #if defined(COMPILER1) || defined(COMPILER2)
1029 STEP_IF("printing pending compilation failure",
1030 _verbose && _thread != nullptr && _thread->is_Compiler_thread())
1031 CompilationFailureInfo::print_pending_compilation_failure(st);
1032 if (CompilationMemoryStatistic::enabled() && CompilationMemoryStatistic::in_oom_crash()) {
1033 st->cr();
1034 st->print_cr(">> Please see below for a detailed breakdown of compiler memory usage.");
1035 st->cr();
1036 }
1037 #endif
1038
1039 STEP_IF("printing registers", _verbose && _context != nullptr)
1040 // printing registers
1041 os::print_context(st, _context);
1042 st->cr();
1043
1044 STEP_IF("printing register info",
1045 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1046 continuation = 0;
1047 ResourceMark rm(_thread);
1048 st->print_cr("Register to memory mapping:");
1049 st->cr();
1050 os::print_register_info(st, _context, continuation);
1051 st->cr();
1052
1053 REATTEMPT_STEP_IF("printing register info, attempt 2",
1054 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1055 ResourceMark rm(_thread);
1056 os::print_register_info(st, _context, continuation);
1057 st->cr();
1058
1059 REATTEMPT_STEP_IF("printing register info, attempt 3",
1060 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1061 ResourceMark rm(_thread);
1062 os::print_register_info(st, _context, continuation);
1063 st->cr();
1064
1065 STEP_IF("printing top of stack, instructions near pc", _verbose && _context != nullptr)
1066 // printing top of stack, instructions near pc
1067 os::print_tos_pc(st, _context);
1068 st->cr();
1069
1070 STEP_IF("inspecting top of stack",
1071 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1072 continuation = 0;
1073 ResourceMark rm(_thread);
1074 st->print_cr("Stack slot to memory mapping:");
1075 st->cr();
1076 print_stack_location(st, _context, continuation);
1077 st->cr();
1078
1079 REATTEMPT_STEP_IF("inspecting top of stack, attempt 2",
1080 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1081 ResourceMark rm(_thread);
1082 print_stack_location(st, _context, continuation);
1083 st->cr();
1084
1085 REATTEMPT_STEP_IF("inspecting top of stack, attempt 3",
1086 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1087 ResourceMark rm(_thread);
1088 print_stack_location(st, _context, continuation);
1089 st->cr();
1090
1091 STEP_IF("printing lock stack", _verbose && _thread != nullptr && _thread->is_Java_thread());
1092 st->print_cr("Lock stack of current Java thread (top to bottom):");
1093 JavaThread::cast(_thread)->lock_stack().print_on(st);
1094 st->cr();
1095
1096 STEP_IF("printing code blobs if possible", _verbose)
1097 const int printed_capacity = max_error_log_print_code;
1098 address printed[printed_capacity];
1099 printed[0] = nullptr;
1100 int printed_len = 0;
1101 // Even though ErrorLogPrintCodeLimit is ranged checked
1102 // during argument parsing, there's no way to prevent it
1103 // subsequently (i.e., after parsing) being set to a
1104 // value outside the range.
1105 int limit = MIN2(ErrorLogPrintCodeLimit, printed_capacity);
1106 if (limit > 0) {
1107 // Check if a pc was found by native stack trace above.
1108 if (lastpc != nullptr) {
1109 if (print_code(st, _thread, lastpc, true, printed, printed_capacity)) {
1110 printed_len++;
1111 }
1112 }
1113
1114 // Scan the native stack
1115 if (!_print_stack_from_frame_used) {
1116 // Only try to print code of the crashing frame since
1117 // the native stack cannot be walked with next_frame.
1118 if (print_code(st, _thread, _pc, true, printed, printed_capacity)) {
1119 printed_len++;
1120 }
1121 } else {
1122 frame fr = _context ? os::fetch_frame_from_context(_context)
1123 : os::current_frame();
1124 while (printed_len < limit && fr.pc() != nullptr) {
1125 if (print_code(st, _thread, fr.pc(), fr.pc() == _pc, printed, printed_capacity)) {
1126 printed_len++;
1127 }
1128 fr = frame::next_frame(fr, _thread);
1129 }
1130 }
1131
1132 // Scan the Java stack
1133 if (_thread != nullptr && _thread->is_Java_thread()) {
1134 JavaThread* jt = JavaThread::cast(_thread);
1135 if (jt->has_last_Java_frame()) {
1136 for (StackFrameStream sfs(jt, true /* update */, true /* process_frames */); printed_len < limit && !sfs.is_done(); sfs.next()) {
1137 address pc = sfs.current()->pc();
1138 if (print_code(st, _thread, pc, pc == _pc, printed, printed_capacity)) {
1139 printed_len++;
1140 }
1141 }
1142 }
1143 }
1144 }
1145
1146 STEP_IF("printing VM operation", _verbose && _thread != nullptr && _thread->is_VM_thread())
1147 VMThread* t = (VMThread*)_thread;
1148 VM_Operation* op = t->vm_operation();
1149 if (op) {
1150 op->print_on_error(st);
1151 st->cr();
1152 st->cr();
1153 }
1154
1155 STEP_IF("printing registered callbacks", _verbose && _thread != nullptr);
1156 size_t count = 0;
1157 for (VMErrorCallback* callback = _thread->_vm_error_callbacks;
1158 callback != nullptr;
1159 callback = callback->_next) {
1160 st->print_cr("VMErrorCallback %zu:", ++count);
1161 callback->call(st);
1162 st->cr();
1163 }
1164
1165 STEP_IF("printing process", _verbose)
1166 st->cr();
1167 st->print_cr("--------------- P R O C E S S ---------------");
1168 st->cr();
1169
1170 STEP_IF("printing user info", ExtensiveErrorReports && _verbose)
1171 os::print_user_info(st);
1172
1173 STEP_IF("printing all threads", _verbose && _thread != nullptr)
1174 // all threads
1175 Threads::print_on_error(st, _thread, buf, sizeof(buf));
1176 st->cr();
1177
1178 STEP_IF("printing VM state", _verbose)
1179 // Safepoint state
1180 st->print("VM state: ");
1181
1182 if (SafepointSynchronize::is_synchronizing()) st->print("synchronizing");
1183 else if (SafepointSynchronize::is_at_safepoint()) st->print("at safepoint");
1184 else st->print("not at safepoint");
1185
1186 // Also see if error occurred during initialization or shutdown
1187 if (!Universe::is_fully_initialized()) {
1188 st->print(" (not fully initialized)");
1189 } else if (VM_Exit::vm_exited()) {
1190 st->print(" (shutting down)");
1191 } else {
1192 st->print(" (normal execution)");
1193 }
1194 st->cr();
1195 st->cr();
1196
1197 STEP_IF("printing owned locks on error", _verbose)
1198 // mutexes/monitors that currently have an owner
1199 Mutex::print_owned_locks_on_error(st);
1200 st->cr();
1201
1202 STEP_IF("printing number of OutOfMemoryError and StackOverflow exceptions",
1203 _verbose && Exceptions::has_exception_counts())
1204 st->print_cr("OutOfMemory and StackOverflow Exception counts:");
1205 Exceptions::print_exception_counts_on_error(st);
1206 st->cr();
1207
1208 #ifdef _LP64
1209 STEP_IF("printing compressed oops mode", _verbose && UseCompressedOops)
1210 CompressedOops::print_mode(st);
1211 st->cr();
1212
1213 STEP_IF("printing compressed klass pointers mode", _verbose && UseCompressedClassPointers)
1214 CDS_ONLY(AOTMetaspace::print_on(st);)
1215 Metaspace::print_compressed_class_space(st);
1216 CompressedKlassPointers::print_mode(st);
1217 st->cr();
1218 #endif
1219
1220 STEP_IF("printing heap information", _verbose)
1221 GCLogPrecious::print_on_error(st);
1222
1223 if (Universe::heap() != nullptr) {
1224 st->print_cr("Heap:");
1225 StreamIndentor si(st, 1);
1226 Universe::heap()->print_heap_on(st);
1227 st->cr();
1228 }
1229
1230 STEP_IF("printing GC information", _verbose)
1231 if (Universe::heap() != nullptr) {
1232 Universe::heap()->print_gc_on(st);
1233 st->cr();
1234 }
1235
1236 if (Universe::is_fully_initialized()) {
1237 st->print_cr("Polling page: " PTR_FORMAT, p2i(SafepointMechanism::get_polling_page()));
1238 st->cr();
1239 }
1240
1241 STEP_IF("printing metaspace information", _verbose && Universe::is_fully_initialized())
1242 st->print_cr("Metaspace:");
1243 MetaspaceUtils::print_on(st);
1244 MetaspaceUtils::print_basic_report(st, 0);
1245
1246 STEP_IF("printing code cache information", _verbose && Universe::is_fully_initialized())
1247 // print code cache information before vm abort
1248 CodeCache::print_summary(st);
1249 st->cr();
1250
1251 STEP_IF("printing ring buffers", _verbose)
1252 Events::print_all(st);
1253 st->cr();
1254
1255 STEP_IF("printing dynamic libraries", _verbose)
1256 // dynamic libraries, or memory map
1257 os::print_dll_info(st);
1258 st->cr();
1259
1260 #if INCLUDE_JVMTI
1261 STEP_IF("printing jvmti agent info", _verbose)
1262 os::print_jvmti_agent_info(st);
1263 st->cr();
1264 #endif
1265
1266 STEP_IF("printing native decoder state", _verbose)
1267 Decoder::print_state_on(st);
1268 st->cr();
1269
1270 STEP_IF("printing VM options", _verbose)
1271 // VM options
1272 Arguments::print_on(st);
1273 st->cr();
1274
1275 STEP_IF("printing flags", _verbose)
1276 JVMFlag::printFlags(
1277 st,
1278 true, // with comments
1279 false, // no ranges
1280 true); // skip defaults
1281 st->cr();
1282
1283 STEP_IF("printing warning if internal testing API used", WhiteBox::used())
1284 st->print_cr("Unsupported internal testing APIs have been used.");
1285 st->cr();
1286
1287 STEP_IF("printing log configuration", _verbose)
1288 st->print_cr("Logging:");
1289 LogConfiguration::describe_current_configuration(st);
1290 st->cr();
1291
1292 STEP_IF("printing release file content", _verbose)
1293 st->print_cr("Release file:");
1294 os::print_image_release_file(st);
1295
1296 STEP_IF("printing all environment variables", _verbose)
1297 os::print_environment_variables(st, env_list);
1298 st->cr();
1299
1300 STEP_IF("printing locale settings", _verbose)
1301 os::print_active_locale(st);
1302 st->cr();
1303
1304 STEP_IF("printing signal handlers", _verbose)
1305 os::print_signal_handlers(st, buf, sizeof(buf));
1306 st->cr();
1307
1308 STEP_IF("Native Memory Tracking", _verbose)
1309 MemTracker::error_report(st);
1310 st->cr();
1311
1312 STEP_IF("printing compiler memory info, if any", _verbose)
1313 CompilationMemoryStatistic::print_error_report(st);
1314 st->cr();
1315
1316 STEP_IF("printing periodic trim state", _verbose)
1317 NativeHeapTrimmer::print_state(st);
1318 st->cr();
1319
1320 STEP_IF("printing system", _verbose)
1321 st->print_cr("--------------- S Y S T E M ---------------");
1322 st->cr();
1323
1324 STEP_IF("printing OS information", _verbose)
1325 os::print_os_info(st);
1326 st->cr();
1327
1328 STEP_IF("printing CPU info", _verbose)
1329 os::print_cpu_info(st, buf, sizeof(buf));
1330 st->cr();
1331
1332 STEP_IF("printing memory info", _verbose)
1333 os::print_memory_info(st);
1334 st->cr();
1335
1336 STEP_IF("printing internal vm info", _verbose)
1337 st->print_cr("vm_info: %s", VM_Version::internal_vm_info_string());
1338 st->cr();
1339
1340 // print a defined marker to show that error handling finished correctly.
1341 STEP_IF("printing end marker", _verbose)
1342 st->print_cr("END.");
1343
1344 END
1345
1346 # undef BEGIN
1347 # undef STEP_IF
1348 # undef STEP
1349 # undef REATTEMPT_STEP_IF
1350 # undef END
1351 }
1352
1353 void VMError::set_handshake_timed_out_thread(Thread* thread) {
1354 // Only preserve the first thread to time-out this way. The atomic operation ensures
1355 // visibility to the target thread.
1356 AtomicAccess::replace_if_null(&_handshake_timed_out_thread, thread);
1357 }
1358
1359 void VMError::set_safepoint_timed_out_thread(Thread* thread) {
1360 // Only preserve the first thread to time-out this way. The atomic operation ensures
1361 // visibility to the target thread.
1362 AtomicAccess::replace_if_null(&_safepoint_timed_out_thread, thread);
1363 }
1364
1365 Thread* VMError::get_handshake_timed_out_thread() {
1366 return AtomicAccess::load(&_handshake_timed_out_thread);
1367 }
1368
1369 Thread* VMError::get_safepoint_timed_out_thread() {
1370 return AtomicAccess::load(&_safepoint_timed_out_thread);
1371 }
1372
1373 // Report for the vm_info_cmd. This prints out the information above omitting
1374 // crash and thread specific information. If output is added above, it should be added
1375 // here also, if it is safe to call during a running process.
1376 void VMError::print_vm_info(outputStream* st) {
1377
1378 char buf[O_BUFLEN];
1379 os::prepare_native_symbols();
1380
1381 report_vm_version(st, buf, sizeof(buf));
1382
1383 // STEP("printing summary")
1384
1385 st->cr();
1386 st->print_cr("--------------- S U M M A R Y ------------");
1387 st->cr();
1388
1389 // STEP("printing VM option summary")
1390
1391 // VM options
1392 Arguments::print_summary_on(st);
1393 st->cr();
1394
1395 // STEP("printing summary machine and OS info")
1396
1397 os::print_summary_info(st, buf, sizeof(buf));
1398
1399 // STEP("printing date and time")
1400
1401 os::print_date_and_time(st, buf, sizeof(buf));
1402
1403 // Skip: STEP("printing thread")
1404
1405 // STEP("printing process")
1406
1407 st->cr();
1408 st->print_cr("--------------- P R O C E S S ---------------");
1409 st->cr();
1410
1411 // STEP("printing number of OutOfMemoryError and StackOverflow exceptions")
1412
1413 if (Exceptions::has_exception_counts()) {
1414 st->print_cr("OutOfMemory and StackOverflow Exception counts:");
1415 Exceptions::print_exception_counts_on_error(st);
1416 st->cr();
1417 }
1418
1419 #ifdef _LP64
1420 // STEP("printing compressed oops mode")
1421 if (UseCompressedOops) {
1422 CompressedOops::print_mode(st);
1423 st->cr();
1424 }
1425 #endif
1426
1427 // STEP("printing compressed class ptrs mode")
1428 if (UseCompressedClassPointers) {
1429 CDS_ONLY(AOTMetaspace::print_on(st);)
1430 Metaspace::print_compressed_class_space(st);
1431 CompressedKlassPointers::print_mode(st);
1432 st->cr();
1433 }
1434
1435 // Take heap lock over heap, GC and metaspace printing so that information
1436 // is consistent.
1437 if (Universe::is_fully_initialized()) {
1438 MutexLocker ml(Heap_lock);
1439
1440 // STEP("printing heap information")
1441
1442 GCLogPrecious::print_on_error(st);
1443
1444 {
1445 st->print_cr("Heap:");
1446 StreamIndentor si(st, 1);
1447 Universe::heap()->print_heap_on(st);
1448 st->cr();
1449 }
1450
1451 // STEP("printing GC information")
1452
1453 Universe::heap()->print_gc_on(st);
1454 st->cr();
1455
1456 st->print_cr("Polling page: " PTR_FORMAT, p2i(SafepointMechanism::get_polling_page()));
1457 st->cr();
1458
1459 // STEP("printing metaspace information")
1460
1461 st->print_cr("Metaspace:");
1462 MetaspaceUtils::print_on(st);
1463 MetaspaceUtils::print_basic_report(st, 0);
1464 }
1465
1466 // STEP("printing code cache information")
1467
1468 if (Universe::is_fully_initialized()) {
1469 // print code cache information before vm abort
1470 CodeCache::print_summary(st);
1471 st->cr();
1472 }
1473
1474 // STEP("printing ring buffers")
1475
1476 Events::print_all(st);
1477 st->cr();
1478
1479 // STEP("printing dynamic libraries")
1480
1481 // dynamic libraries, or memory map
1482 os::print_dll_info(st);
1483 st->cr();
1484
1485 #if INCLUDE_JVMTI
1486 os::print_jvmti_agent_info(st);
1487 st->cr();
1488 #endif
1489
1490 // STEP("printing VM options")
1491
1492 // VM options
1493 Arguments::print_on(st);
1494 st->cr();
1495
1496 // STEP("printing warning if internal testing API used")
1497
1498 if (WhiteBox::used()) {
1499 st->print_cr("Unsupported internal testing APIs have been used.");
1500 st->cr();
1501 }
1502
1503 // STEP("printing log configuration")
1504 st->print_cr("Logging:");
1505 LogConfiguration::describe(st);
1506 st->cr();
1507
1508 // STEP("printing release file content")
1509 st->print_cr("Release file:");
1510 os::print_image_release_file(st);
1511
1512 // STEP("printing all environment variables")
1513
1514 os::print_environment_variables(st, env_list);
1515 st->cr();
1516
1517 // STEP("printing locale settings")
1518
1519 os::print_active_locale(st);
1520 st->cr();
1521
1522
1523 // STEP("printing signal handlers")
1524
1525 os::print_signal_handlers(st, buf, sizeof(buf));
1526 st->cr();
1527
1528 // STEP("Native Memory Tracking")
1529 MemTracker::error_report(st);
1530 st->cr();
1531
1532 // STEP("Compiler Memory Statistic")
1533 CompilationMemoryStatistic::print_final_report(st);
1534
1535 // STEP("printing periodic trim state")
1536 NativeHeapTrimmer::print_state(st);
1537 st->cr();
1538
1539
1540 // STEP("printing system")
1541 st->print_cr("--------------- S Y S T E M ---------------");
1542 st->cr();
1543
1544 // STEP("printing OS information")
1545
1546 os::print_os_info(st);
1547 st->cr();
1548
1549 // STEP("printing CPU info")
1550
1551 os::print_cpu_info(st, buf, sizeof(buf));
1552 st->cr();
1553
1554 // STEP("printing memory info")
1555
1556 os::print_memory_info(st);
1557 st->cr();
1558
1559 // STEP("printing internal vm info")
1560
1561 st->print_cr("vm_info: %s", VM_Version::internal_vm_info_string());
1562 st->cr();
1563
1564 // print a defined marker to show that error handling finished correctly.
1565 // STEP("printing end marker")
1566
1567 st->print_cr("END.");
1568 }
1569
1570 /** Expand a pattern into a buffer starting at pos and open a file using constructed path */
1571 static int expand_and_open(const char* pattern, bool overwrite_existing, char* buf, size_t buflen, size_t pos) {
1572 int fd = -1;
1573 int mode = O_RDWR | O_CREAT;
1574 if (overwrite_existing) {
1575 mode |= O_TRUNC;
1576 } else {
1577 mode |= O_EXCL;
1578 }
1579 if (Arguments::copy_expand_pid(pattern, strlen(pattern), &buf[pos], buflen - pos)) {
1580 fd = open(buf, mode, 0666);
1581 }
1582 return fd;
1583 }
1584
1585 /**
1586 * Construct file name for a log file and return it's file descriptor.
1587 * Name and location depends on pattern, default_pattern params and access
1588 * permissions.
1589 */
1590 int VMError::prepare_log_file(const char* pattern, const char* default_pattern, bool overwrite_existing, char* buf, size_t buflen) {
1591 int fd = -1;
1592
1593 // If possible, use specified pattern to construct log file name
1594 if (pattern != nullptr) {
1595 fd = expand_and_open(pattern, overwrite_existing, buf, buflen, 0);
1596 }
1597
1598 // Either user didn't specify, or the user's location failed,
1599 // so use the default name in the current directory
1600 if (fd == -1) {
1601 const char* cwd = os::get_current_directory(buf, buflen);
1602 if (cwd != nullptr) {
1603 size_t pos = strlen(cwd);
1604 int fsep_len = jio_snprintf(&buf[pos], buflen-pos, "%s", os::file_separator());
1605 pos += fsep_len;
1606 if (fsep_len > 0) {
1607 fd = expand_and_open(default_pattern, overwrite_existing, buf, buflen, pos);
1608 }
1609 }
1610 }
1611
1612 // try temp directory if it exists.
1613 if (fd == -1) {
1614 const char* tmpdir = os::get_temp_directory();
1615 if (tmpdir != nullptr && strlen(tmpdir) > 0) {
1616 int pos = jio_snprintf(buf, buflen, "%s%s", tmpdir, os::file_separator());
1617 if (pos > 0) {
1618 fd = expand_and_open(default_pattern, overwrite_existing, buf, buflen, pos);
1619 }
1620 }
1621 }
1622
1623 return fd;
1624 }
1625
1626 void VMError::report_and_die(Thread* thread, unsigned int sig, address pc, const void* siginfo,
1627 const void* context, const char* detail_fmt, ...)
1628 {
1629 va_list detail_args;
1630 va_start(detail_args, detail_fmt);
1631 report_and_die(sig, nullptr, detail_fmt, detail_args, thread, pc, siginfo, context, nullptr, 0, 0);
1632 va_end(detail_args);
1633 }
1634
1635 void VMError::report_and_die(Thread* thread, const void* context, const char* filename, int lineno, const char* message,
1636 const char* detail_fmt, ...) {
1637 va_list detail_args;
1638 va_start(detail_args, detail_fmt);
1639 report_and_die(thread, context, filename, lineno, message, detail_fmt, detail_args);
1640 va_end(detail_args);
1641 }
1642
1643 void VMError::report_and_die(Thread* thread, unsigned int sig, address pc, const void* siginfo, const void* context)
1644 {
1645 if (ExecutingUnitTests) {
1646 // See TEST_VM_CRASH_SIGNAL gtest macro
1647 char tmp[64];
1648 fprintf(stderr, "signaled: %s", os::exception_name(sig, tmp, sizeof(tmp)));
1649 }
1650
1651 report_and_die(thread, sig, pc, siginfo, context, "%s", "");
1652 }
1653
1654 void VMError::report_and_die(Thread* thread, const void* context, const char* filename, int lineno, const char* message,
1655 const char* detail_fmt, va_list detail_args)
1656 {
1657 report_and_die(INTERNAL_ERROR, message, detail_fmt, detail_args, thread, nullptr, nullptr, context, filename, lineno, 0);
1658 }
1659
1660 void VMError::report_and_die(Thread* thread, const char* filename, int lineno, size_t size,
1661 VMErrorType vm_err_type, const char* detail_fmt, va_list detail_args) {
1662 report_and_die(vm_err_type, nullptr, detail_fmt, detail_args, thread, nullptr, nullptr, nullptr, filename, lineno, size);
1663 }
1664
1665 void VMError::report_and_die(int id, const char* message, const char* detail_fmt, va_list detail_args,
1666 Thread* thread, address pc, const void* siginfo, const void* context, const char* filename,
1667 int lineno, size_t size)
1668 {
1669 // A single scratch buffer to be used from here on.
1670 // Do not rely on it being preserved across function calls.
1671 static char buffer[O_BUFLEN];
1672
1673 // File descriptor to tty to print an error summary to.
1674 // Hard wired to stdout; see JDK-8215004 (compatibility concerns).
1675 static const int fd_out = 1; // stdout
1676
1677 // File descriptor to the error log file.
1678 static int fd_log = -1;
1679
1680 #ifdef CAN_SHOW_REGISTERS_ON_ASSERT
1681 // Disarm assertion poison page, since from this point on we do not need this mechanism anymore and it may
1682 // cause problems in error handling during native OOM, see JDK-8227275.
1683 disarm_assert_poison();
1684 #endif
1685
1686 // Use local fdStream objects only. Do not use global instances whose initialization
1687 // relies on dynamic initialization (see JDK-8214975). Do not rely on these instances
1688 // to carry over into recursions or invocations from other threads.
1689 fdStream out(fd_out);
1690 out.set_scratch_buffer(buffer, sizeof(buffer));
1691
1692 // Depending on the re-entrance depth at this point, fd_log may be -1 or point to an open hs-err file.
1693 fdStream log(fd_log);
1694 log.set_scratch_buffer(buffer, sizeof(buffer));
1695
1696 // How many errors occurred in error handler when reporting first_error.
1697 static int recursive_error_count;
1698
1699 // We will first print a brief message to standard out (verbose = false),
1700 // then save detailed information in log file (verbose = true).
1701 static bool out_done = false; // done printing to standard out
1702 static bool log_done = false; // done saving error log
1703
1704 intptr_t mytid = os::current_thread_id();
1705 if (_first_error_tid == -1 &&
1706 AtomicAccess::cmpxchg(&_first_error_tid, (intptr_t)-1, mytid) == -1) {
1707
1708 if (SuppressFatalErrorMessage) {
1709 os::abort(CreateCoredumpOnCrash);
1710 }
1711
1712 // Initialize time stamps to use the same base.
1713 out.time_stamp().update_to(1);
1714 log.time_stamp().update_to(1);
1715
1716 _id = id;
1717 _message = message;
1718 _thread = thread;
1719 _pc = pc;
1720 _siginfo = siginfo;
1721 _context = context;
1722 _filename = filename;
1723 _lineno = lineno;
1724 _size = size;
1725 jio_vsnprintf(_detail_msg, sizeof(_detail_msg), detail_fmt, detail_args);
1726
1727 reporting_started();
1728 if (!TestUnresponsiveErrorHandler) {
1729 // Record reporting_start_time unless we're running the
1730 // TestUnresponsiveErrorHandler test. For that test we record
1731 // reporting_start_time at the beginning of the test.
1732 record_reporting_start_time();
1733 } else {
1734 out.print_raw_cr("Delaying recording reporting_start_time for TestUnresponsiveErrorHandler.");
1735 }
1736
1737 if (ShowMessageBoxOnError || PauseAtExit) {
1738 show_message_box(buffer, sizeof(buffer));
1739
1740 // User has asked JVM to abort. Reset ShowMessageBoxOnError so the
1741 // WatcherThread can kill JVM if the error handler hangs.
1742 ShowMessageBoxOnError = false;
1743 }
1744
1745 os::check_core_dump_prerequisites(buffer, sizeof(buffer));
1746
1747 // reset signal handlers or exception filter; make sure recursive crashes
1748 // are handled properly.
1749 install_secondary_signal_handler();
1750 } else {
1751 // This is not the first error, see if it happened in a different thread
1752 // or in the same thread during error reporting.
1753 if (_first_error_tid != mytid) {
1754 if (!SuppressFatalErrorMessage) {
1755 char msgbuf[64];
1756 jio_snprintf(msgbuf, sizeof(msgbuf),
1757 "[thread %zd also had an error]",
1758 mytid);
1759 out.print_raw_cr(msgbuf);
1760 }
1761
1762 // Error reporting is not MT-safe, nor can we let the current thread
1763 // proceed, so we block it.
1764 os::infinite_sleep();
1765
1766 } else {
1767 if (recursive_error_count++ > 30) {
1768 if (!SuppressFatalErrorMessage) {
1769 out.print_raw_cr("[Too many errors, abort]");
1770 }
1771 os::die();
1772 }
1773
1774 if (SuppressFatalErrorMessage) {
1775 // If we already hit a secondary error during abort, then calling
1776 // it again is likely to hit another one. But eventually, if we
1777 // don't deadlock somewhere, we will call os::die() above.
1778 os::abort(CreateCoredumpOnCrash);
1779 }
1780
1781 outputStream* const st = log.is_open() ? &log : &out;
1782 st->cr();
1783
1784 // Timeout handling.
1785 if (_step_did_timeout) {
1786 // The current step had a timeout. Lets continue reporting with the next step.
1787 st->print_raw("[timeout occurred during error reporting in step \"");
1788 st->print_raw(_current_step_info);
1789 st->print_cr("\"] after " INT64_FORMAT " s.",
1790 (int64_t)
1791 ((get_current_timestamp() - _step_start_time) / TIMESTAMP_TO_SECONDS_FACTOR));
1792 } else if (_reporting_did_timeout) {
1793 // We hit ErrorLogTimeout. Reporting will stop altogether. Let's wrap things
1794 // up, the process is about to be stopped by the WatcherThread.
1795 st->print_cr("------ Timeout during error reporting after " INT64_FORMAT " s. ------",
1796 (int64_t)
1797 ((get_current_timestamp() - _reporting_start_time) / TIMESTAMP_TO_SECONDS_FACTOR));
1798 st->flush();
1799 // Watcherthread is about to call os::die. Lets just wait.
1800 os::infinite_sleep();
1801 } else {
1802 // A secondary error happened. Print brief information, but take care, since crashing
1803 // here would just recurse endlessly.
1804 // Any information (signal, context, siginfo etc) printed here should use the function
1805 // arguments, not the information stored in *this, since those describe the primary crash.
1806 static char tmp[256]; // cannot use global scratch buffer
1807 // Note: this string does get parsed by a number of jtreg tests,
1808 // see hotspot/jtreg/runtime/ErrorHandling.
1809 st->print("[error occurred during error reporting (%s), id 0x%x",
1810 _current_step_info, id);
1811 if (os::exception_name(id, tmp, sizeof(tmp))) {
1812 st->print(", %s (0x%x) at pc=" PTR_FORMAT, tmp, id, p2i(pc));
1813 } else {
1814 if (should_report_bug(id)) {
1815 st->print(", Internal Error (%s:%d)",
1816 filename == nullptr ? "??" : filename, lineno);
1817 } else {
1818 st->print(", Out of Memory Error (%s:%d)",
1819 filename == nullptr ? "??" : filename, lineno);
1820 }
1821 }
1822 st->print_cr("]");
1823 if (ErrorLogSecondaryErrorDetails) {
1824 static bool recursed = false;
1825 if (!recursed) {
1826 recursed = true;
1827 // Print even more information for secondary errors. This may generate a lot of output
1828 // and possibly disturb error reporting, therefore its optional and only available in debug builds.
1829 if (siginfo != nullptr) {
1830 st->print("[");
1831 os::print_siginfo(st, siginfo);
1832 st->print_cr("]");
1833 }
1834 st->print("[stack: ");
1835 NativeStackPrinter nsp(_thread, context, _filename != nullptr ? get_filename_only() : nullptr, _lineno);
1836 // Subsequent secondary errors build up stack; to avoid flooding the hs-err file with irrelevant
1837 // call stacks, limit the stack we print here (we are only interested in what happened before the
1838 // last assert/fault).
1839 const int max_stack_size = 15;
1840 nsp.print_stack_from_frame(st, tmp, sizeof(tmp), true /* print_source_info */, max_stack_size);
1841 st->print_cr("]");
1842 } // !recursed
1843 recursed = false; // Note: reset outside !recursed
1844 }
1845 }
1846 }
1847 }
1848
1849 // Part 1: print an abbreviated version (the '#' section) to stdout.
1850 if (!out_done) {
1851 // Suppress this output if we plan to print Part 2 to stdout too.
1852 // No need to have the "#" section twice.
1853 if (!(ErrorFileToStdout && out.fd() == 1)) {
1854 report(&out, false);
1855 }
1856
1857 out_done = true;
1858
1859 _current_step = 0;
1860 _current_step_info = "";
1861 }
1862
1863 // Part 2: print a full error log file (optionally to stdout or stderr).
1864 // print to error log file
1865 if (!log_done) {
1866 // see if log file is already open
1867 if (!log.is_open()) {
1868 // open log file
1869 if (ErrorFileToStdout) {
1870 fd_log = 1;
1871 } else if (ErrorFileToStderr) {
1872 fd_log = 2;
1873 } else {
1874 fd_log = prepare_log_file(ErrorFile, "hs_err_pid%p.log", true,
1875 buffer, sizeof(buffer));
1876 if (fd_log != -1) {
1877 out.print_raw("# An error report file with more information is saved as:\n# ");
1878 out.print_raw_cr(buffer);
1879 } else {
1880 out.print_raw_cr("# Can not save log file, dump to screen..");
1881 fd_log = 1;
1882 }
1883 }
1884 log.set_fd(fd_log);
1885 }
1886
1887 report(&log, true);
1888 log_done = true;
1889 _current_step = 0;
1890 _current_step_info = "";
1891
1892 if (fd_log > 3) {
1893 ::close(fd_log);
1894 fd_log = -1;
1895 }
1896
1897 log.set_fd(-1);
1898 }
1899
1900 JFR_ONLY(Jfr::on_vm_shutdown(static_cast<VMErrorType>(_id) == OOM_JAVA_HEAP_FATAL, true);)
1901
1902 if (PrintNMTStatistics) {
1903 fdStream fds(fd_out);
1904 MemTracker::final_report(&fds);
1905 }
1906
1907 static bool skip_replay = ReplayCompiles && !ReplayReduce; // Do not overwrite file during replay
1908 if (DumpReplayDataOnError && _thread && _thread->is_Compiler_thread() && !skip_replay) {
1909 skip_replay = true;
1910 ciEnv* env = ciEnv::current();
1911 if (env != nullptr && env->task() != nullptr) {
1912 const bool overwrite = false; // We do not overwrite an existing replay file.
1913 int fd = prepare_log_file(ReplayDataFile, "replay_pid%p.log", overwrite, buffer, sizeof(buffer));
1914 if (fd != -1) {
1915 FILE* replay_data_file = os::fdopen(fd, "w");
1916 if (replay_data_file != nullptr) {
1917 fileStream replay_data_stream(replay_data_file, /*need_close=*/true);
1918 env->dump_replay_data_unsafe(&replay_data_stream);
1919 out.print_raw("#\n# Compiler replay data is saved as:\n# ");
1920 out.print_raw_cr(buffer);
1921 } else {
1922 int e = errno;
1923 out.print_raw("#\n# Can't open file to dump replay data. Error: ");
1924 out.print_raw_cr(os::strerror(e));
1925 close(fd);
1926 }
1927 }
1928 }
1929 }
1930
1931 #if INCLUDE_JVMCI
1932 if (JVMCI::fatal_log_filename() != nullptr) {
1933 out.print_raw("#\n# The JVMCI shared library error report file is saved as:\n# ");
1934 out.print_raw_cr(JVMCI::fatal_log_filename());
1935 }
1936 #endif
1937
1938 static bool skip_bug_url = !should_submit_bug_report(_id);
1939 if (!skip_bug_url) {
1940 skip_bug_url = true;
1941
1942 out.print_raw_cr("#");
1943 print_bug_submit_message(&out, _thread);
1944 }
1945
1946 static bool skip_OnError = false;
1947 if (!skip_OnError && OnError && OnError[0]) {
1948 skip_OnError = true;
1949
1950 // Flush output and finish logs before running OnError commands.
1951 ostream_abort();
1952
1953 out.print_raw_cr("#");
1954 out.print_raw ("# -XX:OnError=\"");
1955 out.print_raw (OnError);
1956 out.print_raw_cr("\"");
1957
1958 char* cmd;
1959 const char* ptr = OnError;
1960 while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr)) != nullptr){
1961 out.print_raw ("# Executing ");
1962 #if defined(LINUX) || defined(_ALLBSD_SOURCE)
1963 out.print_raw ("/bin/sh -c ");
1964 #elif defined(_WINDOWS)
1965 out.print_raw ("cmd /C ");
1966 #endif
1967 out.print_raw ("\"");
1968 out.print_raw (cmd);
1969 out.print_raw_cr("\" ...");
1970
1971 if (os::fork_and_exec(cmd) < 0) {
1972 out.print_cr("os::fork_and_exec failed: %s (%s=%d)",
1973 os::strerror(errno), os::errno_name(errno), errno);
1974 }
1975 }
1976
1977 // done with OnError
1978 OnError = nullptr;
1979 }
1980
1981 #if defined _WINDOWS
1982 if (UseOSErrorReporting) {
1983 raise_fail_fast(_siginfo, _context);
1984 }
1985 #endif // _WINDOWS
1986
1987 // os::abort() will call abort hooks, try it first.
1988 static bool skip_os_abort = false;
1989 if (!skip_os_abort) {
1990 skip_os_abort = true;
1991 bool dump_core = should_report_bug(_id);
1992 os::abort(dump_core && CreateCoredumpOnCrash, _siginfo, _context);
1993 // if os::abort() doesn't abort, try os::die();
1994 }
1995 os::die();
1996 }
1997
1998 /*
1999 * OnOutOfMemoryError scripts/commands executed while VM is a safepoint - this
2000 * ensures utilities such as jmap can observe the process is a consistent state.
2001 */
2002 class VM_ReportJavaOutOfMemory : public VM_Operation {
2003 private:
2004 const char* _message;
2005 public:
2006 VM_ReportJavaOutOfMemory(const char* message) { _message = message; }
2007 VMOp_Type type() const { return VMOp_ReportJavaOutOfMemory; }
2008 void doit();
2009 };
2010
2011 void VM_ReportJavaOutOfMemory::doit() {
2012 // Don't allocate large buffer on stack
2013 static char buffer[O_BUFLEN];
2014
2015 tty->print_cr("#");
2016 tty->print_cr("# java.lang.OutOfMemoryError: %s", _message);
2017 tty->print_cr("# -XX:OnOutOfMemoryError=\"%s\"", OnOutOfMemoryError);
2018
2019 // make heap parsability
2020 Universe::heap()->ensure_parsability(false); // no need to retire TLABs
2021
2022 char* cmd;
2023 const char* ptr = OnOutOfMemoryError;
2024 while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr)) != nullptr){
2025 tty->print("# Executing ");
2026 #if defined(LINUX)
2027 tty->print ("/bin/sh -c ");
2028 #endif
2029 tty->print_cr("\"%s\"...", cmd);
2030
2031 if (os::fork_and_exec(cmd) < 0) {
2032 tty->print_cr("os::fork_and_exec failed: %s (%s=%d)",
2033 os::strerror(errno), os::errno_name(errno), errno);
2034 }
2035 }
2036 }
2037
2038 void VMError::report_java_out_of_memory(const char* message) {
2039 if (OnOutOfMemoryError && OnOutOfMemoryError[0]) {
2040 MutexLocker ml(Heap_lock);
2041 VM_ReportJavaOutOfMemory op(message);
2042 VMThread::execute(&op);
2043 }
2044 }
2045
2046 void VMError::show_message_box(char *buf, int buflen) {
2047 bool yes;
2048 do {
2049 error_string(buf, buflen);
2050 yes = os::start_debugging(buf,buflen);
2051 } while (yes);
2052 }
2053
2054 // Fatal error handling is subject to several timeouts:
2055 // - a global timeout (controlled via ErrorLogTimeout)
2056 // - local error reporting step timeouts.
2057 //
2058 // The latter aims to "give the JVM a kick" if it gets stuck in one particular place during
2059 // error reporting. This prevents one error reporting step from hogging all the time allotted
2060 // to error reporting under ErrorLogTimeout.
2061 //
2062 // VMError::check_timeout() is called from the watcher thread and checks for either global
2063 // or step timeout. If a timeout happened, we interrupt the reporting thread and set either
2064 // _reporting_did_timeout or _step_did_timeout to signal which timeout fired. Function returns
2065 // true if the *global* timeout fired, which will cause WatcherThread to shut down the JVM
2066 // immediately.
2067 bool VMError::check_timeout() {
2068
2069 // This function is supposed to be called from watcher thread during fatal error handling only.
2070 assert(VMError::is_error_reported(), "Only call during error handling");
2071 assert(Thread::current()->is_Watcher_thread(), "Only call from watcher thread");
2072
2073 if (ErrorLogTimeout == 0) {
2074 return false;
2075 }
2076
2077 // There are three situations where we suppress the *global* error timeout:
2078 // - if the JVM is embedded and the launcher has its abort hook installed.
2079 // That must be allowed to run.
2080 // - if the user specified one or more OnError commands to run, and these
2081 // did not yet run. These must have finished.
2082 // - if the user (typically developer) specified ShowMessageBoxOnError,
2083 // and the error box has not yet been shown
2084 const bool ignore_global_timeout =
2085 (ShowMessageBoxOnError
2086 || (OnError != nullptr && OnError[0] != '\0')
2087 || Arguments::abort_hook() != nullptr);
2088
2089 const jlong now = get_current_timestamp();
2090
2091 // Global timeout hit?
2092 if (!ignore_global_timeout) {
2093 const jlong reporting_start_time = get_reporting_start_time();
2094 // Timestamp is stored in nanos.
2095 if (reporting_start_time > 0) {
2096 const jlong end = reporting_start_time + (jlong)ErrorLogTimeout * TIMESTAMP_TO_SECONDS_FACTOR;
2097 if (end <= now && !_reporting_did_timeout) {
2098 // We hit ErrorLogTimeout and we haven't interrupted the reporting
2099 // thread yet.
2100 _reporting_did_timeout = true;
2101 interrupt_reporting_thread();
2102 return true; // global timeout
2103 }
2104 }
2105 }
2106
2107 // Reporting step timeout?
2108 const jlong step_start_time = get_step_start_time();
2109 if (step_start_time > 0) {
2110 // A step times out after a quarter of the total timeout. Steps are mostly fast unless they
2111 // hang for some reason, so this simple rule allows for three hanging step and still
2112 // hopefully leaves time enough for the rest of the steps to finish.
2113 const int max_step_timeout_secs = 5;
2114 const jlong timeout_duration = MAX2((jlong)max_step_timeout_secs, (jlong)ErrorLogTimeout * TIMESTAMP_TO_SECONDS_FACTOR / 4);
2115 const jlong end = step_start_time + timeout_duration;
2116 if (end <= now && !_step_did_timeout) {
2117 // The step timed out and we haven't interrupted the reporting
2118 // thread yet.
2119 _step_did_timeout = true;
2120 interrupt_reporting_thread();
2121 return false; // (Not a global timeout)
2122 }
2123 }
2124
2125 return false;
2126
2127 }
2128
2129 #ifdef ASSERT
2130 typedef void (*voidfun_t)();
2131
2132 // Crash with an authentic sigfpe; behavior is subtly different from a real signal
2133 // compared to one generated with raise (asynchronous vs synchronous). See JDK-8065895.
2134 volatile int sigfpe_int = 0;
2135
2136 ATTRIBUTE_NO_UBSAN
2137 static void ALWAYSINLINE crash_with_sigfpe() {
2138
2139 // generate a native synchronous SIGFPE where possible;
2140 sigfpe_int = sigfpe_int/sigfpe_int;
2141
2142 // if that did not cause a signal (e.g. on ppc), just
2143 // raise the signal.
2144 #ifndef _WIN32
2145 // OSX implements raise(sig) incorrectly so we need to
2146 // explicitly target the current thread
2147 pthread_kill(pthread_self(), SIGFPE);
2148 #endif
2149
2150 } // end: crash_with_sigfpe
2151
2152 // crash with sigsegv at non-null address.
2153 static void ALWAYSINLINE crash_with_segfault() {
2154
2155 int* crash_addr = reinterpret_cast<int*>(VMError::segfault_address);
2156 *crash_addr = 1;
2157
2158 } // end: crash_with_segfault
2159
2160 // crash in a controlled way:
2161 // 1 - assert
2162 // 2 - guarantee
2163 // 14 - SIGSEGV
2164 // 15 - SIGFPE
2165 void VMError::controlled_crash(int how) {
2166
2167 // Case 14 is tested by test/hotspot/jtreg/runtime/ErrorHandling/SafeFetchInErrorHandlingTest.java.
2168 // Case 15 is tested by test/hotspot/jtreg/runtime/ErrorHandling/SecondaryErrorTest.java.
2169 // Case 16 is tested by test/hotspot/jtreg/runtime/ErrorHandling/ThreadsListHandleInErrorHandlingTest.java.
2170 // Case 17 is tested by test/hotspot/jtreg/runtime/ErrorHandling/NestedThreadsListHandleInErrorHandlingTest.java.
2171
2172 // We try to grab Threads_lock to keep ThreadsSMRSupport::print_info_on()
2173 // from racing with Threads::add() or Threads::remove() as we
2174 // generate the hs_err_pid file. This makes our ErrorHandling tests
2175 // more stable.
2176 if (!Threads_lock->owned_by_self()) {
2177 Threads_lock->try_lock();
2178 // The VM is going to die so no need to unlock Thread_lock.
2179 }
2180
2181 switch (how) {
2182 case 1: assert(how == 0, "test assert"); break;
2183 case 2: guarantee(how == 0, "test guarantee"); break;
2184
2185 // The other cases are unused.
2186 case 14: crash_with_segfault(); break;
2187 case 15: crash_with_sigfpe(); break;
2188 case 16: {
2189 ThreadsListHandle tlh;
2190 fatal("Force crash with an active ThreadsListHandle.");
2191 }
2192 case 17: {
2193 ThreadsListHandle tlh;
2194 {
2195 ThreadsListHandle tlh2;
2196 fatal("Force crash with a nested ThreadsListHandle.");
2197 }
2198 }
2199 case 18: {
2200 // Trigger an error that should cause ASAN to report a double free or use-after-free.
2201 // Please note that this is not 100% bullet-proof since it assumes that this block
2202 // is not immediately repurposed by some other thread after free.
2203 void* const p = os::malloc(4096, mtTest);
2204 os::free(p);
2205 os::free(p);
2206 }
2207 default:
2208 // If another number is given, give a generic crash.
2209 fatal("Crashing with number %d", how);
2210 }
2211 tty->print_cr("controlled_crash: survived intentional crash. Did you suppress the assert?");
2212 ShouldNotReachHere();
2213 }
2214 #endif // !ASSERT
2215
2216 VMErrorCallbackMark::VMErrorCallbackMark(VMErrorCallback* callback)
2217 : _thread(Thread::current()) {
2218 callback->_next = _thread->_vm_error_callbacks;
2219 _thread->_vm_error_callbacks = callback;
2220 }
2221
2222 VMErrorCallbackMark::~VMErrorCallbackMark() {
2223 assert(_thread->_vm_error_callbacks != nullptr, "Popped too far");
2224 _thread->_vm_error_callbacks = _thread->_vm_error_callbacks->_next;
2225 }