1 /*
2 * Copyright (c) 2003, 2026, 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/flags/jvmFlag.hpp"
47 #include "runtime/frame.inline.hpp"
48 #include "runtime/init.hpp"
49 #include "runtime/javaThread.inline.hpp"
50 #include "runtime/os.inline.hpp"
51 #include "runtime/osThread.hpp"
52 #include "runtime/safefetch.hpp"
53 #include "runtime/safepointMechanism.hpp"
54 #include "runtime/stackFrameStream.inline.hpp"
55 #include "runtime/stackOverflow.hpp"
56 #include "runtime/threads.hpp"
57 #include "runtime/threadSMR.hpp"
58 #include "runtime/trimNativeHeap.hpp"
59 #include "runtime/vm_version.hpp"
60 #include "runtime/vmOperations.hpp"
61 #include "runtime/vmThread.hpp"
62 #include "sanitizers/address.hpp"
63 #include "sanitizers/ub.hpp"
64 #include "utilities/debug.hpp"
65 #include "utilities/decoder.hpp"
66 #include "utilities/defaultStream.hpp"
67 #include "utilities/events.hpp"
68 #include "utilities/globalDefinitions.hpp"
69 #include "utilities/macros.hpp"
70 #include "utilities/nativeStackPrinter.hpp"
71 #include "utilities/ostream.hpp"
72 #include "utilities/vmError.hpp"
73 #if INCLUDE_JFR
74 #include "jfr/jfr.hpp"
75 #endif
76 #if INCLUDE_JVMCI
77 #include "jvmci/jvmci.hpp"
78 #endif
79
80 #ifndef PRODUCT
81 #include <signal.h>
82 #endif // PRODUCT
83
84 bool VMError::coredump_status;
85 char VMError::coredump_message[O_BUFLEN];
86 int VMError::_current_step;
87 const char* VMError::_current_step_info;
88 Atomic<jlong> VMError::_reporting_start_time{-1};
89 Atomic<bool> VMError::_reporting_did_timeout{false};
90 Atomic<jlong> VMError::_step_start_time{-1};
91 Atomic<bool> VMError::_step_did_timeout{false};
92 Atomic<bool> VMError::_step_did_succeed{false};
93 Atomic<intptr_t> VMError::_first_error_tid{-1};
94 int VMError::_id;
95 const char* VMError::_message;
96 char VMError::_detail_msg[1024];
97 Thread* VMError::_thread;
98 address VMError::_pc;
99 const void* VMError::_siginfo;
100 const void* VMError::_context;
101 bool VMError::_print_stack_from_frame_used = false;
102 const char* VMError::_filename;
103 int VMError::_lineno;
104 size_t VMError::_size;
105 const size_t VMError::_reattempt_required_stack_headroom = 64 * K;
106 const intptr_t VMError::segfault_address = pd_segfault_address;
107 Atomic<Thread*> VMError::_handshake_timed_out_thread{};
108 Atomic<Thread*> VMError::_safepoint_timed_out_thread{};
109
110 // List of environment variables that should be reported in error log file.
111 static const char* env_list[] = {
112 // All platforms
113 "JAVA_HOME", "JAVA_TOOL_OPTIONS", "_JAVA_OPTIONS", "CLASSPATH",
114 "JDK_AOT_VM_OPTIONS",
115 "JAVA_OPTS", "PATH", "USERNAME",
116
117 "XDG_CACHE_HOME", "XDG_CONFIG_HOME", "FC_LANG", "FONTCONFIG_USE_MMAP",
118
119 // Env variables that are defined on Linux/BSD
120 "LD_LIBRARY_PATH", "LD_PRELOAD", "SHELL", "DISPLAY", "WAYLAND_DISPLAY",
121 "HOSTTYPE", "OSTYPE", "ARCH", "MACHTYPE",
122 "LANG", "LC_ALL", "LC_CTYPE", "LC_NUMERIC", "LC_TIME",
123 "TERM", "TMPDIR", "TZ",
124
125 // defined on AIX
126 "LIBPATH", "LDR_PRELOAD", "LDR_PRELOAD64",
127
128 // defined on Linux/AIX/BSD
129 "_JAVA_SR_SIGNUM",
130
131 // defined on Darwin
132 "DYLD_LIBRARY_PATH", "DYLD_FALLBACK_LIBRARY_PATH",
133 "DYLD_FRAMEWORK_PATH", "DYLD_FALLBACK_FRAMEWORK_PATH",
134 "DYLD_INSERT_LIBRARIES",
135
136 // defined on Windows
137 "OS", "PROCESSOR_IDENTIFIER", "_ALT_JAVA_HOME_DIR", "TMP", "TEMP",
138
139 nullptr // End marker.
140 };
141
142 // A simple parser for lists of commands such as -XX:OnError and -XX:OnOutOfMemoryError
143 // Command list (ptr) is expected to be a sequence of commands delineated by semicolons and/or newlines.
144 // Usage:
145 // ptr = OnError;
146 // while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr) != nullptr)
147 // ... ...
148 static char* next_OnError_command(char* buf, int buflen, const char** ptr) {
149 if (ptr == nullptr || *ptr == nullptr) return nullptr;
150
151 const char* cmd = *ptr;
152
153 // skip leading blanks, ';' or newlines
154 while (*cmd == ' ' || *cmd == ';' || *cmd == '\n') cmd++;
155
156 if (*cmd == '\0') return nullptr;
157
158 const char * cmdend = cmd;
159 while (*cmdend != '\0' && *cmdend != ';' && *cmdend != '\n') cmdend++;
160
161 Arguments::copy_expand_pid(cmd, cmdend - cmd, buf, buflen);
162
163 *ptr = (*cmdend == '\0' ? cmdend : cmdend + 1);
164 return buf;
165 }
166
167 static void print_bug_submit_message(outputStream *out, Thread *thread) {
168 if (out == nullptr) return;
169 const char *url = Arguments::java_vendor_url_bug();
170 if (url == nullptr || *url == '\0')
171 url = JDK_Version::runtime_vendor_vm_bug_url();
172 if (url != nullptr && *url != '\0') {
173 out->print_raw_cr("# If you would like to submit a bug report, please visit:");
174 out->print_raw ("# ");
175 out->print_raw_cr(url);
176 }
177 // If the crash is in native code, encourage user to submit a bug to the
178 // provider of that code.
179 if (thread && thread->is_Java_thread() &&
180 !thread->is_hidden_from_external_view()) {
181 if (JavaThread::cast(thread)->thread_state() == _thread_in_native) {
182 out->print_cr("# The crash happened outside the Java Virtual Machine in native code.\n# See problematic frame for where to report the bug.");
183 }
184 }
185 out->print_raw_cr("#");
186 }
187
188 static bool stack_has_headroom(size_t headroom) {
189 size_t stack_size = 0;
190 address stack_base = nullptr;
191 os::current_stack_base_and_size(&stack_base, &stack_size);
192
193 const size_t guard_size = StackOverflow::stack_guard_zone_size();
194 const size_t unguarded_stack_size = stack_size - guard_size;
195
196 if (unguarded_stack_size < headroom) {
197 return false;
198 }
199
200 const address unguarded_stack_end = stack_base - unguarded_stack_size;
201 const address stack_pointer = os::current_stack_pointer();
202
203 return stack_pointer >= unguarded_stack_end + headroom;
204 }
205
206 #ifdef ASSERT
207 PRAGMA_DIAG_PUSH
208 PRAGMA_INFINITE_RECURSION_IGNORED
209 void VMError::reattempt_test_hit_stack_limit(outputStream* st) {
210 if (stack_has_headroom(_reattempt_required_stack_headroom)) {
211 // Use all but (_reattempt_required_stack_headroom - K) unguarded stack space.
212 size_t stack_size = 0;
213 address stack_base = nullptr;
214 os::current_stack_base_and_size(&stack_base, &stack_size);
215
216 const size_t guard_size = StackOverflow::stack_guard_zone_size();
217 const address stack_pointer = os::current_stack_pointer();
218
219 const size_t unguarded_stack_size = stack_size - guard_size;
220 const address unguarded_stack_end = stack_base - unguarded_stack_size;
221 const size_t available_headroom = stack_pointer - unguarded_stack_end;
222 const size_t allocation_size = available_headroom - _reattempt_required_stack_headroom + K;
223
224 st->print_cr("Current Stack Pointer: " PTR_FORMAT " alloca %zu"
225 " of %zu bytes available unguarded stack space",
226 p2i(stack_pointer), allocation_size, available_headroom);
227
228 // Allocate byte blob on the stack. Make pointer volatile to avoid having
229 // the compiler removing later reads.
230 volatile char* stack_buffer = static_cast<char*>(alloca(allocation_size));
231 // Initialize the last byte.
232 stack_buffer[allocation_size - 1] = '\0';
233 // Recursive call should hit the stack limit.
234 reattempt_test_hit_stack_limit(st);
235 // Perform a volatile read of the last byte to avoid having the complier
236 // remove the allocation.
237 static_cast<void>(stack_buffer[allocation_size - 1] == '\0');
238 }
239 controlled_crash(14);
240 }
241 PRAGMA_DIAG_POP
242 #endif // ASSERT
243
244 bool VMError::can_reattempt_step(const char* &stop_reason) {
245 if (!stack_has_headroom(_reattempt_required_stack_headroom)) {
246 stop_reason = "Stack headroom limit reached";
247 return false;
248 }
249
250 if (_step_did_timeout.load_relaxed()) {
251 stop_reason = "Step time limit reached";
252 return false;
253 }
254
255 return true;
256 }
257
258 void VMError::record_coredump_status(const char* message, bool status) {
259 coredump_status = status;
260 strncpy(coredump_message, message, sizeof(coredump_message));
261 coredump_message[sizeof(coredump_message)-1] = 0;
262 }
263
264 // Return a string to describe the error
265 char* VMError::error_string(char* buf, int buflen) {
266 char signame_buf[64];
267 const char *signame = os::exception_name(_id, signame_buf, sizeof(signame_buf));
268
269 if (signame) {
270 jio_snprintf(buf, buflen,
271 "%s (0x%x) at pc=" PTR_FORMAT ", pid=%d, tid=%zu",
272 signame, _id, p2i(_pc),
273 os::current_process_id(), os::current_thread_id());
274 } else if (_filename != nullptr && _lineno > 0) {
275 // skip directory names
276 int n = jio_snprintf(buf, buflen,
277 "Internal Error at %s:%d, pid=%d, tid=%zu",
278 get_filename_only(), _lineno,
279 os::current_process_id(), os::current_thread_id());
280 if (n >= 0 && n < buflen && _message) {
281 if (strlen(_detail_msg) > 0) {
282 jio_snprintf(buf + n, buflen - n, "%s%s: %s",
283 os::line_separator(), _message, _detail_msg);
284 } else {
285 jio_snprintf(buf + n, buflen - n, "%sError: %s",
286 os::line_separator(), _message);
287 }
288 }
289 } else {
290 jio_snprintf(buf, buflen,
291 "Internal Error (0x%x), pid=%d, tid=%zu",
292 _id, os::current_process_id(), os::current_thread_id());
293 }
294
295 return buf;
296 }
297
298 void VMError::print_stack_trace(outputStream* st, JavaThread* jt,
299 char* buf, int buflen, bool verbose) {
300 #ifdef ZERO
301 if (jt->zero_stack()->sp() && jt->top_zero_frame()) {
302 // StackFrameStream uses the frame anchor, which may not have
303 // been set up. This can be done at any time in Zero, however,
304 // so if it hasn't been set up then we just set it up now and
305 // clear it again when we're done.
306 bool has_last_Java_frame = jt->has_last_Java_frame();
307 if (!has_last_Java_frame)
308 jt->set_last_Java_frame();
309 st->print("Java frames:");
310 st->cr();
311
312 // Print the frames
313 StackFrameStream sfs(jt, true /* update */, true /* process_frames */);
314 for(int i = 0; !sfs.is_done(); sfs.next(), i++) {
315 sfs.current()->zero_print_on_error(i, st, buf, buflen);
316 st->cr();
317 }
318
319 // Reset the frame anchor if necessary
320 if (!has_last_Java_frame)
321 jt->reset_last_Java_frame();
322 }
323 #else
324 if (jt->has_last_Java_frame()) {
325 st->print_cr("Java frames: (J=compiled Java code, j=interpreted, Vv=VM code)");
326 for (StackFrameStream sfs(jt, true /* update */, true /* process_frames */); !sfs.is_done(); sfs.next()) {
327 sfs.current()->print_on_error(st, buf, buflen, verbose);
328 st->cr();
329 }
330 }
331 #endif // ZERO
332 }
333
334 const char* VMError::get_filename_only() {
335 char separator = os::file_separator()[0];
336 const char* p = strrchr(_filename, separator);
337 return p ? p + 1 : _filename;
338 }
339
340 /**
341 * Adds `value` to `list` iff it's not already present and there is sufficient
342 * capacity (i.e. length(list) < `list_capacity`). The length of the list
343 * is the index of the first nullptr entry or `list_capacity` if there are
344 * no nullptr entries.
345 *
346 * @ return true if the value was added, false otherwise
347 */
348 static bool add_if_absent(address value, address* list, int list_capacity) {
349 for (int i = 0; i < list_capacity; i++) {
350 if (list[i] == value) {
351 return false;
352 }
353 if (list[i] == nullptr) {
354 list[i] = value;
355 if (i + 1 < list_capacity) {
356 list[i + 1] = nullptr;
357 }
358 return true;
359 }
360 }
361 return false;
362 }
363
364 /**
365 * Prints the VM generated code unit, if any, containing `pc` if it has not already
366 * been printed. If the code unit is an InterpreterCodelet or StubCodeDesc, it is
367 * only printed if `is_crash_pc` is true.
368 *
369 * @param printed array of code units that have already been printed (delimited by nullptr entry)
370 * @param printed_capacity the capacity of `printed`
371 * @return true if the code unit was printed, false otherwise
372 */
373 static bool print_code(outputStream* st, Thread* thread, address pc, bool is_crash_pc,
374 address* printed, int printed_capacity) {
375 if (Interpreter::contains(pc)) {
376 if (is_crash_pc) {
377 // The interpreter CodeBlob is very large so try to print the codelet instead.
378 InterpreterCodelet* codelet = Interpreter::codelet_containing(pc);
379 if (codelet != nullptr) {
380 if (add_if_absent((address) codelet, printed, printed_capacity)) {
381 codelet->print_on(st);
382 Disassembler::decode(codelet->code_begin(), codelet->code_end(), st);
383 return true;
384 }
385 }
386 }
387 } else {
388 StubCodeDesc* desc = StubCodeDesc::desc_for(pc);
389 if (desc != nullptr) {
390 if (is_crash_pc) {
391 if (add_if_absent((address) desc, printed, printed_capacity)) {
392 desc->print_on(st);
393 Disassembler::decode(desc->begin(), desc->end(), st);
394 return true;
395 }
396 }
397 } else if (thread != nullptr) {
398 CodeBlob* cb = CodeCache::find_blob(pc);
399 if (cb != nullptr && add_if_absent((address) cb, printed, printed_capacity)) {
400 // Disassembling nmethod will incur resource memory allocation,
401 // only do so when thread is valid.
402 ResourceMark rm(thread);
403 Disassembler::decode(cb, st);
404 st->cr();
405 return true;
406 }
407 }
408 }
409 return false;
410 }
411
412 // Like above, but only try to figure out a short name. Return nullptr if not found.
413 static const char* find_code_name(address pc) {
414 if (Interpreter::contains(pc)) {
415 InterpreterCodelet* codelet = Interpreter::codelet_containing(pc);
416 if (codelet != nullptr) {
417 return codelet->description();
418 }
419 } else {
420 StubCodeDesc* desc = StubCodeDesc::desc_for(pc);
421 if (desc != nullptr) {
422 return desc->name();
423 } else {
424 CodeBlob* cb = CodeCache::find_blob(pc);
425 if (cb != nullptr) {
426 return cb->name();
427 }
428 }
429 }
430 return nullptr;
431 }
432
433 static void print_oom_reasons(outputStream* st) {
434 st->print_cr("# Possible reasons:");
435 st->print_cr("# The system is out of physical RAM or swap space");
436 #ifdef LINUX
437 st->print_cr("# This process has exceeded the maximum number of memory mappings (check below");
438 st->print_cr("# for `/proc/sys/vm/max_map_count` and `Total number of mappings`)");
439 #endif
440 if (UseCompressedOops) {
441 st->print_cr("# This process is running with CompressedOops enabled, and the Java Heap may be blocking the growth of the native heap");
442 }
443 if (LogBytesPerWord == 2) {
444 st->print_cr("# In 32 bit mode, the process size limit was hit");
445 }
446 st->print_cr("# Possible solutions:");
447 st->print_cr("# Reduce memory load on the system");
448 st->print_cr("# Increase physical memory or swap space");
449 st->print_cr("# Check if swap backing store is full");
450 if (LogBytesPerWord == 2) {
451 st->print_cr("# Use 64 bit Java on a 64 bit OS");
452 }
453 st->print_cr("# Decrease Java heap size (-Xmx/-Xms)");
454 st->print_cr("# Decrease number of Java threads");
455 st->print_cr("# Decrease Java thread stack sizes (-Xss)");
456 st->print_cr("# Set larger code cache with -XX:ReservedCodeCacheSize=");
457 if (UseCompressedOops) {
458 switch (CompressedOops::mode()) {
459 case CompressedOops::UnscaledNarrowOop:
460 st->print_cr("# JVM is running with Unscaled Compressed Oops mode in which the Java heap is");
461 st->print_cr("# placed in the first 4GB address space. The Java Heap base address is the");
462 st->print_cr("# maximum limit for the native heap growth. Please use -XX:HeapBaseMinAddress");
463 st->print_cr("# to set the Java Heap base and to place the Java Heap above 4GB virtual address.");
464 break;
465 case CompressedOops::ZeroBasedNarrowOop:
466 st->print_cr("# JVM is running with Zero Based Compressed Oops mode in which the Java heap is");
467 st->print_cr("# placed in the first 32GB address space. The Java Heap base address is the");
468 st->print_cr("# maximum limit for the native heap growth. Please use -XX:HeapBaseMinAddress");
469 st->print_cr("# to set the Java Heap base and to place the Java Heap above 32GB virtual address.");
470 break;
471 default:
472 break;
473 }
474 }
475 st->print_cr("# This output file may be truncated or incomplete.");
476 }
477
478 static void print_stack_location(outputStream* st, const void* context, int& continuation) {
479 const int number_of_stack_slots = 8;
480
481 int i = continuation;
482 // Update continuation with next index before fetching frame
483 continuation = i + 1;
484 const frame fr = os::fetch_frame_from_context(context);
485 while (i < number_of_stack_slots) {
486 // Update continuation with next index before printing location
487 continuation = i + 1;
488 // decode stack contents if possible
489 const intptr_t *sp = fr.sp();
490 const intptr_t *slot = sp + i;
491 if (!is_aligned(slot, sizeof(intptr_t))) {
492 st->print_cr("Misaligned sp: " PTR_FORMAT, p2i(sp));
493 break;
494 } else if (os::is_readable_pointer(slot)) {
495 st->print("stack at sp + %d slots: ", i);
496 os::print_location(st, *(slot));
497 } else {
498 st->print_cr("unreadable stack slot at sp + %d", i);
499 }
500 ++i;
501 }
502 }
503
504 static void report_vm_version(outputStream* st, char* buf, int buflen) {
505 // VM version
506 st->print_cr("#");
507 JDK_Version::current().to_string(buf, buflen);
508 const char* runtime_name = JDK_Version::runtime_name() != nullptr ?
509 JDK_Version::runtime_name() : "";
510 const char* runtime_version = JDK_Version::runtime_version() != nullptr ?
511 JDK_Version::runtime_version() : "";
512 const char* vendor_version = JDK_Version::runtime_vendor_version() != nullptr ?
513 JDK_Version::runtime_vendor_version() : "";
514 const char* jdk_debug_level = VM_Version::printable_jdk_debug_level() != nullptr ?
515 VM_Version::printable_jdk_debug_level() : "";
516
517 st->print_cr("# JRE version: %s%s%s (%s) (%sbuild %s)", runtime_name,
518 (*vendor_version != '\0') ? " " : "", vendor_version,
519 buf, jdk_debug_level, runtime_version);
520
521 // This is the long version with some default settings added
522 st->print_cr("# Java VM: %s%s%s (%s%s, %s%s%s%s%s%s, %s, %s)",
523 VM_Version::vm_name(),
524 (*vendor_version != '\0') ? " " : "", vendor_version,
525 jdk_debug_level,
526 VM_Version::vm_release(),
527 VM_Version::vm_info_string(),
528 TieredCompilation ? ", tiered" : "",
529 #if INCLUDE_JVMCI
530 EnableJVMCI ? ", jvmci" : "",
531 UseJVMCICompiler ? ", jvmci compiler" : "",
532 #else
533 "", "",
534 #endif
535 UseCompressedOops ? ", compressed oops" : "",
536 UseCompactObjectHeaders ? ", compact obj headers"
537 : (UseCompressedClassPointers ? ", compressed class ptrs" : ""),
538 GCConfig::hs_err_name(),
539 VM_Version::vm_platform_string()
540 );
541 }
542
543 // Returns true if at least one thread reported a fatal error and fatal error handling is in process.
544 bool VMError::is_error_reported() {
545 return _first_error_tid.load_relaxed() != -1;
546 }
547
548 // Returns true if the current thread reported a fatal error.
549 bool VMError::is_error_reported_in_current_thread() {
550 return _first_error_tid.load_relaxed() == os::current_thread_id();
551 }
552
553 // Helper, return current timestamp for timeout handling.
554 jlong VMError::get_current_timestamp() {
555 return os::javaTimeNanos();
556 }
557 // Factor to translate the timestamp to seconds.
558 #define TIMESTAMP_TO_SECONDS_FACTOR (1000 * 1000 * 1000)
559
560 void VMError::record_reporting_start_time() {
561 const jlong now = get_current_timestamp();
562 _reporting_start_time.store_relaxed(now);
563 }
564
565 jlong VMError::get_reporting_start_time() {
566 return _reporting_start_time.load_relaxed();
567 }
568
569 void VMError::record_step_start_time() {
570 const jlong now = get_current_timestamp();
571 _step_start_time.store_relaxed(now);
572 }
573
574 jlong VMError::get_step_start_time() {
575 return _step_start_time.load_relaxed();
576 }
577
578 void VMError::clear_step_start_time() {
579 return _step_start_time.store_relaxed(0);
580 }
581
582 // This is the main function to report a fatal error. Only one thread can
583 // call this function, so we don't need to worry about MT-safety. But it's
584 // possible that the error handler itself may crash or die on an internal
585 // error, for example, when the stack/heap is badly damaged. We must be
586 // able to handle recursive errors that happen inside error handler.
587 //
588 // Error reporting is done in several steps. If a crash or internal error
589 // occurred when reporting an error, the nested signal/exception handler
590 // can skip steps that are already (or partially) done. Error reporting will
591 // continue from the next step. This allows us to retrieve and print
592 // information that may be unsafe to get after a fatal error. If it happens,
593 // you may find nested report_and_die() frames when you look at the stack
594 // in a debugger.
595 //
596 // In general, a hang in error handler is much worse than a crash or internal
597 // error, as it's harder to recover from a hang. Deadlock can happen if we
598 // try to grab a lock that is already owned by current thread, or if the
599 // owner is blocked forever (e.g. in os::infinite_sleep()). If possible, the
600 // error handler and all the functions it called should avoid grabbing any
601 // lock. An important thing to notice is that memory allocation needs a lock.
602 //
603 // We should avoid using large stack allocated buffers. Many errors happen
604 // when stack space is already low. Making things even worse is that there
605 // could be nested report_and_die() calls on stack (see above). Only one
606 // thread can report error, so large buffers are statically allocated in data
607 // segment.
608 void VMError::report(outputStream* st, bool _verbose) {
609 // Used by reattempt step logic
610 static int continuation = 0;
611 const char* stop_reattempt_reason = nullptr;
612 # define BEGIN \
613 if (_current_step == 0) { \
614 _step_did_succeed.store_relaxed(false); \
615 _current_step = __LINE__; \
616 {
617 // [Begin logic]
618
619 # define STEP_IF(s, cond) \
620 } \
621 _step_did_succeed.store_relaxed(true); \
622 } \
623 if (_current_step < __LINE__) { \
624 _step_did_succeed.store_relaxed(false); \
625 _current_step = __LINE__; \
626 _current_step_info = s; \
627 if ((cond)) { \
628 record_step_start_time(); \
629 _step_did_timeout.store_relaxed(false);
630 // [Step logic]
631
632 # define STEP(s) STEP_IF(s, true)
633
634 # define REATTEMPT_STEP_IF(s, cond) \
635 } \
636 _step_did_succeed.store_relaxed(true); \
637 } \
638 if (_current_step < __LINE__ && !_step_did_succeed.load_relaxed()) { \
639 _current_step = __LINE__; \
640 _current_step_info = s; \
641 const bool cond_value = (cond); \
642 if (cond_value && !can_reattempt_step( \
643 stop_reattempt_reason)) { \
644 st->print_cr("[stop reattempt (%s) reason: %s]", \
645 _current_step_info, \
646 stop_reattempt_reason); \
647 } else if (cond_value) {
648 // [Continue Step logic]
649
650 # define END \
651 } \
652 _step_did_succeed.store_relaxed(true); \
653 clear_step_start_time(); \
654 }
655
656 // don't allocate large buffer on stack
657 static char buf[O_BUFLEN];
658
659 // Native stack trace may get stuck. We try to handle the last pc if it
660 // belongs to VM generated code.
661 address lastpc = nullptr;
662
663 BEGIN
664 if (MemTracker::enabled() &&
665 NmtVirtualMemory_lock != nullptr &&
666 _thread != 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 address sp = (address)fr.sp();
969 if (sp != nullptr) {
970 st->print(", sp=" PTR_FORMAT, p2i(sp));
971 if (sp >= stack_bottom && sp < stack_top) {
972 size_t free_stack_size = pointer_delta(sp, stack_bottom, 1024);
973 st->print(", free space=%zuk", free_stack_size);
974 } else {
975 st->print(" **OUTSIDE STACK**.");
976 }
977 }
978
979 st->cr();
980
981 STEP_IF("printing native stack (with source info)", _verbose)
982
983 NativeStackPrinter nsp(_thread, _context, _filename != nullptr ? get_filename_only() : nullptr, _lineno);
984 if (nsp.print_stack(st, buf, sizeof(buf), lastpc,
985 true /*print_source_info */, -1 /* max stack */)) {
986 // We have printed the native stack in platform-specific code
987 // Windows/x64 needs special handling.
988 // Stack walking may get stuck. Try to find the calling code.
989 if (lastpc != nullptr) {
990 const char* name = find_code_name(lastpc);
991 if (name != nullptr) {
992 st->print_cr("The last pc belongs to %s (printed below).", name);
993 }
994 }
995 } else {
996 _print_stack_from_frame_used = true; // frame-based native stack walk done
997 }
998
999 REATTEMPT_STEP_IF("retry printing native stack (no source info)", _verbose)
1000 st->cr();
1001 st->print_cr("Retrying call stack printing without source information...");
1002 NativeStackPrinter nsp(_thread, _context, get_filename_only(), _lineno);
1003 nsp.print_stack_from_frame(st, buf, sizeof(buf),
1004 false /*print_source_info */, -1 /* max stack */);
1005 _print_stack_from_frame_used = true;
1006
1007 STEP_IF("printing Java stack", _verbose && _thread && _thread->is_Java_thread())
1008 if (_verbose && _thread && _thread->is_Java_thread()) {
1009 print_stack_trace(st, JavaThread::cast(_thread), buf, sizeof(buf));
1010 }
1011
1012 STEP_IF("printing target Java thread stack",
1013 _verbose && _thread != nullptr && (_thread->is_Named_thread()))
1014 // printing Java thread stack trace if it is involved in GC crash
1015 Thread* thread = ((NamedThread *)_thread)->processed_thread();
1016 if (thread != nullptr && thread->is_Java_thread()) {
1017 JavaThread* jt = JavaThread::cast(thread);
1018 st->print_cr("JavaThread " PTR_FORMAT " (nid = %d) was being processed", p2i(jt), jt->osthread()->thread_id());
1019 print_stack_trace(st, jt, buf, sizeof(buf), true);
1020 }
1021
1022 STEP_IF("printing siginfo", _verbose && _siginfo != nullptr)
1023 // signal no, signal code, address that caused the fault
1024 st->cr();
1025 os::print_siginfo(st, _siginfo);
1026 st->cr();
1027
1028 STEP_IF("CDS archive access warning", _verbose && _siginfo != nullptr)
1029 // Print an explicit hint if we crashed on access to the CDS archive.
1030 check_failing_cds_access(st, _siginfo);
1031 st->cr();
1032
1033 #if defined(COMPILER1) || defined(COMPILER2)
1034 STEP_IF("printing pending compilation failure",
1035 _verbose && _thread != nullptr && _thread->is_Compiler_thread())
1036 CompilationFailureInfo::print_pending_compilation_failure(st);
1037 if (CompilationMemoryStatistic::enabled() && CompilationMemoryStatistic::in_oom_crash()) {
1038 st->cr();
1039 st->print_cr(">> Please see below for a detailed breakdown of compiler memory usage.");
1040 st->cr();
1041 }
1042 #endif
1043
1044 STEP_IF("printing registers", _verbose && _context != nullptr)
1045 // printing registers
1046 os::print_context(st, _context);
1047 st->cr();
1048
1049 STEP_IF("printing register info",
1050 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1051 continuation = 0;
1052 ResourceMark rm(_thread);
1053 st->print_cr("Register to memory mapping:");
1054 st->cr();
1055 os::print_register_info(st, _context, continuation);
1056 st->cr();
1057
1058 REATTEMPT_STEP_IF("printing register info, attempt 2",
1059 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1060 ResourceMark rm(_thread);
1061 os::print_register_info(st, _context, continuation);
1062 st->cr();
1063
1064 REATTEMPT_STEP_IF("printing register info, attempt 3",
1065 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1066 ResourceMark rm(_thread);
1067 os::print_register_info(st, _context, continuation);
1068 st->cr();
1069
1070 STEP_IF("printing top of stack, instructions near pc", _verbose && _context != nullptr)
1071 // printing top of stack, instructions near pc
1072 os::print_tos_pc(st, _context);
1073 st->cr();
1074
1075 STEP_IF("inspecting top of stack",
1076 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1077 continuation = 0;
1078 ResourceMark rm(_thread);
1079 st->print_cr("Stack slot to memory mapping:");
1080 st->cr();
1081 print_stack_location(st, _context, continuation);
1082 st->cr();
1083
1084 REATTEMPT_STEP_IF("inspecting top of stack, attempt 2",
1085 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1086 ResourceMark rm(_thread);
1087 print_stack_location(st, _context, continuation);
1088 st->cr();
1089
1090 REATTEMPT_STEP_IF("inspecting top of stack, attempt 3",
1091 _verbose && _context != nullptr && _thread != nullptr && Universe::is_fully_initialized())
1092 ResourceMark rm(_thread);
1093 print_stack_location(st, _context, continuation);
1094 st->cr();
1095
1096 STEP_IF("printing lock stack", _verbose && _thread != nullptr && _thread->is_Java_thread());
1097 st->print_cr("Lock stack of current Java thread (top to bottom):");
1098 JavaThread::cast(_thread)->lock_stack().print_on(st);
1099 st->cr();
1100
1101 STEP_IF("printing code blobs if possible", _verbose)
1102 const int printed_capacity = max_error_log_print_code;
1103 address printed[printed_capacity];
1104 printed[0] = nullptr;
1105 int printed_len = 0;
1106 // Even though ErrorLogPrintCodeLimit is ranged checked
1107 // during argument parsing, there's no way to prevent it
1108 // subsequently (i.e., after parsing) being set to a
1109 // value outside the range.
1110 int limit = MIN2(ErrorLogPrintCodeLimit, printed_capacity);
1111 if (limit > 0) {
1112 // Check if a pc was found by native stack trace above.
1113 if (lastpc != nullptr) {
1114 if (print_code(st, _thread, lastpc, true, printed, printed_capacity)) {
1115 printed_len++;
1116 }
1117 }
1118
1119 // Scan the native stack
1120 if (!_print_stack_from_frame_used) {
1121 // Only try to print code of the crashing frame since
1122 // the native stack cannot be walked with next_frame.
1123 if (print_code(st, _thread, _pc, true, printed, printed_capacity)) {
1124 printed_len++;
1125 }
1126 } else {
1127 frame fr = _context ? os::fetch_frame_from_context(_context)
1128 : os::current_frame();
1129 while (printed_len < limit && fr.pc() != nullptr) {
1130 if (print_code(st, _thread, fr.pc(), fr.pc() == _pc, printed, printed_capacity)) {
1131 printed_len++;
1132 }
1133 fr = frame::next_frame(fr, _thread);
1134 }
1135 }
1136
1137 // Scan the Java stack
1138 if (_thread != nullptr && _thread->is_Java_thread()) {
1139 JavaThread* jt = JavaThread::cast(_thread);
1140 if (jt->has_last_Java_frame()) {
1141 for (StackFrameStream sfs(jt, true /* update */, true /* process_frames */); printed_len < limit && !sfs.is_done(); sfs.next()) {
1142 address pc = sfs.current()->pc();
1143 if (print_code(st, _thread, pc, pc == _pc, printed, printed_capacity)) {
1144 printed_len++;
1145 }
1146 }
1147 }
1148 }
1149 }
1150
1151 STEP_IF("printing VM operation", _verbose && _thread != nullptr && _thread->is_VM_thread())
1152 VMThread* t = (VMThread*)_thread;
1153 VM_Operation* op = t->vm_operation();
1154 if (op) {
1155 op->print_on_error(st);
1156 st->cr();
1157 st->cr();
1158 }
1159
1160 STEP_IF("printing registered callbacks", _verbose && _thread != nullptr);
1161 size_t count = 0;
1162 for (VMErrorCallback* callback = _thread->_vm_error_callbacks;
1163 callback != nullptr;
1164 callback = callback->_next) {
1165 st->print_cr("VMErrorCallback %zu:", ++count);
1166 callback->call(st);
1167 st->cr();
1168 }
1169
1170 STEP_IF("printing process", _verbose)
1171 st->cr();
1172 st->print_cr("--------------- P R O C E S S ---------------");
1173 st->cr();
1174
1175 STEP_IF("printing user info", ExtensiveErrorReports && _verbose)
1176 os::print_user_info(st);
1177
1178 STEP_IF("printing all threads", _verbose && _thread != nullptr)
1179 // all threads
1180 Threads::print_on_error(st, _thread, buf, sizeof(buf));
1181 st->cr();
1182
1183 STEP_IF("printing VM state", _verbose)
1184 // Safepoint state
1185 st->print("VM state: ");
1186
1187 if (SafepointSynchronize::is_synchronizing()) st->print("synchronizing");
1188 else if (SafepointSynchronize::is_at_safepoint()) st->print("at safepoint");
1189 else st->print("not at safepoint");
1190
1191 // Also see if error occurred during initialization or shutdown
1192 if (!Universe::is_fully_initialized()) {
1193 st->print(" (not fully initialized)");
1194 } else if (VM_Exit::vm_exited()) {
1195 st->print(" (shutting down)");
1196 } else {
1197 st->print(" (normal execution)");
1198 }
1199 st->cr();
1200 st->cr();
1201
1202 STEP_IF("printing owned locks on error", _verbose)
1203 // mutexes/monitors that currently have an owner
1204 Mutex::print_owned_locks_on_error(st);
1205 st->cr();
1206
1207 STEP_IF("printing number of OutOfMemoryError and StackOverflow exceptions",
1208 _verbose && Exceptions::has_exception_counts())
1209 st->print_cr("OutOfMemory and StackOverflow Exception counts:");
1210 Exceptions::print_exception_counts_on_error(st);
1211 st->cr();
1212
1213 #ifdef _LP64
1214 STEP_IF("printing compressed oops mode", _verbose && UseCompressedOops)
1215 CompressedOops::print_mode(st);
1216 st->cr();
1217
1218 STEP_IF("printing compressed klass pointers mode", _verbose && UseCompressedClassPointers)
1219 CDS_ONLY(AOTMetaspace::print_on(st);)
1220 Metaspace::print_compressed_class_space(st);
1221 CompressedKlassPointers::print_mode(st);
1222 st->cr();
1223 #endif
1224
1225 STEP_IF("printing heap information", _verbose)
1226 GCLogPrecious::print_on_error(st);
1227
1228 if (Universe::heap() != nullptr) {
1229 st->print_cr("Heap:");
1230 StreamIndentor si(st, 1);
1231 Universe::heap()->print_heap_on(st);
1232 st->cr();
1233 }
1234
1235 STEP_IF("printing GC information", _verbose)
1236 if (Universe::heap() != nullptr) {
1237 Universe::heap()->print_gc_on(st);
1238 st->cr();
1239 }
1240
1241 if (Universe::is_fully_initialized()) {
1242 st->print_cr("Polling page: " PTR_FORMAT, p2i(SafepointMechanism::get_polling_page()));
1243 st->cr();
1244 }
1245
1246 STEP_IF("printing metaspace information", _verbose && Universe::is_fully_initialized())
1247 st->print_cr("Metaspace:");
1248 MetaspaceUtils::print_on(st);
1249 MetaspaceUtils::print_basic_report(st, 0);
1250
1251 STEP_IF("printing code cache information", _verbose && Universe::is_fully_initialized())
1252 // print code cache information before vm abort
1253 CodeCache::print_summary(st);
1254 st->cr();
1255
1256 STEP_IF("printing ring buffers", _verbose)
1257 Events::print_all(st);
1258 st->cr();
1259
1260 STEP_IF("printing dynamic libraries", _verbose)
1261 // dynamic libraries, or memory map
1262 os::print_dll_info(st);
1263 st->cr();
1264
1265 #if INCLUDE_JVMTI
1266 STEP_IF("printing jvmti agent info", _verbose)
1267 os::print_jvmti_agent_info(st);
1268 st->cr();
1269 #endif
1270
1271 STEP_IF("printing native decoder state", _verbose)
1272 Decoder::print_state_on(st);
1273 st->cr();
1274
1275 STEP_IF("printing VM options", _verbose)
1276 // VM options
1277 Arguments::print_on(st);
1278 st->cr();
1279
1280 STEP_IF("printing flags", _verbose)
1281 JVMFlag::printFlags(
1282 st,
1283 true, // with comments
1284 false, // no ranges
1285 true); // skip defaults
1286 st->cr();
1287
1288 STEP_IF("printing warning if internal testing API used", WhiteBox::used())
1289 st->print_cr("Unsupported internal testing APIs have been used.");
1290 st->cr();
1291
1292 STEP_IF("printing log configuration", _verbose)
1293 st->print_cr("Logging:");
1294 LogConfiguration::describe_current_configuration(st);
1295 st->cr();
1296
1297 STEP_IF("printing release file content", _verbose)
1298 st->print_cr("Release file:");
1299 os::print_image_release_file(st);
1300
1301 STEP_IF("printing all environment variables", _verbose)
1302 os::print_environment_variables(st, env_list);
1303 st->cr();
1304
1305 STEP_IF("printing locale settings", _verbose)
1306 os::print_active_locale(st);
1307 st->cr();
1308
1309 STEP_IF("printing signal handlers", _verbose)
1310 os::print_signal_handlers(st, buf, sizeof(buf));
1311 st->cr();
1312
1313 STEP_IF("Native Memory Tracking", _verbose && _thread != nullptr)
1314 MemTracker::error_report(st);
1315 st->cr();
1316
1317 STEP_IF("printing compiler memory info, if any", _verbose)
1318 CompilationMemoryStatistic::print_error_report(st);
1319 st->cr();
1320
1321 STEP_IF("printing periodic trim state", _verbose)
1322 NativeHeapTrimmer::print_state(st);
1323 st->cr();
1324
1325 STEP_IF("printing system", _verbose)
1326 st->print_cr("--------------- S Y S T E M ---------------");
1327 st->cr();
1328
1329 STEP_IF("printing OS information", _verbose)
1330 os::print_os_info(st);
1331 st->cr();
1332 #ifdef __APPLE__
1333 // Avoid large stack allocation on Mac for FD count during signal-handling.
1334 os::Bsd::print_open_file_descriptors(st, buf, sizeof(buf));
1335 st->cr();
1336 #else
1337 os::print_open_file_descriptors(st);
1338 #endif
1339
1340 STEP_IF("printing CPU info", _verbose)
1341 os::print_cpu_info(st, buf, sizeof(buf));
1342 st->cr();
1343
1344 STEP_IF("printing memory info", _verbose)
1345 os::print_memory_info(st);
1346 st->cr();
1347
1348 STEP_IF("printing internal vm info", _verbose)
1349 st->print_cr("vm_info: %s", VM_Version::internal_vm_info_string());
1350 st->cr();
1351
1352 // print a defined marker to show that error handling finished correctly.
1353 STEP_IF("printing end marker", _verbose)
1354 st->print_cr("END.");
1355
1356 END
1357
1358 # undef BEGIN
1359 # undef STEP_IF
1360 # undef STEP
1361 # undef REATTEMPT_STEP_IF
1362 # undef END
1363 }
1364
1365 void VMError::set_handshake_timed_out_thread(Thread* thread) {
1366 // Only preserve the first thread to time-out this way. The atomic operation ensures
1367 // visibility to the target thread.
1368 _handshake_timed_out_thread.compare_exchange(nullptr, thread);
1369 }
1370
1371 void VMError::set_safepoint_timed_out_thread(Thread* thread) {
1372 // Only preserve the first thread to time-out this way. The atomic operation ensures
1373 // visibility to the target thread.
1374 _safepoint_timed_out_thread.compare_exchange(nullptr, thread);
1375 }
1376
1377 Thread* VMError::get_handshake_timed_out_thread() {
1378 return _handshake_timed_out_thread.load_relaxed();
1379 }
1380
1381 Thread* VMError::get_safepoint_timed_out_thread() {
1382 return _safepoint_timed_out_thread.load_relaxed();
1383 }
1384
1385 // Report for the vm_info_cmd. This prints out the information above omitting
1386 // crash and thread specific information. If output is added above, it should be added
1387 // here also, if it is safe to call during a running process.
1388 void VMError::print_vm_info(outputStream* st) {
1389
1390 char buf[O_BUFLEN];
1391 os::prepare_native_symbols();
1392
1393 report_vm_version(st, buf, sizeof(buf));
1394
1395 // STEP("printing summary")
1396
1397 st->cr();
1398 st->print_cr("--------------- S U M M A R Y ------------");
1399 st->cr();
1400
1401 // STEP("printing VM option summary")
1402
1403 // VM options
1404 Arguments::print_summary_on(st);
1405 st->cr();
1406
1407 // STEP("printing summary machine and OS info")
1408
1409 os::print_summary_info(st, buf, sizeof(buf));
1410
1411 // STEP("printing date and time")
1412
1413 os::print_date_and_time(st, buf, sizeof(buf));
1414
1415 // Skip: STEP("printing thread")
1416
1417 // STEP("printing process")
1418
1419 st->cr();
1420 st->print_cr("--------------- P R O C E S S ---------------");
1421 st->cr();
1422
1423 // STEP("printing number of OutOfMemoryError and StackOverflow exceptions")
1424
1425 if (Exceptions::has_exception_counts()) {
1426 st->print_cr("OutOfMemory and StackOverflow Exception counts:");
1427 Exceptions::print_exception_counts_on_error(st);
1428 st->cr();
1429 }
1430
1431 #ifdef _LP64
1432 // STEP("printing compressed oops mode")
1433 if (UseCompressedOops) {
1434 CompressedOops::print_mode(st);
1435 st->cr();
1436 }
1437 #endif
1438
1439 // STEP("printing compressed class ptrs mode")
1440 if (UseCompressedClassPointers) {
1441 CDS_ONLY(AOTMetaspace::print_on(st);)
1442 Metaspace::print_compressed_class_space(st);
1443 CompressedKlassPointers::print_mode(st);
1444 st->cr();
1445 }
1446
1447 // Take heap lock over heap, GC and metaspace printing so that information
1448 // is consistent.
1449 if (Universe::is_fully_initialized()) {
1450 MutexLocker ml(Heap_lock);
1451
1452 // STEP("printing heap information")
1453
1454 GCLogPrecious::print_on_error(st);
1455
1456 {
1457 st->print_cr("Heap:");
1458 StreamIndentor si(st, 1);
1459 Universe::heap()->print_heap_on(st);
1460 st->cr();
1461 }
1462
1463 // STEP("printing GC information")
1464
1465 Universe::heap()->print_gc_on(st);
1466 st->cr();
1467
1468 st->print_cr("Polling page: " PTR_FORMAT, p2i(SafepointMechanism::get_polling_page()));
1469 st->cr();
1470
1471 // STEP("printing metaspace information")
1472
1473 st->print_cr("Metaspace:");
1474 MetaspaceUtils::print_on(st);
1475 MetaspaceUtils::print_basic_report(st, 0);
1476 }
1477
1478 // STEP("printing code cache information")
1479
1480 if (Universe::is_fully_initialized()) {
1481 // print code cache information before vm abort
1482 CodeCache::print_summary(st);
1483 st->cr();
1484 }
1485
1486 // STEP("printing ring buffers")
1487
1488 Events::print_all(st);
1489 st->cr();
1490
1491 // STEP("printing dynamic libraries")
1492
1493 // dynamic libraries, or memory map
1494 os::print_dll_info(st);
1495 st->cr();
1496
1497 #if INCLUDE_JVMTI
1498 os::print_jvmti_agent_info(st);
1499 st->cr();
1500 #endif
1501
1502 // STEP("printing VM options")
1503
1504 // VM options
1505 Arguments::print_on(st);
1506 st->cr();
1507
1508 // STEP("printing warning if internal testing API used")
1509
1510 if (WhiteBox::used()) {
1511 st->print_cr("Unsupported internal testing APIs have been used.");
1512 st->cr();
1513 }
1514
1515 // STEP("printing log configuration")
1516 st->print_cr("Logging:");
1517 LogConfiguration::describe(st);
1518 st->cr();
1519
1520 // STEP("printing release file content")
1521 st->print_cr("Release file:");
1522 os::print_image_release_file(st);
1523
1524 // STEP("printing all environment variables")
1525
1526 os::print_environment_variables(st, env_list);
1527 st->cr();
1528
1529 // STEP("printing locale settings")
1530
1531 os::print_active_locale(st);
1532 st->cr();
1533
1534
1535 // STEP("printing signal handlers")
1536
1537 os::print_signal_handlers(st, buf, sizeof(buf));
1538 st->cr();
1539
1540 // STEP("Native Memory Tracking")
1541 MemTracker::error_report(st);
1542 st->cr();
1543
1544 // STEP("Compiler Memory Statistic")
1545 CompilationMemoryStatistic::print_final_report(st);
1546
1547 // STEP("printing periodic trim state")
1548 NativeHeapTrimmer::print_state(st);
1549 st->cr();
1550
1551
1552 // STEP("printing system")
1553 st->print_cr("--------------- S Y S T E M ---------------");
1554 st->cr();
1555
1556 // STEP("printing OS information")
1557
1558 os::print_os_info(st);
1559 st->cr();
1560 os::print_open_file_descriptors(st);
1561 st->cr();
1562
1563 // STEP("printing CPU info")
1564
1565 os::print_cpu_info(st, buf, sizeof(buf));
1566 st->cr();
1567
1568 // STEP("printing memory info")
1569
1570 os::print_memory_info(st);
1571 st->cr();
1572
1573 // STEP("printing internal vm info")
1574
1575 st->print_cr("vm_info: %s", VM_Version::internal_vm_info_string());
1576 st->cr();
1577
1578 // print a defined marker to show that error handling finished correctly.
1579 // STEP("printing end marker")
1580
1581 st->print_cr("END.");
1582 }
1583
1584 /** Expand a pattern into a buffer starting at pos and open a file using constructed path */
1585 static int expand_and_open(const char* pattern, bool overwrite_existing, char* buf, size_t buflen, size_t pos) {
1586 int fd = -1;
1587 int mode = O_RDWR | O_CREAT;
1588 if (overwrite_existing) {
1589 mode |= O_TRUNC;
1590 } else {
1591 mode |= O_EXCL;
1592 }
1593 if (Arguments::copy_expand_pid(pattern, strlen(pattern), &buf[pos], buflen - pos)) {
1594 fd = open(buf, mode, 0666);
1595 }
1596 return fd;
1597 }
1598
1599 /**
1600 * Construct file name for a log file and return it's file descriptor.
1601 * Name and location depends on pattern, default_pattern params and access
1602 * permissions.
1603 */
1604 int VMError::prepare_log_file(const char* pattern, const char* default_pattern, bool overwrite_existing, char* buf, size_t buflen) {
1605 int fd = -1;
1606
1607 // If possible, use specified pattern to construct log file name
1608 if (pattern != nullptr) {
1609 fd = expand_and_open(pattern, overwrite_existing, buf, buflen, 0);
1610 }
1611
1612 // Either user didn't specify, or the user's location failed,
1613 // so use the default name in the current directory
1614 if (fd == -1) {
1615 const char* cwd = os::get_current_directory(buf, buflen);
1616 if (cwd != nullptr) {
1617 size_t pos = strlen(cwd);
1618 int fsep_len = jio_snprintf(&buf[pos], buflen-pos, "%s", os::file_separator());
1619 pos += fsep_len;
1620 if (fsep_len > 0) {
1621 fd = expand_and_open(default_pattern, overwrite_existing, buf, buflen, pos);
1622 }
1623 }
1624 }
1625
1626 // try temp directory if it exists.
1627 if (fd == -1) {
1628 const char* tmpdir = os::get_temp_directory();
1629 if (tmpdir != nullptr && strlen(tmpdir) > 0) {
1630 int pos = jio_snprintf(buf, buflen, "%s%s", tmpdir, os::file_separator());
1631 if (pos > 0) {
1632 fd = expand_and_open(default_pattern, overwrite_existing, buf, buflen, pos);
1633 }
1634 }
1635 }
1636
1637 return fd;
1638 }
1639
1640 void VMError::report_and_die(Thread* thread, unsigned int sig, address pc, const void* siginfo,
1641 const void* context, const char* detail_fmt, ...)
1642 {
1643 va_list detail_args;
1644 va_start(detail_args, detail_fmt);
1645 report_and_die(sig, nullptr, detail_fmt, detail_args, thread, pc, siginfo, context, nullptr, 0, 0);
1646 va_end(detail_args);
1647 }
1648
1649 void VMError::report_and_die(Thread* thread, const void* context, const char* filename, int lineno, const char* message,
1650 const char* detail_fmt, ...) {
1651 va_list detail_args;
1652 va_start(detail_args, detail_fmt);
1653 report_and_die(thread, context, filename, lineno, message, detail_fmt, detail_args);
1654 va_end(detail_args);
1655 }
1656
1657 void VMError::report_and_die(Thread* thread, unsigned int sig, address pc, const void* siginfo, const void* context)
1658 {
1659 if (ExecutingUnitTests) {
1660 // See TEST_VM_CRASH_SIGNAL gtest macro
1661 char tmp[64];
1662 fprintf(stderr, "signaled: %s", os::exception_name(sig, tmp, sizeof(tmp)));
1663 }
1664
1665 report_and_die(thread, sig, pc, siginfo, context, "%s", "");
1666 }
1667
1668 void VMError::report_and_die(Thread* thread, const void* context, const char* filename, int lineno, const char* message,
1669 const char* detail_fmt, va_list detail_args)
1670 {
1671 report_and_die(INTERNAL_ERROR, message, detail_fmt, detail_args, thread, nullptr, nullptr, context, filename, lineno, 0);
1672 }
1673
1674 void VMError::report_and_die(Thread* thread, const char* filename, int lineno, size_t size,
1675 VMErrorType vm_err_type, const char* detail_fmt, va_list detail_args) {
1676 report_and_die(vm_err_type, nullptr, detail_fmt, detail_args, thread, nullptr, nullptr, nullptr, filename, lineno, size);
1677 }
1678
1679 void VMError::report_and_die(int id, const char* message, const char* detail_fmt, va_list detail_args,
1680 Thread* thread, address pc, const void* siginfo, const void* context, const char* filename,
1681 int lineno, size_t size)
1682 {
1683 // A single scratch buffer to be used from here on.
1684 // Do not rely on it being preserved across function calls.
1685 static char buffer[O_BUFLEN];
1686
1687 // File descriptor to tty to print an error summary to.
1688 // Hard wired to stdout; see JDK-8215004 (compatibility concerns).
1689 static const int fd_out = 1; // stdout
1690
1691 // File descriptor to the error log file.
1692 static int fd_log = -1;
1693
1694 #ifdef CAN_SHOW_REGISTERS_ON_ASSERT
1695 // Disarm assertion poison page, since from this point on we do not need this mechanism anymore and it may
1696 // cause problems in error handling during native OOM, see JDK-8227275.
1697 disarm_assert_poison();
1698 #endif
1699
1700 // Use local fdStream objects only. Do not use global instances whose initialization
1701 // relies on dynamic initialization (see JDK-8214975). Do not rely on these instances
1702 // to carry over into recursions or invocations from other threads.
1703 fdStream out(fd_out);
1704 out.set_scratch_buffer(buffer, sizeof(buffer));
1705
1706 // Depending on the re-entrance depth at this point, fd_log may be -1 or point to an open hs-err file.
1707 fdStream log(fd_log);
1708 log.set_scratch_buffer(buffer, sizeof(buffer));
1709
1710 // How many errors occurred in error handler when reporting first_error.
1711 static int recursive_error_count;
1712
1713 // We will first print a brief message to standard out (verbose = false),
1714 // then save detailed information in log file (verbose = true).
1715 static bool out_done = false; // done printing to standard out
1716 static bool log_done = false; // done saving error log
1717
1718 intptr_t mytid = os::current_thread_id();
1719 if (_first_error_tid.compare_set(-1, mytid)) {
1720
1721 if (SuppressFatalErrorMessage) {
1722 os::abort(CreateCoredumpOnCrash);
1723 }
1724
1725 // Initialize time stamps to use the same base.
1726 out.time_stamp().update_to(1);
1727 log.time_stamp().update_to(1);
1728
1729 _id = id;
1730 _message = message;
1731 _thread = thread;
1732 _pc = pc;
1733 _siginfo = siginfo;
1734 _context = context;
1735 _filename = filename;
1736 _lineno = lineno;
1737 _size = size;
1738 jio_vsnprintf(_detail_msg, sizeof(_detail_msg), detail_fmt, detail_args);
1739
1740 reporting_started();
1741 if (!TestUnresponsiveErrorHandler) {
1742 // Record reporting_start_time unless we're running the
1743 // TestUnresponsiveErrorHandler test. For that test we record
1744 // reporting_start_time at the beginning of the test.
1745 record_reporting_start_time();
1746 } else {
1747 out.print_raw_cr("Delaying recording reporting_start_time for TestUnresponsiveErrorHandler.");
1748 }
1749
1750 if (ShowMessageBoxOnError || PauseAtExit) {
1751 show_message_box(buffer, sizeof(buffer));
1752
1753 // User has asked JVM to abort. Reset ShowMessageBoxOnError so the
1754 // WatcherThread can kill JVM if the error handler hangs.
1755 ShowMessageBoxOnError = false;
1756 }
1757
1758 os::check_core_dump_prerequisites(buffer, sizeof(buffer));
1759
1760 // reset signal handlers or exception filter; make sure recursive crashes
1761 // are handled properly.
1762 install_secondary_signal_handler();
1763 } else {
1764 // This is not the first error, see if it happened in a different thread
1765 // or in the same thread during error reporting.
1766 if (_first_error_tid.load_relaxed() != mytid) {
1767 if (!SuppressFatalErrorMessage) {
1768 char msgbuf[64];
1769 jio_snprintf(msgbuf, sizeof(msgbuf),
1770 "[thread %zd also had an error]",
1771 mytid);
1772 out.print_raw_cr(msgbuf);
1773 }
1774
1775 // Error reporting is not MT-safe, nor can we let the current thread
1776 // proceed, so we block it.
1777 os::infinite_sleep();
1778
1779 } else {
1780 if (recursive_error_count++ > 30) {
1781 if (!SuppressFatalErrorMessage) {
1782 out.print_raw_cr("[Too many errors, abort]");
1783 }
1784 os::die();
1785 }
1786
1787 if (SuppressFatalErrorMessage) {
1788 // If we already hit a secondary error during abort, then calling
1789 // it again is likely to hit another one. But eventually, if we
1790 // don't deadlock somewhere, we will call os::die() above.
1791 os::abort(CreateCoredumpOnCrash);
1792 }
1793
1794 outputStream* const st = log.is_open() ? &log : &out;
1795 st->cr();
1796
1797 // Timeout handling.
1798 if (_step_did_timeout.load_relaxed()) {
1799 // The current step had a timeout. Lets continue reporting with the next step.
1800 st->print_raw("[timeout occurred during error reporting in step \"");
1801 st->print_raw(_current_step_info);
1802 st->print_cr("\"] after " INT64_FORMAT " s.",
1803 (int64_t)
1804 ((get_current_timestamp() - get_step_start_time()) / TIMESTAMP_TO_SECONDS_FACTOR));
1805 } else if (_reporting_did_timeout.load_relaxed()) {
1806 // We hit ErrorLogTimeout. Reporting will stop altogether. Let's wrap things
1807 // up, the process is about to be stopped by the WatcherThread.
1808 st->print_cr("------ Timeout during error reporting after " INT64_FORMAT " s. ------",
1809 (int64_t)
1810 ((get_current_timestamp() - get_reporting_start_time()) / TIMESTAMP_TO_SECONDS_FACTOR));
1811 st->flush();
1812 // Watcherthread is about to call os::die. Lets just wait.
1813 os::infinite_sleep();
1814 } else {
1815 // A secondary error happened. Print brief information, but take care, since crashing
1816 // here would just recurse endlessly.
1817 // Any information (signal, context, siginfo etc) printed here should use the function
1818 // arguments, not the information stored in *this, since those describe the primary crash.
1819 static char tmp[256]; // cannot use global scratch buffer
1820 // Note: this string does get parsed by a number of jtreg tests,
1821 // see hotspot/jtreg/runtime/ErrorHandling.
1822 st->print("[error occurred during error reporting (%s), id 0x%x",
1823 _current_step_info, id);
1824 if (os::exception_name(id, tmp, sizeof(tmp))) {
1825 st->print(", %s (0x%x) at pc=" PTR_FORMAT, tmp, id, p2i(pc));
1826 } else {
1827 if (should_report_bug(id)) {
1828 st->print(", Internal Error (%s:%d)",
1829 filename == nullptr ? "??" : filename, lineno);
1830 } else {
1831 st->print(", Out of Memory Error (%s:%d)",
1832 filename == nullptr ? "??" : filename, lineno);
1833 }
1834 }
1835 st->print_cr("]");
1836 if (ErrorLogSecondaryErrorDetails) {
1837 static bool recursed = false;
1838 if (!recursed) {
1839 recursed = true;
1840 // Print even more information for secondary errors. This may generate a lot of output
1841 // and possibly disturb error reporting, therefore its optional and only available in debug builds.
1842 if (siginfo != nullptr) {
1843 st->print("[");
1844 os::print_siginfo(st, siginfo);
1845 st->print_cr("]");
1846 }
1847 st->print("[stack: ");
1848 NativeStackPrinter nsp(_thread, context, _filename != nullptr ? get_filename_only() : nullptr, _lineno);
1849 // Subsequent secondary errors build up stack; to avoid flooding the hs-err file with irrelevant
1850 // call stacks, limit the stack we print here (we are only interested in what happened before the
1851 // last assert/fault).
1852 const int max_stack_size = 15;
1853 nsp.print_stack_from_frame(st, tmp, sizeof(tmp), true /* print_source_info */, max_stack_size);
1854 st->print_cr("]");
1855 } // !recursed
1856 recursed = false; // Note: reset outside !recursed
1857 }
1858 }
1859 }
1860 }
1861
1862 // Part 1: print an abbreviated version (the '#' section) to stdout.
1863 if (!out_done) {
1864 // Suppress this output if we plan to print Part 2 to stdout too.
1865 // No need to have the "#" section twice.
1866 if (!(ErrorFileToStdout && out.fd() == 1)) {
1867 report(&out, false);
1868 }
1869
1870 out_done = true;
1871
1872 _current_step = 0;
1873 _current_step_info = "";
1874 }
1875
1876 // Part 2: print a full error log file (optionally to stdout or stderr).
1877 // print to error log file
1878 if (!log_done) {
1879 // see if log file is already open
1880 if (!log.is_open()) {
1881 // open log file
1882 if (ErrorFileToStdout) {
1883 fd_log = 1;
1884 } else if (ErrorFileToStderr) {
1885 fd_log = 2;
1886 } else {
1887 fd_log = prepare_log_file(ErrorFile, "hs_err_pid%p.log", true,
1888 buffer, sizeof(buffer));
1889 if (fd_log != -1) {
1890 out.print_raw("# An error report file with more information is saved as:\n# ");
1891 out.print_raw_cr(buffer);
1892 } else {
1893 out.print_raw_cr("# Can not save log file, dump to screen..");
1894 fd_log = 1;
1895 }
1896 }
1897 log.set_fd(fd_log);
1898 }
1899
1900 report(&log, true);
1901 log_done = true;
1902 _current_step = 0;
1903 _current_step_info = "";
1904
1905 if (fd_log > 3) {
1906 ::close(fd_log);
1907 fd_log = -1;
1908 }
1909
1910 log.set_fd(-1);
1911 }
1912
1913 JFR_ONLY(Jfr::on_vm_shutdown(true, false, static_cast<VMErrorType>(_id) == OOM_JAVA_HEAP_FATAL);)
1914
1915 if (PrintNMTStatistics) {
1916 fdStream fds(fd_out);
1917 MemTracker::final_report(&fds);
1918 }
1919
1920 static bool skip_replay = ReplayCompiles && !ReplayReduce; // Do not overwrite file during replay
1921 if (DumpReplayDataOnError && _thread && _thread->is_Compiler_thread() && !skip_replay) {
1922 skip_replay = true;
1923 ciEnv* env = ciEnv::current();
1924 if (env != nullptr && env->task() != nullptr) {
1925 const bool overwrite = false; // We do not overwrite an existing replay file.
1926 int fd = prepare_log_file(ReplayDataFile, "replay_pid%p.log", overwrite, buffer, sizeof(buffer));
1927 if (fd != -1) {
1928 FILE* replay_data_file = os::fdopen(fd, "w");
1929 if (replay_data_file != nullptr) {
1930 fileStream replay_data_stream(replay_data_file, /*need_close=*/true);
1931 env->dump_replay_data_unsafe(&replay_data_stream);
1932 out.print_raw("#\n# Compiler replay data is saved as:\n# ");
1933 out.print_raw_cr(buffer);
1934 } else {
1935 int e = errno;
1936 out.print_raw("#\n# Can't open file to dump replay data. Error: ");
1937 out.print_raw_cr(os::strerror(e));
1938 close(fd);
1939 }
1940 }
1941 }
1942 }
1943
1944 #if INCLUDE_JVMCI
1945 if (JVMCI::fatal_log_filename() != nullptr) {
1946 out.print_raw("#\n# The JVMCI shared library error report file is saved as:\n# ");
1947 out.print_raw_cr(JVMCI::fatal_log_filename());
1948 }
1949 #endif
1950
1951 static bool skip_bug_url = !should_submit_bug_report(_id);
1952 if (!skip_bug_url) {
1953 skip_bug_url = true;
1954
1955 out.print_raw_cr("#");
1956 print_bug_submit_message(&out, _thread);
1957 }
1958
1959 static bool skip_OnError = false;
1960 if (!skip_OnError && OnError && OnError[0]) {
1961 skip_OnError = true;
1962
1963 // Flush output and finish logs before running OnError commands.
1964 ostream_abort();
1965
1966 out.print_raw_cr("#");
1967 out.print_raw ("# -XX:OnError=\"");
1968 out.print_raw (OnError);
1969 out.print_raw_cr("\"");
1970
1971 char* cmd;
1972 const char* ptr = OnError;
1973 while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr)) != nullptr){
1974 out.print_raw ("# Executing ");
1975 #if defined(LINUX) || defined(_ALLBSD_SOURCE)
1976 out.print_raw ("/bin/sh -c ");
1977 #elif defined(_WINDOWS)
1978 out.print_raw ("cmd /C ");
1979 #endif
1980 out.print_raw ("\"");
1981 out.print_raw (cmd);
1982 out.print_raw_cr("\" ...");
1983
1984 if (os::fork_and_exec(cmd) < 0) {
1985 out.print_cr("os::fork_and_exec failed: %s (%s=%d)",
1986 os::strerror(errno), os::errno_name(errno), errno);
1987 }
1988 }
1989
1990 // done with OnError
1991 OnError = nullptr;
1992 }
1993
1994 #if defined _WINDOWS
1995 if (UseOSErrorReporting) {
1996 raise_fail_fast(_siginfo, _context);
1997 }
1998 #endif // _WINDOWS
1999
2000 // os::abort() will call abort hooks, try it first.
2001 static bool skip_os_abort = false;
2002 if (!skip_os_abort) {
2003 skip_os_abort = true;
2004 bool dump_core = should_report_bug(_id);
2005 os::abort(dump_core && CreateCoredumpOnCrash, _siginfo, _context);
2006 // if os::abort() doesn't abort, try os::die();
2007 }
2008 os::die();
2009 }
2010
2011 /*
2012 * OnOutOfMemoryError scripts/commands executed while VM is a safepoint - this
2013 * ensures utilities such as jmap can observe the process is a consistent state.
2014 */
2015 class VM_ReportJavaOutOfMemory : public VM_Operation {
2016 private:
2017 const char* _message;
2018 public:
2019 VM_ReportJavaOutOfMemory(const char* message) { _message = message; }
2020 VMOp_Type type() const { return VMOp_ReportJavaOutOfMemory; }
2021 void doit();
2022 };
2023
2024 void VM_ReportJavaOutOfMemory::doit() {
2025 // Don't allocate large buffer on stack
2026 static char buffer[O_BUFLEN];
2027
2028 tty->print_cr("#");
2029 tty->print_cr("# java.lang.OutOfMemoryError: %s", _message);
2030 tty->print_cr("# -XX:OnOutOfMemoryError=\"%s\"", OnOutOfMemoryError);
2031
2032 // make heap parsability
2033 Universe::heap()->ensure_parsability(false); // no need to retire TLABs
2034
2035 char* cmd;
2036 const char* ptr = OnOutOfMemoryError;
2037 while ((cmd = next_OnError_command(buffer, sizeof(buffer), &ptr)) != nullptr){
2038 tty->print("# Executing ");
2039 #if defined(LINUX)
2040 tty->print ("/bin/sh -c ");
2041 #endif
2042 tty->print_cr("\"%s\"...", cmd);
2043
2044 if (os::fork_and_exec(cmd) < 0) {
2045 tty->print_cr("os::fork_and_exec failed: %s (%s=%d)",
2046 os::strerror(errno), os::errno_name(errno), errno);
2047 }
2048 }
2049 }
2050
2051 void VMError::report_java_out_of_memory(const char* message) {
2052 if (OnOutOfMemoryError && OnOutOfMemoryError[0]) {
2053 MutexLocker ml(Heap_lock);
2054 VM_ReportJavaOutOfMemory op(message);
2055 VMThread::execute(&op);
2056 }
2057 }
2058
2059 void VMError::show_message_box(char *buf, int buflen) {
2060 bool yes;
2061 do {
2062 error_string(buf, buflen);
2063 yes = os::start_debugging(buf,buflen);
2064 } while (yes);
2065 }
2066
2067 // Fatal error handling is subject to several timeouts:
2068 // - a global timeout (controlled via ErrorLogTimeout)
2069 // - local error reporting step timeouts.
2070 //
2071 // The latter aims to "give the JVM a kick" if it gets stuck in one particular place during
2072 // error reporting. This prevents one error reporting step from hogging all the time allotted
2073 // to error reporting under ErrorLogTimeout.
2074 //
2075 // VMError::check_timeout() is called from the watcher thread and checks for either global
2076 // or step timeout. If a timeout happened, we interrupt the reporting thread and set either
2077 // _reporting_did_timeout or _step_did_timeout to signal which timeout fired. Function returns
2078 // true if the *global* timeout fired, which will cause WatcherThread to shut down the JVM
2079 // immediately.
2080 bool VMError::check_timeout() {
2081
2082 // This function is supposed to be called from watcher thread during fatal error handling only.
2083 assert(VMError::is_error_reported(), "Only call during error handling");
2084 assert(Thread::current()->is_Watcher_thread(), "Only call from watcher thread");
2085
2086 if (ErrorLogTimeout == 0) {
2087 return false;
2088 }
2089
2090 // There are three situations where we suppress the *global* error timeout:
2091 // - if the JVM is embedded and the launcher has its abort hook installed.
2092 // That must be allowed to run.
2093 // - if the user specified one or more OnError commands to run, and these
2094 // did not yet run. These must have finished.
2095 // - if the user (typically developer) specified ShowMessageBoxOnError,
2096 // and the error box has not yet been shown
2097 const bool ignore_global_timeout =
2098 (ShowMessageBoxOnError
2099 || (OnError != nullptr && OnError[0] != '\0')
2100 || Arguments::abort_hook() != nullptr);
2101
2102 const jlong now = get_current_timestamp();
2103
2104 // Global timeout hit?
2105 if (!ignore_global_timeout) {
2106 const jlong reporting_start_time = get_reporting_start_time();
2107 // Timestamp is stored in nanos.
2108 if (reporting_start_time > 0) {
2109 const jlong end = reporting_start_time + (jlong)ErrorLogTimeout * TIMESTAMP_TO_SECONDS_FACTOR;
2110 if (end <= now && !_reporting_did_timeout.load_relaxed()) {
2111 // We hit ErrorLogTimeout and we haven't interrupted the reporting
2112 // thread yet.
2113 _reporting_did_timeout.store_relaxed(true);
2114 interrupt_reporting_thread();
2115 return true; // global timeout
2116 }
2117 }
2118 }
2119
2120 // Reporting step timeout?
2121 const jlong step_start_time = get_step_start_time();
2122 if (step_start_time > 0) {
2123 // A step times out after a quarter of the total timeout. Steps are mostly fast unless they
2124 // hang for some reason, so this simple rule allows for three hanging step and still
2125 // hopefully leaves time enough for the rest of the steps to finish.
2126 const int max_step_timeout_secs = 5;
2127 const jlong timeout_duration = MAX2((jlong)max_step_timeout_secs, (jlong)ErrorLogTimeout * TIMESTAMP_TO_SECONDS_FACTOR / 4);
2128 const jlong end = step_start_time + timeout_duration;
2129 if (end <= now && !_step_did_timeout.load_relaxed()) {
2130 // The step timed out and we haven't interrupted the reporting
2131 // thread yet.
2132 _step_did_timeout.store_relaxed(true);
2133 interrupt_reporting_thread();
2134 return false; // (Not a global timeout)
2135 }
2136 }
2137
2138 return false;
2139
2140 }
2141
2142 #ifdef ASSERT
2143 typedef void (*voidfun_t)();
2144
2145 // Crash with an authentic sigfpe; behavior is subtly different from a real signal
2146 // compared to one generated with raise (asynchronous vs synchronous). See JDK-8065895.
2147 volatile int sigfpe_int = 0;
2148
2149 ATTRIBUTE_NO_UBSAN
2150 static void ALWAYSINLINE crash_with_sigfpe() {
2151
2152 // generate a native synchronous SIGFPE where possible;
2153 sigfpe_int = sigfpe_int/sigfpe_int;
2154
2155 // if that did not cause a signal (e.g. on ppc), just
2156 // raise the signal.
2157 #ifndef _WIN32
2158 // OSX implements raise(sig) incorrectly so we need to
2159 // explicitly target the current thread
2160 pthread_kill(pthread_self(), SIGFPE);
2161 #endif
2162
2163 } // end: crash_with_sigfpe
2164
2165 // crash with sigsegv at non-null address.
2166 static void ALWAYSINLINE crash_with_segfault() {
2167
2168 int* crash_addr = reinterpret_cast<int*>(VMError::segfault_address);
2169 *crash_addr = 1;
2170
2171 } // end: crash_with_segfault
2172
2173 // crash in a controlled way:
2174 // 1 - assert
2175 // 2 - guarantee
2176 // 14 - SIGSEGV
2177 // 15 - SIGFPE
2178 void VMError::controlled_crash(int how) {
2179
2180 // Case 14 is tested by test/hotspot/jtreg/runtime/ErrorHandling/SafeFetchInErrorHandlingTest.java.
2181 // Case 15 is tested by test/hotspot/jtreg/runtime/ErrorHandling/SecondaryErrorTest.java.
2182 // Case 16 is tested by test/hotspot/jtreg/runtime/ErrorHandling/ThreadsListHandleInErrorHandlingTest.java.
2183 // Case 17 is tested by test/hotspot/jtreg/runtime/ErrorHandling/NestedThreadsListHandleInErrorHandlingTest.java.
2184
2185 // We try to grab Threads_lock to keep ThreadsSMRSupport::print_info_on()
2186 // from racing with Threads::add() or Threads::remove() as we
2187 // generate the hs_err_pid file. This makes our ErrorHandling tests
2188 // more stable.
2189 if (!Threads_lock->owned_by_self()) {
2190 Threads_lock->try_lock();
2191 // The VM is going to die so no need to unlock Thread_lock.
2192 }
2193
2194 switch (how) {
2195 case 1: assert(how == 0, "test assert"); break;
2196 case 2: guarantee(how == 0, "test guarantee"); break;
2197
2198 // The other cases are unused.
2199 case 14: crash_with_segfault(); break;
2200 case 15: crash_with_sigfpe(); break;
2201 case 16: {
2202 ThreadsListHandle tlh;
2203 fatal("Force crash with an active ThreadsListHandle.");
2204 }
2205 case 17: {
2206 ThreadsListHandle tlh;
2207 {
2208 ThreadsListHandle tlh2;
2209 fatal("Force crash with a nested ThreadsListHandle.");
2210 }
2211 }
2212 case 18: {
2213 // Trigger an error that should cause ASAN to report a double free or use-after-free.
2214 // Please note that this is not 100% bullet-proof since it assumes that this block
2215 // is not immediately repurposed by some other thread after free.
2216 void* const p = os::malloc(4096, mtTest);
2217 os::free(p);
2218 os::free(p);
2219 }
2220 default:
2221 // If another number is given, give a generic crash.
2222 fatal("Crashing with number %d", how);
2223 }
2224 tty->print_cr("controlled_crash: survived intentional crash. Did you suppress the assert?");
2225 ShouldNotReachHere();
2226 }
2227 #endif // !ASSERT
2228
2229 VMErrorCallbackMark::VMErrorCallbackMark(VMErrorCallback* callback)
2230 : _thread(Thread::current()) {
2231 callback->_next = _thread->_vm_error_callbacks;
2232 _thread->_vm_error_callbacks = callback;
2233 }
2234
2235 VMErrorCallbackMark::~VMErrorCallbackMark() {
2236 assert(_thread->_vm_error_callbacks != nullptr, "Popped too far");
2237 _thread->_vm_error_callbacks = _thread->_vm_error_callbacks->_next;
2238 }