1 /*
2 * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef OS_LINUX_OS_LINUX_HPP
26 #define OS_LINUX_OS_LINUX_HPP
27
28 // Linux_OS defines the interface to Linux operating systems
29
30 // Information about the protection of the page at address '0' on this os.
31 static bool zero_page_read_protected() { return true; }
32
33 class Linux {
34 friend class os;
35 friend class OSContainer;
36 friend class TestReserveMemorySpecial;
37
38 static bool libjsig_is_loaded; // libjsig that interposes sigaction(),
39 // __sigaction(), signal() is loaded
40 static struct sigaction *(*get_signal_action)(int);
41 static struct sigaction *get_preinstalled_handler(int);
42 static void save_preinstalled_handler(int, struct sigaction&);
43
44 static void check_signal_handler(int sig);
45
46 static int (*_pthread_getcpuclockid)(pthread_t, clockid_t *);
47 static int (*_pthread_setname_np)(pthread_t, const char*);
48
49 static address _initial_thread_stack_bottom;
50 static uintptr_t _initial_thread_stack_size;
51
52 static const char *_glibc_version;
53 static const char *_libpthread_version;
54
55 static bool _supports_fast_thread_cpu_time;
56
57 static GrowableArray<int>* _cpu_to_node;
58 static GrowableArray<int>* _nindex_to_node;
59
60 // 0x00000000 = uninitialized,
61 // 0x01000000 = kernel version unknown,
62 // otherwise a 32-bit number:
63 // Ox00AABBCC
64 // AA, Major Version
65 // BB, Minor Version
66 // CC, Fix Version
67 static uint32_t _os_version;
68
69 protected:
70
71 static julong _physical_memory;
72 static pthread_t _main_thread;
73 static Mutex* _createThread_lock;
74 static int _page_size;
75
76 static julong available_memory();
77 static julong physical_memory() { return _physical_memory; }
78 static void set_physical_memory(julong phys_mem) { _physical_memory = phys_mem; }
79 static int active_processor_count();
80
81 static void initialize_system_info();
82
83 static int commit_memory_impl(char* addr, size_t bytes, bool exec);
84 static int commit_memory_impl(char* addr, size_t bytes,
85 size_t alignment_hint, bool exec);
86
87 static void set_glibc_version(const char *s) { _glibc_version = s; }
88 static void set_libpthread_version(const char *s) { _libpthread_version = s; }
89
90 static void rebuild_cpu_to_node_map();
91 static void rebuild_nindex_to_node_map();
92 static GrowableArray<int>* cpu_to_node() { return _cpu_to_node; }
93 static GrowableArray<int>* nindex_to_node() { return _nindex_to_node; }
94
95 static size_t find_large_page_size();
96 static size_t setup_large_page_size();
97
98 static bool setup_large_page_type(size_t page_size);
99 static bool transparent_huge_pages_sanity_check(bool warn, size_t pages_size);
100 static bool hugetlbfs_sanity_check(bool warn, size_t page_size);
101
102 static char* reserve_memory_special_shm(size_t bytes, size_t alignment, char* req_addr, bool exec);
103 static char* reserve_memory_special_huge_tlbfs(size_t bytes, size_t alignment, char* req_addr, bool exec);
104 static char* reserve_memory_special_huge_tlbfs_only(size_t bytes, char* req_addr, bool exec);
105 static char* reserve_memory_special_huge_tlbfs_mixed(size_t bytes, size_t alignment, char* req_addr, bool exec);
106
107 static bool release_memory_special_impl(char* base, size_t bytes);
108 static bool release_memory_special_shm(char* base, size_t bytes);
109 static bool release_memory_special_huge_tlbfs(char* base, size_t bytes);
110
111 static void print_full_memory_info(outputStream* st);
112 static void print_container_info(outputStream* st);
113 static void print_virtualization_info(outputStream* st);
114 static void print_distro_info(outputStream* st);
115 static void print_libversion_info(outputStream* st);
116 static void print_proc_sys_info(outputStream* st);
117 static void print_ld_preload_file(outputStream* st);
118
119 public:
120 static bool _stack_is_executable;
121 static void *dlopen_helper(const char *name, char *ebuf, int ebuflen);
122 static void *dll_load_in_vmthread(const char *name, char *ebuf, int ebuflen);
123
124 static void init_thread_fpu_state();
125 static int get_fpu_control_word();
126 static void set_fpu_control_word(int fpu_control);
127 static pthread_t main_thread(void) { return _main_thread; }
128 // returns kernel thread id (similar to LWP id on Solaris), which can be
129 // used to access /proc
130 static pid_t gettid();
131 static void set_createThread_lock(Mutex* lk) { _createThread_lock = lk; }
132 static Mutex* createThread_lock(void) { return _createThread_lock; }
133 static void hotspot_sigmask(Thread* thread);
134
135 static address initial_thread_stack_bottom(void) { return _initial_thread_stack_bottom; }
136 static uintptr_t initial_thread_stack_size(void) { return _initial_thread_stack_size; }
137
138 static int page_size(void) { return _page_size; }
139 static void set_page_size(int val) { _page_size = val; }
140
141 static address ucontext_get_pc(const ucontext_t* uc);
142 static void ucontext_set_pc(ucontext_t* uc, address pc);
143 static intptr_t* ucontext_get_sp(const ucontext_t* uc);
144 static intptr_t* ucontext_get_fp(const ucontext_t* uc);
145
146 // For Analyzer Forte AsyncGetCallTrace profiling support:
147 //
148 // This interface should be declared in os_linux_i486.hpp, but
149 // that file provides extensions to the os class and not the
150 // Linux class.
151 static ExtendedPC fetch_frame_from_ucontext(Thread* thread, const ucontext_t* uc,
152 intptr_t** ret_sp, intptr_t** ret_fp);
153
154 static bool get_frame_at_stack_banging_point(JavaThread* thread, ucontext_t* uc, frame* fr);
155
156 // This boolean allows users to forward their own non-matching signals
157 // to JVM_handle_linux_signal, harmlessly.
158 static bool signal_handlers_are_installed;
159
160 static int get_our_sigflags(int);
161 static void set_our_sigflags(int, int);
162 static void signal_sets_init();
163 static void install_signal_handlers();
164 static void set_signal_handler(int, bool);
165
166 static sigset_t* unblocked_signals();
167 static sigset_t* vm_signals();
168
169 // For signal-chaining
170 static struct sigaction *get_chained_signal_action(int sig);
171 static bool chained_handler(int sig, siginfo_t* siginfo, void* context);
172
173 // GNU libc and libpthread version strings
174 static const char *glibc_version() { return _glibc_version; }
175 static const char *libpthread_version() { return _libpthread_version; }
176
177 static void libpthread_init();
178 static void sched_getcpu_init();
179 static bool libnuma_init();
180 static void* libnuma_dlsym(void* handle, const char* name);
181 // libnuma v2 (libnuma_1.2) symbols
182 static void* libnuma_v2_dlsym(void* handle, const char* name);
183
184 // Return default guard size for the specified thread type
185 static size_t default_guard_size(os::ThreadType thr_type);
186
187 static void capture_initial_stack(size_t max_size);
188
189 // Stack overflow handling
190 static bool manually_expand_stack(JavaThread * t, address addr);
191 static int max_register_window_saves_before_flushing();
192
193 // fast POSIX clocks support
194 static void fast_thread_clock_init(void);
195
196 static int pthread_getcpuclockid(pthread_t tid, clockid_t *clock_id) {
197 return _pthread_getcpuclockid ? _pthread_getcpuclockid(tid, clock_id) : -1;
198 }
199
200 static bool supports_fast_thread_cpu_time() {
201 return _supports_fast_thread_cpu_time;
202 }
203
204 static jlong fast_thread_cpu_time(clockid_t clockid);
205
206 static void initialize_os_info();
207 static bool os_version_is_known();
208 static uint32_t os_version();
209
210 // Stack repair handling
211
212 // none present
213
214 private:
215 static void numa_init();
216 static void expand_stack_to(address bottom);
217
218 typedef int (*sched_getcpu_func_t)(void);
219 typedef int (*numa_node_to_cpus_func_t)(int node, unsigned long *buffer, int bufferlen);
220 typedef int (*numa_max_node_func_t)(void);
221 typedef int (*numa_num_configured_nodes_func_t)(void);
222 typedef int (*numa_available_func_t)(void);
223 typedef int (*numa_tonode_memory_func_t)(void *start, size_t size, int node);
224 typedef void (*numa_interleave_memory_func_t)(void *start, size_t size, unsigned long *nodemask);
225 typedef void (*numa_interleave_memory_v2_func_t)(void *start, size_t size, struct bitmask* mask);
226 typedef struct bitmask* (*numa_get_membind_func_t)(void);
227 typedef struct bitmask* (*numa_get_interleave_mask_func_t)(void);
228
229 typedef void (*numa_set_bind_policy_func_t)(int policy);
230 typedef int (*numa_bitmask_isbitset_func_t)(struct bitmask *bmp, unsigned int n);
231 typedef int (*numa_distance_func_t)(int node1, int node2);
232
233 static sched_getcpu_func_t _sched_getcpu;
234 static numa_node_to_cpus_func_t _numa_node_to_cpus;
235 static numa_max_node_func_t _numa_max_node;
236 static numa_num_configured_nodes_func_t _numa_num_configured_nodes;
237 static numa_available_func_t _numa_available;
238 static numa_tonode_memory_func_t _numa_tonode_memory;
239 static numa_interleave_memory_func_t _numa_interleave_memory;
240 static numa_interleave_memory_v2_func_t _numa_interleave_memory_v2;
241 static numa_set_bind_policy_func_t _numa_set_bind_policy;
242 static numa_bitmask_isbitset_func_t _numa_bitmask_isbitset;
243 static numa_distance_func_t _numa_distance;
244 static numa_get_membind_func_t _numa_get_membind;
245 static numa_get_interleave_mask_func_t _numa_get_interleave_mask;
246 static unsigned long* _numa_all_nodes;
247 static struct bitmask* _numa_all_nodes_ptr;
248 static struct bitmask* _numa_nodes_ptr;
249 static struct bitmask* _numa_interleave_bitmask;
250 static struct bitmask* _numa_membind_bitmask;
251
252 static void set_sched_getcpu(sched_getcpu_func_t func) { _sched_getcpu = func; }
253 static void set_numa_node_to_cpus(numa_node_to_cpus_func_t func) { _numa_node_to_cpus = func; }
254 static void set_numa_max_node(numa_max_node_func_t func) { _numa_max_node = func; }
255 static void set_numa_num_configured_nodes(numa_num_configured_nodes_func_t func) { _numa_num_configured_nodes = func; }
256 static void set_numa_available(numa_available_func_t func) { _numa_available = func; }
257 static void set_numa_tonode_memory(numa_tonode_memory_func_t func) { _numa_tonode_memory = func; }
258 static void set_numa_interleave_memory(numa_interleave_memory_func_t func) { _numa_interleave_memory = func; }
259 static void set_numa_interleave_memory_v2(numa_interleave_memory_v2_func_t func) { _numa_interleave_memory_v2 = func; }
260 static void set_numa_set_bind_policy(numa_set_bind_policy_func_t func) { _numa_set_bind_policy = func; }
261 static void set_numa_bitmask_isbitset(numa_bitmask_isbitset_func_t func) { _numa_bitmask_isbitset = func; }
262 static void set_numa_distance(numa_distance_func_t func) { _numa_distance = func; }
263 static void set_numa_get_membind(numa_get_membind_func_t func) { _numa_get_membind = func; }
264 static void set_numa_get_interleave_mask(numa_get_interleave_mask_func_t func) { _numa_get_interleave_mask = func; }
265 static void set_numa_all_nodes(unsigned long* ptr) { _numa_all_nodes = ptr; }
266 static void set_numa_all_nodes_ptr(struct bitmask **ptr) { _numa_all_nodes_ptr = (ptr == NULL ? NULL : *ptr); }
267 static void set_numa_nodes_ptr(struct bitmask **ptr) { _numa_nodes_ptr = (ptr == NULL ? NULL : *ptr); }
268 static void set_numa_interleave_bitmask(struct bitmask* ptr) { _numa_interleave_bitmask = ptr ; }
269 static void set_numa_membind_bitmask(struct bitmask* ptr) { _numa_membind_bitmask = ptr ; }
270 static int sched_getcpu_syscall(void);
271
272 enum NumaAllocationPolicy{
273 NotInitialized,
274 Membind,
275 Interleave
276 };
277 static NumaAllocationPolicy _current_numa_policy;
278
279 public:
280 static int sched_getcpu() { return _sched_getcpu != NULL ? _sched_getcpu() : -1; }
281 static int numa_node_to_cpus(int node, unsigned long *buffer, int bufferlen) {
282 return _numa_node_to_cpus != NULL ? _numa_node_to_cpus(node, buffer, bufferlen) : -1;
283 }
284 static int numa_max_node() { return _numa_max_node != NULL ? _numa_max_node() : -1; }
285 static int numa_num_configured_nodes() {
286 return _numa_num_configured_nodes != NULL ? _numa_num_configured_nodes() : -1;
287 }
288 static int numa_available() { return _numa_available != NULL ? _numa_available() : -1; }
289 static int numa_tonode_memory(void *start, size_t size, int node) {
290 return _numa_tonode_memory != NULL ? _numa_tonode_memory(start, size, node) : -1;
291 }
292
293 static bool is_running_in_interleave_mode() {
294 return _current_numa_policy == Interleave;
295 }
296
297 static void set_configured_numa_policy(NumaAllocationPolicy numa_policy) {
298 _current_numa_policy = numa_policy;
299 }
300
301 static NumaAllocationPolicy identify_numa_policy() {
302 for (int node = 0; node <= Linux::numa_max_node(); node++) {
303 if (Linux::_numa_bitmask_isbitset(Linux::_numa_interleave_bitmask, node)) {
304 return Interleave;
305 }
306 }
307 return Membind;
308 }
309
310 static void numa_interleave_memory(void *start, size_t size) {
311 // Prefer v2 API
312 if (_numa_interleave_memory_v2 != NULL) {
313 if (is_running_in_interleave_mode()) {
314 _numa_interleave_memory_v2(start, size, _numa_interleave_bitmask);
315 } else if (_numa_membind_bitmask != NULL) {
316 _numa_interleave_memory_v2(start, size, _numa_membind_bitmask);
317 }
318 } else if (_numa_interleave_memory != NULL) {
319 _numa_interleave_memory(start, size, _numa_all_nodes);
320 }
321 }
322 static void numa_set_bind_policy(int policy) {
323 if (_numa_set_bind_policy != NULL) {
324 _numa_set_bind_policy(policy);
325 }
326 }
327 static int numa_distance(int node1, int node2) {
328 return _numa_distance != NULL ? _numa_distance(node1, node2) : -1;
329 }
330 static int get_node_by_cpu(int cpu_id);
331 static int get_existing_num_nodes();
332 // Check if numa node is configured (non-zero memory node).
333 static bool is_node_in_configured_nodes(unsigned int n) {
334 if (_numa_bitmask_isbitset != NULL && _numa_all_nodes_ptr != NULL) {
335 return _numa_bitmask_isbitset(_numa_all_nodes_ptr, n);
336 } else
337 return false;
338 }
339 // Check if numa node exists in the system (including zero memory nodes).
340 static bool is_node_in_existing_nodes(unsigned int n) {
341 if (_numa_bitmask_isbitset != NULL && _numa_nodes_ptr != NULL) {
342 return _numa_bitmask_isbitset(_numa_nodes_ptr, n);
343 } else if (_numa_bitmask_isbitset != NULL && _numa_all_nodes_ptr != NULL) {
344 // Not all libnuma API v2 implement numa_nodes_ptr, so it's not possible
345 // to trust the API version for checking its absence. On the other hand,
346 // numa_nodes_ptr found in libnuma 2.0.9 and above is the only way to get
347 // a complete view of all numa nodes in the system, hence numa_nodes_ptr
348 // is used to handle CPU and nodes on architectures (like PowerPC) where
349 // there can exist nodes with CPUs but no memory or vice-versa and the
350 // nodes may be non-contiguous. For most of the architectures, like
351 // x86_64, numa_node_ptr presents the same node set as found in
352 // numa_all_nodes_ptr so it's possible to use numa_all_nodes_ptr as a
353 // substitute.
354 return _numa_bitmask_isbitset(_numa_all_nodes_ptr, n);
355 } else
356 return false;
357 }
358 // Check if node is in bound node set.
359 static bool is_node_in_bound_nodes(int node) {
360 if (_numa_bitmask_isbitset != NULL) {
361 if (is_running_in_interleave_mode()) {
362 return _numa_bitmask_isbitset(_numa_interleave_bitmask, node);
363 } else {
364 return _numa_membind_bitmask != NULL ? _numa_bitmask_isbitset(_numa_membind_bitmask, node) : false;
365 }
366 }
367 return false;
368 }
369 // Check if bound to only one numa node.
370 // Returns true if bound to a single numa node, otherwise returns false.
371 static bool is_bound_to_single_node() {
372 int nodes = 0;
373 struct bitmask* bmp = NULL;
374 unsigned int node = 0;
375 unsigned int highest_node_number = 0;
376
377 if (_numa_get_membind != NULL && _numa_max_node != NULL && _numa_bitmask_isbitset != NULL) {
378 bmp = _numa_get_membind();
379 highest_node_number = _numa_max_node();
380 } else {
381 return false;
382 }
383
384 for (node = 0; node <= highest_node_number; node++) {
385 if (_numa_bitmask_isbitset(bmp, node)) {
386 nodes++;
387 }
388 }
389
390 if (nodes == 1) {
391 return true;
392 } else {
393 return false;
394 }
395 }
396 };
397
398 #endif // OS_LINUX_OS_LINUX_HPP