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 | |
42 | static void check_signal_handler(int sig); |
43 | |
44 | static int (*_pthread_getcpuclockid)(pthread_t, clockid_t *); |
45 | static int (*_pthread_setname_np)(pthread_t, const char*); |
46 | |
47 | static address _initial_thread_stack_bottom; |
48 | static uintptr_t _initial_thread_stack_size; |
49 | |
50 | static const char *_glibc_version; |
51 | static const char *_libpthread_version; |
52 | |
53 | static bool _supports_fast_thread_cpu_time; |
54 | |
55 | static GrowableArray<int>* _cpu_to_node; |
56 | static GrowableArray<int>* _nindex_to_node; |
57 | |
58 | // 0x00000000 = uninitialized, |
59 | // 0x01000000 = kernel version unknown, |
60 | // otherwise a 32-bit number: |
61 | // Ox00AABBCC |
62 | // AA, Major Version |
63 | // BB, Minor Version |
64 | // CC, Fix Version |
65 | static uint32_t _os_version; |
66 | |
67 | protected: |
68 | |
69 | static julong _physical_memory; |
70 | static pthread_t _main_thread; |
71 | static Mutex* _createThread_lock; |
72 | static int _page_size; |
73 | |
74 | static julong available_memory(); |
75 | static julong physical_memory() { return _physical_memory; } |
76 | static void set_physical_memory(julong phys_mem) { _physical_memory = phys_mem; } |
77 | static int active_processor_count(); |
78 | |
79 | static void initialize_system_info(); |
80 | |
81 | static int commit_memory_impl(char* addr, size_t bytes, bool exec); |
82 | static int commit_memory_impl(char* addr, size_t bytes, |
83 | size_t alignment_hint, bool exec); |
84 | |
85 | static void set_glibc_version(const char *s) { _glibc_version = s; } |
86 | static void set_libpthread_version(const char *s) { _libpthread_version = s; } |
87 | |
88 | static void rebuild_cpu_to_node_map(); |
89 | static void rebuild_nindex_to_node_map(); |
90 | static GrowableArray<int>* cpu_to_node() { return _cpu_to_node; } |
91 | static GrowableArray<int>* nindex_to_node() { return _nindex_to_node; } |
92 | |
93 | static size_t find_large_page_size(); |
94 | static size_t setup_large_page_size(); |
95 | |
96 | static bool setup_large_page_type(size_t page_size); |
97 | static bool transparent_huge_pages_sanity_check(bool warn, size_t pages_size); |
98 | static bool hugetlbfs_sanity_check(bool warn, size_t page_size); |
99 | |
100 | static char* reserve_memory_special_shm(size_t bytes, size_t alignment, char* req_addr, bool exec); |
101 | static char* reserve_memory_special_huge_tlbfs(size_t bytes, size_t alignment, char* req_addr, bool exec); |
102 | static char* reserve_memory_special_huge_tlbfs_only(size_t bytes, char* req_addr, bool exec); |
103 | static char* reserve_memory_special_huge_tlbfs_mixed(size_t bytes, size_t alignment, char* req_addr, bool exec); |
104 | |
105 | static bool release_memory_special_impl(char* base, size_t bytes); |
106 | static bool release_memory_special_shm(char* base, size_t bytes); |
107 | static bool release_memory_special_huge_tlbfs(char* base, size_t bytes); |
108 | |
109 | static void print_full_memory_info(outputStream* st); |
110 | static void print_container_info(outputStream* st); |
111 | static void print_steal_info(outputStream* st); |
112 | static void print_distro_info(outputStream* st); |
113 | static void print_libversion_info(outputStream* st); |
114 | static void print_proc_sys_info(outputStream* st); |
115 | static void print_ld_preload_file(outputStream* st); |
116 | |
117 | public: |
118 | struct CPUPerfTicks { |
119 | uint64_t used; |
120 | uint64_t usedKernel; |
121 | uint64_t total; |
122 | uint64_t steal; |
123 | bool has_steal_ticks; |
124 | }; |
125 | |
126 | // which_logical_cpu=-1 returns accumulated ticks for all cpus. |
127 | static bool get_tick_information(CPUPerfTicks* pticks, int which_logical_cpu); |
128 | static bool _stack_is_executable; |
129 | static void *dlopen_helper(const char *name, char *ebuf, int ebuflen); |
130 | static void *dll_load_in_vmthread(const char *name, char *ebuf, int ebuflen); |
131 | |
132 | static void init_thread_fpu_state(); |
133 | static int get_fpu_control_word(); |
134 | static void set_fpu_control_word(int fpu_control); |
135 | static pthread_t main_thread(void) { return _main_thread; } |
136 | // returns kernel thread id (similar to LWP id on Solaris), which can be |
137 | // used to access /proc |
138 | static pid_t gettid(); |
139 | static void set_createThread_lock(Mutex* lk) { _createThread_lock = lk; } |
140 | static Mutex* createThread_lock(void) { return _createThread_lock; } |
141 | static void hotspot_sigmask(Thread* thread); |
142 | |
143 | static address initial_thread_stack_bottom(void) { return _initial_thread_stack_bottom; } |
144 | static uintptr_t initial_thread_stack_size(void) { return _initial_thread_stack_size; } |
145 | |
146 | static int page_size(void) { return _page_size; } |
147 | static void set_page_size(int val) { _page_size = val; } |
148 | |
149 | static address ucontext_get_pc(const ucontext_t* uc); |
150 | static void ucontext_set_pc(ucontext_t* uc, address pc); |
151 | static intptr_t* ucontext_get_sp(const ucontext_t* uc); |
152 | static intptr_t* ucontext_get_fp(const ucontext_t* uc); |
153 | |
154 | // For Analyzer Forte AsyncGetCallTrace profiling support: |
155 | // |
156 | // This interface should be declared in os_linux_i486.hpp, but |
157 | // that file provides extensions to the os class and not the |
158 | // Linux class. |
159 | static ExtendedPC fetch_frame_from_ucontext(Thread* thread, const ucontext_t* uc, |
160 | intptr_t** ret_sp, intptr_t** ret_fp); |
161 | |
162 | static bool get_frame_at_stack_banging_point(JavaThread* thread, ucontext_t* uc, frame* fr); |
163 | |
164 | // This boolean allows users to forward their own non-matching signals |
165 | // to JVM_handle_linux_signal, harmlessly. |
166 | static bool signal_handlers_are_installed; |
167 | |
168 | static int get_our_sigflags(int); |
169 | static void set_our_sigflags(int, int); |
170 | static void signal_sets_init(); |
171 | static void install_signal_handlers(); |
172 | static void set_signal_handler(int, bool); |
173 | |
174 | static sigset_t* unblocked_signals(); |
175 | static sigset_t* vm_signals(); |
176 | |
177 | // For signal-chaining |
178 | static struct sigaction *get_chained_signal_action(int sig); |
179 | static bool chained_handler(int sig, siginfo_t* siginfo, void* context); |
180 | |
181 | // GNU libc and libpthread version strings |
182 | static const char *glibc_version() { return _glibc_version; } |
183 | static const char *libpthread_version() { return _libpthread_version; } |
184 | |
185 | static void libpthread_init(); |
186 | static void sched_getcpu_init(); |
187 | static bool libnuma_init(); |
188 | static void* libnuma_dlsym(void* handle, const char* name); |
189 | // libnuma v2 (libnuma_1.2) symbols |
190 | static void* libnuma_v2_dlsym(void* handle, const char* name); |
191 | |
192 | // Return default guard size for the specified thread type |
193 | static size_t default_guard_size(os::ThreadType thr_type); |
194 | |
195 | static void capture_initial_stack(size_t max_size); |
196 | |
197 | // Stack overflow handling |
198 | static bool manually_expand_stack(JavaThread * t, address addr); |
199 | static int max_register_window_saves_before_flushing(); |
200 | |
201 | // fast POSIX clocks support |
202 | static void fast_thread_clock_init(void); |
203 | |
204 | static int pthread_getcpuclockid(pthread_t tid, clockid_t *clock_id) { |
205 | return _pthread_getcpuclockid ? _pthread_getcpuclockid(tid, clock_id) : -1; |
206 | } |
207 | |
208 | static bool supports_fast_thread_cpu_time() { |
209 | return _supports_fast_thread_cpu_time; |
210 | } |
211 | |
212 | static jlong fast_thread_cpu_time(clockid_t clockid); |
213 | |
214 | static void initialize_os_info(); |
215 | static bool os_version_is_known(); |
216 | static uint32_t os_version(); |
217 | |
218 | // Stack repair handling |
219 | |
220 | // none present |
221 | |
222 | private: |
223 | static void numa_init(); |
224 | static void expand_stack_to(address bottom); |
225 | |
226 | typedef int (*sched_getcpu_func_t)(void); |
227 | typedef int (*numa_node_to_cpus_func_t)(int node, unsigned long *buffer, int bufferlen); |
228 | typedef int (*numa_max_node_func_t)(void); |
229 | typedef int (*numa_num_configured_nodes_func_t)(void); |
230 | typedef int (*numa_available_func_t)(void); |
231 | typedef int (*numa_tonode_memory_func_t)(void *start, size_t size, int node); |
232 | typedef void (*numa_interleave_memory_func_t)(void *start, size_t size, unsigned long *nodemask); |
233 | typedef void (*numa_interleave_memory_v2_func_t)(void *start, size_t size, struct bitmask* mask); |
234 | typedef struct bitmask* (*numa_get_membind_func_t)(void); |
235 | typedef struct bitmask* (*numa_get_interleave_mask_func_t)(void); |
236 | |
237 | typedef void (*numa_set_bind_policy_func_t)(int policy); |
238 | typedef int (*numa_bitmask_isbitset_func_t)(struct bitmask *bmp, unsigned int n); |
239 | typedef int (*numa_distance_func_t)(int node1, int node2); |
240 | |
241 | static sched_getcpu_func_t _sched_getcpu; |
242 | static numa_node_to_cpus_func_t _numa_node_to_cpus; |
243 | static numa_max_node_func_t _numa_max_node; |
244 | static numa_num_configured_nodes_func_t _numa_num_configured_nodes; |
245 | static numa_available_func_t _numa_available; |
246 | static numa_tonode_memory_func_t _numa_tonode_memory; |
247 | static numa_interleave_memory_func_t _numa_interleave_memory; |
248 | static numa_interleave_memory_v2_func_t _numa_interleave_memory_v2; |
249 | static numa_set_bind_policy_func_t _numa_set_bind_policy; |
250 | static numa_bitmask_isbitset_func_t _numa_bitmask_isbitset; |
251 | static numa_distance_func_t _numa_distance; |
252 | static numa_get_membind_func_t _numa_get_membind; |
253 | static numa_get_interleave_mask_func_t _numa_get_interleave_mask; |
254 | static unsigned long* _numa_all_nodes; |
255 | static struct bitmask* _numa_all_nodes_ptr; |
256 | static struct bitmask* _numa_nodes_ptr; |
257 | static struct bitmask* _numa_interleave_bitmask; |
258 | static struct bitmask* _numa_membind_bitmask; |
259 | |
260 | static void set_sched_getcpu(sched_getcpu_func_t func) { _sched_getcpu = func; } |
261 | static void set_numa_node_to_cpus(numa_node_to_cpus_func_t func) { _numa_node_to_cpus = func; } |
262 | static void set_numa_max_node(numa_max_node_func_t func) { _numa_max_node = func; } |
263 | static void set_numa_num_configured_nodes(numa_num_configured_nodes_func_t func) { _numa_num_configured_nodes = func; } |
264 | static void set_numa_available(numa_available_func_t func) { _numa_available = func; } |
265 | static void set_numa_tonode_memory(numa_tonode_memory_func_t func) { _numa_tonode_memory = func; } |
266 | static void set_numa_interleave_memory(numa_interleave_memory_func_t func) { _numa_interleave_memory = func; } |
267 | static void set_numa_interleave_memory_v2(numa_interleave_memory_v2_func_t func) { _numa_interleave_memory_v2 = func; } |
268 | static void set_numa_set_bind_policy(numa_set_bind_policy_func_t func) { _numa_set_bind_policy = func; } |
269 | static void set_numa_bitmask_isbitset(numa_bitmask_isbitset_func_t func) { _numa_bitmask_isbitset = func; } |
270 | static void set_numa_distance(numa_distance_func_t func) { _numa_distance = func; } |
271 | static void set_numa_get_membind(numa_get_membind_func_t func) { _numa_get_membind = func; } |
272 | static void set_numa_get_interleave_mask(numa_get_interleave_mask_func_t func) { _numa_get_interleave_mask = func; } |
273 | static void set_numa_all_nodes(unsigned long* ptr) { _numa_all_nodes = ptr; } |
274 | static void set_numa_all_nodes_ptr(struct bitmask **ptr) { _numa_all_nodes_ptr = (ptr == NULL ? NULL : *ptr); } |
275 | static void set_numa_nodes_ptr(struct bitmask **ptr) { _numa_nodes_ptr = (ptr == NULL ? NULL : *ptr); } |
276 | static void set_numa_interleave_bitmask(struct bitmask* ptr) { _numa_interleave_bitmask = ptr ; } |
277 | static void set_numa_membind_bitmask(struct bitmask* ptr) { _numa_membind_bitmask = ptr ; } |
278 | static int sched_getcpu_syscall(void); |
279 | |
280 | enum NumaAllocationPolicy{ |
281 | NotInitialized, |
282 | Membind, |
283 | Interleave |
284 | }; |
285 | static NumaAllocationPolicy _current_numa_policy; |
286 | |
287 | public: |
288 | static int sched_getcpu() { return _sched_getcpu != NULL ? _sched_getcpu() : -1; } |
289 | static int numa_node_to_cpus(int node, unsigned long *buffer, int bufferlen) { |
290 | return _numa_node_to_cpus != NULL ? _numa_node_to_cpus(node, buffer, bufferlen) : -1; |
291 | } |
292 | static int numa_max_node() { return _numa_max_node != NULL ? _numa_max_node() : -1; } |
293 | static int numa_num_configured_nodes() { |
294 | return _numa_num_configured_nodes != NULL ? _numa_num_configured_nodes() : -1; |
295 | } |
296 | static int numa_available() { return _numa_available != NULL ? _numa_available() : -1; } |
297 | static int numa_tonode_memory(void *start, size_t size, int node) { |
298 | return _numa_tonode_memory != NULL ? _numa_tonode_memory(start, size, node) : -1; |
299 | } |
300 | |
301 | static bool is_running_in_interleave_mode() { |
302 | return _current_numa_policy == Interleave; |
303 | } |
304 | |
305 | static void set_configured_numa_policy(NumaAllocationPolicy numa_policy) { |
306 | _current_numa_policy = numa_policy; |
307 | } |
308 | |
309 | static NumaAllocationPolicy identify_numa_policy() { |
310 | for (int node = 0; node <= Linux::numa_max_node(); node++) { |
311 | if (Linux::_numa_bitmask_isbitset(Linux::_numa_interleave_bitmask, node)) { |
312 | return Interleave; |
313 | } |
314 | } |
315 | return Membind; |
316 | } |
317 | |
318 | static void numa_interleave_memory(void *start, size_t size) { |
319 | // Prefer v2 API |
320 | if (_numa_interleave_memory_v2 != NULL) { |
321 | if (is_running_in_interleave_mode()) { |
322 | _numa_interleave_memory_v2(start, size, _numa_interleave_bitmask); |
323 | } else if (_numa_membind_bitmask != NULL) { |
324 | _numa_interleave_memory_v2(start, size, _numa_membind_bitmask); |
325 | } |
326 | } else if (_numa_interleave_memory != NULL) { |
327 | _numa_interleave_memory(start, size, _numa_all_nodes); |
328 | } |
329 | } |
330 | static void numa_set_bind_policy(int policy) { |
331 | if (_numa_set_bind_policy != NULL) { |
332 | _numa_set_bind_policy(policy); |
333 | } |
334 | } |
335 | static int numa_distance(int node1, int node2) { |
336 | return _numa_distance != NULL ? _numa_distance(node1, node2) : -1; |
337 | } |
338 | static int get_node_by_cpu(int cpu_id); |
339 | static int get_existing_num_nodes(); |
340 | // Check if numa node is configured (non-zero memory node). |
341 | static bool is_node_in_configured_nodes(unsigned int n) { |
342 | if (_numa_bitmask_isbitset != NULL && _numa_all_nodes_ptr != NULL) { |
343 | return _numa_bitmask_isbitset(_numa_all_nodes_ptr, n); |
344 | } else |
345 | return false; |
346 | } |
347 | // Check if numa node exists in the system (including zero memory nodes). |
348 | static bool is_node_in_existing_nodes(unsigned int n) { |
349 | if (_numa_bitmask_isbitset != NULL && _numa_nodes_ptr != NULL) { |
350 | return _numa_bitmask_isbitset(_numa_nodes_ptr, n); |
351 | } else if (_numa_bitmask_isbitset != NULL && _numa_all_nodes_ptr != NULL) { |
352 | // Not all libnuma API v2 implement numa_nodes_ptr, so it's not possible |
353 | // to trust the API version for checking its absence. On the other hand, |
354 | // numa_nodes_ptr found in libnuma 2.0.9 and above is the only way to get |
355 | // a complete view of all numa nodes in the system, hence numa_nodes_ptr |
356 | // is used to handle CPU and nodes on architectures (like PowerPC) where |
357 | // there can exist nodes with CPUs but no memory or vice-versa and the |
358 | // nodes may be non-contiguous. For most of the architectures, like |
359 | // x86_64, numa_node_ptr presents the same node set as found in |
360 | // numa_all_nodes_ptr so it's possible to use numa_all_nodes_ptr as a |
361 | // substitute. |
362 | return _numa_bitmask_isbitset(_numa_all_nodes_ptr, n); |
363 | } else |
364 | return false; |
365 | } |
366 | // Check if node is in bound node set. |
367 | static bool is_node_in_bound_nodes(int node) { |
368 | if (_numa_bitmask_isbitset != NULL) { |
369 | if (is_running_in_interleave_mode()) { |
370 | return _numa_bitmask_isbitset(_numa_interleave_bitmask, node); |
371 | } else { |
372 | return _numa_membind_bitmask != NULL ? _numa_bitmask_isbitset(_numa_membind_bitmask, node) : false; |
373 | } |
374 | } |
375 | return false; |
376 | } |
377 | // Check if bound to only one numa node. |
378 | // Returns true if bound to a single numa node, otherwise returns false. |
379 | static bool is_bound_to_single_node() { |
380 | int nodes = 0; |
381 | struct bitmask* bmp = NULL; |
382 | unsigned int node = 0; |
383 | unsigned int highest_node_number = 0; |
384 | |
385 | if (_numa_get_membind != NULL && _numa_max_node != NULL && _numa_bitmask_isbitset != NULL) { |
386 | bmp = _numa_get_membind(); |
387 | highest_node_number = _numa_max_node(); |
388 | } else { |
389 | return false; |
390 | } |
391 | |
392 | for (node = 0; node <= highest_node_number; node++) { |
393 | if (_numa_bitmask_isbitset(bmp, node)) { |
394 | nodes++; |
395 | } |
396 | } |
397 | |
398 | if (nodes == 1) { |
399 | return true; |
400 | } else { |
401 | return false; |
402 | } |
403 | } |
404 | }; |
405 | |
406 | #endif // OS_LINUX_OS_LINUX_HPP |
407 | |