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