1 /*
2 * Copyright (C) 2008 The Android Open Source Project
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * * Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in
12 * the documentation and/or other materials provided with the
13 * distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
16 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
17 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
18 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
19 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
21 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
22 * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
23 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
24 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
25 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29 #pragma once
30
31 #include <pthread.h>
32 #include <stdatomic.h>
33
34 #if __has_feature(hwaddress_sanitizer)
35 #include <sanitizer/hwasan_interface.h>
36 #else
37 #define __hwasan_thread_enter()
38 #define __hwasan_thread_exit()
39 #endif
40
41 #include "private/bionic_elf_tls.h"
42 #include "private/bionic_lock.h"
43 #include "private/bionic_tls.h"
44
45 // Has the thread been detached by a pthread_join or pthread_detach call?
46 #define PTHREAD_ATTR_FLAG_DETACHED 0x00000001
47
48 // Has the thread been joined by another thread?
49 #define PTHREAD_ATTR_FLAG_JOINED 0x00000002
50
51 // Used for pthread_attr_setinheritsched. We need two flags for this apparent
52 // boolean because our historical behavior matches neither of the POSIX choices.
53 #define PTHREAD_ATTR_FLAG_INHERIT 0x00000004
54 #define PTHREAD_ATTR_FLAG_EXPLICIT 0x00000008
55
56 enum ThreadJoinState {
57 THREAD_NOT_JOINED,
58 THREAD_EXITED_NOT_JOINED,
59 THREAD_JOINED,
60 THREAD_DETACHED
61 };
62
63 class thread_local_dtor;
64
65 class pthread_internal_t {
66 public:
67 class pthread_internal_t* next;
68 class pthread_internal_t* prev;
69
70 pid_t tid;
71
72 private:
73 pid_t cached_pid_;
74
75 public:
invalidate_cached_pid()76 pid_t invalidate_cached_pid() {
77 pid_t old_value;
78 get_cached_pid(&old_value);
79 set_cached_pid(0);
80 return old_value;
81 }
82
set_cached_pid(pid_t value)83 void set_cached_pid(pid_t value) {
84 cached_pid_ = value;
85 }
86
get_cached_pid(pid_t * cached_pid)87 bool get_cached_pid(pid_t* cached_pid) {
88 *cached_pid = cached_pid_;
89 return (*cached_pid != 0);
90 }
91
92 pthread_attr_t attr;
93
94 _Atomic(ThreadJoinState) join_state;
95
96 __pthread_cleanup_t* cleanup_stack;
97
98 void* (*start_routine)(void*);
99 void* start_routine_arg;
100 void* return_value;
101
102 void* alternate_signal_stack;
103
104 // The start address of the shadow call stack's guard region (arm64 only).
105 // This address is only used to deallocate the shadow call stack on thread
106 // exit; the address of the stack itself is stored only in the x18 register.
107 // Because the protection offered by SCS relies on the secrecy of the stack
108 // address, storing the address here weakens the protection, but only
109 // slightly, because it is relatively easy for an attacker to discover the
110 // address of the guard region anyway (e.g. it can be discovered by reference
111 // to other allocations), but not the stack itself, which is <0.1% of the size
112 // of the guard region.
113 //
114 // There are at least two other options for discovering the start address of
115 // the guard region on thread exit, but they are not as simple as storing in
116 // TLS.
117 // 1) Derive it from the value of the x18 register. This is only possible in
118 // processes that do not contain legacy code that might clobber x18,
119 // therefore each process must declare early during process startup whether
120 // it might load legacy code.
121 // 2) Mark the guard region as such using prctl(PR_SET_VMA_ANON_NAME) and
122 // discover its address by reading /proc/self/maps. One issue with this is
123 // that reading /proc/self/maps can race with allocations, so we may need
124 // code to handle retries.
125 void* shadow_call_stack_guard_region;
126
127 Lock startup_handshake_lock;
128
129 void* mmap_base;
130 size_t mmap_size;
131
132 thread_local_dtor* thread_local_dtors;
133
134 /*
135 * The dynamic linker implements dlerror(3), which makes it hard for us to implement this
136 * per-thread buffer by simply using malloc(3) and free(3).
137 */
138 char* current_dlerror;
139 #define __BIONIC_DLERROR_BUFFER_SIZE 512
140 char dlerror_buffer[__BIONIC_DLERROR_BUFFER_SIZE];
141
142 bionic_tls* bionic_tls;
143
144 int errno_value;
145 };
146
147 struct ThreadMapping {
148 char* mmap_base;
149 size_t mmap_size;
150
151 char* static_tls;
152 char* stack_base;
153 char* stack_top;
154 };
155
156 __LIBC_HIDDEN__ void __init_tcb(bionic_tcb* tcb, pthread_internal_t* thread);
157 __LIBC_HIDDEN__ void __init_tcb_stack_guard(bionic_tcb* tcb);
158 __LIBC_HIDDEN__ void __init_tcb_dtv(bionic_tcb* tcb);
159 __LIBC_HIDDEN__ void __init_bionic_tls_ptrs(bionic_tcb* tcb, bionic_tls* tls);
160 __LIBC_HIDDEN__ bionic_tls* __allocate_temp_bionic_tls();
161 __LIBC_HIDDEN__ void __free_temp_bionic_tls(bionic_tls* tls);
162 __LIBC_HIDDEN__ void __init_additional_stacks(pthread_internal_t*);
163 __LIBC_HIDDEN__ int __init_thread(pthread_internal_t* thread);
164 __LIBC_HIDDEN__ ThreadMapping __allocate_thread_mapping(size_t stack_size, size_t stack_guard_size);
165
166 __LIBC_HIDDEN__ pthread_t __pthread_internal_add(pthread_internal_t* thread);
167 __LIBC_HIDDEN__ pthread_internal_t* __pthread_internal_find(pthread_t pthread_id, const char* caller);
168 __LIBC_HIDDEN__ pid_t __pthread_internal_gettid(pthread_t pthread_id, const char* caller);
169 __LIBC_HIDDEN__ void __pthread_internal_remove(pthread_internal_t* thread);
170 __LIBC_HIDDEN__ void __pthread_internal_remove_and_free(pthread_internal_t* thread);
171
__get_bionic_tcb()172 static inline __always_inline bionic_tcb* __get_bionic_tcb() {
173 return reinterpret_cast<bionic_tcb*>(&__get_tls()[MIN_TLS_SLOT]);
174 }
175
176 // Make __get_thread() inlined for performance reason. See http://b/19825434.
__get_thread()177 static inline __always_inline pthread_internal_t* __get_thread() {
178 return static_cast<pthread_internal_t*>(__get_tls()[TLS_SLOT_THREAD_ID]);
179 }
180
__get_bionic_tls()181 static inline __always_inline bionic_tls& __get_bionic_tls() {
182 return *static_cast<bionic_tls*>(__get_tls()[TLS_SLOT_BIONIC_TLS]);
183 }
184
__get_tcb_dtv(bionic_tcb * tcb)185 static inline __always_inline TlsDtv* __get_tcb_dtv(bionic_tcb* tcb) {
186 uintptr_t dtv_slot = reinterpret_cast<uintptr_t>(tcb->tls_slot(TLS_SLOT_DTV));
187 return reinterpret_cast<TlsDtv*>(dtv_slot - offsetof(TlsDtv, generation));
188 }
189
__set_tcb_dtv(bionic_tcb * tcb,TlsDtv * val)190 static inline void __set_tcb_dtv(bionic_tcb* tcb, TlsDtv* val) {
191 tcb->tls_slot(TLS_SLOT_DTV) = &val->generation;
192 }
193
194 extern "C" __LIBC_HIDDEN__ int __set_tls(void* ptr);
195
196 __LIBC_HIDDEN__ void pthread_key_clean_all(void);
197
198 // Address space is precious on LP32, so use the minimum unit: one page.
199 // On LP64, we could use more but there's no obvious advantage to doing
200 // so, and the various media processes use RLIMIT_AS as a way to limit
201 // the amount of allocation they'll do.
202 #define PTHREAD_GUARD_SIZE PAGE_SIZE
203
204 // SIGSTKSZ (8KiB) is not big enough.
205 // An snprintf to a stack buffer of size PATH_MAX consumes ~7KiB of stack.
206 // On 64-bit, logging uses more than 8KiB by itself, ucontext is comically
207 // large on aarch64, and we have effectively infinite address space, so double
208 // the signal stack size.
209 #if defined(__LP64__)
210 #define SIGNAL_STACK_SIZE_WITHOUT_GUARD (32 * 1024)
211 #else
212 #define SIGNAL_STACK_SIZE_WITHOUT_GUARD (16 * 1024)
213 #endif
214
215 // Traditionally we gave threads a 1MiB stack. When we started
216 // allocating per-thread alternate signal stacks to ease debugging of
217 // stack overflows, we subtracted the same amount we were using there
218 // from the default thread stack size. This should keep memory usage
219 // roughly constant.
220 #define PTHREAD_STACK_SIZE_DEFAULT ((1 * 1024 * 1024) - SIGNAL_STACK_SIZE_WITHOUT_GUARD)
221
222 // Leave room for a guard page in the internally created signal stacks.
223 #define SIGNAL_STACK_SIZE (SIGNAL_STACK_SIZE_WITHOUT_GUARD + PTHREAD_GUARD_SIZE)
224
225 // Needed by fork.
226 __LIBC_HIDDEN__ extern void __bionic_atfork_run_prepare();
227 __LIBC_HIDDEN__ extern void __bionic_atfork_run_child();
228 __LIBC_HIDDEN__ extern void __bionic_atfork_run_parent();
229