Commit | Line | Data |
---|---|---|
0d1fb0a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
995473ae | 2 | /* |
2eb5f31b AI |
3 | * Copyright (C) 2015 Anton Ivanov (aivanov@{brocade.com,kot-begemot.co.uk}) |
4 | * Copyright (C) 2015 Thomas Meyer (thomas@m3y3r.de) | |
ba180fd4 | 5 | * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) |
1da177e4 | 6 | * Copyright 2003 PathScale, Inc. |
1da177e4 LT |
7 | */ |
8 | ||
c5d4bb17 JD |
9 | #include <linux/stddef.h> |
10 | #include <linux/err.h> | |
11 | #include <linux/hardirq.h> | |
c5d4bb17 | 12 | #include <linux/mm.h> |
6613c5e8 | 13 | #include <linux/module.h> |
c5d4bb17 JD |
14 | #include <linux/personality.h> |
15 | #include <linux/proc_fs.h> | |
16 | #include <linux/ptrace.h> | |
17 | #include <linux/random.h> | |
5a0e3ad6 | 18 | #include <linux/slab.h> |
c5d4bb17 | 19 | #include <linux/sched.h> |
b17b0153 | 20 | #include <linux/sched/debug.h> |
29930025 | 21 | #include <linux/sched/task.h> |
68db0cf1 | 22 | #include <linux/sched/task_stack.h> |
6613c5e8 | 23 | #include <linux/seq_file.h> |
c5d4bb17 JD |
24 | #include <linux/tick.h> |
25 | #include <linux/threads.h> | |
03248add | 26 | #include <linux/resume_user_mode.h> |
c5d4bb17 | 27 | #include <asm/current.h> |
445c5786 | 28 | #include <asm/mmu_context.h> |
7c0f6ba6 | 29 | #include <linux/uaccess.h> |
37185b33 AV |
30 | #include <as-layout.h> |
31 | #include <kern_util.h> | |
32 | #include <os.h> | |
33 | #include <skas.h> | |
dbba7f70 | 34 | #include <registers.h> |
f185063b | 35 | #include <linux/time-internal.h> |
1da177e4 | 36 | |
ba180fd4 JD |
37 | /* |
38 | * This is a per-cpu array. A processor only modifies its entry and it only | |
1da177e4 LT |
39 | * cares about its entry, so it's OK if another processor is modifying its |
40 | * entry. | |
41 | */ | |
42 | struct cpu_task cpu_tasks[NR_CPUS] = { [0 ... NR_CPUS - 1] = { -1, NULL } }; | |
43 | ||
2dc5802a | 44 | static inline int external_pid(void) |
1da177e4 | 45 | { |
77bf4400 | 46 | /* FIXME: Need to look up userspace_pid by cpu */ |
ba180fd4 | 47 | return userspace_pid[0]; |
1da177e4 LT |
48 | } |
49 | ||
50 | int pid_to_processor_id(int pid) | |
51 | { | |
52 | int i; | |
53 | ||
c5d4bb17 | 54 | for (i = 0; i < ncpus; i++) { |
ba180fd4 | 55 | if (cpu_tasks[i].pid == pid) |
6e21aec3 | 56 | return i; |
1da177e4 | 57 | } |
6e21aec3 | 58 | return -1; |
1da177e4 LT |
59 | } |
60 | ||
61 | void free_stack(unsigned long stack, int order) | |
62 | { | |
63 | free_pages(stack, order); | |
64 | } | |
65 | ||
2fcb4090 | 66 | unsigned long alloc_stack(int order, int atomic) |
1da177e4 | 67 | { |
2fcb4090 | 68 | unsigned long page; |
53f9fc93 | 69 | gfp_t flags = GFP_KERNEL; |
1da177e4 | 70 | |
46db4a42 PBG |
71 | if (atomic) |
72 | flags = GFP_ATOMIC; | |
2fcb4090 | 73 | page = __get_free_pages(flags, order); |
5c8aacea | 74 | |
2fcb4090 | 75 | return page; |
1da177e4 LT |
76 | } |
77 | ||
6e21aec3 | 78 | static inline void set_current(struct task_struct *task) |
1da177e4 | 79 | { |
ca9bc0bb | 80 | cpu_tasks[task_thread_info(task)->cpu] = ((struct cpu_task) |
2dc5802a | 81 | { external_pid(), task }); |
1da177e4 LT |
82 | } |
83 | ||
291248fd | 84 | extern void arch_switch_to(struct task_struct *to); |
77bf4400 | 85 | |
76b278ed | 86 | void *__switch_to(struct task_struct *from, struct task_struct *to) |
1da177e4 | 87 | { |
995473ae JD |
88 | to->thread.prev_sched = from; |
89 | set_current(to); | |
f6e34c6a | 90 | |
a1850e9c RW |
91 | switch_threads(&from->thread.switch_buf, &to->thread.switch_buf); |
92 | arch_switch_to(current); | |
f6e34c6a | 93 | |
6e21aec3 | 94 | return current->thread.prev_sched; |
1da177e4 LT |
95 | } |
96 | ||
97 | void interrupt_end(void) | |
98 | { | |
ccaee5f8 IM |
99 | struct pt_regs *regs = ¤t->thread.regs; |
100 | ||
ba180fd4 | 101 | if (need_resched()) |
6e21aec3 | 102 | schedule(); |
09041c92 JA |
103 | if (test_thread_flag(TIF_SIGPENDING) || |
104 | test_thread_flag(TIF_NOTIFY_SIGNAL)) | |
ccaee5f8 | 105 | do_signal(regs); |
3c532798 | 106 | if (test_thread_flag(TIF_NOTIFY_RESUME)) |
03248add | 107 | resume_user_mode_work(regs); |
1da177e4 LT |
108 | } |
109 | ||
c2220b2a | 110 | int get_current_pid(void) |
1da177e4 | 111 | { |
c2220b2a | 112 | return task_pid_nr(current); |
1da177e4 LT |
113 | } |
114 | ||
ba180fd4 JD |
115 | /* |
116 | * This is called magically, by its address being stuffed in a jmp_buf | |
77bf4400 JD |
117 | * and being longjmp-d to. |
118 | */ | |
119 | void new_thread_handler(void) | |
120 | { | |
121 | int (*fn)(void *), n; | |
122 | void *arg; | |
123 | ||
ba180fd4 | 124 | if (current->thread.prev_sched != NULL) |
77bf4400 JD |
125 | schedule_tail(current->thread.prev_sched); |
126 | current->thread.prev_sched = NULL; | |
127 | ||
128 | fn = current->thread.request.u.thread.proc; | |
129 | arg = current->thread.request.u.thread.arg; | |
130 | ||
ba180fd4 | 131 | /* |
22e2430d | 132 | * callback returns only if the kernel thread execs a process |
77bf4400 | 133 | */ |
22e2430d | 134 | n = fn(arg); |
6f602afd | 135 | userspace(¤t->thread.regs.regs, current_thread_info()->aux_fp_regs); |
77bf4400 JD |
136 | } |
137 | ||
138 | /* Called magically, see new_thread_handler above */ | |
139 | void fork_handler(void) | |
140 | { | |
141 | force_flush_all(); | |
77bf4400 JD |
142 | |
143 | schedule_tail(current->thread.prev_sched); | |
144 | ||
ba180fd4 JD |
145 | /* |
146 | * XXX: if interrupt_end() calls schedule, this call to | |
77bf4400 | 147 | * arch_switch_to isn't needed. We could want to apply this to |
ba180fd4 JD |
148 | * improve performance. -bb |
149 | */ | |
291248fd | 150 | arch_switch_to(current); |
77bf4400 JD |
151 | |
152 | current->thread.prev_sched = NULL; | |
153 | ||
6f602afd | 154 | userspace(¤t->thread.regs.regs, current_thread_info()->aux_fp_regs); |
77bf4400 JD |
155 | } |
156 | ||
c5febea0 | 157 | int copy_thread(struct task_struct * p, const struct kernel_clone_args *args) |
1da177e4 | 158 | { |
c5febea0 EB |
159 | unsigned long clone_flags = args->flags; |
160 | unsigned long sp = args->stack; | |
c5febea0 | 161 | unsigned long tls = args->tls; |
77bf4400 JD |
162 | void (*handler)(void); |
163 | int ret = 0; | |
aa6758d4 | 164 | |
1da177e4 | 165 | p->thread = (struct thread_struct) INIT_THREAD; |
aa6758d4 | 166 | |
5bd2e97c | 167 | if (!args->fn) { |
2b067fc9 | 168 | memcpy(&p->thread.regs.regs, current_pt_regs(), |
77bf4400 | 169 | sizeof(p->thread.regs.regs)); |
a3170d2e | 170 | PT_REGS_SET_SYSCALL_RETURN(&p->thread.regs, 0); |
ba180fd4 | 171 | if (sp != 0) |
18baddda | 172 | REGS_SP(p->thread.regs.regs.gp) = sp; |
aa6758d4 | 173 | |
77bf4400 | 174 | handler = fork_handler; |
aa6758d4 | 175 | |
77bf4400 | 176 | arch_copy_thread(¤t->thread.arch, &p->thread.arch); |
d2ce4e92 | 177 | } else { |
fbfe9c84 | 178 | get_safe_registers(p->thread.regs.regs.gp, p->thread.regs.regs.fp); |
5bd2e97c EB |
179 | p->thread.request.u.thread.proc = args->fn; |
180 | p->thread.request.u.thread.arg = args->fn_arg; | |
77bf4400 JD |
181 | handler = new_thread_handler; |
182 | } | |
183 | ||
184 | new_thread(task_stack_page(p), &p->thread.switch_buf, handler); | |
185 | ||
5bd2e97c | 186 | if (!args->fn) { |
77bf4400 JD |
187 | clear_flushed_tls(p); |
188 | ||
189 | /* | |
190 | * Set a new TLS for the child thread? | |
191 | */ | |
192 | if (clone_flags & CLONE_SETTLS) | |
457677c7 | 193 | ret = arch_set_tls(p, tls); |
77bf4400 | 194 | } |
aa6758d4 | 195 | |
aa6758d4 | 196 | return ret; |
1da177e4 LT |
197 | } |
198 | ||
199 | void initial_thread_cb(void (*proc)(void *), void *arg) | |
200 | { | |
201 | int save_kmalloc_ok = kmalloc_ok; | |
202 | ||
203 | kmalloc_ok = 0; | |
6aa802ce | 204 | initial_thread_cb_skas(proc, arg); |
1da177e4 LT |
205 | kmalloc_ok = save_kmalloc_ok; |
206 | } | |
995473ae | 207 | |
a374b7cb | 208 | void um_idle_sleep(void) |
06503870 | 209 | { |
49da38a3 JB |
210 | if (time_travel_mode != TT_MODE_OFF) |
211 | time_travel_sleep(); | |
212 | else | |
213 | os_idle_sleep(); | |
06503870 JB |
214 | } |
215 | ||
8198c169 | 216 | void arch_cpu_idle(void) |
1da177e4 | 217 | { |
a5a678c8 | 218 | cpu_tasks[current_thread_info()->cpu].pid = os_getpid(); |
06503870 | 219 | um_idle_sleep(); |
58c644ba | 220 | raw_local_irq_enable(); |
1da177e4 LT |
221 | } |
222 | ||
b6316293 PBG |
223 | int __cant_sleep(void) { |
224 | return in_atomic() || irqs_disabled() || in_interrupt(); | |
225 | /* Is in_interrupt() really needed? */ | |
1da177e4 LT |
226 | } |
227 | ||
1da177e4 LT |
228 | int user_context(unsigned long sp) |
229 | { | |
230 | unsigned long stack; | |
231 | ||
232 | stack = sp & (PAGE_MASK << CONFIG_KERNEL_STACK_ORDER); | |
a5a678c8 | 233 | return stack != (unsigned long) current_thread_info(); |
1da177e4 LT |
234 | } |
235 | ||
1da177e4 LT |
236 | extern exitcall_t __uml_exitcall_begin, __uml_exitcall_end; |
237 | ||
238 | void do_uml_exitcalls(void) | |
239 | { | |
240 | exitcall_t *call; | |
241 | ||
242 | call = &__uml_exitcall_end; | |
243 | while (--call >= &__uml_exitcall_begin) | |
244 | (*call)(); | |
245 | } | |
246 | ||
c0a9290e | 247 | char *uml_strdup(const char *string) |
1da177e4 | 248 | { |
dfe52244 | 249 | return kstrdup(string, GFP_KERNEL); |
1da177e4 | 250 | } |
73395a00 | 251 | EXPORT_SYMBOL(uml_strdup); |
1da177e4 | 252 | |
1da177e4 LT |
253 | int copy_to_user_proc(void __user *to, void *from, int size) |
254 | { | |
6e21aec3 | 255 | return copy_to_user(to, from, size); |
1da177e4 LT |
256 | } |
257 | ||
258 | int copy_from_user_proc(void *to, void __user *from, int size) | |
259 | { | |
6e21aec3 | 260 | return copy_from_user(to, from, size); |
1da177e4 LT |
261 | } |
262 | ||
263 | int clear_user_proc(void __user *buf, int size) | |
264 | { | |
6e21aec3 | 265 | return clear_user(buf, size); |
1da177e4 LT |
266 | } |
267 | ||
1da177e4 LT |
268 | static atomic_t using_sysemu = ATOMIC_INIT(0); |
269 | int sysemu_supported; | |
270 | ||
271 | void set_using_sysemu(int value) | |
272 | { | |
273 | if (value > sysemu_supported) | |
274 | return; | |
275 | atomic_set(&using_sysemu, value); | |
276 | } | |
277 | ||
278 | int get_using_sysemu(void) | |
279 | { | |
280 | return atomic_read(&using_sysemu); | |
281 | } | |
282 | ||
6613c5e8 | 283 | static int sysemu_proc_show(struct seq_file *m, void *v) |
1da177e4 | 284 | { |
6613c5e8 AD |
285 | seq_printf(m, "%d\n", get_using_sysemu()); |
286 | return 0; | |
287 | } | |
1da177e4 | 288 | |
6613c5e8 AD |
289 | static int sysemu_proc_open(struct inode *inode, struct file *file) |
290 | { | |
291 | return single_open(file, sysemu_proc_show, NULL); | |
1da177e4 LT |
292 | } |
293 | ||
6613c5e8 AD |
294 | static ssize_t sysemu_proc_write(struct file *file, const char __user *buf, |
295 | size_t count, loff_t *pos) | |
1da177e4 LT |
296 | { |
297 | char tmp[2]; | |
298 | ||
299 | if (copy_from_user(tmp, buf, 1)) | |
300 | return -EFAULT; | |
301 | ||
302 | if (tmp[0] >= '0' && tmp[0] <= '2') | |
303 | set_using_sysemu(tmp[0] - '0'); | |
ba180fd4 JD |
304 | /* We use the first char, but pretend to write everything */ |
305 | return count; | |
1da177e4 LT |
306 | } |
307 | ||
97a32539 AD |
308 | static const struct proc_ops sysemu_proc_ops = { |
309 | .proc_open = sysemu_proc_open, | |
310 | .proc_read = seq_read, | |
311 | .proc_lseek = seq_lseek, | |
312 | .proc_release = single_release, | |
313 | .proc_write = sysemu_proc_write, | |
6613c5e8 AD |
314 | }; |
315 | ||
1da177e4 LT |
316 | int __init make_proc_sysemu(void) |
317 | { | |
318 | struct proc_dir_entry *ent; | |
319 | if (!sysemu_supported) | |
320 | return 0; | |
321 | ||
97a32539 | 322 | ent = proc_create("sysemu", 0600, NULL, &sysemu_proc_ops); |
1da177e4 LT |
323 | |
324 | if (ent == NULL) | |
325 | { | |
30f417c6 | 326 | printk(KERN_WARNING "Failed to register /proc/sysemu\n"); |
6e21aec3 | 327 | return 0; |
1da177e4 LT |
328 | } |
329 | ||
1da177e4 LT |
330 | return 0; |
331 | } | |
332 | ||
333 | late_initcall(make_proc_sysemu); | |
334 | ||
335 | int singlestepping(void * t) | |
336 | { | |
337 | struct task_struct *task = t ? t : current; | |
338 | ||
c200e4bb | 339 | if (!test_thread_flag(TIF_SINGLESTEP)) |
ba180fd4 | 340 | return 0; |
1da177e4 LT |
341 | |
342 | if (task->thread.singlestep_syscall) | |
ba180fd4 | 343 | return 1; |
1da177e4 LT |
344 | |
345 | return 2; | |
346 | } | |
347 | ||
b8bd0220 BS |
348 | /* |
349 | * Only x86 and x86_64 have an arch_align_stack(). | |
350 | * All other arches have "#define arch_align_stack(x) (x)" | |
cf7bc58f | 351 | * in their asm/exec.h |
b8bd0220 BS |
352 | * As this is included in UML from asm-um/system-generic.h, |
353 | * we can use it to behave as the subarch does. | |
354 | */ | |
355 | #ifndef arch_align_stack | |
1da177e4 LT |
356 | unsigned long arch_align_stack(unsigned long sp) |
357 | { | |
8f80e946 | 358 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
81895a65 | 359 | sp -= prandom_u32_max(8192); |
1da177e4 LT |
360 | return sp & ~0xf; |
361 | } | |
b8bd0220 | 362 | #endif |
c1127465 | 363 | |
42a20f86 | 364 | unsigned long __get_wchan(struct task_struct *p) |
c1127465 JD |
365 | { |
366 | unsigned long stack_page, sp, ip; | |
367 | bool seen_sched = 0; | |
368 | ||
c1127465 JD |
369 | stack_page = (unsigned long) task_stack_page(p); |
370 | /* Bail if the process has no kernel stack for some reason */ | |
371 | if (stack_page == 0) | |
372 | return 0; | |
373 | ||
374 | sp = p->thread.switch_buf->JB_SP; | |
375 | /* | |
376 | * Bail if the stack pointer is below the bottom of the kernel | |
377 | * stack for some reason | |
378 | */ | |
379 | if (sp < stack_page) | |
380 | return 0; | |
381 | ||
382 | while (sp < stack_page + THREAD_SIZE) { | |
383 | ip = *((unsigned long *) sp); | |
384 | if (in_sched_functions(ip)) | |
385 | /* Ignore everything until we're above the scheduler */ | |
386 | seen_sched = 1; | |
387 | else if (kernel_text_address(ip) && seen_sched) | |
388 | return ip; | |
389 | ||
390 | sp += sizeof(unsigned long); | |
391 | } | |
392 | ||
393 | return 0; | |
394 | } | |
8192ab42 JD |
395 | |
396 | int elf_core_copy_fpregs(struct task_struct *t, elf_fpregset_t *fpu) | |
397 | { | |
398 | int cpu = current_thread_info()->cpu; | |
399 | ||
a78ff111 | 400 | return save_i387_registers(userspace_pid[cpu], (unsigned long *) fpu); |
8192ab42 JD |
401 | } |
402 |