1 /*
2 * Based on arch/arm/kernel/process.c
3 *
4 * Original Copyright (C) 1995 Linus Torvalds
5 * Copyright (C) 1996-2000 Russell King - Converted to ARM.
6 * Copyright (C) 2012 ARM Ltd.
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program. If not, see <http://www.gnu.org/licenses/>.
19 */
20
21 #include <stdarg.h>
22
23 #include <linux/compat.h>
24 #include <linux/efi.h>
25 #include <linux/export.h>
26 #include <linux/sched.h>
27 #include <linux/kernel.h>
28 #include <linux/mm.h>
29 #include <linux/stddef.h>
30 #include <linux/unistd.h>
31 #include <linux/user.h>
32 #include <linux/delay.h>
33 #include <linux/reboot.h>
34 #include <linux/interrupt.h>
35 #include <linux/kallsyms.h>
36 #include <linux/init.h>
37 #include <linux/cpu.h>
38 #include <linux/elfcore.h>
39 #include <linux/pm.h>
40 #include <linux/tick.h>
41 #include <linux/utsname.h>
42 #include <linux/uaccess.h>
43 #include <linux/random.h>
44 #include <linux/hw_breakpoint.h>
45 #include <linux/personality.h>
46 #include <linux/notifier.h>
47 #include <trace/events/power.h>
48 #ifdef CONFIG_THREAD_INFO_IN_TASK
49 #include <linux/percpu.h>
50 #endif
51
52 #include <asm/alternative.h>
53 #include <asm/compat.h>
54 #include <asm/cacheflush.h>
55 #include <asm/exec.h>
56 #include <asm/fpsimd.h>
57 #include <asm/mmu_context.h>
58 #include <asm/processor.h>
59 #include <asm/stacktrace.h>
60
61 #ifdef CONFIG_CC_STACKPROTECTOR
62 #include <linux/stackprotector.h>
63 unsigned long __stack_chk_guard __read_mostly;
64 EXPORT_SYMBOL(__stack_chk_guard);
65 #endif
66
67 /*
68 * Function pointers to optional machine specific functions
69 */
70 void (*pm_power_off)(void);
71 EXPORT_SYMBOL_GPL(pm_power_off);
72
73 void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
74
75 /*
76 * This is our default idle handler.
77 */
arch_cpu_idle(void)78 void arch_cpu_idle(void)
79 {
80 /*
81 * This should do all the clock switching and wait for interrupt
82 * tricks
83 */
84 trace_cpu_idle_rcuidle(1, smp_processor_id());
85 cpu_do_idle();
86 local_irq_enable();
87 trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
88 }
89
90 #ifdef CONFIG_HOTPLUG_CPU
arch_cpu_idle_dead(void)91 void arch_cpu_idle_dead(void)
92 {
93 cpu_die();
94 }
95 #endif
96
97 /*
98 * Called by kexec, immediately prior to machine_kexec().
99 *
100 * This must completely disable all secondary CPUs; simply causing those CPUs
101 * to execute e.g. a RAM-based pin loop is not sufficient. This allows the
102 * kexec'd kernel to use any and all RAM as it sees fit, without having to
103 * avoid any code or data used by any SW CPU pin loop. The CPU hotplug
104 * functionality embodied in disable_nonboot_cpus() to achieve this.
105 */
machine_shutdown(void)106 void machine_shutdown(void)
107 {
108 disable_nonboot_cpus();
109 }
110
111 /*
112 * Halting simply requires that the secondary CPUs stop performing any
113 * activity (executing tasks, handling interrupts). smp_send_stop()
114 * achieves this.
115 */
machine_halt(void)116 void machine_halt(void)
117 {
118 local_irq_disable();
119 smp_send_stop();
120 while (1);
121 }
122
123 /*
124 * Power-off simply requires that the secondary CPUs stop performing any
125 * activity (executing tasks, handling interrupts). smp_send_stop()
126 * achieves this. When the system power is turned off, it will take all CPUs
127 * with it.
128 */
machine_power_off(void)129 void machine_power_off(void)
130 {
131 local_irq_disable();
132 smp_send_stop();
133 if (pm_power_off)
134 pm_power_off();
135 }
136
137 /*
138 * Restart requires that the secondary CPUs stop performing any activity
139 * while the primary CPU resets the system. Systems with multiple CPUs must
140 * provide a HW restart implementation, to ensure that all CPUs reset at once.
141 * This is required so that any code running after reset on the primary CPU
142 * doesn't have to co-ordinate with other CPUs to ensure they aren't still
143 * executing pre-reset code, and using RAM that the primary CPU's code wishes
144 * to use. Implementing such co-ordination would be essentially impossible.
145 */
machine_restart(char * cmd)146 void machine_restart(char *cmd)
147 {
148 /* Disable interrupts first */
149 local_irq_disable();
150 smp_send_stop();
151
152 /*
153 * UpdateCapsule() depends on the system being reset via
154 * ResetSystem().
155 */
156 if (efi_enabled(EFI_RUNTIME_SERVICES))
157 efi_reboot(reboot_mode, NULL);
158
159 /* Now call the architecture specific reboot code. */
160 if (arm_pm_restart)
161 arm_pm_restart(reboot_mode, cmd);
162 else
163 do_kernel_restart(cmd);
164
165 /*
166 * Whoops - the architecture was unable to reboot.
167 */
168 printk("Reboot failed -- System halted\n");
169 while (1);
170 }
171
172 /*
173 * dump a block of kernel memory from around the given address
174 */
show_data(unsigned long addr,int nbytes,const char * name)175 static void show_data(unsigned long addr, int nbytes, const char *name)
176 {
177 int i, j;
178 int nlines;
179 u32 *p;
180
181 /*
182 * don't attempt to dump non-kernel addresses or
183 * values that are probably just small negative numbers
184 */
185 if (addr < PAGE_OFFSET || addr > -256UL)
186 return;
187
188 printk("\n%s: %#lx:\n", name, addr);
189
190 /*
191 * round address down to a 32 bit boundary
192 * and always dump a multiple of 32 bytes
193 */
194 p = (u32 *)(addr & ~(sizeof(u32) - 1));
195 nbytes += (addr & (sizeof(u32) - 1));
196 nlines = (nbytes + 31) / 32;
197
198
199 for (i = 0; i < nlines; i++) {
200 /*
201 * just display low 16 bits of address to keep
202 * each line of the dump < 80 characters
203 */
204 printk("%04lx ", (unsigned long)p & 0xffff);
205 for (j = 0; j < 8; j++) {
206 u32 data;
207 if (probe_kernel_address(p, data)) {
208 printk(" ********");
209 } else {
210 printk(" %08x", data);
211 }
212 ++p;
213 }
214 printk("\n");
215 }
216 }
217
show_extra_register_data(struct pt_regs * regs,int nbytes)218 static void show_extra_register_data(struct pt_regs *regs, int nbytes)
219 {
220 mm_segment_t fs;
221 unsigned int i;
222
223 fs = get_fs();
224 set_fs(KERNEL_DS);
225 show_data(regs->pc - nbytes, nbytes * 2, "PC");
226 show_data(regs->regs[30] - nbytes, nbytes * 2, "LR");
227 show_data(regs->sp - nbytes, nbytes * 2, "SP");
228 for (i = 0; i < 30; i++) {
229 char name[4];
230 snprintf(name, sizeof(name), "X%u", i);
231 show_data(regs->regs[i] - nbytes, nbytes * 2, name);
232 }
233 set_fs(fs);
234 }
235
__show_regs(struct pt_regs * regs)236 void __show_regs(struct pt_regs *regs)
237 {
238 int i, top_reg;
239 u64 lr, sp;
240
241 if (compat_user_mode(regs)) {
242 lr = regs->compat_lr;
243 sp = regs->compat_sp;
244 top_reg = 12;
245 } else {
246 lr = regs->regs[30];
247 sp = regs->sp;
248 top_reg = 29;
249 }
250
251 show_regs_print_info(KERN_DEFAULT);
252 print_symbol("PC is at %s\n", instruction_pointer(regs));
253 print_symbol("LR is at %s\n", lr);
254 printk("pc : [<%016llx>] lr : [<%016llx>] pstate: %08llx\n",
255 regs->pc, lr, regs->pstate);
256 printk("sp : %016llx\n", sp);
257 for (i = top_reg; i >= 0; i--) {
258 printk("x%-2d: %016llx ", i, regs->regs[i]);
259 if (i % 2 == 0)
260 printk("\n");
261 }
262 if (!user_mode(regs))
263 show_extra_register_data(regs, 128);
264 printk("\n");
265 }
266
show_regs(struct pt_regs * regs)267 void show_regs(struct pt_regs * regs)
268 {
269 printk("\n");
270 __show_regs(regs);
271 }
272
tls_thread_flush(void)273 static void tls_thread_flush(void)
274 {
275 asm ("msr tpidr_el0, xzr");
276
277 if (is_compat_task()) {
278 current->thread.tp_value = 0;
279
280 /*
281 * We need to ensure ordering between the shadow state and the
282 * hardware state, so that we don't corrupt the hardware state
283 * with a stale shadow state during context switch.
284 */
285 barrier();
286 asm ("msr tpidrro_el0, xzr");
287 }
288 }
289
flush_thread(void)290 void flush_thread(void)
291 {
292 fpsimd_flush_thread();
293 tls_thread_flush();
294 flush_ptrace_hw_breakpoint(current);
295 }
296
release_thread(struct task_struct * dead_task)297 void release_thread(struct task_struct *dead_task)
298 {
299 }
300
arch_dup_task_struct(struct task_struct * dst,struct task_struct * src)301 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
302 {
303 if (current->mm)
304 fpsimd_preserve_current_state();
305 *dst = *src;
306 return 0;
307 }
308
309 asmlinkage void ret_from_fork(void) asm("ret_from_fork");
310
copy_thread(unsigned long clone_flags,unsigned long stack_start,unsigned long stk_sz,struct task_struct * p)311 int copy_thread(unsigned long clone_flags, unsigned long stack_start,
312 unsigned long stk_sz, struct task_struct *p)
313 {
314 struct pt_regs *childregs = task_pt_regs(p);
315
316 memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
317
318 /*
319 * In case p was allocated the same task_struct pointer as some
320 * other recently-exited task, make sure p is disassociated from
321 * any cpu that may have run that now-exited task recently.
322 * Otherwise we could erroneously skip reloading the FPSIMD
323 * registers for p.
324 */
325 fpsimd_flush_task_state(p);
326
327 if (likely(!(p->flags & PF_KTHREAD))) {
328 *childregs = *current_pt_regs();
329 childregs->regs[0] = 0;
330
331 /*
332 * Read the current TLS pointer from tpidr_el0 as it may be
333 * out-of-sync with the saved value.
334 */
335 asm("mrs %0, tpidr_el0" : "=r" (*task_user_tls(p)));
336
337 if (stack_start) {
338 if (is_compat_thread(task_thread_info(p)))
339 childregs->compat_sp = stack_start;
340 /* 16-byte aligned stack mandatory on AArch64 */
341 else if (stack_start & 15)
342 return -EINVAL;
343 else
344 childregs->sp = stack_start;
345 }
346
347 /*
348 * If a TLS pointer was passed to clone (4th argument), use it
349 * for the new thread.
350 */
351 if (clone_flags & CLONE_SETTLS)
352 p->thread.tp_value = childregs->regs[3];
353 } else {
354 memset(childregs, 0, sizeof(struct pt_regs));
355 childregs->pstate = PSR_MODE_EL1h;
356 if (IS_ENABLED(CONFIG_ARM64_UAO) &&
357 cpus_have_cap(ARM64_HAS_UAO))
358 childregs->pstate |= PSR_UAO_BIT;
359 p->thread.cpu_context.x19 = stack_start;
360 p->thread.cpu_context.x20 = stk_sz;
361 }
362 p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
363 p->thread.cpu_context.sp = (unsigned long)childregs;
364
365 ptrace_hw_copy_thread(p);
366
367 return 0;
368 }
369
tls_thread_switch(struct task_struct * next)370 static void tls_thread_switch(struct task_struct *next)
371 {
372 unsigned long tpidr;
373
374 asm("mrs %0, tpidr_el0" : "=r" (tpidr));
375 *task_user_tls(current) = tpidr;
376
377 if (is_compat_thread(task_thread_info(next)))
378 write_sysreg(next->thread.tp_value, tpidrro_el0);
379 else if (!arm64_kernel_unmapped_at_el0())
380 write_sysreg(0, tpidrro_el0);
381
382 write_sysreg(*task_user_tls(next), tpidr_el0);
383 }
384
385 /* Restore the UAO state depending on next's addr_limit */
uao_thread_switch(struct task_struct * next)386 void uao_thread_switch(struct task_struct *next)
387 {
388 if (IS_ENABLED(CONFIG_ARM64_UAO)) {
389 if (task_thread_info(next)->addr_limit == KERNEL_DS)
390 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
391 else
392 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
393 }
394 }
395
396 #ifdef CONFIG_THREAD_INFO_IN_TASK
397 /*
398 * We store our current task in sp_el0, which is clobbered by userspace. Keep a
399 * shadow copy so that we can restore this upon entry from userspace.
400 *
401 * This is *only* for exception entry from EL0, and is not valid until we
402 * __switch_to() a user task.
403 */
404 DEFINE_PER_CPU(struct task_struct *, __entry_task);
405
entry_task_switch(struct task_struct * next)406 static void entry_task_switch(struct task_struct *next)
407 {
408 __this_cpu_write(__entry_task, next);
409 }
410 #endif
411
412 /*
413 * Thread switching.
414 */
__switch_to(struct task_struct * prev,struct task_struct * next)415 struct task_struct *__switch_to(struct task_struct *prev,
416 struct task_struct *next)
417 {
418 struct task_struct *last;
419
420 fpsimd_thread_switch(next);
421 tls_thread_switch(next);
422 hw_breakpoint_thread_switch(next);
423 contextidr_thread_switch(next);
424 #ifdef CONFIG_THREAD_INFO_IN_TASK
425 entry_task_switch(next);
426 #endif
427 uao_thread_switch(next);
428
429 /*
430 * Complete any pending TLB or cache maintenance on this CPU in case
431 * the thread migrates to a different CPU.
432 */
433 dsb(ish);
434
435 /* the actual thread switch */
436 last = cpu_switch_to(prev, next);
437
438 return last;
439 }
440
get_wchan(struct task_struct * p)441 unsigned long get_wchan(struct task_struct *p)
442 {
443 struct stackframe frame;
444 unsigned long stack_page, ret = 0;
445 int count = 0;
446 if (!p || p == current || p->state == TASK_RUNNING)
447 return 0;
448
449 stack_page = (unsigned long)try_get_task_stack(p);
450 if (!stack_page)
451 return 0;
452
453 frame.fp = thread_saved_fp(p);
454 frame.sp = thread_saved_sp(p);
455 frame.pc = thread_saved_pc(p);
456 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
457 frame.graph = p->curr_ret_stack;
458 #endif
459 do {
460 if (frame.sp < stack_page ||
461 frame.sp >= stack_page + THREAD_SIZE ||
462 unwind_frame(p, &frame))
463 goto out;
464 if (!in_sched_functions(frame.pc)) {
465 ret = frame.pc;
466 goto out;
467 }
468 } while (count ++ < 16);
469
470 out:
471 put_task_stack(p);
472 return ret;
473 }
474
arch_align_stack(unsigned long sp)475 unsigned long arch_align_stack(unsigned long sp)
476 {
477 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
478 sp -= get_random_int() & ~PAGE_MASK;
479 return sp & ~0xf;
480 }
481
randomize_base(unsigned long base)482 static unsigned long randomize_base(unsigned long base)
483 {
484 unsigned long range_end = base + (STACK_RND_MASK << PAGE_SHIFT) + 1;
485 return randomize_range(base, range_end, 0) ? : base;
486 }
487
arch_randomize_brk(struct mm_struct * mm)488 unsigned long arch_randomize_brk(struct mm_struct *mm)
489 {
490 return randomize_base(mm->brk);
491 }
492