2 * Based on arch/arm/kernel/process.c
4 * Original Copyright (C) 1995 Linus Torvalds
5 * Copyright (C) 1996-2000 Russell King - Converted to ARM.
6 * Copyright (C) 2012 ARM Ltd.
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.
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.
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/>.
23 #include <linux/compat.h>
24 #include <linux/efi.h>
25 #include <linux/export.h>
26 #include <linux/sched.h>
27 #include <linux/sched/debug.h>
28 #include <linux/sched/task.h>
29 #include <linux/sched/task_stack.h>
30 #include <linux/kernel.h>
32 #include <linux/stddef.h>
33 #include <linux/unistd.h>
34 #include <linux/user.h>
35 #include <linux/delay.h>
36 #include <linux/reboot.h>
37 #include <linux/interrupt.h>
38 #include <linux/kallsyms.h>
39 #include <linux/init.h>
40 #include <linux/cpu.h>
41 #include <linux/elfcore.h>
43 #include <linux/tick.h>
44 #include <linux/utsname.h>
45 #include <linux/uaccess.h>
46 #include <linux/random.h>
47 #include <linux/hw_breakpoint.h>
48 #include <linux/personality.h>
49 #include <linux/notifier.h>
50 #include <trace/events/power.h>
51 #include <linux/percpu.h>
53 #include <asm/alternative.h>
54 #include <asm/compat.h>
55 #include <asm/cacheflush.h>
57 #include <asm/fpsimd.h>
58 #include <asm/mmu_context.h>
59 #include <asm/processor.h>
60 #include <asm/stacktrace.h>
62 #ifdef CONFIG_CC_STACKPROTECTOR
63 #include <linux/stackprotector.h>
64 unsigned long __stack_chk_guard __ro_after_init;
65 EXPORT_SYMBOL(__stack_chk_guard);
69 * Function pointers to optional machine specific functions
71 void (*pm_power_off)(void);
72 EXPORT_SYMBOL_GPL(pm_power_off);
74 void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
77 * This is our default idle handler.
79 void arch_cpu_idle(void)
82 * This should do all the clock switching and wait for interrupt
85 trace_cpu_idle_rcuidle(1, smp_processor_id());
88 trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
91 #ifdef CONFIG_HOTPLUG_CPU
92 void arch_cpu_idle_dead(void)
99 * Called by kexec, immediately prior to machine_kexec().
101 * This must completely disable all secondary CPUs; simply causing those CPUs
102 * to execute e.g. a RAM-based pin loop is not sufficient. This allows the
103 * kexec'd kernel to use any and all RAM as it sees fit, without having to
104 * avoid any code or data used by any SW CPU pin loop. The CPU hotplug
105 * functionality embodied in disable_nonboot_cpus() to achieve this.
107 void machine_shutdown(void)
109 disable_nonboot_cpus();
113 * Halting simply requires that the secondary CPUs stop performing any
114 * activity (executing tasks, handling interrupts). smp_send_stop()
117 void machine_halt(void)
125 * Power-off simply requires that the secondary CPUs stop performing any
126 * activity (executing tasks, handling interrupts). smp_send_stop()
127 * achieves this. When the system power is turned off, it will take all CPUs
130 void machine_power_off(void)
139 * Restart requires that the secondary CPUs stop performing any activity
140 * while the primary CPU resets the system. Systems with multiple CPUs must
141 * provide a HW restart implementation, to ensure that all CPUs reset at once.
142 * This is required so that any code running after reset on the primary CPU
143 * doesn't have to co-ordinate with other CPUs to ensure they aren't still
144 * executing pre-reset code, and using RAM that the primary CPU's code wishes
145 * to use. Implementing such co-ordination would be essentially impossible.
147 void machine_restart(char *cmd)
149 /* Disable interrupts first */
154 * UpdateCapsule() depends on the system being reset via
157 if (efi_enabled(EFI_RUNTIME_SERVICES))
158 efi_reboot(reboot_mode, NULL);
160 /* Now call the architecture specific reboot code. */
162 arm_pm_restart(reboot_mode, cmd);
164 do_kernel_restart(cmd);
167 * Whoops - the architecture was unable to reboot.
169 printk("Reboot failed -- System halted\n");
173 void __show_regs(struct pt_regs *regs)
178 if (compat_user_mode(regs)) {
179 lr = regs->compat_lr;
180 sp = regs->compat_sp;
188 show_regs_print_info(KERN_DEFAULT);
189 print_symbol("pc : %s\n", regs->pc);
190 print_symbol("lr : %s\n", lr);
191 printk("sp : %016llx pstate : %08llx\n", sp, regs->pstate);
196 printk("x%-2d: %016llx ", i, regs->regs[i]);
200 pr_cont("x%-2d: %016llx ", i, regs->regs[i]);
208 void show_regs(struct pt_regs * regs)
211 dump_backtrace(regs, NULL);
214 static void tls_thread_flush(void)
216 write_sysreg(0, tpidr_el0);
218 if (is_compat_task()) {
219 current->thread.tp_value = 0;
222 * We need to ensure ordering between the shadow state and the
223 * hardware state, so that we don't corrupt the hardware state
224 * with a stale shadow state during context switch.
227 write_sysreg(0, tpidrro_el0);
231 void flush_thread(void)
233 fpsimd_flush_thread();
235 flush_ptrace_hw_breakpoint(current);
238 void release_thread(struct task_struct *dead_task)
242 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
245 fpsimd_preserve_current_state();
250 asmlinkage void ret_from_fork(void) asm("ret_from_fork");
252 int copy_thread(unsigned long clone_flags, unsigned long stack_start,
253 unsigned long stk_sz, struct task_struct *p)
255 struct pt_regs *childregs = task_pt_regs(p);
257 memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
260 * In case p was allocated the same task_struct pointer as some
261 * other recently-exited task, make sure p is disassociated from
262 * any cpu that may have run that now-exited task recently.
263 * Otherwise we could erroneously skip reloading the FPSIMD
266 fpsimd_flush_task_state(p);
268 if (likely(!(p->flags & PF_KTHREAD))) {
269 *childregs = *current_pt_regs();
270 childregs->regs[0] = 0;
273 * Read the current TLS pointer from tpidr_el0 as it may be
274 * out-of-sync with the saved value.
276 *task_user_tls(p) = read_sysreg(tpidr_el0);
279 if (is_compat_thread(task_thread_info(p)))
280 childregs->compat_sp = stack_start;
282 childregs->sp = stack_start;
286 * If a TLS pointer was passed to clone (4th argument), use it
287 * for the new thread.
289 if (clone_flags & CLONE_SETTLS)
290 p->thread.tp_value = childregs->regs[3];
292 memset(childregs, 0, sizeof(struct pt_regs));
293 childregs->pstate = PSR_MODE_EL1h;
294 if (IS_ENABLED(CONFIG_ARM64_UAO) &&
295 cpus_have_const_cap(ARM64_HAS_UAO))
296 childregs->pstate |= PSR_UAO_BIT;
298 if (arm64_get_ssbd_state() == ARM64_SSBD_FORCE_DISABLE)
299 set_ssbs_bit(childregs);
301 p->thread.cpu_context.x19 = stack_start;
302 p->thread.cpu_context.x20 = stk_sz;
304 p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
305 p->thread.cpu_context.sp = (unsigned long)childregs;
307 ptrace_hw_copy_thread(p);
312 void tls_preserve_current_state(void)
314 *task_user_tls(current) = read_sysreg(tpidr_el0);
317 static void tls_thread_switch(struct task_struct *next)
319 tls_preserve_current_state();
321 if (is_compat_thread(task_thread_info(next)))
322 write_sysreg(next->thread.tp_value, tpidrro_el0);
323 else if (!arm64_kernel_unmapped_at_el0())
324 write_sysreg(0, tpidrro_el0);
326 write_sysreg(*task_user_tls(next), tpidr_el0);
329 /* Restore the UAO state depending on next's addr_limit */
330 void uao_thread_switch(struct task_struct *next)
332 if (IS_ENABLED(CONFIG_ARM64_UAO)) {
333 if (task_thread_info(next)->addr_limit == KERNEL_DS)
334 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
336 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
341 * Force SSBS state on context-switch, since it may be lost after migrating
342 * from a CPU which treats the bit as RES0 in a heterogeneous system.
344 static void ssbs_thread_switch(struct task_struct *next)
346 struct pt_regs *regs = task_pt_regs(next);
349 * Nothing to do for kernel threads, but 'regs' may be junk
350 * (e.g. idle task) so check the flags and bail early.
352 if (unlikely(next->flags & PF_KTHREAD))
356 * If all CPUs implement the SSBS extension, then we just need to
357 * context-switch the PSTATE field.
359 if (cpu_have_feature(cpu_feature(SSBS)))
362 /* If the mitigation is enabled, then we leave SSBS clear. */
363 if ((arm64_get_ssbd_state() == ARM64_SSBD_FORCE_ENABLE) ||
364 test_tsk_thread_flag(next, TIF_SSBD))
367 if (compat_user_mode(regs))
368 set_compat_ssbs_bit(regs);
369 else if (user_mode(regs))
374 * We store our current task in sp_el0, which is clobbered by userspace. Keep a
375 * shadow copy so that we can restore this upon entry from userspace.
377 * This is *only* for exception entry from EL0, and is not valid until we
378 * __switch_to() a user task.
380 DEFINE_PER_CPU(struct task_struct *, __entry_task);
382 static void entry_task_switch(struct task_struct *next)
384 __this_cpu_write(__entry_task, next);
390 __notrace_funcgraph struct task_struct *__switch_to(struct task_struct *prev,
391 struct task_struct *next)
393 struct task_struct *last;
395 fpsimd_thread_switch(next);
396 tls_thread_switch(next);
397 hw_breakpoint_thread_switch(next);
398 contextidr_thread_switch(next);
399 entry_task_switch(next);
400 uao_thread_switch(next);
401 ssbs_thread_switch(next);
404 * Complete any pending TLB or cache maintenance on this CPU in case
405 * the thread migrates to a different CPU.
406 * This full barrier is also required by the membarrier system
411 /* the actual thread switch */
412 last = cpu_switch_to(prev, next);
417 unsigned long get_wchan(struct task_struct *p)
419 struct stackframe frame;
420 unsigned long stack_page, ret = 0;
422 if (!p || p == current || p->state == TASK_RUNNING)
425 stack_page = (unsigned long)try_get_task_stack(p);
429 frame.fp = thread_saved_fp(p);
430 frame.pc = thread_saved_pc(p);
431 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
432 frame.graph = p->curr_ret_stack;
435 if (unwind_frame(p, &frame))
437 if (!in_sched_functions(frame.pc)) {
441 } while (count ++ < 16);
448 unsigned long arch_align_stack(unsigned long sp)
450 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
451 sp -= get_random_int() & ~PAGE_MASK;
455 unsigned long arch_randomize_brk(struct mm_struct *mm)
457 if (is_compat_task())
458 return randomize_page(mm->brk, SZ_32M);
460 return randomize_page(mm->brk, SZ_1G);
464 * Called from setup_new_exec() after (COMPAT_)SET_PERSONALITY.
466 void arch_setup_new_exec(void)
468 current->mm->context.flags = is_compat_task() ? MMCF_AARCH32 : 0;