1 // SPDX-License-Identifier: GPL-2.0-only
3 * Based on arch/arm/kernel/process.c
5 * Original Copyright (C) 1995 Linus Torvalds
6 * Copyright (C) 1996-2000 Russell King - Converted to ARM.
7 * Copyright (C) 2012 ARM Ltd.
12 #include <linux/compat.h>
13 #include <linux/efi.h>
14 #include <linux/export.h>
15 #include <linux/sched.h>
16 #include <linux/sched/debug.h>
17 #include <linux/sched/task.h>
18 #include <linux/sched/task_stack.h>
19 #include <linux/kernel.h>
20 #include <linux/lockdep.h>
22 #include <linux/stddef.h>
23 #include <linux/sysctl.h>
24 #include <linux/unistd.h>
25 #include <linux/user.h>
26 #include <linux/delay.h>
27 #include <linux/reboot.h>
28 #include <linux/interrupt.h>
29 #include <linux/init.h>
30 #include <linux/cpu.h>
31 #include <linux/elfcore.h>
33 #include <linux/tick.h>
34 #include <linux/utsname.h>
35 #include <linux/uaccess.h>
36 #include <linux/random.h>
37 #include <linux/hw_breakpoint.h>
38 #include <linux/personality.h>
39 #include <linux/notifier.h>
40 #include <trace/events/power.h>
41 #include <linux/percpu.h>
42 #include <linux/thread_info.h>
43 #include <linux/prctl.h>
45 #include <asm/alternative.h>
46 #include <asm/arch_gicv3.h>
47 #include <asm/compat.h>
48 #include <asm/cpufeature.h>
49 #include <asm/cacheflush.h>
51 #include <asm/fpsimd.h>
52 #include <asm/mmu_context.h>
53 #include <asm/processor.h>
54 #include <asm/pointer_auth.h>
55 #include <asm/stacktrace.h>
57 #if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_STACKPROTECTOR_PER_TASK)
58 #include <linux/stackprotector.h>
59 unsigned long __stack_chk_guard __read_mostly;
60 EXPORT_SYMBOL(__stack_chk_guard);
64 * Function pointers to optional machine specific functions
66 void (*pm_power_off)(void);
67 EXPORT_SYMBOL_GPL(pm_power_off);
69 void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
71 static void __cpu_do_idle(void)
77 static void __cpu_do_idle_irqprio(void)
80 unsigned long daif_bits;
82 daif_bits = read_sysreg(daif);
83 write_sysreg(daif_bits | PSR_I_BIT, daif);
86 * Unmask PMR before going idle to make sure interrupts can
90 gic_write_pmr(GIC_PRIO_IRQON | GIC_PRIO_PSR_I_SET);
95 write_sysreg(daif_bits, daif);
101 * Idle the processor (wait for interrupt).
103 * If the CPU supports priority masking we must do additional work to
104 * ensure that interrupts are not masked at the PMR (because the core will
105 * not wake up if we block the wake up signal in the interrupt controller).
107 void cpu_do_idle(void)
109 if (system_uses_irq_prio_masking())
110 __cpu_do_idle_irqprio();
116 * This is our default idle handler.
118 void arch_cpu_idle(void)
121 * This should do all the clock switching and wait for interrupt
124 trace_cpu_idle_rcuidle(1, smp_processor_id());
127 trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
130 #ifdef CONFIG_HOTPLUG_CPU
131 void arch_cpu_idle_dead(void)
138 * Called by kexec, immediately prior to machine_kexec().
140 * This must completely disable all secondary CPUs; simply causing those CPUs
141 * to execute e.g. a RAM-based pin loop is not sufficient. This allows the
142 * kexec'd kernel to use any and all RAM as it sees fit, without having to
143 * avoid any code or data used by any SW CPU pin loop. The CPU hotplug
144 * functionality embodied in smpt_shutdown_nonboot_cpus() to achieve this.
146 void machine_shutdown(void)
148 smp_shutdown_nonboot_cpus(reboot_cpu);
152 * Halting simply requires that the secondary CPUs stop performing any
153 * activity (executing tasks, handling interrupts). smp_send_stop()
156 void machine_halt(void)
164 * Power-off simply requires that the secondary CPUs stop performing any
165 * activity (executing tasks, handling interrupts). smp_send_stop()
166 * achieves this. When the system power is turned off, it will take all CPUs
169 void machine_power_off(void)
178 * Restart requires that the secondary CPUs stop performing any activity
179 * while the primary CPU resets the system. Systems with multiple CPUs must
180 * provide a HW restart implementation, to ensure that all CPUs reset at once.
181 * This is required so that any code running after reset on the primary CPU
182 * doesn't have to co-ordinate with other CPUs to ensure they aren't still
183 * executing pre-reset code, and using RAM that the primary CPU's code wishes
184 * to use. Implementing such co-ordination would be essentially impossible.
186 void machine_restart(char *cmd)
188 /* Disable interrupts first */
193 * UpdateCapsule() depends on the system being reset via
196 if (efi_enabled(EFI_RUNTIME_SERVICES))
197 efi_reboot(reboot_mode, NULL);
199 /* Now call the architecture specific reboot code. */
201 arm_pm_restart(reboot_mode, cmd);
203 do_kernel_restart(cmd);
206 * Whoops - the architecture was unable to reboot.
208 printk("Reboot failed -- System halted\n");
212 static void print_pstate(struct pt_regs *regs)
214 u64 pstate = regs->pstate;
216 if (compat_user_mode(regs)) {
217 printk("pstate: %08llx (%c%c%c%c %c %s %s %c%c%c)\n",
219 pstate & PSR_AA32_N_BIT ? 'N' : 'n',
220 pstate & PSR_AA32_Z_BIT ? 'Z' : 'z',
221 pstate & PSR_AA32_C_BIT ? 'C' : 'c',
222 pstate & PSR_AA32_V_BIT ? 'V' : 'v',
223 pstate & PSR_AA32_Q_BIT ? 'Q' : 'q',
224 pstate & PSR_AA32_T_BIT ? "T32" : "A32",
225 pstate & PSR_AA32_E_BIT ? "BE" : "LE",
226 pstate & PSR_AA32_A_BIT ? 'A' : 'a',
227 pstate & PSR_AA32_I_BIT ? 'I' : 'i',
228 pstate & PSR_AA32_F_BIT ? 'F' : 'f');
230 printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n",
232 pstate & PSR_N_BIT ? 'N' : 'n',
233 pstate & PSR_Z_BIT ? 'Z' : 'z',
234 pstate & PSR_C_BIT ? 'C' : 'c',
235 pstate & PSR_V_BIT ? 'V' : 'v',
236 pstate & PSR_D_BIT ? 'D' : 'd',
237 pstate & PSR_A_BIT ? 'A' : 'a',
238 pstate & PSR_I_BIT ? 'I' : 'i',
239 pstate & PSR_F_BIT ? 'F' : 'f',
240 pstate & PSR_PAN_BIT ? '+' : '-',
241 pstate & PSR_UAO_BIT ? '+' : '-');
245 void __show_regs(struct pt_regs *regs)
250 if (compat_user_mode(regs)) {
251 lr = regs->compat_lr;
252 sp = regs->compat_sp;
260 show_regs_print_info(KERN_DEFAULT);
263 if (!user_mode(regs)) {
264 printk("pc : %pS\n", (void *)regs->pc);
265 printk("lr : %pS\n", (void *)ptrauth_strip_insn_pac(lr));
267 printk("pc : %016llx\n", regs->pc);
268 printk("lr : %016llx\n", lr);
271 printk("sp : %016llx\n", sp);
273 if (system_uses_irq_prio_masking())
274 printk("pmr_save: %08llx\n", regs->pmr_save);
279 printk("x%-2d: %016llx ", i, regs->regs[i]);
283 pr_cont("x%-2d: %016llx ", i, regs->regs[i]);
291 void show_regs(struct pt_regs * regs)
294 dump_backtrace(regs, NULL);
297 static void tls_thread_flush(void)
299 write_sysreg(0, tpidr_el0);
301 if (is_compat_task()) {
302 current->thread.uw.tp_value = 0;
305 * We need to ensure ordering between the shadow state and the
306 * hardware state, so that we don't corrupt the hardware state
307 * with a stale shadow state during context switch.
310 write_sysreg(0, tpidrro_el0);
314 static void flush_tagged_addr_state(void)
316 if (IS_ENABLED(CONFIG_ARM64_TAGGED_ADDR_ABI))
317 clear_thread_flag(TIF_TAGGED_ADDR);
320 void flush_thread(void)
322 fpsimd_flush_thread();
324 flush_ptrace_hw_breakpoint(current);
325 flush_tagged_addr_state();
328 void release_thread(struct task_struct *dead_task)
332 void arch_release_task_struct(struct task_struct *tsk)
334 fpsimd_release_task(tsk);
337 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
340 fpsimd_preserve_current_state();
343 /* We rely on the above assignment to initialize dst's thread_flags: */
344 BUILD_BUG_ON(!IS_ENABLED(CONFIG_THREAD_INFO_IN_TASK));
347 * Detach src's sve_state (if any) from dst so that it does not
348 * get erroneously used or freed prematurely. dst's sve_state
349 * will be allocated on demand later on if dst uses SVE.
350 * For consistency, also clear TIF_SVE here: this could be done
351 * later in copy_process(), but to avoid tripping up future
352 * maintainers it is best not to leave TIF_SVE and sve_state in
353 * an inconsistent state, even temporarily.
355 dst->thread.sve_state = NULL;
356 clear_tsk_thread_flag(dst, TIF_SVE);
361 asmlinkage void ret_from_fork(void) asm("ret_from_fork");
363 int copy_thread_tls(unsigned long clone_flags, unsigned long stack_start,
364 unsigned long stk_sz, struct task_struct *p, unsigned long tls)
366 struct pt_regs *childregs = task_pt_regs(p);
368 memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
371 * In case p was allocated the same task_struct pointer as some
372 * other recently-exited task, make sure p is disassociated from
373 * any cpu that may have run that now-exited task recently.
374 * Otherwise we could erroneously skip reloading the FPSIMD
377 fpsimd_flush_task_state(p);
379 ptrauth_thread_init_kernel(p);
381 if (likely(!(p->flags & PF_KTHREAD))) {
382 *childregs = *current_pt_regs();
383 childregs->regs[0] = 0;
386 * Read the current TLS pointer from tpidr_el0 as it may be
387 * out-of-sync with the saved value.
389 *task_user_tls(p) = read_sysreg(tpidr_el0);
392 if (is_compat_thread(task_thread_info(p)))
393 childregs->compat_sp = stack_start;
395 childregs->sp = stack_start;
399 * If a TLS pointer was passed to clone, use it for the new
402 if (clone_flags & CLONE_SETTLS)
403 p->thread.uw.tp_value = tls;
405 memset(childregs, 0, sizeof(struct pt_regs));
406 childregs->pstate = PSR_MODE_EL1h;
407 if (IS_ENABLED(CONFIG_ARM64_UAO) &&
408 cpus_have_const_cap(ARM64_HAS_UAO))
409 childregs->pstate |= PSR_UAO_BIT;
411 if (arm64_get_ssbd_state() == ARM64_SSBD_FORCE_DISABLE)
412 set_ssbs_bit(childregs);
414 if (system_uses_irq_prio_masking())
415 childregs->pmr_save = GIC_PRIO_IRQON;
417 p->thread.cpu_context.x19 = stack_start;
418 p->thread.cpu_context.x20 = stk_sz;
420 p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
421 p->thread.cpu_context.sp = (unsigned long)childregs;
423 ptrace_hw_copy_thread(p);
428 void tls_preserve_current_state(void)
430 *task_user_tls(current) = read_sysreg(tpidr_el0);
433 static void tls_thread_switch(struct task_struct *next)
435 tls_preserve_current_state();
437 if (is_compat_thread(task_thread_info(next)))
438 write_sysreg(next->thread.uw.tp_value, tpidrro_el0);
439 else if (!arm64_kernel_unmapped_at_el0())
440 write_sysreg(0, tpidrro_el0);
442 write_sysreg(*task_user_tls(next), tpidr_el0);
445 /* Restore the UAO state depending on next's addr_limit */
446 void uao_thread_switch(struct task_struct *next)
448 if (IS_ENABLED(CONFIG_ARM64_UAO)) {
449 if (task_thread_info(next)->addr_limit == KERNEL_DS)
450 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
452 asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
457 * Force SSBS state on context-switch, since it may be lost after migrating
458 * from a CPU which treats the bit as RES0 in a heterogeneous system.
460 static void ssbs_thread_switch(struct task_struct *next)
462 struct pt_regs *regs = task_pt_regs(next);
465 * Nothing to do for kernel threads, but 'regs' may be junk
466 * (e.g. idle task) so check the flags and bail early.
468 if (unlikely(next->flags & PF_KTHREAD))
472 * If all CPUs implement the SSBS extension, then we just need to
473 * context-switch the PSTATE field.
475 if (cpu_have_feature(cpu_feature(SSBS)))
478 /* If the mitigation is enabled, then we leave SSBS clear. */
479 if ((arm64_get_ssbd_state() == ARM64_SSBD_FORCE_ENABLE) ||
480 test_tsk_thread_flag(next, TIF_SSBD))
483 if (compat_user_mode(regs))
484 set_compat_ssbs_bit(regs);
485 else if (user_mode(regs))
490 * We store our current task in sp_el0, which is clobbered by userspace. Keep a
491 * shadow copy so that we can restore this upon entry from userspace.
493 * This is *only* for exception entry from EL0, and is not valid until we
494 * __switch_to() a user task.
496 DEFINE_PER_CPU(struct task_struct *, __entry_task);
498 static void entry_task_switch(struct task_struct *next)
500 __this_cpu_write(__entry_task, next);
506 __notrace_funcgraph struct task_struct *__switch_to(struct task_struct *prev,
507 struct task_struct *next)
509 struct task_struct *last;
511 fpsimd_thread_switch(next);
512 tls_thread_switch(next);
513 hw_breakpoint_thread_switch(next);
514 contextidr_thread_switch(next);
515 entry_task_switch(next);
516 uao_thread_switch(next);
517 ssbs_thread_switch(next);
520 * Complete any pending TLB or cache maintenance on this CPU in case
521 * the thread migrates to a different CPU.
522 * This full barrier is also required by the membarrier system
527 /* the actual thread switch */
528 last = cpu_switch_to(prev, next);
533 unsigned long get_wchan(struct task_struct *p)
535 struct stackframe frame;
536 unsigned long stack_page, ret = 0;
538 if (!p || p == current || p->state == TASK_RUNNING)
541 stack_page = (unsigned long)try_get_task_stack(p);
545 start_backtrace(&frame, thread_saved_fp(p), thread_saved_pc(p));
548 if (unwind_frame(p, &frame))
550 if (!in_sched_functions(frame.pc)) {
554 } while (count ++ < 16);
561 unsigned long arch_align_stack(unsigned long sp)
563 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
564 sp -= get_random_int() & ~PAGE_MASK;
569 * Called from setup_new_exec() after (COMPAT_)SET_PERSONALITY.
571 void arch_setup_new_exec(void)
573 current->mm->context.flags = is_compat_task() ? MMCF_AARCH32 : 0;
575 ptrauth_thread_init_user(current);
578 #ifdef CONFIG_ARM64_TAGGED_ADDR_ABI
580 * Control the relaxed ABI allowing tagged user addresses into the kernel.
582 static unsigned int tagged_addr_disabled;
584 long set_tagged_addr_ctrl(unsigned long arg)
586 if (is_compat_task())
588 if (arg & ~PR_TAGGED_ADDR_ENABLE)
592 * Do not allow the enabling of the tagged address ABI if globally
593 * disabled via sysctl abi.tagged_addr_disabled.
595 if (arg & PR_TAGGED_ADDR_ENABLE && tagged_addr_disabled)
598 update_thread_flag(TIF_TAGGED_ADDR, arg & PR_TAGGED_ADDR_ENABLE);
603 long get_tagged_addr_ctrl(void)
605 if (is_compat_task())
608 if (test_thread_flag(TIF_TAGGED_ADDR))
609 return PR_TAGGED_ADDR_ENABLE;
615 * Global sysctl to disable the tagged user addresses support. This control
616 * only prevents the tagged address ABI enabling via prctl() and does not
617 * disable it for tasks that already opted in to the relaxed ABI.
620 static struct ctl_table tagged_addr_sysctl_table[] = {
622 .procname = "tagged_addr_disabled",
624 .data = &tagged_addr_disabled,
625 .maxlen = sizeof(int),
626 .proc_handler = proc_dointvec_minmax,
627 .extra1 = SYSCTL_ZERO,
628 .extra2 = SYSCTL_ONE,
633 static int __init tagged_addr_init(void)
635 if (!register_sysctl("abi", tagged_addr_sysctl_table))
640 core_initcall(tagged_addr_init);
641 #endif /* CONFIG_ARM64_TAGGED_ADDR_ABI */
643 asmlinkage void __sched arm64_preempt_schedule_irq(void)
645 lockdep_assert_irqs_disabled();
648 * Preempting a task from an IRQ means we leave copies of PSTATE
649 * on the stack. cpufeature's enable calls may modify PSTATE, but
650 * resuming one of these preempted tasks would undo those changes.
652 * Only allow a task to be preempted once cpufeatures have been
655 if (system_capabilities_finalized())
656 preempt_schedule_irq();