arch/arm64/kernel/process.c

   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 #include <linux/percpu.h>
  49
  50 #include <asm/alternative.h>
  51 #include <asm/compat.h>
  52 #include <asm/cacheflush.h>
  53 #include <asm/exec.h>
  54 #include <asm/fpsimd.h>
  55 #include <asm/mmu_context.h>
  56 #include <asm/processor.h>
  57 #include <asm/stacktrace.h>
  58
  59 #ifdef CONFIG_CC_STACKPROTECTOR
  60 #include <linux/stackprotector.h>
  61 unsigned long __stack_chk_guard __read_mostly;
  62 EXPORT_SYMBOL(__stack_chk_guard);
  63 #endif
  64
  65 /*
  66  * Function pointers to optional machine specific functions
  67  */
  68 void (*pm_power_off)(void);
  69 EXPORT_SYMBOL_GPL(pm_power_off);
  70
  71 void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
  72
  73 /*
  74  * This is our default idle handler.
  75  */
  76 void arch_cpu_idle(void)
  77 {
  78         /*
  79          * This should do all the clock switching and wait for interrupt
  80          * tricks
  81          */
  82         trace_cpu_idle_rcuidle(1, smp_processor_id());
  83         cpu_do_idle();
  84         local_irq_enable();
  85         trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
  86 }
  87
  88 #ifdef CONFIG_HOTPLUG_CPU
  89 void arch_cpu_idle_dead(void)
  90 {
  91        cpu_die();
  92 }
  93 #endif
  94
  95 /*
  96  * Called by kexec, immediately prior to machine_kexec().
  97  *
  98  * This must completely disable all secondary CPUs; simply causing those CPUs
  99  * to execute e.g. a RAM-based pin loop is not sufficient. This allows the
 100  * kexec'd kernel to use any and all RAM as it sees fit, without having to
 101  * avoid any code or data used by any SW CPU pin loop. The CPU hotplug
 102  * functionality embodied in disable_nonboot_cpus() to achieve this.
 103  */
 104 void machine_shutdown(void)
 105 {
 106         disable_nonboot_cpus();
 107 }
 108
 109 /*
 110  * Halting simply requires that the secondary CPUs stop performing any
 111  * activity (executing tasks, handling interrupts). smp_send_stop()
 112  * achieves this.
 113  */
 114 void machine_halt(void)
 115 {
 116         local_irq_disable();
 117         smp_send_stop();
 118         while (1);
 119 }
 120
 121 /*
 122  * Power-off simply requires that the secondary CPUs stop performing any
 123  * activity (executing tasks, handling interrupts). smp_send_stop()
 124  * achieves this. When the system power is turned off, it will take all CPUs
 125  * with it.
 126  */
 127 void machine_power_off(void)
 128 {
 129         local_irq_disable();
 130         smp_send_stop();
 131         if (pm_power_off)
 132                 pm_power_off();
 133 }
 134
 135 /*
 136  * Restart requires that the secondary CPUs stop performing any activity
 137  * while the primary CPU resets the system. Systems with multiple CPUs must
 138  * provide a HW restart implementation, to ensure that all CPUs reset at once.
 139  * This is required so that any code running after reset on the primary CPU
 140  * doesn't have to co-ordinate with other CPUs to ensure they aren't still
 141  * executing pre-reset code, and using RAM that the primary CPU's code wishes
 142  * to use. Implementing such co-ordination would be essentially impossible.
 143  */
 144 void machine_restart(char *cmd)
 145 {
 146         /* Disable interrupts first */
 147         local_irq_disable();
 148         smp_send_stop();
 149
 150         /*
 151          * UpdateCapsule() depends on the system being reset via
 152          * ResetSystem().
 153          */
 154         if (efi_enabled(EFI_RUNTIME_SERVICES))
 155                 efi_reboot(reboot_mode, NULL);
 156
 157         /* Now call the architecture specific reboot code. */
 158         if (arm_pm_restart)
 159                 arm_pm_restart(reboot_mode, cmd);
 160         else
 161                 do_kernel_restart(cmd);
 162
 163         /*
 164          * Whoops - the architecture was unable to reboot.
 165          */
 166         printk("Reboot failed -- System halted\n");
 167         while (1);
 168 }
 169
 170 void __show_regs(struct pt_regs *regs)
 171 {
 172         int i, top_reg;
 173         u64 lr, sp;
 174
 175         if (compat_user_mode(regs)) {
 176                 lr = regs->compat_lr;
 177                 sp = regs->compat_sp;
 178                 top_reg = 12;
 179         } else {
 180                 lr = regs->regs[30];
 181                 sp = regs->sp;
 182                 top_reg = 29;
 183         }
 184
 185         show_regs_print_info(KERN_DEFAULT);
 186         print_symbol("PC is at %s\n", instruction_pointer(regs));
 187         print_symbol("LR is at %s\n", lr);
 188         printk("pc : [<%016llx>] lr : [<%016llx>] pstate: %08llx\n",
 189                regs->pc, lr, regs->pstate);
 190         printk("sp : %016llx\n", sp);
 191
 192         i = top_reg;
 193
 194         while (i >= 0) {
 195                 printk("x%-2d: %016llx ", i, regs->regs[i]);
 196                 i--;
 197
 198                 if (i % 2 == 0) {
 199                         pr_cont("x%-2d: %016llx ", i, regs->regs[i]);
 200                         i--;
 201                 }
 202
 203                 pr_cont("\n");
 204         }
 205         printk("\n");
 206 }
 207
 208 void show_regs(struct pt_regs * regs)
 209 {
 210         printk("\n");
 211         __show_regs(regs);
 212 }
 213
 214 static void tls_thread_flush(void)
 215 {
 216         write_sysreg(0, tpidr_el0);
 217
 218         if (is_compat_task()) {
 219                 current->thread.tp_value = 0;
 220
 221                 /*
 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.
 225                  */
 226                 barrier();
 227                 write_sysreg(0, tpidrro_el0);
 228         }
 229 }
 230
 231 void flush_thread(void)
 232 {
 233         fpsimd_flush_thread();
 234         tls_thread_flush();
 235         flush_ptrace_hw_breakpoint(current);
 236 }
 237
 238 void release_thread(struct task_struct *dead_task)
 239 {
 240 }
 241
 242 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 243 {
 244         if (current->mm)
 245                 fpsimd_preserve_current_state();
 246         *dst = *src;
 247         return 0;
 248 }
 249
 250 asmlinkage void ret_from_fork(void) asm("ret_from_fork");
 251
 252 int copy_thread(unsigned long clone_flags, unsigned long stack_start,
 253                 unsigned long stk_sz, struct task_struct *p)
 254 {
 255         struct pt_regs *childregs = task_pt_regs(p);
 256
 257         memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
 258
 259         if (likely(!(p->flags & PF_KTHREAD))) {
 260                 *childregs = *current_pt_regs();
 261                 childregs->regs[0] = 0;
 262
 263                 /*
 264                  * Read the current TLS pointer from tpidr_el0 as it may be
 265                  * out-of-sync with the saved value.
 266                  */
 267                 *task_user_tls(p) = read_sysreg(tpidr_el0);
 268
 269                 if (stack_start) {
 270                         if (is_compat_thread(task_thread_info(p)))
 271                                 childregs->compat_sp = stack_start;
 272                         else
 273                                 childregs->sp = stack_start;
 274                 }
 275
 276                 /*
 277                  * If a TLS pointer was passed to clone (4th argument), use it
 278                  * for the new thread.
 279                  */
 280                 if (clone_flags & CLONE_SETTLS)
 281                         p->thread.tp_value = childregs->regs[3];
 282         } else {
 283                 memset(childregs, 0, sizeof(struct pt_regs));
 284                 childregs->pstate = PSR_MODE_EL1h;
 285                 if (IS_ENABLED(CONFIG_ARM64_UAO) &&
 286                     cpus_have_const_cap(ARM64_HAS_UAO))
 287                         childregs->pstate |= PSR_UAO_BIT;
 288                 p->thread.cpu_context.x19 = stack_start;
 289                 p->thread.cpu_context.x20 = stk_sz;
 290         }
 291         p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
 292         p->thread.cpu_context.sp = (unsigned long)childregs;
 293
 294         ptrace_hw_copy_thread(p);
 295
 296         return 0;
 297 }
 298
 299 static void tls_thread_switch(struct task_struct *next)
 300 {
 301         unsigned long tpidr, tpidrro;
 302
 303         tpidr = read_sysreg(tpidr_el0);
 304         *task_user_tls(current) = tpidr;
 305
 306         tpidr = *task_user_tls(next);
 307         tpidrro = is_compat_thread(task_thread_info(next)) ?
 308                   next->thread.tp_value : 0;
 309
 310         write_sysreg(tpidr, tpidr_el0);
 311         write_sysreg(tpidrro, tpidrro_el0);
 312 }
 313
 314 /* Restore the UAO state depending on next's addr_limit */
 315 void uao_thread_switch(struct task_struct *next)
 316 {
 317         if (IS_ENABLED(CONFIG_ARM64_UAO)) {
 318                 if (task_thread_info(next)->addr_limit == KERNEL_DS)
 319                         asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
 320                 else
 321                         asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
 322         }
 323 }
 324
 325 /*
 326  * We store our current task in sp_el0, which is clobbered by userspace. Keep a
 327  * shadow copy so that we can restore this upon entry from userspace.
 328  *
 329  * This is *only* for exception entry from EL0, and is not valid until we
 330  * __switch_to() a user task.
 331  */
 332 DEFINE_PER_CPU(struct task_struct *, __entry_task);
 333
 334 static void entry_task_switch(struct task_struct *next)
 335 {
 336         __this_cpu_write(__entry_task, next);
 337 }
 338
 339 /*
 340  * Thread switching.
 341  */
 342 struct task_struct *__switch_to(struct task_struct *prev,
 343                                 struct task_struct *next)
 344 {
 345         struct task_struct *last;
 346
 347         fpsimd_thread_switch(next);
 348         tls_thread_switch(next);
 349         hw_breakpoint_thread_switch(next);
 350         contextidr_thread_switch(next);
 351         entry_task_switch(next);
 352         uao_thread_switch(next);
 353
 354         /*
 355          * Complete any pending TLB or cache maintenance on this CPU in case
 356          * the thread migrates to a different CPU.
 357          */
 358         dsb(ish);
 359
 360         /* the actual thread switch */
 361         last = cpu_switch_to(prev, next);
 362
 363         return last;
 364 }
 365
 366 unsigned long get_wchan(struct task_struct *p)
 367 {
 368         struct stackframe frame;
 369         unsigned long stack_page, ret = 0;
 370         int count = 0;
 371         if (!p || p == current || p->state == TASK_RUNNING)
 372                 return 0;
 373
 374         stack_page = (unsigned long)try_get_task_stack(p);
 375         if (!stack_page)
 376                 return 0;
 377
 378         frame.fp = thread_saved_fp(p);
 379         frame.sp = thread_saved_sp(p);
 380         frame.pc = thread_saved_pc(p);
 381 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 382         frame.graph = p->curr_ret_stack;
 383 #endif
 384         do {
 385                 if (frame.sp < stack_page ||
 386                     frame.sp >= stack_page + THREAD_SIZE ||
 387                     unwind_frame(p, &frame))
 388                         goto out;
 389                 if (!in_sched_functions(frame.pc)) {
 390                         ret = frame.pc;
 391                         goto out;
 392                 }
 393         } while (count ++ < 16);
 394
 395 out:
 396         put_task_stack(p);
 397         return ret;
 398 }
 399
 400 unsigned long arch_align_stack(unsigned long sp)
 401 {
 402         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 403                 sp -= get_random_int() & ~PAGE_MASK;
 404         return sp & ~0xf;
 405 }
 406
 407 unsigned long arch_randomize_brk(struct mm_struct *mm)
 408 {
 409         if (is_compat_task())
 410                 return randomize_page(mm->brk, 0x02000000);
 411         else
 412                 return randomize_page(mm->brk, 0x40000000);
 413 }