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/sched/debug.h>
  28 #include <linux/sched/task.h>
  29 #include <linux/sched/task_stack.h>
  30 #include <linux/kernel.h>
  31 #include <linux/mm.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>
  42 #include <linux/pm.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>
  52
  53 #include <asm/alternative.h>
  54 #include <asm/compat.h>
  55 #include <asm/cacheflush.h>
  56 #include <asm/exec.h>
  57 #include <asm/fpsimd.h>
  58 #include <asm/mmu_context.h>
  59 #include <asm/processor.h>
  60 #include <asm/stacktrace.h>
  61
  62 #ifdef CONFIG_CC_STACKPROTECTOR
  63 #include <linux/stackprotector.h>
  64 unsigned long __stack_chk_guard __read_mostly;
  65 EXPORT_SYMBOL(__stack_chk_guard);
  66 #endif
  67
  68 /*
  69  * Function pointers to optional machine specific functions
  70  */
  71 void (*pm_power_off)(void);
  72 EXPORT_SYMBOL_GPL(pm_power_off);
  73
  74 void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
  75
  76 /*
  77  * This is our default idle handler.
  78  */
  79 void arch_cpu_idle(void)
  80 {
  81         /*
  82          * This should do all the clock switching and wait for interrupt
  83          * tricks
  84          */
  85         trace_cpu_idle_rcuidle(1, smp_processor_id());
  86         cpu_do_idle();
  87         local_irq_enable();
  88         trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
  89 }
  90
  91 #ifdef CONFIG_HOTPLUG_CPU
  92 void arch_cpu_idle_dead(void)
  93 {
  94        cpu_die();
  95 }
  96 #endif
  97
  98 /*
  99  * Called by kexec, immediately prior to machine_kexec().
 100  *
 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.
 106  */
 107 void machine_shutdown(void)
 108 {
 109         disable_nonboot_cpus();
 110 }
 111
 112 /*
 113  * Halting simply requires that the secondary CPUs stop performing any
 114  * activity (executing tasks, handling interrupts). smp_send_stop()
 115  * achieves this.
 116  */
 117 void machine_halt(void)
 118 {
 119         local_irq_disable();
 120         smp_send_stop();
 121         while (1);
 122 }
 123
 124 /*
 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
 128  * with it.
 129  */
 130 void machine_power_off(void)
 131 {
 132         local_irq_disable();
 133         smp_send_stop();
 134         if (pm_power_off)
 135                 pm_power_off();
 136 }
 137
 138 /*
 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.
 146  */
 147 void machine_restart(char *cmd)
 148 {
 149         /* Disable interrupts first */
 150         local_irq_disable();
 151         smp_send_stop();
 152
 153         /*
 154          * UpdateCapsule() depends on the system being reset via
 155          * ResetSystem().
 156          */
 157         if (efi_enabled(EFI_RUNTIME_SERVICES))
 158                 efi_reboot(reboot_mode, NULL);
 159
 160         /* Now call the architecture specific reboot code. */
 161         if (arm_pm_restart)
 162                 arm_pm_restart(reboot_mode, cmd);
 163         else
 164                 do_kernel_restart(cmd);
 165
 166         /*
 167          * Whoops - the architecture was unable to reboot.
 168          */
 169         printk("Reboot failed -- System halted\n");
 170         while (1);
 171 }
 172
 173 void __show_regs(struct pt_regs *regs)
 174 {
 175         int i, top_reg;
 176         u64 lr, sp;
 177
 178         if (compat_user_mode(regs)) {
 179                 lr = regs->compat_lr;
 180                 sp = regs->compat_sp;
 181                 top_reg = 12;
 182         } else {
 183                 lr = regs->regs[30];
 184                 sp = regs->sp;
 185                 top_reg = 29;
 186         }
 187
 188         show_regs_print_info(KERN_DEFAULT);
 189         print_symbol("PC is at %s\n", instruction_pointer(regs));
 190         print_symbol("LR is at %s\n", lr);
 191         printk("pc : [<%016llx>] lr : [<%016llx>] pstate: %08llx\n",
 192                regs->pc, lr, regs->pstate);
 193         printk("sp : %016llx\n", sp);
 194
 195         i = top_reg;
 196
 197         while (i >= 0) {
 198                 printk("x%-2d: %016llx ", i, regs->regs[i]);
 199                 i--;
 200
 201                 if (i % 2 == 0) {
 202                         pr_cont("x%-2d: %016llx ", i, regs->regs[i]);
 203                         i--;
 204                 }
 205
 206                 pr_cont("\n");
 207         }
 208 }
 209
 210 void show_regs(struct pt_regs * regs)
 211 {
 212         __show_regs(regs);
 213         dump_backtrace(regs, NULL);
 214 }
 215
 216 static void tls_thread_flush(void)
 217 {
 218         write_sysreg(0, tpidr_el0);
 219
 220         if (is_compat_task()) {
 221                 current->thread.tp_value = 0;
 222
 223                 /*
 224                  * We need to ensure ordering between the shadow state and the
 225                  * hardware state, so that we don't corrupt the hardware state
 226                  * with a stale shadow state during context switch.
 227                  */
 228                 barrier();
 229                 write_sysreg(0, tpidrro_el0);
 230         }
 231 }
 232
 233 void flush_thread(void)
 234 {
 235         fpsimd_flush_thread();
 236         tls_thread_flush();
 237         flush_ptrace_hw_breakpoint(current);
 238 }
 239
 240 void release_thread(struct task_struct *dead_task)
 241 {
 242 }
 243
 244 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 245 {
 246         if (current->mm)
 247                 fpsimd_preserve_current_state();
 248         *dst = *src;
 249         return 0;
 250 }
 251
 252 asmlinkage void ret_from_fork(void) asm("ret_from_fork");
 253
 254 int copy_thread(unsigned long clone_flags, unsigned long stack_start,
 255                 unsigned long stk_sz, struct task_struct *p)
 256 {
 257         struct pt_regs *childregs = task_pt_regs(p);
 258
 259         memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
 260
 261         if (likely(!(p->flags & PF_KTHREAD))) {
 262                 *childregs = *current_pt_regs();
 263                 childregs->regs[0] = 0;
 264
 265                 /*
 266                  * Read the current TLS pointer from tpidr_el0 as it may be
 267                  * out-of-sync with the saved value.
 268                  */
 269                 *task_user_tls(p) = read_sysreg(tpidr_el0);
 270
 271                 if (stack_start) {
 272                         if (is_compat_thread(task_thread_info(p)))
 273                                 childregs->compat_sp = stack_start;
 274                         else
 275                                 childregs->sp = stack_start;
 276                 }
 277
 278                 /*
 279                  * If a TLS pointer was passed to clone (4th argument), use it
 280                  * for the new thread.
 281                  */
 282                 if (clone_flags & CLONE_SETTLS)
 283                         p->thread.tp_value = childregs->regs[3];
 284         } else {
 285                 memset(childregs, 0, sizeof(struct pt_regs));
 286                 childregs->pstate = PSR_MODE_EL1h;
 287                 if (IS_ENABLED(CONFIG_ARM64_UAO) &&
 288                     cpus_have_const_cap(ARM64_HAS_UAO))
 289                         childregs->pstate |= PSR_UAO_BIT;
 290                 p->thread.cpu_context.x19 = stack_start;
 291                 p->thread.cpu_context.x20 = stk_sz;
 292         }
 293         p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
 294         p->thread.cpu_context.sp = (unsigned long)childregs;
 295
 296         ptrace_hw_copy_thread(p);
 297
 298         return 0;
 299 }
 300
 301 void tls_preserve_current_state(void)
 302 {
 303         *task_user_tls(current) = read_sysreg(tpidr_el0);
 304 }
 305
 306 static void tls_thread_switch(struct task_struct *next)
 307 {
 308         unsigned long tpidr, tpidrro;
 309
 310         tls_preserve_current_state();
 311
 312         tpidr = *task_user_tls(next);
 313         tpidrro = is_compat_thread(task_thread_info(next)) ?
 314                   next->thread.tp_value : 0;
 315
 316         write_sysreg(tpidr, tpidr_el0);
 317         write_sysreg(tpidrro, tpidrro_el0);
 318 }
 319
 320 /* Restore the UAO state depending on next's addr_limit */
 321 void uao_thread_switch(struct task_struct *next)
 322 {
 323         if (IS_ENABLED(CONFIG_ARM64_UAO)) {
 324                 if (task_thread_info(next)->addr_limit == KERNEL_DS)
 325                         asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
 326                 else
 327                         asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
 328         }
 329 }
 330
 331 /*
 332  * We store our current task in sp_el0, which is clobbered by userspace. Keep a
 333  * shadow copy so that we can restore this upon entry from userspace.
 334  *
 335  * This is *only* for exception entry from EL0, and is not valid until we
 336  * __switch_to() a user task.
 337  */
 338 DEFINE_PER_CPU(struct task_struct *, __entry_task);
 339
 340 static void entry_task_switch(struct task_struct *next)
 341 {
 342         __this_cpu_write(__entry_task, next);
 343 }
 344
 345 /*
 346  * Thread switching.
 347  */
 348 __notrace_funcgraph struct task_struct *__switch_to(struct task_struct *prev,
 349                                 struct task_struct *next)
 350 {
 351         struct task_struct *last;
 352
 353         fpsimd_thread_switch(next);
 354         tls_thread_switch(next);
 355         hw_breakpoint_thread_switch(next);
 356         contextidr_thread_switch(next);
 357         entry_task_switch(next);
 358         uao_thread_switch(next);
 359
 360         /*
 361          * Complete any pending TLB or cache maintenance on this CPU in case
 362          * the thread migrates to a different CPU.
 363          */
 364         dsb(ish);
 365
 366         /* the actual thread switch */
 367         last = cpu_switch_to(prev, next);
 368
 369         return last;
 370 }
 371
 372 unsigned long get_wchan(struct task_struct *p)
 373 {
 374         struct stackframe frame;
 375         unsigned long stack_page, ret = 0;
 376         int count = 0;
 377         if (!p || p == current || p->state == TASK_RUNNING)
 378                 return 0;
 379
 380         stack_page = (unsigned long)try_get_task_stack(p);
 381         if (!stack_page)
 382                 return 0;
 383
 384         frame.fp = thread_saved_fp(p);
 385         frame.sp = thread_saved_sp(p);
 386         frame.pc = thread_saved_pc(p);
 387 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
 388         frame.graph = p->curr_ret_stack;
 389 #endif
 390         do {
 391                 if (frame.sp < stack_page ||
 392                     frame.sp >= stack_page + THREAD_SIZE ||
 393                     unwind_frame(p, &frame))
 394                         goto out;
 395                 if (!in_sched_functions(frame.pc)) {
 396                         ret = frame.pc;
 397                         goto out;
 398                 }
 399         } while (count ++ < 16);
 400
 401 out:
 402         put_task_stack(p);
 403         return ret;
 404 }
 405
 406 unsigned long arch_align_stack(unsigned long sp)
 407 {
 408         if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 409                 sp -= get_random_int() & ~PAGE_MASK;
 410         return sp & ~0xf;
 411 }
 412
 413 unsigned long arch_randomize_brk(struct mm_struct *mm)
 414 {
 415         if (is_compat_task())
 416                 return randomize_page(mm->brk, SZ_32M);
 417         else
 418                 return randomize_page(mm->brk, SZ_1G);
 419 }