linux/arch/arm64/kernel/process.c
<<
>>
Prefs
   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/init.h>
  39#include <linux/cpu.h>
  40#include <linux/elfcore.h>
  41#include <linux/pm.h>
  42#include <linux/tick.h>
  43#include <linux/utsname.h>
  44#include <linux/uaccess.h>
  45#include <linux/random.h>
  46#include <linux/hw_breakpoint.h>
  47#include <linux/personality.h>
  48#include <linux/notifier.h>
  49#include <trace/events/power.h>
  50#include <linux/percpu.h>
  51#include <linux/thread_info.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_STACKPROTECTOR
  63#include <linux/stackprotector.h>
  64unsigned long __stack_chk_guard __read_mostly;
  65EXPORT_SYMBOL(__stack_chk_guard);
  66#endif
  67
  68/*
  69 * Function pointers to optional machine specific functions
  70 */
  71void (*pm_power_off)(void);
  72EXPORT_SYMBOL_GPL(pm_power_off);
  73
  74void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd);
  75
  76/*
  77 * This is our default idle handler.
  78 */
  79void 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
  92void 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 */
 107void 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 */
 117void 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 */
 130void 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 */
 147void 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
 173static void print_pstate(struct pt_regs *regs)
 174{
 175        u64 pstate = regs->pstate;
 176
 177        if (compat_user_mode(regs)) {
 178                printk("pstate: %08llx (%c%c%c%c %c %s %s %c%c%c)\n",
 179                        pstate,
 180                        pstate & COMPAT_PSR_N_BIT ? 'N' : 'n',
 181                        pstate & COMPAT_PSR_Z_BIT ? 'Z' : 'z',
 182                        pstate & COMPAT_PSR_C_BIT ? 'C' : 'c',
 183                        pstate & COMPAT_PSR_V_BIT ? 'V' : 'v',
 184                        pstate & COMPAT_PSR_Q_BIT ? 'Q' : 'q',
 185                        pstate & COMPAT_PSR_T_BIT ? "T32" : "A32",
 186                        pstate & COMPAT_PSR_E_BIT ? "BE" : "LE",
 187                        pstate & COMPAT_PSR_A_BIT ? 'A' : 'a',
 188                        pstate & COMPAT_PSR_I_BIT ? 'I' : 'i',
 189                        pstate & COMPAT_PSR_F_BIT ? 'F' : 'f');
 190        } else {
 191                printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n",
 192                        pstate,
 193                        pstate & PSR_N_BIT ? 'N' : 'n',
 194                        pstate & PSR_Z_BIT ? 'Z' : 'z',
 195                        pstate & PSR_C_BIT ? 'C' : 'c',
 196                        pstate & PSR_V_BIT ? 'V' : 'v',
 197                        pstate & PSR_D_BIT ? 'D' : 'd',
 198                        pstate & PSR_A_BIT ? 'A' : 'a',
 199                        pstate & PSR_I_BIT ? 'I' : 'i',
 200                        pstate & PSR_F_BIT ? 'F' : 'f',
 201                        pstate & PSR_PAN_BIT ? '+' : '-',
 202                        pstate & PSR_UAO_BIT ? '+' : '-');
 203        }
 204}
 205
 206void __show_regs(struct pt_regs *regs)
 207{
 208        int i, top_reg;
 209        u64 lr, sp;
 210
 211        if (compat_user_mode(regs)) {
 212                lr = regs->compat_lr;
 213                sp = regs->compat_sp;
 214                top_reg = 12;
 215        } else {
 216                lr = regs->regs[30];
 217                sp = regs->sp;
 218                top_reg = 29;
 219        }
 220
 221        show_regs_print_info(KERN_DEFAULT);
 222        print_pstate(regs);
 223
 224        if (!user_mode(regs)) {
 225                printk("pc : %pS\n", (void *)regs->pc);
 226                printk("lr : %pS\n", (void *)lr);
 227        } else {
 228                printk("pc : %016llx\n", regs->pc);
 229                printk("lr : %016llx\n", lr);
 230        }
 231
 232        printk("sp : %016llx\n", sp);
 233
 234        i = top_reg;
 235
 236        while (i >= 0) {
 237                printk("x%-2d: %016llx ", i, regs->regs[i]);
 238                i--;
 239
 240                if (i % 2 == 0) {
 241                        pr_cont("x%-2d: %016llx ", i, regs->regs[i]);
 242                        i--;
 243                }
 244
 245                pr_cont("\n");
 246        }
 247}
 248
 249void show_regs(struct pt_regs * regs)
 250{
 251        __show_regs(regs);
 252        dump_backtrace(regs, NULL);
 253}
 254
 255static void tls_thread_flush(void)
 256{
 257        write_sysreg(0, tpidr_el0);
 258
 259        if (is_compat_task()) {
 260                current->thread.uw.tp_value = 0;
 261
 262                /*
 263                 * We need to ensure ordering between the shadow state and the
 264                 * hardware state, so that we don't corrupt the hardware state
 265                 * with a stale shadow state during context switch.
 266                 */
 267                barrier();
 268                write_sysreg(0, tpidrro_el0);
 269        }
 270}
 271
 272void flush_thread(void)
 273{
 274        fpsimd_flush_thread();
 275        tls_thread_flush();
 276        flush_ptrace_hw_breakpoint(current);
 277}
 278
 279void release_thread(struct task_struct *dead_task)
 280{
 281}
 282
 283void arch_release_task_struct(struct task_struct *tsk)
 284{
 285        fpsimd_release_task(tsk);
 286}
 287
 288/*
 289 * src and dst may temporarily have aliased sve_state after task_struct
 290 * is copied.  We cannot fix this properly here, because src may have
 291 * live SVE state and dst's thread_info may not exist yet, so tweaking
 292 * either src's or dst's TIF_SVE is not safe.
 293 *
 294 * The unaliasing is done in copy_thread() instead.  This works because
 295 * dst is not schedulable or traceable until both of these functions
 296 * have been called.
 297 */
 298int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 299{
 300        if (current->mm)
 301                fpsimd_preserve_current_state();
 302        *dst = *src;
 303
 304        return 0;
 305}
 306
 307asmlinkage void ret_from_fork(void) asm("ret_from_fork");
 308
 309int copy_thread(unsigned long clone_flags, unsigned long stack_start,
 310                unsigned long stk_sz, struct task_struct *p)
 311{
 312        struct pt_regs *childregs = task_pt_regs(p);
 313
 314        memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
 315
 316        /*
 317         * Unalias p->thread.sve_state (if any) from the parent task
 318         * and disable discard SVE state for p:
 319         */
 320        clear_tsk_thread_flag(p, TIF_SVE);
 321        p->thread.sve_state = NULL;
 322
 323        /*
 324         * In case p was allocated the same task_struct pointer as some
 325         * other recently-exited task, make sure p is disassociated from
 326         * any cpu that may have run that now-exited task recently.
 327         * Otherwise we could erroneously skip reloading the FPSIMD
 328         * registers for p.
 329         */
 330        fpsimd_flush_task_state(p);
 331
 332        if (likely(!(p->flags & PF_KTHREAD))) {
 333                *childregs = *current_pt_regs();
 334                childregs->regs[0] = 0;
 335
 336                /*
 337                 * Read the current TLS pointer from tpidr_el0 as it may be
 338                 * out-of-sync with the saved value.
 339                 */
 340                *task_user_tls(p) = read_sysreg(tpidr_el0);
 341
 342                if (stack_start) {
 343                        if (is_compat_thread(task_thread_info(p)))
 344                                childregs->compat_sp = stack_start;
 345                        else
 346                                childregs->sp = stack_start;
 347                }
 348
 349                /*
 350                 * If a TLS pointer was passed to clone (4th argument), use it
 351                 * for the new thread.
 352                 */
 353                if (clone_flags & CLONE_SETTLS)
 354                        p->thread.uw.tp_value = childregs->regs[3];
 355        } else {
 356                memset(childregs, 0, sizeof(struct pt_regs));
 357                childregs->pstate = PSR_MODE_EL1h;
 358                if (IS_ENABLED(CONFIG_ARM64_UAO) &&
 359                    cpus_have_const_cap(ARM64_HAS_UAO))
 360                        childregs->pstate |= PSR_UAO_BIT;
 361                p->thread.cpu_context.x19 = stack_start;
 362                p->thread.cpu_context.x20 = stk_sz;
 363        }
 364        p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
 365        p->thread.cpu_context.sp = (unsigned long)childregs;
 366
 367        ptrace_hw_copy_thread(p);
 368
 369        return 0;
 370}
 371
 372void tls_preserve_current_state(void)
 373{
 374        *task_user_tls(current) = read_sysreg(tpidr_el0);
 375}
 376
 377static void tls_thread_switch(struct task_struct *next)
 378{
 379        tls_preserve_current_state();
 380
 381        if (is_compat_thread(task_thread_info(next)))
 382                write_sysreg(next->thread.uw.tp_value, tpidrro_el0);
 383        else if (!arm64_kernel_unmapped_at_el0())
 384                write_sysreg(0, tpidrro_el0);
 385
 386        write_sysreg(*task_user_tls(next), tpidr_el0);
 387}
 388
 389/* Restore the UAO state depending on next's addr_limit */
 390void uao_thread_switch(struct task_struct *next)
 391{
 392        if (IS_ENABLED(CONFIG_ARM64_UAO)) {
 393                if (task_thread_info(next)->addr_limit == KERNEL_DS)
 394                        asm(ALTERNATIVE("nop", SET_PSTATE_UAO(1), ARM64_HAS_UAO));
 395                else
 396                        asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
 397        }
 398}
 399
 400/*
 401 * We store our current task in sp_el0, which is clobbered by userspace. Keep a
 402 * shadow copy so that we can restore this upon entry from userspace.
 403 *
 404 * This is *only* for exception entry from EL0, and is not valid until we
 405 * __switch_to() a user task.
 406 */
 407DEFINE_PER_CPU(struct task_struct *, __entry_task);
 408
 409static void entry_task_switch(struct task_struct *next)
 410{
 411        __this_cpu_write(__entry_task, next);
 412}
 413
 414/*
 415 * Thread switching.
 416 */
 417__notrace_funcgraph struct task_struct *__switch_to(struct task_struct *prev,
 418                                struct task_struct *next)
 419{
 420        struct task_struct *last;
 421
 422        fpsimd_thread_switch(next);
 423        tls_thread_switch(next);
 424        hw_breakpoint_thread_switch(next);
 425        contextidr_thread_switch(next);
 426        entry_task_switch(next);
 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         * This full barrier is also required by the membarrier system
 433         * call.
 434         */
 435        dsb(ish);
 436
 437        /* the actual thread switch */
 438        last = cpu_switch_to(prev, next);
 439
 440        return last;
 441}
 442
 443unsigned long get_wchan(struct task_struct *p)
 444{
 445        struct stackframe frame;
 446        unsigned long stack_page, ret = 0;
 447        int count = 0;
 448        if (!p || p == current || p->state == TASK_RUNNING)
 449                return 0;
 450
 451        stack_page = (unsigned long)try_get_task_stack(p);
 452        if (!stack_page)
 453                return 0;
 454
 455        frame.fp = thread_saved_fp(p);
 456        frame.pc = thread_saved_pc(p);
 457#ifdef CONFIG_FUNCTION_GRAPH_TRACER
 458        frame.graph = p->curr_ret_stack;
 459#endif
 460        do {
 461                if (unwind_frame(p, &frame))
 462                        goto out;
 463                if (!in_sched_functions(frame.pc)) {
 464                        ret = frame.pc;
 465                        goto out;
 466                }
 467        } while (count ++ < 16);
 468
 469out:
 470        put_task_stack(p);
 471        return ret;
 472}
 473
 474unsigned long arch_align_stack(unsigned long sp)
 475{
 476        if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
 477                sp -= get_random_int() & ~PAGE_MASK;
 478        return sp & ~0xf;
 479}
 480
 481unsigned long arch_randomize_brk(struct mm_struct *mm)
 482{
 483        if (is_compat_task())
 484                return randomize_page(mm->brk, SZ_32M);
 485        else
 486                return randomize_page(mm->brk, SZ_1G);
 487}
 488
 489/*
 490 * Called from setup_new_exec() after (COMPAT_)SET_PERSONALITY.
 491 */
 492void arch_setup_new_exec(void)
 493{
 494        current->mm->context.flags = is_compat_task() ? MMCF_AARCH32 : 0;
 495}
 496