| .. | .. |
|---|
| 1 | +// SPDX-License-Identifier: GPL-2.0-only |
|---|
| 1 | 2 | /* |
|---|
| 2 | 3 | * Based on arch/arm/kernel/process.c |
|---|
| 3 | 4 | * |
|---|
| 4 | 5 | * Original Copyright (C) 1995 Linus Torvalds |
|---|
| 5 | 6 | * Copyright (C) 1996-2000 Russell King - Converted to ARM. |
|---|
| 6 | 7 | * 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 | 8 | */ |
|---|
| 20 | 9 | |
|---|
| 21 | 10 | #include <stdarg.h> |
|---|
| 22 | 11 | |
|---|
| 23 | 12 | #include <linux/compat.h> |
|---|
| 24 | 13 | #include <linux/efi.h> |
|---|
| 14 | +#include <linux/elf.h> |
|---|
| 25 | 15 | #include <linux/export.h> |
|---|
| 26 | 16 | #include <linux/sched.h> |
|---|
| 27 | 17 | #include <linux/sched/debug.h> |
|---|
| 28 | 18 | #include <linux/sched/task.h> |
|---|
| 29 | 19 | #include <linux/sched/task_stack.h> |
|---|
| 30 | 20 | #include <linux/kernel.h> |
|---|
| 21 | +#include <linux/lockdep.h> |
|---|
| 22 | +#include <linux/mman.h> |
|---|
| 31 | 23 | #include <linux/mm.h> |
|---|
| 24 | +#include <linux/nospec.h> |
|---|
| 32 | 25 | #include <linux/stddef.h> |
|---|
| 33 | 26 | #include <linux/sysctl.h> |
|---|
| 34 | 27 | #include <linux/unistd.h> |
|---|
| .. | .. |
|---|
| 51 | 44 | #include <linux/percpu.h> |
|---|
| 52 | 45 | #include <linux/thread_info.h> |
|---|
| 53 | 46 | #include <linux/prctl.h> |
|---|
| 47 | +#include <trace/hooks/fpsimd.h> |
|---|
| 54 | 48 | |
|---|
| 55 | 49 | #include <asm/alternative.h> |
|---|
| 50 | +#include <asm/arch_gicv3.h> |
|---|
| 56 | 51 | #include <asm/compat.h> |
|---|
| 52 | +#include <asm/cpufeature.h> |
|---|
| 57 | 53 | #include <asm/cacheflush.h> |
|---|
| 58 | 54 | #include <asm/exec.h> |
|---|
| 59 | 55 | #include <asm/fpsimd.h> |
|---|
| 60 | 56 | #include <asm/mmu_context.h> |
|---|
| 57 | +#include <asm/mte.h> |
|---|
| 61 | 58 | #include <asm/processor.h> |
|---|
| 62 | | -#include <asm/scs.h> |
|---|
| 59 | +#include <asm/pointer_auth.h> |
|---|
| 63 | 60 | #include <asm/stacktrace.h> |
|---|
| 64 | 61 | |
|---|
| 65 | | -#ifdef CONFIG_STACKPROTECTOR |
|---|
| 62 | +#if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_STACKPROTECTOR_PER_TASK) |
|---|
| 66 | 63 | #include <linux/stackprotector.h> |
|---|
| 67 | 64 | unsigned long __stack_chk_guard __ro_after_init; |
|---|
| 68 | 65 | EXPORT_SYMBOL(__stack_chk_guard); |
|---|
| .. | .. |
|---|
| 74 | 71 | void (*pm_power_off)(void); |
|---|
| 75 | 72 | EXPORT_SYMBOL_GPL(pm_power_off); |
|---|
| 76 | 73 | |
|---|
| 77 | | -void (*arm_pm_restart)(enum reboot_mode reboot_mode, const char *cmd); |
|---|
| 78 | | -EXPORT_SYMBOL_GPL(arm_pm_restart); |
|---|
| 74 | +static void noinstr __cpu_do_idle(void) |
|---|
| 75 | +{ |
|---|
| 76 | + dsb(sy); |
|---|
| 77 | + wfi(); |
|---|
| 78 | +} |
|---|
| 79 | + |
|---|
| 80 | +static void noinstr __cpu_do_idle_irqprio(void) |
|---|
| 81 | +{ |
|---|
| 82 | + unsigned long pmr; |
|---|
| 83 | + unsigned long daif_bits; |
|---|
| 84 | + |
|---|
| 85 | + daif_bits = read_sysreg(daif); |
|---|
| 86 | + write_sysreg(daif_bits | PSR_I_BIT, daif); |
|---|
| 87 | + |
|---|
| 88 | + /* |
|---|
| 89 | + * Unmask PMR before going idle to make sure interrupts can |
|---|
| 90 | + * be raised. |
|---|
| 91 | + */ |
|---|
| 92 | + pmr = gic_read_pmr(); |
|---|
| 93 | + gic_write_pmr(GIC_PRIO_IRQON | GIC_PRIO_PSR_I_SET); |
|---|
| 94 | + |
|---|
| 95 | + __cpu_do_idle(); |
|---|
| 96 | + |
|---|
| 97 | + gic_write_pmr(pmr); |
|---|
| 98 | + write_sysreg(daif_bits, daif); |
|---|
| 99 | +} |
|---|
| 100 | + |
|---|
| 101 | +/* |
|---|
| 102 | + * cpu_do_idle() |
|---|
| 103 | + * |
|---|
| 104 | + * Idle the processor (wait for interrupt). |
|---|
| 105 | + * |
|---|
| 106 | + * If the CPU supports priority masking we must do additional work to |
|---|
| 107 | + * ensure that interrupts are not masked at the PMR (because the core will |
|---|
| 108 | + * not wake up if we block the wake up signal in the interrupt controller). |
|---|
| 109 | + */ |
|---|
| 110 | +void noinstr cpu_do_idle(void) |
|---|
| 111 | +{ |
|---|
| 112 | + if (system_uses_irq_prio_masking()) |
|---|
| 113 | + __cpu_do_idle_irqprio(); |
|---|
| 114 | + else |
|---|
| 115 | + __cpu_do_idle(); |
|---|
| 116 | +} |
|---|
| 79 | 117 | |
|---|
| 80 | 118 | /* |
|---|
| 81 | 119 | * This is our default idle handler. |
|---|
| 82 | 120 | */ |
|---|
| 83 | | -void arch_cpu_idle(void) |
|---|
| 121 | +void noinstr arch_cpu_idle(void) |
|---|
| 84 | 122 | { |
|---|
| 85 | 123 | /* |
|---|
| 86 | 124 | * This should do all the clock switching and wait for interrupt |
|---|
| 87 | 125 | * tricks |
|---|
| 88 | 126 | */ |
|---|
| 89 | | - trace_cpu_idle_rcuidle(1, smp_processor_id()); |
|---|
| 90 | 127 | cpu_do_idle(); |
|---|
| 91 | | - local_irq_enable(); |
|---|
| 92 | | - trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id()); |
|---|
| 128 | + raw_local_irq_enable(); |
|---|
| 93 | 129 | } |
|---|
| 94 | 130 | |
|---|
| 95 | 131 | void arch_cpu_idle_enter(void) |
|---|
| .. | .. |
|---|
| 116 | 152 | * to execute e.g. a RAM-based pin loop is not sufficient. This allows the |
|---|
| 117 | 153 | * kexec'd kernel to use any and all RAM as it sees fit, without having to |
|---|
| 118 | 154 | * avoid any code or data used by any SW CPU pin loop. The CPU hotplug |
|---|
| 119 | | - * functionality embodied in disable_nonboot_cpus() to achieve this. |
|---|
| 155 | + * functionality embodied in smpt_shutdown_nonboot_cpus() to achieve this. |
|---|
| 120 | 156 | */ |
|---|
| 121 | 157 | void machine_shutdown(void) |
|---|
| 122 | 158 | { |
|---|
| 123 | | - disable_nonboot_cpus(); |
|---|
| 159 | + smp_shutdown_nonboot_cpus(reboot_cpu); |
|---|
| 124 | 160 | } |
|---|
| 125 | 161 | |
|---|
| 126 | 162 | /* |
|---|
| .. | .. |
|---|
| 174 | 210 | efi_reboot(reboot_mode, NULL); |
|---|
| 175 | 211 | |
|---|
| 176 | 212 | /* Now call the architecture specific reboot code. */ |
|---|
| 177 | | - if (arm_pm_restart) |
|---|
| 178 | | - arm_pm_restart(reboot_mode, cmd); |
|---|
| 179 | | - else |
|---|
| 180 | | - do_kernel_restart(cmd); |
|---|
| 213 | + do_kernel_restart(cmd); |
|---|
| 181 | 214 | |
|---|
| 182 | 215 | /* |
|---|
| 183 | 216 | * Whoops - the architecture was unable to reboot. |
|---|
| .. | .. |
|---|
| 185 | 218 | printk("Reboot failed -- System halted\n"); |
|---|
| 186 | 219 | while (1); |
|---|
| 187 | 220 | } |
|---|
| 221 | + |
|---|
| 222 | +#define bstr(suffix, str) [PSR_BTYPE_ ## suffix >> PSR_BTYPE_SHIFT] = str |
|---|
| 223 | +static const char *const btypes[] = { |
|---|
| 224 | + bstr(NONE, "--"), |
|---|
| 225 | + bstr( JC, "jc"), |
|---|
| 226 | + bstr( C, "-c"), |
|---|
| 227 | + bstr( J , "j-") |
|---|
| 228 | +}; |
|---|
| 229 | +#undef bstr |
|---|
| 188 | 230 | |
|---|
| 189 | 231 | static void print_pstate(struct pt_regs *regs) |
|---|
| 190 | 232 | { |
|---|
| .. | .. |
|---|
| 204 | 246 | pstate & PSR_AA32_I_BIT ? 'I' : 'i', |
|---|
| 205 | 247 | pstate & PSR_AA32_F_BIT ? 'F' : 'f'); |
|---|
| 206 | 248 | } else { |
|---|
| 207 | | - printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO)\n", |
|---|
| 249 | + const char *btype_str = btypes[(pstate & PSR_BTYPE_MASK) >> |
|---|
| 250 | + PSR_BTYPE_SHIFT]; |
|---|
| 251 | + |
|---|
| 252 | + printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO %cTCO BTYPE=%s)\n", |
|---|
| 208 | 253 | pstate, |
|---|
| 209 | 254 | pstate & PSR_N_BIT ? 'N' : 'n', |
|---|
| 210 | 255 | pstate & PSR_Z_BIT ? 'Z' : 'z', |
|---|
| .. | .. |
|---|
| 215 | 260 | pstate & PSR_I_BIT ? 'I' : 'i', |
|---|
| 216 | 261 | pstate & PSR_F_BIT ? 'F' : 'f', |
|---|
| 217 | 262 | pstate & PSR_PAN_BIT ? '+' : '-', |
|---|
| 218 | | - pstate & PSR_UAO_BIT ? '+' : '-'); |
|---|
| 263 | + pstate & PSR_UAO_BIT ? '+' : '-', |
|---|
| 264 | + pstate & PSR_TCO_BIT ? '+' : '-', |
|---|
| 265 | + btype_str); |
|---|
| 219 | 266 | } |
|---|
| 220 | 267 | } |
|---|
| 221 | 268 | |
|---|
| .. | .. |
|---|
| 232 | 279 | * don't attempt to dump non-kernel addresses or |
|---|
| 233 | 280 | * values that are probably just small negative numbers |
|---|
| 234 | 281 | */ |
|---|
| 235 | | - if (addr < VA_START || addr > -4096UL) |
|---|
| 282 | + if (addr < PAGE_OFFSET || addr > -4096UL) |
|---|
| 236 | 283 | return; |
|---|
| 237 | 284 | |
|---|
| 238 | 285 | printk("\n%s: %#lx:\n", name, addr + nbytes / 2); |
|---|
| .. | .. |
|---|
| 258 | 305 | |
|---|
| 259 | 306 | for (j = 0; j < 8; j++) { |
|---|
| 260 | 307 | u32 data; |
|---|
| 261 | | - if (probe_kernel_address(p, data)) { |
|---|
| 308 | + |
|---|
| 309 | + if (aarch64_insn_read((void *)p, &data)) { |
|---|
| 262 | 310 | pr_cont(" ********"); |
|---|
| 263 | 311 | } else { |
|---|
| 264 | 312 | pr_cont(" %08x", data); |
|---|
| .. | .. |
|---|
| 307 | 355 | |
|---|
| 308 | 356 | if (!user_mode(regs)) { |
|---|
| 309 | 357 | printk("pc : %pS\n", (void *)regs->pc); |
|---|
| 310 | | - printk("lr : %pS\n", (void *)lr); |
|---|
| 358 | + printk("lr : %pS\n", (void *)ptrauth_strip_insn_pac(lr)); |
|---|
| 311 | 359 | } else { |
|---|
| 312 | 360 | printk("pc : %016llx\n", regs->pc); |
|---|
| 313 | 361 | printk("lr : %016llx\n", lr); |
|---|
| 314 | 362 | } |
|---|
| 315 | 363 | |
|---|
| 316 | 364 | printk("sp : %016llx\n", sp); |
|---|
| 365 | + |
|---|
| 366 | + if (system_uses_irq_prio_masking()) |
|---|
| 367 | + printk("pmr_save: %08llx\n", regs->pmr_save); |
|---|
| 317 | 368 | |
|---|
| 318 | 369 | i = top_reg; |
|---|
| 319 | 370 | |
|---|
| .. | .. |
|---|
| 333 | 384 | void show_regs(struct pt_regs * regs) |
|---|
| 334 | 385 | { |
|---|
| 335 | 386 | __show_regs(regs); |
|---|
| 336 | | - dump_backtrace(regs, NULL); |
|---|
| 387 | + dump_backtrace(regs, NULL, KERN_DEFAULT); |
|---|
| 337 | 388 | |
|---|
| 338 | 389 | if (!user_mode(regs)) |
|---|
| 339 | 390 | show_extra_register_data(regs, 512); |
|---|
| 340 | 391 | } |
|---|
| 392 | +EXPORT_SYMBOL_GPL(show_regs); |
|---|
| 341 | 393 | |
|---|
| 342 | 394 | static void tls_thread_flush(void) |
|---|
| 343 | 395 | { |
|---|
| .. | .. |
|---|
| 400 | 452 | dst->thread.sve_state = NULL; |
|---|
| 401 | 453 | clear_tsk_thread_flag(dst, TIF_SVE); |
|---|
| 402 | 454 | |
|---|
| 455 | + /* clear any pending asynchronous tag fault raised by the parent */ |
|---|
| 456 | + clear_tsk_thread_flag(dst, TIF_MTE_ASYNC_FAULT); |
|---|
| 457 | + |
|---|
| 403 | 458 | return 0; |
|---|
| 404 | 459 | } |
|---|
| 405 | 460 | |
|---|
| 406 | 461 | asmlinkage void ret_from_fork(void) asm("ret_from_fork"); |
|---|
| 407 | 462 | |
|---|
| 408 | 463 | int copy_thread(unsigned long clone_flags, unsigned long stack_start, |
|---|
| 409 | | - unsigned long stk_sz, struct task_struct *p) |
|---|
| 464 | + unsigned long stk_sz, struct task_struct *p, unsigned long tls) |
|---|
| 410 | 465 | { |
|---|
| 411 | 466 | struct pt_regs *childregs = task_pt_regs(p); |
|---|
| 412 | 467 | |
|---|
| .. | .. |
|---|
| 421 | 476 | */ |
|---|
| 422 | 477 | fpsimd_flush_task_state(p); |
|---|
| 423 | 478 | |
|---|
| 424 | | - if (likely(!(p->flags & PF_KTHREAD))) { |
|---|
| 479 | + ptrauth_thread_init_kernel(p); |
|---|
| 480 | + |
|---|
| 481 | + if (likely(!(p->flags & (PF_KTHREAD | PF_IO_WORKER)))) { |
|---|
| 425 | 482 | *childregs = *current_pt_regs(); |
|---|
| 426 | 483 | childregs->regs[0] = 0; |
|---|
| 427 | 484 | |
|---|
| .. | .. |
|---|
| 439 | 496 | } |
|---|
| 440 | 497 | |
|---|
| 441 | 498 | /* |
|---|
| 442 | | - * If a TLS pointer was passed to clone (4th argument), use it |
|---|
| 443 | | - * for the new thread. |
|---|
| 499 | + * If a TLS pointer was passed to clone, use it for the new |
|---|
| 500 | + * thread. |
|---|
| 444 | 501 | */ |
|---|
| 445 | 502 | if (clone_flags & CLONE_SETTLS) |
|---|
| 446 | | - p->thread.uw.tp_value = childregs->regs[3]; |
|---|
| 503 | + p->thread.uw.tp_value = tls; |
|---|
| 447 | 504 | } else { |
|---|
| 505 | + /* |
|---|
| 506 | + * A kthread has no context to ERET to, so ensure any buggy |
|---|
| 507 | + * ERET is treated as an illegal exception return. |
|---|
| 508 | + * |
|---|
| 509 | + * When a user task is created from a kthread, childregs will |
|---|
| 510 | + * be initialized by start_thread() or start_compat_thread(). |
|---|
| 511 | + */ |
|---|
| 448 | 512 | memset(childregs, 0, sizeof(struct pt_regs)); |
|---|
| 449 | | - childregs->pstate = PSR_MODE_EL1h; |
|---|
| 450 | | - if (IS_ENABLED(CONFIG_ARM64_UAO) && |
|---|
| 451 | | - cpus_have_const_cap(ARM64_HAS_UAO)) |
|---|
| 452 | | - childregs->pstate |= PSR_UAO_BIT; |
|---|
| 453 | | - |
|---|
| 454 | | - if (arm64_get_ssbd_state() == ARM64_SSBD_FORCE_DISABLE) |
|---|
| 455 | | - set_ssbs_bit(childregs); |
|---|
| 513 | + childregs->pstate = PSR_MODE_EL1h | PSR_IL_BIT; |
|---|
| 456 | 514 | |
|---|
| 457 | 515 | p->thread.cpu_context.x19 = stack_start; |
|---|
| 458 | 516 | p->thread.cpu_context.x20 = stk_sz; |
|---|
| .. | .. |
|---|
| 499 | 557 | */ |
|---|
| 500 | 558 | static void ssbs_thread_switch(struct task_struct *next) |
|---|
| 501 | 559 | { |
|---|
| 502 | | - struct pt_regs *regs = task_pt_regs(next); |
|---|
| 503 | | - |
|---|
| 504 | 560 | /* |
|---|
| 505 | 561 | * Nothing to do for kernel threads, but 'regs' may be junk |
|---|
| 506 | 562 | * (e.g. idle task) so check the flags and bail early. |
|---|
| .. | .. |
|---|
| 512 | 568 | * If all CPUs implement the SSBS extension, then we just need to |
|---|
| 513 | 569 | * context-switch the PSTATE field. |
|---|
| 514 | 570 | */ |
|---|
| 515 | | - if (cpu_have_feature(cpu_feature(SSBS))) |
|---|
| 571 | + if (cpus_have_const_cap(ARM64_SSBS)) |
|---|
| 516 | 572 | return; |
|---|
| 517 | 573 | |
|---|
| 518 | | - /* If the mitigation is enabled, then we leave SSBS clear. */ |
|---|
| 519 | | - if ((arm64_get_ssbd_state() == ARM64_SSBD_FORCE_ENABLE) || |
|---|
| 520 | | - test_tsk_thread_flag(next, TIF_SSBD)) |
|---|
| 521 | | - return; |
|---|
| 522 | | - |
|---|
| 523 | | - if (compat_user_mode(regs)) |
|---|
| 524 | | - set_compat_ssbs_bit(regs); |
|---|
| 525 | | - else if (user_mode(regs)) |
|---|
| 526 | | - set_ssbs_bit(regs); |
|---|
| 574 | + spectre_v4_enable_task_mitigation(next); |
|---|
| 527 | 575 | } |
|---|
| 528 | 576 | |
|---|
| 529 | 577 | /* |
|---|
| .. | .. |
|---|
| 541 | 589 | } |
|---|
| 542 | 590 | |
|---|
| 543 | 591 | /* |
|---|
| 592 | + * ARM erratum 1418040 handling, affecting the 32bit view of CNTVCT. |
|---|
| 593 | + * Ensure access is disabled when switching to a 32bit task, ensure |
|---|
| 594 | + * access is enabled when switching to a 64bit task. |
|---|
| 595 | + */ |
|---|
| 596 | +static void erratum_1418040_thread_switch(struct task_struct *next) |
|---|
| 597 | +{ |
|---|
| 598 | + if (!IS_ENABLED(CONFIG_ARM64_ERRATUM_1418040) || |
|---|
| 599 | + !this_cpu_has_cap(ARM64_WORKAROUND_1418040)) |
|---|
| 600 | + return; |
|---|
| 601 | + |
|---|
| 602 | + if (is_compat_thread(task_thread_info(next))) |
|---|
| 603 | + sysreg_clear_set(cntkctl_el1, ARCH_TIMER_USR_VCT_ACCESS_EN, 0); |
|---|
| 604 | + else |
|---|
| 605 | + sysreg_clear_set(cntkctl_el1, 0, ARCH_TIMER_USR_VCT_ACCESS_EN); |
|---|
| 606 | +} |
|---|
| 607 | + |
|---|
| 608 | +static void erratum_1418040_new_exec(void) |
|---|
| 609 | +{ |
|---|
| 610 | + preempt_disable(); |
|---|
| 611 | + erratum_1418040_thread_switch(current); |
|---|
| 612 | + preempt_enable(); |
|---|
| 613 | +} |
|---|
| 614 | + |
|---|
| 615 | +/* |
|---|
| 616 | + * __switch_to() checks current->thread.sctlr_user as an optimisation. Therefore |
|---|
| 617 | + * this function must be called with preemption disabled and the update to |
|---|
| 618 | + * sctlr_user must be made in the same preemption disabled block so that |
|---|
| 619 | + * __switch_to() does not see the variable update before the SCTLR_EL1 one. |
|---|
| 620 | + */ |
|---|
| 621 | +void update_sctlr_el1(u64 sctlr) |
|---|
| 622 | +{ |
|---|
| 623 | + /* |
|---|
| 624 | + * EnIA must not be cleared while in the kernel as this is necessary for |
|---|
| 625 | + * in-kernel PAC. It will be cleared on kernel exit if needed. |
|---|
| 626 | + */ |
|---|
| 627 | + sysreg_clear_set(sctlr_el1, SCTLR_USER_MASK & ~SCTLR_ELx_ENIA, sctlr); |
|---|
| 628 | + |
|---|
| 629 | + /* ISB required for the kernel uaccess routines when setting TCF0. */ |
|---|
| 630 | + isb(); |
|---|
| 631 | +} |
|---|
| 632 | + |
|---|
| 633 | +/* |
|---|
| 544 | 634 | * Thread switching. |
|---|
| 545 | 635 | */ |
|---|
| 546 | 636 | __notrace_funcgraph struct task_struct *__switch_to(struct task_struct *prev, |
|---|
| .. | .. |
|---|
| 555 | 645 | entry_task_switch(next); |
|---|
| 556 | 646 | uao_thread_switch(next); |
|---|
| 557 | 647 | ssbs_thread_switch(next); |
|---|
| 558 | | - scs_overflow_check(next); |
|---|
| 648 | + erratum_1418040_thread_switch(next); |
|---|
| 649 | + ptrauth_thread_switch_user(next); |
|---|
| 559 | 650 | |
|---|
| 560 | 651 | /* |
|---|
| 561 | 652 | * Complete any pending TLB or cache maintenance on this CPU in case |
|---|
| .. | .. |
|---|
| 564 | 655 | * call. |
|---|
| 565 | 656 | */ |
|---|
| 566 | 657 | dsb(ish); |
|---|
| 658 | + |
|---|
| 659 | + /* |
|---|
| 660 | + * MTE thread switching must happen after the DSB above to ensure that |
|---|
| 661 | + * any asynchronous tag check faults have been logged in the TFSR*_EL1 |
|---|
| 662 | + * registers. |
|---|
| 663 | + */ |
|---|
| 664 | + mte_thread_switch(next); |
|---|
| 665 | + /* avoid expensive SCTLR_EL1 accesses if no change */ |
|---|
| 666 | + if (prev->thread.sctlr_user != next->thread.sctlr_user) |
|---|
| 667 | + update_sctlr_el1(next->thread.sctlr_user); |
|---|
| 668 | + |
|---|
| 669 | + trace_android_vh_is_fpsimd_save(prev, next); |
|---|
| 567 | 670 | |
|---|
| 568 | 671 | /* the actual thread switch */ |
|---|
| 569 | 672 | last = cpu_switch_to(prev, next); |
|---|
| .. | .. |
|---|
| 583 | 686 | if (!stack_page) |
|---|
| 584 | 687 | return 0; |
|---|
| 585 | 688 | |
|---|
| 586 | | - frame.fp = thread_saved_fp(p); |
|---|
| 587 | | - frame.pc = thread_saved_pc(p); |
|---|
| 588 | | -#ifdef CONFIG_FUNCTION_GRAPH_TRACER |
|---|
| 589 | | - frame.graph = p->curr_ret_stack; |
|---|
| 590 | | -#endif |
|---|
| 689 | + start_backtrace(&frame, thread_saved_fp(p), thread_saved_pc(p)); |
|---|
| 690 | + |
|---|
| 591 | 691 | do { |
|---|
| 592 | 692 | if (unwind_frame(p, &frame)) |
|---|
| 593 | 693 | goto out; |
|---|
| .. | .. |
|---|
| 601 | 701 | put_task_stack(p); |
|---|
| 602 | 702 | return ret; |
|---|
| 603 | 703 | } |
|---|
| 704 | +EXPORT_SYMBOL_GPL(get_wchan); |
|---|
| 604 | 705 | |
|---|
| 605 | 706 | unsigned long arch_align_stack(unsigned long sp) |
|---|
| 606 | 707 | { |
|---|
| .. | .. |
|---|
| 609 | 710 | return sp & ~0xf; |
|---|
| 610 | 711 | } |
|---|
| 611 | 712 | |
|---|
| 612 | | -unsigned long arch_randomize_brk(struct mm_struct *mm) |
|---|
| 613 | | -{ |
|---|
| 614 | | - if (is_compat_task()) |
|---|
| 615 | | - return randomize_page(mm->brk, SZ_32M); |
|---|
| 616 | | - else |
|---|
| 617 | | - return randomize_page(mm->brk, SZ_1G); |
|---|
| 618 | | -} |
|---|
| 619 | | - |
|---|
| 620 | 713 | /* |
|---|
| 621 | 714 | * Called from setup_new_exec() after (COMPAT_)SET_PERSONALITY. |
|---|
| 622 | 715 | */ |
|---|
| 623 | 716 | void arch_setup_new_exec(void) |
|---|
| 624 | 717 | { |
|---|
| 625 | | - current->mm->context.flags = is_compat_task() ? MMCF_AARCH32 : 0; |
|---|
| 718 | + unsigned long mmflags = 0; |
|---|
| 719 | + |
|---|
| 720 | + if (is_compat_task()) { |
|---|
| 721 | + mmflags = MMCF_AARCH32; |
|---|
| 722 | + |
|---|
| 723 | + /* |
|---|
| 724 | + * Restrict the CPU affinity mask for a 32-bit task so that |
|---|
| 725 | + * it contains only 32-bit-capable CPUs. |
|---|
| 726 | + * |
|---|
| 727 | + * From the perspective of the task, this looks similar to |
|---|
| 728 | + * what would happen if the 64-bit-only CPUs were hot-unplugged |
|---|
| 729 | + * at the point of execve(), although we try a bit harder to |
|---|
| 730 | + * honour the cpuset hierarchy. |
|---|
| 731 | + */ |
|---|
| 732 | + if (static_branch_unlikely(&arm64_mismatched_32bit_el0)) |
|---|
| 733 | + force_compatible_cpus_allowed_ptr(current); |
|---|
| 734 | + } |
|---|
| 735 | + |
|---|
| 736 | + current->mm->context.flags = mmflags; |
|---|
| 737 | + ptrauth_thread_init_user(); |
|---|
| 738 | + mte_thread_init_user(); |
|---|
| 739 | + erratum_1418040_new_exec(); |
|---|
| 740 | + |
|---|
| 741 | + if (task_spec_ssb_noexec(current)) { |
|---|
| 742 | + arch_prctl_spec_ctrl_set(current, PR_SPEC_STORE_BYPASS, |
|---|
| 743 | + PR_SPEC_ENABLE); |
|---|
| 744 | + } |
|---|
| 626 | 745 | } |
|---|
| 627 | | - |
|---|
| 628 | | -#ifdef CONFIG_GCC_PLUGIN_STACKLEAK |
|---|
| 629 | | -void __used stackleak_check_alloca(unsigned long size) |
|---|
| 630 | | -{ |
|---|
| 631 | | - unsigned long stack_left; |
|---|
| 632 | | - unsigned long current_sp = current_stack_pointer; |
|---|
| 633 | | - struct stack_info info; |
|---|
| 634 | | - |
|---|
| 635 | | - BUG_ON(!on_accessible_stack(current, current_sp, &info)); |
|---|
| 636 | | - |
|---|
| 637 | | - stack_left = current_sp - info.low; |
|---|
| 638 | | - |
|---|
| 639 | | - /* |
|---|
| 640 | | - * There's a good chance we're almost out of stack space if this |
|---|
| 641 | | - * is true. Using panic() over BUG() is more likely to give |
|---|
| 642 | | - * reliable debugging output. |
|---|
| 643 | | - */ |
|---|
| 644 | | - if (size >= stack_left) |
|---|
| 645 | | - panic("alloca() over the kernel stack boundary\n"); |
|---|
| 646 | | -} |
|---|
| 647 | | -EXPORT_SYMBOL(stackleak_check_alloca); |
|---|
| 648 | | -#endif |
|---|
| 649 | 746 | |
|---|
| 650 | 747 | #ifdef CONFIG_ARM64_TAGGED_ADDR_ABI |
|---|
| 651 | 748 | /* |
|---|
| .. | .. |
|---|
| 653 | 750 | */ |
|---|
| 654 | 751 | static unsigned int tagged_addr_disabled; |
|---|
| 655 | 752 | |
|---|
| 656 | | -long set_tagged_addr_ctrl(unsigned long arg) |
|---|
| 753 | +long set_tagged_addr_ctrl(struct task_struct *task, unsigned long arg) |
|---|
| 657 | 754 | { |
|---|
| 658 | | - if (is_compat_task()) |
|---|
| 755 | + unsigned long valid_mask = PR_TAGGED_ADDR_ENABLE; |
|---|
| 756 | + struct thread_info *ti = task_thread_info(task); |
|---|
| 757 | + |
|---|
| 758 | + if (is_compat_thread(ti)) |
|---|
| 659 | 759 | return -EINVAL; |
|---|
| 660 | | - if (arg & ~PR_TAGGED_ADDR_ENABLE) |
|---|
| 760 | + |
|---|
| 761 | + if (system_supports_mte()) |
|---|
| 762 | + valid_mask |= PR_MTE_TCF_MASK | PR_MTE_TAG_MASK; |
|---|
| 763 | + |
|---|
| 764 | + if (arg & ~valid_mask) |
|---|
| 661 | 765 | return -EINVAL; |
|---|
| 662 | 766 | |
|---|
| 663 | 767 | /* |
|---|
| .. | .. |
|---|
| 667 | 771 | if (arg & PR_TAGGED_ADDR_ENABLE && tagged_addr_disabled) |
|---|
| 668 | 772 | return -EINVAL; |
|---|
| 669 | 773 | |
|---|
| 670 | | - update_thread_flag(TIF_TAGGED_ADDR, arg & PR_TAGGED_ADDR_ENABLE); |
|---|
| 774 | + if (set_mte_ctrl(task, arg) != 0) |
|---|
| 775 | + return -EINVAL; |
|---|
| 776 | + |
|---|
| 777 | + update_ti_thread_flag(ti, TIF_TAGGED_ADDR, arg & PR_TAGGED_ADDR_ENABLE); |
|---|
| 671 | 778 | |
|---|
| 672 | 779 | return 0; |
|---|
| 673 | 780 | } |
|---|
| 674 | 781 | |
|---|
| 675 | | -long get_tagged_addr_ctrl(void) |
|---|
| 782 | +long get_tagged_addr_ctrl(struct task_struct *task) |
|---|
| 676 | 783 | { |
|---|
| 677 | | - if (is_compat_task()) |
|---|
| 784 | + long ret = 0; |
|---|
| 785 | + struct thread_info *ti = task_thread_info(task); |
|---|
| 786 | + |
|---|
| 787 | + if (is_compat_thread(ti)) |
|---|
| 678 | 788 | return -EINVAL; |
|---|
| 679 | 789 | |
|---|
| 680 | | - if (test_thread_flag(TIF_TAGGED_ADDR)) |
|---|
| 681 | | - return PR_TAGGED_ADDR_ENABLE; |
|---|
| 790 | + if (test_ti_thread_flag(ti, TIF_TAGGED_ADDR)) |
|---|
| 791 | + ret = PR_TAGGED_ADDR_ENABLE; |
|---|
| 682 | 792 | |
|---|
| 683 | | - return 0; |
|---|
| 793 | + ret |= get_mte_ctrl(task); |
|---|
| 794 | + |
|---|
| 795 | + return ret; |
|---|
| 684 | 796 | } |
|---|
| 685 | 797 | |
|---|
| 686 | 798 | /* |
|---|
| .. | .. |
|---|
| 688 | 800 | * only prevents the tagged address ABI enabling via prctl() and does not |
|---|
| 689 | 801 | * disable it for tasks that already opted in to the relaxed ABI. |
|---|
| 690 | 802 | */ |
|---|
| 691 | | -static int zero; |
|---|
| 692 | | -static int one = 1; |
|---|
| 693 | 803 | |
|---|
| 694 | 804 | static struct ctl_table tagged_addr_sysctl_table[] = { |
|---|
| 695 | 805 | { |
|---|
| .. | .. |
|---|
| 698 | 808 | .data = &tagged_addr_disabled, |
|---|
| 699 | 809 | .maxlen = sizeof(int), |
|---|
| 700 | 810 | .proc_handler = proc_dointvec_minmax, |
|---|
| 701 | | - .extra1 = &zero, |
|---|
| 702 | | - .extra2 = &one, |
|---|
| 811 | + .extra1 = SYSCTL_ZERO, |
|---|
| 812 | + .extra2 = SYSCTL_ONE, |
|---|
| 703 | 813 | }, |
|---|
| 704 | 814 | { } |
|---|
| 705 | 815 | }; |
|---|
| .. | .. |
|---|
| 713 | 823 | |
|---|
| 714 | 824 | core_initcall(tagged_addr_init); |
|---|
| 715 | 825 | #endif /* CONFIG_ARM64_TAGGED_ADDR_ABI */ |
|---|
| 826 | + |
|---|
| 827 | +asmlinkage void __sched arm64_preempt_schedule_irq(void) |
|---|
| 828 | +{ |
|---|
| 829 | + lockdep_assert_irqs_disabled(); |
|---|
| 830 | + |
|---|
| 831 | + /* |
|---|
| 832 | + * Preempting a task from an IRQ means we leave copies of PSTATE |
|---|
| 833 | + * on the stack. cpufeature's enable calls may modify PSTATE, but |
|---|
| 834 | + * resuming one of these preempted tasks would undo those changes. |
|---|
| 835 | + * |
|---|
| 836 | + * Only allow a task to be preempted once cpufeatures have been |
|---|
| 837 | + * enabled. |
|---|
| 838 | + */ |
|---|
| 839 | + if (system_capabilities_finalized()) |
|---|
| 840 | + preempt_schedule_irq(); |
|---|
| 841 | +} |
|---|
| 842 | + |
|---|
| 843 | +#ifdef CONFIG_BINFMT_ELF |
|---|
| 844 | +int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state, |
|---|
| 845 | + bool has_interp, bool is_interp) |
|---|
| 846 | +{ |
|---|
| 847 | + /* |
|---|
| 848 | + * For dynamically linked executables the interpreter is |
|---|
| 849 | + * responsible for setting PROT_BTI on everything except |
|---|
| 850 | + * itself. |
|---|
| 851 | + */ |
|---|
| 852 | + if (is_interp != has_interp) |
|---|
| 853 | + return prot; |
|---|
| 854 | + |
|---|
| 855 | + if (!(state->flags & ARM64_ELF_BTI)) |
|---|
| 856 | + return prot; |
|---|
| 857 | + |
|---|
| 858 | + if (prot & PROT_EXEC) |
|---|
| 859 | + prot |= PROT_BTI; |
|---|
| 860 | + |
|---|
| 861 | + return prot; |
|---|
| 862 | +} |
|---|
| 863 | +#endif |
|---|