| 1 | // SPDX-License-Identifier: GPL-2.0 |
| 2 | /* |
| 3 | * Author: Huacai Chen <chenhuacai@loongson.cn> |
| 4 | * Copyright (C) 2020-2022 Loongson Technology Corporation Limited |
| 5 | * |
| 6 | * Derived from MIPS: |
| 7 | * Copyright (C) 1994 - 1999, 2000 by Ralf Baechle and others. |
| 8 | * Copyright (C) 2005, 2006 by Ralf Baechle (ralf@linux-mips.org) |
| 9 | * Copyright (C) 1999, 2000 Silicon Graphics, Inc. |
| 10 | * Copyright (C) 2004 Thiemo Seufer |
| 11 | * Copyright (C) 2013 Imagination Technologies Ltd. |
| 12 | */ |
| 13 | #include <linux/cpu.h> |
| 14 | #include <linux/init.h> |
| 15 | #include <linux/kernel.h> |
| 16 | #include <linux/entry-common.h> |
| 17 | #include <linux/errno.h> |
| 18 | #include <linux/sched.h> |
| 19 | #include <linux/sched/debug.h> |
| 20 | #include <linux/sched/task.h> |
| 21 | #include <linux/sched/task_stack.h> |
| 22 | #include <linux/hw_breakpoint.h> |
| 23 | #include <linux/mm.h> |
| 24 | #include <linux/stddef.h> |
| 25 | #include <linux/unistd.h> |
| 26 | #include <linux/export.h> |
| 27 | #include <linux/ptrace.h> |
| 28 | #include <linux/mman.h> |
| 29 | #include <linux/personality.h> |
| 30 | #include <linux/sys.h> |
| 31 | #include <linux/completion.h> |
| 32 | #include <linux/kallsyms.h> |
| 33 | #include <linux/random.h> |
| 34 | #include <linux/prctl.h> |
| 35 | #include <linux/nmi.h> |
| 36 | |
| 37 | #include <asm/asm.h> |
| 38 | #include <asm/asm-prototypes.h> |
| 39 | #include <asm/bootinfo.h> |
| 40 | #include <asm/cpu.h> |
| 41 | #include <asm/elf.h> |
| 42 | #include <asm/exec.h> |
| 43 | #include <asm/fpu.h> |
| 44 | #include <asm/lbt.h> |
| 45 | #include <asm/io.h> |
| 46 | #include <asm/irq.h> |
| 47 | #include <asm/irq_regs.h> |
| 48 | #include <asm/loongarch.h> |
| 49 | #include <asm/pgtable.h> |
| 50 | #include <asm/processor.h> |
| 51 | #include <asm/reg.h> |
| 52 | #include <asm/switch_to.h> |
| 53 | #include <asm/unwind.h> |
| 54 | #include <asm/vdso.h> |
| 55 | |
| 56 | #ifdef CONFIG_STACKPROTECTOR |
| 57 | #include <linux/stackprotector.h> |
| 58 | unsigned long __stack_chk_guard __read_mostly; |
| 59 | EXPORT_SYMBOL(__stack_chk_guard); |
| 60 | #endif |
| 61 | |
| 62 | /* |
| 63 | * Idle related variables and functions |
| 64 | */ |
| 65 | |
| 66 | unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE; |
| 67 | EXPORT_SYMBOL(boot_option_idle_override); |
| 68 | |
| 69 | asmlinkage void restore_and_ret(void); |
| 70 | asmlinkage void ret_from_fork_asm(void); |
| 71 | asmlinkage void ret_from_kernel_thread_asm(void); |
| 72 | |
| 73 | void start_thread(struct pt_regs *regs, unsigned long pc, unsigned long sp) |
| 74 | { |
| 75 | unsigned long crmd; |
| 76 | unsigned long prmd; |
| 77 | unsigned long euen; |
| 78 | |
| 79 | /* New thread loses kernel privileges. */ |
| 80 | crmd = regs->csr_crmd & ~(PLV_MASK); |
| 81 | crmd |= PLV_USER; |
| 82 | regs->csr_crmd = crmd; |
| 83 | |
| 84 | prmd = regs->csr_prmd & ~(PLV_MASK); |
| 85 | prmd |= PLV_USER; |
| 86 | regs->csr_prmd = prmd; |
| 87 | |
| 88 | euen = regs->csr_euen & ~(CSR_EUEN_FPEN); |
| 89 | regs->csr_euen = euen; |
| 90 | lose_fpu(0); |
| 91 | lose_lbt(0); |
| 92 | current->thread.fpu.fcsr = boot_cpu_data.fpu_csr0; |
| 93 | |
| 94 | clear_thread_flag(TIF_LSX_CTX_LIVE); |
| 95 | clear_thread_flag(TIF_LASX_CTX_LIVE); |
| 96 | clear_thread_flag(TIF_LBT_CTX_LIVE); |
| 97 | clear_used_math(); |
| 98 | regs->csr_era = pc; |
| 99 | regs->regs[3] = sp; |
| 100 | } |
| 101 | |
| 102 | void flush_thread(void) |
| 103 | { |
| 104 | flush_ptrace_hw_breakpoint(current); |
| 105 | } |
| 106 | |
| 107 | void exit_thread(struct task_struct *tsk) |
| 108 | { |
| 109 | } |
| 110 | |
| 111 | int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) |
| 112 | { |
| 113 | /* |
| 114 | * Save any process state which is live in hardware registers to the |
| 115 | * parent context prior to duplication. This prevents the new child |
| 116 | * state becoming stale if the parent is preempted before copy_thread() |
| 117 | * gets a chance to save the parent's live hardware registers to the |
| 118 | * child context. |
| 119 | */ |
| 120 | preempt_disable(); |
| 121 | |
| 122 | if (is_fpu_owner()) { |
| 123 | if (is_lasx_enabled()) |
| 124 | save_lasx(current); |
| 125 | else if (is_lsx_enabled()) |
| 126 | save_lsx(current); |
| 127 | else |
| 128 | save_fp(current); |
| 129 | } |
| 130 | |
| 131 | preempt_enable(); |
| 132 | |
| 133 | if (IS_ENABLED(CONFIG_RANDSTRUCT)) { |
| 134 | memcpy(dst, src, sizeof(struct task_struct)); |
| 135 | return 0; |
| 136 | } |
| 137 | |
| 138 | if (!used_math()) |
| 139 | memcpy(dst, src, offsetof(struct task_struct, thread.fpu.fpr)); |
| 140 | else |
| 141 | memcpy(dst, src, offsetof(struct task_struct, thread.lbt.scr0)); |
| 142 | |
| 143 | #ifdef CONFIG_CPU_HAS_LBT |
| 144 | memcpy(&dst->thread.lbt, &src->thread.lbt, sizeof(struct loongarch_lbt)); |
| 145 | #endif |
| 146 | |
| 147 | return 0; |
| 148 | } |
| 149 | |
| 150 | asmlinkage void noinstr __no_stack_protector ret_from_fork(struct task_struct *prev, |
| 151 | struct pt_regs *regs) |
| 152 | { |
| 153 | schedule_tail(prev); |
| 154 | syscall_exit_to_user_mode(regs); |
| 155 | } |
| 156 | |
| 157 | asmlinkage void noinstr __no_stack_protector ret_from_kernel_thread(struct task_struct *prev, |
| 158 | struct pt_regs *regs, |
| 159 | int (*fn)(void *), |
| 160 | void *fn_arg) |
| 161 | { |
| 162 | schedule_tail(prev); |
| 163 | fn(fn_arg); |
| 164 | syscall_exit_to_user_mode(regs); |
| 165 | } |
| 166 | |
| 167 | /* |
| 168 | * Copy architecture-specific thread state |
| 169 | */ |
| 170 | int copy_thread(struct task_struct *p, const struct kernel_clone_args *args) |
| 171 | { |
| 172 | unsigned long childksp; |
| 173 | unsigned long tls = args->tls; |
| 174 | unsigned long usp = args->stack; |
| 175 | u64 clone_flags = args->flags; |
| 176 | struct pt_regs *childregs, *regs = current_pt_regs(); |
| 177 | |
| 178 | childksp = (unsigned long)task_stack_page(task: p) + THREAD_SIZE; |
| 179 | |
| 180 | /* set up new TSS. */ |
| 181 | childregs = (struct pt_regs *) childksp - 1; |
| 182 | /* Put the stack after the struct pt_regs. */ |
| 183 | childksp = (unsigned long) childregs; |
| 184 | p->thread.sched_cfa = 0; |
| 185 | p->thread.csr_euen = 0; |
| 186 | p->thread.csr_crmd = csr_read32(LOONGARCH_CSR_CRMD); |
| 187 | p->thread.csr_prmd = csr_read32(LOONGARCH_CSR_PRMD); |
| 188 | p->thread.csr_ecfg = csr_read32(LOONGARCH_CSR_ECFG); |
| 189 | if (unlikely(args->fn)) { |
| 190 | /* kernel thread */ |
| 191 | p->thread.reg03 = childksp; |
| 192 | p->thread.reg23 = (unsigned long)args->fn; |
| 193 | p->thread.reg24 = (unsigned long)args->fn_arg; |
| 194 | p->thread.reg01 = (unsigned long)ret_from_kernel_thread_asm; |
| 195 | p->thread.sched_ra = (unsigned long)ret_from_kernel_thread_asm; |
| 196 | memset(childregs, 0, sizeof(struct pt_regs)); |
| 197 | childregs->csr_euen = p->thread.csr_euen; |
| 198 | childregs->csr_crmd = p->thread.csr_crmd; |
| 199 | childregs->csr_prmd = p->thread.csr_prmd; |
| 200 | childregs->csr_ecfg = p->thread.csr_ecfg; |
| 201 | goto out; |
| 202 | } |
| 203 | |
| 204 | /* user thread */ |
| 205 | *childregs = *regs; |
| 206 | childregs->regs[4] = 0; /* Child gets zero as return value */ |
| 207 | if (usp) |
| 208 | childregs->regs[3] = usp; |
| 209 | |
| 210 | p->thread.reg03 = (unsigned long) childregs; |
| 211 | p->thread.reg01 = (unsigned long) ret_from_fork_asm; |
| 212 | p->thread.sched_ra = (unsigned long) ret_from_fork_asm; |
| 213 | |
| 214 | /* |
| 215 | * New tasks lose permission to use the fpu. This accelerates context |
| 216 | * switching for most programs since they don't use the fpu. |
| 217 | */ |
| 218 | childregs->csr_euen = 0; |
| 219 | |
| 220 | if (clone_flags & CLONE_SETTLS) |
| 221 | childregs->regs[2] = tls; |
| 222 | |
| 223 | out: |
| 224 | ptrace_hw_copy_thread(p); |
| 225 | clear_tsk_thread_flag(p, TIF_USEDFPU); |
| 226 | clear_tsk_thread_flag(p, TIF_USEDSIMD); |
| 227 | clear_tsk_thread_flag(p, TIF_USEDLBT); |
| 228 | clear_tsk_thread_flag(p, TIF_LSX_CTX_LIVE); |
| 229 | clear_tsk_thread_flag(p, TIF_LASX_CTX_LIVE); |
| 230 | clear_tsk_thread_flag(p, TIF_LBT_CTX_LIVE); |
| 231 | |
| 232 | return 0; |
| 233 | } |
| 234 | |
| 235 | unsigned long __get_wchan(struct task_struct *task) |
| 236 | { |
| 237 | unsigned long pc = 0; |
| 238 | struct unwind_state state; |
| 239 | |
| 240 | if (!try_get_task_stack(tsk: task)) |
| 241 | return 0; |
| 242 | |
| 243 | for (unwind_start(&state, task, NULL); |
| 244 | !unwind_done(state: &state); unwind_next_frame(state: &state)) { |
| 245 | pc = unwind_get_return_address(state: &state); |
| 246 | if (!pc) |
| 247 | break; |
| 248 | if (in_sched_functions(addr: pc)) |
| 249 | continue; |
| 250 | break; |
| 251 | } |
| 252 | |
| 253 | put_task_stack(tsk: task); |
| 254 | |
| 255 | return pc; |
| 256 | } |
| 257 | |
| 258 | bool in_irq_stack(unsigned long stack, struct stack_info *info) |
| 259 | { |
| 260 | unsigned long nextsp; |
| 261 | unsigned long begin = (unsigned long)this_cpu_read(irq_stack); |
| 262 | unsigned long end = begin + IRQ_STACK_START; |
| 263 | |
| 264 | if (stack < begin || stack >= end) |
| 265 | return false; |
| 266 | |
| 267 | nextsp = *(unsigned long *)end; |
| 268 | if (nextsp & (SZREG - 1)) |
| 269 | return false; |
| 270 | |
| 271 | info->begin = begin; |
| 272 | info->end = end; |
| 273 | info->next_sp = nextsp; |
| 274 | info->type = STACK_TYPE_IRQ; |
| 275 | |
| 276 | return true; |
| 277 | } |
| 278 | |
| 279 | bool in_task_stack(unsigned long stack, struct task_struct *task, |
| 280 | struct stack_info *info) |
| 281 | { |
| 282 | unsigned long begin = (unsigned long)task_stack_page(task); |
| 283 | unsigned long end = begin + THREAD_SIZE; |
| 284 | |
| 285 | if (stack < begin || stack >= end) |
| 286 | return false; |
| 287 | |
| 288 | info->begin = begin; |
| 289 | info->end = end; |
| 290 | info->next_sp = 0; |
| 291 | info->type = STACK_TYPE_TASK; |
| 292 | |
| 293 | return true; |
| 294 | } |
| 295 | |
| 296 | int get_stack_info(unsigned long stack, struct task_struct *task, |
| 297 | struct stack_info *info) |
| 298 | { |
| 299 | task = task ? : current; |
| 300 | |
| 301 | if (!stack || stack & (SZREG - 1)) |
| 302 | goto unknown; |
| 303 | |
| 304 | if (in_task_stack(stack, task, info)) |
| 305 | return 0; |
| 306 | |
| 307 | if (task != current) |
| 308 | goto unknown; |
| 309 | |
| 310 | if (in_irq_stack(stack, info)) |
| 311 | return 0; |
| 312 | |
| 313 | unknown: |
| 314 | info->type = STACK_TYPE_UNKNOWN; |
| 315 | return -EINVAL; |
| 316 | } |
| 317 | |
| 318 | unsigned long stack_top(void) |
| 319 | { |
| 320 | unsigned long top = TASK_SIZE & PAGE_MASK; |
| 321 | |
| 322 | if (current->thread.vdso) { |
| 323 | /* Space for the VDSO & data page */ |
| 324 | top -= PAGE_ALIGN(current->thread.vdso->size); |
| 325 | top -= VVAR_SIZE; |
| 326 | |
| 327 | /* Space to randomize the VDSO base */ |
| 328 | if (current->flags & PF_RANDOMIZE) |
| 329 | top -= VDSO_RANDOMIZE_SIZE; |
| 330 | } |
| 331 | |
| 332 | return top; |
| 333 | } |
| 334 | |
| 335 | /* |
| 336 | * Don't forget that the stack pointer must be aligned on a 8 bytes |
| 337 | * boundary for 32-bits ABI and 16 bytes for 64-bits ABI. |
| 338 | */ |
| 339 | unsigned long arch_align_stack(unsigned long sp) |
| 340 | { |
| 341 | if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space) |
| 342 | sp -= get_random_u32_below(PAGE_SIZE); |
| 343 | |
| 344 | return sp & STACK_ALIGN; |
| 345 | } |
| 346 | |
| 347 | static DEFINE_PER_CPU(call_single_data_t, backtrace_csd); |
| 348 | static struct cpumask backtrace_csd_busy; |
| 349 | |
| 350 | static void handle_backtrace(void *info) |
| 351 | { |
| 352 | nmi_cpu_backtrace(regs: get_irq_regs()); |
| 353 | cpumask_clear_cpu(smp_processor_id(), dstp: &backtrace_csd_busy); |
| 354 | } |
| 355 | |
| 356 | static void raise_backtrace(cpumask_t *mask) |
| 357 | { |
| 358 | call_single_data_t *csd; |
| 359 | int cpu; |
| 360 | |
| 361 | for_each_cpu(cpu, mask) { |
| 362 | /* |
| 363 | * If we previously sent an IPI to the target CPU & it hasn't |
| 364 | * cleared its bit in the busy cpumask then it didn't handle |
| 365 | * our previous IPI & it's not safe for us to reuse the |
| 366 | * call_single_data_t. |
| 367 | */ |
| 368 | if (cpumask_test_and_set_cpu(cpu, cpumask: &backtrace_csd_busy)) { |
| 369 | pr_warn("Unable to send backtrace IPI to CPU%u - perhaps it hung?\n" , |
| 370 | cpu); |
| 371 | continue; |
| 372 | } |
| 373 | |
| 374 | csd = &per_cpu(backtrace_csd, cpu); |
| 375 | csd->func = handle_backtrace; |
| 376 | smp_call_function_single_async(cpu, csd); |
| 377 | } |
| 378 | } |
| 379 | |
| 380 | void arch_trigger_cpumask_backtrace(const cpumask_t *mask, int exclude_cpu) |
| 381 | { |
| 382 | nmi_trigger_cpumask_backtrace(mask, exclude_cpu, raise: raise_backtrace); |
| 383 | } |
| 384 | |
| 385 | #ifdef CONFIG_32BIT |
| 386 | void loongarch_dump_regs32(u32 *uregs, const struct pt_regs *regs) |
| 387 | #else |
| 388 | void loongarch_dump_regs64(u64 *uregs, const struct pt_regs *regs) |
| 389 | #endif |
| 390 | { |
| 391 | unsigned int i; |
| 392 | |
| 393 | for (i = LOONGARCH_EF_R1; i <= LOONGARCH_EF_R31; i++) { |
| 394 | uregs[i] = regs->regs[i - LOONGARCH_EF_R0]; |
| 395 | } |
| 396 | |
| 397 | uregs[LOONGARCH_EF_ORIG_A0] = regs->orig_a0; |
| 398 | uregs[LOONGARCH_EF_CSR_ERA] = regs->csr_era; |
| 399 | uregs[LOONGARCH_EF_CSR_BADV] = regs->csr_badvaddr; |
| 400 | uregs[LOONGARCH_EF_CSR_CRMD] = regs->csr_crmd; |
| 401 | uregs[LOONGARCH_EF_CSR_PRMD] = regs->csr_prmd; |
| 402 | uregs[LOONGARCH_EF_CSR_EUEN] = regs->csr_euen; |
| 403 | uregs[LOONGARCH_EF_CSR_ECFG] = regs->csr_ecfg; |
| 404 | uregs[LOONGARCH_EF_CSR_ESTAT] = regs->csr_estat; |
| 405 | } |
| 406 | |