1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Architecture-specific setup. 4 * 5 * Copyright (C) 1998-2003 Hewlett-Packard Co 6 * David Mosberger-Tang <davidm@hpl.hp.com> 7 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support 8 * 9 * 2005-10-07 Keith Owens <kaos@sgi.com> 10 * Add notify_die() hooks. 11 */ 12 #include <linux/cpu.h> 13 #include <linux/pm.h> 14 #include <linux/elf.h> 15 #include <linux/errno.h> 16 #include <linux/kernel.h> 17 #include <linux/mm.h> 18 #include <linux/slab.h> 19 #include <linux/module.h> 20 #include <linux/notifier.h> 21 #include <linux/personality.h> 22 #include <linux/reboot.h> 23 #include <linux/sched.h> 24 #include <linux/sched/debug.h> 25 #include <linux/sched/hotplug.h> 26 #include <linux/sched/task.h> 27 #include <linux/sched/task_stack.h> 28 #include <linux/stddef.h> 29 #include <linux/thread_info.h> 30 #include <linux/unistd.h> 31 #include <linux/efi.h> 32 #include <linux/interrupt.h> 33 #include <linux/delay.h> 34 #include <linux/kdebug.h> 35 #include <linux/utsname.h> 36 #include <linux/resume_user_mode.h> 37 #include <linux/rcupdate.h> 38 39 #include <asm/cpu.h> 40 #include <asm/delay.h> 41 #include <asm/elf.h> 42 #include <asm/irq.h> 43 #include <asm/kexec.h> 44 #include <asm/processor.h> 45 #include <asm/sal.h> 46 #include <asm/switch_to.h> 47 #include <asm/tlbflush.h> 48 #include <linux/uaccess.h> 49 #include <asm/unwind.h> 50 #include <asm/user.h> 51 #include <asm/xtp.h> 52 53 #include "entry.h" 54 55 #include "sigframe.h" 56 57 void (*ia64_mark_idle)(int); 58 59 unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE; 60 EXPORT_SYMBOL(boot_option_idle_override); 61 void (*pm_power_off) (void); 62 EXPORT_SYMBOL(pm_power_off); 63 64 static void 65 ia64_do_show_stack (struct unw_frame_info *info, void *arg) 66 { 67 unsigned long ip, sp, bsp; 68 const char *loglvl = arg; 69 70 printk("%s\nCall Trace:\n", loglvl); 71 do { 72 unw_get_ip(info, &ip); 73 if (ip == 0) 74 break; 75 76 unw_get_sp(info, &sp); 77 unw_get_bsp(info, &bsp); 78 printk("%s [<%016lx>] %pS\n" 79 " sp=%016lx bsp=%016lx\n", 80 loglvl, ip, (void *)ip, sp, bsp); 81 } while (unw_unwind(info) >= 0); 82 } 83 84 void 85 show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl) 86 { 87 if (!task) 88 unw_init_running(ia64_do_show_stack, (void *)loglvl); 89 else { 90 struct unw_frame_info info; 91 92 unw_init_from_blocked_task(&info, task); 93 ia64_do_show_stack(&info, (void *)loglvl); 94 } 95 } 96 97 void 98 show_regs (struct pt_regs *regs) 99 { 100 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri; 101 102 print_modules(); 103 printk("\n"); 104 show_regs_print_info(KERN_DEFAULT); 105 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n", 106 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(), 107 init_utsname()->release); 108 printk("ip is at %pS\n", (void *)ip); 109 printk("unat: %016lx pfs : %016lx rsc : %016lx\n", 110 regs->ar_unat, regs->ar_pfs, regs->ar_rsc); 111 printk("rnat: %016lx bsps: %016lx pr : %016lx\n", 112 regs->ar_rnat, regs->ar_bspstore, regs->pr); 113 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n", 114 regs->loadrs, regs->ar_ccv, regs->ar_fpsr); 115 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd); 116 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7); 117 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n", 118 regs->f6.u.bits[1], regs->f6.u.bits[0], 119 regs->f7.u.bits[1], regs->f7.u.bits[0]); 120 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n", 121 regs->f8.u.bits[1], regs->f8.u.bits[0], 122 regs->f9.u.bits[1], regs->f9.u.bits[0]); 123 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n", 124 regs->f10.u.bits[1], regs->f10.u.bits[0], 125 regs->f11.u.bits[1], regs->f11.u.bits[0]); 126 127 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3); 128 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10); 129 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13); 130 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16); 131 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19); 132 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22); 133 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25); 134 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28); 135 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31); 136 137 if (user_mode(regs)) { 138 /* print the stacked registers */ 139 unsigned long val, *bsp, ndirty; 140 int i, sof, is_nat = 0; 141 142 sof = regs->cr_ifs & 0x7f; /* size of frame */ 143 ndirty = (regs->loadrs >> 19); 144 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty); 145 for (i = 0; i < sof; ++i) { 146 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i)); 147 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val, 148 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " "); 149 } 150 } else 151 show_stack(NULL, NULL, KERN_DEFAULT); 152 } 153 154 /* local support for deprecated console_print */ 155 void 156 console_print(const char *s) 157 { 158 printk(KERN_EMERG "%s", s); 159 } 160 161 void 162 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall) 163 { 164 if (fsys_mode(current, &scr->pt)) { 165 /* 166 * defer signal-handling etc. until we return to 167 * privilege-level 0. 168 */ 169 if (!ia64_psr(&scr->pt)->lp) 170 ia64_psr(&scr->pt)->lp = 1; 171 return; 172 } 173 174 /* deal with pending signal delivery */ 175 if (test_thread_flag(TIF_SIGPENDING) || 176 test_thread_flag(TIF_NOTIFY_SIGNAL)) { 177 local_irq_enable(); /* force interrupt enable */ 178 ia64_do_signal(scr, in_syscall); 179 } 180 181 if (test_thread_flag(TIF_NOTIFY_RESUME)) { 182 local_irq_enable(); /* force interrupt enable */ 183 resume_user_mode_work(&scr->pt); 184 } 185 186 /* copy user rbs to kernel rbs */ 187 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) { 188 local_irq_enable(); /* force interrupt enable */ 189 ia64_sync_krbs(); 190 } 191 192 local_irq_disable(); /* force interrupt disable */ 193 } 194 195 static int __init nohalt_setup(char * str) 196 { 197 cpu_idle_poll_ctrl(true); 198 return 1; 199 } 200 __setup("nohalt", nohalt_setup); 201 202 #ifdef CONFIG_HOTPLUG_CPU 203 /* We don't actually take CPU down, just spin without interrupts. */ 204 static inline void play_dead(void) 205 { 206 unsigned int this_cpu = smp_processor_id(); 207 208 /* Ack it */ 209 __this_cpu_write(cpu_state, CPU_DEAD); 210 211 max_xtp(); 212 local_irq_disable(); 213 idle_task_exit(); 214 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]); 215 /* 216 * The above is a point of no-return, the processor is 217 * expected to be in SAL loop now. 218 */ 219 BUG(); 220 } 221 #else 222 static inline void play_dead(void) 223 { 224 BUG(); 225 } 226 #endif /* CONFIG_HOTPLUG_CPU */ 227 228 void arch_cpu_idle_dead(void) 229 { 230 play_dead(); 231 } 232 233 void arch_cpu_idle(void) 234 { 235 void (*mark_idle)(int) = ia64_mark_idle; 236 237 #ifdef CONFIG_SMP 238 min_xtp(); 239 #endif 240 rmb(); 241 if (mark_idle) 242 (*mark_idle)(1); 243 244 raw_safe_halt(); 245 246 if (mark_idle) 247 (*mark_idle)(0); 248 #ifdef CONFIG_SMP 249 normal_xtp(); 250 #endif 251 } 252 253 void 254 ia64_save_extra (struct task_struct *task) 255 { 256 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) 257 ia64_save_debug_regs(&task->thread.dbr[0]); 258 } 259 260 void 261 ia64_load_extra (struct task_struct *task) 262 { 263 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) 264 ia64_load_debug_regs(&task->thread.dbr[0]); 265 } 266 267 /* 268 * Copy the state of an ia-64 thread. 269 * 270 * We get here through the following call chain: 271 * 272 * from user-level: from kernel: 273 * 274 * <clone syscall> <some kernel call frames> 275 * sys_clone : 276 * kernel_clone kernel_clone 277 * copy_thread copy_thread 278 * 279 * This means that the stack layout is as follows: 280 * 281 * +---------------------+ (highest addr) 282 * | struct pt_regs | 283 * +---------------------+ 284 * | struct switch_stack | 285 * +---------------------+ 286 * | | 287 * | memory stack | 288 * | | <-- sp (lowest addr) 289 * +---------------------+ 290 * 291 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an 292 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register, 293 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the 294 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since 295 * the stack is page aligned and the page size is at least 4KB, this is always the case, 296 * so there is nothing to worry about. 297 */ 298 int 299 copy_thread(unsigned long clone_flags, unsigned long user_stack_base, 300 unsigned long user_stack_size, struct task_struct *p, unsigned long tls) 301 { 302 extern char ia64_ret_from_clone; 303 struct switch_stack *child_stack, *stack; 304 unsigned long rbs, child_rbs, rbs_size; 305 struct pt_regs *child_ptregs; 306 struct pt_regs *regs = current_pt_regs(); 307 int retval = 0; 308 309 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1; 310 child_stack = (struct switch_stack *) child_ptregs - 1; 311 312 rbs = (unsigned long) current + IA64_RBS_OFFSET; 313 child_rbs = (unsigned long) p + IA64_RBS_OFFSET; 314 315 /* copy parts of thread_struct: */ 316 p->thread.ksp = (unsigned long) child_stack - 16; 317 318 /* 319 * NOTE: The calling convention considers all floating point 320 * registers in the high partition (fph) to be scratch. Since 321 * the only way to get to this point is through a system call, 322 * we know that the values in fph are all dead. Hence, there 323 * is no need to inherit the fph state from the parent to the 324 * child and all we have to do is to make sure that 325 * IA64_THREAD_FPH_VALID is cleared in the child. 326 * 327 * XXX We could push this optimization a bit further by 328 * clearing IA64_THREAD_FPH_VALID on ANY system call. 329 * However, it's not clear this is worth doing. Also, it 330 * would be a slight deviation from the normal Linux system 331 * call behavior where scratch registers are preserved across 332 * system calls (unless used by the system call itself). 333 */ 334 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \ 335 | IA64_THREAD_PM_VALID) 336 # define THREAD_FLAGS_TO_SET 0 337 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR) 338 | THREAD_FLAGS_TO_SET); 339 340 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */ 341 342 if (unlikely(p->flags & (PF_KTHREAD | PF_IO_WORKER))) { 343 if (unlikely(!user_stack_base)) { 344 /* fork_idle() called us */ 345 return 0; 346 } 347 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack)); 348 child_stack->r4 = user_stack_base; /* payload */ 349 child_stack->r5 = user_stack_size; /* argument */ 350 /* 351 * Preserve PSR bits, except for bits 32-34 and 37-45, 352 * which we can't read. 353 */ 354 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN; 355 /* mark as valid, empty frame */ 356 child_ptregs->cr_ifs = 1UL << 63; 357 child_stack->ar_fpsr = child_ptregs->ar_fpsr 358 = ia64_getreg(_IA64_REG_AR_FPSR); 359 child_stack->pr = (1 << PRED_KERNEL_STACK); 360 child_stack->ar_bspstore = child_rbs; 361 child_stack->b0 = (unsigned long) &ia64_ret_from_clone; 362 363 /* stop some PSR bits from being inherited. 364 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve() 365 * therefore we must specify them explicitly here and not include them in 366 * IA64_PSR_BITS_TO_CLEAR. 367 */ 368 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET) 369 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP)); 370 371 return 0; 372 } 373 stack = ((struct switch_stack *) regs) - 1; 374 /* copy parent's switch_stack & pt_regs to child: */ 375 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack)); 376 377 /* copy the parent's register backing store to the child: */ 378 rbs_size = stack->ar_bspstore - rbs; 379 memcpy((void *) child_rbs, (void *) rbs, rbs_size); 380 if (clone_flags & CLONE_SETTLS) 381 child_ptregs->r13 = tls; 382 if (user_stack_base) { 383 child_ptregs->r12 = user_stack_base + user_stack_size - 16; 384 child_ptregs->ar_bspstore = user_stack_base; 385 child_ptregs->ar_rnat = 0; 386 child_ptregs->loadrs = 0; 387 } 388 child_stack->ar_bspstore = child_rbs + rbs_size; 389 child_stack->b0 = (unsigned long) &ia64_ret_from_clone; 390 391 /* stop some PSR bits from being inherited. 392 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve() 393 * therefore we must specify them explicitly here and not include them in 394 * IA64_PSR_BITS_TO_CLEAR. 395 */ 396 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET) 397 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP)); 398 return retval; 399 } 400 401 asmlinkage long ia64_clone(unsigned long clone_flags, unsigned long stack_start, 402 unsigned long stack_size, unsigned long parent_tidptr, 403 unsigned long child_tidptr, unsigned long tls) 404 { 405 struct kernel_clone_args args = { 406 .flags = (lower_32_bits(clone_flags) & ~CSIGNAL), 407 .pidfd = (int __user *)parent_tidptr, 408 .child_tid = (int __user *)child_tidptr, 409 .parent_tid = (int __user *)parent_tidptr, 410 .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL), 411 .stack = stack_start, 412 .stack_size = stack_size, 413 .tls = tls, 414 }; 415 416 return kernel_clone(&args); 417 } 418 419 static void 420 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg) 421 { 422 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm; 423 unsigned long ip; 424 elf_greg_t *dst = arg; 425 struct pt_regs *pt; 426 char nat; 427 int i; 428 429 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */ 430 431 if (unw_unwind_to_user(info) < 0) 432 return; 433 434 unw_get_sp(info, &sp); 435 pt = (struct pt_regs *) (sp + 16); 436 437 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm); 438 439 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0) 440 return; 441 442 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end), 443 &ar_rnat); 444 445 /* 446 * coredump format: 447 * r0-r31 448 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT) 449 * predicate registers (p0-p63) 450 * b0-b7 451 * ip cfm user-mask 452 * ar.rsc ar.bsp ar.bspstore ar.rnat 453 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec 454 */ 455 456 /* r0 is zero */ 457 for (i = 1, mask = (1UL << i); i < 32; ++i) { 458 unw_get_gr(info, i, &dst[i], &nat); 459 if (nat) 460 nat_bits |= mask; 461 mask <<= 1; 462 } 463 dst[32] = nat_bits; 464 unw_get_pr(info, &dst[33]); 465 466 for (i = 0; i < 8; ++i) 467 unw_get_br(info, i, &dst[34 + i]); 468 469 unw_get_rp(info, &ip); 470 dst[42] = ip + ia64_psr(pt)->ri; 471 dst[43] = cfm; 472 dst[44] = pt->cr_ipsr & IA64_PSR_UM; 473 474 unw_get_ar(info, UNW_AR_RSC, &dst[45]); 475 /* 476 * For bsp and bspstore, unw_get_ar() would return the kernel 477 * addresses, but we need the user-level addresses instead: 478 */ 479 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */ 480 dst[47] = pt->ar_bspstore; 481 dst[48] = ar_rnat; 482 unw_get_ar(info, UNW_AR_CCV, &dst[49]); 483 unw_get_ar(info, UNW_AR_UNAT, &dst[50]); 484 unw_get_ar(info, UNW_AR_FPSR, &dst[51]); 485 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */ 486 unw_get_ar(info, UNW_AR_LC, &dst[53]); 487 unw_get_ar(info, UNW_AR_EC, &dst[54]); 488 unw_get_ar(info, UNW_AR_CSD, &dst[55]); 489 unw_get_ar(info, UNW_AR_SSD, &dst[56]); 490 } 491 492 static void 493 do_copy_regs (struct unw_frame_info *info, void *arg) 494 { 495 do_copy_task_regs(current, info, arg); 496 } 497 498 void 499 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst) 500 { 501 unw_init_running(do_copy_regs, dst); 502 } 503 504 /* 505 * Flush thread state. This is called when a thread does an execve(). 506 */ 507 void 508 flush_thread (void) 509 { 510 /* drop floating-point and debug-register state if it exists: */ 511 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID); 512 ia64_drop_fpu(current); 513 } 514 515 /* 516 * Clean up state associated with a thread. This is called when 517 * the thread calls exit(). 518 */ 519 void 520 exit_thread (struct task_struct *tsk) 521 { 522 523 ia64_drop_fpu(tsk); 524 } 525 526 unsigned long 527 __get_wchan (struct task_struct *p) 528 { 529 struct unw_frame_info info; 530 unsigned long ip; 531 int count = 0; 532 533 /* 534 * Note: p may not be a blocked task (it could be current or 535 * another process running on some other CPU. Rather than 536 * trying to determine if p is really blocked, we just assume 537 * it's blocked and rely on the unwind routines to fail 538 * gracefully if the process wasn't really blocked after all. 539 * --davidm 99/12/15 540 */ 541 unw_init_from_blocked_task(&info, p); 542 do { 543 if (task_is_running(p)) 544 return 0; 545 if (unw_unwind(&info) < 0) 546 return 0; 547 unw_get_ip(&info, &ip); 548 if (!in_sched_functions(ip)) 549 return ip; 550 } while (count++ < 16); 551 return 0; 552 } 553 554 void 555 cpu_halt (void) 556 { 557 pal_power_mgmt_info_u_t power_info[8]; 558 unsigned long min_power; 559 int i, min_power_state; 560 561 if (ia64_pal_halt_info(power_info) != 0) 562 return; 563 564 min_power_state = 0; 565 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption; 566 for (i = 1; i < 8; ++i) 567 if (power_info[i].pal_power_mgmt_info_s.im 568 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) { 569 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption; 570 min_power_state = i; 571 } 572 573 while (1) 574 ia64_pal_halt(min_power_state); 575 } 576 577 void machine_shutdown(void) 578 { 579 smp_shutdown_nonboot_cpus(reboot_cpu); 580 581 #ifdef CONFIG_KEXEC 582 kexec_disable_iosapic(); 583 #endif 584 } 585 586 void 587 machine_restart (char *restart_cmd) 588 { 589 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0); 590 efi_reboot(REBOOT_WARM, NULL); 591 } 592 593 void 594 machine_halt (void) 595 { 596 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0); 597 cpu_halt(); 598 } 599 600 void 601 machine_power_off (void) 602 { 603 do_kernel_power_off(); 604 machine_halt(); 605 } 606 607 EXPORT_SYMBOL(ia64_delay_loop); 608