1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Common boot and setup code for both 32-bit and 64-bit. 4 * Extracted from arch/powerpc/kernel/setup_64.c. 5 * 6 * Copyright (C) 2001 PPC64 Team, IBM Corp 7 */ 8 9 #undef DEBUG 10 11 #include <linux/export.h> 12 #include <linux/panic_notifier.h> 13 #include <linux/string.h> 14 #include <linux/sched.h> 15 #include <linux/init.h> 16 #include <linux/kernel.h> 17 #include <linux/reboot.h> 18 #include <linux/delay.h> 19 #include <linux/initrd.h> 20 #include <linux/platform_device.h> 21 #include <linux/printk.h> 22 #include <linux/seq_file.h> 23 #include <linux/ioport.h> 24 #include <linux/console.h> 25 #include <linux/screen_info.h> 26 #include <linux/root_dev.h> 27 #include <linux/cpu.h> 28 #include <linux/unistd.h> 29 #include <linux/seq_buf.h> 30 #include <linux/serial.h> 31 #include <linux/serial_8250.h> 32 #include <linux/percpu.h> 33 #include <linux/memblock.h> 34 #include <linux/of_irq.h> 35 #include <linux/of_fdt.h> 36 #include <linux/of_platform.h> 37 #include <linux/hugetlb.h> 38 #include <linux/pgtable.h> 39 #include <asm/io.h> 40 #include <asm/paca.h> 41 #include <asm/processor.h> 42 #include <asm/vdso_datapage.h> 43 #include <asm/smp.h> 44 #include <asm/elf.h> 45 #include <asm/machdep.h> 46 #include <asm/time.h> 47 #include <asm/cputable.h> 48 #include <asm/sections.h> 49 #include <asm/firmware.h> 50 #include <asm/btext.h> 51 #include <asm/nvram.h> 52 #include <asm/setup.h> 53 #include <asm/rtas.h> 54 #include <asm/iommu.h> 55 #include <asm/serial.h> 56 #include <asm/cache.h> 57 #include <asm/page.h> 58 #include <asm/mmu.h> 59 #include <asm/xmon.h> 60 #include <asm/cputhreads.h> 61 #include <mm/mmu_decl.h> 62 #include <asm/fadump.h> 63 #include <asm/udbg.h> 64 #include <asm/hugetlb.h> 65 #include <asm/livepatch.h> 66 #include <asm/mmu_context.h> 67 #include <asm/cpu_has_feature.h> 68 #include <asm/kasan.h> 69 #include <asm/mce.h> 70 71 #include "setup.h" 72 73 #ifdef DEBUG 74 #define DBG(fmt...) udbg_printf(fmt) 75 #else 76 #define DBG(fmt...) 77 #endif 78 79 /* The main machine-dep calls structure 80 */ 81 struct machdep_calls ppc_md; 82 EXPORT_SYMBOL(ppc_md); 83 struct machdep_calls *machine_id; 84 EXPORT_SYMBOL(machine_id); 85 86 int boot_cpuid = -1; 87 EXPORT_SYMBOL_GPL(boot_cpuid); 88 89 /* 90 * These are used in binfmt_elf.c to put aux entries on the stack 91 * for each elf executable being started. 92 */ 93 int dcache_bsize; 94 int icache_bsize; 95 96 /* 97 * This still seems to be needed... -- paulus 98 */ 99 struct screen_info screen_info = { 100 .orig_x = 0, 101 .orig_y = 25, 102 .orig_video_cols = 80, 103 .orig_video_lines = 25, 104 .orig_video_isVGA = 1, 105 .orig_video_points = 16 106 }; 107 #if defined(CONFIG_FB_VGA16_MODULE) 108 EXPORT_SYMBOL(screen_info); 109 #endif 110 111 /* Variables required to store legacy IO irq routing */ 112 int of_i8042_kbd_irq; 113 EXPORT_SYMBOL_GPL(of_i8042_kbd_irq); 114 int of_i8042_aux_irq; 115 EXPORT_SYMBOL_GPL(of_i8042_aux_irq); 116 117 #ifdef __DO_IRQ_CANON 118 /* XXX should go elsewhere eventually */ 119 int ppc_do_canonicalize_irqs; 120 EXPORT_SYMBOL(ppc_do_canonicalize_irqs); 121 #endif 122 123 #ifdef CONFIG_CRASH_CORE 124 /* This keeps a track of which one is the crashing cpu. */ 125 int crashing_cpu = -1; 126 #endif 127 128 /* also used by kexec */ 129 void machine_shutdown(void) 130 { 131 /* 132 * if fadump is active, cleanup the fadump registration before we 133 * shutdown. 134 */ 135 fadump_cleanup(); 136 137 if (ppc_md.machine_shutdown) 138 ppc_md.machine_shutdown(); 139 } 140 141 static void machine_hang(void) 142 { 143 pr_emerg("System Halted, OK to turn off power\n"); 144 local_irq_disable(); 145 while (1) 146 ; 147 } 148 149 void machine_restart(char *cmd) 150 { 151 machine_shutdown(); 152 if (ppc_md.restart) 153 ppc_md.restart(cmd); 154 155 smp_send_stop(); 156 157 do_kernel_restart(cmd); 158 mdelay(1000); 159 160 machine_hang(); 161 } 162 163 void machine_power_off(void) 164 { 165 machine_shutdown(); 166 do_kernel_power_off(); 167 smp_send_stop(); 168 machine_hang(); 169 } 170 /* Used by the G5 thermal driver */ 171 EXPORT_SYMBOL_GPL(machine_power_off); 172 173 void (*pm_power_off)(void); 174 EXPORT_SYMBOL_GPL(pm_power_off); 175 176 size_t __must_check arch_get_random_seed_longs(unsigned long *v, size_t max_longs) 177 { 178 if (max_longs && ppc_md.get_random_seed && ppc_md.get_random_seed(v)) 179 return 1; 180 return 0; 181 } 182 EXPORT_SYMBOL(arch_get_random_seed_longs); 183 184 void machine_halt(void) 185 { 186 machine_shutdown(); 187 if (ppc_md.halt) 188 ppc_md.halt(); 189 190 smp_send_stop(); 191 machine_hang(); 192 } 193 194 #ifdef CONFIG_SMP 195 DEFINE_PER_CPU(unsigned int, cpu_pvr); 196 #endif 197 198 static void show_cpuinfo_summary(struct seq_file *m) 199 { 200 struct device_node *root; 201 const char *model = NULL; 202 unsigned long bogosum = 0; 203 int i; 204 205 if (IS_ENABLED(CONFIG_SMP) && IS_ENABLED(CONFIG_PPC32)) { 206 for_each_online_cpu(i) 207 bogosum += loops_per_jiffy; 208 seq_printf(m, "total bogomips\t: %lu.%02lu\n", 209 bogosum / (500000 / HZ), bogosum / (5000 / HZ) % 100); 210 } 211 seq_printf(m, "timebase\t: %lu\n", ppc_tb_freq); 212 if (ppc_md.name) 213 seq_printf(m, "platform\t: %s\n", ppc_md.name); 214 root = of_find_node_by_path("/"); 215 if (root) 216 model = of_get_property(root, "model", NULL); 217 if (model) 218 seq_printf(m, "model\t\t: %s\n", model); 219 of_node_put(root); 220 221 if (ppc_md.show_cpuinfo != NULL) 222 ppc_md.show_cpuinfo(m); 223 224 /* Display the amount of memory */ 225 if (IS_ENABLED(CONFIG_PPC32)) 226 seq_printf(m, "Memory\t\t: %d MB\n", 227 (unsigned int)(total_memory / (1024 * 1024))); 228 } 229 230 static int show_cpuinfo(struct seq_file *m, void *v) 231 { 232 unsigned long cpu_id = (unsigned long)v - 1; 233 unsigned int pvr; 234 unsigned long proc_freq; 235 unsigned short maj; 236 unsigned short min; 237 238 #ifdef CONFIG_SMP 239 pvr = per_cpu(cpu_pvr, cpu_id); 240 #else 241 pvr = mfspr(SPRN_PVR); 242 #endif 243 maj = (pvr >> 8) & 0xFF; 244 min = pvr & 0xFF; 245 246 seq_printf(m, "processor\t: %lu\ncpu\t\t: ", cpu_id); 247 248 if (cur_cpu_spec->pvr_mask && cur_cpu_spec->cpu_name) 249 seq_puts(m, cur_cpu_spec->cpu_name); 250 else 251 seq_printf(m, "unknown (%08x)", pvr); 252 253 if (cpu_has_feature(CPU_FTR_ALTIVEC)) 254 seq_puts(m, ", altivec supported"); 255 256 seq_putc(m, '\n'); 257 258 #ifdef CONFIG_TAU 259 if (cpu_has_feature(CPU_FTR_TAU)) { 260 if (IS_ENABLED(CONFIG_TAU_AVERAGE)) { 261 /* more straightforward, but potentially misleading */ 262 seq_printf(m, "temperature \t: %u C (uncalibrated)\n", 263 cpu_temp(cpu_id)); 264 } else { 265 /* show the actual temp sensor range */ 266 u32 temp; 267 temp = cpu_temp_both(cpu_id); 268 seq_printf(m, "temperature \t: %u-%u C (uncalibrated)\n", 269 temp & 0xff, temp >> 16); 270 } 271 } 272 #endif /* CONFIG_TAU */ 273 274 /* 275 * Platforms that have variable clock rates, should implement 276 * the method ppc_md.get_proc_freq() that reports the clock 277 * rate of a given cpu. The rest can use ppc_proc_freq to 278 * report the clock rate that is same across all cpus. 279 */ 280 if (ppc_md.get_proc_freq) 281 proc_freq = ppc_md.get_proc_freq(cpu_id); 282 else 283 proc_freq = ppc_proc_freq; 284 285 if (proc_freq) 286 seq_printf(m, "clock\t\t: %lu.%06luMHz\n", 287 proc_freq / 1000000, proc_freq % 1000000); 288 289 /* If we are a Freescale core do a simple check so 290 * we don't have to keep adding cases in the future */ 291 if (PVR_VER(pvr) & 0x8000) { 292 switch (PVR_VER(pvr)) { 293 case 0x8000: /* 7441/7450/7451, Voyager */ 294 case 0x8001: /* 7445/7455, Apollo 6 */ 295 case 0x8002: /* 7447/7457, Apollo 7 */ 296 case 0x8003: /* 7447A, Apollo 7 PM */ 297 case 0x8004: /* 7448, Apollo 8 */ 298 case 0x800c: /* 7410, Nitro */ 299 maj = ((pvr >> 8) & 0xF); 300 min = PVR_MIN(pvr); 301 break; 302 default: /* e500/book-e */ 303 maj = PVR_MAJ(pvr); 304 min = PVR_MIN(pvr); 305 break; 306 } 307 } else { 308 switch (PVR_VER(pvr)) { 309 case 0x1008: /* 740P/750P ?? */ 310 maj = ((pvr >> 8) & 0xFF) - 1; 311 min = pvr & 0xFF; 312 break; 313 case 0x004e: /* POWER9 bits 12-15 give chip type */ 314 case 0x0080: /* POWER10 bit 12 gives SMT8/4 */ 315 maj = (pvr >> 8) & 0x0F; 316 min = pvr & 0xFF; 317 break; 318 default: 319 maj = (pvr >> 8) & 0xFF; 320 min = pvr & 0xFF; 321 break; 322 } 323 } 324 325 seq_printf(m, "revision\t: %hd.%hd (pvr %04x %04x)\n", 326 maj, min, PVR_VER(pvr), PVR_REV(pvr)); 327 328 if (IS_ENABLED(CONFIG_PPC32)) 329 seq_printf(m, "bogomips\t: %lu.%02lu\n", loops_per_jiffy / (500000 / HZ), 330 (loops_per_jiffy / (5000 / HZ)) % 100); 331 332 seq_putc(m, '\n'); 333 334 /* If this is the last cpu, print the summary */ 335 if (cpumask_next(cpu_id, cpu_online_mask) >= nr_cpu_ids) 336 show_cpuinfo_summary(m); 337 338 return 0; 339 } 340 341 static void *c_start(struct seq_file *m, loff_t *pos) 342 { 343 if (*pos == 0) /* just in case, cpu 0 is not the first */ 344 *pos = cpumask_first(cpu_online_mask); 345 else 346 *pos = cpumask_next(*pos - 1, cpu_online_mask); 347 if ((*pos) < nr_cpu_ids) 348 return (void *)(unsigned long)(*pos + 1); 349 return NULL; 350 } 351 352 static void *c_next(struct seq_file *m, void *v, loff_t *pos) 353 { 354 (*pos)++; 355 return c_start(m, pos); 356 } 357 358 static void c_stop(struct seq_file *m, void *v) 359 { 360 } 361 362 const struct seq_operations cpuinfo_op = { 363 .start = c_start, 364 .next = c_next, 365 .stop = c_stop, 366 .show = show_cpuinfo, 367 }; 368 369 void __init check_for_initrd(void) 370 { 371 #ifdef CONFIG_BLK_DEV_INITRD 372 DBG(" -> check_for_initrd() initrd_start=0x%lx initrd_end=0x%lx\n", 373 initrd_start, initrd_end); 374 375 /* If we were passed an initrd, set the ROOT_DEV properly if the values 376 * look sensible. If not, clear initrd reference. 377 */ 378 if (is_kernel_addr(initrd_start) && is_kernel_addr(initrd_end) && 379 initrd_end > initrd_start) 380 ROOT_DEV = Root_RAM0; 381 else 382 initrd_start = initrd_end = 0; 383 384 if (initrd_start) 385 pr_info("Found initrd at 0x%lx:0x%lx\n", initrd_start, initrd_end); 386 387 DBG(" <- check_for_initrd()\n"); 388 #endif /* CONFIG_BLK_DEV_INITRD */ 389 } 390 391 #ifdef CONFIG_SMP 392 393 int threads_per_core, threads_per_subcore, threads_shift __read_mostly; 394 cpumask_t threads_core_mask __read_mostly; 395 EXPORT_SYMBOL_GPL(threads_per_core); 396 EXPORT_SYMBOL_GPL(threads_per_subcore); 397 EXPORT_SYMBOL_GPL(threads_shift); 398 EXPORT_SYMBOL_GPL(threads_core_mask); 399 400 static void __init cpu_init_thread_core_maps(int tpc) 401 { 402 int i; 403 404 threads_per_core = tpc; 405 threads_per_subcore = tpc; 406 cpumask_clear(&threads_core_mask); 407 408 /* This implementation only supports power of 2 number of threads 409 * for simplicity and performance 410 */ 411 threads_shift = ilog2(tpc); 412 BUG_ON(tpc != (1 << threads_shift)); 413 414 for (i = 0; i < tpc; i++) 415 cpumask_set_cpu(i, &threads_core_mask); 416 417 printk(KERN_INFO "CPU maps initialized for %d thread%s per core\n", 418 tpc, tpc > 1 ? "s" : ""); 419 printk(KERN_DEBUG " (thread shift is %d)\n", threads_shift); 420 } 421 422 423 u32 *cpu_to_phys_id = NULL; 424 425 /** 426 * setup_cpu_maps - initialize the following cpu maps: 427 * cpu_possible_mask 428 * cpu_present_mask 429 * 430 * Having the possible map set up early allows us to restrict allocations 431 * of things like irqstacks to nr_cpu_ids rather than NR_CPUS. 432 * 433 * We do not initialize the online map here; cpus set their own bits in 434 * cpu_online_mask as they come up. 435 * 436 * This function is valid only for Open Firmware systems. finish_device_tree 437 * must be called before using this. 438 * 439 * While we're here, we may as well set the "physical" cpu ids in the paca. 440 * 441 * NOTE: This must match the parsing done in early_init_dt_scan_cpus. 442 */ 443 void __init smp_setup_cpu_maps(void) 444 { 445 struct device_node *dn; 446 int cpu = 0; 447 int nthreads = 1; 448 449 DBG("smp_setup_cpu_maps()\n"); 450 451 cpu_to_phys_id = memblock_alloc(nr_cpu_ids * sizeof(u32), 452 __alignof__(u32)); 453 if (!cpu_to_phys_id) 454 panic("%s: Failed to allocate %zu bytes align=0x%zx\n", 455 __func__, nr_cpu_ids * sizeof(u32), __alignof__(u32)); 456 457 for_each_node_by_type(dn, "cpu") { 458 const __be32 *intserv; 459 __be32 cpu_be; 460 int j, len; 461 462 DBG(" * %pOF...\n", dn); 463 464 intserv = of_get_property(dn, "ibm,ppc-interrupt-server#s", 465 &len); 466 if (intserv) { 467 DBG(" ibm,ppc-interrupt-server#s -> %lu threads\n", 468 (len / sizeof(int))); 469 } else { 470 DBG(" no ibm,ppc-interrupt-server#s -> 1 thread\n"); 471 intserv = of_get_property(dn, "reg", &len); 472 if (!intserv) { 473 cpu_be = cpu_to_be32(cpu); 474 /* XXX: what is this? uninitialized?? */ 475 intserv = &cpu_be; /* assume logical == phys */ 476 len = 4; 477 } 478 } 479 480 nthreads = len / sizeof(int); 481 482 for (j = 0; j < nthreads && cpu < nr_cpu_ids; j++) { 483 bool avail; 484 485 DBG(" thread %d -> cpu %d (hard id %d)\n", 486 j, cpu, be32_to_cpu(intserv[j])); 487 488 avail = of_device_is_available(dn); 489 if (!avail) 490 avail = !of_property_match_string(dn, 491 "enable-method", "spin-table"); 492 493 set_cpu_present(cpu, avail); 494 set_cpu_possible(cpu, true); 495 cpu_to_phys_id[cpu] = be32_to_cpu(intserv[j]); 496 cpu++; 497 } 498 499 if (cpu >= nr_cpu_ids) { 500 of_node_put(dn); 501 break; 502 } 503 } 504 505 /* If no SMT supported, nthreads is forced to 1 */ 506 if (!cpu_has_feature(CPU_FTR_SMT)) { 507 DBG(" SMT disabled ! nthreads forced to 1\n"); 508 nthreads = 1; 509 } 510 511 #ifdef CONFIG_PPC64 512 /* 513 * On pSeries LPAR, we need to know how many cpus 514 * could possibly be added to this partition. 515 */ 516 if (firmware_has_feature(FW_FEATURE_LPAR) && 517 (dn = of_find_node_by_path("/rtas"))) { 518 int num_addr_cell, num_size_cell, maxcpus; 519 const __be32 *ireg; 520 521 num_addr_cell = of_n_addr_cells(dn); 522 num_size_cell = of_n_size_cells(dn); 523 524 ireg = of_get_property(dn, "ibm,lrdr-capacity", NULL); 525 526 if (!ireg) 527 goto out; 528 529 maxcpus = be32_to_cpup(ireg + num_addr_cell + num_size_cell); 530 531 /* Double maxcpus for processors which have SMT capability */ 532 if (cpu_has_feature(CPU_FTR_SMT)) 533 maxcpus *= nthreads; 534 535 if (maxcpus > nr_cpu_ids) { 536 printk(KERN_WARNING 537 "Partition configured for %d cpus, " 538 "operating system maximum is %u.\n", 539 maxcpus, nr_cpu_ids); 540 maxcpus = nr_cpu_ids; 541 } else 542 printk(KERN_INFO "Partition configured for %d cpus.\n", 543 maxcpus); 544 545 for (cpu = 0; cpu < maxcpus; cpu++) 546 set_cpu_possible(cpu, true); 547 out: 548 of_node_put(dn); 549 } 550 vdso_data->processorCount = num_present_cpus(); 551 #endif /* CONFIG_PPC64 */ 552 553 /* Initialize CPU <=> thread mapping/ 554 * 555 * WARNING: We assume that the number of threads is the same for 556 * every CPU in the system. If that is not the case, then some code 557 * here will have to be reworked 558 */ 559 cpu_init_thread_core_maps(nthreads); 560 561 /* Now that possible cpus are set, set nr_cpu_ids for later use */ 562 setup_nr_cpu_ids(); 563 564 free_unused_pacas(); 565 } 566 #endif /* CONFIG_SMP */ 567 568 #ifdef CONFIG_PCSPKR_PLATFORM 569 static __init int add_pcspkr(void) 570 { 571 struct device_node *np; 572 struct platform_device *pd; 573 int ret; 574 575 np = of_find_compatible_node(NULL, NULL, "pnpPNP,100"); 576 of_node_put(np); 577 if (!np) 578 return -ENODEV; 579 580 pd = platform_device_alloc("pcspkr", -1); 581 if (!pd) 582 return -ENOMEM; 583 584 ret = platform_device_add(pd); 585 if (ret) 586 platform_device_put(pd); 587 588 return ret; 589 } 590 device_initcall(add_pcspkr); 591 #endif /* CONFIG_PCSPKR_PLATFORM */ 592 593 static char ppc_hw_desc_buf[128] __initdata; 594 595 struct seq_buf ppc_hw_desc __initdata = { 596 .buffer = ppc_hw_desc_buf, 597 .size = sizeof(ppc_hw_desc_buf), 598 .len = 0, 599 .readpos = 0, 600 }; 601 602 static __init void probe_machine(void) 603 { 604 extern struct machdep_calls __machine_desc_start; 605 extern struct machdep_calls __machine_desc_end; 606 unsigned int i; 607 608 /* 609 * Iterate all ppc_md structures until we find the proper 610 * one for the current machine type 611 */ 612 DBG("Probing machine type ...\n"); 613 614 /* 615 * Check ppc_md is empty, if not we have a bug, ie, we setup an 616 * entry before probe_machine() which will be overwritten 617 */ 618 for (i = 0; i < (sizeof(ppc_md) / sizeof(void *)); i++) { 619 if (((void **)&ppc_md)[i]) { 620 printk(KERN_ERR "Entry %d in ppc_md non empty before" 621 " machine probe !\n", i); 622 } 623 } 624 625 for (machine_id = &__machine_desc_start; 626 machine_id < &__machine_desc_end; 627 machine_id++) { 628 DBG(" %s ...", machine_id->name); 629 memcpy(&ppc_md, machine_id, sizeof(struct machdep_calls)); 630 if (ppc_md.probe()) { 631 DBG(" match !\n"); 632 break; 633 } 634 DBG("\n"); 635 } 636 /* What can we do if we didn't find ? */ 637 if (machine_id >= &__machine_desc_end) { 638 pr_err("No suitable machine description found !\n"); 639 for (;;); 640 } 641 642 // Append the machine name to other info we've gathered 643 seq_buf_puts(&ppc_hw_desc, ppc_md.name); 644 645 // Set the generic hardware description shown in oopses 646 dump_stack_set_arch_desc(ppc_hw_desc.buffer); 647 648 pr_info("Hardware name: %s\n", ppc_hw_desc.buffer); 649 } 650 651 /* Match a class of boards, not a specific device configuration. */ 652 int check_legacy_ioport(unsigned long base_port) 653 { 654 struct device_node *parent, *np = NULL; 655 int ret = -ENODEV; 656 657 switch(base_port) { 658 case I8042_DATA_REG: 659 if (!(np = of_find_compatible_node(NULL, NULL, "pnpPNP,303"))) 660 np = of_find_compatible_node(NULL, NULL, "pnpPNP,f03"); 661 if (np) { 662 parent = of_get_parent(np); 663 664 of_i8042_kbd_irq = irq_of_parse_and_map(parent, 0); 665 if (!of_i8042_kbd_irq) 666 of_i8042_kbd_irq = 1; 667 668 of_i8042_aux_irq = irq_of_parse_and_map(parent, 1); 669 if (!of_i8042_aux_irq) 670 of_i8042_aux_irq = 12; 671 672 of_node_put(np); 673 np = parent; 674 break; 675 } 676 np = of_find_node_by_type(NULL, "8042"); 677 /* Pegasos has no device_type on its 8042 node, look for the 678 * name instead */ 679 if (!np) 680 np = of_find_node_by_name(NULL, "8042"); 681 if (np) { 682 of_i8042_kbd_irq = 1; 683 of_i8042_aux_irq = 12; 684 } 685 break; 686 case FDC_BASE: /* FDC1 */ 687 np = of_find_node_by_type(NULL, "fdc"); 688 break; 689 default: 690 /* ipmi is supposed to fail here */ 691 break; 692 } 693 if (!np) 694 return ret; 695 parent = of_get_parent(np); 696 if (parent) { 697 if (of_node_is_type(parent, "isa")) 698 ret = 0; 699 of_node_put(parent); 700 } 701 of_node_put(np); 702 return ret; 703 } 704 EXPORT_SYMBOL(check_legacy_ioport); 705 706 /* 707 * Panic notifiers setup 708 * 709 * We have 3 notifiers for powerpc, each one from a different "nature": 710 * 711 * - ppc_panic_fadump_handler() is a hypervisor notifier, which hard-disables 712 * IRQs and deal with the Firmware-Assisted dump, when it is configured; 713 * should run early in the panic path. 714 * 715 * - dump_kernel_offset() is an informative notifier, just showing the KASLR 716 * offset if we have RANDOMIZE_BASE set. 717 * 718 * - ppc_panic_platform_handler() is a low-level handler that's registered 719 * only if the platform wishes to perform final actions in the panic path, 720 * hence it should run late and might not even return. Currently, only 721 * pseries and ps3 platforms register callbacks. 722 */ 723 static int ppc_panic_fadump_handler(struct notifier_block *this, 724 unsigned long event, void *ptr) 725 { 726 /* 727 * panic does a local_irq_disable, but we really 728 * want interrupts to be hard disabled. 729 */ 730 hard_irq_disable(); 731 732 /* 733 * If firmware-assisted dump has been registered then trigger 734 * its callback and let the firmware handles everything else. 735 */ 736 crash_fadump(NULL, ptr); 737 738 return NOTIFY_DONE; 739 } 740 741 static int dump_kernel_offset(struct notifier_block *self, unsigned long v, 742 void *p) 743 { 744 pr_emerg("Kernel Offset: 0x%lx from 0x%lx\n", 745 kaslr_offset(), KERNELBASE); 746 747 return NOTIFY_DONE; 748 } 749 750 static int ppc_panic_platform_handler(struct notifier_block *this, 751 unsigned long event, void *ptr) 752 { 753 /* 754 * This handler is only registered if we have a panic callback 755 * on ppc_md, hence NULL check is not needed. 756 * Also, it may not return, so it runs really late on panic path. 757 */ 758 ppc_md.panic(ptr); 759 760 return NOTIFY_DONE; 761 } 762 763 static struct notifier_block ppc_fadump_block = { 764 .notifier_call = ppc_panic_fadump_handler, 765 .priority = INT_MAX, /* run early, to notify the firmware ASAP */ 766 }; 767 768 static struct notifier_block kernel_offset_notifier = { 769 .notifier_call = dump_kernel_offset, 770 }; 771 772 static struct notifier_block ppc_panic_block = { 773 .notifier_call = ppc_panic_platform_handler, 774 .priority = INT_MIN, /* may not return; must be done last */ 775 }; 776 777 void __init setup_panic(void) 778 { 779 /* Hard-disables IRQs + deal with FW-assisted dump (fadump) */ 780 atomic_notifier_chain_register(&panic_notifier_list, 781 &ppc_fadump_block); 782 783 if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && kaslr_offset() > 0) 784 atomic_notifier_chain_register(&panic_notifier_list, 785 &kernel_offset_notifier); 786 787 /* Low-level platform-specific routines that should run on panic */ 788 if (ppc_md.panic) 789 atomic_notifier_chain_register(&panic_notifier_list, 790 &ppc_panic_block); 791 } 792 793 #ifdef CONFIG_CHECK_CACHE_COHERENCY 794 /* 795 * For platforms that have configurable cache-coherency. This function 796 * checks that the cache coherency setting of the kernel matches the setting 797 * left by the firmware, as indicated in the device tree. Since a mismatch 798 * will eventually result in DMA failures, we print * and error and call 799 * BUG() in that case. 800 */ 801 802 #define KERNEL_COHERENCY (!IS_ENABLED(CONFIG_NOT_COHERENT_CACHE)) 803 804 static int __init check_cache_coherency(void) 805 { 806 struct device_node *np; 807 const void *prop; 808 bool devtree_coherency; 809 810 np = of_find_node_by_path("/"); 811 prop = of_get_property(np, "coherency-off", NULL); 812 of_node_put(np); 813 814 devtree_coherency = prop ? false : true; 815 816 if (devtree_coherency != KERNEL_COHERENCY) { 817 printk(KERN_ERR 818 "kernel coherency:%s != device tree_coherency:%s\n", 819 KERNEL_COHERENCY ? "on" : "off", 820 devtree_coherency ? "on" : "off"); 821 BUG(); 822 } 823 824 return 0; 825 } 826 827 late_initcall(check_cache_coherency); 828 #endif /* CONFIG_CHECK_CACHE_COHERENCY */ 829 830 void ppc_printk_progress(char *s, unsigned short hex) 831 { 832 pr_info("%s\n", s); 833 } 834 835 static __init void print_system_info(void) 836 { 837 pr_info("-----------------------------------------------------\n"); 838 pr_info("phys_mem_size = 0x%llx\n", 839 (unsigned long long)memblock_phys_mem_size()); 840 841 pr_info("dcache_bsize = 0x%x\n", dcache_bsize); 842 pr_info("icache_bsize = 0x%x\n", icache_bsize); 843 844 pr_info("cpu_features = 0x%016lx\n", cur_cpu_spec->cpu_features); 845 pr_info(" possible = 0x%016lx\n", 846 (unsigned long)CPU_FTRS_POSSIBLE); 847 pr_info(" always = 0x%016lx\n", 848 (unsigned long)CPU_FTRS_ALWAYS); 849 pr_info("cpu_user_features = 0x%08x 0x%08x\n", 850 cur_cpu_spec->cpu_user_features, 851 cur_cpu_spec->cpu_user_features2); 852 pr_info("mmu_features = 0x%08x\n", cur_cpu_spec->mmu_features); 853 #ifdef CONFIG_PPC64 854 pr_info("firmware_features = 0x%016lx\n", powerpc_firmware_features); 855 #ifdef CONFIG_PPC_BOOK3S 856 pr_info("vmalloc start = 0x%lx\n", KERN_VIRT_START); 857 pr_info("IO start = 0x%lx\n", KERN_IO_START); 858 pr_info("vmemmap start = 0x%lx\n", (unsigned long)vmemmap); 859 #endif 860 #endif 861 862 if (!early_radix_enabled()) 863 print_system_hash_info(); 864 865 if (PHYSICAL_START > 0) 866 pr_info("physical_start = 0x%llx\n", 867 (unsigned long long)PHYSICAL_START); 868 pr_info("-----------------------------------------------------\n"); 869 } 870 871 #ifdef CONFIG_SMP 872 static void __init smp_setup_pacas(void) 873 { 874 int cpu; 875 876 for_each_possible_cpu(cpu) { 877 if (cpu == smp_processor_id()) 878 continue; 879 allocate_paca(cpu); 880 set_hard_smp_processor_id(cpu, cpu_to_phys_id[cpu]); 881 } 882 883 memblock_free(cpu_to_phys_id, nr_cpu_ids * sizeof(u32)); 884 cpu_to_phys_id = NULL; 885 } 886 #endif 887 888 /* 889 * Called into from start_kernel this initializes memblock, which is used 890 * to manage page allocation until mem_init is called. 891 */ 892 void __init setup_arch(char **cmdline_p) 893 { 894 kasan_init(); 895 896 *cmdline_p = boot_command_line; 897 898 /* Set a half-reasonable default so udelay does something sensible */ 899 loops_per_jiffy = 500000000 / HZ; 900 901 /* Unflatten the device-tree passed by prom_init or kexec */ 902 unflatten_device_tree(); 903 904 /* 905 * Initialize cache line/block info from device-tree (on ppc64) or 906 * just cputable (on ppc32). 907 */ 908 initialize_cache_info(); 909 910 /* Initialize RTAS if available. */ 911 rtas_initialize(); 912 913 /* Check if we have an initrd provided via the device-tree. */ 914 check_for_initrd(); 915 916 /* Probe the machine type, establish ppc_md. */ 917 probe_machine(); 918 919 /* Setup panic notifier if requested by the platform. */ 920 setup_panic(); 921 922 /* 923 * Configure ppc_md.power_save (ppc32 only, 64-bit machines do 924 * it from their respective probe() function. 925 */ 926 setup_power_save(); 927 928 /* Discover standard serial ports. */ 929 find_legacy_serial_ports(); 930 931 /* Register early console with the printk subsystem. */ 932 register_early_udbg_console(); 933 934 /* Setup the various CPU maps based on the device-tree. */ 935 smp_setup_cpu_maps(); 936 937 /* Initialize xmon. */ 938 xmon_setup(); 939 940 /* Check the SMT related command line arguments (ppc64). */ 941 check_smt_enabled(); 942 943 /* Parse memory topology */ 944 mem_topology_setup(); 945 946 /* 947 * Release secondary cpus out of their spinloops at 0x60 now that 948 * we can map physical -> logical CPU ids. 949 * 950 * Freescale Book3e parts spin in a loop provided by firmware, 951 * so smp_release_cpus() does nothing for them. 952 */ 953 #ifdef CONFIG_SMP 954 smp_setup_pacas(); 955 956 /* On BookE, setup per-core TLB data structures. */ 957 setup_tlb_core_data(); 958 #endif 959 960 /* Print various info about the machine that has been gathered so far. */ 961 print_system_info(); 962 963 klp_init_thread_info(&init_task); 964 965 setup_initial_init_mm(_stext, _etext, _edata, _end); 966 967 mm_iommu_init(&init_mm); 968 irqstack_early_init(); 969 exc_lvl_early_init(); 970 emergency_stack_init(); 971 972 mce_init(); 973 smp_release_cpus(); 974 975 initmem_init(); 976 977 /* 978 * Reserve large chunks of memory for use by CMA for KVM and hugetlb. These must 979 * be called after initmem_init(), so that pageblock_order is initialised. 980 */ 981 kvm_cma_reserve(); 982 gigantic_hugetlb_cma_reserve(); 983 984 early_memtest(min_low_pfn << PAGE_SHIFT, max_low_pfn << PAGE_SHIFT); 985 986 if (ppc_md.setup_arch) 987 ppc_md.setup_arch(); 988 989 setup_barrier_nospec(); 990 setup_spectre_v2(); 991 992 paging_init(); 993 994 /* Initialize the MMU context management stuff. */ 995 mmu_context_init(); 996 997 /* Interrupt code needs to be 64K-aligned. */ 998 if (IS_ENABLED(CONFIG_PPC64) && (unsigned long)_stext & 0xffff) 999 panic("Kernelbase not 64K-aligned (0x%lx)!\n", 1000 (unsigned long)_stext); 1001 } 1002