1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * S390 version 4 * Copyright IBM Corp. 1999, 2012 5 * Author(s): Hartmut Penner (hp@de.ibm.com), 6 * Martin Schwidefsky (schwidefsky@de.ibm.com) 7 * 8 * Derived from "arch/i386/kernel/setup.c" 9 * Copyright (C) 1995, Linus Torvalds 10 */ 11 12 /* 13 * This file handles the architecture-dependent parts of initialization 14 */ 15 16 #define KMSG_COMPONENT "setup" 17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 18 19 #include <linux/errno.h> 20 #include <linux/export.h> 21 #include <linux/sched.h> 22 #include <linux/sched/task.h> 23 #include <linux/cpu.h> 24 #include <linux/kernel.h> 25 #include <linux/memblock.h> 26 #include <linux/mm.h> 27 #include <linux/stddef.h> 28 #include <linux/unistd.h> 29 #include <linux/ptrace.h> 30 #include <linux/random.h> 31 #include <linux/user.h> 32 #include <linux/tty.h> 33 #include <linux/ioport.h> 34 #include <linux/delay.h> 35 #include <linux/init.h> 36 #include <linux/initrd.h> 37 #include <linux/root_dev.h> 38 #include <linux/console.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/dma-contiguous.h> 41 #include <linux/device.h> 42 #include <linux/notifier.h> 43 #include <linux/pfn.h> 44 #include <linux/ctype.h> 45 #include <linux/reboot.h> 46 #include <linux/topology.h> 47 #include <linux/kexec.h> 48 #include <linux/crash_dump.h> 49 #include <linux/memory.h> 50 #include <linux/compat.h> 51 #include <linux/start_kernel.h> 52 53 #include <asm/ipl.h> 54 #include <asm/facility.h> 55 #include <asm/smp.h> 56 #include <asm/mmu_context.h> 57 #include <asm/cpcmd.h> 58 #include <asm/lowcore.h> 59 #include <asm/nmi.h> 60 #include <asm/irq.h> 61 #include <asm/page.h> 62 #include <asm/ptrace.h> 63 #include <asm/sections.h> 64 #include <asm/ebcdic.h> 65 #include <asm/diag.h> 66 #include <asm/os_info.h> 67 #include <asm/sclp.h> 68 #include <asm/sysinfo.h> 69 #include <asm/numa.h> 70 #include <asm/alternative.h> 71 #include <asm/nospec-branch.h> 72 #include <asm/mem_detect.h> 73 #include "entry.h" 74 75 /* 76 * Machine setup.. 77 */ 78 unsigned int console_mode = 0; 79 EXPORT_SYMBOL(console_mode); 80 81 unsigned int console_devno = -1; 82 EXPORT_SYMBOL(console_devno); 83 84 unsigned int console_irq = -1; 85 EXPORT_SYMBOL(console_irq); 86 87 unsigned long elf_hwcap __read_mostly = 0; 88 char elf_platform[ELF_PLATFORM_SIZE]; 89 90 unsigned long int_hwcap = 0; 91 92 int __bootdata(noexec_disabled); 93 int __bootdata(memory_end_set); 94 unsigned long __bootdata(memory_end); 95 unsigned long __bootdata(max_physmem_end); 96 struct mem_detect_info __bootdata(mem_detect); 97 98 unsigned long VMALLOC_START; 99 EXPORT_SYMBOL(VMALLOC_START); 100 101 unsigned long VMALLOC_END; 102 EXPORT_SYMBOL(VMALLOC_END); 103 104 struct page *vmemmap; 105 EXPORT_SYMBOL(vmemmap); 106 107 unsigned long MODULES_VADDR; 108 unsigned long MODULES_END; 109 110 /* An array with a pointer to the lowcore of every CPU. */ 111 struct lowcore *lowcore_ptr[NR_CPUS]; 112 EXPORT_SYMBOL(lowcore_ptr); 113 114 /* 115 * This is set up by the setup-routine at boot-time 116 * for S390 need to find out, what we have to setup 117 * using address 0x10400 ... 118 */ 119 120 #include <asm/setup.h> 121 122 /* 123 * condev= and conmode= setup parameter. 124 */ 125 126 static int __init condev_setup(char *str) 127 { 128 int vdev; 129 130 vdev = simple_strtoul(str, &str, 0); 131 if (vdev >= 0 && vdev < 65536) { 132 console_devno = vdev; 133 console_irq = -1; 134 } 135 return 1; 136 } 137 138 __setup("condev=", condev_setup); 139 140 static void __init set_preferred_console(void) 141 { 142 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 143 add_preferred_console("ttyS", 0, NULL); 144 else if (CONSOLE_IS_3270) 145 add_preferred_console("tty3270", 0, NULL); 146 else if (CONSOLE_IS_VT220) 147 add_preferred_console("ttyS", 1, NULL); 148 else if (CONSOLE_IS_HVC) 149 add_preferred_console("hvc", 0, NULL); 150 } 151 152 static int __init conmode_setup(char *str) 153 { 154 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 155 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0) 156 SET_CONSOLE_SCLP; 157 #endif 158 #if defined(CONFIG_TN3215_CONSOLE) 159 if (strncmp(str, "3215", 5) == 0) 160 SET_CONSOLE_3215; 161 #endif 162 #if defined(CONFIG_TN3270_CONSOLE) 163 if (strncmp(str, "3270", 5) == 0) 164 SET_CONSOLE_3270; 165 #endif 166 set_preferred_console(); 167 return 1; 168 } 169 170 __setup("conmode=", conmode_setup); 171 172 static void __init conmode_default(void) 173 { 174 char query_buffer[1024]; 175 char *ptr; 176 177 if (MACHINE_IS_VM) { 178 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 179 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 180 ptr = strstr(query_buffer, "SUBCHANNEL ="); 181 console_irq = simple_strtoul(ptr + 13, NULL, 16); 182 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 183 ptr = strstr(query_buffer, "CONMODE"); 184 /* 185 * Set the conmode to 3215 so that the device recognition 186 * will set the cu_type of the console to 3215. If the 187 * conmode is 3270 and we don't set it back then both 188 * 3215 and the 3270 driver will try to access the console 189 * device (3215 as console and 3270 as normal tty). 190 */ 191 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 192 if (ptr == NULL) { 193 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 194 SET_CONSOLE_SCLP; 195 #endif 196 return; 197 } 198 if (strncmp(ptr + 8, "3270", 4) == 0) { 199 #if defined(CONFIG_TN3270_CONSOLE) 200 SET_CONSOLE_3270; 201 #elif defined(CONFIG_TN3215_CONSOLE) 202 SET_CONSOLE_3215; 203 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 204 SET_CONSOLE_SCLP; 205 #endif 206 } else if (strncmp(ptr + 8, "3215", 4) == 0) { 207 #if defined(CONFIG_TN3215_CONSOLE) 208 SET_CONSOLE_3215; 209 #elif defined(CONFIG_TN3270_CONSOLE) 210 SET_CONSOLE_3270; 211 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 212 SET_CONSOLE_SCLP; 213 #endif 214 } 215 } else if (MACHINE_IS_KVM) { 216 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 217 SET_CONSOLE_VT220; 218 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 219 SET_CONSOLE_SCLP; 220 else 221 SET_CONSOLE_HVC; 222 } else { 223 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 224 SET_CONSOLE_SCLP; 225 #endif 226 } 227 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE)) 228 conswitchp = &dummy_con; 229 } 230 231 #ifdef CONFIG_CRASH_DUMP 232 static void __init setup_zfcpdump(void) 233 { 234 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 235 return; 236 if (OLDMEM_BASE) 237 return; 238 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 239 console_loglevel = 2; 240 } 241 #else 242 static inline void setup_zfcpdump(void) {} 243 #endif /* CONFIG_CRASH_DUMP */ 244 245 /* 246 * Reboot, halt and power_off stubs. They just call _machine_restart, 247 * _machine_halt or _machine_power_off. 248 */ 249 250 void machine_restart(char *command) 251 { 252 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 253 /* 254 * Only unblank the console if we are called in enabled 255 * context or a bust_spinlocks cleared the way for us. 256 */ 257 console_unblank(); 258 _machine_restart(command); 259 } 260 261 void machine_halt(void) 262 { 263 if (!in_interrupt() || oops_in_progress) 264 /* 265 * Only unblank the console if we are called in enabled 266 * context or a bust_spinlocks cleared the way for us. 267 */ 268 console_unblank(); 269 _machine_halt(); 270 } 271 272 void machine_power_off(void) 273 { 274 if (!in_interrupt() || oops_in_progress) 275 /* 276 * Only unblank the console if we are called in enabled 277 * context or a bust_spinlocks cleared the way for us. 278 */ 279 console_unblank(); 280 _machine_power_off(); 281 } 282 283 /* 284 * Dummy power off function. 285 */ 286 void (*pm_power_off)(void) = machine_power_off; 287 EXPORT_SYMBOL_GPL(pm_power_off); 288 289 static int __init parse_vmalloc(char *arg) 290 { 291 if (!arg) 292 return -EINVAL; 293 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK; 294 return 0; 295 } 296 early_param("vmalloc", parse_vmalloc); 297 298 void *restart_stack __section(.data); 299 300 unsigned long stack_alloc(void) 301 { 302 #ifdef CONFIG_VMAP_STACK 303 return (unsigned long) 304 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE, 305 VMALLOC_START, VMALLOC_END, 306 THREADINFO_GFP, 307 PAGE_KERNEL, 0, NUMA_NO_NODE, 308 __builtin_return_address(0)); 309 #else 310 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 311 #endif 312 } 313 314 void stack_free(unsigned long stack) 315 { 316 #ifdef CONFIG_VMAP_STACK 317 vfree((void *) stack); 318 #else 319 free_pages(stack, THREAD_SIZE_ORDER); 320 #endif 321 } 322 323 int __init arch_early_irq_init(void) 324 { 325 unsigned long stack; 326 327 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 328 if (!stack) 329 panic("Couldn't allocate async stack"); 330 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 331 return 0; 332 } 333 334 static int __init async_stack_realloc(void) 335 { 336 unsigned long old, new; 337 338 old = S390_lowcore.async_stack - STACK_INIT_OFFSET; 339 new = stack_alloc(); 340 if (!new) 341 panic("Couldn't allocate async stack"); 342 S390_lowcore.async_stack = new + STACK_INIT_OFFSET; 343 free_pages(old, THREAD_SIZE_ORDER); 344 return 0; 345 } 346 early_initcall(async_stack_realloc); 347 348 void __init arch_call_rest_init(void) 349 { 350 struct stack_frame *frame; 351 unsigned long stack; 352 353 stack = stack_alloc(); 354 if (!stack) 355 panic("Couldn't allocate kernel stack"); 356 current->stack = (void *) stack; 357 #ifdef CONFIG_VMAP_STACK 358 current->stack_vm_area = (void *) stack; 359 #endif 360 set_task_stack_end_magic(current); 361 stack += STACK_INIT_OFFSET; 362 S390_lowcore.kernel_stack = stack; 363 frame = (struct stack_frame *) stack; 364 memset(frame, 0, sizeof(*frame)); 365 /* Branch to rest_init on the new stack, never returns */ 366 asm volatile( 367 " la 15,0(%[_frame])\n" 368 " jg rest_init\n" 369 : : [_frame] "a" (frame)); 370 } 371 372 static void __init setup_lowcore(void) 373 { 374 struct lowcore *lc; 375 376 /* 377 * Setup lowcore for boot cpu 378 */ 379 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 380 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 381 lc->restart_psw.mask = PSW_KERNEL_BITS; 382 lc->restart_psw.addr = (unsigned long) restart_int_handler; 383 lc->external_new_psw.mask = PSW_KERNEL_BITS | 384 PSW_MASK_DAT | PSW_MASK_MCHECK; 385 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 386 lc->svc_new_psw.mask = PSW_KERNEL_BITS | 387 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 388 lc->svc_new_psw.addr = (unsigned long) system_call; 389 lc->program_new_psw.mask = PSW_KERNEL_BITS | 390 PSW_MASK_DAT | PSW_MASK_MCHECK; 391 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 392 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 393 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 394 lc->io_new_psw.mask = PSW_KERNEL_BITS | 395 PSW_MASK_DAT | PSW_MASK_MCHECK; 396 lc->io_new_psw.addr = (unsigned long) io_int_handler; 397 lc->clock_comparator = clock_comparator_max; 398 lc->nodat_stack = ((unsigned long) &init_thread_union) 399 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 400 lc->current_task = (unsigned long)&init_task; 401 lc->lpp = LPP_MAGIC; 402 lc->machine_flags = S390_lowcore.machine_flags; 403 lc->preempt_count = S390_lowcore.preempt_count; 404 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 405 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 406 sizeof(lc->stfle_fac_list)); 407 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list, 408 sizeof(lc->alt_stfle_fac_list)); 409 nmi_alloc_boot_cpu(lc); 410 vdso_alloc_boot_cpu(lc); 411 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 412 lc->async_enter_timer = S390_lowcore.async_enter_timer; 413 lc->exit_timer = S390_lowcore.exit_timer; 414 lc->user_timer = S390_lowcore.user_timer; 415 lc->system_timer = S390_lowcore.system_timer; 416 lc->steal_timer = S390_lowcore.steal_timer; 417 lc->last_update_timer = S390_lowcore.last_update_timer; 418 lc->last_update_clock = S390_lowcore.last_update_clock; 419 420 /* 421 * Allocate the global restart stack which is the same for 422 * all CPUs in cast *one* of them does a PSW restart. 423 */ 424 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE); 425 restart_stack += STACK_INIT_OFFSET; 426 427 /* 428 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 429 * restart data to the absolute zero lowcore. This is necessary if 430 * PSW restart is done on an offline CPU that has lowcore zero. 431 */ 432 lc->restart_stack = (unsigned long) restart_stack; 433 lc->restart_fn = (unsigned long) do_restart; 434 lc->restart_data = 0; 435 lc->restart_source = -1UL; 436 437 /* Setup absolute zero lowcore */ 438 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 439 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 440 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 441 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 442 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 443 444 #ifdef CONFIG_SMP 445 lc->spinlock_lockval = arch_spin_lockval(0); 446 lc->spinlock_index = 0; 447 arch_spin_lock_setup(0); 448 #endif 449 lc->br_r1_trampoline = 0x07f1; /* br %r1 */ 450 451 set_prefix((u32)(unsigned long) lc); 452 lowcore_ptr[0] = lc; 453 } 454 455 static struct resource code_resource = { 456 .name = "Kernel code", 457 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 458 }; 459 460 static struct resource data_resource = { 461 .name = "Kernel data", 462 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 463 }; 464 465 static struct resource bss_resource = { 466 .name = "Kernel bss", 467 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 468 }; 469 470 static struct resource __initdata *standard_resources[] = { 471 &code_resource, 472 &data_resource, 473 &bss_resource, 474 }; 475 476 static void __init setup_resources(void) 477 { 478 struct resource *res, *std_res, *sub_res; 479 struct memblock_region *reg; 480 int j; 481 482 code_resource.start = (unsigned long) _text; 483 code_resource.end = (unsigned long) _etext - 1; 484 data_resource.start = (unsigned long) _etext; 485 data_resource.end = (unsigned long) _edata - 1; 486 bss_resource.start = (unsigned long) __bss_start; 487 bss_resource.end = (unsigned long) __bss_stop - 1; 488 489 for_each_memblock(memory, reg) { 490 res = memblock_alloc(sizeof(*res), 8); 491 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 492 493 res->name = "System RAM"; 494 res->start = reg->base; 495 res->end = reg->base + reg->size - 1; 496 request_resource(&iomem_resource, res); 497 498 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 499 std_res = standard_resources[j]; 500 if (std_res->start < res->start || 501 std_res->start > res->end) 502 continue; 503 if (std_res->end > res->end) { 504 sub_res = memblock_alloc(sizeof(*sub_res), 8); 505 *sub_res = *std_res; 506 sub_res->end = res->end; 507 std_res->start = res->end + 1; 508 request_resource(res, sub_res); 509 } else { 510 request_resource(res, std_res); 511 } 512 } 513 } 514 #ifdef CONFIG_CRASH_DUMP 515 /* 516 * Re-add removed crash kernel memory as reserved memory. This makes 517 * sure it will be mapped with the identity mapping and struct pages 518 * will be created, so it can be resized later on. 519 * However add it later since the crash kernel resource should not be 520 * part of the System RAM resource. 521 */ 522 if (crashk_res.end) { 523 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 524 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 525 insert_resource(&iomem_resource, &crashk_res); 526 } 527 #endif 528 } 529 530 static void __init setup_memory_end(void) 531 { 532 unsigned long vmax, vmalloc_size, tmp; 533 534 /* Choose kernel address space layout: 3 or 4 levels. */ 535 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN; 536 if (IS_ENABLED(CONFIG_KASAN)) { 537 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING) 538 ? _REGION1_SIZE 539 : _REGION2_SIZE; 540 } else { 541 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 542 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 543 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE) 544 vmax = _REGION2_SIZE; /* 3-level kernel page table */ 545 else 546 vmax = _REGION1_SIZE; /* 4-level kernel page table */ 547 } 548 549 /* module area is at the end of the kernel address space. */ 550 MODULES_END = vmax; 551 MODULES_VADDR = MODULES_END - MODULES_LEN; 552 VMALLOC_END = MODULES_VADDR; 553 VMALLOC_START = VMALLOC_END - vmalloc_size; 554 555 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 556 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 557 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 558 tmp = SECTION_ALIGN_UP(tmp); 559 tmp = VMALLOC_START - tmp * sizeof(struct page); 560 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 561 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 562 vmemmap = (struct page *) tmp; 563 564 /* Take care that memory_end is set and <= vmemmap */ 565 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap); 566 #ifdef CONFIG_KASAN 567 /* fit in kasan shadow memory region between 1:1 and vmemmap */ 568 memory_end = min(memory_end, KASAN_SHADOW_START); 569 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END); 570 #endif 571 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 572 memblock_remove(memory_end, ULONG_MAX); 573 574 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 575 } 576 577 #ifdef CONFIG_CRASH_DUMP 578 579 /* 580 * When kdump is enabled, we have to ensure that no memory from 581 * the area [0 - crashkernel memory size] and 582 * [crashk_res.start - crashk_res.end] is set offline. 583 */ 584 static int kdump_mem_notifier(struct notifier_block *nb, 585 unsigned long action, void *data) 586 { 587 struct memory_notify *arg = data; 588 589 if (action != MEM_GOING_OFFLINE) 590 return NOTIFY_OK; 591 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 592 return NOTIFY_BAD; 593 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 594 return NOTIFY_OK; 595 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 596 return NOTIFY_OK; 597 return NOTIFY_BAD; 598 } 599 600 static struct notifier_block kdump_mem_nb = { 601 .notifier_call = kdump_mem_notifier, 602 }; 603 604 #endif 605 606 /* 607 * Make sure that the area behind memory_end is protected 608 */ 609 static void reserve_memory_end(void) 610 { 611 if (memory_end_set) 612 memblock_reserve(memory_end, ULONG_MAX); 613 } 614 615 /* 616 * Make sure that oldmem, where the dump is stored, is protected 617 */ 618 static void reserve_oldmem(void) 619 { 620 #ifdef CONFIG_CRASH_DUMP 621 if (OLDMEM_BASE) 622 /* Forget all memory above the running kdump system */ 623 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 624 #endif 625 } 626 627 /* 628 * Make sure that oldmem, where the dump is stored, is protected 629 */ 630 static void remove_oldmem(void) 631 { 632 #ifdef CONFIG_CRASH_DUMP 633 if (OLDMEM_BASE) 634 /* Forget all memory above the running kdump system */ 635 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 636 #endif 637 } 638 639 /* 640 * Reserve memory for kdump kernel to be loaded with kexec 641 */ 642 static void __init reserve_crashkernel(void) 643 { 644 #ifdef CONFIG_CRASH_DUMP 645 unsigned long long crash_base, crash_size; 646 phys_addr_t low, high; 647 int rc; 648 649 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 650 &crash_base); 651 652 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 653 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 654 if (rc || crash_size == 0) 655 return; 656 657 if (memblock.memory.regions[0].size < crash_size) { 658 pr_info("crashkernel reservation failed: %s\n", 659 "first memory chunk must be at least crashkernel size"); 660 return; 661 } 662 663 low = crash_base ?: OLDMEM_BASE; 664 high = low + crash_size; 665 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 666 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 667 crash_base = low; 668 } else { 669 /* Find suitable area in free memory */ 670 low = max_t(unsigned long, crash_size, sclp.hsa_size); 671 high = crash_base ? crash_base + crash_size : ULONG_MAX; 672 673 if (crash_base && crash_base < low) { 674 pr_info("crashkernel reservation failed: %s\n", 675 "crash_base too low"); 676 return; 677 } 678 low = crash_base ?: low; 679 crash_base = memblock_find_in_range(low, high, crash_size, 680 KEXEC_CRASH_MEM_ALIGN); 681 } 682 683 if (!crash_base) { 684 pr_info("crashkernel reservation failed: %s\n", 685 "no suitable area found"); 686 return; 687 } 688 689 if (register_memory_notifier(&kdump_mem_nb)) 690 return; 691 692 if (!OLDMEM_BASE && MACHINE_IS_VM) 693 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 694 crashk_res.start = crash_base; 695 crashk_res.end = crash_base + crash_size - 1; 696 memblock_remove(crash_base, crash_size); 697 pr_info("Reserving %lluMB of memory at %lluMB " 698 "for crashkernel (System RAM: %luMB)\n", 699 crash_size >> 20, crash_base >> 20, 700 (unsigned long)memblock.memory.total_size >> 20); 701 os_info_crashkernel_add(crash_base, crash_size); 702 #endif 703 } 704 705 /* 706 * Reserve the initrd from being used by memblock 707 */ 708 static void __init reserve_initrd(void) 709 { 710 #ifdef CONFIG_BLK_DEV_INITRD 711 if (!INITRD_START || !INITRD_SIZE) 712 return; 713 initrd_start = INITRD_START; 714 initrd_end = initrd_start + INITRD_SIZE; 715 memblock_reserve(INITRD_START, INITRD_SIZE); 716 #endif 717 } 718 719 static void __init reserve_mem_detect_info(void) 720 { 721 unsigned long start, size; 722 723 get_mem_detect_reserved(&start, &size); 724 if (size) 725 memblock_reserve(start, size); 726 } 727 728 static void __init free_mem_detect_info(void) 729 { 730 unsigned long start, size; 731 732 get_mem_detect_reserved(&start, &size); 733 if (size) 734 memblock_free(start, size); 735 } 736 737 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size) 738 { 739 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n", 740 start, start + size - 1); 741 memblock_add_range(&memblock.memory, start, size, 0, 0); 742 memblock_add_range(&memblock.physmem, start, size, 0, 0); 743 } 744 745 static const char * __init get_mem_info_source(void) 746 { 747 switch (mem_detect.info_source) { 748 case MEM_DETECT_SCLP_STOR_INFO: 749 return "sclp storage info"; 750 case MEM_DETECT_DIAG260: 751 return "diag260"; 752 case MEM_DETECT_SCLP_READ_INFO: 753 return "sclp read info"; 754 case MEM_DETECT_BIN_SEARCH: 755 return "binary search"; 756 } 757 return "none"; 758 } 759 760 static void __init memblock_add_mem_detect_info(void) 761 { 762 unsigned long start, end; 763 int i; 764 765 memblock_dbg("physmem info source: %s (%hhd)\n", 766 get_mem_info_source(), mem_detect.info_source); 767 /* keep memblock lists close to the kernel */ 768 memblock_set_bottom_up(true); 769 for_each_mem_detect_block(i, &start, &end) 770 memblock_physmem_add(start, end - start); 771 memblock_set_bottom_up(false); 772 memblock_dump_all(); 773 } 774 775 /* 776 * Check for initrd being in usable memory 777 */ 778 static void __init check_initrd(void) 779 { 780 #ifdef CONFIG_BLK_DEV_INITRD 781 if (INITRD_START && INITRD_SIZE && 782 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 783 pr_err("The initial RAM disk does not fit into the memory\n"); 784 memblock_free(INITRD_START, INITRD_SIZE); 785 initrd_start = initrd_end = 0; 786 } 787 #endif 788 } 789 790 /* 791 * Reserve memory used for lowcore/command line/kernel image. 792 */ 793 static void __init reserve_kernel(void) 794 { 795 unsigned long start_pfn = PFN_UP(__pa(_end)); 796 797 #ifdef CONFIG_DMA_API_DEBUG 798 /* 799 * DMA_API_DEBUG code stumbles over addresses from the 800 * range [PARMAREA_END, _stext]. Mark the memory as reserved 801 * so it is not used for CONFIG_DMA_API_DEBUG=y. 802 */ 803 memblock_reserve(0, PFN_PHYS(start_pfn)); 804 #else 805 memblock_reserve(0, PARMAREA_END); 806 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 807 - (unsigned long)_stext); 808 #endif 809 } 810 811 static void __init setup_memory(void) 812 { 813 struct memblock_region *reg; 814 815 /* 816 * Init storage key for present memory 817 */ 818 for_each_memblock(memory, reg) { 819 storage_key_init_range(reg->base, reg->base + reg->size); 820 } 821 psw_set_key(PAGE_DEFAULT_KEY); 822 823 /* Only cosmetics */ 824 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 825 } 826 827 /* 828 * Setup hardware capabilities. 829 */ 830 static int __init setup_hwcaps(void) 831 { 832 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 833 struct cpuid cpu_id; 834 int i; 835 836 /* 837 * The store facility list bits numbers as found in the principles 838 * of operation are numbered with bit 1UL<<31 as number 0 to 839 * bit 1UL<<0 as number 31. 840 * Bit 0: instructions named N3, "backported" to esa-mode 841 * Bit 2: z/Architecture mode is active 842 * Bit 7: the store-facility-list-extended facility is installed 843 * Bit 17: the message-security assist is installed 844 * Bit 19: the long-displacement facility is installed 845 * Bit 21: the extended-immediate facility is installed 846 * Bit 22: extended-translation facility 3 is installed 847 * Bit 30: extended-translation facility 3 enhancement facility 848 * These get translated to: 849 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 850 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 851 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 852 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 853 */ 854 for (i = 0; i < 6; i++) 855 if (test_facility(stfl_bits[i])) 856 elf_hwcap |= 1UL << i; 857 858 if (test_facility(22) && test_facility(30)) 859 elf_hwcap |= HWCAP_S390_ETF3EH; 860 861 /* 862 * Check for additional facilities with store-facility-list-extended. 863 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 864 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 865 * as stored by stfl, bits 32-xxx contain additional facilities. 866 * How many facility words are stored depends on the number of 867 * doublewords passed to the instruction. The additional facilities 868 * are: 869 * Bit 42: decimal floating point facility is installed 870 * Bit 44: perform floating point operation facility is installed 871 * translated to: 872 * HWCAP_S390_DFP bit 6 (42 && 44). 873 */ 874 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 875 elf_hwcap |= HWCAP_S390_DFP; 876 877 /* 878 * Huge page support HWCAP_S390_HPAGE is bit 7. 879 */ 880 if (MACHINE_HAS_EDAT1) 881 elf_hwcap |= HWCAP_S390_HPAGE; 882 883 /* 884 * 64-bit register support for 31-bit processes 885 * HWCAP_S390_HIGH_GPRS is bit 9. 886 */ 887 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 888 889 /* 890 * Transactional execution support HWCAP_S390_TE is bit 10. 891 */ 892 if (MACHINE_HAS_TE) 893 elf_hwcap |= HWCAP_S390_TE; 894 895 /* 896 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 897 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 898 * instead of facility bit 129. 899 */ 900 if (MACHINE_HAS_VX) { 901 elf_hwcap |= HWCAP_S390_VXRS; 902 if (test_facility(134)) 903 elf_hwcap |= HWCAP_S390_VXRS_EXT; 904 if (test_facility(135)) 905 elf_hwcap |= HWCAP_S390_VXRS_BCD; 906 } 907 908 /* 909 * Guarded storage support HWCAP_S390_GS is bit 12. 910 */ 911 if (MACHINE_HAS_GS) 912 elf_hwcap |= HWCAP_S390_GS; 913 914 get_cpu_id(&cpu_id); 915 add_device_randomness(&cpu_id, sizeof(cpu_id)); 916 switch (cpu_id.machine) { 917 case 0x2064: 918 case 0x2066: 919 default: /* Use "z900" as default for 64 bit kernels. */ 920 strcpy(elf_platform, "z900"); 921 break; 922 case 0x2084: 923 case 0x2086: 924 strcpy(elf_platform, "z990"); 925 break; 926 case 0x2094: 927 case 0x2096: 928 strcpy(elf_platform, "z9-109"); 929 break; 930 case 0x2097: 931 case 0x2098: 932 strcpy(elf_platform, "z10"); 933 break; 934 case 0x2817: 935 case 0x2818: 936 strcpy(elf_platform, "z196"); 937 break; 938 case 0x2827: 939 case 0x2828: 940 strcpy(elf_platform, "zEC12"); 941 break; 942 case 0x2964: 943 case 0x2965: 944 strcpy(elf_platform, "z13"); 945 break; 946 case 0x3906: 947 case 0x3907: 948 strcpy(elf_platform, "z14"); 949 break; 950 } 951 952 /* 953 * Virtualization support HWCAP_INT_SIE is bit 0. 954 */ 955 if (sclp.has_sief2) 956 int_hwcap |= HWCAP_INT_SIE; 957 958 return 0; 959 } 960 arch_initcall(setup_hwcaps); 961 962 /* 963 * Add system information as device randomness 964 */ 965 static void __init setup_randomness(void) 966 { 967 struct sysinfo_3_2_2 *vmms; 968 969 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE, 970 PAGE_SIZE); 971 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 972 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 973 memblock_free((unsigned long) vmms, PAGE_SIZE); 974 } 975 976 /* 977 * Find the correct size for the task_struct. This depends on 978 * the size of the struct fpu at the end of the thread_struct 979 * which is embedded in the task_struct. 980 */ 981 static void __init setup_task_size(void) 982 { 983 int task_size = sizeof(struct task_struct); 984 985 if (!MACHINE_HAS_VX) { 986 task_size -= sizeof(__vector128) * __NUM_VXRS; 987 task_size += sizeof(freg_t) * __NUM_FPRS; 988 } 989 arch_task_struct_size = task_size; 990 } 991 992 /* 993 * Setup function called from init/main.c just after the banner 994 * was printed. 995 */ 996 997 void __init setup_arch(char **cmdline_p) 998 { 999 /* 1000 * print what head.S has found out about the machine 1001 */ 1002 if (MACHINE_IS_VM) 1003 pr_info("Linux is running as a z/VM " 1004 "guest operating system in 64-bit mode\n"); 1005 else if (MACHINE_IS_KVM) 1006 pr_info("Linux is running under KVM in 64-bit mode\n"); 1007 else if (MACHINE_IS_LPAR) 1008 pr_info("Linux is running natively in 64-bit mode\n"); 1009 1010 /* Have one command line that is parsed and saved in /proc/cmdline */ 1011 /* boot_command_line has been already set up in early.c */ 1012 *cmdline_p = boot_command_line; 1013 1014 ROOT_DEV = Root_RAM0; 1015 1016 /* Is init_mm really needed? */ 1017 init_mm.start_code = PAGE_OFFSET; 1018 init_mm.end_code = (unsigned long) _etext; 1019 init_mm.end_data = (unsigned long) _edata; 1020 init_mm.brk = (unsigned long) _end; 1021 1022 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 1023 nospec_auto_detect(); 1024 1025 parse_early_param(); 1026 #ifdef CONFIG_CRASH_DUMP 1027 /* Deactivate elfcorehdr= kernel parameter */ 1028 elfcorehdr_addr = ELFCORE_ADDR_MAX; 1029 #endif 1030 1031 os_info_init(); 1032 setup_ipl(); 1033 setup_task_size(); 1034 1035 /* Do some memory reservations *before* memory is added to memblock */ 1036 reserve_memory_end(); 1037 reserve_oldmem(); 1038 reserve_kernel(); 1039 reserve_initrd(); 1040 reserve_mem_detect_info(); 1041 memblock_allow_resize(); 1042 1043 /* Get information about *all* installed memory */ 1044 memblock_add_mem_detect_info(); 1045 1046 free_mem_detect_info(); 1047 remove_oldmem(); 1048 1049 /* 1050 * Make sure all chunks are MAX_ORDER aligned so we don't need the 1051 * extra checks that HOLES_IN_ZONE would require. 1052 * 1053 * Is this still required? 1054 */ 1055 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 1056 1057 setup_memory_end(); 1058 setup_memory(); 1059 dma_contiguous_reserve(memory_end); 1060 vmcp_cma_reserve(); 1061 1062 check_initrd(); 1063 reserve_crashkernel(); 1064 #ifdef CONFIG_CRASH_DUMP 1065 /* 1066 * Be aware that smp_save_dump_cpus() triggers a system reset. 1067 * Therefore CPU and device initialization should be done afterwards. 1068 */ 1069 smp_save_dump_cpus(); 1070 #endif 1071 1072 setup_resources(); 1073 setup_lowcore(); 1074 smp_fill_possible_mask(); 1075 cpu_detect_mhz_feature(); 1076 cpu_init(); 1077 numa_setup(); 1078 smp_detect_cpus(); 1079 topology_init_early(); 1080 1081 /* 1082 * Create kernel page tables and switch to virtual addressing. 1083 */ 1084 paging_init(); 1085 1086 /* Setup default console */ 1087 conmode_default(); 1088 set_preferred_console(); 1089 1090 apply_alternative_instructions(); 1091 if (IS_ENABLED(CONFIG_EXPOLINE)) 1092 nospec_init_branches(); 1093 1094 /* Setup zfcpdump support */ 1095 setup_zfcpdump(); 1096 1097 /* Add system specific data to the random pool */ 1098 setup_randomness(); 1099 } 1100