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/boot_data.h> 54 #include <asm/ipl.h> 55 #include <asm/facility.h> 56 #include <asm/smp.h> 57 #include <asm/mmu_context.h> 58 #include <asm/cpcmd.h> 59 #include <asm/lowcore.h> 60 #include <asm/nmi.h> 61 #include <asm/irq.h> 62 #include <asm/page.h> 63 #include <asm/ptrace.h> 64 #include <asm/sections.h> 65 #include <asm/ebcdic.h> 66 #include <asm/diag.h> 67 #include <asm/os_info.h> 68 #include <asm/sclp.h> 69 #include <asm/stacktrace.h> 70 #include <asm/sysinfo.h> 71 #include <asm/numa.h> 72 #include <asm/alternative.h> 73 #include <asm/nospec-branch.h> 74 #include <asm/mem_detect.h> 75 #include <asm/uv.h> 76 #include "entry.h" 77 78 /* 79 * Machine setup.. 80 */ 81 unsigned int console_mode = 0; 82 EXPORT_SYMBOL(console_mode); 83 84 unsigned int console_devno = -1; 85 EXPORT_SYMBOL(console_devno); 86 87 unsigned int console_irq = -1; 88 EXPORT_SYMBOL(console_irq); 89 90 unsigned long elf_hwcap __read_mostly = 0; 91 char elf_platform[ELF_PLATFORM_SIZE]; 92 93 unsigned long int_hwcap = 0; 94 95 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST 96 int __bootdata_preserved(prot_virt_guest); 97 #endif 98 99 int __bootdata(noexec_disabled); 100 int __bootdata(memory_end_set); 101 unsigned long __bootdata(memory_end); 102 unsigned long __bootdata(vmalloc_size); 103 unsigned long __bootdata(max_physmem_end); 104 struct mem_detect_info __bootdata(mem_detect); 105 106 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table); 107 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table); 108 unsigned long __bootdata_preserved(__swsusp_reset_dma); 109 unsigned long __bootdata_preserved(__stext_dma); 110 unsigned long __bootdata_preserved(__etext_dma); 111 unsigned long __bootdata_preserved(__sdma); 112 unsigned long __bootdata_preserved(__edma); 113 unsigned long __bootdata_preserved(__kaslr_offset); 114 115 unsigned long VMALLOC_START; 116 EXPORT_SYMBOL(VMALLOC_START); 117 118 unsigned long VMALLOC_END; 119 EXPORT_SYMBOL(VMALLOC_END); 120 121 struct page *vmemmap; 122 EXPORT_SYMBOL(vmemmap); 123 124 unsigned long MODULES_VADDR; 125 unsigned long MODULES_END; 126 127 /* An array with a pointer to the lowcore of every CPU. */ 128 struct lowcore *lowcore_ptr[NR_CPUS]; 129 EXPORT_SYMBOL(lowcore_ptr); 130 131 /* 132 * This is set up by the setup-routine at boot-time 133 * for S390 need to find out, what we have to setup 134 * using address 0x10400 ... 135 */ 136 137 #include <asm/setup.h> 138 139 /* 140 * condev= and conmode= setup parameter. 141 */ 142 143 static int __init condev_setup(char *str) 144 { 145 int vdev; 146 147 vdev = simple_strtoul(str, &str, 0); 148 if (vdev >= 0 && vdev < 65536) { 149 console_devno = vdev; 150 console_irq = -1; 151 } 152 return 1; 153 } 154 155 __setup("condev=", condev_setup); 156 157 static void __init set_preferred_console(void) 158 { 159 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 160 add_preferred_console("ttyS", 0, NULL); 161 else if (CONSOLE_IS_3270) 162 add_preferred_console("tty3270", 0, NULL); 163 else if (CONSOLE_IS_VT220) 164 add_preferred_console("ttyS", 1, NULL); 165 else if (CONSOLE_IS_HVC) 166 add_preferred_console("hvc", 0, NULL); 167 } 168 169 static int __init conmode_setup(char *str) 170 { 171 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 172 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 173 SET_CONSOLE_SCLP; 174 #endif 175 #if defined(CONFIG_TN3215_CONSOLE) 176 if (!strcmp(str, "3215")) 177 SET_CONSOLE_3215; 178 #endif 179 #if defined(CONFIG_TN3270_CONSOLE) 180 if (!strcmp(str, "3270")) 181 SET_CONSOLE_3270; 182 #endif 183 set_preferred_console(); 184 return 1; 185 } 186 187 __setup("conmode=", conmode_setup); 188 189 static void __init conmode_default(void) 190 { 191 char query_buffer[1024]; 192 char *ptr; 193 194 if (MACHINE_IS_VM) { 195 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 196 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 197 ptr = strstr(query_buffer, "SUBCHANNEL ="); 198 console_irq = simple_strtoul(ptr + 13, NULL, 16); 199 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 200 ptr = strstr(query_buffer, "CONMODE"); 201 /* 202 * Set the conmode to 3215 so that the device recognition 203 * will set the cu_type of the console to 3215. If the 204 * conmode is 3270 and we don't set it back then both 205 * 3215 and the 3270 driver will try to access the console 206 * device (3215 as console and 3270 as normal tty). 207 */ 208 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 209 if (ptr == NULL) { 210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 211 SET_CONSOLE_SCLP; 212 #endif 213 return; 214 } 215 if (str_has_prefix(ptr + 8, "3270")) { 216 #if defined(CONFIG_TN3270_CONSOLE) 217 SET_CONSOLE_3270; 218 #elif defined(CONFIG_TN3215_CONSOLE) 219 SET_CONSOLE_3215; 220 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 221 SET_CONSOLE_SCLP; 222 #endif 223 } else if (str_has_prefix(ptr + 8, "3215")) { 224 #if defined(CONFIG_TN3215_CONSOLE) 225 SET_CONSOLE_3215; 226 #elif defined(CONFIG_TN3270_CONSOLE) 227 SET_CONSOLE_3270; 228 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 229 SET_CONSOLE_SCLP; 230 #endif 231 } 232 } else if (MACHINE_IS_KVM) { 233 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 234 SET_CONSOLE_VT220; 235 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 236 SET_CONSOLE_SCLP; 237 else 238 SET_CONSOLE_HVC; 239 } else { 240 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 241 SET_CONSOLE_SCLP; 242 #endif 243 } 244 } 245 246 #ifdef CONFIG_CRASH_DUMP 247 static void __init setup_zfcpdump(void) 248 { 249 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 250 return; 251 if (OLDMEM_BASE) 252 return; 253 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 254 console_loglevel = 2; 255 } 256 #else 257 static inline void setup_zfcpdump(void) {} 258 #endif /* CONFIG_CRASH_DUMP */ 259 260 /* 261 * Reboot, halt and power_off stubs. They just call _machine_restart, 262 * _machine_halt or _machine_power_off. 263 */ 264 265 void machine_restart(char *command) 266 { 267 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 268 /* 269 * Only unblank the console if we are called in enabled 270 * context or a bust_spinlocks cleared the way for us. 271 */ 272 console_unblank(); 273 _machine_restart(command); 274 } 275 276 void machine_halt(void) 277 { 278 if (!in_interrupt() || oops_in_progress) 279 /* 280 * Only unblank the console if we are called in enabled 281 * context or a bust_spinlocks cleared the way for us. 282 */ 283 console_unblank(); 284 _machine_halt(); 285 } 286 287 void machine_power_off(void) 288 { 289 if (!in_interrupt() || oops_in_progress) 290 /* 291 * Only unblank the console if we are called in enabled 292 * context or a bust_spinlocks cleared the way for us. 293 */ 294 console_unblank(); 295 _machine_power_off(); 296 } 297 298 /* 299 * Dummy power off function. 300 */ 301 void (*pm_power_off)(void) = machine_power_off; 302 EXPORT_SYMBOL_GPL(pm_power_off); 303 304 void *restart_stack __section(.data); 305 306 unsigned long stack_alloc(void) 307 { 308 #ifdef CONFIG_VMAP_STACK 309 return (unsigned long) 310 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE, 311 VMALLOC_START, VMALLOC_END, 312 THREADINFO_GFP, 313 PAGE_KERNEL, 0, NUMA_NO_NODE, 314 __builtin_return_address(0)); 315 #else 316 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 317 #endif 318 } 319 320 void stack_free(unsigned long stack) 321 { 322 #ifdef CONFIG_VMAP_STACK 323 vfree((void *) stack); 324 #else 325 free_pages(stack, THREAD_SIZE_ORDER); 326 #endif 327 } 328 329 int __init arch_early_irq_init(void) 330 { 331 unsigned long stack; 332 333 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 334 if (!stack) 335 panic("Couldn't allocate async stack"); 336 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 337 return 0; 338 } 339 340 static int __init async_stack_realloc(void) 341 { 342 unsigned long old, new; 343 344 old = S390_lowcore.async_stack - STACK_INIT_OFFSET; 345 new = stack_alloc(); 346 if (!new) 347 panic("Couldn't allocate async stack"); 348 S390_lowcore.async_stack = new + STACK_INIT_OFFSET; 349 free_pages(old, THREAD_SIZE_ORDER); 350 return 0; 351 } 352 early_initcall(async_stack_realloc); 353 354 void __init arch_call_rest_init(void) 355 { 356 unsigned long stack; 357 358 stack = stack_alloc(); 359 if (!stack) 360 panic("Couldn't allocate kernel stack"); 361 current->stack = (void *) stack; 362 #ifdef CONFIG_VMAP_STACK 363 current->stack_vm_area = (void *) stack; 364 #endif 365 set_task_stack_end_magic(current); 366 stack += STACK_INIT_OFFSET; 367 S390_lowcore.kernel_stack = stack; 368 CALL_ON_STACK_NORETURN(rest_init, stack); 369 } 370 371 static void __init setup_lowcore_dat_off(void) 372 { 373 struct lowcore *lc; 374 375 /* 376 * Setup lowcore for boot cpu 377 */ 378 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 379 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 380 if (!lc) 381 panic("%s: Failed to allocate %zu bytes align=%zx\n", 382 __func__, sizeof(*lc), sizeof(*lc)); 383 384 lc->restart_psw.mask = PSW_KERNEL_BITS; 385 lc->restart_psw.addr = (unsigned long) restart_int_handler; 386 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 387 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 388 lc->svc_new_psw.mask = PSW_KERNEL_BITS | 389 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 390 lc->svc_new_psw.addr = (unsigned long) system_call; 391 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 392 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 393 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 394 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 395 lc->io_new_psw.mask = PSW_KERNEL_BITS | 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 if (!restart_stack) 426 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 427 __func__, THREAD_SIZE, THREAD_SIZE); 428 restart_stack += STACK_INIT_OFFSET; 429 430 /* 431 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 432 * restart data to the absolute zero lowcore. This is necessary if 433 * PSW restart is done on an offline CPU that has lowcore zero. 434 */ 435 lc->restart_stack = (unsigned long) restart_stack; 436 lc->restart_fn = (unsigned long) do_restart; 437 lc->restart_data = 0; 438 lc->restart_source = -1UL; 439 440 /* Setup absolute zero lowcore */ 441 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 442 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 443 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 444 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 445 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 446 447 lc->spinlock_lockval = arch_spin_lockval(0); 448 lc->spinlock_index = 0; 449 arch_spin_lock_setup(0); 450 lc->br_r1_trampoline = 0x07f1; /* br %r1 */ 451 452 set_prefix((u32)(unsigned long) lc); 453 lowcore_ptr[0] = lc; 454 } 455 456 static void __init setup_lowcore_dat_on(void) 457 { 458 __ctl_clear_bit(0, 28); 459 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT; 460 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT; 461 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT; 462 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT; 463 __ctl_set_bit(0, 28); 464 } 465 466 static struct resource code_resource = { 467 .name = "Kernel code", 468 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 469 }; 470 471 static struct resource data_resource = { 472 .name = "Kernel data", 473 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 474 }; 475 476 static struct resource bss_resource = { 477 .name = "Kernel bss", 478 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 479 }; 480 481 static struct resource __initdata *standard_resources[] = { 482 &code_resource, 483 &data_resource, 484 &bss_resource, 485 }; 486 487 static void __init setup_resources(void) 488 { 489 struct resource *res, *std_res, *sub_res; 490 struct memblock_region *reg; 491 int j; 492 493 code_resource.start = (unsigned long) _text; 494 code_resource.end = (unsigned long) _etext - 1; 495 data_resource.start = (unsigned long) _etext; 496 data_resource.end = (unsigned long) _edata - 1; 497 bss_resource.start = (unsigned long) __bss_start; 498 bss_resource.end = (unsigned long) __bss_stop - 1; 499 500 for_each_memblock(memory, reg) { 501 res = memblock_alloc(sizeof(*res), 8); 502 if (!res) 503 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 504 __func__, sizeof(*res), 8); 505 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 506 507 res->name = "System RAM"; 508 res->start = reg->base; 509 res->end = reg->base + reg->size - 1; 510 request_resource(&iomem_resource, res); 511 512 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 513 std_res = standard_resources[j]; 514 if (std_res->start < res->start || 515 std_res->start > res->end) 516 continue; 517 if (std_res->end > res->end) { 518 sub_res = memblock_alloc(sizeof(*sub_res), 8); 519 if (!sub_res) 520 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 521 __func__, sizeof(*sub_res), 8); 522 *sub_res = *std_res; 523 sub_res->end = res->end; 524 std_res->start = res->end + 1; 525 request_resource(res, sub_res); 526 } else { 527 request_resource(res, std_res); 528 } 529 } 530 } 531 #ifdef CONFIG_CRASH_DUMP 532 /* 533 * Re-add removed crash kernel memory as reserved memory. This makes 534 * sure it will be mapped with the identity mapping and struct pages 535 * will be created, so it can be resized later on. 536 * However add it later since the crash kernel resource should not be 537 * part of the System RAM resource. 538 */ 539 if (crashk_res.end) { 540 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 541 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 542 insert_resource(&iomem_resource, &crashk_res); 543 } 544 #endif 545 } 546 547 static void __init setup_memory_end(void) 548 { 549 unsigned long vmax, tmp; 550 551 /* Choose kernel address space layout: 3 or 4 levels. */ 552 if (IS_ENABLED(CONFIG_KASAN)) { 553 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING) 554 ? _REGION1_SIZE 555 : _REGION2_SIZE; 556 } else { 557 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 558 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 559 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE) 560 vmax = _REGION2_SIZE; /* 3-level kernel page table */ 561 else 562 vmax = _REGION1_SIZE; /* 4-level kernel page table */ 563 } 564 565 /* module area is at the end of the kernel address space. */ 566 MODULES_END = vmax; 567 MODULES_VADDR = MODULES_END - MODULES_LEN; 568 VMALLOC_END = MODULES_VADDR; 569 VMALLOC_START = VMALLOC_END - vmalloc_size; 570 571 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 572 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 573 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 574 tmp = SECTION_ALIGN_UP(tmp); 575 tmp = VMALLOC_START - tmp * sizeof(struct page); 576 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 577 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 578 vmemmap = (struct page *) tmp; 579 580 /* Take care that memory_end is set and <= vmemmap */ 581 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap); 582 #ifdef CONFIG_KASAN 583 /* fit in kasan shadow memory region between 1:1 and vmemmap */ 584 memory_end = min(memory_end, KASAN_SHADOW_START); 585 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END); 586 #endif 587 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 588 memblock_remove(memory_end, ULONG_MAX); 589 590 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 591 } 592 593 #ifdef CONFIG_CRASH_DUMP 594 595 /* 596 * When kdump is enabled, we have to ensure that no memory from 597 * the area [0 - crashkernel memory size] and 598 * [crashk_res.start - crashk_res.end] is set offline. 599 */ 600 static int kdump_mem_notifier(struct notifier_block *nb, 601 unsigned long action, void *data) 602 { 603 struct memory_notify *arg = data; 604 605 if (action != MEM_GOING_OFFLINE) 606 return NOTIFY_OK; 607 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 608 return NOTIFY_BAD; 609 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 610 return NOTIFY_OK; 611 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 612 return NOTIFY_OK; 613 return NOTIFY_BAD; 614 } 615 616 static struct notifier_block kdump_mem_nb = { 617 .notifier_call = kdump_mem_notifier, 618 }; 619 620 #endif 621 622 /* 623 * Make sure that the area behind memory_end is protected 624 */ 625 static void reserve_memory_end(void) 626 { 627 if (memory_end_set) 628 memblock_reserve(memory_end, ULONG_MAX); 629 } 630 631 /* 632 * Make sure that oldmem, where the dump is stored, is protected 633 */ 634 static void reserve_oldmem(void) 635 { 636 #ifdef CONFIG_CRASH_DUMP 637 if (OLDMEM_BASE) 638 /* Forget all memory above the running kdump system */ 639 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 640 #endif 641 } 642 643 /* 644 * Make sure that oldmem, where the dump is stored, is protected 645 */ 646 static void remove_oldmem(void) 647 { 648 #ifdef CONFIG_CRASH_DUMP 649 if (OLDMEM_BASE) 650 /* Forget all memory above the running kdump system */ 651 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 652 #endif 653 } 654 655 /* 656 * Reserve memory for kdump kernel to be loaded with kexec 657 */ 658 static void __init reserve_crashkernel(void) 659 { 660 #ifdef CONFIG_CRASH_DUMP 661 unsigned long long crash_base, crash_size; 662 phys_addr_t low, high; 663 int rc; 664 665 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 666 &crash_base); 667 668 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 669 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 670 if (rc || crash_size == 0) 671 return; 672 673 if (memblock.memory.regions[0].size < crash_size) { 674 pr_info("crashkernel reservation failed: %s\n", 675 "first memory chunk must be at least crashkernel size"); 676 return; 677 } 678 679 low = crash_base ?: OLDMEM_BASE; 680 high = low + crash_size; 681 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 682 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 683 crash_base = low; 684 } else { 685 /* Find suitable area in free memory */ 686 low = max_t(unsigned long, crash_size, sclp.hsa_size); 687 high = crash_base ? crash_base + crash_size : ULONG_MAX; 688 689 if (crash_base && crash_base < low) { 690 pr_info("crashkernel reservation failed: %s\n", 691 "crash_base too low"); 692 return; 693 } 694 low = crash_base ?: low; 695 crash_base = memblock_find_in_range(low, high, crash_size, 696 KEXEC_CRASH_MEM_ALIGN); 697 } 698 699 if (!crash_base) { 700 pr_info("crashkernel reservation failed: %s\n", 701 "no suitable area found"); 702 return; 703 } 704 705 if (register_memory_notifier(&kdump_mem_nb)) 706 return; 707 708 if (!OLDMEM_BASE && MACHINE_IS_VM) 709 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 710 crashk_res.start = crash_base; 711 crashk_res.end = crash_base + crash_size - 1; 712 memblock_remove(crash_base, crash_size); 713 pr_info("Reserving %lluMB of memory at %lluMB " 714 "for crashkernel (System RAM: %luMB)\n", 715 crash_size >> 20, crash_base >> 20, 716 (unsigned long)memblock.memory.total_size >> 20); 717 os_info_crashkernel_add(crash_base, crash_size); 718 #endif 719 } 720 721 /* 722 * Reserve the initrd from being used by memblock 723 */ 724 static void __init reserve_initrd(void) 725 { 726 #ifdef CONFIG_BLK_DEV_INITRD 727 if (!INITRD_START || !INITRD_SIZE) 728 return; 729 initrd_start = INITRD_START; 730 initrd_end = initrd_start + INITRD_SIZE; 731 memblock_reserve(INITRD_START, INITRD_SIZE); 732 #endif 733 } 734 735 /* 736 * Reserve the memory area used to pass the certificate lists 737 */ 738 static void __init reserve_certificate_list(void) 739 { 740 if (ipl_cert_list_addr) 741 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 742 } 743 744 static void __init reserve_mem_detect_info(void) 745 { 746 unsigned long start, size; 747 748 get_mem_detect_reserved(&start, &size); 749 if (size) 750 memblock_reserve(start, size); 751 } 752 753 static void __init free_mem_detect_info(void) 754 { 755 unsigned long start, size; 756 757 get_mem_detect_reserved(&start, &size); 758 if (size) 759 memblock_free(start, size); 760 } 761 762 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size) 763 { 764 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n", 765 start, start + size - 1); 766 memblock_add_range(&memblock.memory, start, size, 0, 0); 767 memblock_add_range(&memblock.physmem, start, size, 0, 0); 768 } 769 770 static const char * __init get_mem_info_source(void) 771 { 772 switch (mem_detect.info_source) { 773 case MEM_DETECT_SCLP_STOR_INFO: 774 return "sclp storage info"; 775 case MEM_DETECT_DIAG260: 776 return "diag260"; 777 case MEM_DETECT_SCLP_READ_INFO: 778 return "sclp read info"; 779 case MEM_DETECT_BIN_SEARCH: 780 return "binary search"; 781 } 782 return "none"; 783 } 784 785 static void __init memblock_add_mem_detect_info(void) 786 { 787 unsigned long start, end; 788 int i; 789 790 memblock_dbg("physmem info source: %s (%hhd)\n", 791 get_mem_info_source(), mem_detect.info_source); 792 /* keep memblock lists close to the kernel */ 793 memblock_set_bottom_up(true); 794 for_each_mem_detect_block(i, &start, &end) 795 memblock_physmem_add(start, end - start); 796 memblock_set_bottom_up(false); 797 memblock_dump_all(); 798 } 799 800 /* 801 * Check for initrd being in usable memory 802 */ 803 static void __init check_initrd(void) 804 { 805 #ifdef CONFIG_BLK_DEV_INITRD 806 if (INITRD_START && INITRD_SIZE && 807 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 808 pr_err("The initial RAM disk does not fit into the memory\n"); 809 memblock_free(INITRD_START, INITRD_SIZE); 810 initrd_start = initrd_end = 0; 811 } 812 #endif 813 } 814 815 /* 816 * Reserve memory used for lowcore/command line/kernel image. 817 */ 818 static void __init reserve_kernel(void) 819 { 820 unsigned long start_pfn = PFN_UP(__pa(_end)); 821 822 memblock_reserve(0, HEAD_END); 823 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 824 - (unsigned long)_stext); 825 memblock_reserve(__sdma, __edma - __sdma); 826 } 827 828 static void __init setup_memory(void) 829 { 830 struct memblock_region *reg; 831 832 /* 833 * Init storage key for present memory 834 */ 835 for_each_memblock(memory, reg) { 836 storage_key_init_range(reg->base, reg->base + reg->size); 837 } 838 psw_set_key(PAGE_DEFAULT_KEY); 839 840 /* Only cosmetics */ 841 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 842 } 843 844 /* 845 * Setup hardware capabilities. 846 */ 847 static int __init setup_hwcaps(void) 848 { 849 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 850 struct cpuid cpu_id; 851 int i; 852 853 /* 854 * The store facility list bits numbers as found in the principles 855 * of operation are numbered with bit 1UL<<31 as number 0 to 856 * bit 1UL<<0 as number 31. 857 * Bit 0: instructions named N3, "backported" to esa-mode 858 * Bit 2: z/Architecture mode is active 859 * Bit 7: the store-facility-list-extended facility is installed 860 * Bit 17: the message-security assist is installed 861 * Bit 19: the long-displacement facility is installed 862 * Bit 21: the extended-immediate facility is installed 863 * Bit 22: extended-translation facility 3 is installed 864 * Bit 30: extended-translation facility 3 enhancement facility 865 * These get translated to: 866 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 867 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 868 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 869 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 870 */ 871 for (i = 0; i < 6; i++) 872 if (test_facility(stfl_bits[i])) 873 elf_hwcap |= 1UL << i; 874 875 if (test_facility(22) && test_facility(30)) 876 elf_hwcap |= HWCAP_S390_ETF3EH; 877 878 /* 879 * Check for additional facilities with store-facility-list-extended. 880 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 881 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 882 * as stored by stfl, bits 32-xxx contain additional facilities. 883 * How many facility words are stored depends on the number of 884 * doublewords passed to the instruction. The additional facilities 885 * are: 886 * Bit 42: decimal floating point facility is installed 887 * Bit 44: perform floating point operation facility is installed 888 * translated to: 889 * HWCAP_S390_DFP bit 6 (42 && 44). 890 */ 891 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 892 elf_hwcap |= HWCAP_S390_DFP; 893 894 /* 895 * Huge page support HWCAP_S390_HPAGE is bit 7. 896 */ 897 if (MACHINE_HAS_EDAT1) 898 elf_hwcap |= HWCAP_S390_HPAGE; 899 900 /* 901 * 64-bit register support for 31-bit processes 902 * HWCAP_S390_HIGH_GPRS is bit 9. 903 */ 904 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 905 906 /* 907 * Transactional execution support HWCAP_S390_TE is bit 10. 908 */ 909 if (MACHINE_HAS_TE) 910 elf_hwcap |= HWCAP_S390_TE; 911 912 /* 913 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 914 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 915 * instead of facility bit 129. 916 */ 917 if (MACHINE_HAS_VX) { 918 elf_hwcap |= HWCAP_S390_VXRS; 919 if (test_facility(134)) 920 elf_hwcap |= HWCAP_S390_VXRS_EXT; 921 if (test_facility(135)) 922 elf_hwcap |= HWCAP_S390_VXRS_BCD; 923 if (test_facility(148)) 924 elf_hwcap |= HWCAP_S390_VXRS_EXT2; 925 if (test_facility(152)) 926 elf_hwcap |= HWCAP_S390_VXRS_PDE; 927 } 928 if (test_facility(150)) 929 elf_hwcap |= HWCAP_S390_SORT; 930 if (test_facility(151)) 931 elf_hwcap |= HWCAP_S390_DFLT; 932 933 /* 934 * Guarded storage support HWCAP_S390_GS is bit 12. 935 */ 936 if (MACHINE_HAS_GS) 937 elf_hwcap |= HWCAP_S390_GS; 938 939 get_cpu_id(&cpu_id); 940 add_device_randomness(&cpu_id, sizeof(cpu_id)); 941 switch (cpu_id.machine) { 942 case 0x2064: 943 case 0x2066: 944 default: /* Use "z900" as default for 64 bit kernels. */ 945 strcpy(elf_platform, "z900"); 946 break; 947 case 0x2084: 948 case 0x2086: 949 strcpy(elf_platform, "z990"); 950 break; 951 case 0x2094: 952 case 0x2096: 953 strcpy(elf_platform, "z9-109"); 954 break; 955 case 0x2097: 956 case 0x2098: 957 strcpy(elf_platform, "z10"); 958 break; 959 case 0x2817: 960 case 0x2818: 961 strcpy(elf_platform, "z196"); 962 break; 963 case 0x2827: 964 case 0x2828: 965 strcpy(elf_platform, "zEC12"); 966 break; 967 case 0x2964: 968 case 0x2965: 969 strcpy(elf_platform, "z13"); 970 break; 971 case 0x3906: 972 case 0x3907: 973 strcpy(elf_platform, "z14"); 974 break; 975 case 0x8561: 976 case 0x8562: 977 strcpy(elf_platform, "z15"); 978 break; 979 } 980 981 /* 982 * Virtualization support HWCAP_INT_SIE is bit 0. 983 */ 984 if (sclp.has_sief2) 985 int_hwcap |= HWCAP_INT_SIE; 986 987 return 0; 988 } 989 arch_initcall(setup_hwcaps); 990 991 /* 992 * Add system information as device randomness 993 */ 994 static void __init setup_randomness(void) 995 { 996 struct sysinfo_3_2_2 *vmms; 997 998 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE, 999 PAGE_SIZE); 1000 if (!vmms) 1001 panic("Failed to allocate memory for sysinfo structure\n"); 1002 1003 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 1004 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 1005 memblock_free((unsigned long) vmms, PAGE_SIZE); 1006 } 1007 1008 /* 1009 * Find the correct size for the task_struct. This depends on 1010 * the size of the struct fpu at the end of the thread_struct 1011 * which is embedded in the task_struct. 1012 */ 1013 static void __init setup_task_size(void) 1014 { 1015 int task_size = sizeof(struct task_struct); 1016 1017 if (!MACHINE_HAS_VX) { 1018 task_size -= sizeof(__vector128) * __NUM_VXRS; 1019 task_size += sizeof(freg_t) * __NUM_FPRS; 1020 } 1021 arch_task_struct_size = task_size; 1022 } 1023 1024 /* 1025 * Issue diagnose 318 to set the control program name and 1026 * version codes. 1027 */ 1028 static void __init setup_control_program_code(void) 1029 { 1030 union diag318_info diag318_info = { 1031 .cpnc = CPNC_LINUX, 1032 .cpvc_linux = 0, 1033 .cpvc_distro = {0}, 1034 }; 1035 1036 if (!sclp.has_diag318) 1037 return; 1038 1039 diag_stat_inc(DIAG_STAT_X318); 1040 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 1041 } 1042 1043 /* 1044 * Print the component list from the IPL report 1045 */ 1046 static void __init log_component_list(void) 1047 { 1048 struct ipl_rb_component_entry *ptr, *end; 1049 char *str; 1050 1051 if (!early_ipl_comp_list_addr) 1052 return; 1053 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 1054 pr_info("Linux is running with Secure-IPL enabled\n"); 1055 else 1056 pr_info("Linux is running with Secure-IPL disabled\n"); 1057 ptr = (void *) early_ipl_comp_list_addr; 1058 end = (void *) ptr + early_ipl_comp_list_size; 1059 pr_info("The IPL report contains the following components:\n"); 1060 while (ptr < end) { 1061 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 1062 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 1063 str = "signed, verified"; 1064 else 1065 str = "signed, verification failed"; 1066 } else { 1067 str = "not signed"; 1068 } 1069 pr_info("%016llx - %016llx (%s)\n", 1070 ptr->addr, ptr->addr + ptr->len, str); 1071 ptr++; 1072 } 1073 } 1074 1075 /* 1076 * Setup function called from init/main.c just after the banner 1077 * was printed. 1078 */ 1079 1080 void __init setup_arch(char **cmdline_p) 1081 { 1082 /* 1083 * print what head.S has found out about the machine 1084 */ 1085 if (MACHINE_IS_VM) 1086 pr_info("Linux is running as a z/VM " 1087 "guest operating system in 64-bit mode\n"); 1088 else if (MACHINE_IS_KVM) 1089 pr_info("Linux is running under KVM in 64-bit mode\n"); 1090 else if (MACHINE_IS_LPAR) 1091 pr_info("Linux is running natively in 64-bit mode\n"); 1092 else 1093 pr_info("Linux is running as a guest in 64-bit mode\n"); 1094 1095 log_component_list(); 1096 1097 /* Have one command line that is parsed and saved in /proc/cmdline */ 1098 /* boot_command_line has been already set up in early.c */ 1099 *cmdline_p = boot_command_line; 1100 1101 ROOT_DEV = Root_RAM0; 1102 1103 init_mm.start_code = (unsigned long) _text; 1104 init_mm.end_code = (unsigned long) _etext; 1105 init_mm.end_data = (unsigned long) _edata; 1106 init_mm.brk = (unsigned long) _end; 1107 1108 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 1109 nospec_auto_detect(); 1110 1111 parse_early_param(); 1112 #ifdef CONFIG_CRASH_DUMP 1113 /* Deactivate elfcorehdr= kernel parameter */ 1114 elfcorehdr_addr = ELFCORE_ADDR_MAX; 1115 #endif 1116 1117 os_info_init(); 1118 setup_ipl(); 1119 setup_task_size(); 1120 setup_control_program_code(); 1121 1122 /* Do some memory reservations *before* memory is added to memblock */ 1123 reserve_memory_end(); 1124 reserve_oldmem(); 1125 reserve_kernel(); 1126 reserve_initrd(); 1127 reserve_certificate_list(); 1128 reserve_mem_detect_info(); 1129 memblock_allow_resize(); 1130 1131 /* Get information about *all* installed memory */ 1132 memblock_add_mem_detect_info(); 1133 1134 free_mem_detect_info(); 1135 remove_oldmem(); 1136 1137 /* 1138 * Make sure all chunks are MAX_ORDER aligned so we don't need the 1139 * extra checks that HOLES_IN_ZONE would require. 1140 * 1141 * Is this still required? 1142 */ 1143 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 1144 1145 setup_memory_end(); 1146 setup_memory(); 1147 dma_contiguous_reserve(memory_end); 1148 vmcp_cma_reserve(); 1149 1150 check_initrd(); 1151 reserve_crashkernel(); 1152 #ifdef CONFIG_CRASH_DUMP 1153 /* 1154 * Be aware that smp_save_dump_cpus() triggers a system reset. 1155 * Therefore CPU and device initialization should be done afterwards. 1156 */ 1157 smp_save_dump_cpus(); 1158 #endif 1159 1160 setup_resources(); 1161 setup_lowcore_dat_off(); 1162 smp_fill_possible_mask(); 1163 cpu_detect_mhz_feature(); 1164 cpu_init(); 1165 numa_setup(); 1166 smp_detect_cpus(); 1167 topology_init_early(); 1168 1169 /* 1170 * Create kernel page tables and switch to virtual addressing. 1171 */ 1172 paging_init(); 1173 1174 /* 1175 * After paging_init created the kernel page table, the new PSWs 1176 * in lowcore can now run with DAT enabled. 1177 */ 1178 setup_lowcore_dat_on(); 1179 1180 /* Setup default console */ 1181 conmode_default(); 1182 set_preferred_console(); 1183 1184 apply_alternative_instructions(); 1185 if (IS_ENABLED(CONFIG_EXPOLINE)) 1186 nospec_init_branches(); 1187 1188 /* Setup zfcpdump support */ 1189 setup_zfcpdump(); 1190 1191 /* Add system specific data to the random pool */ 1192 setup_randomness(); 1193 } 1194