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-map-ops.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 #include <linux/hugetlb.h> 53 #include <linux/kmemleak.h> 54 55 #include <asm/archrandom.h> 56 #include <asm/boot_data.h> 57 #include <asm/ipl.h> 58 #include <asm/facility.h> 59 #include <asm/smp.h> 60 #include <asm/mmu_context.h> 61 #include <asm/cpcmd.h> 62 #include <asm/abs_lowcore.h> 63 #include <asm/nmi.h> 64 #include <asm/irq.h> 65 #include <asm/page.h> 66 #include <asm/ptrace.h> 67 #include <asm/sections.h> 68 #include <asm/ebcdic.h> 69 #include <asm/diag.h> 70 #include <asm/os_info.h> 71 #include <asm/sclp.h> 72 #include <asm/stacktrace.h> 73 #include <asm/sysinfo.h> 74 #include <asm/numa.h> 75 #include <asm/alternative.h> 76 #include <asm/nospec-branch.h> 77 #include <asm/physmem_info.h> 78 #include <asm/maccess.h> 79 #include <asm/uv.h> 80 #include <asm/asm-offsets.h> 81 #include "entry.h" 82 83 /* 84 * Machine setup.. 85 */ 86 unsigned int console_mode = 0; 87 EXPORT_SYMBOL(console_mode); 88 89 unsigned int console_devno = -1; 90 EXPORT_SYMBOL(console_devno); 91 92 unsigned int console_irq = -1; 93 EXPORT_SYMBOL(console_irq); 94 95 /* 96 * Some code and data needs to stay below 2 GB, even when the kernel would be 97 * relocated above 2 GB, because it has to use 31 bit addresses. 98 * Such code and data is part of the .amode31 section. 99 */ 100 char __amode31_ref *__samode31 = _samode31; 101 char __amode31_ref *__eamode31 = _eamode31; 102 char __amode31_ref *__stext_amode31 = _stext_amode31; 103 char __amode31_ref *__etext_amode31 = _etext_amode31; 104 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table; 105 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table; 106 107 /* 108 * Control registers CR2, CR5 and CR15 are initialized with addresses 109 * of tables that must be placed below 2G which is handled by the AMODE31 110 * sections. 111 * Because the AMODE31 sections are relocated below 2G at startup, 112 * the content of control registers CR2, CR5 and CR15 must be updated 113 * with new addresses after the relocation. The initial initialization of 114 * control registers occurs in head64.S and then gets updated again after AMODE31 115 * relocation. We must access the relevant AMODE31 tables indirectly via 116 * pointers placed in the .amode31.refs linker section. Those pointers get 117 * updated automatically during AMODE31 relocation and always contain a valid 118 * address within AMODE31 sections. 119 */ 120 121 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64); 122 123 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = { 124 [1] = 0xffffffffffffffff 125 }; 126 127 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = { 128 0x80000000, 0, 0, 0, 129 0x80000000, 0, 0, 0, 130 0x80000000, 0, 0, 0, 131 0x80000000, 0, 0, 0, 132 0x80000000, 0, 0, 0, 133 0x80000000, 0, 0, 0, 134 0x80000000, 0, 0, 0, 135 0x80000000, 0, 0, 0 136 }; 137 138 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = { 139 0, 0, 0x89000000, 0, 140 0, 0, 0x8a000000, 0 141 }; 142 143 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31; 144 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31; 145 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31; 146 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31; 147 148 unsigned long __bootdata_preserved(max_mappable); 149 struct physmem_info __bootdata(physmem_info); 150 151 struct vm_layout __bootdata_preserved(vm_layout); 152 EXPORT_SYMBOL(vm_layout); 153 int __bootdata_preserved(__kaslr_enabled); 154 unsigned int __bootdata_preserved(zlib_dfltcc_support); 155 EXPORT_SYMBOL(zlib_dfltcc_support); 156 u64 __bootdata_preserved(stfle_fac_list[16]); 157 EXPORT_SYMBOL(stfle_fac_list); 158 struct oldmem_data __bootdata_preserved(oldmem_data); 159 160 unsigned long VMALLOC_START; 161 EXPORT_SYMBOL(VMALLOC_START); 162 163 unsigned long VMALLOC_END; 164 EXPORT_SYMBOL(VMALLOC_END); 165 166 struct page *vmemmap; 167 EXPORT_SYMBOL(vmemmap); 168 unsigned long vmemmap_size; 169 170 unsigned long MODULES_VADDR; 171 unsigned long MODULES_END; 172 173 /* An array with a pointer to the lowcore of every CPU. */ 174 struct lowcore *lowcore_ptr[NR_CPUS]; 175 EXPORT_SYMBOL(lowcore_ptr); 176 177 DEFINE_STATIC_KEY_FALSE(cpu_has_bear); 178 179 /* 180 * The Write Back bit position in the physaddr is given by the SLPC PCI. 181 * Leaving the mask zero always uses write through which is safe 182 */ 183 unsigned long mio_wb_bit_mask __ro_after_init; 184 185 /* 186 * This is set up by the setup-routine at boot-time 187 * for S390 need to find out, what we have to setup 188 * using address 0x10400 ... 189 */ 190 191 #include <asm/setup.h> 192 193 /* 194 * condev= and conmode= setup parameter. 195 */ 196 197 static int __init condev_setup(char *str) 198 { 199 int vdev; 200 201 vdev = simple_strtoul(str, &str, 0); 202 if (vdev >= 0 && vdev < 65536) { 203 console_devno = vdev; 204 console_irq = -1; 205 } 206 return 1; 207 } 208 209 __setup("condev=", condev_setup); 210 211 static void __init set_preferred_console(void) 212 { 213 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 214 add_preferred_console("ttyS", 0, NULL); 215 else if (CONSOLE_IS_3270) 216 add_preferred_console("tty3270", 0, NULL); 217 else if (CONSOLE_IS_VT220) 218 add_preferred_console("ttysclp", 0, NULL); 219 else if (CONSOLE_IS_HVC) 220 add_preferred_console("hvc", 0, NULL); 221 } 222 223 static int __init conmode_setup(char *str) 224 { 225 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 226 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 227 SET_CONSOLE_SCLP; 228 #endif 229 #if defined(CONFIG_TN3215_CONSOLE) 230 if (!strcmp(str, "3215")) 231 SET_CONSOLE_3215; 232 #endif 233 #if defined(CONFIG_TN3270_CONSOLE) 234 if (!strcmp(str, "3270")) 235 SET_CONSOLE_3270; 236 #endif 237 set_preferred_console(); 238 return 1; 239 } 240 241 __setup("conmode=", conmode_setup); 242 243 static void __init conmode_default(void) 244 { 245 char query_buffer[1024]; 246 char *ptr; 247 248 if (MACHINE_IS_VM) { 249 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 250 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 251 ptr = strstr(query_buffer, "SUBCHANNEL ="); 252 console_irq = simple_strtoul(ptr + 13, NULL, 16); 253 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 254 ptr = strstr(query_buffer, "CONMODE"); 255 /* 256 * Set the conmode to 3215 so that the device recognition 257 * will set the cu_type of the console to 3215. If the 258 * conmode is 3270 and we don't set it back then both 259 * 3215 and the 3270 driver will try to access the console 260 * device (3215 as console and 3270 as normal tty). 261 */ 262 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 263 if (ptr == NULL) { 264 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 265 SET_CONSOLE_SCLP; 266 #endif 267 return; 268 } 269 if (str_has_prefix(ptr + 8, "3270")) { 270 #if defined(CONFIG_TN3270_CONSOLE) 271 SET_CONSOLE_3270; 272 #elif defined(CONFIG_TN3215_CONSOLE) 273 SET_CONSOLE_3215; 274 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 275 SET_CONSOLE_SCLP; 276 #endif 277 } else if (str_has_prefix(ptr + 8, "3215")) { 278 #if defined(CONFIG_TN3215_CONSOLE) 279 SET_CONSOLE_3215; 280 #elif defined(CONFIG_TN3270_CONSOLE) 281 SET_CONSOLE_3270; 282 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 283 SET_CONSOLE_SCLP; 284 #endif 285 } 286 } else if (MACHINE_IS_KVM) { 287 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 288 SET_CONSOLE_VT220; 289 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 290 SET_CONSOLE_SCLP; 291 else 292 SET_CONSOLE_HVC; 293 } else { 294 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 295 SET_CONSOLE_SCLP; 296 #endif 297 } 298 } 299 300 #ifdef CONFIG_CRASH_DUMP 301 static void __init setup_zfcpdump(void) 302 { 303 if (!is_ipl_type_dump()) 304 return; 305 if (oldmem_data.start) 306 return; 307 strlcat(boot_command_line, " cio_ignore=all,!ipldev,!condev", COMMAND_LINE_SIZE); 308 console_loglevel = 2; 309 } 310 #else 311 static inline void setup_zfcpdump(void) {} 312 #endif /* CONFIG_CRASH_DUMP */ 313 314 /* 315 * Reboot, halt and power_off stubs. They just call _machine_restart, 316 * _machine_halt or _machine_power_off. 317 */ 318 319 void machine_restart(char *command) 320 { 321 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 322 /* 323 * Only unblank the console if we are called in enabled 324 * context or a bust_spinlocks cleared the way for us. 325 */ 326 console_unblank(); 327 _machine_restart(command); 328 } 329 330 void machine_halt(void) 331 { 332 if (!in_interrupt() || oops_in_progress) 333 /* 334 * Only unblank the console if we are called in enabled 335 * context or a bust_spinlocks cleared the way for us. 336 */ 337 console_unblank(); 338 _machine_halt(); 339 } 340 341 void machine_power_off(void) 342 { 343 if (!in_interrupt() || oops_in_progress) 344 /* 345 * Only unblank the console if we are called in enabled 346 * context or a bust_spinlocks cleared the way for us. 347 */ 348 console_unblank(); 349 _machine_power_off(); 350 } 351 352 /* 353 * Dummy power off function. 354 */ 355 void (*pm_power_off)(void) = machine_power_off; 356 EXPORT_SYMBOL_GPL(pm_power_off); 357 358 void *restart_stack; 359 360 unsigned long stack_alloc(void) 361 { 362 #ifdef CONFIG_VMAP_STACK 363 void *ret; 364 365 ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP, 366 NUMA_NO_NODE, __builtin_return_address(0)); 367 kmemleak_not_leak(ret); 368 return (unsigned long)ret; 369 #else 370 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 371 #endif 372 } 373 374 void stack_free(unsigned long stack) 375 { 376 #ifdef CONFIG_VMAP_STACK 377 vfree((void *) stack); 378 #else 379 free_pages(stack, THREAD_SIZE_ORDER); 380 #endif 381 } 382 383 static unsigned long __init stack_alloc_early(void) 384 { 385 unsigned long stack; 386 387 stack = (unsigned long)memblock_alloc_or_panic(THREAD_SIZE, THREAD_SIZE); 388 return stack; 389 } 390 391 static void __init setup_lowcore(void) 392 { 393 struct lowcore *lc, *abs_lc; 394 395 /* 396 * Setup lowcore for boot cpu 397 */ 398 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 399 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 400 if (!lc) 401 panic("%s: Failed to allocate %zu bytes align=%zx\n", 402 __func__, sizeof(*lc), sizeof(*lc)); 403 404 lc->pcpu = (unsigned long)per_cpu_ptr(&pcpu_devices, 0); 405 lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT; 406 lc->restart_psw.addr = __pa(restart_int_handler); 407 lc->external_new_psw.mask = PSW_KERNEL_BITS; 408 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 409 lc->svc_new_psw.mask = PSW_KERNEL_BITS; 410 lc->svc_new_psw.addr = (unsigned long) system_call; 411 lc->program_new_psw.mask = PSW_KERNEL_BITS; 412 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 413 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 414 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 415 lc->io_new_psw.mask = PSW_KERNEL_BITS; 416 lc->io_new_psw.addr = (unsigned long) io_int_handler; 417 lc->clock_comparator = clock_comparator_max; 418 lc->current_task = (unsigned long)&init_task; 419 lc->lpp = LPP_MAGIC; 420 lc->machine_flags = get_lowcore()->machine_flags; 421 lc->preempt_count = get_lowcore()->preempt_count; 422 nmi_alloc_mcesa_early(&lc->mcesad); 423 lc->sys_enter_timer = get_lowcore()->sys_enter_timer; 424 lc->exit_timer = get_lowcore()->exit_timer; 425 lc->user_timer = get_lowcore()->user_timer; 426 lc->system_timer = get_lowcore()->system_timer; 427 lc->steal_timer = get_lowcore()->steal_timer; 428 lc->last_update_timer = get_lowcore()->last_update_timer; 429 lc->last_update_clock = get_lowcore()->last_update_clock; 430 /* 431 * Allocate the global restart stack which is the same for 432 * all CPUs in case *one* of them does a PSW restart. 433 */ 434 restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET); 435 lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET; 436 lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET; 437 lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET; 438 lc->kernel_stack = get_lowcore()->kernel_stack; 439 /* 440 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 441 * restart data to the absolute zero lowcore. This is necessary if 442 * PSW restart is done on an offline CPU that has lowcore zero. 443 */ 444 lc->restart_stack = (unsigned long) restart_stack; 445 lc->restart_fn = (unsigned long) do_restart; 446 lc->restart_data = 0; 447 lc->restart_source = -1U; 448 lc->spinlock_lockval = arch_spin_lockval(0); 449 lc->spinlock_index = 0; 450 arch_spin_lock_setup(0); 451 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 452 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 453 lc->preempt_count = PREEMPT_DISABLED; 454 lc->kernel_asce = get_lowcore()->kernel_asce; 455 lc->user_asce = get_lowcore()->user_asce; 456 457 system_ctlreg_init_save_area(lc); 458 abs_lc = get_abs_lowcore(); 459 abs_lc->restart_stack = lc->restart_stack; 460 abs_lc->restart_fn = lc->restart_fn; 461 abs_lc->restart_data = lc->restart_data; 462 abs_lc->restart_source = lc->restart_source; 463 abs_lc->restart_psw = lc->restart_psw; 464 abs_lc->restart_flags = RESTART_FLAG_CTLREGS; 465 abs_lc->program_new_psw = lc->program_new_psw; 466 abs_lc->mcesad = lc->mcesad; 467 put_abs_lowcore(abs_lc); 468 469 set_prefix(__pa(lc)); 470 lowcore_ptr[0] = lc; 471 if (abs_lowcore_map(0, lowcore_ptr[0], false)) 472 panic("Couldn't setup absolute lowcore"); 473 } 474 475 static struct resource code_resource = { 476 .name = "Kernel code", 477 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 478 }; 479 480 static struct resource data_resource = { 481 .name = "Kernel data", 482 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 483 }; 484 485 static struct resource bss_resource = { 486 .name = "Kernel bss", 487 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 488 }; 489 490 static struct resource __initdata *standard_resources[] = { 491 &code_resource, 492 &data_resource, 493 &bss_resource, 494 }; 495 496 static void __init setup_resources(void) 497 { 498 struct resource *res, *std_res, *sub_res; 499 phys_addr_t start, end; 500 int j; 501 u64 i; 502 503 code_resource.start = __pa_symbol(_text); 504 code_resource.end = __pa_symbol(_etext) - 1; 505 data_resource.start = __pa_symbol(_etext); 506 data_resource.end = __pa_symbol(_edata) - 1; 507 bss_resource.start = __pa_symbol(__bss_start); 508 bss_resource.end = __pa_symbol(__bss_stop) - 1; 509 510 for_each_mem_range(i, &start, &end) { 511 res = memblock_alloc_or_panic(sizeof(*res), 8); 512 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 513 514 res->name = "System RAM"; 515 res->start = start; 516 /* 517 * In memblock, end points to the first byte after the 518 * range while in resources, end points to the last byte in 519 * the range. 520 */ 521 res->end = end - 1; 522 request_resource(&iomem_resource, res); 523 524 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 525 std_res = standard_resources[j]; 526 if (std_res->start < res->start || 527 std_res->start > res->end) 528 continue; 529 if (std_res->end > res->end) { 530 sub_res = memblock_alloc_or_panic(sizeof(*sub_res), 8); 531 *sub_res = *std_res; 532 sub_res->end = res->end; 533 std_res->start = res->end + 1; 534 request_resource(res, sub_res); 535 } else { 536 request_resource(res, std_res); 537 } 538 } 539 } 540 #ifdef CONFIG_CRASH_DUMP 541 /* 542 * Re-add removed crash kernel memory as reserved memory. This makes 543 * sure it will be mapped with the identity mapping and struct pages 544 * will be created, so it can be resized later on. 545 * However add it later since the crash kernel resource should not be 546 * part of the System RAM resource. 547 */ 548 if (crashk_res.end) { 549 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 550 0, MEMBLOCK_NONE); 551 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 552 insert_resource(&iomem_resource, &crashk_res); 553 } 554 #endif 555 } 556 557 static void __init setup_memory_end(void) 558 { 559 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size); 560 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20); 561 } 562 563 #ifdef CONFIG_CRASH_DUMP 564 565 /* 566 * When kdump is enabled, we have to ensure that no memory from the area 567 * [0 - crashkernel memory size] is set offline - it will be exchanged with 568 * the crashkernel memory region when kdump is triggered. The crashkernel 569 * memory region can never get offlined (pages are unmovable). 570 */ 571 static int kdump_mem_notifier(struct notifier_block *nb, 572 unsigned long action, void *data) 573 { 574 struct memory_notify *arg = data; 575 576 if (action != MEM_GOING_OFFLINE) 577 return NOTIFY_OK; 578 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 579 return NOTIFY_BAD; 580 return NOTIFY_OK; 581 } 582 583 static struct notifier_block kdump_mem_nb = { 584 .notifier_call = kdump_mem_notifier, 585 }; 586 587 #endif 588 589 /* 590 * Reserve page tables created by decompressor 591 */ 592 static void __init reserve_pgtables(void) 593 { 594 unsigned long start, end; 595 struct reserved_range *range; 596 597 for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end) 598 memblock_reserve(start, end - start); 599 } 600 601 /* 602 * Reserve memory for kdump kernel to be loaded with kexec 603 */ 604 static void __init reserve_crashkernel(void) 605 { 606 #ifdef CONFIG_CRASH_DUMP 607 unsigned long long crash_base, crash_size; 608 phys_addr_t low, high; 609 int rc; 610 611 rc = parse_crashkernel(boot_command_line, ident_map_size, 612 &crash_size, &crash_base, NULL, NULL); 613 614 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 615 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 616 if (rc || crash_size == 0) 617 return; 618 619 if (memblock.memory.regions[0].size < crash_size) { 620 pr_info("crashkernel reservation failed: %s\n", 621 "first memory chunk must be at least crashkernel size"); 622 return; 623 } 624 625 low = crash_base ?: oldmem_data.start; 626 high = low + crash_size; 627 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) { 628 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 629 crash_base = low; 630 } else { 631 /* Find suitable area in free memory */ 632 low = max_t(unsigned long, crash_size, sclp.hsa_size); 633 high = crash_base ? crash_base + crash_size : ULONG_MAX; 634 635 if (crash_base && crash_base < low) { 636 pr_info("crashkernel reservation failed: %s\n", 637 "crash_base too low"); 638 return; 639 } 640 low = crash_base ?: low; 641 crash_base = memblock_phys_alloc_range(crash_size, 642 KEXEC_CRASH_MEM_ALIGN, 643 low, high); 644 } 645 646 if (!crash_base) { 647 pr_info("crashkernel reservation failed: %s\n", 648 "no suitable area found"); 649 return; 650 } 651 652 if (register_memory_notifier(&kdump_mem_nb)) { 653 memblock_phys_free(crash_base, crash_size); 654 return; 655 } 656 657 if (!oldmem_data.start && MACHINE_IS_VM) 658 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 659 crashk_res.start = crash_base; 660 crashk_res.end = crash_base + crash_size - 1; 661 memblock_remove(crash_base, crash_size); 662 pr_info("Reserving %lluMB of memory at %lluMB " 663 "for crashkernel (System RAM: %luMB)\n", 664 crash_size >> 20, crash_base >> 20, 665 (unsigned long)memblock.memory.total_size >> 20); 666 os_info_crashkernel_add(crash_base, crash_size); 667 #endif 668 } 669 670 /* 671 * Reserve the initrd from being used by memblock 672 */ 673 static void __init reserve_initrd(void) 674 { 675 unsigned long addr, size; 676 677 if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size)) 678 return; 679 initrd_start = (unsigned long)__va(addr); 680 initrd_end = initrd_start + size; 681 memblock_reserve(addr, size); 682 } 683 684 /* 685 * Reserve the memory area used to pass the certificate lists 686 */ 687 static void __init reserve_certificate_list(void) 688 { 689 if (ipl_cert_list_addr) 690 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 691 } 692 693 static void __init reserve_physmem_info(void) 694 { 695 unsigned long addr, size; 696 697 if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size)) 698 memblock_reserve(addr, size); 699 } 700 701 static void __init free_physmem_info(void) 702 { 703 unsigned long addr, size; 704 705 if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size)) 706 memblock_phys_free(addr, size); 707 } 708 709 static void __init memblock_add_physmem_info(void) 710 { 711 unsigned long start, end; 712 int i; 713 714 pr_debug("physmem info source: %s (%hhd)\n", 715 get_physmem_info_source(), physmem_info.info_source); 716 /* keep memblock lists close to the kernel */ 717 memblock_set_bottom_up(true); 718 for_each_physmem_usable_range(i, &start, &end) 719 memblock_add(start, end - start); 720 for_each_physmem_online_range(i, &start, &end) 721 memblock_physmem_add(start, end - start); 722 memblock_set_bottom_up(false); 723 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 724 } 725 726 /* 727 * Reserve memory used for lowcore. 728 */ 729 static void __init reserve_lowcore(void) 730 { 731 void *lowcore_start = get_lowcore(); 732 void *lowcore_end = lowcore_start + sizeof(struct lowcore); 733 void *start, *end; 734 735 if ((void *)__identity_base < lowcore_end) { 736 start = max(lowcore_start, (void *)__identity_base); 737 end = min(lowcore_end, (void *)(__identity_base + ident_map_size)); 738 memblock_reserve(__pa(start), __pa(end)); 739 } 740 } 741 742 /* 743 * Reserve memory used for absolute lowcore/command line/kernel image. 744 */ 745 static void __init reserve_kernel(void) 746 { 747 memblock_reserve(0, STARTUP_NORMAL_OFFSET); 748 memblock_reserve(OLDMEM_BASE, sizeof(unsigned long)); 749 memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long)); 750 memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31); 751 memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP); 752 memblock_reserve(__pa(_stext), _end - _stext); 753 } 754 755 static void __init setup_memory(void) 756 { 757 phys_addr_t start, end; 758 u64 i; 759 760 /* 761 * Init storage key for present memory 762 */ 763 for_each_mem_range(i, &start, &end) 764 storage_key_init_range(start, end); 765 766 psw_set_key(PAGE_DEFAULT_KEY); 767 } 768 769 static void __init relocate_amode31_section(void) 770 { 771 unsigned long amode31_size = __eamode31 - __samode31; 772 long amode31_offset, *ptr; 773 774 amode31_offset = AMODE31_START - (unsigned long)__samode31; 775 pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size); 776 777 /* Move original AMODE31 section to the new one */ 778 memmove((void *)physmem_info.reserved[RR_AMODE31].start, __samode31, amode31_size); 779 /* Zero out the old AMODE31 section to catch invalid accesses within it */ 780 memset(__samode31, 0, amode31_size); 781 782 /* Update all AMODE31 region references */ 783 for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++) 784 *ptr += amode31_offset; 785 } 786 787 /* This must be called after AMODE31 relocation */ 788 static void __init setup_cr(void) 789 { 790 union ctlreg2 cr2; 791 union ctlreg5 cr5; 792 union ctlreg15 cr15; 793 794 __ctl_duct[1] = (unsigned long)__ctl_aste; 795 __ctl_duct[2] = (unsigned long)__ctl_aste; 796 __ctl_duct[4] = (unsigned long)__ctl_duald; 797 798 /* Update control registers CR2, CR5 and CR15 */ 799 local_ctl_store(2, &cr2.reg); 800 local_ctl_store(5, &cr5.reg); 801 local_ctl_store(15, &cr15.reg); 802 cr2.ducto = (unsigned long)__ctl_duct >> 6; 803 cr5.pasteo = (unsigned long)__ctl_duct >> 6; 804 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3; 805 system_ctl_load(2, &cr2.reg); 806 system_ctl_load(5, &cr5.reg); 807 system_ctl_load(15, &cr15.reg); 808 } 809 810 /* 811 * Add system information as device randomness 812 */ 813 static void __init setup_randomness(void) 814 { 815 struct sysinfo_3_2_2 *vmms; 816 817 vmms = memblock_alloc_or_panic(PAGE_SIZE, PAGE_SIZE); 818 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 819 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 820 memblock_free(vmms, PAGE_SIZE); 821 822 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG)) 823 static_branch_enable(&s390_arch_random_available); 824 } 825 826 /* 827 * Issue diagnose 318 to set the control program name and 828 * version codes. 829 */ 830 static void __init setup_control_program_code(void) 831 { 832 union diag318_info diag318_info = { 833 .cpnc = CPNC_LINUX, 834 .cpvc = 0, 835 }; 836 837 if (!sclp.has_diag318) 838 return; 839 840 diag_stat_inc(DIAG_STAT_X318); 841 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 842 } 843 844 /* 845 * Print the component list from the IPL report 846 */ 847 static void __init log_component_list(void) 848 { 849 struct ipl_rb_component_entry *ptr, *end; 850 char *str; 851 852 if (!early_ipl_comp_list_addr) 853 return; 854 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 855 pr_info("Linux is running with Secure-IPL enabled\n"); 856 else 857 pr_info("Linux is running with Secure-IPL disabled\n"); 858 ptr = __va(early_ipl_comp_list_addr); 859 end = (void *) ptr + early_ipl_comp_list_size; 860 pr_info("The IPL report contains the following components:\n"); 861 while (ptr < end) { 862 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 863 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 864 str = "signed, verified"; 865 else 866 str = "signed, verification failed"; 867 } else { 868 str = "not signed"; 869 } 870 pr_info("%016llx - %016llx (%s)\n", 871 ptr->addr, ptr->addr + ptr->len, str); 872 ptr++; 873 } 874 } 875 876 /* 877 * Setup function called from init/main.c just after the banner 878 * was printed. 879 */ 880 881 void __init setup_arch(char **cmdline_p) 882 { 883 /* 884 * print what head.S has found out about the machine 885 */ 886 if (MACHINE_IS_VM) 887 pr_info("Linux is running as a z/VM " 888 "guest operating system in 64-bit mode\n"); 889 else if (MACHINE_IS_KVM) 890 pr_info("Linux is running under KVM in 64-bit mode\n"); 891 else if (MACHINE_IS_LPAR) 892 pr_info("Linux is running natively in 64-bit mode\n"); 893 else 894 pr_info("Linux is running as a guest in 64-bit mode\n"); 895 896 if (have_relocated_lowcore()) 897 pr_info("Lowcore relocated to 0x%px\n", get_lowcore()); 898 899 log_component_list(); 900 901 /* Have one command line that is parsed and saved in /proc/cmdline */ 902 /* boot_command_line has been already set up in early.c */ 903 *cmdline_p = boot_command_line; 904 905 ROOT_DEV = Root_RAM0; 906 907 setup_initial_init_mm(_text, _etext, _edata, _end); 908 909 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 910 nospec_auto_detect(); 911 912 jump_label_init(); 913 parse_early_param(); 914 #ifdef CONFIG_CRASH_DUMP 915 /* Deactivate elfcorehdr= kernel parameter */ 916 elfcorehdr_addr = ELFCORE_ADDR_MAX; 917 #endif 918 919 os_info_init(); 920 setup_ipl(); 921 setup_control_program_code(); 922 923 /* Do some memory reservations *before* memory is added to memblock */ 924 reserve_pgtables(); 925 reserve_lowcore(); 926 reserve_kernel(); 927 reserve_initrd(); 928 reserve_certificate_list(); 929 reserve_physmem_info(); 930 memblock_set_current_limit(ident_map_size); 931 memblock_allow_resize(); 932 933 /* Get information about *all* installed memory */ 934 memblock_add_physmem_info(); 935 936 free_physmem_info(); 937 setup_memory_end(); 938 memblock_dump_all(); 939 setup_memory(); 940 941 relocate_amode31_section(); 942 setup_cr(); 943 setup_uv(); 944 dma_contiguous_reserve(ident_map_size); 945 vmcp_cma_reserve(); 946 if (MACHINE_HAS_EDAT2) 947 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); 948 949 reserve_crashkernel(); 950 #ifdef CONFIG_CRASH_DUMP 951 /* 952 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset. 953 * Therefore CPU and device initialization should be done afterwards. 954 */ 955 smp_save_dump_secondary_cpus(); 956 #endif 957 958 setup_resources(); 959 setup_lowcore(); 960 smp_fill_possible_mask(); 961 cpu_detect_mhz_feature(); 962 cpu_init(); 963 numa_setup(); 964 smp_detect_cpus(); 965 topology_init_early(); 966 967 if (test_facility(193)) 968 static_branch_enable(&cpu_has_bear); 969 970 /* 971 * Create kernel page tables. 972 */ 973 paging_init(); 974 975 /* 976 * After paging_init created the kernel page table, the new PSWs 977 * in lowcore can now run with DAT enabled. 978 */ 979 #ifdef CONFIG_CRASH_DUMP 980 smp_save_dump_ipl_cpu(); 981 #endif 982 983 /* Setup default console */ 984 conmode_default(); 985 set_preferred_console(); 986 987 apply_alternative_instructions(); 988 if (IS_ENABLED(CONFIG_EXPOLINE)) 989 nospec_init_branches(); 990 991 /* Setup zfcp/nvme dump support */ 992 setup_zfcpdump(); 993 994 /* Add system specific data to the random pool */ 995 setup_randomness(); 996 } 997