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 <asm/asm-offsets.h> 77 #include "entry.h" 78 79 /* 80 * Machine setup.. 81 */ 82 unsigned int console_mode = 0; 83 EXPORT_SYMBOL(console_mode); 84 85 unsigned int console_devno = -1; 86 EXPORT_SYMBOL(console_devno); 87 88 unsigned int console_irq = -1; 89 EXPORT_SYMBOL(console_irq); 90 91 unsigned long elf_hwcap __read_mostly = 0; 92 char elf_platform[ELF_PLATFORM_SIZE]; 93 94 unsigned long int_hwcap = 0; 95 96 int __bootdata(noexec_disabled); 97 int __bootdata(memory_end_set); 98 unsigned long __bootdata(memory_end); 99 unsigned long __bootdata(vmalloc_size); 100 unsigned long __bootdata(max_physmem_end); 101 struct mem_detect_info __bootdata(mem_detect); 102 103 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table); 104 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table); 105 unsigned long __bootdata_preserved(__swsusp_reset_dma); 106 unsigned long __bootdata_preserved(__stext_dma); 107 unsigned long __bootdata_preserved(__etext_dma); 108 unsigned long __bootdata_preserved(__sdma); 109 unsigned long __bootdata_preserved(__edma); 110 unsigned long __bootdata_preserved(__kaslr_offset); 111 unsigned int __bootdata_preserved(zlib_dfltcc_support); 112 EXPORT_SYMBOL(zlib_dfltcc_support); 113 114 unsigned long VMALLOC_START; 115 EXPORT_SYMBOL(VMALLOC_START); 116 117 unsigned long VMALLOC_END; 118 EXPORT_SYMBOL(VMALLOC_END); 119 120 struct page *vmemmap; 121 EXPORT_SYMBOL(vmemmap); 122 123 unsigned long MODULES_VADDR; 124 unsigned long MODULES_END; 125 126 /* An array with a pointer to the lowcore of every CPU. */ 127 struct lowcore *lowcore_ptr[NR_CPUS]; 128 EXPORT_SYMBOL(lowcore_ptr); 129 130 /* 131 * This is set up by the setup-routine at boot-time 132 * for S390 need to find out, what we have to setup 133 * using address 0x10400 ... 134 */ 135 136 #include <asm/setup.h> 137 138 /* 139 * condev= and conmode= setup parameter. 140 */ 141 142 static int __init condev_setup(char *str) 143 { 144 int vdev; 145 146 vdev = simple_strtoul(str, &str, 0); 147 if (vdev >= 0 && vdev < 65536) { 148 console_devno = vdev; 149 console_irq = -1; 150 } 151 return 1; 152 } 153 154 __setup("condev=", condev_setup); 155 156 static void __init set_preferred_console(void) 157 { 158 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 159 add_preferred_console("ttyS", 0, NULL); 160 else if (CONSOLE_IS_3270) 161 add_preferred_console("tty3270", 0, NULL); 162 else if (CONSOLE_IS_VT220) 163 add_preferred_console("ttyS", 1, NULL); 164 else if (CONSOLE_IS_HVC) 165 add_preferred_console("hvc", 0, NULL); 166 } 167 168 static int __init conmode_setup(char *str) 169 { 170 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 171 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 172 SET_CONSOLE_SCLP; 173 #endif 174 #if defined(CONFIG_TN3215_CONSOLE) 175 if (!strcmp(str, "3215")) 176 SET_CONSOLE_3215; 177 #endif 178 #if defined(CONFIG_TN3270_CONSOLE) 179 if (!strcmp(str, "3270")) 180 SET_CONSOLE_3270; 181 #endif 182 set_preferred_console(); 183 return 1; 184 } 185 186 __setup("conmode=", conmode_setup); 187 188 static void __init conmode_default(void) 189 { 190 char query_buffer[1024]; 191 char *ptr; 192 193 if (MACHINE_IS_VM) { 194 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 195 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 196 ptr = strstr(query_buffer, "SUBCHANNEL ="); 197 console_irq = simple_strtoul(ptr + 13, NULL, 16); 198 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 199 ptr = strstr(query_buffer, "CONMODE"); 200 /* 201 * Set the conmode to 3215 so that the device recognition 202 * will set the cu_type of the console to 3215. If the 203 * conmode is 3270 and we don't set it back then both 204 * 3215 and the 3270 driver will try to access the console 205 * device (3215 as console and 3270 as normal tty). 206 */ 207 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 208 if (ptr == NULL) { 209 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 210 SET_CONSOLE_SCLP; 211 #endif 212 return; 213 } 214 if (str_has_prefix(ptr + 8, "3270")) { 215 #if defined(CONFIG_TN3270_CONSOLE) 216 SET_CONSOLE_3270; 217 #elif defined(CONFIG_TN3215_CONSOLE) 218 SET_CONSOLE_3215; 219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 220 SET_CONSOLE_SCLP; 221 #endif 222 } else if (str_has_prefix(ptr + 8, "3215")) { 223 #if defined(CONFIG_TN3215_CONSOLE) 224 SET_CONSOLE_3215; 225 #elif defined(CONFIG_TN3270_CONSOLE) 226 SET_CONSOLE_3270; 227 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 228 SET_CONSOLE_SCLP; 229 #endif 230 } 231 } else if (MACHINE_IS_KVM) { 232 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 233 SET_CONSOLE_VT220; 234 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 235 SET_CONSOLE_SCLP; 236 else 237 SET_CONSOLE_HVC; 238 } else { 239 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 240 SET_CONSOLE_SCLP; 241 #endif 242 } 243 } 244 245 #ifdef CONFIG_CRASH_DUMP 246 static void __init setup_zfcpdump(void) 247 { 248 if (ipl_info.type != IPL_TYPE_FCP_DUMP) 249 return; 250 if (OLDMEM_BASE) 251 return; 252 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 253 console_loglevel = 2; 254 } 255 #else 256 static inline void setup_zfcpdump(void) {} 257 #endif /* CONFIG_CRASH_DUMP */ 258 259 /* 260 * Reboot, halt and power_off stubs. They just call _machine_restart, 261 * _machine_halt or _machine_power_off. 262 */ 263 264 void machine_restart(char *command) 265 { 266 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 267 /* 268 * Only unblank the console if we are called in enabled 269 * context or a bust_spinlocks cleared the way for us. 270 */ 271 console_unblank(); 272 _machine_restart(command); 273 } 274 275 void machine_halt(void) 276 { 277 if (!in_interrupt() || oops_in_progress) 278 /* 279 * Only unblank the console if we are called in enabled 280 * context or a bust_spinlocks cleared the way for us. 281 */ 282 console_unblank(); 283 _machine_halt(); 284 } 285 286 void machine_power_off(void) 287 { 288 if (!in_interrupt() || oops_in_progress) 289 /* 290 * Only unblank the console if we are called in enabled 291 * context or a bust_spinlocks cleared the way for us. 292 */ 293 console_unblank(); 294 _machine_power_off(); 295 } 296 297 /* 298 * Dummy power off function. 299 */ 300 void (*pm_power_off)(void) = machine_power_off; 301 EXPORT_SYMBOL_GPL(pm_power_off); 302 303 void *restart_stack __section(.data); 304 305 unsigned long stack_alloc(void) 306 { 307 #ifdef CONFIG_VMAP_STACK 308 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE, 309 THREADINFO_GFP, NUMA_NO_NODE, 310 __builtin_return_address(0)); 311 #else 312 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 313 #endif 314 } 315 316 void stack_free(unsigned long stack) 317 { 318 #ifdef CONFIG_VMAP_STACK 319 vfree((void *) stack); 320 #else 321 free_pages(stack, THREAD_SIZE_ORDER); 322 #endif 323 } 324 325 int __init arch_early_irq_init(void) 326 { 327 unsigned long stack; 328 329 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 330 if (!stack) 331 panic("Couldn't allocate async stack"); 332 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 333 return 0; 334 } 335 336 static int __init async_stack_realloc(void) 337 { 338 unsigned long old, new; 339 340 old = S390_lowcore.async_stack - STACK_INIT_OFFSET; 341 new = stack_alloc(); 342 if (!new) 343 panic("Couldn't allocate async stack"); 344 S390_lowcore.async_stack = new + STACK_INIT_OFFSET; 345 free_pages(old, THREAD_SIZE_ORDER); 346 return 0; 347 } 348 early_initcall(async_stack_realloc); 349 350 void __init arch_call_rest_init(void) 351 { 352 unsigned long stack; 353 354 stack = stack_alloc(); 355 if (!stack) 356 panic("Couldn't allocate kernel stack"); 357 current->stack = (void *) stack; 358 #ifdef CONFIG_VMAP_STACK 359 current->stack_vm_area = (void *) stack; 360 #endif 361 set_task_stack_end_magic(current); 362 stack += STACK_INIT_OFFSET; 363 S390_lowcore.kernel_stack = stack; 364 CALL_ON_STACK_NORETURN(rest_init, stack); 365 } 366 367 static void __init setup_lowcore_dat_off(void) 368 { 369 struct lowcore *lc; 370 371 /* 372 * Setup lowcore for boot cpu 373 */ 374 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 375 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 376 if (!lc) 377 panic("%s: Failed to allocate %zu bytes align=%zx\n", 378 __func__, sizeof(*lc), sizeof(*lc)); 379 380 lc->restart_psw.mask = PSW_KERNEL_BITS; 381 lc->restart_psw.addr = (unsigned long) restart_int_handler; 382 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 383 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 384 lc->svc_new_psw.mask = PSW_KERNEL_BITS | 385 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK; 386 lc->svc_new_psw.addr = (unsigned long) system_call; 387 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 388 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 389 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 390 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 391 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK; 392 lc->io_new_psw.addr = (unsigned long) io_int_handler; 393 lc->clock_comparator = clock_comparator_max; 394 lc->nodat_stack = ((unsigned long) &init_thread_union) 395 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 396 lc->current_task = (unsigned long)&init_task; 397 lc->lpp = LPP_MAGIC; 398 lc->machine_flags = S390_lowcore.machine_flags; 399 lc->preempt_count = S390_lowcore.preempt_count; 400 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 401 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 402 sizeof(lc->stfle_fac_list)); 403 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list, 404 sizeof(lc->alt_stfle_fac_list)); 405 nmi_alloc_boot_cpu(lc); 406 vdso_alloc_boot_cpu(lc); 407 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 408 lc->async_enter_timer = S390_lowcore.async_enter_timer; 409 lc->exit_timer = S390_lowcore.exit_timer; 410 lc->user_timer = S390_lowcore.user_timer; 411 lc->system_timer = S390_lowcore.system_timer; 412 lc->steal_timer = S390_lowcore.steal_timer; 413 lc->last_update_timer = S390_lowcore.last_update_timer; 414 lc->last_update_clock = S390_lowcore.last_update_clock; 415 416 /* 417 * Allocate the global restart stack which is the same for 418 * all CPUs in cast *one* of them does a PSW restart. 419 */ 420 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE); 421 if (!restart_stack) 422 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 423 __func__, THREAD_SIZE, THREAD_SIZE); 424 restart_stack += STACK_INIT_OFFSET; 425 426 /* 427 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 428 * restart data to the absolute zero lowcore. This is necessary if 429 * PSW restart is done on an offline CPU that has lowcore zero. 430 */ 431 lc->restart_stack = (unsigned long) restart_stack; 432 lc->restart_fn = (unsigned long) do_restart; 433 lc->restart_data = 0; 434 lc->restart_source = -1UL; 435 436 /* Setup absolute zero lowcore */ 437 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 438 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 439 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 440 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 441 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 442 443 lc->spinlock_lockval = arch_spin_lockval(0); 444 lc->spinlock_index = 0; 445 arch_spin_lock_setup(0); 446 lc->br_r1_trampoline = 0x07f1; /* br %r1 */ 447 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 448 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 449 450 set_prefix((u32)(unsigned long) lc); 451 lowcore_ptr[0] = lc; 452 } 453 454 static void __init setup_lowcore_dat_on(void) 455 { 456 __ctl_clear_bit(0, 28); 457 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT; 458 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT; 459 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT; 460 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT; 461 __ctl_set_bit(0, 28); 462 } 463 464 static struct resource code_resource = { 465 .name = "Kernel code", 466 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 467 }; 468 469 static struct resource data_resource = { 470 .name = "Kernel data", 471 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 472 }; 473 474 static struct resource bss_resource = { 475 .name = "Kernel bss", 476 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 477 }; 478 479 static struct resource __initdata *standard_resources[] = { 480 &code_resource, 481 &data_resource, 482 &bss_resource, 483 }; 484 485 static void __init setup_resources(void) 486 { 487 struct resource *res, *std_res, *sub_res; 488 struct memblock_region *reg; 489 int j; 490 491 code_resource.start = (unsigned long) _text; 492 code_resource.end = (unsigned long) _etext - 1; 493 data_resource.start = (unsigned long) _etext; 494 data_resource.end = (unsigned long) _edata - 1; 495 bss_resource.start = (unsigned long) __bss_start; 496 bss_resource.end = (unsigned long) __bss_stop - 1; 497 498 for_each_memblock(memory, reg) { 499 res = memblock_alloc(sizeof(*res), 8); 500 if (!res) 501 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 502 __func__, sizeof(*res), 8); 503 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 504 505 res->name = "System RAM"; 506 res->start = reg->base; 507 res->end = reg->base + reg->size - 1; 508 request_resource(&iomem_resource, res); 509 510 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 511 std_res = standard_resources[j]; 512 if (std_res->start < res->start || 513 std_res->start > res->end) 514 continue; 515 if (std_res->end > res->end) { 516 sub_res = memblock_alloc(sizeof(*sub_res), 8); 517 if (!sub_res) 518 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 519 __func__, sizeof(*sub_res), 8); 520 *sub_res = *std_res; 521 sub_res->end = res->end; 522 std_res->start = res->end + 1; 523 request_resource(res, sub_res); 524 } else { 525 request_resource(res, std_res); 526 } 527 } 528 } 529 #ifdef CONFIG_CRASH_DUMP 530 /* 531 * Re-add removed crash kernel memory as reserved memory. This makes 532 * sure it will be mapped with the identity mapping and struct pages 533 * will be created, so it can be resized later on. 534 * However add it later since the crash kernel resource should not be 535 * part of the System RAM resource. 536 */ 537 if (crashk_res.end) { 538 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 539 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 540 insert_resource(&iomem_resource, &crashk_res); 541 } 542 #endif 543 } 544 545 static void __init setup_memory_end(void) 546 { 547 unsigned long vmax, tmp; 548 549 /* Choose kernel address space layout: 3 or 4 levels. */ 550 if (IS_ENABLED(CONFIG_KASAN)) { 551 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING) 552 ? _REGION1_SIZE 553 : _REGION2_SIZE; 554 } else { 555 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE; 556 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 557 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE) 558 vmax = _REGION2_SIZE; /* 3-level kernel page table */ 559 else 560 vmax = _REGION1_SIZE; /* 4-level kernel page table */ 561 } 562 563 if (is_prot_virt_host()) 564 adjust_to_uv_max(&vmax); 565 566 /* module area is at the end of the kernel address space. */ 567 MODULES_END = vmax; 568 MODULES_VADDR = MODULES_END - MODULES_LEN; 569 VMALLOC_END = MODULES_VADDR; 570 VMALLOC_START = VMALLOC_END - vmalloc_size; 571 572 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 573 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 574 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 575 tmp = SECTION_ALIGN_UP(tmp); 576 tmp = VMALLOC_START - tmp * sizeof(struct page); 577 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 578 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 579 vmemmap = (struct page *) tmp; 580 581 /* Take care that memory_end is set and <= vmemmap */ 582 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap); 583 #ifdef CONFIG_KASAN 584 /* fit in kasan shadow memory region between 1:1 and vmemmap */ 585 memory_end = min(memory_end, KASAN_SHADOW_START); 586 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END); 587 #endif 588 max_pfn = max_low_pfn = PFN_DOWN(memory_end); 589 memblock_remove(memory_end, ULONG_MAX); 590 591 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20); 592 } 593 594 #ifdef CONFIG_CRASH_DUMP 595 596 /* 597 * When kdump is enabled, we have to ensure that no memory from 598 * the area [0 - crashkernel memory size] and 599 * [crashk_res.start - crashk_res.end] is set offline. 600 */ 601 static int kdump_mem_notifier(struct notifier_block *nb, 602 unsigned long action, void *data) 603 { 604 struct memory_notify *arg = data; 605 606 if (action != MEM_GOING_OFFLINE) 607 return NOTIFY_OK; 608 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 609 return NOTIFY_BAD; 610 if (arg->start_pfn > PFN_DOWN(crashk_res.end)) 611 return NOTIFY_OK; 612 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start)) 613 return NOTIFY_OK; 614 return NOTIFY_BAD; 615 } 616 617 static struct notifier_block kdump_mem_nb = { 618 .notifier_call = kdump_mem_notifier, 619 }; 620 621 #endif 622 623 /* 624 * Make sure that the area behind memory_end is protected 625 */ 626 static void reserve_memory_end(void) 627 { 628 if (memory_end_set) 629 memblock_reserve(memory_end, ULONG_MAX); 630 } 631 632 /* 633 * Make sure that oldmem, where the dump is stored, is protected 634 */ 635 static void reserve_oldmem(void) 636 { 637 #ifdef CONFIG_CRASH_DUMP 638 if (OLDMEM_BASE) 639 /* Forget all memory above the running kdump system */ 640 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 641 #endif 642 } 643 644 /* 645 * Make sure that oldmem, where the dump is stored, is protected 646 */ 647 static void remove_oldmem(void) 648 { 649 #ifdef CONFIG_CRASH_DUMP 650 if (OLDMEM_BASE) 651 /* Forget all memory above the running kdump system */ 652 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 653 #endif 654 } 655 656 /* 657 * Reserve memory for kdump kernel to be loaded with kexec 658 */ 659 static void __init reserve_crashkernel(void) 660 { 661 #ifdef CONFIG_CRASH_DUMP 662 unsigned long long crash_base, crash_size; 663 phys_addr_t low, high; 664 int rc; 665 666 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size, 667 &crash_base); 668 669 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 670 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 671 if (rc || crash_size == 0) 672 return; 673 674 if (memblock.memory.regions[0].size < crash_size) { 675 pr_info("crashkernel reservation failed: %s\n", 676 "first memory chunk must be at least crashkernel size"); 677 return; 678 } 679 680 low = crash_base ?: OLDMEM_BASE; 681 high = low + crash_size; 682 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 683 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 684 crash_base = low; 685 } else { 686 /* Find suitable area in free memory */ 687 low = max_t(unsigned long, crash_size, sclp.hsa_size); 688 high = crash_base ? crash_base + crash_size : ULONG_MAX; 689 690 if (crash_base && crash_base < low) { 691 pr_info("crashkernel reservation failed: %s\n", 692 "crash_base too low"); 693 return; 694 } 695 low = crash_base ?: low; 696 crash_base = memblock_find_in_range(low, high, crash_size, 697 KEXEC_CRASH_MEM_ALIGN); 698 } 699 700 if (!crash_base) { 701 pr_info("crashkernel reservation failed: %s\n", 702 "no suitable area found"); 703 return; 704 } 705 706 if (register_memory_notifier(&kdump_mem_nb)) 707 return; 708 709 if (!OLDMEM_BASE && MACHINE_IS_VM) 710 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 711 crashk_res.start = crash_base; 712 crashk_res.end = crash_base + crash_size - 1; 713 memblock_remove(crash_base, crash_size); 714 pr_info("Reserving %lluMB of memory at %lluMB " 715 "for crashkernel (System RAM: %luMB)\n", 716 crash_size >> 20, crash_base >> 20, 717 (unsigned long)memblock.memory.total_size >> 20); 718 os_info_crashkernel_add(crash_base, crash_size); 719 #endif 720 } 721 722 /* 723 * Reserve the initrd from being used by memblock 724 */ 725 static void __init reserve_initrd(void) 726 { 727 #ifdef CONFIG_BLK_DEV_INITRD 728 if (!INITRD_START || !INITRD_SIZE) 729 return; 730 initrd_start = INITRD_START; 731 initrd_end = initrd_start + INITRD_SIZE; 732 memblock_reserve(INITRD_START, INITRD_SIZE); 733 #endif 734 } 735 736 /* 737 * Reserve the memory area used to pass the certificate lists 738 */ 739 static void __init reserve_certificate_list(void) 740 { 741 if (ipl_cert_list_addr) 742 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 743 } 744 745 static void __init reserve_mem_detect_info(void) 746 { 747 unsigned long start, size; 748 749 get_mem_detect_reserved(&start, &size); 750 if (size) 751 memblock_reserve(start, size); 752 } 753 754 static void __init free_mem_detect_info(void) 755 { 756 unsigned long start, size; 757 758 get_mem_detect_reserved(&start, &size); 759 if (size) 760 memblock_free(start, size); 761 } 762 763 static const char * __init get_mem_info_source(void) 764 { 765 switch (mem_detect.info_source) { 766 case MEM_DETECT_SCLP_STOR_INFO: 767 return "sclp storage info"; 768 case MEM_DETECT_DIAG260: 769 return "diag260"; 770 case MEM_DETECT_SCLP_READ_INFO: 771 return "sclp read info"; 772 case MEM_DETECT_BIN_SEARCH: 773 return "binary search"; 774 } 775 return "none"; 776 } 777 778 static void __init memblock_add_mem_detect_info(void) 779 { 780 unsigned long start, end; 781 int i; 782 783 memblock_dbg("physmem info source: %s (%hhd)\n", 784 get_mem_info_source(), mem_detect.info_source); 785 /* keep memblock lists close to the kernel */ 786 memblock_set_bottom_up(true); 787 for_each_mem_detect_block(i, &start, &end) { 788 memblock_add(start, end - start); 789 memblock_physmem_add(start, end - start); 790 } 791 memblock_set_bottom_up(false); 792 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 793 memblock_dump_all(); 794 } 795 796 /* 797 * Check for initrd being in usable memory 798 */ 799 static void __init check_initrd(void) 800 { 801 #ifdef CONFIG_BLK_DEV_INITRD 802 if (INITRD_START && INITRD_SIZE && 803 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 804 pr_err("The initial RAM disk does not fit into the memory\n"); 805 memblock_free(INITRD_START, INITRD_SIZE); 806 initrd_start = initrd_end = 0; 807 } 808 #endif 809 } 810 811 /* 812 * Reserve memory used for lowcore/command line/kernel image. 813 */ 814 static void __init reserve_kernel(void) 815 { 816 unsigned long start_pfn = PFN_UP(__pa(_end)); 817 818 memblock_reserve(0, HEAD_END); 819 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 820 - (unsigned long)_stext); 821 memblock_reserve(__sdma, __edma - __sdma); 822 } 823 824 static void __init setup_memory(void) 825 { 826 struct memblock_region *reg; 827 828 /* 829 * Init storage key for present memory 830 */ 831 for_each_memblock(memory, reg) { 832 storage_key_init_range(reg->base, reg->base + reg->size); 833 } 834 psw_set_key(PAGE_DEFAULT_KEY); 835 836 /* Only cosmetics */ 837 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 838 } 839 840 /* 841 * Setup hardware capabilities. 842 */ 843 static int __init setup_hwcaps(void) 844 { 845 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 846 struct cpuid cpu_id; 847 int i; 848 849 /* 850 * The store facility list bits numbers as found in the principles 851 * of operation are numbered with bit 1UL<<31 as number 0 to 852 * bit 1UL<<0 as number 31. 853 * Bit 0: instructions named N3, "backported" to esa-mode 854 * Bit 2: z/Architecture mode is active 855 * Bit 7: the store-facility-list-extended facility is installed 856 * Bit 17: the message-security assist is installed 857 * Bit 19: the long-displacement facility is installed 858 * Bit 21: the extended-immediate facility is installed 859 * Bit 22: extended-translation facility 3 is installed 860 * Bit 30: extended-translation facility 3 enhancement facility 861 * These get translated to: 862 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 863 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 864 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 865 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 866 */ 867 for (i = 0; i < 6; i++) 868 if (test_facility(stfl_bits[i])) 869 elf_hwcap |= 1UL << i; 870 871 if (test_facility(22) && test_facility(30)) 872 elf_hwcap |= HWCAP_S390_ETF3EH; 873 874 /* 875 * Check for additional facilities with store-facility-list-extended. 876 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 877 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 878 * as stored by stfl, bits 32-xxx contain additional facilities. 879 * How many facility words are stored depends on the number of 880 * doublewords passed to the instruction. The additional facilities 881 * are: 882 * Bit 42: decimal floating point facility is installed 883 * Bit 44: perform floating point operation facility is installed 884 * translated to: 885 * HWCAP_S390_DFP bit 6 (42 && 44). 886 */ 887 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 888 elf_hwcap |= HWCAP_S390_DFP; 889 890 /* 891 * Huge page support HWCAP_S390_HPAGE is bit 7. 892 */ 893 if (MACHINE_HAS_EDAT1) 894 elf_hwcap |= HWCAP_S390_HPAGE; 895 896 /* 897 * 64-bit register support for 31-bit processes 898 * HWCAP_S390_HIGH_GPRS is bit 9. 899 */ 900 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 901 902 /* 903 * Transactional execution support HWCAP_S390_TE is bit 10. 904 */ 905 if (MACHINE_HAS_TE) 906 elf_hwcap |= HWCAP_S390_TE; 907 908 /* 909 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 910 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 911 * instead of facility bit 129. 912 */ 913 if (MACHINE_HAS_VX) { 914 elf_hwcap |= HWCAP_S390_VXRS; 915 if (test_facility(134)) 916 elf_hwcap |= HWCAP_S390_VXRS_EXT; 917 if (test_facility(135)) 918 elf_hwcap |= HWCAP_S390_VXRS_BCD; 919 if (test_facility(148)) 920 elf_hwcap |= HWCAP_S390_VXRS_EXT2; 921 if (test_facility(152)) 922 elf_hwcap |= HWCAP_S390_VXRS_PDE; 923 } 924 if (test_facility(150)) 925 elf_hwcap |= HWCAP_S390_SORT; 926 if (test_facility(151)) 927 elf_hwcap |= HWCAP_S390_DFLT; 928 929 /* 930 * Guarded storage support HWCAP_S390_GS is bit 12. 931 */ 932 if (MACHINE_HAS_GS) 933 elf_hwcap |= HWCAP_S390_GS; 934 935 get_cpu_id(&cpu_id); 936 add_device_randomness(&cpu_id, sizeof(cpu_id)); 937 switch (cpu_id.machine) { 938 case 0x2064: 939 case 0x2066: 940 default: /* Use "z900" as default for 64 bit kernels. */ 941 strcpy(elf_platform, "z900"); 942 break; 943 case 0x2084: 944 case 0x2086: 945 strcpy(elf_platform, "z990"); 946 break; 947 case 0x2094: 948 case 0x2096: 949 strcpy(elf_platform, "z9-109"); 950 break; 951 case 0x2097: 952 case 0x2098: 953 strcpy(elf_platform, "z10"); 954 break; 955 case 0x2817: 956 case 0x2818: 957 strcpy(elf_platform, "z196"); 958 break; 959 case 0x2827: 960 case 0x2828: 961 strcpy(elf_platform, "zEC12"); 962 break; 963 case 0x2964: 964 case 0x2965: 965 strcpy(elf_platform, "z13"); 966 break; 967 case 0x3906: 968 case 0x3907: 969 strcpy(elf_platform, "z14"); 970 break; 971 case 0x8561: 972 case 0x8562: 973 strcpy(elf_platform, "z15"); 974 break; 975 } 976 977 /* 978 * Virtualization support HWCAP_INT_SIE is bit 0. 979 */ 980 if (sclp.has_sief2) 981 int_hwcap |= HWCAP_INT_SIE; 982 983 return 0; 984 } 985 arch_initcall(setup_hwcaps); 986 987 /* 988 * Add system information as device randomness 989 */ 990 static void __init setup_randomness(void) 991 { 992 struct sysinfo_3_2_2 *vmms; 993 994 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE, 995 PAGE_SIZE); 996 if (!vmms) 997 panic("Failed to allocate memory for sysinfo structure\n"); 998 999 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 1000 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 1001 memblock_free((unsigned long) vmms, PAGE_SIZE); 1002 } 1003 1004 /* 1005 * Find the correct size for the task_struct. This depends on 1006 * the size of the struct fpu at the end of the thread_struct 1007 * which is embedded in the task_struct. 1008 */ 1009 static void __init setup_task_size(void) 1010 { 1011 int task_size = sizeof(struct task_struct); 1012 1013 if (!MACHINE_HAS_VX) { 1014 task_size -= sizeof(__vector128) * __NUM_VXRS; 1015 task_size += sizeof(freg_t) * __NUM_FPRS; 1016 } 1017 arch_task_struct_size = task_size; 1018 } 1019 1020 /* 1021 * Issue diagnose 318 to set the control program name and 1022 * version codes. 1023 */ 1024 static void __init setup_control_program_code(void) 1025 { 1026 union diag318_info diag318_info = { 1027 .cpnc = CPNC_LINUX, 1028 .cpvc_linux = 0, 1029 .cpvc_distro = {0}, 1030 }; 1031 1032 if (!sclp.has_diag318) 1033 return; 1034 1035 diag_stat_inc(DIAG_STAT_X318); 1036 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 1037 } 1038 1039 /* 1040 * Print the component list from the IPL report 1041 */ 1042 static void __init log_component_list(void) 1043 { 1044 struct ipl_rb_component_entry *ptr, *end; 1045 char *str; 1046 1047 if (!early_ipl_comp_list_addr) 1048 return; 1049 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 1050 pr_info("Linux is running with Secure-IPL enabled\n"); 1051 else 1052 pr_info("Linux is running with Secure-IPL disabled\n"); 1053 ptr = (void *) early_ipl_comp_list_addr; 1054 end = (void *) ptr + early_ipl_comp_list_size; 1055 pr_info("The IPL report contains the following components:\n"); 1056 while (ptr < end) { 1057 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 1058 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 1059 str = "signed, verified"; 1060 else 1061 str = "signed, verification failed"; 1062 } else { 1063 str = "not signed"; 1064 } 1065 pr_info("%016llx - %016llx (%s)\n", 1066 ptr->addr, ptr->addr + ptr->len, str); 1067 ptr++; 1068 } 1069 } 1070 1071 /* 1072 * Setup function called from init/main.c just after the banner 1073 * was printed. 1074 */ 1075 1076 void __init setup_arch(char **cmdline_p) 1077 { 1078 /* 1079 * print what head.S has found out about the machine 1080 */ 1081 if (MACHINE_IS_VM) 1082 pr_info("Linux is running as a z/VM " 1083 "guest operating system in 64-bit mode\n"); 1084 else if (MACHINE_IS_KVM) 1085 pr_info("Linux is running under KVM in 64-bit mode\n"); 1086 else if (MACHINE_IS_LPAR) 1087 pr_info("Linux is running natively in 64-bit mode\n"); 1088 else 1089 pr_info("Linux is running as a guest in 64-bit mode\n"); 1090 1091 log_component_list(); 1092 1093 /* Have one command line that is parsed and saved in /proc/cmdline */ 1094 /* boot_command_line has been already set up in early.c */ 1095 *cmdline_p = boot_command_line; 1096 1097 ROOT_DEV = Root_RAM0; 1098 1099 init_mm.start_code = (unsigned long) _text; 1100 init_mm.end_code = (unsigned long) _etext; 1101 init_mm.end_data = (unsigned long) _edata; 1102 init_mm.brk = (unsigned long) _end; 1103 1104 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 1105 nospec_auto_detect(); 1106 1107 parse_early_param(); 1108 #ifdef CONFIG_CRASH_DUMP 1109 /* Deactivate elfcorehdr= kernel parameter */ 1110 elfcorehdr_addr = ELFCORE_ADDR_MAX; 1111 #endif 1112 1113 os_info_init(); 1114 setup_ipl(); 1115 setup_task_size(); 1116 setup_control_program_code(); 1117 1118 /* Do some memory reservations *before* memory is added to memblock */ 1119 reserve_memory_end(); 1120 reserve_oldmem(); 1121 reserve_kernel(); 1122 reserve_initrd(); 1123 reserve_certificate_list(); 1124 reserve_mem_detect_info(); 1125 memblock_allow_resize(); 1126 1127 /* Get information about *all* installed memory */ 1128 memblock_add_mem_detect_info(); 1129 1130 free_mem_detect_info(); 1131 remove_oldmem(); 1132 1133 /* 1134 * Make sure all chunks are MAX_ORDER aligned so we don't need the 1135 * extra checks that HOLES_IN_ZONE would require. 1136 * 1137 * Is this still required? 1138 */ 1139 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT)); 1140 1141 if (is_prot_virt_host()) 1142 setup_uv(); 1143 setup_memory_end(); 1144 setup_memory(); 1145 dma_contiguous_reserve(memory_end); 1146 vmcp_cma_reserve(); 1147 1148 check_initrd(); 1149 reserve_crashkernel(); 1150 #ifdef CONFIG_CRASH_DUMP 1151 /* 1152 * Be aware that smp_save_dump_cpus() triggers a system reset. 1153 * Therefore CPU and device initialization should be done afterwards. 1154 */ 1155 smp_save_dump_cpus(); 1156 #endif 1157 1158 setup_resources(); 1159 setup_lowcore_dat_off(); 1160 smp_fill_possible_mask(); 1161 cpu_detect_mhz_feature(); 1162 cpu_init(); 1163 numa_setup(); 1164 smp_detect_cpus(); 1165 topology_init_early(); 1166 1167 /* 1168 * Create kernel page tables and switch to virtual addressing. 1169 */ 1170 paging_init(); 1171 1172 /* 1173 * After paging_init created the kernel page table, the new PSWs 1174 * in lowcore can now run with DAT enabled. 1175 */ 1176 setup_lowcore_dat_on(); 1177 1178 /* Setup default console */ 1179 conmode_default(); 1180 set_preferred_console(); 1181 1182 apply_alternative_instructions(); 1183 if (IS_ENABLED(CONFIG_EXPOLINE)) 1184 nospec_init_branches(); 1185 1186 /* Setup zfcpdump support */ 1187 setup_zfcpdump(); 1188 1189 /* Add system specific data to the random pool */ 1190 setup_randomness(); 1191 } 1192