arch/s390/kernel/setup.c

   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
  54 #include <asm/boot_data.h>
  55 #include <asm/ipl.h>
  56 #include <asm/facility.h>
  57 #include <asm/smp.h>
  58 #include <asm/mmu_context.h>
  59 #include <asm/cpcmd.h>
  60 #include <asm/lowcore.h>
  61 #include <asm/nmi.h>
  62 #include <asm/irq.h>
  63 #include <asm/page.h>
  64 #include <asm/ptrace.h>
  65 #include <asm/sections.h>
  66 #include <asm/ebcdic.h>
  67 #include <asm/diag.h>
  68 #include <asm/os_info.h>
  69 #include <asm/sclp.h>
  70 #include <asm/stacktrace.h>
  71 #include <asm/sysinfo.h>
  72 #include <asm/numa.h>
  73 #include <asm/alternative.h>
  74 #include <asm/nospec-branch.h>
  75 #include <asm/mem_detect.h>
  76 #include <asm/uv.h>
  77 #include <asm/asm-offsets.h>
  78 #include "entry.h"
  79
  80 /*
  81  * Machine setup..
  82  */
  83 unsigned int console_mode = 0;
  84 EXPORT_SYMBOL(console_mode);
  85
  86 unsigned int console_devno = -1;
  87 EXPORT_SYMBOL(console_devno);
  88
  89 unsigned int console_irq = -1;
  90 EXPORT_SYMBOL(console_irq);
  91
  92 unsigned long elf_hwcap __read_mostly = 0;
  93 char elf_platform[ELF_PLATFORM_SIZE];
  94
  95 unsigned long int_hwcap = 0;
  96
  97 int __bootdata(noexec_disabled);
  98 unsigned long __bootdata(ident_map_size);
  99 struct mem_detect_info __bootdata(mem_detect);
 100
 101 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
 102 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
 103 unsigned long __bootdata_preserved(__stext_dma);
 104 unsigned long __bootdata_preserved(__etext_dma);
 105 unsigned long __bootdata_preserved(__sdma);
 106 unsigned long __bootdata_preserved(__edma);
 107 unsigned long __bootdata_preserved(__kaslr_offset);
 108 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 109 EXPORT_SYMBOL(zlib_dfltcc_support);
 110 u64 __bootdata_preserved(stfle_fac_list[16]);
 111 EXPORT_SYMBOL(stfle_fac_list);
 112 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
 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 unsigned long vmemmap_size;
 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  * The Write Back bit position in the physaddr is given by the SLPC PCI.
 133  * Leaving the mask zero always uses write through which is safe
 134  */
 135 unsigned long mio_wb_bit_mask __ro_after_init;
 136
 137 /*
 138  * This is set up by the setup-routine at boot-time
 139  * for S390 need to find out, what we have to setup
 140  * using address 0x10400 ...
 141  */
 142
 143 #include <asm/setup.h>
 144
 145 /*
 146  * condev= and conmode= setup parameter.
 147  */
 148
 149 static int __init condev_setup(char *str)
 150 {
 151         int vdev;
 152
 153         vdev = simple_strtoul(str, &str, 0);
 154         if (vdev >= 0 && vdev < 65536) {
 155                 console_devno = vdev;
 156                 console_irq = -1;
 157         }
 158         return 1;
 159 }
 160
 161 __setup("condev=", condev_setup);
 162
 163 static void __init set_preferred_console(void)
 164 {
 165         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 166                 add_preferred_console("ttyS", 0, NULL);
 167         else if (CONSOLE_IS_3270)
 168                 add_preferred_console("tty3270", 0, NULL);
 169         else if (CONSOLE_IS_VT220)
 170                 add_preferred_console("ttysclp", 0, NULL);
 171         else if (CONSOLE_IS_HVC)
 172                 add_preferred_console("hvc", 0, NULL);
 173 }
 174
 175 static int __init conmode_setup(char *str)
 176 {
 177 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 178         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 179                 SET_CONSOLE_SCLP;
 180 #endif
 181 #if defined(CONFIG_TN3215_CONSOLE)
 182         if (!strcmp(str, "3215"))
 183                 SET_CONSOLE_3215;
 184 #endif
 185 #if defined(CONFIG_TN3270_CONSOLE)
 186         if (!strcmp(str, "3270"))
 187                 SET_CONSOLE_3270;
 188 #endif
 189         set_preferred_console();
 190         return 1;
 191 }
 192
 193 __setup("conmode=", conmode_setup);
 194
 195 static void __init conmode_default(void)
 196 {
 197         char query_buffer[1024];
 198         char *ptr;
 199
 200         if (MACHINE_IS_VM) {
 201                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 202                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 203                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 204                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 205                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 206                 ptr = strstr(query_buffer, "CONMODE");
 207                 /*
 208                  * Set the conmode to 3215 so that the device recognition
 209                  * will set the cu_type of the console to 3215. If the
 210                  * conmode is 3270 and we don't set it back then both
 211                  * 3215 and the 3270 driver will try to access the console
 212                  * device (3215 as console and 3270 as normal tty).
 213                  */
 214                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 215                 if (ptr == NULL) {
 216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 217                         SET_CONSOLE_SCLP;
 218 #endif
 219                         return;
 220                 }
 221                 if (str_has_prefix(ptr + 8, "3270")) {
 222 #if defined(CONFIG_TN3270_CONSOLE)
 223                         SET_CONSOLE_3270;
 224 #elif defined(CONFIG_TN3215_CONSOLE)
 225                         SET_CONSOLE_3215;
 226 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 227                         SET_CONSOLE_SCLP;
 228 #endif
 229                 } else if (str_has_prefix(ptr + 8, "3215")) {
 230 #if defined(CONFIG_TN3215_CONSOLE)
 231                         SET_CONSOLE_3215;
 232 #elif defined(CONFIG_TN3270_CONSOLE)
 233                         SET_CONSOLE_3270;
 234 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 235                         SET_CONSOLE_SCLP;
 236 #endif
 237                 }
 238         } else if (MACHINE_IS_KVM) {
 239                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 240                         SET_CONSOLE_VT220;
 241                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 242                         SET_CONSOLE_SCLP;
 243                 else
 244                         SET_CONSOLE_HVC;
 245         } else {
 246 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 247                 SET_CONSOLE_SCLP;
 248 #endif
 249         }
 250 }
 251
 252 #ifdef CONFIG_CRASH_DUMP
 253 static void __init setup_zfcpdump(void)
 254 {
 255         if (!is_ipl_type_dump())
 256                 return;
 257         if (OLDMEM_BASE)
 258                 return;
 259         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 260         console_loglevel = 2;
 261 }
 262 #else
 263 static inline void setup_zfcpdump(void) {}
 264 #endif /* CONFIG_CRASH_DUMP */
 265
 266  /*
 267  * Reboot, halt and power_off stubs. They just call _machine_restart,
 268  * _machine_halt or _machine_power_off.
 269  */
 270
 271 void machine_restart(char *command)
 272 {
 273         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 274                 /*
 275                  * Only unblank the console if we are called in enabled
 276                  * context or a bust_spinlocks cleared the way for us.
 277                  */
 278                 console_unblank();
 279         _machine_restart(command);
 280 }
 281
 282 void machine_halt(void)
 283 {
 284         if (!in_interrupt() || oops_in_progress)
 285                 /*
 286                  * Only unblank the console if we are called in enabled
 287                  * context or a bust_spinlocks cleared the way for us.
 288                  */
 289                 console_unblank();
 290         _machine_halt();
 291 }
 292
 293 void machine_power_off(void)
 294 {
 295         if (!in_interrupt() || oops_in_progress)
 296                 /*
 297                  * Only unblank the console if we are called in enabled
 298                  * context or a bust_spinlocks cleared the way for us.
 299                  */
 300                 console_unblank();
 301         _machine_power_off();
 302 }
 303
 304 /*
 305  * Dummy power off function.
 306  */
 307 void (*pm_power_off)(void) = machine_power_off;
 308 EXPORT_SYMBOL_GPL(pm_power_off);
 309
 310 void *restart_stack;
 311
 312 unsigned long stack_alloc(void)
 313 {
 314 #ifdef CONFIG_VMAP_STACK
 315         return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
 316                         THREADINFO_GFP, NUMA_NO_NODE,
 317                         __builtin_return_address(0));
 318 #else
 319         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 320 #endif
 321 }
 322
 323 void stack_free(unsigned long stack)
 324 {
 325 #ifdef CONFIG_VMAP_STACK
 326         vfree((void *) stack);
 327 #else
 328         free_pages(stack, THREAD_SIZE_ORDER);
 329 #endif
 330 }
 331
 332 int __init arch_early_irq_init(void)
 333 {
 334         unsigned long stack;
 335
 336         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 337         if (!stack)
 338                 panic("Couldn't allocate async stack");
 339         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 340         return 0;
 341 }
 342
 343 void __init arch_call_rest_init(void)
 344 {
 345         unsigned long stack;
 346
 347         stack = stack_alloc();
 348         if (!stack)
 349                 panic("Couldn't allocate kernel stack");
 350         current->stack = (void *) stack;
 351 #ifdef CONFIG_VMAP_STACK
 352         current->stack_vm_area = (void *) stack;
 353 #endif
 354         set_task_stack_end_magic(current);
 355         stack += STACK_INIT_OFFSET;
 356         S390_lowcore.kernel_stack = stack;
 357         call_on_stack_noreturn(rest_init, stack);
 358 }
 359
 360 static void __init setup_lowcore_dat_off(void)
 361 {
 362         unsigned long int_psw_mask = PSW_KERNEL_BITS;
 363         unsigned long mcck_stack;
 364         struct lowcore *lc;
 365
 366         if (IS_ENABLED(CONFIG_KASAN))
 367                 int_psw_mask |= PSW_MASK_DAT;
 368
 369         /*
 370          * Setup lowcore for boot cpu
 371          */
 372         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 373         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 374         if (!lc)
 375                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 376                       __func__, sizeof(*lc), sizeof(*lc));
 377
 378         lc->restart_psw.mask = PSW_KERNEL_BITS;
 379         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 380         lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 381         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 382         lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 383         lc->svc_new_psw.addr = (unsigned long) system_call;
 384         lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 385         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 386         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 387         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 388         lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 389         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 390         lc->clock_comparator = clock_comparator_max;
 391         lc->nodat_stack = ((unsigned long) &init_thread_union)
 392                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 393         lc->current_task = (unsigned long)&init_task;
 394         lc->lpp = LPP_MAGIC;
 395         lc->machine_flags = S390_lowcore.machine_flags;
 396         lc->preempt_count = S390_lowcore.preempt_count;
 397         nmi_alloc_boot_cpu(lc);
 398         lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
 399         lc->exit_timer = S390_lowcore.exit_timer;
 400         lc->user_timer = S390_lowcore.user_timer;
 401         lc->system_timer = S390_lowcore.system_timer;
 402         lc->steal_timer = S390_lowcore.steal_timer;
 403         lc->last_update_timer = S390_lowcore.last_update_timer;
 404         lc->last_update_clock = S390_lowcore.last_update_clock;
 405
 406         /*
 407          * Allocate the global restart stack which is the same for
 408          * all CPUs in cast *one* of them does a PSW restart.
 409          */
 410         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 411         if (!restart_stack)
 412                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 413                       __func__, THREAD_SIZE, THREAD_SIZE);
 414         restart_stack += STACK_INIT_OFFSET;
 415
 416         /*
 417          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 418          * restart data to the absolute zero lowcore. This is necessary if
 419          * PSW restart is done on an offline CPU that has lowcore zero.
 420          */
 421         lc->restart_stack = (unsigned long) restart_stack;
 422         lc->restart_fn = (unsigned long) do_restart;
 423         lc->restart_data = 0;
 424         lc->restart_source = -1UL;
 425
 426         mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 427         if (!mcck_stack)
 428                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 429                       __func__, THREAD_SIZE, THREAD_SIZE);
 430         lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
 431
 432         /* Setup absolute zero lowcore */
 433         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 434         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 435         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 436         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 437         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 438
 439         lc->spinlock_lockval = arch_spin_lockval(0);
 440         lc->spinlock_index = 0;
 441         arch_spin_lock_setup(0);
 442         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 443         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 444         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 445         lc->preempt_count = PREEMPT_DISABLED;
 446
 447         set_prefix((u32)(unsigned long) lc);
 448         lowcore_ptr[0] = lc;
 449 }
 450
 451 static void __init setup_lowcore_dat_on(void)
 452 {
 453         __ctl_clear_bit(0, 28);
 454         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 455         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 456         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 457         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 458         __ctl_set_bit(0, 28);
 459 }
 460
 461 static struct resource code_resource = {
 462         .name  = "Kernel code",
 463         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 464 };
 465
 466 static struct resource data_resource = {
 467         .name = "Kernel data",
 468         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 469 };
 470
 471 static struct resource bss_resource = {
 472         .name = "Kernel bss",
 473         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 474 };
 475
 476 static struct resource __initdata *standard_resources[] = {
 477         &code_resource,
 478         &data_resource,
 479         &bss_resource,
 480 };
 481
 482 static void __init setup_resources(void)
 483 {
 484         struct resource *res, *std_res, *sub_res;
 485         phys_addr_t start, end;
 486         int j;
 487         u64 i;
 488
 489         code_resource.start = (unsigned long) _text;
 490         code_resource.end = (unsigned long) _etext - 1;
 491         data_resource.start = (unsigned long) _etext;
 492         data_resource.end = (unsigned long) _edata - 1;
 493         bss_resource.start = (unsigned long) __bss_start;
 494         bss_resource.end = (unsigned long) __bss_stop - 1;
 495
 496         for_each_mem_range(i, &start, &end) {
 497                 res = memblock_alloc(sizeof(*res), 8);
 498                 if (!res)
 499                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 500                               __func__, sizeof(*res), 8);
 501                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 502
 503                 res->name = "System RAM";
 504                 res->start = start;
 505                 /*
 506                  * In memblock, end points to the first byte after the
 507                  * range while in resourses, end points to the last byte in
 508                  * the range.
 509                  */
 510                 res->end = end - 1;
 511                 request_resource(&iomem_resource, res);
 512
 513                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 514                         std_res = standard_resources[j];
 515                         if (std_res->start < res->start ||
 516                             std_res->start > res->end)
 517                                 continue;
 518                         if (std_res->end > res->end) {
 519                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 520                                 if (!sub_res)
 521                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 522                                               __func__, sizeof(*sub_res), 8);
 523                                 *sub_res = *std_res;
 524                                 sub_res->end = res->end;
 525                                 std_res->start = res->end + 1;
 526                                 request_resource(res, sub_res);
 527                         } else {
 528                                 request_resource(res, std_res);
 529                         }
 530                 }
 531         }
 532 #ifdef CONFIG_CRASH_DUMP
 533         /*
 534          * Re-add removed crash kernel memory as reserved memory. This makes
 535          * sure it will be mapped with the identity mapping and struct pages
 536          * will be created, so it can be resized later on.
 537          * However add it later since the crash kernel resource should not be
 538          * part of the System RAM resource.
 539          */
 540         if (crashk_res.end) {
 541                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 542                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 543                 insert_resource(&iomem_resource, &crashk_res);
 544         }
 545 #endif
 546 }
 547
 548 static void __init setup_memory_end(void)
 549 {
 550         memblock_remove(ident_map_size, ULONG_MAX);
 551         max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
 552         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
 553 }
 554
 555 #ifdef CONFIG_CRASH_DUMP
 556
 557 /*
 558  * When kdump is enabled, we have to ensure that no memory from the area
 559  * [0 - crashkernel memory size] is set offline - it will be exchanged with
 560  * the crashkernel memory region when kdump is triggered. The crashkernel
 561  * memory region can never get offlined (pages are unmovable).
 562  */
 563 static int kdump_mem_notifier(struct notifier_block *nb,
 564                               unsigned long action, void *data)
 565 {
 566         struct memory_notify *arg = data;
 567
 568         if (action != MEM_GOING_OFFLINE)
 569                 return NOTIFY_OK;
 570         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 571                 return NOTIFY_BAD;
 572         return NOTIFY_OK;
 573 }
 574
 575 static struct notifier_block kdump_mem_nb = {
 576         .notifier_call = kdump_mem_notifier,
 577 };
 578
 579 #endif
 580
 581 /*
 582  * Make sure that the area above identity mapping is protected
 583  */
 584 static void __init reserve_above_ident_map(void)
 585 {
 586         memblock_reserve(ident_map_size, ULONG_MAX);
 587 }
 588
 589 /*
 590  * Reserve memory for kdump kernel to be loaded with kexec
 591  */
 592 static void __init reserve_crashkernel(void)
 593 {
 594 #ifdef CONFIG_CRASH_DUMP
 595         unsigned long long crash_base, crash_size;
 596         phys_addr_t low, high;
 597         int rc;
 598
 599         rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
 600                                &crash_base);
 601
 602         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 603         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 604         if (rc || crash_size == 0)
 605                 return;
 606
 607         if (memblock.memory.regions[0].size < crash_size) {
 608                 pr_info("crashkernel reservation failed: %s\n",
 609                         "first memory chunk must be at least crashkernel size");
 610                 return;
 611         }
 612
 613         low = crash_base ?: OLDMEM_BASE;
 614         high = low + crash_size;
 615         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 616                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 617                 crash_base = low;
 618         } else {
 619                 /* Find suitable area in free memory */
 620                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 621                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 622
 623                 if (crash_base && crash_base < low) {
 624                         pr_info("crashkernel reservation failed: %s\n",
 625                                 "crash_base too low");
 626                         return;
 627                 }
 628                 low = crash_base ?: low;
 629                 crash_base = memblock_phys_alloc_range(crash_size,
 630                                                        KEXEC_CRASH_MEM_ALIGN,
 631                                                        low, high);
 632         }
 633
 634         if (!crash_base) {
 635                 pr_info("crashkernel reservation failed: %s\n",
 636                         "no suitable area found");
 637                 return;
 638         }
 639
 640         if (register_memory_notifier(&kdump_mem_nb)) {
 641                 memblock_free(crash_base, crash_size);
 642                 return;
 643         }
 644
 645         if (!OLDMEM_BASE && MACHINE_IS_VM)
 646                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 647         crashk_res.start = crash_base;
 648         crashk_res.end = crash_base + crash_size - 1;
 649         memblock_remove(crash_base, crash_size);
 650         pr_info("Reserving %lluMB of memory at %lluMB "
 651                 "for crashkernel (System RAM: %luMB)\n",
 652                 crash_size >> 20, crash_base >> 20,
 653                 (unsigned long)memblock.memory.total_size >> 20);
 654         os_info_crashkernel_add(crash_base, crash_size);
 655 #endif
 656 }
 657
 658 /*
 659  * Reserve the initrd from being used by memblock
 660  */
 661 static void __init reserve_initrd(void)
 662 {
 663 #ifdef CONFIG_BLK_DEV_INITRD
 664         if (!INITRD_START || !INITRD_SIZE)
 665                 return;
 666         initrd_start = INITRD_START;
 667         initrd_end = initrd_start + INITRD_SIZE;
 668         memblock_reserve(INITRD_START, INITRD_SIZE);
 669 #endif
 670 }
 671
 672 /*
 673  * Reserve the memory area used to pass the certificate lists
 674  */
 675 static void __init reserve_certificate_list(void)
 676 {
 677         if (ipl_cert_list_addr)
 678                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 679 }
 680
 681 static void __init reserve_mem_detect_info(void)
 682 {
 683         unsigned long start, size;
 684
 685         get_mem_detect_reserved(&start, &size);
 686         if (size)
 687                 memblock_reserve(start, size);
 688 }
 689
 690 static void __init free_mem_detect_info(void)
 691 {
 692         unsigned long start, size;
 693
 694         get_mem_detect_reserved(&start, &size);
 695         if (size)
 696                 memblock_free(start, size);
 697 }
 698
 699 static const char * __init get_mem_info_source(void)
 700 {
 701         switch (mem_detect.info_source) {
 702         case MEM_DETECT_SCLP_STOR_INFO:
 703                 return "sclp storage info";
 704         case MEM_DETECT_DIAG260:
 705                 return "diag260";
 706         case MEM_DETECT_SCLP_READ_INFO:
 707                 return "sclp read info";
 708         case MEM_DETECT_BIN_SEARCH:
 709                 return "binary search";
 710         }
 711         return "none";
 712 }
 713
 714 static void __init memblock_add_mem_detect_info(void)
 715 {
 716         unsigned long start, end;
 717         int i;
 718
 719         pr_debug("physmem info source: %s (%hhd)\n",
 720                  get_mem_info_source(), mem_detect.info_source);
 721         /* keep memblock lists close to the kernel */
 722         memblock_set_bottom_up(true);
 723         for_each_mem_detect_block(i, &start, &end) {
 724                 memblock_add(start, end - start);
 725                 memblock_physmem_add(start, end - start);
 726         }
 727         memblock_set_bottom_up(false);
 728         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 729         memblock_dump_all();
 730 }
 731
 732 /*
 733  * Check for initrd being in usable memory
 734  */
 735 static void __init check_initrd(void)
 736 {
 737 #ifdef CONFIG_BLK_DEV_INITRD
 738         if (INITRD_START && INITRD_SIZE &&
 739             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 740                 pr_err("The initial RAM disk does not fit into the memory\n");
 741                 memblock_free(INITRD_START, INITRD_SIZE);
 742                 initrd_start = initrd_end = 0;
 743         }
 744 #endif
 745 }
 746
 747 /*
 748  * Reserve memory used for lowcore/command line/kernel image.
 749  */
 750 static void __init reserve_kernel(void)
 751 {
 752         unsigned long start_pfn = PFN_UP(__pa(_end));
 753
 754         memblock_reserve(0, HEAD_END);
 755         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 756                          - (unsigned long)_stext);
 757         memblock_reserve(__sdma, __edma - __sdma);
 758 }
 759
 760 static void __init setup_memory(void)
 761 {
 762         phys_addr_t start, end;
 763         u64 i;
 764
 765         /*
 766          * Init storage key for present memory
 767          */
 768         for_each_mem_range(i, &start, &end)
 769                 storage_key_init_range(start, end);
 770
 771         psw_set_key(PAGE_DEFAULT_KEY);
 772
 773         /* Only cosmetics */
 774         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 775 }
 776
 777 /*
 778  * Setup hardware capabilities.
 779  */
 780 static int __init setup_hwcaps(void)
 781 {
 782         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 783         struct cpuid cpu_id;
 784         int i;
 785
 786         /*
 787          * The store facility list bits numbers as found in the principles
 788          * of operation are numbered with bit 1UL<<31 as number 0 to
 789          * bit 1UL<<0 as number 31.
 790          *   Bit 0: instructions named N3, "backported" to esa-mode
 791          *   Bit 2: z/Architecture mode is active
 792          *   Bit 7: the store-facility-list-extended facility is installed
 793          *   Bit 17: the message-security assist is installed
 794          *   Bit 19: the long-displacement facility is installed
 795          *   Bit 21: the extended-immediate facility is installed
 796          *   Bit 22: extended-translation facility 3 is installed
 797          *   Bit 30: extended-translation facility 3 enhancement facility
 798          * These get translated to:
 799          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 800          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 801          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 802          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 803          */
 804         for (i = 0; i < 6; i++)
 805                 if (test_facility(stfl_bits[i]))
 806                         elf_hwcap |= 1UL << i;
 807
 808         if (test_facility(22) && test_facility(30))
 809                 elf_hwcap |= HWCAP_S390_ETF3EH;
 810
 811         /*
 812          * Check for additional facilities with store-facility-list-extended.
 813          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 814          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 815          * as stored by stfl, bits 32-xxx contain additional facilities.
 816          * How many facility words are stored depends on the number of
 817          * doublewords passed to the instruction. The additional facilities
 818          * are:
 819          *   Bit 42: decimal floating point facility is installed
 820          *   Bit 44: perform floating point operation facility is installed
 821          * translated to:
 822          *   HWCAP_S390_DFP bit 6 (42 && 44).
 823          */
 824         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 825                 elf_hwcap |= HWCAP_S390_DFP;
 826
 827         /*
 828          * Huge page support HWCAP_S390_HPAGE is bit 7.
 829          */
 830         if (MACHINE_HAS_EDAT1)
 831                 elf_hwcap |= HWCAP_S390_HPAGE;
 832
 833         /*
 834          * 64-bit register support for 31-bit processes
 835          * HWCAP_S390_HIGH_GPRS is bit 9.
 836          */
 837         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 838
 839         /*
 840          * Transactional execution support HWCAP_S390_TE is bit 10.
 841          */
 842         if (MACHINE_HAS_TE)
 843                 elf_hwcap |= HWCAP_S390_TE;
 844
 845         /*
 846          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 847          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 848          * instead of facility bit 129.
 849          */
 850         if (MACHINE_HAS_VX) {
 851                 elf_hwcap |= HWCAP_S390_VXRS;
 852                 if (test_facility(134))
 853                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 854                 if (test_facility(135))
 855                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 856                 if (test_facility(148))
 857                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 858                 if (test_facility(152))
 859                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 860         }
 861         if (test_facility(150))
 862                 elf_hwcap |= HWCAP_S390_SORT;
 863         if (test_facility(151))
 864                 elf_hwcap |= HWCAP_S390_DFLT;
 865
 866         /*
 867          * Guarded storage support HWCAP_S390_GS is bit 12.
 868          */
 869         if (MACHINE_HAS_GS)
 870                 elf_hwcap |= HWCAP_S390_GS;
 871
 872         get_cpu_id(&cpu_id);
 873         add_device_randomness(&cpu_id, sizeof(cpu_id));
 874         switch (cpu_id.machine) {
 875         case 0x2064:
 876         case 0x2066:
 877         default:        /* Use "z900" as default for 64 bit kernels. */
 878                 strcpy(elf_platform, "z900");
 879                 break;
 880         case 0x2084:
 881         case 0x2086:
 882                 strcpy(elf_platform, "z990");
 883                 break;
 884         case 0x2094:
 885         case 0x2096:
 886                 strcpy(elf_platform, "z9-109");
 887                 break;
 888         case 0x2097:
 889         case 0x2098:
 890                 strcpy(elf_platform, "z10");
 891                 break;
 892         case 0x2817:
 893         case 0x2818:
 894                 strcpy(elf_platform, "z196");
 895                 break;
 896         case 0x2827:
 897         case 0x2828:
 898                 strcpy(elf_platform, "zEC12");
 899                 break;
 900         case 0x2964:
 901         case 0x2965:
 902                 strcpy(elf_platform, "z13");
 903                 break;
 904         case 0x3906:
 905         case 0x3907:
 906                 strcpy(elf_platform, "z14");
 907                 break;
 908         case 0x8561:
 909         case 0x8562:
 910                 strcpy(elf_platform, "z15");
 911                 break;
 912         }
 913
 914         /*
 915          * Virtualization support HWCAP_INT_SIE is bit 0.
 916          */
 917         if (sclp.has_sief2)
 918                 int_hwcap |= HWCAP_INT_SIE;
 919
 920         return 0;
 921 }
 922 arch_initcall(setup_hwcaps);
 923
 924 /*
 925  * Add system information as device randomness
 926  */
 927 static void __init setup_randomness(void)
 928 {
 929         struct sysinfo_3_2_2 *vmms;
 930
 931         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 932                                                             PAGE_SIZE);
 933         if (!vmms)
 934                 panic("Failed to allocate memory for sysinfo structure\n");
 935
 936         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 937                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 938         memblock_free((unsigned long) vmms, PAGE_SIZE);
 939 }
 940
 941 /*
 942  * Find the correct size for the task_struct. This depends on
 943  * the size of the struct fpu at the end of the thread_struct
 944  * which is embedded in the task_struct.
 945  */
 946 static void __init setup_task_size(void)
 947 {
 948         int task_size = sizeof(struct task_struct);
 949
 950         if (!MACHINE_HAS_VX) {
 951                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 952                 task_size += sizeof(freg_t) * __NUM_FPRS;
 953         }
 954         arch_task_struct_size = task_size;
 955 }
 956
 957 /*
 958  * Issue diagnose 318 to set the control program name and
 959  * version codes.
 960  */
 961 static void __init setup_control_program_code(void)
 962 {
 963         union diag318_info diag318_info = {
 964                 .cpnc = CPNC_LINUX,
 965                 .cpvc = 0,
 966         };
 967
 968         if (!sclp.has_diag318)
 969                 return;
 970
 971         diag_stat_inc(DIAG_STAT_X318);
 972         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
 973 }
 974
 975 /*
 976  * Print the component list from the IPL report
 977  */
 978 static void __init log_component_list(void)
 979 {
 980         struct ipl_rb_component_entry *ptr, *end;
 981         char *str;
 982
 983         if (!early_ipl_comp_list_addr)
 984                 return;
 985         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
 986                 pr_info("Linux is running with Secure-IPL enabled\n");
 987         else
 988                 pr_info("Linux is running with Secure-IPL disabled\n");
 989         ptr = (void *) early_ipl_comp_list_addr;
 990         end = (void *) ptr + early_ipl_comp_list_size;
 991         pr_info("The IPL report contains the following components:\n");
 992         while (ptr < end) {
 993                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
 994                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
 995                                 str = "signed, verified";
 996                         else
 997                                 str = "signed, verification failed";
 998                 } else {
 999                         str = "not signed";
1000                 }
1001                 pr_info("%016llx - %016llx (%s)\n",
1002                         ptr->addr, ptr->addr + ptr->len, str);
1003                 ptr++;
1004         }
1005 }
1006
1007 /*
1008  * Setup function called from init/main.c just after the banner
1009  * was printed.
1010  */
1011
1012 void __init setup_arch(char **cmdline_p)
1013 {
1014         /*
1015          * print what head.S has found out about the machine
1016          */
1017         if (MACHINE_IS_VM)
1018                 pr_info("Linux is running as a z/VM "
1019                         "guest operating system in 64-bit mode\n");
1020         else if (MACHINE_IS_KVM)
1021                 pr_info("Linux is running under KVM in 64-bit mode\n");
1022         else if (MACHINE_IS_LPAR)
1023                 pr_info("Linux is running natively in 64-bit mode\n");
1024         else
1025                 pr_info("Linux is running as a guest in 64-bit mode\n");
1026
1027         log_component_list();
1028
1029         /* Have one command line that is parsed and saved in /proc/cmdline */
1030         /* boot_command_line has been already set up in early.c */
1031         *cmdline_p = boot_command_line;
1032
1033         ROOT_DEV = Root_RAM0;
1034
1035         setup_initial_init_mm(_text, _etext, _edata, _end);
1036
1037         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1038                 nospec_auto_detect();
1039
1040         jump_label_init();
1041         parse_early_param();
1042 #ifdef CONFIG_CRASH_DUMP
1043         /* Deactivate elfcorehdr= kernel parameter */
1044         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1045 #endif
1046
1047         os_info_init();
1048         setup_ipl();
1049         setup_task_size();
1050         setup_control_program_code();
1051
1052         /* Do some memory reservations *before* memory is added to memblock */
1053         reserve_above_ident_map();
1054         reserve_kernel();
1055         reserve_initrd();
1056         reserve_certificate_list();
1057         reserve_mem_detect_info();
1058         memblock_allow_resize();
1059
1060         /* Get information about *all* installed memory */
1061         memblock_add_mem_detect_info();
1062
1063         free_mem_detect_info();
1064
1065         setup_uv();
1066         setup_memory_end();
1067         setup_memory();
1068         dma_contiguous_reserve(ident_map_size);
1069         vmcp_cma_reserve();
1070         if (MACHINE_HAS_EDAT2)
1071                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1072
1073         check_initrd();
1074         reserve_crashkernel();
1075 #ifdef CONFIG_CRASH_DUMP
1076         /*
1077          * Be aware that smp_save_dump_cpus() triggers a system reset.
1078          * Therefore CPU and device initialization should be done afterwards.
1079          */
1080         smp_save_dump_cpus();
1081 #endif
1082
1083         setup_resources();
1084         setup_lowcore_dat_off();
1085         smp_fill_possible_mask();
1086         cpu_detect_mhz_feature();
1087         cpu_init();
1088         numa_setup();
1089         smp_detect_cpus();
1090         topology_init_early();
1091
1092         /*
1093          * Create kernel page tables and switch to virtual addressing.
1094          */
1095         paging_init();
1096
1097         /*
1098          * After paging_init created the kernel page table, the new PSWs
1099          * in lowcore can now run with DAT enabled.
1100          */
1101         setup_lowcore_dat_on();
1102
1103         /* Setup default console */
1104         conmode_default();
1105         set_preferred_console();
1106
1107         apply_alternative_instructions();
1108         if (IS_ENABLED(CONFIG_EXPOLINE))
1109                 nospec_init_branches();
1110
1111         /* Setup zfcp/nvme dump support */
1112         setup_zfcpdump();
1113
1114         /* Add system specific data to the random pool */
1115         setup_randomness();
1116 }