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 #include <linux/kmemleak.h>
  54
  55 #include <asm/boot_data.h>
  56 #include <asm/ipl.h>
  57 #include <asm/facility.h>
  58 #include <asm/smp.h>
  59 #include <asm/mmu_context.h>
  60 #include <asm/cpcmd.h>
  61 #include <asm/lowcore.h>
  62 #include <asm/nmi.h>
  63 #include <asm/irq.h>
  64 #include <asm/page.h>
  65 #include <asm/ptrace.h>
  66 #include <asm/sections.h>
  67 #include <asm/ebcdic.h>
  68 #include <asm/diag.h>
  69 #include <asm/os_info.h>
  70 #include <asm/sclp.h>
  71 #include <asm/stacktrace.h>
  72 #include <asm/sysinfo.h>
  73 #include <asm/numa.h>
  74 #include <asm/alternative.h>
  75 #include <asm/nospec-branch.h>
  76 #include <asm/mem_detect.h>
  77 #include <asm/uv.h>
  78 #include <asm/asm-offsets.h>
  79 #include "entry.h"
  80
  81 /*
  82  * Machine setup..
  83  */
  84 unsigned int console_mode = 0;
  85 EXPORT_SYMBOL(console_mode);
  86
  87 unsigned int console_devno = -1;
  88 EXPORT_SYMBOL(console_devno);
  89
  90 unsigned int console_irq = -1;
  91 EXPORT_SYMBOL(console_irq);
  92
  93 /*
  94  * Some code and data needs to stay below 2 GB, even when the kernel would be
  95  * relocated above 2 GB, because it has to use 31 bit addresses.
  96  * Such code and data is part of the .amode31 section.
  97  */
  98 unsigned long __amode31_ref __samode31 = __pa(&_samode31);
  99 unsigned long __amode31_ref __eamode31 = __pa(&_eamode31);
 100 unsigned long __amode31_ref __stext_amode31 = __pa(&_stext_amode31);
 101 unsigned long __amode31_ref __etext_amode31 = __pa(&_etext_amode31);
 102 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
 103 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
 104
 105 /*
 106  * Control registers CR2, CR5 and CR15 are initialized with addresses
 107  * of tables that must be placed below 2G which is handled by the AMODE31
 108  * sections.
 109  * Because the AMODE31 sections are relocated below 2G at startup,
 110  * the content of control registers CR2, CR5 and CR15 must be updated
 111  * with new addresses after the relocation. The initial initialization of
 112  * control registers occurs in head64.S and then gets updated again after AMODE31
 113  * relocation. We must access the relevant AMODE31 tables indirectly via
 114  * pointers placed in the .amode31.refs linker section. Those pointers get
 115  * updated automatically during AMODE31 relocation and always contain a valid
 116  * address within AMODE31 sections.
 117  */
 118
 119 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
 120
 121 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
 122         [1] = 0xffffffffffffffff
 123 };
 124
 125 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
 126         0x80000000, 0, 0, 0,
 127         0x80000000, 0, 0, 0,
 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 };
 135
 136 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
 137         0, 0, 0x89000000, 0,
 138         0, 0, 0x8a000000, 0
 139 };
 140
 141 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
 142 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
 143 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
 144 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
 145
 146 int __bootdata(noexec_disabled);
 147 unsigned long __bootdata(ident_map_size);
 148 struct mem_detect_info __bootdata(mem_detect);
 149 struct initrd_data __bootdata(initrd_data);
 150
 151 unsigned long __bootdata_preserved(__kaslr_offset);
 152 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 153 EXPORT_SYMBOL(zlib_dfltcc_support);
 154 u64 __bootdata_preserved(stfle_fac_list[16]);
 155 EXPORT_SYMBOL(stfle_fac_list);
 156 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
 157 struct oldmem_data __bootdata_preserved(oldmem_data);
 158
 159 unsigned long VMALLOC_START;
 160 EXPORT_SYMBOL(VMALLOC_START);
 161
 162 unsigned long VMALLOC_END;
 163 EXPORT_SYMBOL(VMALLOC_END);
 164
 165 struct page *vmemmap;
 166 EXPORT_SYMBOL(vmemmap);
 167 unsigned long vmemmap_size;
 168
 169 unsigned long MODULES_VADDR;
 170 unsigned long MODULES_END;
 171
 172 /* An array with a pointer to the lowcore of every CPU. */
 173 struct lowcore *lowcore_ptr[NR_CPUS];
 174 EXPORT_SYMBOL(lowcore_ptr);
 175
 176 /*
 177  * The Write Back bit position in the physaddr is given by the SLPC PCI.
 178  * Leaving the mask zero always uses write through which is safe
 179  */
 180 unsigned long mio_wb_bit_mask __ro_after_init;
 181
 182 /*
 183  * This is set up by the setup-routine at boot-time
 184  * for S390 need to find out, what we have to setup
 185  * using address 0x10400 ...
 186  */
 187
 188 #include <asm/setup.h>
 189
 190 /*
 191  * condev= and conmode= setup parameter.
 192  */
 193
 194 static int __init condev_setup(char *str)
 195 {
 196         int vdev;
 197
 198         vdev = simple_strtoul(str, &str, 0);
 199         if (vdev >= 0 && vdev < 65536) {
 200                 console_devno = vdev;
 201                 console_irq = -1;
 202         }
 203         return 1;
 204 }
 205
 206 __setup("condev=", condev_setup);
 207
 208 static void __init set_preferred_console(void)
 209 {
 210         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 211                 add_preferred_console("ttyS", 0, NULL);
 212         else if (CONSOLE_IS_3270)
 213                 add_preferred_console("tty3270", 0, NULL);
 214         else if (CONSOLE_IS_VT220)
 215                 add_preferred_console("ttysclp", 0, NULL);
 216         else if (CONSOLE_IS_HVC)
 217                 add_preferred_console("hvc", 0, NULL);
 218 }
 219
 220 static int __init conmode_setup(char *str)
 221 {
 222 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 223         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 224                 SET_CONSOLE_SCLP;
 225 #endif
 226 #if defined(CONFIG_TN3215_CONSOLE)
 227         if (!strcmp(str, "3215"))
 228                 SET_CONSOLE_3215;
 229 #endif
 230 #if defined(CONFIG_TN3270_CONSOLE)
 231         if (!strcmp(str, "3270"))
 232                 SET_CONSOLE_3270;
 233 #endif
 234         set_preferred_console();
 235         return 1;
 236 }
 237
 238 __setup("conmode=", conmode_setup);
 239
 240 static void __init conmode_default(void)
 241 {
 242         char query_buffer[1024];
 243         char *ptr;
 244
 245         if (MACHINE_IS_VM) {
 246                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 247                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 248                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 249                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 250                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 251                 ptr = strstr(query_buffer, "CONMODE");
 252                 /*
 253                  * Set the conmode to 3215 so that the device recognition
 254                  * will set the cu_type of the console to 3215. If the
 255                  * conmode is 3270 and we don't set it back then both
 256                  * 3215 and the 3270 driver will try to access the console
 257                  * device (3215 as console and 3270 as normal tty).
 258                  */
 259                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 260                 if (ptr == NULL) {
 261 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 262                         SET_CONSOLE_SCLP;
 263 #endif
 264                         return;
 265                 }
 266                 if (str_has_prefix(ptr + 8, "3270")) {
 267 #if defined(CONFIG_TN3270_CONSOLE)
 268                         SET_CONSOLE_3270;
 269 #elif defined(CONFIG_TN3215_CONSOLE)
 270                         SET_CONSOLE_3215;
 271 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 272                         SET_CONSOLE_SCLP;
 273 #endif
 274                 } else if (str_has_prefix(ptr + 8, "3215")) {
 275 #if defined(CONFIG_TN3215_CONSOLE)
 276                         SET_CONSOLE_3215;
 277 #elif defined(CONFIG_TN3270_CONSOLE)
 278                         SET_CONSOLE_3270;
 279 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 280                         SET_CONSOLE_SCLP;
 281 #endif
 282                 }
 283         } else if (MACHINE_IS_KVM) {
 284                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 285                         SET_CONSOLE_VT220;
 286                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 287                         SET_CONSOLE_SCLP;
 288                 else
 289                         SET_CONSOLE_HVC;
 290         } else {
 291 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 292                 SET_CONSOLE_SCLP;
 293 #endif
 294         }
 295 }
 296
 297 #ifdef CONFIG_CRASH_DUMP
 298 static void __init setup_zfcpdump(void)
 299 {
 300         if (!is_ipl_type_dump())
 301                 return;
 302         if (oldmem_data.start)
 303                 return;
 304         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 305         console_loglevel = 2;
 306 }
 307 #else
 308 static inline void setup_zfcpdump(void) {}
 309 #endif /* CONFIG_CRASH_DUMP */
 310
 311  /*
 312  * Reboot, halt and power_off stubs. They just call _machine_restart,
 313  * _machine_halt or _machine_power_off.
 314  */
 315
 316 void machine_restart(char *command)
 317 {
 318         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 319                 /*
 320                  * Only unblank the console if we are called in enabled
 321                  * context or a bust_spinlocks cleared the way for us.
 322                  */
 323                 console_unblank();
 324         _machine_restart(command);
 325 }
 326
 327 void machine_halt(void)
 328 {
 329         if (!in_interrupt() || oops_in_progress)
 330                 /*
 331                  * Only unblank the console if we are called in enabled
 332                  * context or a bust_spinlocks cleared the way for us.
 333                  */
 334                 console_unblank();
 335         _machine_halt();
 336 }
 337
 338 void machine_power_off(void)
 339 {
 340         if (!in_interrupt() || oops_in_progress)
 341                 /*
 342                  * Only unblank the console if we are called in enabled
 343                  * context or a bust_spinlocks cleared the way for us.
 344                  */
 345                 console_unblank();
 346         _machine_power_off();
 347 }
 348
 349 /*
 350  * Dummy power off function.
 351  */
 352 void (*pm_power_off)(void) = machine_power_off;
 353 EXPORT_SYMBOL_GPL(pm_power_off);
 354
 355 void *restart_stack;
 356
 357 unsigned long stack_alloc(void)
 358 {
 359 #ifdef CONFIG_VMAP_STACK
 360         void *ret;
 361
 362         ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
 363                              NUMA_NO_NODE, __builtin_return_address(0));
 364         kmemleak_not_leak(ret);
 365         return (unsigned long)ret;
 366 #else
 367         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 368 #endif
 369 }
 370
 371 void stack_free(unsigned long stack)
 372 {
 373 #ifdef CONFIG_VMAP_STACK
 374         vfree((void *) stack);
 375 #else
 376         free_pages(stack, THREAD_SIZE_ORDER);
 377 #endif
 378 }
 379
 380 int __init arch_early_irq_init(void)
 381 {
 382         unsigned long stack;
 383
 384         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 385         if (!stack)
 386                 panic("Couldn't allocate async stack");
 387         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 388         return 0;
 389 }
 390
 391 void __init arch_call_rest_init(void)
 392 {
 393         unsigned long stack;
 394
 395         stack = stack_alloc();
 396         if (!stack)
 397                 panic("Couldn't allocate kernel stack");
 398         current->stack = (void *) stack;
 399 #ifdef CONFIG_VMAP_STACK
 400         current->stack_vm_area = (void *) stack;
 401 #endif
 402         set_task_stack_end_magic(current);
 403         stack += STACK_INIT_OFFSET;
 404         S390_lowcore.kernel_stack = stack;
 405         call_on_stack_noreturn(rest_init, stack);
 406 }
 407
 408 static void __init setup_lowcore_dat_off(void)
 409 {
 410         unsigned long int_psw_mask = PSW_KERNEL_BITS;
 411         unsigned long mcck_stack;
 412         struct lowcore *lc;
 413
 414         if (IS_ENABLED(CONFIG_KASAN))
 415                 int_psw_mask |= PSW_MASK_DAT;
 416
 417         /*
 418          * Setup lowcore for boot cpu
 419          */
 420         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 421         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 422         if (!lc)
 423                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 424                       __func__, sizeof(*lc), sizeof(*lc));
 425
 426         lc->restart_psw.mask = PSW_KERNEL_BITS;
 427         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 428         lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 429         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 430         lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 431         lc->svc_new_psw.addr = (unsigned long) system_call;
 432         lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 433         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 434         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 435         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 436         lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 437         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 438         lc->clock_comparator = clock_comparator_max;
 439         lc->nodat_stack = ((unsigned long) &init_thread_union)
 440                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 441         lc->current_task = (unsigned long)&init_task;
 442         lc->lpp = LPP_MAGIC;
 443         lc->machine_flags = S390_lowcore.machine_flags;
 444         lc->preempt_count = S390_lowcore.preempt_count;
 445         nmi_alloc_boot_cpu(lc);
 446         lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
 447         lc->exit_timer = S390_lowcore.exit_timer;
 448         lc->user_timer = S390_lowcore.user_timer;
 449         lc->system_timer = S390_lowcore.system_timer;
 450         lc->steal_timer = S390_lowcore.steal_timer;
 451         lc->last_update_timer = S390_lowcore.last_update_timer;
 452         lc->last_update_clock = S390_lowcore.last_update_clock;
 453
 454         /*
 455          * Allocate the global restart stack which is the same for
 456          * all CPUs in cast *one* of them does a PSW restart.
 457          */
 458         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 459         if (!restart_stack)
 460                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 461                       __func__, THREAD_SIZE, THREAD_SIZE);
 462         restart_stack += STACK_INIT_OFFSET;
 463
 464         /*
 465          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 466          * restart data to the absolute zero lowcore. This is necessary if
 467          * PSW restart is done on an offline CPU that has lowcore zero.
 468          */
 469         lc->restart_stack = (unsigned long) restart_stack;
 470         lc->restart_fn = (unsigned long) do_restart;
 471         lc->restart_data = 0;
 472         lc->restart_source = -1U;
 473
 474         mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 475         if (!mcck_stack)
 476                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 477                       __func__, THREAD_SIZE, THREAD_SIZE);
 478         lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
 479
 480         /* Setup absolute zero lowcore */
 481         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 482         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 483         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 484         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 485         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 486
 487         lc->spinlock_lockval = arch_spin_lockval(0);
 488         lc->spinlock_index = 0;
 489         arch_spin_lock_setup(0);
 490         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 491         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 492         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 493         lc->preempt_count = PREEMPT_DISABLED;
 494
 495         set_prefix((u32)(unsigned long) lc);
 496         lowcore_ptr[0] = lc;
 497 }
 498
 499 static void __init setup_lowcore_dat_on(void)
 500 {
 501         struct lowcore *lc = lowcore_ptr[0];
 502
 503         __ctl_clear_bit(0, 28);
 504         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 505         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 506         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 507         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 508         __ctl_store(S390_lowcore.cregs_save_area, 0, 15);
 509         __ctl_set_bit(0, 28);
 510         mem_assign_absolute(S390_lowcore.restart_flags, RESTART_FLAG_CTLREGS);
 511         mem_assign_absolute(S390_lowcore.program_new_psw, lc->program_new_psw);
 512         memcpy_absolute(&S390_lowcore.cregs_save_area, lc->cregs_save_area,
 513                         sizeof(S390_lowcore.cregs_save_area));
 514 }
 515
 516 static struct resource code_resource = {
 517         .name  = "Kernel code",
 518         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 519 };
 520
 521 static struct resource data_resource = {
 522         .name = "Kernel data",
 523         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 524 };
 525
 526 static struct resource bss_resource = {
 527         .name = "Kernel bss",
 528         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 529 };
 530
 531 static struct resource __initdata *standard_resources[] = {
 532         &code_resource,
 533         &data_resource,
 534         &bss_resource,
 535 };
 536
 537 static void __init setup_resources(void)
 538 {
 539         struct resource *res, *std_res, *sub_res;
 540         phys_addr_t start, end;
 541         int j;
 542         u64 i;
 543
 544         code_resource.start = (unsigned long) _text;
 545         code_resource.end = (unsigned long) _etext - 1;
 546         data_resource.start = (unsigned long) _etext;
 547         data_resource.end = (unsigned long) _edata - 1;
 548         bss_resource.start = (unsigned long) __bss_start;
 549         bss_resource.end = (unsigned long) __bss_stop - 1;
 550
 551         for_each_mem_range(i, &start, &end) {
 552                 res = memblock_alloc(sizeof(*res), 8);
 553                 if (!res)
 554                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 555                               __func__, sizeof(*res), 8);
 556                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 557
 558                 res->name = "System RAM";
 559                 res->start = start;
 560                 /*
 561                  * In memblock, end points to the first byte after the
 562                  * range while in resourses, end points to the last byte in
 563                  * the range.
 564                  */
 565                 res->end = end - 1;
 566                 request_resource(&iomem_resource, res);
 567
 568                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 569                         std_res = standard_resources[j];
 570                         if (std_res->start < res->start ||
 571                             std_res->start > res->end)
 572                                 continue;
 573                         if (std_res->end > res->end) {
 574                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 575                                 if (!sub_res)
 576                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 577                                               __func__, sizeof(*sub_res), 8);
 578                                 *sub_res = *std_res;
 579                                 sub_res->end = res->end;
 580                                 std_res->start = res->end + 1;
 581                                 request_resource(res, sub_res);
 582                         } else {
 583                                 request_resource(res, std_res);
 584                         }
 585                 }
 586         }
 587 #ifdef CONFIG_CRASH_DUMP
 588         /*
 589          * Re-add removed crash kernel memory as reserved memory. This makes
 590          * sure it will be mapped with the identity mapping and struct pages
 591          * will be created, so it can be resized later on.
 592          * However add it later since the crash kernel resource should not be
 593          * part of the System RAM resource.
 594          */
 595         if (crashk_res.end) {
 596                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 597                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 598                 insert_resource(&iomem_resource, &crashk_res);
 599         }
 600 #endif
 601 }
 602
 603 static void __init setup_memory_end(void)
 604 {
 605         memblock_remove(ident_map_size, ULONG_MAX);
 606         max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
 607         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
 608 }
 609
 610 #ifdef CONFIG_CRASH_DUMP
 611
 612 /*
 613  * When kdump is enabled, we have to ensure that no memory from the area
 614  * [0 - crashkernel memory size] is set offline - it will be exchanged with
 615  * the crashkernel memory region when kdump is triggered. The crashkernel
 616  * memory region can never get offlined (pages are unmovable).
 617  */
 618 static int kdump_mem_notifier(struct notifier_block *nb,
 619                               unsigned long action, void *data)
 620 {
 621         struct memory_notify *arg = data;
 622
 623         if (action != MEM_GOING_OFFLINE)
 624                 return NOTIFY_OK;
 625         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 626                 return NOTIFY_BAD;
 627         return NOTIFY_OK;
 628 }
 629
 630 static struct notifier_block kdump_mem_nb = {
 631         .notifier_call = kdump_mem_notifier,
 632 };
 633
 634 #endif
 635
 636 /*
 637  * Reserve memory for kdump kernel to be loaded with kexec
 638  */
 639 static void __init reserve_crashkernel(void)
 640 {
 641 #ifdef CONFIG_CRASH_DUMP
 642         unsigned long long crash_base, crash_size;
 643         phys_addr_t low, high;
 644         int rc;
 645
 646         rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
 647                                &crash_base);
 648
 649         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 650         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 651         if (rc || crash_size == 0)
 652                 return;
 653
 654         if (memblock.memory.regions[0].size < crash_size) {
 655                 pr_info("crashkernel reservation failed: %s\n",
 656                         "first memory chunk must be at least crashkernel size");
 657                 return;
 658         }
 659
 660         low = crash_base ?: oldmem_data.start;
 661         high = low + crash_size;
 662         if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
 663                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 664                 crash_base = low;
 665         } else {
 666                 /* Find suitable area in free memory */
 667                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 668                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 669
 670                 if (crash_base && crash_base < low) {
 671                         pr_info("crashkernel reservation failed: %s\n",
 672                                 "crash_base too low");
 673                         return;
 674                 }
 675                 low = crash_base ?: low;
 676                 crash_base = memblock_phys_alloc_range(crash_size,
 677                                                        KEXEC_CRASH_MEM_ALIGN,
 678                                                        low, high);
 679         }
 680
 681         if (!crash_base) {
 682                 pr_info("crashkernel reservation failed: %s\n",
 683                         "no suitable area found");
 684                 return;
 685         }
 686
 687         if (register_memory_notifier(&kdump_mem_nb)) {
 688                 memblock_free(crash_base, crash_size);
 689                 return;
 690         }
 691
 692         if (!oldmem_data.start && MACHINE_IS_VM)
 693                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 694         crashk_res.start = crash_base;
 695         crashk_res.end = crash_base + crash_size - 1;
 696         memblock_remove(crash_base, crash_size);
 697         pr_info("Reserving %lluMB of memory at %lluMB "
 698                 "for crashkernel (System RAM: %luMB)\n",
 699                 crash_size >> 20, crash_base >> 20,
 700                 (unsigned long)memblock.memory.total_size >> 20);
 701         os_info_crashkernel_add(crash_base, crash_size);
 702 #endif
 703 }
 704
 705 /*
 706  * Reserve the initrd from being used by memblock
 707  */
 708 static void __init reserve_initrd(void)
 709 {
 710 #ifdef CONFIG_BLK_DEV_INITRD
 711         if (!initrd_data.start || !initrd_data.size)
 712                 return;
 713         initrd_start = initrd_data.start;
 714         initrd_end = initrd_start + initrd_data.size;
 715         memblock_reserve(initrd_data.start, initrd_data.size);
 716 #endif
 717 }
 718
 719 /*
 720  * Reserve the memory area used to pass the certificate lists
 721  */
 722 static void __init reserve_certificate_list(void)
 723 {
 724         if (ipl_cert_list_addr)
 725                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 726 }
 727
 728 static void __init reserve_mem_detect_info(void)
 729 {
 730         unsigned long start, size;
 731
 732         get_mem_detect_reserved(&start, &size);
 733         if (size)
 734                 memblock_reserve(start, size);
 735 }
 736
 737 static void __init free_mem_detect_info(void)
 738 {
 739         unsigned long start, size;
 740
 741         get_mem_detect_reserved(&start, &size);
 742         if (size)
 743                 memblock_free(start, size);
 744 }
 745
 746 static const char * __init get_mem_info_source(void)
 747 {
 748         switch (mem_detect.info_source) {
 749         case MEM_DETECT_SCLP_STOR_INFO:
 750                 return "sclp storage info";
 751         case MEM_DETECT_DIAG260:
 752                 return "diag260";
 753         case MEM_DETECT_SCLP_READ_INFO:
 754                 return "sclp read info";
 755         case MEM_DETECT_BIN_SEARCH:
 756                 return "binary search";
 757         }
 758         return "none";
 759 }
 760
 761 static void __init memblock_add_mem_detect_info(void)
 762 {
 763         unsigned long start, end;
 764         int i;
 765
 766         pr_debug("physmem info source: %s (%hhd)\n",
 767                  get_mem_info_source(), mem_detect.info_source);
 768         /* keep memblock lists close to the kernel */
 769         memblock_set_bottom_up(true);
 770         for_each_mem_detect_block(i, &start, &end) {
 771                 memblock_add(start, end - start);
 772                 memblock_physmem_add(start, end - start);
 773         }
 774         memblock_set_bottom_up(false);
 775         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 776         memblock_dump_all();
 777 }
 778
 779 /*
 780  * Check for initrd being in usable memory
 781  */
 782 static void __init check_initrd(void)
 783 {
 784 #ifdef CONFIG_BLK_DEV_INITRD
 785         if (initrd_data.start && initrd_data.size &&
 786             !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
 787                 pr_err("The initial RAM disk does not fit into the memory\n");
 788                 memblock_free(initrd_data.start, initrd_data.size);
 789                 initrd_start = initrd_end = 0;
 790         }
 791 #endif
 792 }
 793
 794 /*
 795  * Reserve memory used for lowcore/command line/kernel image.
 796  */
 797 static void __init reserve_kernel(void)
 798 {
 799         unsigned long start_pfn = PFN_UP(__pa(_end));
 800
 801         memblock_reserve(0, STARTUP_NORMAL_OFFSET);
 802         memblock_reserve((unsigned long)sclp_early_sccb, EXT_SCCB_READ_SCP);
 803         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 804                          - (unsigned long)_stext);
 805 }
 806
 807 static void __init setup_memory(void)
 808 {
 809         phys_addr_t start, end;
 810         u64 i;
 811
 812         /*
 813          * Init storage key for present memory
 814          */
 815         for_each_mem_range(i, &start, &end)
 816                 storage_key_init_range(start, end);
 817
 818         psw_set_key(PAGE_DEFAULT_KEY);
 819 }
 820
 821 static void __init relocate_amode31_section(void)
 822 {
 823         unsigned long amode31_addr, amode31_size;
 824         long amode31_offset;
 825         long *ptr;
 826
 827         /* Allocate a new AMODE31 capable memory region */
 828         amode31_size = __eamode31 - __samode31;
 829         pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
 830         amode31_addr = (unsigned long)memblock_alloc_low(amode31_size, PAGE_SIZE);
 831         if (!amode31_addr)
 832                 panic("Failed to allocate memory for AMODE31 section\n");
 833         amode31_offset = amode31_addr - __samode31;
 834
 835         /* Move original AMODE31 section to the new one */
 836         memmove((void *)amode31_addr, (void *)__samode31, amode31_size);
 837         /* Zero out the old AMODE31 section to catch invalid accesses within it */
 838         memset((void *)__samode31, 0, amode31_size);
 839
 840         /* Update all AMODE31 region references */
 841         for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
 842                 *ptr += amode31_offset;
 843 }
 844
 845 /* This must be called after AMODE31 relocation */
 846 static void __init setup_cr(void)
 847 {
 848         union ctlreg2 cr2;
 849         union ctlreg5 cr5;
 850         union ctlreg15 cr15;
 851
 852         __ctl_duct[1] = (unsigned long)__ctl_aste;
 853         __ctl_duct[2] = (unsigned long)__ctl_aste;
 854         __ctl_duct[4] = (unsigned long)__ctl_duald;
 855
 856         /* Update control registers CR2, CR5 and CR15 */
 857         __ctl_store(cr2.val, 2, 2);
 858         __ctl_store(cr5.val, 5, 5);
 859         __ctl_store(cr15.val, 15, 15);
 860         cr2.ducto = (unsigned long)__ctl_duct >> 6;
 861         cr5.pasteo = (unsigned long)__ctl_duct >> 6;
 862         cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
 863         __ctl_load(cr2.val, 2, 2);
 864         __ctl_load(cr5.val, 5, 5);
 865         __ctl_load(cr15.val, 15, 15);
 866 }
 867
 868 /*
 869  * Add system information as device randomness
 870  */
 871 static void __init setup_randomness(void)
 872 {
 873         struct sysinfo_3_2_2 *vmms;
 874
 875         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 876                                                             PAGE_SIZE);
 877         if (!vmms)
 878                 panic("Failed to allocate memory for sysinfo structure\n");
 879
 880         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 881                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 882         memblock_free((unsigned long) vmms, PAGE_SIZE);
 883
 884 #ifdef CONFIG_ARCH_RANDOM
 885         if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
 886                 static_branch_enable(&s390_arch_random_available);
 887 #endif
 888 }
 889
 890 /*
 891  * Find the correct size for the task_struct. This depends on
 892  * the size of the struct fpu at the end of the thread_struct
 893  * which is embedded in the task_struct.
 894  */
 895 static void __init setup_task_size(void)
 896 {
 897         int task_size = sizeof(struct task_struct);
 898
 899         if (!MACHINE_HAS_VX) {
 900                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 901                 task_size += sizeof(freg_t) * __NUM_FPRS;
 902         }
 903         arch_task_struct_size = task_size;
 904 }
 905
 906 /*
 907  * Issue diagnose 318 to set the control program name and
 908  * version codes.
 909  */
 910 static void __init setup_control_program_code(void)
 911 {
 912         union diag318_info diag318_info = {
 913                 .cpnc = CPNC_LINUX,
 914                 .cpvc = 0,
 915         };
 916
 917         if (!sclp.has_diag318)
 918                 return;
 919
 920         diag_stat_inc(DIAG_STAT_X318);
 921         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
 922 }
 923
 924 /*
 925  * Print the component list from the IPL report
 926  */
 927 static void __init log_component_list(void)
 928 {
 929         struct ipl_rb_component_entry *ptr, *end;
 930         char *str;
 931
 932         if (!early_ipl_comp_list_addr)
 933                 return;
 934         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
 935                 pr_info("Linux is running with Secure-IPL enabled\n");
 936         else
 937                 pr_info("Linux is running with Secure-IPL disabled\n");
 938         ptr = (void *) early_ipl_comp_list_addr;
 939         end = (void *) ptr + early_ipl_comp_list_size;
 940         pr_info("The IPL report contains the following components:\n");
 941         while (ptr < end) {
 942                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
 943                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
 944                                 str = "signed, verified";
 945                         else
 946                                 str = "signed, verification failed";
 947                 } else {
 948                         str = "not signed";
 949                 }
 950                 pr_info("%016llx - %016llx (%s)\n",
 951                         ptr->addr, ptr->addr + ptr->len, str);
 952                 ptr++;
 953         }
 954 }
 955
 956 /*
 957  * Setup function called from init/main.c just after the banner
 958  * was printed.
 959  */
 960
 961 void __init setup_arch(char **cmdline_p)
 962 {
 963         /*
 964          * print what head.S has found out about the machine
 965          */
 966         if (MACHINE_IS_VM)
 967                 pr_info("Linux is running as a z/VM "
 968                         "guest operating system in 64-bit mode\n");
 969         else if (MACHINE_IS_KVM)
 970                 pr_info("Linux is running under KVM in 64-bit mode\n");
 971         else if (MACHINE_IS_LPAR)
 972                 pr_info("Linux is running natively in 64-bit mode\n");
 973         else
 974                 pr_info("Linux is running as a guest in 64-bit mode\n");
 975
 976         log_component_list();
 977
 978         /* Have one command line that is parsed and saved in /proc/cmdline */
 979         /* boot_command_line has been already set up in early.c */
 980         *cmdline_p = boot_command_line;
 981
 982         ROOT_DEV = Root_RAM0;
 983
 984         setup_initial_init_mm(_text, _etext, _edata, _end);
 985
 986         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
 987                 nospec_auto_detect();
 988
 989         jump_label_init();
 990         parse_early_param();
 991 #ifdef CONFIG_CRASH_DUMP
 992         /* Deactivate elfcorehdr= kernel parameter */
 993         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 994 #endif
 995
 996         os_info_init();
 997         setup_ipl();
 998         setup_task_size();
 999         setup_control_program_code();
1000
1001         /* Do some memory reservations *before* memory is added to memblock */
1002         reserve_kernel();
1003         reserve_initrd();
1004         reserve_certificate_list();
1005         reserve_mem_detect_info();
1006         memblock_set_current_limit(ident_map_size);
1007         memblock_allow_resize();
1008
1009         /* Get information about *all* installed memory */
1010         memblock_add_mem_detect_info();
1011
1012         free_mem_detect_info();
1013
1014         relocate_amode31_section();
1015         setup_cr();
1016
1017         setup_uv();
1018         setup_memory_end();
1019         setup_memory();
1020         dma_contiguous_reserve(ident_map_size);
1021         vmcp_cma_reserve();
1022         if (MACHINE_HAS_EDAT2)
1023                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1024
1025         check_initrd();
1026         reserve_crashkernel();
1027 #ifdef CONFIG_CRASH_DUMP
1028         /*
1029          * Be aware that smp_save_dump_cpus() triggers a system reset.
1030          * Therefore CPU and device initialization should be done afterwards.
1031          */
1032         smp_save_dump_cpus();
1033 #endif
1034
1035         setup_resources();
1036         setup_lowcore_dat_off();
1037         smp_fill_possible_mask();
1038         cpu_detect_mhz_feature();
1039         cpu_init();
1040         numa_setup();
1041         smp_detect_cpus();
1042         topology_init_early();
1043
1044         /*
1045          * Create kernel page tables and switch to virtual addressing.
1046          */
1047         paging_init();
1048
1049         /*
1050          * After paging_init created the kernel page table, the new PSWs
1051          * in lowcore can now run with DAT enabled.
1052          */
1053         setup_lowcore_dat_on();
1054
1055         /* Setup default console */
1056         conmode_default();
1057         set_preferred_console();
1058
1059         apply_alternative_instructions();
1060         if (IS_ENABLED(CONFIG_EXPOLINE))
1061                 nospec_init_branches();
1062
1063         /* Setup zfcp/nvme dump support */
1064         setup_zfcpdump();
1065
1066         /* Add system specific data to the random pool */
1067         setup_randomness();
1068 }