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 /*
  93  * Some code and data needs to stay below 2 GB, even when the kernel would be
  94  * relocated above 2 GB, because it has to use 31 bit addresses.
  95  * Such code and data is part of the .dma section.
  96  */
  97 unsigned long __dma_ref __sdma = __pa(&_sdma);
  98 unsigned long __dma_ref __edma = __pa(&_edma);
  99 unsigned long __dma_ref __stext_dma = __pa(&_stext_dma);
 100 unsigned long __dma_ref __etext_dma = __pa(&_etext_dma);
 101 struct exception_table_entry __dma_ref *__start_dma_ex_table = _start_dma_ex_table;
 102 struct exception_table_entry __dma_ref *__stop_dma_ex_table = _stop_dma_ex_table;
 103
 104 /*
 105  * Control registers CR2, CR5 and CR15 are initialized with addresses
 106  * of tables that must be placed below 2G which is handled by the DMA
 107  * sections.
 108  * Because the DMA sections are relocated below 2G at startup,
 109  * the content of control registers CR2, CR5 and CR15 must be updated
 110  * with new addresses after the relocation. The initial initialization of
 111  * control registers occurs in head64.S and then gets updated again after DMA
 112  * relocation. We must access the relevant DMA tables indirectly via
 113  * pointers placed in the .dma.refs linker section. Those pointers get
 114  * updated automatically during DMA relocation and always contain a valid
 115  * address within DMA sections.
 116  */
 117
 118 static __dma_data u32 __ctl_duct_dma[16] __aligned(64);
 119
 120 static __dma_data u64 __ctl_aste_dma[8] __aligned(64) = {
 121         [1] = 0xffffffffffffffff
 122 };
 123
 124 static __dma_data u32 __ctl_duald_dma[32] __aligned(128) = {
 125         0x80000000, 0, 0, 0,
 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 };
 134
 135 static __dma_data u32 __ctl_linkage_stack_dma[8] __aligned(64) = {
 136         0, 0, 0x89000000, 0,
 137         0, 0, 0x8a000000, 0
 138 };
 139
 140 static u64 __dma_ref *__ctl_aste = __ctl_aste_dma;
 141 static u32 __dma_ref *__ctl_duald = __ctl_duald_dma;
 142 static u32 __dma_ref *__ctl_linkage_stack = __ctl_linkage_stack_dma;
 143 static u32 __dma_ref *__ctl_duct = __ctl_duct_dma;
 144
 145 int __bootdata(noexec_disabled);
 146 unsigned long __bootdata(ident_map_size);
 147 struct mem_detect_info __bootdata(mem_detect);
 148 struct initrd_data __bootdata(initrd_data);
 149
 150 unsigned long __bootdata_preserved(__kaslr_offset);
 151 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 152 EXPORT_SYMBOL(zlib_dfltcc_support);
 153 u64 __bootdata_preserved(stfle_fac_list[16]);
 154 EXPORT_SYMBOL(stfle_fac_list);
 155 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
 156 struct oldmem_data __bootdata_preserved(oldmem_data);
 157
 158 unsigned long VMALLOC_START;
 159 EXPORT_SYMBOL(VMALLOC_START);
 160
 161 unsigned long VMALLOC_END;
 162 EXPORT_SYMBOL(VMALLOC_END);
 163
 164 struct page *vmemmap;
 165 EXPORT_SYMBOL(vmemmap);
 166 unsigned long vmemmap_size;
 167
 168 unsigned long MODULES_VADDR;
 169 unsigned long MODULES_END;
 170
 171 /* An array with a pointer to the lowcore of every CPU. */
 172 struct lowcore *lowcore_ptr[NR_CPUS];
 173 EXPORT_SYMBOL(lowcore_ptr);
 174
 175 /*
 176  * The Write Back bit position in the physaddr is given by the SLPC PCI.
 177  * Leaving the mask zero always uses write through which is safe
 178  */
 179 unsigned long mio_wb_bit_mask __ro_after_init;
 180
 181 /*
 182  * This is set up by the setup-routine at boot-time
 183  * for S390 need to find out, what we have to setup
 184  * using address 0x10400 ...
 185  */
 186
 187 #include <asm/setup.h>
 188
 189 /*
 190  * condev= and conmode= setup parameter.
 191  */
 192
 193 static int __init condev_setup(char *str)
 194 {
 195         int vdev;
 196
 197         vdev = simple_strtoul(str, &str, 0);
 198         if (vdev >= 0 && vdev < 65536) {
 199                 console_devno = vdev;
 200                 console_irq = -1;
 201         }
 202         return 1;
 203 }
 204
 205 __setup("condev=", condev_setup);
 206
 207 static void __init set_preferred_console(void)
 208 {
 209         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 210                 add_preferred_console("ttyS", 0, NULL);
 211         else if (CONSOLE_IS_3270)
 212                 add_preferred_console("tty3270", 0, NULL);
 213         else if (CONSOLE_IS_VT220)
 214                 add_preferred_console("ttysclp", 0, NULL);
 215         else if (CONSOLE_IS_HVC)
 216                 add_preferred_console("hvc", 0, NULL);
 217 }
 218
 219 static int __init conmode_setup(char *str)
 220 {
 221 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 222         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 223                 SET_CONSOLE_SCLP;
 224 #endif
 225 #if defined(CONFIG_TN3215_CONSOLE)
 226         if (!strcmp(str, "3215"))
 227                 SET_CONSOLE_3215;
 228 #endif
 229 #if defined(CONFIG_TN3270_CONSOLE)
 230         if (!strcmp(str, "3270"))
 231                 SET_CONSOLE_3270;
 232 #endif
 233         set_preferred_console();
 234         return 1;
 235 }
 236
 237 __setup("conmode=", conmode_setup);
 238
 239 static void __init conmode_default(void)
 240 {
 241         char query_buffer[1024];
 242         char *ptr;
 243
 244         if (MACHINE_IS_VM) {
 245                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 246                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 247                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 248                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 249                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 250                 ptr = strstr(query_buffer, "CONMODE");
 251                 /*
 252                  * Set the conmode to 3215 so that the device recognition
 253                  * will set the cu_type of the console to 3215. If the
 254                  * conmode is 3270 and we don't set it back then both
 255                  * 3215 and the 3270 driver will try to access the console
 256                  * device (3215 as console and 3270 as normal tty).
 257                  */
 258                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 259                 if (ptr == NULL) {
 260 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 261                         SET_CONSOLE_SCLP;
 262 #endif
 263                         return;
 264                 }
 265                 if (str_has_prefix(ptr + 8, "3270")) {
 266 #if defined(CONFIG_TN3270_CONSOLE)
 267                         SET_CONSOLE_3270;
 268 #elif defined(CONFIG_TN3215_CONSOLE)
 269                         SET_CONSOLE_3215;
 270 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 271                         SET_CONSOLE_SCLP;
 272 #endif
 273                 } else if (str_has_prefix(ptr + 8, "3215")) {
 274 #if defined(CONFIG_TN3215_CONSOLE)
 275                         SET_CONSOLE_3215;
 276 #elif defined(CONFIG_TN3270_CONSOLE)
 277                         SET_CONSOLE_3270;
 278 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 279                         SET_CONSOLE_SCLP;
 280 #endif
 281                 }
 282         } else if (MACHINE_IS_KVM) {
 283                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 284                         SET_CONSOLE_VT220;
 285                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 286                         SET_CONSOLE_SCLP;
 287                 else
 288                         SET_CONSOLE_HVC;
 289         } else {
 290 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 291                 SET_CONSOLE_SCLP;
 292 #endif
 293         }
 294 }
 295
 296 #ifdef CONFIG_CRASH_DUMP
 297 static void __init setup_zfcpdump(void)
 298 {
 299         if (!is_ipl_type_dump())
 300                 return;
 301         if (oldmem_data.start)
 302                 return;
 303         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 304         console_loglevel = 2;
 305 }
 306 #else
 307 static inline void setup_zfcpdump(void) {}
 308 #endif /* CONFIG_CRASH_DUMP */
 309
 310  /*
 311  * Reboot, halt and power_off stubs. They just call _machine_restart,
 312  * _machine_halt or _machine_power_off.
 313  */
 314
 315 void machine_restart(char *command)
 316 {
 317         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 318                 /*
 319                  * Only unblank the console if we are called in enabled
 320                  * context or a bust_spinlocks cleared the way for us.
 321                  */
 322                 console_unblank();
 323         _machine_restart(command);
 324 }
 325
 326 void machine_halt(void)
 327 {
 328         if (!in_interrupt() || oops_in_progress)
 329                 /*
 330                  * Only unblank the console if we are called in enabled
 331                  * context or a bust_spinlocks cleared the way for us.
 332                  */
 333                 console_unblank();
 334         _machine_halt();
 335 }
 336
 337 void machine_power_off(void)
 338 {
 339         if (!in_interrupt() || oops_in_progress)
 340                 /*
 341                  * Only unblank the console if we are called in enabled
 342                  * context or a bust_spinlocks cleared the way for us.
 343                  */
 344                 console_unblank();
 345         _machine_power_off();
 346 }
 347
 348 /*
 349  * Dummy power off function.
 350  */
 351 void (*pm_power_off)(void) = machine_power_off;
 352 EXPORT_SYMBOL_GPL(pm_power_off);
 353
 354 void *restart_stack;
 355
 356 unsigned long stack_alloc(void)
 357 {
 358 #ifdef CONFIG_VMAP_STACK
 359         return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
 360                         THREADINFO_GFP, NUMA_NO_NODE,
 361                         __builtin_return_address(0));
 362 #else
 363         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 364 #endif
 365 }
 366
 367 void stack_free(unsigned long stack)
 368 {
 369 #ifdef CONFIG_VMAP_STACK
 370         vfree((void *) stack);
 371 #else
 372         free_pages(stack, THREAD_SIZE_ORDER);
 373 #endif
 374 }
 375
 376 int __init arch_early_irq_init(void)
 377 {
 378         unsigned long stack;
 379
 380         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 381         if (!stack)
 382                 panic("Couldn't allocate async stack");
 383         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 384         return 0;
 385 }
 386
 387 void __init arch_call_rest_init(void)
 388 {
 389         unsigned long stack;
 390
 391         stack = stack_alloc();
 392         if (!stack)
 393                 panic("Couldn't allocate kernel stack");
 394         current->stack = (void *) stack;
 395 #ifdef CONFIG_VMAP_STACK
 396         current->stack_vm_area = (void *) stack;
 397 #endif
 398         set_task_stack_end_magic(current);
 399         stack += STACK_INIT_OFFSET;
 400         S390_lowcore.kernel_stack = stack;
 401         call_on_stack_noreturn(rest_init, stack);
 402 }
 403
 404 static void __init setup_lowcore_dat_off(void)
 405 {
 406         unsigned long int_psw_mask = PSW_KERNEL_BITS;
 407         unsigned long mcck_stack;
 408         struct lowcore *lc;
 409
 410         if (IS_ENABLED(CONFIG_KASAN))
 411                 int_psw_mask |= PSW_MASK_DAT;
 412
 413         /*
 414          * Setup lowcore for boot cpu
 415          */
 416         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 417         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 418         if (!lc)
 419                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 420                       __func__, sizeof(*lc), sizeof(*lc));
 421
 422         lc->restart_psw.mask = PSW_KERNEL_BITS;
 423         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 424         lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 425         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 426         lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 427         lc->svc_new_psw.addr = (unsigned long) system_call;
 428         lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 429         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 430         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 431         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 432         lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
 433         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 434         lc->clock_comparator = clock_comparator_max;
 435         lc->nodat_stack = ((unsigned long) &init_thread_union)
 436                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 437         lc->current_task = (unsigned long)&init_task;
 438         lc->lpp = LPP_MAGIC;
 439         lc->machine_flags = S390_lowcore.machine_flags;
 440         lc->preempt_count = S390_lowcore.preempt_count;
 441         nmi_alloc_boot_cpu(lc);
 442         lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
 443         lc->exit_timer = S390_lowcore.exit_timer;
 444         lc->user_timer = S390_lowcore.user_timer;
 445         lc->system_timer = S390_lowcore.system_timer;
 446         lc->steal_timer = S390_lowcore.steal_timer;
 447         lc->last_update_timer = S390_lowcore.last_update_timer;
 448         lc->last_update_clock = S390_lowcore.last_update_clock;
 449
 450         /*
 451          * Allocate the global restart stack which is the same for
 452          * all CPUs in cast *one* of them does a PSW restart.
 453          */
 454         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 455         if (!restart_stack)
 456                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 457                       __func__, THREAD_SIZE, THREAD_SIZE);
 458         restart_stack += STACK_INIT_OFFSET;
 459
 460         /*
 461          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 462          * restart data to the absolute zero lowcore. This is necessary if
 463          * PSW restart is done on an offline CPU that has lowcore zero.
 464          */
 465         lc->restart_stack = (unsigned long) restart_stack;
 466         lc->restart_fn = (unsigned long) do_restart;
 467         lc->restart_data = 0;
 468         lc->restart_source = -1UL;
 469
 470         mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 471         if (!mcck_stack)
 472                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 473                       __func__, THREAD_SIZE, THREAD_SIZE);
 474         lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
 475
 476         /* Setup absolute zero lowcore */
 477         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 478         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 479         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 480         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 481         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 482
 483         lc->spinlock_lockval = arch_spin_lockval(0);
 484         lc->spinlock_index = 0;
 485         arch_spin_lock_setup(0);
 486         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 487         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 488         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 489         lc->preempt_count = PREEMPT_DISABLED;
 490
 491         set_prefix((u32)(unsigned long) lc);
 492         lowcore_ptr[0] = lc;
 493 }
 494
 495 static void __init setup_lowcore_dat_on(void)
 496 {
 497         __ctl_clear_bit(0, 28);
 498         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 499         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 500         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 501         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 502         __ctl_set_bit(0, 28);
 503 }
 504
 505 static struct resource code_resource = {
 506         .name  = "Kernel code",
 507         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 508 };
 509
 510 static struct resource data_resource = {
 511         .name = "Kernel data",
 512         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 513 };
 514
 515 static struct resource bss_resource = {
 516         .name = "Kernel bss",
 517         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 518 };
 519
 520 static struct resource __initdata *standard_resources[] = {
 521         &code_resource,
 522         &data_resource,
 523         &bss_resource,
 524 };
 525
 526 static void __init setup_resources(void)
 527 {
 528         struct resource *res, *std_res, *sub_res;
 529         phys_addr_t start, end;
 530         int j;
 531         u64 i;
 532
 533         code_resource.start = (unsigned long) _text;
 534         code_resource.end = (unsigned long) _etext - 1;
 535         data_resource.start = (unsigned long) _etext;
 536         data_resource.end = (unsigned long) _edata - 1;
 537         bss_resource.start = (unsigned long) __bss_start;
 538         bss_resource.end = (unsigned long) __bss_stop - 1;
 539
 540         for_each_mem_range(i, &start, &end) {
 541                 res = memblock_alloc(sizeof(*res), 8);
 542                 if (!res)
 543                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 544                               __func__, sizeof(*res), 8);
 545                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 546
 547                 res->name = "System RAM";
 548                 res->start = start;
 549                 /*
 550                  * In memblock, end points to the first byte after the
 551                  * range while in resourses, end points to the last byte in
 552                  * the range.
 553                  */
 554                 res->end = end - 1;
 555                 request_resource(&iomem_resource, res);
 556
 557                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 558                         std_res = standard_resources[j];
 559                         if (std_res->start < res->start ||
 560                             std_res->start > res->end)
 561                                 continue;
 562                         if (std_res->end > res->end) {
 563                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 564                                 if (!sub_res)
 565                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 566                                               __func__, sizeof(*sub_res), 8);
 567                                 *sub_res = *std_res;
 568                                 sub_res->end = res->end;
 569                                 std_res->start = res->end + 1;
 570                                 request_resource(res, sub_res);
 571                         } else {
 572                                 request_resource(res, std_res);
 573                         }
 574                 }
 575         }
 576 #ifdef CONFIG_CRASH_DUMP
 577         /*
 578          * Re-add removed crash kernel memory as reserved memory. This makes
 579          * sure it will be mapped with the identity mapping and struct pages
 580          * will be created, so it can be resized later on.
 581          * However add it later since the crash kernel resource should not be
 582          * part of the System RAM resource.
 583          */
 584         if (crashk_res.end) {
 585                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 586                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 587                 insert_resource(&iomem_resource, &crashk_res);
 588         }
 589 #endif
 590 }
 591
 592 static void __init setup_memory_end(void)
 593 {
 594         memblock_remove(ident_map_size, ULONG_MAX);
 595         max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
 596         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
 597 }
 598
 599 #ifdef CONFIG_CRASH_DUMP
 600
 601 /*
 602  * When kdump is enabled, we have to ensure that no memory from the area
 603  * [0 - crashkernel memory size] is set offline - it will be exchanged with
 604  * the crashkernel memory region when kdump is triggered. The crashkernel
 605  * memory region can never get offlined (pages are unmovable).
 606  */
 607 static int kdump_mem_notifier(struct notifier_block *nb,
 608                               unsigned long action, void *data)
 609 {
 610         struct memory_notify *arg = data;
 611
 612         if (action != MEM_GOING_OFFLINE)
 613                 return NOTIFY_OK;
 614         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 615                 return NOTIFY_BAD;
 616         return NOTIFY_OK;
 617 }
 618
 619 static struct notifier_block kdump_mem_nb = {
 620         .notifier_call = kdump_mem_notifier,
 621 };
 622
 623 #endif
 624
 625 /*
 626  * Make sure that the area above identity mapping is protected
 627  */
 628 static void __init reserve_above_ident_map(void)
 629 {
 630         memblock_reserve(ident_map_size, ULONG_MAX);
 631 }
 632
 633 /*
 634  * Reserve memory for kdump kernel to be loaded with kexec
 635  */
 636 static void __init reserve_crashkernel(void)
 637 {
 638 #ifdef CONFIG_CRASH_DUMP
 639         unsigned long long crash_base, crash_size;
 640         phys_addr_t low, high;
 641         int rc;
 642
 643         rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
 644                                &crash_base);
 645
 646         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 647         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 648         if (rc || crash_size == 0)
 649                 return;
 650
 651         if (memblock.memory.regions[0].size < crash_size) {
 652                 pr_info("crashkernel reservation failed: %s\n",
 653                         "first memory chunk must be at least crashkernel size");
 654                 return;
 655         }
 656
 657         low = crash_base ?: oldmem_data.start;
 658         high = low + crash_size;
 659         if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
 660                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 661                 crash_base = low;
 662         } else {
 663                 /* Find suitable area in free memory */
 664                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 665                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 666
 667                 if (crash_base && crash_base < low) {
 668                         pr_info("crashkernel reservation failed: %s\n",
 669                                 "crash_base too low");
 670                         return;
 671                 }
 672                 low = crash_base ?: low;
 673                 crash_base = memblock_find_in_range(low, high, crash_size,
 674                                                     KEXEC_CRASH_MEM_ALIGN);
 675         }
 676
 677         if (!crash_base) {
 678                 pr_info("crashkernel reservation failed: %s\n",
 679                         "no suitable area found");
 680                 return;
 681         }
 682
 683         if (register_memory_notifier(&kdump_mem_nb))
 684                 return;
 685
 686         if (!oldmem_data.start && MACHINE_IS_VM)
 687                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 688         crashk_res.start = crash_base;
 689         crashk_res.end = crash_base + crash_size - 1;
 690         memblock_remove(crash_base, crash_size);
 691         pr_info("Reserving %lluMB of memory at %lluMB "
 692                 "for crashkernel (System RAM: %luMB)\n",
 693                 crash_size >> 20, crash_base >> 20,
 694                 (unsigned long)memblock.memory.total_size >> 20);
 695         os_info_crashkernel_add(crash_base, crash_size);
 696 #endif
 697 }
 698
 699 /*
 700  * Reserve the initrd from being used by memblock
 701  */
 702 static void __init reserve_initrd(void)
 703 {
 704 #ifdef CONFIG_BLK_DEV_INITRD
 705         if (!initrd_data.start || !initrd_data.size)
 706                 return;
 707         initrd_start = initrd_data.start;
 708         initrd_end = initrd_start + initrd_data.size;
 709         memblock_reserve(initrd_data.start, initrd_data.size);
 710 #endif
 711 }
 712
 713 /*
 714  * Reserve the memory area used to pass the certificate lists
 715  */
 716 static void __init reserve_certificate_list(void)
 717 {
 718         if (ipl_cert_list_addr)
 719                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 720 }
 721
 722 static void __init reserve_mem_detect_info(void)
 723 {
 724         unsigned long start, size;
 725
 726         get_mem_detect_reserved(&start, &size);
 727         if (size)
 728                 memblock_reserve(start, size);
 729 }
 730
 731 static void __init free_mem_detect_info(void)
 732 {
 733         unsigned long start, size;
 734
 735         get_mem_detect_reserved(&start, &size);
 736         if (size)
 737                 memblock_free(start, size);
 738 }
 739
 740 static const char * __init get_mem_info_source(void)
 741 {
 742         switch (mem_detect.info_source) {
 743         case MEM_DETECT_SCLP_STOR_INFO:
 744                 return "sclp storage info";
 745         case MEM_DETECT_DIAG260:
 746                 return "diag260";
 747         case MEM_DETECT_SCLP_READ_INFO:
 748                 return "sclp read info";
 749         case MEM_DETECT_BIN_SEARCH:
 750                 return "binary search";
 751         }
 752         return "none";
 753 }
 754
 755 static void __init memblock_add_mem_detect_info(void)
 756 {
 757         unsigned long start, end;
 758         int i;
 759
 760         pr_debug("physmem info source: %s (%hhd)\n",
 761                  get_mem_info_source(), mem_detect.info_source);
 762         /* keep memblock lists close to the kernel */
 763         memblock_set_bottom_up(true);
 764         for_each_mem_detect_block(i, &start, &end) {
 765                 memblock_add(start, end - start);
 766                 memblock_physmem_add(start, end - start);
 767         }
 768         memblock_set_bottom_up(false);
 769         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 770         memblock_dump_all();
 771 }
 772
 773 /*
 774  * Check for initrd being in usable memory
 775  */
 776 static void __init check_initrd(void)
 777 {
 778 #ifdef CONFIG_BLK_DEV_INITRD
 779         if (initrd_data.start && initrd_data.size &&
 780             !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
 781                 pr_err("The initial RAM disk does not fit into the memory\n");
 782                 memblock_free(initrd_data.start, initrd_data.size);
 783                 initrd_start = initrd_end = 0;
 784         }
 785 #endif
 786 }
 787
 788 /*
 789  * Reserve memory used for lowcore/command line/kernel image.
 790  */
 791 static void __init reserve_kernel(void)
 792 {
 793         unsigned long start_pfn = PFN_UP(__pa(_end));
 794
 795         memblock_reserve(0, STARTUP_NORMAL_OFFSET);
 796         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 797                          - (unsigned long)_stext);
 798 }
 799
 800 static void __init setup_memory(void)
 801 {
 802         phys_addr_t start, end;
 803         u64 i;
 804
 805         /*
 806          * Init storage key for present memory
 807          */
 808         for_each_mem_range(i, &start, &end)
 809                 storage_key_init_range(start, end);
 810
 811         psw_set_key(PAGE_DEFAULT_KEY);
 812
 813         /* Only cosmetics */
 814         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 815 }
 816
 817 static void __init relocate_dma_section(void)
 818 {
 819         unsigned long dma_addr, dma_size;
 820         long dma_offset;
 821         long *ptr;
 822
 823         /* Allocate a new DMA capable memory region */
 824         dma_size = __edma - __sdma;
 825         pr_info("Relocating DMA section of size 0x%08lx\n", dma_size);
 826         dma_addr = (unsigned long)memblock_alloc_low(dma_size, PAGE_SIZE);
 827         if (!dma_addr)
 828                 panic("Failed to allocate memory for DMA section\n");
 829         dma_offset = dma_addr - __sdma;
 830
 831         /* Move original DMA section to the new one */
 832         memmove((void *)dma_addr, (void *)__sdma, dma_size);
 833         /* Zero out the old DMA section to catch invalid accesses within it */
 834         memset((void *)__sdma, 0, dma_size);
 835
 836         /* Update all DMA region references */
 837         for (ptr = _start_dma_refs; ptr != _end_dma_refs; ptr++)
 838                 *ptr += dma_offset;
 839 }
 840
 841 /* This must be called after DMA relocation */
 842 static void __init setup_cr(void)
 843 {
 844         union ctlreg2 cr2;
 845         union ctlreg5 cr5;
 846         union ctlreg15 cr15;
 847
 848         __ctl_duct[1] = (unsigned long)__ctl_aste;
 849         __ctl_duct[2] = (unsigned long)__ctl_aste;
 850         __ctl_duct[4] = (unsigned long)__ctl_duald;
 851
 852         /* Update control registers CR2, CR5 and CR15 */
 853         __ctl_store(cr2.val, 2, 2);
 854         __ctl_store(cr5.val, 5, 5);
 855         __ctl_store(cr15.val, 15, 15);
 856         cr2.ducto = (unsigned long)__ctl_duct >> 6;
 857         cr5.pasteo = (unsigned long)__ctl_duct >> 6;
 858         cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
 859         __ctl_load(cr2.val, 2, 2);
 860         __ctl_load(cr5.val, 5, 5);
 861         __ctl_load(cr15.val, 15, 15);
 862 }
 863
 864 /*
 865  * Add system information as device randomness
 866  */
 867 static void __init setup_randomness(void)
 868 {
 869         struct sysinfo_3_2_2 *vmms;
 870
 871         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 872                                                             PAGE_SIZE);
 873         if (!vmms)
 874                 panic("Failed to allocate memory for sysinfo structure\n");
 875
 876         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 877                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 878         memblock_free((unsigned long) vmms, PAGE_SIZE);
 879 }
 880
 881 /*
 882  * Find the correct size for the task_struct. This depends on
 883  * the size of the struct fpu at the end of the thread_struct
 884  * which is embedded in the task_struct.
 885  */
 886 static void __init setup_task_size(void)
 887 {
 888         int task_size = sizeof(struct task_struct);
 889
 890         if (!MACHINE_HAS_VX) {
 891                 task_size -= sizeof(__vector128) * __NUM_VXRS;
 892                 task_size += sizeof(freg_t) * __NUM_FPRS;
 893         }
 894         arch_task_struct_size = task_size;
 895 }
 896
 897 /*
 898  * Issue diagnose 318 to set the control program name and
 899  * version codes.
 900  */
 901 static void __init setup_control_program_code(void)
 902 {
 903         union diag318_info diag318_info = {
 904                 .cpnc = CPNC_LINUX,
 905                 .cpvc = 0,
 906         };
 907
 908         if (!sclp.has_diag318)
 909                 return;
 910
 911         diag_stat_inc(DIAG_STAT_X318);
 912         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
 913 }
 914
 915 /*
 916  * Print the component list from the IPL report
 917  */
 918 static void __init log_component_list(void)
 919 {
 920         struct ipl_rb_component_entry *ptr, *end;
 921         char *str;
 922
 923         if (!early_ipl_comp_list_addr)
 924                 return;
 925         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
 926                 pr_info("Linux is running with Secure-IPL enabled\n");
 927         else
 928                 pr_info("Linux is running with Secure-IPL disabled\n");
 929         ptr = (void *) early_ipl_comp_list_addr;
 930         end = (void *) ptr + early_ipl_comp_list_size;
 931         pr_info("The IPL report contains the following components:\n");
 932         while (ptr < end) {
 933                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
 934                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
 935                                 str = "signed, verified";
 936                         else
 937                                 str = "signed, verification failed";
 938                 } else {
 939                         str = "not signed";
 940                 }
 941                 pr_info("%016llx - %016llx (%s)\n",
 942                         ptr->addr, ptr->addr + ptr->len, str);
 943                 ptr++;
 944         }
 945 }
 946
 947 /*
 948  * Setup function called from init/main.c just after the banner
 949  * was printed.
 950  */
 951
 952 void __init setup_arch(char **cmdline_p)
 953 {
 954         /*
 955          * print what head.S has found out about the machine
 956          */
 957         if (MACHINE_IS_VM)
 958                 pr_info("Linux is running as a z/VM "
 959                         "guest operating system in 64-bit mode\n");
 960         else if (MACHINE_IS_KVM)
 961                 pr_info("Linux is running under KVM in 64-bit mode\n");
 962         else if (MACHINE_IS_LPAR)
 963                 pr_info("Linux is running natively in 64-bit mode\n");
 964         else
 965                 pr_info("Linux is running as a guest in 64-bit mode\n");
 966
 967         log_component_list();
 968
 969         /* Have one command line that is parsed and saved in /proc/cmdline */
 970         /* boot_command_line has been already set up in early.c */
 971         *cmdline_p = boot_command_line;
 972
 973         ROOT_DEV = Root_RAM0;
 974
 975         setup_initial_init_mm(_text, _etext, _edata, _end);
 976
 977         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
 978                 nospec_auto_detect();
 979
 980         jump_label_init();
 981         parse_early_param();
 982 #ifdef CONFIG_CRASH_DUMP
 983         /* Deactivate elfcorehdr= kernel parameter */
 984         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 985 #endif
 986
 987         os_info_init();
 988         setup_ipl();
 989         setup_task_size();
 990         setup_control_program_code();
 991
 992         /* Do some memory reservations *before* memory is added to memblock */
 993         reserve_above_ident_map();
 994         reserve_kernel();
 995         reserve_initrd();
 996         reserve_certificate_list();
 997         reserve_mem_detect_info();
 998         memblock_allow_resize();
 999
1000         /* Get information about *all* installed memory */
1001         memblock_add_mem_detect_info();
1002
1003         free_mem_detect_info();
1004
1005         relocate_dma_section();
1006         setup_cr();
1007
1008         setup_uv();
1009         setup_memory_end();
1010         setup_memory();
1011         dma_contiguous_reserve(ident_map_size);
1012         vmcp_cma_reserve();
1013         if (MACHINE_HAS_EDAT2)
1014                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1015
1016         check_initrd();
1017         reserve_crashkernel();
1018 #ifdef CONFIG_CRASH_DUMP
1019         /*
1020          * Be aware that smp_save_dump_cpus() triggers a system reset.
1021          * Therefore CPU and device initialization should be done afterwards.
1022          */
1023         smp_save_dump_cpus();
1024 #endif
1025
1026         setup_resources();
1027         setup_lowcore_dat_off();
1028         smp_fill_possible_mask();
1029         cpu_detect_mhz_feature();
1030         cpu_init();
1031         numa_setup();
1032         smp_detect_cpus();
1033         topology_init_early();
1034
1035         /*
1036          * Create kernel page tables and switch to virtual addressing.
1037          */
1038         paging_init();
1039
1040         /*
1041          * After paging_init created the kernel page table, the new PSWs
1042          * in lowcore can now run with DAT enabled.
1043          */
1044         setup_lowcore_dat_on();
1045
1046         /* Setup default console */
1047         conmode_default();
1048         set_preferred_console();
1049
1050         apply_alternative_instructions();
1051         if (IS_ENABLED(CONFIG_EXPOLINE))
1052                 nospec_init_branches();
1053
1054         /* Setup zfcp/nvme dump support */
1055         setup_zfcpdump();
1056
1057         /* Add system specific data to the random pool */
1058         setup_randomness();
1059 }