3 * Copyright IBM Corp. 1999, 2012
4 * Author(s): Hartmut Penner (hp@de.ibm.com),
5 * Martin Schwidefsky (schwidefsky@de.ibm.com)
7 * Derived from "arch/i386/kernel/setup.c"
8 * Copyright (C) 1995, Linus Torvalds
12 * This file handles the architecture-dependent parts of initialization
15 #define KMSG_COMPONENT "setup"
16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18 #include <linux/errno.h>
19 #include <linux/export.h>
20 #include <linux/sched.h>
21 #include <linux/kernel.h>
22 #include <linux/memblock.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/ptrace.h>
27 #include <linux/user.h>
28 #include <linux/tty.h>
29 #include <linux/ioport.h>
30 #include <linux/delay.h>
31 #include <linux/init.h>
32 #include <linux/initrd.h>
33 #include <linux/bootmem.h>
34 #include <linux/root_dev.h>
35 #include <linux/console.h>
36 #include <linux/kernel_stat.h>
37 #include <linux/device.h>
38 #include <linux/notifier.h>
39 #include <linux/pfn.h>
40 #include <linux/ctype.h>
41 #include <linux/reboot.h>
42 #include <linux/topology.h>
43 #include <linux/ftrace.h>
44 #include <linux/kexec.h>
45 #include <linux/crash_dump.h>
46 #include <linux/memory.h>
47 #include <linux/compat.h>
50 #include <asm/uaccess.h>
51 #include <asm/facility.h>
53 #include <asm/mmu_context.h>
54 #include <asm/cpcmd.h>
55 #include <asm/lowcore.h>
58 #include <asm/ptrace.h>
59 #include <asm/sections.h>
60 #include <asm/ebcdic.h>
61 #include <asm/kvm_virtio.h>
63 #include <asm/os_info.h>
67 long psw_kernel_bits = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
68 PSW_MASK_EA | PSW_MASK_BA;
69 long psw_user_bits = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
70 PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
71 PSW_MASK_PSTATE | PSW_ASC_HOME;
74 * User copy operations.
76 struct uaccess_ops uaccess;
77 EXPORT_SYMBOL(uaccess);
82 unsigned int console_mode = 0;
83 EXPORT_SYMBOL(console_mode);
85 unsigned int console_devno = -1;
86 EXPORT_SYMBOL(console_devno);
88 unsigned int console_irq = -1;
89 EXPORT_SYMBOL(console_irq);
91 unsigned long elf_hwcap = 0;
92 char elf_platform[ELF_PLATFORM_SIZE];
94 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
96 int __initdata memory_end_set;
97 unsigned long __initdata memory_end;
99 unsigned long VMALLOC_START;
100 EXPORT_SYMBOL(VMALLOC_START);
102 unsigned long VMALLOC_END;
103 EXPORT_SYMBOL(VMALLOC_END);
105 struct page *vmemmap;
106 EXPORT_SYMBOL(vmemmap);
109 unsigned long MODULES_VADDR;
110 unsigned long MODULES_END;
113 /* An array with a pointer to the lowcore of every CPU. */
114 struct _lowcore *lowcore_ptr[NR_CPUS];
115 EXPORT_SYMBOL(lowcore_ptr);
118 * This is set up by the setup-routine at boot-time
119 * for S390 need to find out, what we have to setup
120 * using address 0x10400 ...
123 #include <asm/setup.h>
126 * condev= and conmode= setup parameter.
129 static int __init condev_setup(char *str)
133 vdev = simple_strtoul(str, &str, 0);
134 if (vdev >= 0 && vdev < 65536) {
135 console_devno = vdev;
141 __setup("condev=", condev_setup);
143 static void __init set_preferred_console(void)
145 if (MACHINE_IS_KVM) {
146 if (sclp_has_vt220())
147 add_preferred_console("ttyS", 1, NULL);
148 else if (sclp_has_linemode())
149 add_preferred_console("ttyS", 0, NULL);
151 add_preferred_console("hvc", 0, NULL);
152 } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
153 add_preferred_console("ttyS", 0, NULL);
154 else if (CONSOLE_IS_3270)
155 add_preferred_console("tty3270", 0, NULL);
158 static int __init conmode_setup(char *str)
160 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
161 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
164 #if defined(CONFIG_TN3215_CONSOLE)
165 if (strncmp(str, "3215", 5) == 0)
168 #if defined(CONFIG_TN3270_CONSOLE)
169 if (strncmp(str, "3270", 5) == 0)
172 set_preferred_console();
176 __setup("conmode=", conmode_setup);
178 static void __init conmode_default(void)
180 char query_buffer[1024];
184 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
185 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
186 ptr = strstr(query_buffer, "SUBCHANNEL =");
187 console_irq = simple_strtoul(ptr + 13, NULL, 16);
188 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
189 ptr = strstr(query_buffer, "CONMODE");
191 * Set the conmode to 3215 so that the device recognition
192 * will set the cu_type of the console to 3215. If the
193 * conmode is 3270 and we don't set it back then both
194 * 3215 and the 3270 driver will try to access the console
195 * device (3215 as console and 3270 as normal tty).
197 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
199 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 if (strncmp(ptr + 8, "3270", 4) == 0) {
205 #if defined(CONFIG_TN3270_CONSOLE)
207 #elif defined(CONFIG_TN3215_CONSOLE)
209 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
212 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
213 #if defined(CONFIG_TN3215_CONSOLE)
215 #elif defined(CONFIG_TN3270_CONSOLE)
217 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
222 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
228 #ifdef CONFIG_ZFCPDUMP
229 static void __init setup_zfcpdump(unsigned int console_devno)
233 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
237 if (console_devno != -1)
238 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
239 ipl_info.data.fcp.dev_id.devno, console_devno);
241 sprintf(str, " cio_ignore=all,!0.0.%04x",
242 ipl_info.data.fcp.dev_id.devno);
243 strcat(boot_command_line, str);
244 console_loglevel = 2;
247 static inline void setup_zfcpdump(unsigned int console_devno) {}
248 #endif /* CONFIG_ZFCPDUMP */
251 * Reboot, halt and power_off stubs. They just call _machine_restart,
252 * _machine_halt or _machine_power_off.
255 void machine_restart(char *command)
257 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
259 * Only unblank the console if we are called in enabled
260 * context or a bust_spinlocks cleared the way for us.
263 _machine_restart(command);
266 void machine_halt(void)
268 if (!in_interrupt() || oops_in_progress)
270 * Only unblank the console if we are called in enabled
271 * context or a bust_spinlocks cleared the way for us.
277 void machine_power_off(void)
279 if (!in_interrupt() || oops_in_progress)
281 * Only unblank the console if we are called in enabled
282 * context or a bust_spinlocks cleared the way for us.
285 _machine_power_off();
289 * Dummy power off function.
291 void (*pm_power_off)(void) = machine_power_off;
292 EXPORT_SYMBOL_GPL(pm_power_off);
294 static int __init early_parse_mem(char *p)
296 memory_end = memparse(p, &p);
300 early_param("mem", early_parse_mem);
302 static int __init parse_vmalloc(char *arg)
306 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
309 early_param("vmalloc", parse_vmalloc);
311 unsigned int s390_user_mode = PRIMARY_SPACE_MODE;
312 EXPORT_SYMBOL_GPL(s390_user_mode);
314 static void __init set_user_mode_primary(void)
316 psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
317 psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
320 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
322 uaccess = MACHINE_HAS_MVCOS ? uaccess_mvcos_switch : uaccess_pt;
325 static int __init early_parse_user_mode(char *p)
327 if (p && strcmp(p, "primary") == 0)
328 s390_user_mode = PRIMARY_SPACE_MODE;
329 else if (!p || strcmp(p, "home") == 0)
330 s390_user_mode = HOME_SPACE_MODE;
335 early_param("user_mode", early_parse_user_mode);
337 static void __init setup_addressing_mode(void)
339 if (s390_user_mode != PRIMARY_SPACE_MODE)
341 set_user_mode_primary();
342 if (MACHINE_HAS_MVCOS)
343 pr_info("Address spaces switched, mvcos available\n");
345 pr_info("Address spaces switched, mvcos not available\n");
348 void *restart_stack __attribute__((__section__(".data")));
350 static void __init setup_lowcore(void)
355 * Setup lowcore for boot cpu
357 BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
358 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
359 lc->restart_psw.mask = psw_kernel_bits;
360 lc->restart_psw.addr =
361 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
362 lc->external_new_psw.mask = psw_kernel_bits |
363 PSW_MASK_DAT | PSW_MASK_MCHECK;
364 lc->external_new_psw.addr =
365 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
366 lc->svc_new_psw.mask = psw_kernel_bits |
367 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
368 lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
369 lc->program_new_psw.mask = psw_kernel_bits |
370 PSW_MASK_DAT | PSW_MASK_MCHECK;
371 lc->program_new_psw.addr =
372 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
373 lc->mcck_new_psw.mask = psw_kernel_bits;
374 lc->mcck_new_psw.addr =
375 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
376 lc->io_new_psw.mask = psw_kernel_bits |
377 PSW_MASK_DAT | PSW_MASK_MCHECK;
378 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
379 lc->clock_comparator = -1ULL;
380 lc->kernel_stack = ((unsigned long) &init_thread_union)
381 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
382 lc->async_stack = (unsigned long)
383 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
384 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
385 lc->panic_stack = (unsigned long)
386 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
387 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
388 lc->current_task = (unsigned long) init_thread_union.thread_info.task;
389 lc->thread_info = (unsigned long) &init_thread_union;
390 lc->machine_flags = S390_lowcore.machine_flags;
391 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
392 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
395 if (MACHINE_HAS_IEEE) {
396 lc->extended_save_area_addr = (__u32)
397 __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
398 /* enable extended save area */
399 __ctl_set_bit(14, 29);
402 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
404 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
405 lc->async_enter_timer = S390_lowcore.async_enter_timer;
406 lc->exit_timer = S390_lowcore.exit_timer;
407 lc->user_timer = S390_lowcore.user_timer;
408 lc->system_timer = S390_lowcore.system_timer;
409 lc->steal_timer = S390_lowcore.steal_timer;
410 lc->last_update_timer = S390_lowcore.last_update_timer;
411 lc->last_update_clock = S390_lowcore.last_update_clock;
412 lc->ftrace_func = S390_lowcore.ftrace_func;
414 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
415 restart_stack += ASYNC_SIZE;
418 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
419 * restart data to the absolute zero lowcore. This is necesary if
420 * PSW restart is done on an offline CPU that has lowcore zero.
422 lc->restart_stack = (unsigned long) restart_stack;
423 lc->restart_fn = (unsigned long) do_restart;
424 lc->restart_data = 0;
425 lc->restart_source = -1UL;
427 /* Setup absolute zero lowcore */
428 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
429 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
430 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
431 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
432 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
434 set_prefix((u32)(unsigned long) lc);
438 static struct resource code_resource = {
439 .name = "Kernel code",
440 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
443 static struct resource data_resource = {
444 .name = "Kernel data",
445 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
448 static struct resource bss_resource = {
449 .name = "Kernel bss",
450 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
453 static struct resource __initdata *standard_resources[] = {
459 static void __init setup_resources(void)
461 struct resource *res, *std_res, *sub_res;
464 code_resource.start = (unsigned long) &_text;
465 code_resource.end = (unsigned long) &_etext - 1;
466 data_resource.start = (unsigned long) &_etext;
467 data_resource.end = (unsigned long) &_edata - 1;
468 bss_resource.start = (unsigned long) &__bss_start;
469 bss_resource.end = (unsigned long) &__bss_stop - 1;
471 for (i = 0; i < MEMORY_CHUNKS; i++) {
472 if (!memory_chunk[i].size)
474 if (memory_chunk[i].type == CHUNK_OLDMEM ||
475 memory_chunk[i].type == CHUNK_CRASHK)
477 res = alloc_bootmem_low(sizeof(*res));
478 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
479 switch (memory_chunk[i].type) {
480 case CHUNK_READ_WRITE:
482 res->name = "System RAM";
484 case CHUNK_READ_ONLY:
485 res->name = "System ROM";
486 res->flags |= IORESOURCE_READONLY;
489 res->name = "reserved";
491 res->start = memory_chunk[i].addr;
492 res->end = res->start + memory_chunk[i].size - 1;
493 request_resource(&iomem_resource, res);
495 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
496 std_res = standard_resources[j];
497 if (std_res->start < res->start ||
498 std_res->start > res->end)
500 if (std_res->end > res->end) {
501 sub_res = alloc_bootmem_low(sizeof(*sub_res));
503 sub_res->end = res->end;
504 std_res->start = res->end + 1;
505 request_resource(res, sub_res);
507 request_resource(res, std_res);
513 unsigned long real_memory_size;
514 EXPORT_SYMBOL_GPL(real_memory_size);
516 static void __init setup_memory_end(void)
518 unsigned long vmax, vmalloc_size, tmp;
522 #ifdef CONFIG_ZFCPDUMP
523 if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
524 memory_end = ZFCPDUMP_HSA_SIZE;
528 real_memory_size = 0;
529 memory_end &= PAGE_MASK;
532 * Make sure all chunks are MAX_ORDER aligned so we don't need the
533 * extra checks that HOLES_IN_ZONE would require.
535 for (i = 0; i < MEMORY_CHUNKS; i++) {
536 unsigned long start, end;
537 struct mem_chunk *chunk;
540 chunk = &memory_chunk[i];
541 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
542 start = (chunk->addr + align - 1) & ~(align - 1);
543 end = (chunk->addr + chunk->size) & ~(align - 1);
545 memset(chunk, 0, sizeof(*chunk));
548 chunk->size = end - start;
550 real_memory_size = max(real_memory_size,
551 chunk->addr + chunk->size);
554 /* Choose kernel address space layout: 2, 3, or 4 levels. */
556 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
557 tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
558 tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
559 if (tmp <= (1UL << 42))
560 vmax = 1UL << 42; /* 3-level kernel page table */
562 vmax = 1UL << 53; /* 4-level kernel page table */
563 /* module area is at the end of the kernel address space. */
565 MODULES_VADDR = MODULES_END - MODULES_LEN;
566 VMALLOC_END = MODULES_VADDR;
568 vmalloc_size = VMALLOC_END ?: 96UL << 20;
569 vmax = 1UL << 31; /* 2-level kernel page table */
570 /* vmalloc area is at the end of the kernel address space. */
573 VMALLOC_START = vmax - vmalloc_size;
575 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
576 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
577 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
578 tmp = SECTION_ALIGN_UP(tmp);
579 tmp = VMALLOC_START - tmp * sizeof(struct page);
580 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
581 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
582 vmemmap = (struct page *) tmp;
584 /* Take care that memory_end is set and <= vmemmap */
585 memory_end = min(memory_end ?: real_memory_size, tmp);
587 /* Fixup memory chunk array to fit into 0..memory_end */
588 for (i = 0; i < MEMORY_CHUNKS; i++) {
589 struct mem_chunk *chunk = &memory_chunk[i];
591 if (chunk->addr >= memory_end) {
592 memset(chunk, 0, sizeof(*chunk));
595 if (chunk->addr + chunk->size > memory_end)
596 chunk->size = memory_end - chunk->addr;
600 static void __init setup_vmcoreinfo(void)
602 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
605 #ifdef CONFIG_CRASH_DUMP
608 * Find suitable location for crashkernel memory
610 static unsigned long __init find_crash_base(unsigned long crash_size,
613 unsigned long crash_base;
614 struct mem_chunk *chunk;
617 if (memory_chunk[0].size < crash_size) {
618 *msg = "first memory chunk must be at least crashkernel size";
621 if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
624 for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
625 chunk = &memory_chunk[i];
626 if (chunk->size == 0)
628 if (chunk->type != CHUNK_READ_WRITE)
630 if (chunk->size < crash_size)
632 crash_base = (chunk->addr + chunk->size) - crash_size;
633 if (crash_base < crash_size)
635 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
637 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
641 *msg = "no suitable area found";
646 * Check if crash_base and crash_size is valid
648 static int __init verify_crash_base(unsigned long crash_base,
649 unsigned long crash_size,
652 struct mem_chunk *chunk;
656 * Because we do the swap to zero, we must have at least 'crash_size'
657 * bytes free space before crash_base
659 if (crash_size > crash_base) {
660 *msg = "crashkernel offset must be greater than size";
664 /* First memory chunk must be at least crash_size */
665 if (memory_chunk[0].size < crash_size) {
666 *msg = "first memory chunk must be at least crashkernel size";
669 /* Check if we fit into the respective memory chunk */
670 for (i = 0; i < MEMORY_CHUNKS; i++) {
671 chunk = &memory_chunk[i];
672 if (chunk->size == 0)
674 if (crash_base < chunk->addr)
676 if (crash_base >= chunk->addr + chunk->size)
678 /* we have found the memory chunk */
679 if (crash_base + crash_size > chunk->addr + chunk->size) {
680 *msg = "selected memory chunk is too small for "
681 "crashkernel memory";
686 *msg = "invalid memory range specified";
691 * Reserve kdump memory by creating a memory hole in the mem_chunk array
693 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
696 create_mem_hole(memory_chunk, addr, size, type);
700 * When kdump is enabled, we have to ensure that no memory from
701 * the area [0 - crashkernel memory size] and
702 * [crashk_res.start - crashk_res.end] is set offline.
704 static int kdump_mem_notifier(struct notifier_block *nb,
705 unsigned long action, void *data)
707 struct memory_notify *arg = data;
709 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
711 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
713 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
718 static struct notifier_block kdump_mem_nb = {
719 .notifier_call = kdump_mem_notifier,
725 * Make sure that oldmem, where the dump is stored, is protected
727 static void reserve_oldmem(void)
729 #ifdef CONFIG_CRASH_DUMP
733 reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
734 reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
736 if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
737 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
739 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
744 * Reserve memory for kdump kernel to be loaded with kexec
746 static void __init reserve_crashkernel(void)
748 #ifdef CONFIG_CRASH_DUMP
749 unsigned long long crash_base, crash_size;
753 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
755 if (rc || crash_size == 0)
757 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
758 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
759 if (register_memory_notifier(&kdump_mem_nb))
762 crash_base = find_crash_base(crash_size, &msg);
764 pr_info("crashkernel reservation failed: %s\n", msg);
765 unregister_memory_notifier(&kdump_mem_nb);
768 if (verify_crash_base(crash_base, crash_size, &msg)) {
769 pr_info("crashkernel reservation failed: %s\n", msg);
770 unregister_memory_notifier(&kdump_mem_nb);
773 if (!OLDMEM_BASE && MACHINE_IS_VM)
774 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
775 crashk_res.start = crash_base;
776 crashk_res.end = crash_base + crash_size - 1;
777 insert_resource(&iomem_resource, &crashk_res);
778 reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
779 pr_info("Reserving %lluMB of memory at %lluMB "
780 "for crashkernel (System RAM: %luMB)\n",
781 crash_size >> 20, crash_base >> 20, memory_end >> 20);
782 os_info_crashkernel_add(crash_base, crash_size);
786 static void __init setup_memory(void)
788 unsigned long bootmap_size;
789 unsigned long start_pfn, end_pfn;
793 * partially used pages are not usable - thus
794 * we are rounding upwards:
796 start_pfn = PFN_UP(__pa(&_end));
797 end_pfn = max_pfn = PFN_DOWN(memory_end);
799 #ifdef CONFIG_BLK_DEV_INITRD
801 * Move the initrd in case the bitmap of the bootmem allocater
802 * would overwrite it.
805 if (INITRD_START && INITRD_SIZE) {
806 unsigned long bmap_size;
809 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
810 bmap_size = PFN_PHYS(bmap_size);
812 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
813 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
815 #ifdef CONFIG_CRASH_DUMP
817 /* Move initrd behind kdump oldmem */
818 if (start + INITRD_SIZE > OLDMEM_BASE &&
819 start < OLDMEM_BASE + OLDMEM_SIZE)
820 start = OLDMEM_BASE + OLDMEM_SIZE;
823 if (start + INITRD_SIZE > memory_end) {
824 pr_err("initrd extends beyond end of "
825 "memory (0x%08lx > 0x%08lx) "
826 "disabling initrd\n",
827 start + INITRD_SIZE, memory_end);
828 INITRD_START = INITRD_SIZE = 0;
830 pr_info("Moving initrd (0x%08lx -> "
831 "0x%08lx, size: %ld)\n",
832 INITRD_START, start, INITRD_SIZE);
833 memmove((void *) start, (void *) INITRD_START,
835 INITRD_START = start;
842 * Initialize the boot-time allocator
844 bootmap_size = init_bootmem(start_pfn, end_pfn);
847 * Register RAM areas with the bootmem allocator.
850 for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
851 unsigned long start_chunk, end_chunk, pfn;
853 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
854 memory_chunk[i].type != CHUNK_CRASHK)
856 start_chunk = PFN_DOWN(memory_chunk[i].addr);
857 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
858 end_chunk = min(end_chunk, end_pfn);
859 if (start_chunk >= end_chunk)
861 memblock_add_node(PFN_PHYS(start_chunk),
862 PFN_PHYS(end_chunk - start_chunk), 0);
863 pfn = max(start_chunk, start_pfn);
864 storage_key_init_range(PFN_PHYS(pfn), PFN_PHYS(end_chunk));
867 psw_set_key(PAGE_DEFAULT_KEY);
869 free_bootmem_with_active_regions(0, max_pfn);
872 * Reserve memory used for lowcore/command line/kernel image.
874 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
875 reserve_bootmem((unsigned long)_stext,
876 PFN_PHYS(start_pfn) - (unsigned long)_stext,
879 * Reserve the bootmem bitmap itself as well. We do this in two
880 * steps (first step was init_bootmem()) because this catches
881 * the (very unlikely) case of us accidentally initializing the
882 * bootmem allocator with an invalid RAM area.
884 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
887 #ifdef CONFIG_CRASH_DUMP
888 if (crashk_res.start)
889 reserve_bootmem(crashk_res.start,
890 crashk_res.end - crashk_res.start + 1,
892 if (is_kdump_kernel())
893 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
894 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
896 #ifdef CONFIG_BLK_DEV_INITRD
897 if (INITRD_START && INITRD_SIZE) {
898 if (INITRD_START + INITRD_SIZE <= memory_end) {
899 reserve_bootmem(INITRD_START, INITRD_SIZE,
901 initrd_start = INITRD_START;
902 initrd_end = initrd_start + INITRD_SIZE;
904 pr_err("initrd extends beyond end of "
905 "memory (0x%08lx > 0x%08lx) "
906 "disabling initrd\n",
907 initrd_start + INITRD_SIZE, memory_end);
908 initrd_start = initrd_end = 0;
915 * Setup hardware capabilities.
917 static void __init setup_hwcaps(void)
919 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
924 * The store facility list bits numbers as found in the principles
925 * of operation are numbered with bit 1UL<<31 as number 0 to
926 * bit 1UL<<0 as number 31.
927 * Bit 0: instructions named N3, "backported" to esa-mode
928 * Bit 2: z/Architecture mode is active
929 * Bit 7: the store-facility-list-extended facility is installed
930 * Bit 17: the message-security assist is installed
931 * Bit 19: the long-displacement facility is installed
932 * Bit 21: the extended-immediate facility is installed
933 * Bit 22: extended-translation facility 3 is installed
934 * Bit 30: extended-translation facility 3 enhancement facility
935 * These get translated to:
936 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
937 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
938 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
939 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
941 for (i = 0; i < 6; i++)
942 if (test_facility(stfl_bits[i]))
943 elf_hwcap |= 1UL << i;
945 if (test_facility(22) && test_facility(30))
946 elf_hwcap |= HWCAP_S390_ETF3EH;
949 * Check for additional facilities with store-facility-list-extended.
950 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
951 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
952 * as stored by stfl, bits 32-xxx contain additional facilities.
953 * How many facility words are stored depends on the number of
954 * doublewords passed to the instruction. The additional facilities
956 * Bit 42: decimal floating point facility is installed
957 * Bit 44: perform floating point operation facility is installed
959 * HWCAP_S390_DFP bit 6 (42 && 44).
961 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
962 elf_hwcap |= HWCAP_S390_DFP;
965 * Huge page support HWCAP_S390_HPAGE is bit 7.
967 if (MACHINE_HAS_HPAGE)
968 elf_hwcap |= HWCAP_S390_HPAGE;
970 #if defined(CONFIG_64BIT)
972 * 64-bit register support for 31-bit processes
973 * HWCAP_S390_HIGH_GPRS is bit 9.
975 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
978 * Transactional execution support HWCAP_S390_TE is bit 10.
980 if (test_facility(50) && test_facility(73))
981 elf_hwcap |= HWCAP_S390_TE;
985 switch (cpu_id.machine) {
987 #if !defined(CONFIG_64BIT)
988 default: /* Use "g5" as default for 31 bit kernels. */
990 strcpy(elf_platform, "g5");
994 #if defined(CONFIG_64BIT)
995 default: /* Use "z900" as default for 64 bit kernels. */
997 strcpy(elf_platform, "z900");
1001 strcpy(elf_platform, "z990");
1005 strcpy(elf_platform, "z9-109");
1009 strcpy(elf_platform, "z10");
1013 strcpy(elf_platform, "z196");
1016 strcpy(elf_platform, "zEC12");
1022 * Setup function called from init/main.c just after the banner
1026 void __init setup_arch(char **cmdline_p)
1029 * print what head.S has found out about the machine
1031 #ifndef CONFIG_64BIT
1033 pr_info("Linux is running as a z/VM "
1034 "guest operating system in 31-bit mode\n");
1035 else if (MACHINE_IS_LPAR)
1036 pr_info("Linux is running natively in 31-bit mode\n");
1037 if (MACHINE_HAS_IEEE)
1038 pr_info("The hardware system has IEEE compatible "
1039 "floating point units\n");
1041 pr_info("The hardware system has no IEEE compatible "
1042 "floating point units\n");
1043 #else /* CONFIG_64BIT */
1045 pr_info("Linux is running as a z/VM "
1046 "guest operating system in 64-bit mode\n");
1047 else if (MACHINE_IS_KVM)
1048 pr_info("Linux is running under KVM in 64-bit mode\n");
1049 else if (MACHINE_IS_LPAR)
1050 pr_info("Linux is running natively in 64-bit mode\n");
1051 #endif /* CONFIG_64BIT */
1053 /* Have one command line that is parsed and saved in /proc/cmdline */
1054 /* boot_command_line has been already set up in early.c */
1055 *cmdline_p = boot_command_line;
1057 ROOT_DEV = Root_RAM0;
1059 init_mm.start_code = PAGE_OFFSET;
1060 init_mm.end_code = (unsigned long) &_etext;
1061 init_mm.end_data = (unsigned long) &_edata;
1062 init_mm.brk = (unsigned long) &_end;
1064 if (MACHINE_HAS_MVCOS)
1065 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1067 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1069 parse_early_param();
1074 setup_addressing_mode();
1076 reserve_crashkernel();
1083 s390_init_cpu_topology();
1086 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1091 * Create kernel page tables and switch to virtual addressing.
1095 /* Setup default console */
1097 set_preferred_console();
1099 /* Setup zfcpdump support */
1100 setup_zfcpdump(console_devno);