1/*
2 *  S390 version
3 *    Copyright IBM Corp. 1999, 2012
4 *    Author(s): Hartmut Penner (hp@de.ibm.com),
5 *               Martin Schwidefsky (schwidefsky@de.ibm.com)
6 *
7 *  Derived from "arch/i386/kernel/setup.c"
8 *    Copyright (C) 1995, Linus Torvalds
9 */
10
11/*
12 * This file handles the architecture-dependent parts of initialization
13 */
14
15#define KMSG_COMPONENT "setup"
16#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17
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>
23#include <linux/mm.h>
24#include <linux/stddef.h>
25#include <linux/unistd.h>
26#include <linux/ptrace.h>
27#include <linux/random.h>
28#include <linux/user.h>
29#include <linux/tty.h>
30#include <linux/ioport.h>
31#include <linux/delay.h>
32#include <linux/init.h>
33#include <linux/initrd.h>
34#include <linux/bootmem.h>
35#include <linux/root_dev.h>
36#include <linux/console.h>
37#include <linux/kernel_stat.h>
38#include <linux/device.h>
39#include <linux/notifier.h>
40#include <linux/pfn.h>
41#include <linux/ctype.h>
42#include <linux/reboot.h>
43#include <linux/topology.h>
44#include <linux/ftrace.h>
45#include <linux/kexec.h>
46#include <linux/crash_dump.h>
47#include <linux/memory.h>
48#include <linux/compat.h>
49
50#include <asm/ipl.h>
51#include <asm/facility.h>
52#include <asm/smp.h>
53#include <asm/mmu_context.h>
54#include <asm/cpcmd.h>
55#include <asm/lowcore.h>
56#include <asm/irq.h>
57#include <asm/page.h>
58#include <asm/ptrace.h>
59#include <asm/sections.h>
60#include <asm/ebcdic.h>
61#include <asm/kvm_virtio.h>
62#include <asm/diag.h>
63#include <asm/os_info.h>
64#include <asm/sclp.h>
65#include <asm/sysinfo.h>
66#include "entry.h"
67
68/*
69 * Machine setup..
70 */
71unsigned int console_mode = 0;
72EXPORT_SYMBOL(console_mode);
73
74unsigned int console_devno = -1;
75EXPORT_SYMBOL(console_devno);
76
77unsigned int console_irq = -1;
78EXPORT_SYMBOL(console_irq);
79
80unsigned long elf_hwcap = 0;
81char elf_platform[ELF_PLATFORM_SIZE];
82
83int __initdata memory_end_set;
84unsigned long __initdata memory_end;
85unsigned long __initdata max_physmem_end;
86
87unsigned long VMALLOC_START;
88EXPORT_SYMBOL(VMALLOC_START);
89
90unsigned long VMALLOC_END;
91EXPORT_SYMBOL(VMALLOC_END);
92
93struct page *vmemmap;
94EXPORT_SYMBOL(vmemmap);
95
96#ifdef CONFIG_64BIT
97unsigned long MODULES_VADDR;
98unsigned long MODULES_END;
99#endif
100
101/* An array with a pointer to the lowcore of every CPU. */
102struct _lowcore *lowcore_ptr[NR_CPUS];
103EXPORT_SYMBOL(lowcore_ptr);
104
105/*
106 * This is set up by the setup-routine at boot-time
107 * for S390 need to find out, what we have to setup
108 * using address 0x10400 ...
109 */
110
111#include <asm/setup.h>
112
113/*
114 * condev= and conmode= setup parameter.
115 */
116
117static int __init condev_setup(char *str)
118{
119	int vdev;
120
121	vdev = simple_strtoul(str, &str, 0);
122	if (vdev >= 0 && vdev < 65536) {
123		console_devno = vdev;
124		console_irq = -1;
125	}
126	return 1;
127}
128
129__setup("condev=", condev_setup);
130
131static void __init set_preferred_console(void)
132{
133	if (MACHINE_IS_KVM) {
134		if (sclp_has_vt220())
135			add_preferred_console("ttyS", 1, NULL);
136		else if (sclp_has_linemode())
137			add_preferred_console("ttyS", 0, NULL);
138		else
139			add_preferred_console("hvc", 0, NULL);
140	} else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
141		add_preferred_console("ttyS", 0, NULL);
142	else if (CONSOLE_IS_3270)
143		add_preferred_console("tty3270", 0, NULL);
144}
145
146static int __init conmode_setup(char *str)
147{
148#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
149	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
150                SET_CONSOLE_SCLP;
151#endif
152#if defined(CONFIG_TN3215_CONSOLE)
153	if (strncmp(str, "3215", 5) == 0)
154		SET_CONSOLE_3215;
155#endif
156#if defined(CONFIG_TN3270_CONSOLE)
157	if (strncmp(str, "3270", 5) == 0)
158		SET_CONSOLE_3270;
159#endif
160	set_preferred_console();
161        return 1;
162}
163
164__setup("conmode=", conmode_setup);
165
166static void __init conmode_default(void)
167{
168	char query_buffer[1024];
169	char *ptr;
170
171        if (MACHINE_IS_VM) {
172		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
173		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
174		ptr = strstr(query_buffer, "SUBCHANNEL =");
175		console_irq = simple_strtoul(ptr + 13, NULL, 16);
176		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
177		ptr = strstr(query_buffer, "CONMODE");
178		/*
179		 * Set the conmode to 3215 so that the device recognition
180		 * will set the cu_type of the console to 3215. If the
181		 * conmode is 3270 and we don't set it back then both
182		 * 3215 and the 3270 driver will try to access the console
183		 * device (3215 as console and 3270 as normal tty).
184		 */
185		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
186		if (ptr == NULL) {
187#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
188			SET_CONSOLE_SCLP;
189#endif
190			return;
191		}
192		if (strncmp(ptr + 8, "3270", 4) == 0) {
193#if defined(CONFIG_TN3270_CONSOLE)
194			SET_CONSOLE_3270;
195#elif defined(CONFIG_TN3215_CONSOLE)
196			SET_CONSOLE_3215;
197#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198			SET_CONSOLE_SCLP;
199#endif
200		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
201#if defined(CONFIG_TN3215_CONSOLE)
202			SET_CONSOLE_3215;
203#elif defined(CONFIG_TN3270_CONSOLE)
204			SET_CONSOLE_3270;
205#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
206			SET_CONSOLE_SCLP;
207#endif
208		}
209	} else {
210#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
211		SET_CONSOLE_SCLP;
212#endif
213	}
214}
215
216#ifdef CONFIG_CRASH_DUMP
217static void __init setup_zfcpdump(void)
218{
219	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
220		return;
221	if (OLDMEM_BASE)
222		return;
223	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
224	console_loglevel = 2;
225}
226#else
227static inline void setup_zfcpdump(void) {}
228#endif /* CONFIG_CRASH_DUMP */
229
230 /*
231 * Reboot, halt and power_off stubs. They just call _machine_restart,
232 * _machine_halt or _machine_power_off.
233 */
234
235void machine_restart(char *command)
236{
237	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
238		/*
239		 * Only unblank the console if we are called in enabled
240		 * context or a bust_spinlocks cleared the way for us.
241		 */
242		console_unblank();
243	_machine_restart(command);
244}
245
246void machine_halt(void)
247{
248	if (!in_interrupt() || oops_in_progress)
249		/*
250		 * Only unblank the console if we are called in enabled
251		 * context or a bust_spinlocks cleared the way for us.
252		 */
253		console_unblank();
254	_machine_halt();
255}
256
257void machine_power_off(void)
258{
259	if (!in_interrupt() || oops_in_progress)
260		/*
261		 * Only unblank the console if we are called in enabled
262		 * context or a bust_spinlocks cleared the way for us.
263		 */
264		console_unblank();
265	_machine_power_off();
266}
267
268/*
269 * Dummy power off function.
270 */
271void (*pm_power_off)(void) = machine_power_off;
272EXPORT_SYMBOL_GPL(pm_power_off);
273
274static int __init early_parse_mem(char *p)
275{
276	memory_end = memparse(p, &p);
277	memory_end &= PAGE_MASK;
278	memory_end_set = 1;
279	return 0;
280}
281early_param("mem", early_parse_mem);
282
283static int __init parse_vmalloc(char *arg)
284{
285	if (!arg)
286		return -EINVAL;
287	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
288	return 0;
289}
290early_param("vmalloc", parse_vmalloc);
291
292void *restart_stack __attribute__((__section__(".data")));
293
294static void __init setup_lowcore(void)
295{
296	struct _lowcore *lc;
297
298	/*
299	 * Setup lowcore for boot cpu
300	 */
301	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
302	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
303	lc->restart_psw.mask = PSW_KERNEL_BITS;
304	lc->restart_psw.addr =
305		PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
306	lc->external_new_psw.mask = PSW_KERNEL_BITS |
307		PSW_MASK_DAT | PSW_MASK_MCHECK;
308	lc->external_new_psw.addr =
309		PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
310	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
311		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
312	lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
313	lc->program_new_psw.mask = PSW_KERNEL_BITS |
314		PSW_MASK_DAT | PSW_MASK_MCHECK;
315	lc->program_new_psw.addr =
316		PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
317	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
318	lc->mcck_new_psw.addr =
319		PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
320	lc->io_new_psw.mask = PSW_KERNEL_BITS |
321		PSW_MASK_DAT | PSW_MASK_MCHECK;
322	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
323	lc->clock_comparator = -1ULL;
324	lc->kernel_stack = ((unsigned long) &init_thread_union)
325		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
326	lc->async_stack = (unsigned long)
327		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
328		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
329	lc->panic_stack = (unsigned long)
330		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
331		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
332	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
333	lc->thread_info = (unsigned long) &init_thread_union;
334	lc->machine_flags = S390_lowcore.machine_flags;
335	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
336	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
337	       MAX_FACILITY_BIT/8);
338#ifndef CONFIG_64BIT
339	if (MACHINE_HAS_IEEE) {
340		lc->extended_save_area_addr = (__u32)
341			__alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
342		/* enable extended save area */
343		__ctl_set_bit(14, 29);
344	}
345#else
346	if (MACHINE_HAS_VX)
347		lc->vector_save_area_addr =
348			(unsigned long) &lc->vector_save_area;
349	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
350#endif
351	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
352	lc->async_enter_timer = S390_lowcore.async_enter_timer;
353	lc->exit_timer = S390_lowcore.exit_timer;
354	lc->user_timer = S390_lowcore.user_timer;
355	lc->system_timer = S390_lowcore.system_timer;
356	lc->steal_timer = S390_lowcore.steal_timer;
357	lc->last_update_timer = S390_lowcore.last_update_timer;
358	lc->last_update_clock = S390_lowcore.last_update_clock;
359	lc->ftrace_func = S390_lowcore.ftrace_func;
360
361	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
362	restart_stack += ASYNC_SIZE;
363
364	/*
365	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
366	 * restart data to the absolute zero lowcore. This is necessary if
367	 * PSW restart is done on an offline CPU that has lowcore zero.
368	 */
369	lc->restart_stack = (unsigned long) restart_stack;
370	lc->restart_fn = (unsigned long) do_restart;
371	lc->restart_data = 0;
372	lc->restart_source = -1UL;
373
374	/* Setup absolute zero lowcore */
375	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
376	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
377	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
378	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
379	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
380
381#ifdef CONFIG_SMP
382	lc->spinlock_lockval = arch_spin_lockval(0);
383#endif
384
385	set_prefix((u32)(unsigned long) lc);
386	lowcore_ptr[0] = lc;
387}
388
389static struct resource code_resource = {
390	.name  = "Kernel code",
391	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
392};
393
394static struct resource data_resource = {
395	.name = "Kernel data",
396	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
397};
398
399static struct resource bss_resource = {
400	.name = "Kernel bss",
401	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
402};
403
404static struct resource __initdata *standard_resources[] = {
405	&code_resource,
406	&data_resource,
407	&bss_resource,
408};
409
410static void __init setup_resources(void)
411{
412	struct resource *res, *std_res, *sub_res;
413	struct memblock_region *reg;
414	int j;
415
416	code_resource.start = (unsigned long) &_text;
417	code_resource.end = (unsigned long) &_etext - 1;
418	data_resource.start = (unsigned long) &_etext;
419	data_resource.end = (unsigned long) &_edata - 1;
420	bss_resource.start = (unsigned long) &__bss_start;
421	bss_resource.end = (unsigned long) &__bss_stop - 1;
422
423	for_each_memblock(memory, reg) {
424		res = alloc_bootmem_low(sizeof(*res));
425		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
426
427		res->name = "System RAM";
428		res->start = reg->base;
429		res->end = reg->base + reg->size - 1;
430		request_resource(&iomem_resource, res);
431
432		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
433			std_res = standard_resources[j];
434			if (std_res->start < res->start ||
435			    std_res->start > res->end)
436				continue;
437			if (std_res->end > res->end) {
438				sub_res = alloc_bootmem_low(sizeof(*sub_res));
439				*sub_res = *std_res;
440				sub_res->end = res->end;
441				std_res->start = res->end + 1;
442				request_resource(res, sub_res);
443			} else {
444				request_resource(res, std_res);
445			}
446		}
447	}
448}
449
450static void __init setup_memory_end(void)
451{
452	unsigned long vmax, vmalloc_size, tmp;
453
454	/* Choose kernel address space layout: 2, 3, or 4 levels. */
455#ifdef CONFIG_64BIT
456	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
457	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
458	tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
459	if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
460		vmax = 1UL << 42;	/* 3-level kernel page table */
461	else
462		vmax = 1UL << 53;	/* 4-level kernel page table */
463	/* module area is at the end of the kernel address space. */
464	MODULES_END = vmax;
465	MODULES_VADDR = MODULES_END - MODULES_LEN;
466	VMALLOC_END = MODULES_VADDR;
467#else
468	vmalloc_size = VMALLOC_END ?: 96UL << 20;
469	vmax = 1UL << 31;		/* 2-level kernel page table */
470	/* vmalloc area is at the end of the kernel address space. */
471	VMALLOC_END = vmax;
472#endif
473	VMALLOC_START = vmax - vmalloc_size;
474
475	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
476	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
477	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
478	tmp = SECTION_ALIGN_UP(tmp);
479	tmp = VMALLOC_START - tmp * sizeof(struct page);
480	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
481	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
482	vmemmap = (struct page *) tmp;
483
484	/* Take care that memory_end is set and <= vmemmap */
485	memory_end = min(memory_end ?: max_physmem_end, tmp);
486	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
487	memblock_remove(memory_end, ULONG_MAX);
488
489	pr_notice("Max memory size: %luMB\n", memory_end >> 20);
490}
491
492static void __init setup_vmcoreinfo(void)
493{
494	mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
495}
496
497#ifdef CONFIG_CRASH_DUMP
498
499/*
500 * When kdump is enabled, we have to ensure that no memory from
501 * the area [0 - crashkernel memory size] and
502 * [crashk_res.start - crashk_res.end] is set offline.
503 */
504static int kdump_mem_notifier(struct notifier_block *nb,
505			      unsigned long action, void *data)
506{
507	struct memory_notify *arg = data;
508
509	if (action != MEM_GOING_OFFLINE)
510		return NOTIFY_OK;
511	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
512		return NOTIFY_BAD;
513	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
514		return NOTIFY_OK;
515	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
516		return NOTIFY_OK;
517	return NOTIFY_BAD;
518}
519
520static struct notifier_block kdump_mem_nb = {
521	.notifier_call = kdump_mem_notifier,
522};
523
524#endif
525
526/*
527 * Make sure that the area behind memory_end is protected
528 */
529static void reserve_memory_end(void)
530{
531#ifdef CONFIG_CRASH_DUMP
532	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
533	    !OLDMEM_BASE && sclp_get_hsa_size()) {
534		memory_end = sclp_get_hsa_size();
535		memory_end &= PAGE_MASK;
536		memory_end_set = 1;
537	}
538#endif
539	if (!memory_end_set)
540		return;
541	memblock_reserve(memory_end, ULONG_MAX);
542}
543
544/*
545 * Make sure that oldmem, where the dump is stored, is protected
546 */
547static void reserve_oldmem(void)
548{
549#ifdef CONFIG_CRASH_DUMP
550	if (OLDMEM_BASE)
551		/* Forget all memory above the running kdump system */
552		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
553#endif
554}
555
556/*
557 * Make sure that oldmem, where the dump is stored, is protected
558 */
559static void remove_oldmem(void)
560{
561#ifdef CONFIG_CRASH_DUMP
562	if (OLDMEM_BASE)
563		/* Forget all memory above the running kdump system */
564		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
565#endif
566}
567
568/*
569 * Reserve memory for kdump kernel to be loaded with kexec
570 */
571static void __init reserve_crashkernel(void)
572{
573#ifdef CONFIG_CRASH_DUMP
574	unsigned long long crash_base, crash_size;
575	phys_addr_t low, high;
576	int rc;
577
578	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
579			       &crash_base);
580
581	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
582	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
583	if (rc || crash_size == 0)
584		return;
585
586	if (memblock.memory.regions[0].size < crash_size) {
587		pr_info("crashkernel reservation failed: %s\n",
588			"first memory chunk must be at least crashkernel size");
589		return;
590	}
591
592	low = crash_base ?: OLDMEM_BASE;
593	high = low + crash_size;
594	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
595		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
596		crash_base = low;
597	} else {
598		/* Find suitable area in free memory */
599		low = max_t(unsigned long, crash_size, sclp_get_hsa_size());
600		high = crash_base ? crash_base + crash_size : ULONG_MAX;
601
602		if (crash_base && crash_base < low) {
603			pr_info("crashkernel reservation failed: %s\n",
604				"crash_base too low");
605			return;
606		}
607		low = crash_base ?: low;
608		crash_base = memblock_find_in_range(low, high, crash_size,
609						    KEXEC_CRASH_MEM_ALIGN);
610	}
611
612	if (!crash_base) {
613		pr_info("crashkernel reservation failed: %s\n",
614			"no suitable area found");
615		return;
616	}
617
618	if (register_memory_notifier(&kdump_mem_nb))
619		return;
620
621	if (!OLDMEM_BASE && MACHINE_IS_VM)
622		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
623	crashk_res.start = crash_base;
624	crashk_res.end = crash_base + crash_size - 1;
625	insert_resource(&iomem_resource, &crashk_res);
626	memblock_remove(crash_base, crash_size);
627	pr_info("Reserving %lluMB of memory at %lluMB "
628		"for crashkernel (System RAM: %luMB)\n",
629		crash_size >> 20, crash_base >> 20,
630		(unsigned long)memblock.memory.total_size >> 20);
631	os_info_crashkernel_add(crash_base, crash_size);
632#endif
633}
634
635/*
636 * Reserve the initrd from being used by memblock
637 */
638static void __init reserve_initrd(void)
639{
640#ifdef CONFIG_BLK_DEV_INITRD
641	initrd_start = INITRD_START;
642	initrd_end = initrd_start + INITRD_SIZE;
643	memblock_reserve(INITRD_START, INITRD_SIZE);
644#endif
645}
646
647/*
648 * Check for initrd being in usable memory
649 */
650static void __init check_initrd(void)
651{
652#ifdef CONFIG_BLK_DEV_INITRD
653	if (INITRD_START && INITRD_SIZE &&
654	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
655		pr_err("initrd does not fit memory.\n");
656		memblock_free(INITRD_START, INITRD_SIZE);
657		initrd_start = initrd_end = 0;
658	}
659#endif
660}
661
662/*
663 * Reserve all kernel text
664 */
665static void __init reserve_kernel(void)
666{
667	unsigned long start_pfn;
668	start_pfn = PFN_UP(__pa(&_end));
669
670	/*
671	 * Reserve memory used for lowcore/command line/kernel image.
672	 */
673	memblock_reserve(0, (unsigned long)_ehead);
674	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
675			 - (unsigned long)_stext);
676}
677
678static void __init reserve_elfcorehdr(void)
679{
680#ifdef CONFIG_CRASH_DUMP
681	if (is_kdump_kernel())
682		memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
683				 PAGE_ALIGN(elfcorehdr_size));
684#endif
685}
686
687static void __init setup_memory(void)
688{
689	struct memblock_region *reg;
690
691	/*
692	 * Init storage key for present memory
693	 */
694	for_each_memblock(memory, reg) {
695		storage_key_init_range(reg->base, reg->base + reg->size);
696	}
697	psw_set_key(PAGE_DEFAULT_KEY);
698
699	/* Only cosmetics */
700	memblock_enforce_memory_limit(memblock_end_of_DRAM());
701}
702
703/*
704 * Setup hardware capabilities.
705 */
706static void __init setup_hwcaps(void)
707{
708	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
709	struct cpuid cpu_id;
710	int i;
711
712	/*
713	 * The store facility list bits numbers as found in the principles
714	 * of operation are numbered with bit 1UL<<31 as number 0 to
715	 * bit 1UL<<0 as number 31.
716	 *   Bit 0: instructions named N3, "backported" to esa-mode
717	 *   Bit 2: z/Architecture mode is active
718	 *   Bit 7: the store-facility-list-extended facility is installed
719	 *   Bit 17: the message-security assist is installed
720	 *   Bit 19: the long-displacement facility is installed
721	 *   Bit 21: the extended-immediate facility is installed
722	 *   Bit 22: extended-translation facility 3 is installed
723	 *   Bit 30: extended-translation facility 3 enhancement facility
724	 * These get translated to:
725	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
726	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
727	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
728	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
729	 */
730	for (i = 0; i < 6; i++)
731		if (test_facility(stfl_bits[i]))
732			elf_hwcap |= 1UL << i;
733
734	if (test_facility(22) && test_facility(30))
735		elf_hwcap |= HWCAP_S390_ETF3EH;
736
737	/*
738	 * Check for additional facilities with store-facility-list-extended.
739	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
740	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
741	 * as stored by stfl, bits 32-xxx contain additional facilities.
742	 * How many facility words are stored depends on the number of
743	 * doublewords passed to the instruction. The additional facilities
744	 * are:
745	 *   Bit 42: decimal floating point facility is installed
746	 *   Bit 44: perform floating point operation facility is installed
747	 * translated to:
748	 *   HWCAP_S390_DFP bit 6 (42 && 44).
749	 */
750	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
751		elf_hwcap |= HWCAP_S390_DFP;
752
753	/*
754	 * Huge page support HWCAP_S390_HPAGE is bit 7.
755	 */
756	if (MACHINE_HAS_HPAGE)
757		elf_hwcap |= HWCAP_S390_HPAGE;
758
759#if defined(CONFIG_64BIT)
760	/*
761	 * 64-bit register support for 31-bit processes
762	 * HWCAP_S390_HIGH_GPRS is bit 9.
763	 */
764	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
765
766	/*
767	 * Transactional execution support HWCAP_S390_TE is bit 10.
768	 */
769	if (test_facility(50) && test_facility(73))
770		elf_hwcap |= HWCAP_S390_TE;
771
772	/*
773	 * Vector extension HWCAP_S390_VXRS is bit 11.
774	 */
775	if (test_facility(129))
776		elf_hwcap |= HWCAP_S390_VXRS;
777#endif
778
779	get_cpu_id(&cpu_id);
780	add_device_randomness(&cpu_id, sizeof(cpu_id));
781	switch (cpu_id.machine) {
782	case 0x9672:
783#if !defined(CONFIG_64BIT)
784	default:	/* Use "g5" as default for 31 bit kernels. */
785#endif
786		strcpy(elf_platform, "g5");
787		break;
788	case 0x2064:
789	case 0x2066:
790#if defined(CONFIG_64BIT)
791	default:	/* Use "z900" as default for 64 bit kernels. */
792#endif
793		strcpy(elf_platform, "z900");
794		break;
795	case 0x2084:
796	case 0x2086:
797		strcpy(elf_platform, "z990");
798		break;
799	case 0x2094:
800	case 0x2096:
801		strcpy(elf_platform, "z9-109");
802		break;
803	case 0x2097:
804	case 0x2098:
805		strcpy(elf_platform, "z10");
806		break;
807	case 0x2817:
808	case 0x2818:
809		strcpy(elf_platform, "z196");
810		break;
811	case 0x2827:
812	case 0x2828:
813		strcpy(elf_platform, "zEC12");
814		break;
815	}
816}
817
818/*
819 * Add system information as device randomness
820 */
821static void __init setup_randomness(void)
822{
823	struct sysinfo_3_2_2 *vmms;
824
825	vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
826	if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
827		add_device_randomness(&vmms, vmms->count);
828	free_page((unsigned long) vmms);
829}
830
831/*
832 * Setup function called from init/main.c just after the banner
833 * was printed.
834 */
835
836void __init setup_arch(char **cmdline_p)
837{
838        /*
839         * print what head.S has found out about the machine
840         */
841#ifndef CONFIG_64BIT
842	if (MACHINE_IS_VM)
843		pr_info("Linux is running as a z/VM "
844			"guest operating system in 31-bit mode\n");
845	else if (MACHINE_IS_LPAR)
846		pr_info("Linux is running natively in 31-bit mode\n");
847	if (MACHINE_HAS_IEEE)
848		pr_info("The hardware system has IEEE compatible "
849			"floating point units\n");
850	else
851		pr_info("The hardware system has no IEEE compatible "
852			"floating point units\n");
853#else /* CONFIG_64BIT */
854	if (MACHINE_IS_VM)
855		pr_info("Linux is running as a z/VM "
856			"guest operating system in 64-bit mode\n");
857	else if (MACHINE_IS_KVM)
858		pr_info("Linux is running under KVM in 64-bit mode\n");
859	else if (MACHINE_IS_LPAR)
860		pr_info("Linux is running natively in 64-bit mode\n");
861#endif /* CONFIG_64BIT */
862
863	/* Have one command line that is parsed and saved in /proc/cmdline */
864	/* boot_command_line has been already set up in early.c */
865	*cmdline_p = boot_command_line;
866
867        ROOT_DEV = Root_RAM0;
868
869	/* Is init_mm really needed? */
870	init_mm.start_code = PAGE_OFFSET;
871	init_mm.end_code = (unsigned long) &_etext;
872	init_mm.end_data = (unsigned long) &_edata;
873	init_mm.brk = (unsigned long) &_end;
874
875	parse_early_param();
876	os_info_init();
877	setup_ipl();
878
879	/* Do some memory reservations *before* memory is added to memblock */
880	reserve_memory_end();
881	reserve_oldmem();
882	reserve_kernel();
883	reserve_initrd();
884	reserve_elfcorehdr();
885	memblock_allow_resize();
886
887	/* Get information about *all* installed memory */
888	detect_memory_memblock();
889
890	remove_oldmem();
891
892	/*
893	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
894	 * extra checks that HOLES_IN_ZONE would require.
895	 *
896	 * Is this still required?
897	 */
898	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
899
900	setup_memory_end();
901	setup_memory();
902
903	check_initrd();
904	reserve_crashkernel();
905
906	setup_resources();
907	setup_vmcoreinfo();
908	setup_lowcore();
909	smp_fill_possible_mask();
910        cpu_init();
911	s390_init_cpu_topology();
912
913	/*
914	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
915	 */
916	setup_hwcaps();
917
918	/*
919	 * Create kernel page tables and switch to virtual addressing.
920	 */
921        paging_init();
922
923        /* Setup default console */
924	conmode_default();
925	set_preferred_console();
926
927	/* Setup zfcpdump support */
928	setup_zfcpdump();
929
930	/* Add system specific data to the random pool */
931	setup_randomness();
932}
933
934#ifdef CONFIG_32BIT
935static int no_removal_warning __initdata;
936
937static int __init parse_no_removal_warning(char *str)
938{
939	no_removal_warning = 1;
940	return 0;
941}
942__setup("no_removal_warning", parse_no_removal_warning);
943
944static int __init removal_warning(void)
945{
946	if (no_removal_warning)
947		return 0;
948	printk(KERN_ALERT "\n\n");
949	printk(KERN_CONT "Warning - you are using a 31 bit kernel!\n\n");
950	printk(KERN_CONT "We plan to remove 31 bit kernel support from the kernel sources in March 2015.\n");
951	printk(KERN_CONT "Currently we assume that nobody is using the 31 bit kernel on old 31 bit\n");
952	printk(KERN_CONT "hardware anymore. If you think that the code should not be removed and also\n");
953	printk(KERN_CONT "future versions of the Linux kernel should be able to run in 31 bit mode\n");
954	printk(KERN_CONT "please let us know. Please write to:\n");
955	printk(KERN_CONT "linux390@de.ibm.com (mail address) and/or\n");
956	printk(KERN_CONT "linux-s390@vger.kernel.org (mailing list).\n\n");
957	printk(KERN_CONT "Thank you!\n\n");
958	printk(KERN_CONT "If this kernel runs on a 64 bit machine you may consider using a 64 bit kernel.\n");
959	printk(KERN_CONT "This message can be disabled with the \"no_removal_warning\" kernel parameter.\n");
960	schedule_timeout_uninterruptible(300 * HZ);
961	return 0;
962}
963early_initcall(removal_warning);
964#endif
965