xref: /linux/arch/s390/kernel/setup.c (revision 3932b9ca55b0be314a36d3e84faff3e823c081f5)
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  */
71 unsigned int console_mode = 0;
72 EXPORT_SYMBOL(console_mode);
73 
74 unsigned int console_devno = -1;
75 EXPORT_SYMBOL(console_devno);
76 
77 unsigned int console_irq = -1;
78 EXPORT_SYMBOL(console_irq);
79 
80 unsigned long elf_hwcap = 0;
81 char elf_platform[ELF_PLATFORM_SIZE];
82 
83 int __initdata memory_end_set;
84 unsigned long __initdata memory_end;
85 unsigned long __initdata max_physmem_end;
86 
87 unsigned long VMALLOC_START;
88 EXPORT_SYMBOL(VMALLOC_START);
89 
90 unsigned long VMALLOC_END;
91 EXPORT_SYMBOL(VMALLOC_END);
92 
93 struct page *vmemmap;
94 EXPORT_SYMBOL(vmemmap);
95 
96 #ifdef CONFIG_64BIT
97 unsigned long MODULES_VADDR;
98 unsigned long MODULES_END;
99 #endif
100 
101 /* An array with a pointer to the lowcore of every CPU. */
102 struct _lowcore *lowcore_ptr[NR_CPUS];
103 EXPORT_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 
117 static 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 
131 static 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 
146 static 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 
166 static 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
217 static 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
227 static 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 
235 void 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 
246 void 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 
257 void 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  */
271 void (*pm_power_off)(void) = machine_power_off;
272 EXPORT_SYMBOL_GPL(pm_power_off);
273 
274 static 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 }
281 early_param("mem", early_parse_mem);
282 
283 static 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 }
290 early_param("vmalloc", parse_vmalloc);
291 
292 void *restart_stack __attribute__((__section__(".data")));
293 
294 static 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 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
347 #endif
348 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
349 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
350 	lc->exit_timer = S390_lowcore.exit_timer;
351 	lc->user_timer = S390_lowcore.user_timer;
352 	lc->system_timer = S390_lowcore.system_timer;
353 	lc->steal_timer = S390_lowcore.steal_timer;
354 	lc->last_update_timer = S390_lowcore.last_update_timer;
355 	lc->last_update_clock = S390_lowcore.last_update_clock;
356 	lc->ftrace_func = S390_lowcore.ftrace_func;
357 
358 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
359 	restart_stack += ASYNC_SIZE;
360 
361 	/*
362 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
363 	 * restart data to the absolute zero lowcore. This is necessary if
364 	 * PSW restart is done on an offline CPU that has lowcore zero.
365 	 */
366 	lc->restart_stack = (unsigned long) restart_stack;
367 	lc->restart_fn = (unsigned long) do_restart;
368 	lc->restart_data = 0;
369 	lc->restart_source = -1UL;
370 
371 	/* Setup absolute zero lowcore */
372 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
373 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
374 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
375 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
376 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
377 
378 #ifdef CONFIG_SMP
379 	lc->spinlock_lockval = arch_spin_lockval(0);
380 #endif
381 
382 	set_prefix((u32)(unsigned long) lc);
383 	lowcore_ptr[0] = lc;
384 }
385 
386 static struct resource code_resource = {
387 	.name  = "Kernel code",
388 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
389 };
390 
391 static struct resource data_resource = {
392 	.name = "Kernel data",
393 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
394 };
395 
396 static struct resource bss_resource = {
397 	.name = "Kernel bss",
398 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
399 };
400 
401 static struct resource __initdata *standard_resources[] = {
402 	&code_resource,
403 	&data_resource,
404 	&bss_resource,
405 };
406 
407 static void __init setup_resources(void)
408 {
409 	struct resource *res, *std_res, *sub_res;
410 	struct memblock_region *reg;
411 	int j;
412 
413 	code_resource.start = (unsigned long) &_text;
414 	code_resource.end = (unsigned long) &_etext - 1;
415 	data_resource.start = (unsigned long) &_etext;
416 	data_resource.end = (unsigned long) &_edata - 1;
417 	bss_resource.start = (unsigned long) &__bss_start;
418 	bss_resource.end = (unsigned long) &__bss_stop - 1;
419 
420 	for_each_memblock(memory, reg) {
421 		res = alloc_bootmem_low(sizeof(*res));
422 		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
423 
424 		res->name = "System RAM";
425 		res->start = reg->base;
426 		res->end = reg->base + reg->size - 1;
427 		request_resource(&iomem_resource, res);
428 
429 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
430 			std_res = standard_resources[j];
431 			if (std_res->start < res->start ||
432 			    std_res->start > res->end)
433 				continue;
434 			if (std_res->end > res->end) {
435 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
436 				*sub_res = *std_res;
437 				sub_res->end = res->end;
438 				std_res->start = res->end + 1;
439 				request_resource(res, sub_res);
440 			} else {
441 				request_resource(res, std_res);
442 			}
443 		}
444 	}
445 }
446 
447 static void __init setup_memory_end(void)
448 {
449 	unsigned long vmax, vmalloc_size, tmp;
450 
451 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
452 #ifdef CONFIG_64BIT
453 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
454 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
455 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
456 	if (tmp <= (1UL << 42))
457 		vmax = 1UL << 42;	/* 3-level kernel page table */
458 	else
459 		vmax = 1UL << 53;	/* 4-level kernel page table */
460 	/* module area is at the end of the kernel address space. */
461 	MODULES_END = vmax;
462 	MODULES_VADDR = MODULES_END - MODULES_LEN;
463 	VMALLOC_END = MODULES_VADDR;
464 #else
465 	vmalloc_size = VMALLOC_END ?: 96UL << 20;
466 	vmax = 1UL << 31;		/* 2-level kernel page table */
467 	/* vmalloc area is at the end of the kernel address space. */
468 	VMALLOC_END = vmax;
469 #endif
470 	VMALLOC_START = vmax - vmalloc_size;
471 
472 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
473 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
474 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
475 	tmp = SECTION_ALIGN_UP(tmp);
476 	tmp = VMALLOC_START - tmp * sizeof(struct page);
477 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
478 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
479 	vmemmap = (struct page *) tmp;
480 
481 	/* Take care that memory_end is set and <= vmemmap */
482 	memory_end = min(memory_end ?: max_physmem_end, tmp);
483 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
484 	memblock_remove(memory_end, ULONG_MAX);
485 
486 	pr_notice("Max memory size: %luMB\n", memory_end >> 20);
487 }
488 
489 static void __init setup_vmcoreinfo(void)
490 {
491 	mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
492 }
493 
494 #ifdef CONFIG_CRASH_DUMP
495 
496 /*
497  * When kdump is enabled, we have to ensure that no memory from
498  * the area [0 - crashkernel memory size] and
499  * [crashk_res.start - crashk_res.end] is set offline.
500  */
501 static int kdump_mem_notifier(struct notifier_block *nb,
502 			      unsigned long action, void *data)
503 {
504 	struct memory_notify *arg = data;
505 
506 	if (action != MEM_GOING_OFFLINE)
507 		return NOTIFY_OK;
508 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
509 		return NOTIFY_BAD;
510 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
511 		return NOTIFY_OK;
512 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
513 		return NOTIFY_OK;
514 	return NOTIFY_BAD;
515 }
516 
517 static struct notifier_block kdump_mem_nb = {
518 	.notifier_call = kdump_mem_notifier,
519 };
520 
521 #endif
522 
523 /*
524  * Make sure that the area behind memory_end is protected
525  */
526 static void reserve_memory_end(void)
527 {
528 #ifdef CONFIG_CRASH_DUMP
529 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
530 	    !OLDMEM_BASE && sclp_get_hsa_size()) {
531 		memory_end = sclp_get_hsa_size();
532 		memory_end &= PAGE_MASK;
533 		memory_end_set = 1;
534 	}
535 #endif
536 	if (!memory_end_set)
537 		return;
538 	memblock_reserve(memory_end, ULONG_MAX);
539 }
540 
541 /*
542  * Make sure that oldmem, where the dump is stored, is protected
543  */
544 static void reserve_oldmem(void)
545 {
546 #ifdef CONFIG_CRASH_DUMP
547 	if (OLDMEM_BASE)
548 		/* Forget all memory above the running kdump system */
549 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
550 #endif
551 }
552 
553 /*
554  * Make sure that oldmem, where the dump is stored, is protected
555  */
556 static void remove_oldmem(void)
557 {
558 #ifdef CONFIG_CRASH_DUMP
559 	if (OLDMEM_BASE)
560 		/* Forget all memory above the running kdump system */
561 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
562 #endif
563 }
564 
565 /*
566  * Reserve memory for kdump kernel to be loaded with kexec
567  */
568 static void __init reserve_crashkernel(void)
569 {
570 #ifdef CONFIG_CRASH_DUMP
571 	unsigned long long crash_base, crash_size;
572 	phys_addr_t low, high;
573 	int rc;
574 
575 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
576 			       &crash_base);
577 
578 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
579 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
580 	if (rc || crash_size == 0)
581 		return;
582 
583 	if (memblock.memory.regions[0].size < crash_size) {
584 		pr_info("crashkernel reservation failed: %s\n",
585 			"first memory chunk must be at least crashkernel size");
586 		return;
587 	}
588 
589 	low = crash_base ?: OLDMEM_BASE;
590 	high = low + crash_size;
591 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
592 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
593 		crash_base = low;
594 	} else {
595 		/* Find suitable area in free memory */
596 		low = max_t(unsigned long, crash_size, sclp_get_hsa_size());
597 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
598 
599 		if (crash_base && crash_base < low) {
600 			pr_info("crashkernel reservation failed: %s\n",
601 				"crash_base too low");
602 			return;
603 		}
604 		low = crash_base ?: low;
605 		crash_base = memblock_find_in_range(low, high, crash_size,
606 						    KEXEC_CRASH_MEM_ALIGN);
607 	}
608 
609 	if (!crash_base) {
610 		pr_info("crashkernel reservation failed: %s\n",
611 			"no suitable area found");
612 		return;
613 	}
614 
615 	if (register_memory_notifier(&kdump_mem_nb))
616 		return;
617 
618 	if (!OLDMEM_BASE && MACHINE_IS_VM)
619 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
620 	crashk_res.start = crash_base;
621 	crashk_res.end = crash_base + crash_size - 1;
622 	insert_resource(&iomem_resource, &crashk_res);
623 	memblock_remove(crash_base, crash_size);
624 	pr_info("Reserving %lluMB of memory at %lluMB "
625 		"for crashkernel (System RAM: %luMB)\n",
626 		crash_size >> 20, crash_base >> 20,
627 		(unsigned long)memblock.memory.total_size >> 20);
628 	os_info_crashkernel_add(crash_base, crash_size);
629 #endif
630 }
631 
632 /*
633  * Reserve the initrd from being used by memblock
634  */
635 static void __init reserve_initrd(void)
636 {
637 #ifdef CONFIG_BLK_DEV_INITRD
638 	initrd_start = INITRD_START;
639 	initrd_end = initrd_start + INITRD_SIZE;
640 	memblock_reserve(INITRD_START, INITRD_SIZE);
641 #endif
642 }
643 
644 /*
645  * Check for initrd being in usable memory
646  */
647 static void __init check_initrd(void)
648 {
649 #ifdef CONFIG_BLK_DEV_INITRD
650 	if (INITRD_START && INITRD_SIZE &&
651 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
652 		pr_err("initrd does not fit memory.\n");
653 		memblock_free(INITRD_START, INITRD_SIZE);
654 		initrd_start = initrd_end = 0;
655 	}
656 #endif
657 }
658 
659 /*
660  * Reserve all kernel text
661  */
662 static void __init reserve_kernel(void)
663 {
664 	unsigned long start_pfn;
665 	start_pfn = PFN_UP(__pa(&_end));
666 
667 	/*
668 	 * Reserve memory used for lowcore/command line/kernel image.
669 	 */
670 	memblock_reserve(0, (unsigned long)_ehead);
671 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
672 			 - (unsigned long)_stext);
673 }
674 
675 static void __init reserve_elfcorehdr(void)
676 {
677 #ifdef CONFIG_CRASH_DUMP
678 	if (is_kdump_kernel())
679 		memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
680 				 PAGE_ALIGN(elfcorehdr_size));
681 #endif
682 }
683 
684 static void __init setup_memory(void)
685 {
686 	struct memblock_region *reg;
687 
688 	/*
689 	 * Init storage key for present memory
690 	 */
691 	for_each_memblock(memory, reg) {
692 		storage_key_init_range(reg->base, reg->base + reg->size);
693 	}
694 	psw_set_key(PAGE_DEFAULT_KEY);
695 
696 	/* Only cosmetics */
697 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
698 }
699 
700 /*
701  * Setup hardware capabilities.
702  */
703 static void __init setup_hwcaps(void)
704 {
705 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
706 	struct cpuid cpu_id;
707 	int i;
708 
709 	/*
710 	 * The store facility list bits numbers as found in the principles
711 	 * of operation are numbered with bit 1UL<<31 as number 0 to
712 	 * bit 1UL<<0 as number 31.
713 	 *   Bit 0: instructions named N3, "backported" to esa-mode
714 	 *   Bit 2: z/Architecture mode is active
715 	 *   Bit 7: the store-facility-list-extended facility is installed
716 	 *   Bit 17: the message-security assist is installed
717 	 *   Bit 19: the long-displacement facility is installed
718 	 *   Bit 21: the extended-immediate facility is installed
719 	 *   Bit 22: extended-translation facility 3 is installed
720 	 *   Bit 30: extended-translation facility 3 enhancement facility
721 	 * These get translated to:
722 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
723 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
724 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
725 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
726 	 */
727 	for (i = 0; i < 6; i++)
728 		if (test_facility(stfl_bits[i]))
729 			elf_hwcap |= 1UL << i;
730 
731 	if (test_facility(22) && test_facility(30))
732 		elf_hwcap |= HWCAP_S390_ETF3EH;
733 
734 	/*
735 	 * Check for additional facilities with store-facility-list-extended.
736 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
737 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
738 	 * as stored by stfl, bits 32-xxx contain additional facilities.
739 	 * How many facility words are stored depends on the number of
740 	 * doublewords passed to the instruction. The additional facilities
741 	 * are:
742 	 *   Bit 42: decimal floating point facility is installed
743 	 *   Bit 44: perform floating point operation facility is installed
744 	 * translated to:
745 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
746 	 */
747 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
748 		elf_hwcap |= HWCAP_S390_DFP;
749 
750 	/*
751 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
752 	 */
753 	if (MACHINE_HAS_HPAGE)
754 		elf_hwcap |= HWCAP_S390_HPAGE;
755 
756 #if defined(CONFIG_64BIT)
757 	/*
758 	 * 64-bit register support for 31-bit processes
759 	 * HWCAP_S390_HIGH_GPRS is bit 9.
760 	 */
761 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
762 
763 	/*
764 	 * Transactional execution support HWCAP_S390_TE is bit 10.
765 	 */
766 	if (test_facility(50) && test_facility(73))
767 		elf_hwcap |= HWCAP_S390_TE;
768 #endif
769 
770 	get_cpu_id(&cpu_id);
771 	add_device_randomness(&cpu_id, sizeof(cpu_id));
772 	switch (cpu_id.machine) {
773 	case 0x9672:
774 #if !defined(CONFIG_64BIT)
775 	default:	/* Use "g5" as default for 31 bit kernels. */
776 #endif
777 		strcpy(elf_platform, "g5");
778 		break;
779 	case 0x2064:
780 	case 0x2066:
781 #if defined(CONFIG_64BIT)
782 	default:	/* Use "z900" as default for 64 bit kernels. */
783 #endif
784 		strcpy(elf_platform, "z900");
785 		break;
786 	case 0x2084:
787 	case 0x2086:
788 		strcpy(elf_platform, "z990");
789 		break;
790 	case 0x2094:
791 	case 0x2096:
792 		strcpy(elf_platform, "z9-109");
793 		break;
794 	case 0x2097:
795 	case 0x2098:
796 		strcpy(elf_platform, "z10");
797 		break;
798 	case 0x2817:
799 	case 0x2818:
800 		strcpy(elf_platform, "z196");
801 		break;
802 	case 0x2827:
803 	case 0x2828:
804 		strcpy(elf_platform, "zEC12");
805 		break;
806 	}
807 }
808 
809 /*
810  * Add system information as device randomness
811  */
812 static void __init setup_randomness(void)
813 {
814 	struct sysinfo_3_2_2 *vmms;
815 
816 	vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
817 	if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
818 		add_device_randomness(&vmms, vmms->count);
819 	free_page((unsigned long) vmms);
820 }
821 
822 /*
823  * Setup function called from init/main.c just after the banner
824  * was printed.
825  */
826 
827 void __init setup_arch(char **cmdline_p)
828 {
829         /*
830          * print what head.S has found out about the machine
831          */
832 #ifndef CONFIG_64BIT
833 	if (MACHINE_IS_VM)
834 		pr_info("Linux is running as a z/VM "
835 			"guest operating system in 31-bit mode\n");
836 	else if (MACHINE_IS_LPAR)
837 		pr_info("Linux is running natively in 31-bit mode\n");
838 	if (MACHINE_HAS_IEEE)
839 		pr_info("The hardware system has IEEE compatible "
840 			"floating point units\n");
841 	else
842 		pr_info("The hardware system has no IEEE compatible "
843 			"floating point units\n");
844 #else /* CONFIG_64BIT */
845 	if (MACHINE_IS_VM)
846 		pr_info("Linux is running as a z/VM "
847 			"guest operating system in 64-bit mode\n");
848 	else if (MACHINE_IS_KVM)
849 		pr_info("Linux is running under KVM in 64-bit mode\n");
850 	else if (MACHINE_IS_LPAR)
851 		pr_info("Linux is running natively in 64-bit mode\n");
852 #endif /* CONFIG_64BIT */
853 
854 	/* Have one command line that is parsed and saved in /proc/cmdline */
855 	/* boot_command_line has been already set up in early.c */
856 	*cmdline_p = boot_command_line;
857 
858         ROOT_DEV = Root_RAM0;
859 
860 	/* Is init_mm really needed? */
861 	init_mm.start_code = PAGE_OFFSET;
862 	init_mm.end_code = (unsigned long) &_etext;
863 	init_mm.end_data = (unsigned long) &_edata;
864 	init_mm.brk = (unsigned long) &_end;
865 
866 	parse_early_param();
867 	os_info_init();
868 	setup_ipl();
869 
870 	/* Do some memory reservations *before* memory is added to memblock */
871 	reserve_memory_end();
872 	reserve_oldmem();
873 	reserve_kernel();
874 	reserve_initrd();
875 	reserve_elfcorehdr();
876 	memblock_allow_resize();
877 
878 	/* Get information about *all* installed memory */
879 	detect_memory_memblock();
880 
881 	remove_oldmem();
882 
883 	/*
884 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
885 	 * extra checks that HOLES_IN_ZONE would require.
886 	 *
887 	 * Is this still required?
888 	 */
889 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
890 
891 	setup_memory_end();
892 	setup_memory();
893 
894 	check_initrd();
895 	reserve_crashkernel();
896 
897 	setup_resources();
898 	setup_vmcoreinfo();
899 	setup_lowcore();
900 	smp_fill_possible_mask();
901         cpu_init();
902 	s390_init_cpu_topology();
903 
904 	/*
905 	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
906 	 */
907 	setup_hwcaps();
908 
909 	/*
910 	 * Create kernel page tables and switch to virtual addressing.
911 	 */
912         paging_init();
913 
914         /* Setup default console */
915 	conmode_default();
916 	set_preferred_console();
917 
918 	/* Setup zfcpdump support */
919 	setup_zfcpdump();
920 
921 	/* Add system specific data to the random pool */
922 	setup_randomness();
923 }
924 
925 #ifdef CONFIG_32BIT
926 static int no_removal_warning __initdata;
927 
928 static int __init parse_no_removal_warning(char *str)
929 {
930 	no_removal_warning = 1;
931 	return 0;
932 }
933 __setup("no_removal_warning", parse_no_removal_warning);
934 
935 static int __init removal_warning(void)
936 {
937 	if (no_removal_warning)
938 		return 0;
939 	printk(KERN_ALERT "\n\n");
940 	printk(KERN_CONT "Warning - you are using a 31 bit kernel!\n\n");
941 	printk(KERN_CONT "We plan to remove 31 bit kernel support from the kernel sources in March 2015.\n");
942 	printk(KERN_CONT "Currently we assume that nobody is using the 31 bit kernel on old 31 bit\n");
943 	printk(KERN_CONT "hardware anymore. If you think that the code should not be removed and also\n");
944 	printk(KERN_CONT "future versions of the Linux kernel should be able to run in 31 bit mode\n");
945 	printk(KERN_CONT "please let us know. Please write to:\n");
946 	printk(KERN_CONT "linux390@de.ibm.com (mail address) and/or\n");
947 	printk(KERN_CONT "linux-s390@vger.kernel.org (mailing list).\n\n");
948 	printk(KERN_CONT "Thank you!\n\n");
949 	printk(KERN_CONT "If this kernel runs on a 64 bit machine you may consider using a 64 bit kernel.\n");
950 	printk(KERN_CONT "This message can be disabled with the \"no_removal_warning\" kernel parameter.\n");
951 	schedule_timeout_uninterruptible(300 * HZ);
952 	return 0;
953 }
954 early_initcall(removal_warning);
955 #endif
956