xref: /openbmc/linux/arch/s390/kernel/setup.c (revision a8a28aff)
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/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>
48 
49 #include <asm/ipl.h>
50 #include <asm/facility.h>
51 #include <asm/smp.h>
52 #include <asm/mmu_context.h>
53 #include <asm/cpcmd.h>
54 #include <asm/lowcore.h>
55 #include <asm/irq.h>
56 #include <asm/page.h>
57 #include <asm/ptrace.h>
58 #include <asm/sections.h>
59 #include <asm/ebcdic.h>
60 #include <asm/kvm_virtio.h>
61 #include <asm/diag.h>
62 #include <asm/os_info.h>
63 #include <asm/sclp.h>
64 #include "entry.h"
65 
66 /*
67  * Machine setup..
68  */
69 unsigned int console_mode = 0;
70 EXPORT_SYMBOL(console_mode);
71 
72 unsigned int console_devno = -1;
73 EXPORT_SYMBOL(console_devno);
74 
75 unsigned int console_irq = -1;
76 EXPORT_SYMBOL(console_irq);
77 
78 unsigned long elf_hwcap = 0;
79 char elf_platform[ELF_PLATFORM_SIZE];
80 
81 int __initdata memory_end_set;
82 unsigned long __initdata memory_end;
83 unsigned long __initdata max_physmem_end;
84 
85 unsigned long VMALLOC_START;
86 EXPORT_SYMBOL(VMALLOC_START);
87 
88 unsigned long VMALLOC_END;
89 EXPORT_SYMBOL(VMALLOC_END);
90 
91 struct page *vmemmap;
92 EXPORT_SYMBOL(vmemmap);
93 
94 #ifdef CONFIG_64BIT
95 unsigned long MODULES_VADDR;
96 unsigned long MODULES_END;
97 #endif
98 
99 /* An array with a pointer to the lowcore of every CPU. */
100 struct _lowcore *lowcore_ptr[NR_CPUS];
101 EXPORT_SYMBOL(lowcore_ptr);
102 
103 /*
104  * This is set up by the setup-routine at boot-time
105  * for S390 need to find out, what we have to setup
106  * using address 0x10400 ...
107  */
108 
109 #include <asm/setup.h>
110 
111 /*
112  * condev= and conmode= setup parameter.
113  */
114 
115 static int __init condev_setup(char *str)
116 {
117 	int vdev;
118 
119 	vdev = simple_strtoul(str, &str, 0);
120 	if (vdev >= 0 && vdev < 65536) {
121 		console_devno = vdev;
122 		console_irq = -1;
123 	}
124 	return 1;
125 }
126 
127 __setup("condev=", condev_setup);
128 
129 static void __init set_preferred_console(void)
130 {
131 	if (MACHINE_IS_KVM) {
132 		if (sclp_has_vt220())
133 			add_preferred_console("ttyS", 1, NULL);
134 		else if (sclp_has_linemode())
135 			add_preferred_console("ttyS", 0, NULL);
136 		else
137 			add_preferred_console("hvc", 0, NULL);
138 	} else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
139 		add_preferred_console("ttyS", 0, NULL);
140 	else if (CONSOLE_IS_3270)
141 		add_preferred_console("tty3270", 0, NULL);
142 }
143 
144 static int __init conmode_setup(char *str)
145 {
146 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
147 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
148                 SET_CONSOLE_SCLP;
149 #endif
150 #if defined(CONFIG_TN3215_CONSOLE)
151 	if (strncmp(str, "3215", 5) == 0)
152 		SET_CONSOLE_3215;
153 #endif
154 #if defined(CONFIG_TN3270_CONSOLE)
155 	if (strncmp(str, "3270", 5) == 0)
156 		SET_CONSOLE_3270;
157 #endif
158 	set_preferred_console();
159         return 1;
160 }
161 
162 __setup("conmode=", conmode_setup);
163 
164 static void __init conmode_default(void)
165 {
166 	char query_buffer[1024];
167 	char *ptr;
168 
169         if (MACHINE_IS_VM) {
170 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
171 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
172 		ptr = strstr(query_buffer, "SUBCHANNEL =");
173 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
174 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
175 		ptr = strstr(query_buffer, "CONMODE");
176 		/*
177 		 * Set the conmode to 3215 so that the device recognition
178 		 * will set the cu_type of the console to 3215. If the
179 		 * conmode is 3270 and we don't set it back then both
180 		 * 3215 and the 3270 driver will try to access the console
181 		 * device (3215 as console and 3270 as normal tty).
182 		 */
183 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
184 		if (ptr == NULL) {
185 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
186 			SET_CONSOLE_SCLP;
187 #endif
188 			return;
189 		}
190 		if (strncmp(ptr + 8, "3270", 4) == 0) {
191 #if defined(CONFIG_TN3270_CONSOLE)
192 			SET_CONSOLE_3270;
193 #elif defined(CONFIG_TN3215_CONSOLE)
194 			SET_CONSOLE_3215;
195 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
196 			SET_CONSOLE_SCLP;
197 #endif
198 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
199 #if defined(CONFIG_TN3215_CONSOLE)
200 			SET_CONSOLE_3215;
201 #elif defined(CONFIG_TN3270_CONSOLE)
202 			SET_CONSOLE_3270;
203 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 			SET_CONSOLE_SCLP;
205 #endif
206 		}
207 	} else {
208 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
209 		SET_CONSOLE_SCLP;
210 #endif
211 	}
212 }
213 
214 #ifdef CONFIG_CRASH_DUMP
215 static void __init setup_zfcpdump(void)
216 {
217 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
218 		return;
219 	if (OLDMEM_BASE)
220 		return;
221 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
222 	console_loglevel = 2;
223 }
224 #else
225 static inline void setup_zfcpdump(void) {}
226 #endif /* CONFIG_CRASH_DUMP */
227 
228  /*
229  * Reboot, halt and power_off stubs. They just call _machine_restart,
230  * _machine_halt or _machine_power_off.
231  */
232 
233 void machine_restart(char *command)
234 {
235 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
236 		/*
237 		 * Only unblank the console if we are called in enabled
238 		 * context or a bust_spinlocks cleared the way for us.
239 		 */
240 		console_unblank();
241 	_machine_restart(command);
242 }
243 
244 void machine_halt(void)
245 {
246 	if (!in_interrupt() || oops_in_progress)
247 		/*
248 		 * Only unblank the console if we are called in enabled
249 		 * context or a bust_spinlocks cleared the way for us.
250 		 */
251 		console_unblank();
252 	_machine_halt();
253 }
254 
255 void machine_power_off(void)
256 {
257 	if (!in_interrupt() || oops_in_progress)
258 		/*
259 		 * Only unblank the console if we are called in enabled
260 		 * context or a bust_spinlocks cleared the way for us.
261 		 */
262 		console_unblank();
263 	_machine_power_off();
264 }
265 
266 /*
267  * Dummy power off function.
268  */
269 void (*pm_power_off)(void) = machine_power_off;
270 EXPORT_SYMBOL_GPL(pm_power_off);
271 
272 static int __init early_parse_mem(char *p)
273 {
274 	memory_end = memparse(p, &p);
275 	memory_end &= PAGE_MASK;
276 	memory_end_set = 1;
277 	return 0;
278 }
279 early_param("mem", early_parse_mem);
280 
281 static int __init parse_vmalloc(char *arg)
282 {
283 	if (!arg)
284 		return -EINVAL;
285 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
286 	return 0;
287 }
288 early_param("vmalloc", parse_vmalloc);
289 
290 void *restart_stack __attribute__((__section__(".data")));
291 
292 static void __init setup_lowcore(void)
293 {
294 	struct _lowcore *lc;
295 
296 	/*
297 	 * Setup lowcore for boot cpu
298 	 */
299 	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
300 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
301 	lc->restart_psw.mask = PSW_KERNEL_BITS;
302 	lc->restart_psw.addr =
303 		PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
304 	lc->external_new_psw.mask = PSW_KERNEL_BITS |
305 		PSW_MASK_DAT | PSW_MASK_MCHECK;
306 	lc->external_new_psw.addr =
307 		PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
308 	lc->svc_new_psw.mask = PSW_KERNEL_BITS |
309 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
310 	lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
311 	lc->program_new_psw.mask = PSW_KERNEL_BITS |
312 		PSW_MASK_DAT | PSW_MASK_MCHECK;
313 	lc->program_new_psw.addr =
314 		PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
315 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
316 	lc->mcck_new_psw.addr =
317 		PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
318 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
319 		PSW_MASK_DAT | PSW_MASK_MCHECK;
320 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
321 	lc->clock_comparator = -1ULL;
322 	lc->kernel_stack = ((unsigned long) &init_thread_union)
323 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
324 	lc->async_stack = (unsigned long)
325 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
326 		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
327 	lc->panic_stack = (unsigned long)
328 		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
329 		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
330 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
331 	lc->thread_info = (unsigned long) &init_thread_union;
332 	lc->machine_flags = S390_lowcore.machine_flags;
333 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
334 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
335 	       MAX_FACILITY_BIT/8);
336 #ifndef CONFIG_64BIT
337 	if (MACHINE_HAS_IEEE) {
338 		lc->extended_save_area_addr = (__u32)
339 			__alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
340 		/* enable extended save area */
341 		__ctl_set_bit(14, 29);
342 	}
343 #else
344 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
345 #endif
346 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
347 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
348 	lc->exit_timer = S390_lowcore.exit_timer;
349 	lc->user_timer = S390_lowcore.user_timer;
350 	lc->system_timer = S390_lowcore.system_timer;
351 	lc->steal_timer = S390_lowcore.steal_timer;
352 	lc->last_update_timer = S390_lowcore.last_update_timer;
353 	lc->last_update_clock = S390_lowcore.last_update_clock;
354 	lc->ftrace_func = S390_lowcore.ftrace_func;
355 
356 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
357 	restart_stack += ASYNC_SIZE;
358 
359 	/*
360 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
361 	 * restart data to the absolute zero lowcore. This is necessary if
362 	 * PSW restart is done on an offline CPU that has lowcore zero.
363 	 */
364 	lc->restart_stack = (unsigned long) restart_stack;
365 	lc->restart_fn = (unsigned long) do_restart;
366 	lc->restart_data = 0;
367 	lc->restart_source = -1UL;
368 
369 	/* Setup absolute zero lowcore */
370 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
371 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
372 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
373 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
374 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
375 
376 #ifdef CONFIG_SMP
377 	lc->spinlock_lockval = arch_spin_lockval(0);
378 #endif
379 
380 	set_prefix((u32)(unsigned long) lc);
381 	lowcore_ptr[0] = lc;
382 }
383 
384 static struct resource code_resource = {
385 	.name  = "Kernel code",
386 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
387 };
388 
389 static struct resource data_resource = {
390 	.name = "Kernel data",
391 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
392 };
393 
394 static struct resource bss_resource = {
395 	.name = "Kernel bss",
396 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
397 };
398 
399 static struct resource __initdata *standard_resources[] = {
400 	&code_resource,
401 	&data_resource,
402 	&bss_resource,
403 };
404 
405 static void __init setup_resources(void)
406 {
407 	struct resource *res, *std_res, *sub_res;
408 	struct memblock_region *reg;
409 	int j;
410 
411 	code_resource.start = (unsigned long) &_text;
412 	code_resource.end = (unsigned long) &_etext - 1;
413 	data_resource.start = (unsigned long) &_etext;
414 	data_resource.end = (unsigned long) &_edata - 1;
415 	bss_resource.start = (unsigned long) &__bss_start;
416 	bss_resource.end = (unsigned long) &__bss_stop - 1;
417 
418 	for_each_memblock(memory, reg) {
419 		res = alloc_bootmem_low(sizeof(*res));
420 		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
421 
422 		res->name = "System RAM";
423 		res->start = reg->base;
424 		res->end = reg->base + reg->size - 1;
425 		request_resource(&iomem_resource, res);
426 
427 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
428 			std_res = standard_resources[j];
429 			if (std_res->start < res->start ||
430 			    std_res->start > res->end)
431 				continue;
432 			if (std_res->end > res->end) {
433 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
434 				*sub_res = *std_res;
435 				sub_res->end = res->end;
436 				std_res->start = res->end + 1;
437 				request_resource(res, sub_res);
438 			} else {
439 				request_resource(res, std_res);
440 			}
441 		}
442 	}
443 }
444 
445 static void __init setup_memory_end(void)
446 {
447 	unsigned long vmax, vmalloc_size, tmp;
448 
449 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
450 #ifdef CONFIG_64BIT
451 	vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
452 	tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
453 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
454 	if (tmp <= (1UL << 42))
455 		vmax = 1UL << 42;	/* 3-level kernel page table */
456 	else
457 		vmax = 1UL << 53;	/* 4-level kernel page table */
458 	/* module area is at the end of the kernel address space. */
459 	MODULES_END = vmax;
460 	MODULES_VADDR = MODULES_END - MODULES_LEN;
461 	VMALLOC_END = MODULES_VADDR;
462 #else
463 	vmalloc_size = VMALLOC_END ?: 96UL << 20;
464 	vmax = 1UL << 31;		/* 2-level kernel page table */
465 	/* vmalloc area is at the end of the kernel address space. */
466 	VMALLOC_END = vmax;
467 #endif
468 	VMALLOC_START = vmax - vmalloc_size;
469 
470 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
471 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
472 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
473 	tmp = SECTION_ALIGN_UP(tmp);
474 	tmp = VMALLOC_START - tmp * sizeof(struct page);
475 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
476 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
477 	vmemmap = (struct page *) tmp;
478 
479 	/* Take care that memory_end is set and <= vmemmap */
480 	memory_end = min(memory_end ?: max_physmem_end, tmp);
481 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
482 	memblock_remove(memory_end, ULONG_MAX);
483 
484 	pr_notice("Max memory size: %luMB\n", memory_end >> 20);
485 }
486 
487 static void __init setup_vmcoreinfo(void)
488 {
489 	mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
490 }
491 
492 #ifdef CONFIG_CRASH_DUMP
493 
494 /*
495  * When kdump is enabled, we have to ensure that no memory from
496  * the area [0 - crashkernel memory size] and
497  * [crashk_res.start - crashk_res.end] is set offline.
498  */
499 static int kdump_mem_notifier(struct notifier_block *nb,
500 			      unsigned long action, void *data)
501 {
502 	struct memory_notify *arg = data;
503 
504 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
505 		return NOTIFY_BAD;
506 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
507 		return NOTIFY_OK;
508 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
509 		return NOTIFY_OK;
510 	return NOTIFY_BAD;
511 }
512 
513 static struct notifier_block kdump_mem_nb = {
514 	.notifier_call = kdump_mem_notifier,
515 };
516 
517 #endif
518 
519 /*
520  * Make sure that the area behind memory_end is protected
521  */
522 static void reserve_memory_end(void)
523 {
524 #ifdef CONFIG_CRASH_DUMP
525 	if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
526 	    !OLDMEM_BASE && sclp_get_hsa_size()) {
527 		memory_end = sclp_get_hsa_size();
528 		memory_end &= PAGE_MASK;
529 		memory_end_set = 1;
530 	}
531 #endif
532 	if (!memory_end_set)
533 		return;
534 	memblock_reserve(memory_end, ULONG_MAX);
535 }
536 
537 /*
538  * Make sure that oldmem, where the dump is stored, is protected
539  */
540 static void reserve_oldmem(void)
541 {
542 #ifdef CONFIG_CRASH_DUMP
543 	if (OLDMEM_BASE)
544 		/* Forget all memory above the running kdump system */
545 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
546 #endif
547 }
548 
549 /*
550  * Make sure that oldmem, where the dump is stored, is protected
551  */
552 static void remove_oldmem(void)
553 {
554 #ifdef CONFIG_CRASH_DUMP
555 	if (OLDMEM_BASE)
556 		/* Forget all memory above the running kdump system */
557 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
558 #endif
559 }
560 
561 /*
562  * Reserve memory for kdump kernel to be loaded with kexec
563  */
564 static void __init reserve_crashkernel(void)
565 {
566 #ifdef CONFIG_CRASH_DUMP
567 	unsigned long long crash_base, crash_size;
568 	phys_addr_t low, high;
569 	int rc;
570 
571 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
572 			       &crash_base);
573 
574 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
575 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
576 	if (rc || crash_size == 0)
577 		return;
578 
579 	if (memblock.memory.regions[0].size < crash_size) {
580 		pr_info("crashkernel reservation failed: %s\n",
581 			"first memory chunk must be at least crashkernel size");
582 		return;
583 	}
584 
585 	low = crash_base ?: OLDMEM_BASE;
586 	high = low + crash_size;
587 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
588 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
589 		crash_base = low;
590 	} else {
591 		/* Find suitable area in free memory */
592 		low = max_t(unsigned long, crash_size, sclp_get_hsa_size());
593 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
594 
595 		if (crash_base && crash_base < low) {
596 			pr_info("crashkernel reservation failed: %s\n",
597 				"crash_base too low");
598 			return;
599 		}
600 		low = crash_base ?: low;
601 		crash_base = memblock_find_in_range(low, high, crash_size,
602 						    KEXEC_CRASH_MEM_ALIGN);
603 	}
604 
605 	if (!crash_base) {
606 		pr_info("crashkernel reservation failed: %s\n",
607 			"no suitable area found");
608 		return;
609 	}
610 
611 	if (register_memory_notifier(&kdump_mem_nb))
612 		return;
613 
614 	if (!OLDMEM_BASE && MACHINE_IS_VM)
615 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
616 	crashk_res.start = crash_base;
617 	crashk_res.end = crash_base + crash_size - 1;
618 	insert_resource(&iomem_resource, &crashk_res);
619 	memblock_remove(crash_base, crash_size);
620 	pr_info("Reserving %lluMB of memory at %lluMB "
621 		"for crashkernel (System RAM: %luMB)\n",
622 		crash_size >> 20, crash_base >> 20,
623 		(unsigned long)memblock.memory.total_size >> 20);
624 	os_info_crashkernel_add(crash_base, crash_size);
625 #endif
626 }
627 
628 /*
629  * Reserve the initrd from being used by memblock
630  */
631 static void __init reserve_initrd(void)
632 {
633 #ifdef CONFIG_BLK_DEV_INITRD
634 	initrd_start = INITRD_START;
635 	initrd_end = initrd_start + INITRD_SIZE;
636 	memblock_reserve(INITRD_START, INITRD_SIZE);
637 #endif
638 }
639 
640 /*
641  * Check for initrd being in usable memory
642  */
643 static void __init check_initrd(void)
644 {
645 #ifdef CONFIG_BLK_DEV_INITRD
646 	if (INITRD_START && INITRD_SIZE &&
647 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
648 		pr_err("initrd does not fit memory.\n");
649 		memblock_free(INITRD_START, INITRD_SIZE);
650 		initrd_start = initrd_end = 0;
651 	}
652 #endif
653 }
654 
655 /*
656  * Reserve all kernel text
657  */
658 static void __init reserve_kernel(void)
659 {
660 	unsigned long start_pfn;
661 	start_pfn = PFN_UP(__pa(&_end));
662 
663 	/*
664 	 * Reserve memory used for lowcore/command line/kernel image.
665 	 */
666 	memblock_reserve(0, (unsigned long)_ehead);
667 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
668 			 - (unsigned long)_stext);
669 }
670 
671 static void __init reserve_elfcorehdr(void)
672 {
673 #ifdef CONFIG_CRASH_DUMP
674 	if (is_kdump_kernel())
675 		memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
676 				 PAGE_ALIGN(elfcorehdr_size));
677 #endif
678 }
679 
680 static void __init setup_memory(void)
681 {
682 	struct memblock_region *reg;
683 
684 	/*
685 	 * Init storage key for present memory
686 	 */
687 	for_each_memblock(memory, reg) {
688 		storage_key_init_range(reg->base, reg->base + reg->size);
689 	}
690 	psw_set_key(PAGE_DEFAULT_KEY);
691 
692 	/* Only cosmetics */
693 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
694 }
695 
696 /*
697  * Setup hardware capabilities.
698  */
699 static void __init setup_hwcaps(void)
700 {
701 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
702 	struct cpuid cpu_id;
703 	int i;
704 
705 	/*
706 	 * The store facility list bits numbers as found in the principles
707 	 * of operation are numbered with bit 1UL<<31 as number 0 to
708 	 * bit 1UL<<0 as number 31.
709 	 *   Bit 0: instructions named N3, "backported" to esa-mode
710 	 *   Bit 2: z/Architecture mode is active
711 	 *   Bit 7: the store-facility-list-extended facility is installed
712 	 *   Bit 17: the message-security assist is installed
713 	 *   Bit 19: the long-displacement facility is installed
714 	 *   Bit 21: the extended-immediate facility is installed
715 	 *   Bit 22: extended-translation facility 3 is installed
716 	 *   Bit 30: extended-translation facility 3 enhancement facility
717 	 * These get translated to:
718 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
719 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
720 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
721 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
722 	 */
723 	for (i = 0; i < 6; i++)
724 		if (test_facility(stfl_bits[i]))
725 			elf_hwcap |= 1UL << i;
726 
727 	if (test_facility(22) && test_facility(30))
728 		elf_hwcap |= HWCAP_S390_ETF3EH;
729 
730 	/*
731 	 * Check for additional facilities with store-facility-list-extended.
732 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
733 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
734 	 * as stored by stfl, bits 32-xxx contain additional facilities.
735 	 * How many facility words are stored depends on the number of
736 	 * doublewords passed to the instruction. The additional facilities
737 	 * are:
738 	 *   Bit 42: decimal floating point facility is installed
739 	 *   Bit 44: perform floating point operation facility is installed
740 	 * translated to:
741 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
742 	 */
743 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
744 		elf_hwcap |= HWCAP_S390_DFP;
745 
746 	/*
747 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
748 	 */
749 	if (MACHINE_HAS_HPAGE)
750 		elf_hwcap |= HWCAP_S390_HPAGE;
751 
752 #if defined(CONFIG_64BIT)
753 	/*
754 	 * 64-bit register support for 31-bit processes
755 	 * HWCAP_S390_HIGH_GPRS is bit 9.
756 	 */
757 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
758 
759 	/*
760 	 * Transactional execution support HWCAP_S390_TE is bit 10.
761 	 */
762 	if (test_facility(50) && test_facility(73))
763 		elf_hwcap |= HWCAP_S390_TE;
764 #endif
765 
766 	get_cpu_id(&cpu_id);
767 	switch (cpu_id.machine) {
768 	case 0x9672:
769 #if !defined(CONFIG_64BIT)
770 	default:	/* Use "g5" as default for 31 bit kernels. */
771 #endif
772 		strcpy(elf_platform, "g5");
773 		break;
774 	case 0x2064:
775 	case 0x2066:
776 #if defined(CONFIG_64BIT)
777 	default:	/* Use "z900" as default for 64 bit kernels. */
778 #endif
779 		strcpy(elf_platform, "z900");
780 		break;
781 	case 0x2084:
782 	case 0x2086:
783 		strcpy(elf_platform, "z990");
784 		break;
785 	case 0x2094:
786 	case 0x2096:
787 		strcpy(elf_platform, "z9-109");
788 		break;
789 	case 0x2097:
790 	case 0x2098:
791 		strcpy(elf_platform, "z10");
792 		break;
793 	case 0x2817:
794 	case 0x2818:
795 		strcpy(elf_platform, "z196");
796 		break;
797 	case 0x2827:
798 	case 0x2828:
799 		strcpy(elf_platform, "zEC12");
800 		break;
801 	}
802 }
803 
804 /*
805  * Setup function called from init/main.c just after the banner
806  * was printed.
807  */
808 
809 void __init setup_arch(char **cmdline_p)
810 {
811         /*
812          * print what head.S has found out about the machine
813          */
814 #ifndef CONFIG_64BIT
815 	if (MACHINE_IS_VM)
816 		pr_info("Linux is running as a z/VM "
817 			"guest operating system in 31-bit mode\n");
818 	else if (MACHINE_IS_LPAR)
819 		pr_info("Linux is running natively in 31-bit mode\n");
820 	if (MACHINE_HAS_IEEE)
821 		pr_info("The hardware system has IEEE compatible "
822 			"floating point units\n");
823 	else
824 		pr_info("The hardware system has no IEEE compatible "
825 			"floating point units\n");
826 #else /* CONFIG_64BIT */
827 	if (MACHINE_IS_VM)
828 		pr_info("Linux is running as a z/VM "
829 			"guest operating system in 64-bit mode\n");
830 	else if (MACHINE_IS_KVM)
831 		pr_info("Linux is running under KVM in 64-bit mode\n");
832 	else if (MACHINE_IS_LPAR)
833 		pr_info("Linux is running natively in 64-bit mode\n");
834 #endif /* CONFIG_64BIT */
835 
836 	/* Have one command line that is parsed and saved in /proc/cmdline */
837 	/* boot_command_line has been already set up in early.c */
838 	*cmdline_p = boot_command_line;
839 
840         ROOT_DEV = Root_RAM0;
841 
842 	/* Is init_mm really needed? */
843 	init_mm.start_code = PAGE_OFFSET;
844 	init_mm.end_code = (unsigned long) &_etext;
845 	init_mm.end_data = (unsigned long) &_edata;
846 	init_mm.brk = (unsigned long) &_end;
847 
848 	parse_early_param();
849 	os_info_init();
850 	setup_ipl();
851 
852 	/* Do some memory reservations *before* memory is added to memblock */
853 	reserve_memory_end();
854 	reserve_oldmem();
855 	reserve_kernel();
856 	reserve_initrd();
857 	reserve_elfcorehdr();
858 	memblock_allow_resize();
859 
860 	/* Get information about *all* installed memory */
861 	detect_memory_memblock();
862 
863 	remove_oldmem();
864 
865 	/*
866 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
867 	 * extra checks that HOLES_IN_ZONE would require.
868 	 *
869 	 * Is this still required?
870 	 */
871 	memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
872 
873 	setup_memory_end();
874 	setup_memory();
875 
876 	check_initrd();
877 	reserve_crashkernel();
878 
879 	setup_resources();
880 	setup_vmcoreinfo();
881 	setup_lowcore();
882 	smp_fill_possible_mask();
883         cpu_init();
884 	s390_init_cpu_topology();
885 
886 	/*
887 	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
888 	 */
889 	setup_hwcaps();
890 
891 	/*
892 	 * Create kernel page tables and switch to virtual addressing.
893 	 */
894         paging_init();
895 
896         /* Setup default console */
897 	conmode_default();
898 	set_preferred_console();
899 
900 	/* Setup zfcpdump support */
901 	setup_zfcpdump();
902 }
903 
904 #ifdef CONFIG_32BIT
905 static int no_removal_warning __initdata;
906 
907 static int __init parse_no_removal_warning(char *str)
908 {
909 	no_removal_warning = 1;
910 	return 0;
911 }
912 __setup("no_removal_warning", parse_no_removal_warning);
913 
914 static int __init removal_warning(void)
915 {
916 	if (no_removal_warning)
917 		return 0;
918 	printk(KERN_ALERT "\n\n");
919 	printk(KERN_CONT "Warning - you are using a 31 bit kernel!\n\n");
920 	printk(KERN_CONT "We plan to remove 31 bit kernel support from the kernel sources in March 2015.\n");
921 	printk(KERN_CONT "Currently we assume that nobody is using the 31 bit kernel on old 31 bit\n");
922 	printk(KERN_CONT "hardware anymore. If you think that the code should not be removed and also\n");
923 	printk(KERN_CONT "future versions of the Linux kernel should be able to run in 31 bit mode\n");
924 	printk(KERN_CONT "please let us know. Please write to:\n");
925 	printk(KERN_CONT "linux390@de.ibm.com (mail address) and/or\n");
926 	printk(KERN_CONT "linux-s390@vger.kernel.org (mailing list).\n\n");
927 	printk(KERN_CONT "Thank you!\n\n");
928 	printk(KERN_CONT "If this kernel runs on a 64 bit machine you may consider using a 64 bit kernel.\n");
929 	printk(KERN_CONT "This message can be disabled with the \"no_removal_warning\" kernel parameter.\n");
930 	schedule_timeout_uninterruptible(300 * HZ);
931 	return 0;
932 }
933 early_initcall(removal_warning);
934 #endif
935