xref: /linux/arch/s390/kernel/setup.c (revision 171f1bc77c2d34308392841bcffa69b8a22c2e09)
1 /*
2  *  arch/s390/kernel/setup.c
3  *
4  *  S390 version
5  *    Copyright (C) IBM Corp. 1999,2010
6  *    Author(s): Hartmut Penner (hp@de.ibm.com),
7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
8  *
9  *  Derived from "arch/i386/kernel/setup.c"
10  *    Copyright (C) 1995, Linus Torvalds
11  */
12 
13 /*
14  * This file handles the architecture-dependent parts of initialization
15  */
16 
17 #define KMSG_COMPONENT "setup"
18 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19 
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/sched.h>
23 #include <linux/kernel.h>
24 #include <linux/mm.h>
25 #include <linux/stddef.h>
26 #include <linux/unistd.h>
27 #include <linux/ptrace.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 
49 #include <asm/ipl.h>
50 #include <asm/uaccess.h>
51 #include <asm/system.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/compat.h>
62 #include <asm/kvm_virtio.h>
63 #include <asm/diag.h>
64 
65 long psw_kernel_bits	= PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
66 			  PSW_MASK_EA | PSW_MASK_BA;
67 long psw_user_bits	= PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
68 			  PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
69 			  PSW_MASK_PSTATE | PSW_ASC_HOME;
70 
71 /*
72  * User copy operations.
73  */
74 struct uaccess_ops uaccess;
75 EXPORT_SYMBOL(uaccess);
76 
77 /*
78  * Machine setup..
79  */
80 unsigned int console_mode = 0;
81 EXPORT_SYMBOL(console_mode);
82 
83 unsigned int console_devno = -1;
84 EXPORT_SYMBOL(console_devno);
85 
86 unsigned int console_irq = -1;
87 EXPORT_SYMBOL(console_irq);
88 
89 unsigned long elf_hwcap = 0;
90 char elf_platform[ELF_PLATFORM_SIZE];
91 
92 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
93 
94 int __initdata memory_end_set;
95 unsigned long __initdata memory_end;
96 
97 /* An array with a pointer to the lowcore of every CPU. */
98 struct _lowcore *lowcore_ptr[NR_CPUS];
99 EXPORT_SYMBOL(lowcore_ptr);
100 
101 /*
102  * This is set up by the setup-routine at boot-time
103  * for S390 need to find out, what we have to setup
104  * using address 0x10400 ...
105  */
106 
107 #include <asm/setup.h>
108 
109 /*
110  * condev= and conmode= setup parameter.
111  */
112 
113 static int __init condev_setup(char *str)
114 {
115 	int vdev;
116 
117 	vdev = simple_strtoul(str, &str, 0);
118 	if (vdev >= 0 && vdev < 65536) {
119 		console_devno = vdev;
120 		console_irq = -1;
121 	}
122 	return 1;
123 }
124 
125 __setup("condev=", condev_setup);
126 
127 static void __init set_preferred_console(void)
128 {
129 	if (MACHINE_IS_KVM)
130 		add_preferred_console("hvc", 0, NULL);
131 	else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
132 		add_preferred_console("ttyS", 0, NULL);
133 	else if (CONSOLE_IS_3270)
134 		add_preferred_console("tty3270", 0, NULL);
135 }
136 
137 static int __init conmode_setup(char *str)
138 {
139 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
140 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
141                 SET_CONSOLE_SCLP;
142 #endif
143 #if defined(CONFIG_TN3215_CONSOLE)
144 	if (strncmp(str, "3215", 5) == 0)
145 		SET_CONSOLE_3215;
146 #endif
147 #if defined(CONFIG_TN3270_CONSOLE)
148 	if (strncmp(str, "3270", 5) == 0)
149 		SET_CONSOLE_3270;
150 #endif
151 	set_preferred_console();
152         return 1;
153 }
154 
155 __setup("conmode=", conmode_setup);
156 
157 static void __init conmode_default(void)
158 {
159 	char query_buffer[1024];
160 	char *ptr;
161 
162         if (MACHINE_IS_VM) {
163 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
164 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
165 		ptr = strstr(query_buffer, "SUBCHANNEL =");
166 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
167 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
168 		ptr = strstr(query_buffer, "CONMODE");
169 		/*
170 		 * Set the conmode to 3215 so that the device recognition
171 		 * will set the cu_type of the console to 3215. If the
172 		 * conmode is 3270 and we don't set it back then both
173 		 * 3215 and the 3270 driver will try to access the console
174 		 * device (3215 as console and 3270 as normal tty).
175 		 */
176 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
177 		if (ptr == NULL) {
178 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
179 			SET_CONSOLE_SCLP;
180 #endif
181 			return;
182 		}
183 		if (strncmp(ptr + 8, "3270", 4) == 0) {
184 #if defined(CONFIG_TN3270_CONSOLE)
185 			SET_CONSOLE_3270;
186 #elif defined(CONFIG_TN3215_CONSOLE)
187 			SET_CONSOLE_3215;
188 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
189 			SET_CONSOLE_SCLP;
190 #endif
191 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
192 #if defined(CONFIG_TN3215_CONSOLE)
193 			SET_CONSOLE_3215;
194 #elif defined(CONFIG_TN3270_CONSOLE)
195 			SET_CONSOLE_3270;
196 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
197 			SET_CONSOLE_SCLP;
198 #endif
199 		}
200 	} else {
201 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
202 		SET_CONSOLE_SCLP;
203 #endif
204 	}
205 }
206 
207 #ifdef CONFIG_ZFCPDUMP
208 static void __init setup_zfcpdump(unsigned int console_devno)
209 {
210 	static char str[41];
211 
212 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
213 		return;
214 	if (OLDMEM_BASE)
215 		return;
216 	if (console_devno != -1)
217 		sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
218 			ipl_info.data.fcp.dev_id.devno, console_devno);
219 	else
220 		sprintf(str, " cio_ignore=all,!0.0.%04x",
221 			ipl_info.data.fcp.dev_id.devno);
222 	strcat(boot_command_line, str);
223 	console_loglevel = 2;
224 }
225 #else
226 static inline void setup_zfcpdump(unsigned int console_devno) {}
227 #endif /* CONFIG_ZFCPDUMP */
228 
229  /*
230  * Reboot, halt and power_off stubs. They just call _machine_restart,
231  * _machine_halt or _machine_power_off.
232  */
233 
234 void machine_restart(char *command)
235 {
236 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
237 		/*
238 		 * Only unblank the console if we are called in enabled
239 		 * context or a bust_spinlocks cleared the way for us.
240 		 */
241 		console_unblank();
242 	_machine_restart(command);
243 }
244 
245 void machine_halt(void)
246 {
247 	if (!in_interrupt() || oops_in_progress)
248 		/*
249 		 * Only unblank the console if we are called in enabled
250 		 * context or a bust_spinlocks cleared the way for us.
251 		 */
252 		console_unblank();
253 	_machine_halt();
254 }
255 
256 void machine_power_off(void)
257 {
258 	if (!in_interrupt() || oops_in_progress)
259 		/*
260 		 * Only unblank the console if we are called in enabled
261 		 * context or a bust_spinlocks cleared the way for us.
262 		 */
263 		console_unblank();
264 	_machine_power_off();
265 }
266 
267 /*
268  * Dummy power off function.
269  */
270 void (*pm_power_off)(void) = machine_power_off;
271 
272 static int __init early_parse_mem(char *p)
273 {
274 	memory_end = memparse(p, &p);
275 	memory_end_set = 1;
276 	return 0;
277 }
278 early_param("mem", early_parse_mem);
279 
280 unsigned int user_mode = HOME_SPACE_MODE;
281 EXPORT_SYMBOL_GPL(user_mode);
282 
283 static int set_amode_primary(void)
284 {
285 	psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
286 	psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
287 #ifdef CONFIG_COMPAT
288 	psw32_user_bits =
289 		(psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
290 #endif
291 
292 	if (MACHINE_HAS_MVCOS) {
293 		memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
294 		return 1;
295 	} else {
296 		memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
297 		return 0;
298 	}
299 }
300 
301 /*
302  * Switch kernel/user addressing modes?
303  */
304 static int __init early_parse_switch_amode(char *p)
305 {
306 	user_mode = PRIMARY_SPACE_MODE;
307 	return 0;
308 }
309 early_param("switch_amode", early_parse_switch_amode);
310 
311 static int __init early_parse_user_mode(char *p)
312 {
313 	if (p && strcmp(p, "primary") == 0)
314 		user_mode = PRIMARY_SPACE_MODE;
315 	else if (!p || strcmp(p, "home") == 0)
316 		user_mode = HOME_SPACE_MODE;
317 	else
318 		return 1;
319 	return 0;
320 }
321 early_param("user_mode", early_parse_user_mode);
322 
323 static void setup_addressing_mode(void)
324 {
325 	if (user_mode == PRIMARY_SPACE_MODE) {
326 		if (set_amode_primary())
327 			pr_info("Address spaces switched, "
328 				"mvcos available\n");
329 		else
330 			pr_info("Address spaces switched, "
331 				"mvcos not available\n");
332 	}
333 }
334 
335 static void __init
336 setup_lowcore(void)
337 {
338 	struct _lowcore *lc;
339 
340 	/*
341 	 * Setup lowcore for boot cpu
342 	 */
343 	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
344 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
345 	lc->restart_psw.mask = psw_kernel_bits;
346 	lc->restart_psw.addr =
347 		PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
348 	lc->external_new_psw.mask = psw_kernel_bits |
349 		PSW_MASK_DAT | PSW_MASK_MCHECK;
350 	lc->external_new_psw.addr =
351 		PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
352 	lc->svc_new_psw.mask = psw_kernel_bits |
353 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
354 	lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
355 	lc->program_new_psw.mask = psw_kernel_bits |
356 		PSW_MASK_DAT | PSW_MASK_MCHECK;
357 	lc->program_new_psw.addr =
358 		PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
359 	lc->mcck_new_psw.mask = psw_kernel_bits;
360 	lc->mcck_new_psw.addr =
361 		PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
362 	lc->io_new_psw.mask = psw_kernel_bits |
363 		PSW_MASK_DAT | PSW_MASK_MCHECK;
364 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
365 	lc->clock_comparator = -1ULL;
366 	lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
367 	lc->async_stack = (unsigned long)
368 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
369 	lc->panic_stack = (unsigned long)
370 		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
371 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
372 	lc->thread_info = (unsigned long) &init_thread_union;
373 	lc->machine_flags = S390_lowcore.machine_flags;
374 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
375 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
376 	       MAX_FACILITY_BIT/8);
377 #ifndef CONFIG_64BIT
378 	if (MACHINE_HAS_IEEE) {
379 		lc->extended_save_area_addr = (__u32)
380 			__alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
381 		/* enable extended save area */
382 		__ctl_set_bit(14, 29);
383 	}
384 #else
385 	lc->cmf_hpp = -1ULL;
386 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
387 #endif
388 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
389 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
390 	lc->exit_timer = S390_lowcore.exit_timer;
391 	lc->user_timer = S390_lowcore.user_timer;
392 	lc->system_timer = S390_lowcore.system_timer;
393 	lc->steal_timer = S390_lowcore.steal_timer;
394 	lc->last_update_timer = S390_lowcore.last_update_timer;
395 	lc->last_update_clock = S390_lowcore.last_update_clock;
396 	lc->ftrace_func = S390_lowcore.ftrace_func;
397 	set_prefix((u32)(unsigned long) lc);
398 	lowcore_ptr[0] = lc;
399 }
400 
401 static struct resource code_resource = {
402 	.name  = "Kernel code",
403 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
404 };
405 
406 static struct resource data_resource = {
407 	.name = "Kernel data",
408 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
409 };
410 
411 static struct resource bss_resource = {
412 	.name = "Kernel bss",
413 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
414 };
415 
416 static struct resource __initdata *standard_resources[] = {
417 	&code_resource,
418 	&data_resource,
419 	&bss_resource,
420 };
421 
422 static void __init setup_resources(void)
423 {
424 	struct resource *res, *std_res, *sub_res;
425 	int i, j;
426 
427 	code_resource.start = (unsigned long) &_text;
428 	code_resource.end = (unsigned long) &_etext - 1;
429 	data_resource.start = (unsigned long) &_etext;
430 	data_resource.end = (unsigned long) &_edata - 1;
431 	bss_resource.start = (unsigned long) &__bss_start;
432 	bss_resource.end = (unsigned long) &__bss_stop - 1;
433 
434 	for (i = 0; i < MEMORY_CHUNKS; i++) {
435 		if (!memory_chunk[i].size)
436 			continue;
437 		if (memory_chunk[i].type == CHUNK_OLDMEM ||
438 		    memory_chunk[i].type == CHUNK_CRASHK)
439 			continue;
440 		res = alloc_bootmem_low(sizeof(*res));
441 		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
442 		switch (memory_chunk[i].type) {
443 		case CHUNK_READ_WRITE:
444 		case CHUNK_CRASHK:
445 			res->name = "System RAM";
446 			break;
447 		case CHUNK_READ_ONLY:
448 			res->name = "System ROM";
449 			res->flags |= IORESOURCE_READONLY;
450 			break;
451 		default:
452 			res->name = "reserved";
453 		}
454 		res->start = memory_chunk[i].addr;
455 		res->end = res->start + memory_chunk[i].size - 1;
456 		request_resource(&iomem_resource, res);
457 
458 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
459 			std_res = standard_resources[j];
460 			if (std_res->start < res->start ||
461 			    std_res->start > res->end)
462 				continue;
463 			if (std_res->end > res->end) {
464 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
465 				*sub_res = *std_res;
466 				sub_res->end = res->end;
467 				std_res->start = res->end + 1;
468 				request_resource(res, sub_res);
469 			} else {
470 				request_resource(res, std_res);
471 			}
472 		}
473 	}
474 }
475 
476 unsigned long real_memory_size;
477 EXPORT_SYMBOL_GPL(real_memory_size);
478 
479 static void __init setup_memory_end(void)
480 {
481 	unsigned long memory_size;
482 	unsigned long max_mem;
483 	int i;
484 
485 
486 #ifdef CONFIG_ZFCPDUMP
487 	if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
488 		memory_end = ZFCPDUMP_HSA_SIZE;
489 		memory_end_set = 1;
490 	}
491 #endif
492 	memory_size = 0;
493 	memory_end &= PAGE_MASK;
494 
495 	max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
496 	memory_end = min(max_mem, memory_end);
497 
498 	/*
499 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
500 	 * extra checks that HOLES_IN_ZONE would require.
501 	 */
502 	for (i = 0; i < MEMORY_CHUNKS; i++) {
503 		unsigned long start, end;
504 		struct mem_chunk *chunk;
505 		unsigned long align;
506 
507 		chunk = &memory_chunk[i];
508 		align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
509 		start = (chunk->addr + align - 1) & ~(align - 1);
510 		end = (chunk->addr + chunk->size) & ~(align - 1);
511 		if (start >= end)
512 			memset(chunk, 0, sizeof(*chunk));
513 		else {
514 			chunk->addr = start;
515 			chunk->size = end - start;
516 		}
517 	}
518 
519 	for (i = 0; i < MEMORY_CHUNKS; i++) {
520 		struct mem_chunk *chunk = &memory_chunk[i];
521 
522 		real_memory_size = max(real_memory_size,
523 				       chunk->addr + chunk->size);
524 		if (chunk->addr >= max_mem) {
525 			memset(chunk, 0, sizeof(*chunk));
526 			continue;
527 		}
528 		if (chunk->addr + chunk->size > max_mem)
529 			chunk->size = max_mem - chunk->addr;
530 		memory_size = max(memory_size, chunk->addr + chunk->size);
531 	}
532 	if (!memory_end)
533 		memory_end = memory_size;
534 }
535 
536 void *restart_stack __attribute__((__section__(".data")));
537 
538 /*
539  * Setup new PSW and allocate stack for PSW restart interrupt
540  */
541 static void __init setup_restart_psw(void)
542 {
543 	psw_t psw;
544 
545 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
546 	restart_stack += ASYNC_SIZE;
547 
548 	/*
549 	 * Setup restart PSW for absolute zero lowcore. This is necesary
550 	 * if PSW restart is done on an offline CPU that has lowcore zero
551 	 */
552 	psw.mask = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_EA | PSW_MASK_BA;
553 	psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
554 	copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
555 }
556 
557 static void __init setup_vmcoreinfo(void)
558 {
559 #ifdef CONFIG_KEXEC
560 	unsigned long ptr = paddr_vmcoreinfo_note();
561 
562 	copy_to_absolute_zero(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
563 #endif
564 }
565 
566 #ifdef CONFIG_CRASH_DUMP
567 
568 /*
569  * Find suitable location for crashkernel memory
570  */
571 static unsigned long __init find_crash_base(unsigned long crash_size,
572 					    char **msg)
573 {
574 	unsigned long crash_base;
575 	struct mem_chunk *chunk;
576 	int i;
577 
578 	if (memory_chunk[0].size < crash_size) {
579 		*msg = "first memory chunk must be at least crashkernel size";
580 		return 0;
581 	}
582 	if (is_kdump_kernel() && (crash_size == OLDMEM_SIZE))
583 		return OLDMEM_BASE;
584 
585 	for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
586 		chunk = &memory_chunk[i];
587 		if (chunk->size == 0)
588 			continue;
589 		if (chunk->type != CHUNK_READ_WRITE)
590 			continue;
591 		if (chunk->size < crash_size)
592 			continue;
593 		crash_base = (chunk->addr + chunk->size) - crash_size;
594 		if (crash_base < crash_size)
595 			continue;
596 		if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
597 			continue;
598 		if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
599 			continue;
600 		return crash_base;
601 	}
602 	*msg = "no suitable area found";
603 	return 0;
604 }
605 
606 /*
607  * Check if crash_base and crash_size is valid
608  */
609 static int __init verify_crash_base(unsigned long crash_base,
610 				    unsigned long crash_size,
611 				    char **msg)
612 {
613 	struct mem_chunk *chunk;
614 	int i;
615 
616 	/*
617 	 * Because we do the swap to zero, we must have at least 'crash_size'
618 	 * bytes free space before crash_base
619 	 */
620 	if (crash_size > crash_base) {
621 		*msg = "crashkernel offset must be greater than size";
622 		return -EINVAL;
623 	}
624 
625 	/* First memory chunk must be at least crash_size */
626 	if (memory_chunk[0].size < crash_size) {
627 		*msg = "first memory chunk must be at least crashkernel size";
628 		return -EINVAL;
629 	}
630 	/* Check if we fit into the respective memory chunk */
631 	for (i = 0; i < MEMORY_CHUNKS; i++) {
632 		chunk = &memory_chunk[i];
633 		if (chunk->size == 0)
634 			continue;
635 		if (crash_base < chunk->addr)
636 			continue;
637 		if (crash_base >= chunk->addr + chunk->size)
638 			continue;
639 		/* we have found the memory chunk */
640 		if (crash_base + crash_size > chunk->addr + chunk->size) {
641 			*msg = "selected memory chunk is too small for "
642 				"crashkernel memory";
643 			return -EINVAL;
644 		}
645 		return 0;
646 	}
647 	*msg = "invalid memory range specified";
648 	return -EINVAL;
649 }
650 
651 /*
652  * Reserve kdump memory by creating a memory hole in the mem_chunk array
653  */
654 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
655 					 int type)
656 {
657 
658 	create_mem_hole(memory_chunk, addr, size, type);
659 }
660 
661 /*
662  * When kdump is enabled, we have to ensure that no memory from
663  * the area [0 - crashkernel memory size] and
664  * [crashk_res.start - crashk_res.end] is set offline.
665  */
666 static int kdump_mem_notifier(struct notifier_block *nb,
667 			      unsigned long action, void *data)
668 {
669 	struct memory_notify *arg = data;
670 
671 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
672 		return NOTIFY_BAD;
673 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
674 		return NOTIFY_OK;
675 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
676 		return NOTIFY_OK;
677 	return NOTIFY_BAD;
678 }
679 
680 static struct notifier_block kdump_mem_nb = {
681 	.notifier_call = kdump_mem_notifier,
682 };
683 
684 #endif
685 
686 /*
687  * Make sure that oldmem, where the dump is stored, is protected
688  */
689 static void reserve_oldmem(void)
690 {
691 #ifdef CONFIG_CRASH_DUMP
692 	if (!OLDMEM_BASE)
693 		return;
694 
695 	reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
696 	reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
697 			      CHUNK_OLDMEM);
698 	if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
699 		saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
700 	else
701 		saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
702 #endif
703 }
704 
705 /*
706  * Reserve memory for kdump kernel to be loaded with kexec
707  */
708 static void __init reserve_crashkernel(void)
709 {
710 #ifdef CONFIG_CRASH_DUMP
711 	unsigned long long crash_base, crash_size;
712 	char *msg;
713 	int rc;
714 
715 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
716 			       &crash_base);
717 	if (rc || crash_size == 0)
718 		return;
719 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
720 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
721 	if (register_memory_notifier(&kdump_mem_nb))
722 		return;
723 	if (!crash_base)
724 		crash_base = find_crash_base(crash_size, &msg);
725 	if (!crash_base) {
726 		pr_info("crashkernel reservation failed: %s\n", msg);
727 		unregister_memory_notifier(&kdump_mem_nb);
728 		return;
729 	}
730 	if (verify_crash_base(crash_base, crash_size, &msg)) {
731 		pr_info("crashkernel reservation failed: %s\n", msg);
732 		unregister_memory_notifier(&kdump_mem_nb);
733 		return;
734 	}
735 	if (!OLDMEM_BASE && MACHINE_IS_VM)
736 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
737 	crashk_res.start = crash_base;
738 	crashk_res.end = crash_base + crash_size - 1;
739 	insert_resource(&iomem_resource, &crashk_res);
740 	reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
741 	pr_info("Reserving %lluMB of memory at %lluMB "
742 		"for crashkernel (System RAM: %luMB)\n",
743 		crash_size >> 20, crash_base >> 20, memory_end >> 20);
744 #endif
745 }
746 
747 static void __init
748 setup_memory(void)
749 {
750         unsigned long bootmap_size;
751 	unsigned long start_pfn, end_pfn;
752 	int i;
753 
754 	/*
755 	 * partially used pages are not usable - thus
756 	 * we are rounding upwards:
757 	 */
758 	start_pfn = PFN_UP(__pa(&_end));
759 	end_pfn = max_pfn = PFN_DOWN(memory_end);
760 
761 #ifdef CONFIG_BLK_DEV_INITRD
762 	/*
763 	 * Move the initrd in case the bitmap of the bootmem allocater
764 	 * would overwrite it.
765 	 */
766 
767 	if (INITRD_START && INITRD_SIZE) {
768 		unsigned long bmap_size;
769 		unsigned long start;
770 
771 		bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
772 		bmap_size = PFN_PHYS(bmap_size);
773 
774 		if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
775 			start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
776 
777 #ifdef CONFIG_CRASH_DUMP
778 			if (OLDMEM_BASE) {
779 				/* Move initrd behind kdump oldmem */
780 				if (start + INITRD_SIZE > OLDMEM_BASE &&
781 				    start < OLDMEM_BASE + OLDMEM_SIZE)
782 					start = OLDMEM_BASE + OLDMEM_SIZE;
783 			}
784 #endif
785 			if (start + INITRD_SIZE > memory_end) {
786 				pr_err("initrd extends beyond end of "
787 				       "memory (0x%08lx > 0x%08lx) "
788 				       "disabling initrd\n",
789 				       start + INITRD_SIZE, memory_end);
790 				INITRD_START = INITRD_SIZE = 0;
791 			} else {
792 				pr_info("Moving initrd (0x%08lx -> "
793 					"0x%08lx, size: %ld)\n",
794 					INITRD_START, start, INITRD_SIZE);
795 				memmove((void *) start, (void *) INITRD_START,
796 					INITRD_SIZE);
797 				INITRD_START = start;
798 			}
799 		}
800 	}
801 #endif
802 
803 	/*
804 	 * Initialize the boot-time allocator
805 	 */
806 	bootmap_size = init_bootmem(start_pfn, end_pfn);
807 
808 	/*
809 	 * Register RAM areas with the bootmem allocator.
810 	 */
811 
812 	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
813 		unsigned long start_chunk, end_chunk, pfn;
814 
815 		if (memory_chunk[i].type != CHUNK_READ_WRITE &&
816 		    memory_chunk[i].type != CHUNK_CRASHK)
817 			continue;
818 		start_chunk = PFN_DOWN(memory_chunk[i].addr);
819 		end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
820 		end_chunk = min(end_chunk, end_pfn);
821 		if (start_chunk >= end_chunk)
822 			continue;
823 		add_active_range(0, start_chunk, end_chunk);
824 		pfn = max(start_chunk, start_pfn);
825 		for (; pfn < end_chunk; pfn++)
826 			page_set_storage_key(PFN_PHYS(pfn),
827 					     PAGE_DEFAULT_KEY, 0);
828 	}
829 
830 	psw_set_key(PAGE_DEFAULT_KEY);
831 
832 	free_bootmem_with_active_regions(0, max_pfn);
833 
834 	/*
835 	 * Reserve memory used for lowcore/command line/kernel image.
836 	 */
837 	reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
838 	reserve_bootmem((unsigned long)_stext,
839 			PFN_PHYS(start_pfn) - (unsigned long)_stext,
840 			BOOTMEM_DEFAULT);
841 	/*
842 	 * Reserve the bootmem bitmap itself as well. We do this in two
843 	 * steps (first step was init_bootmem()) because this catches
844 	 * the (very unlikely) case of us accidentally initializing the
845 	 * bootmem allocator with an invalid RAM area.
846 	 */
847 	reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
848 			BOOTMEM_DEFAULT);
849 
850 #ifdef CONFIG_CRASH_DUMP
851 	if (crashk_res.start)
852 		reserve_bootmem(crashk_res.start,
853 				crashk_res.end - crashk_res.start + 1,
854 				BOOTMEM_DEFAULT);
855 	if (is_kdump_kernel())
856 		reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
857 				PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
858 #endif
859 #ifdef CONFIG_BLK_DEV_INITRD
860 	if (INITRD_START && INITRD_SIZE) {
861 		if (INITRD_START + INITRD_SIZE <= memory_end) {
862 			reserve_bootmem(INITRD_START, INITRD_SIZE,
863 					BOOTMEM_DEFAULT);
864 			initrd_start = INITRD_START;
865 			initrd_end = initrd_start + INITRD_SIZE;
866 		} else {
867 			pr_err("initrd extends beyond end of "
868 			       "memory (0x%08lx > 0x%08lx) "
869 			       "disabling initrd\n",
870 			       initrd_start + INITRD_SIZE, memory_end);
871 			initrd_start = initrd_end = 0;
872 		}
873 	}
874 #endif
875 }
876 
877 /*
878  * Setup hardware capabilities.
879  */
880 static void __init setup_hwcaps(void)
881 {
882 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
883 	struct cpuid cpu_id;
884 	int i;
885 
886 	/*
887 	 * The store facility list bits numbers as found in the principles
888 	 * of operation are numbered with bit 1UL<<31 as number 0 to
889 	 * bit 1UL<<0 as number 31.
890 	 *   Bit 0: instructions named N3, "backported" to esa-mode
891 	 *   Bit 2: z/Architecture mode is active
892 	 *   Bit 7: the store-facility-list-extended facility is installed
893 	 *   Bit 17: the message-security assist is installed
894 	 *   Bit 19: the long-displacement facility is installed
895 	 *   Bit 21: the extended-immediate facility is installed
896 	 *   Bit 22: extended-translation facility 3 is installed
897 	 *   Bit 30: extended-translation facility 3 enhancement facility
898 	 * These get translated to:
899 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
900 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
901 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
902 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
903 	 */
904 	for (i = 0; i < 6; i++)
905 		if (test_facility(stfl_bits[i]))
906 			elf_hwcap |= 1UL << i;
907 
908 	if (test_facility(22) && test_facility(30))
909 		elf_hwcap |= HWCAP_S390_ETF3EH;
910 
911 	/*
912 	 * Check for additional facilities with store-facility-list-extended.
913 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
914 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
915 	 * as stored by stfl, bits 32-xxx contain additional facilities.
916 	 * How many facility words are stored depends on the number of
917 	 * doublewords passed to the instruction. The additional facilities
918 	 * are:
919 	 *   Bit 42: decimal floating point facility is installed
920 	 *   Bit 44: perform floating point operation facility is installed
921 	 * translated to:
922 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
923 	 */
924 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
925 		elf_hwcap |= HWCAP_S390_DFP;
926 
927 	/*
928 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
929 	 */
930 	if (MACHINE_HAS_HPAGE)
931 		elf_hwcap |= HWCAP_S390_HPAGE;
932 
933 	/*
934 	 * 64-bit register support for 31-bit processes
935 	 * HWCAP_S390_HIGH_GPRS is bit 9.
936 	 */
937 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
938 
939 	get_cpu_id(&cpu_id);
940 	switch (cpu_id.machine) {
941 	case 0x9672:
942 #if !defined(CONFIG_64BIT)
943 	default:	/* Use "g5" as default for 31 bit kernels. */
944 #endif
945 		strcpy(elf_platform, "g5");
946 		break;
947 	case 0x2064:
948 	case 0x2066:
949 #if defined(CONFIG_64BIT)
950 	default:	/* Use "z900" as default for 64 bit kernels. */
951 #endif
952 		strcpy(elf_platform, "z900");
953 		break;
954 	case 0x2084:
955 	case 0x2086:
956 		strcpy(elf_platform, "z990");
957 		break;
958 	case 0x2094:
959 	case 0x2096:
960 		strcpy(elf_platform, "z9-109");
961 		break;
962 	case 0x2097:
963 	case 0x2098:
964 		strcpy(elf_platform, "z10");
965 		break;
966 	case 0x2817:
967 	case 0x2818:
968 		strcpy(elf_platform, "z196");
969 		break;
970 	}
971 }
972 
973 /*
974  * Setup function called from init/main.c just after the banner
975  * was printed.
976  */
977 
978 void __init
979 setup_arch(char **cmdline_p)
980 {
981         /*
982          * print what head.S has found out about the machine
983          */
984 #ifndef CONFIG_64BIT
985 	if (MACHINE_IS_VM)
986 		pr_info("Linux is running as a z/VM "
987 			"guest operating system in 31-bit mode\n");
988 	else if (MACHINE_IS_LPAR)
989 		pr_info("Linux is running natively in 31-bit mode\n");
990 	if (MACHINE_HAS_IEEE)
991 		pr_info("The hardware system has IEEE compatible "
992 			"floating point units\n");
993 	else
994 		pr_info("The hardware system has no IEEE compatible "
995 			"floating point units\n");
996 #else /* CONFIG_64BIT */
997 	if (MACHINE_IS_VM)
998 		pr_info("Linux is running as a z/VM "
999 			"guest operating system in 64-bit mode\n");
1000 	else if (MACHINE_IS_KVM)
1001 		pr_info("Linux is running under KVM in 64-bit mode\n");
1002 	else if (MACHINE_IS_LPAR)
1003 		pr_info("Linux is running natively in 64-bit mode\n");
1004 #endif /* CONFIG_64BIT */
1005 
1006 	/* Have one command line that is parsed and saved in /proc/cmdline */
1007 	/* boot_command_line has been already set up in early.c */
1008 	*cmdline_p = boot_command_line;
1009 
1010         ROOT_DEV = Root_RAM0;
1011 
1012 	init_mm.start_code = PAGE_OFFSET;
1013 	init_mm.end_code = (unsigned long) &_etext;
1014 	init_mm.end_data = (unsigned long) &_edata;
1015 	init_mm.brk = (unsigned long) &_end;
1016 
1017 	if (MACHINE_HAS_MVCOS)
1018 		memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1019 	else
1020 		memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1021 
1022 	parse_early_param();
1023 
1024 	setup_ipl();
1025 	setup_memory_end();
1026 	setup_addressing_mode();
1027 	reserve_oldmem();
1028 	reserve_crashkernel();
1029 	setup_memory();
1030 	setup_resources();
1031 	setup_vmcoreinfo();
1032 	setup_restart_psw();
1033 	setup_lowcore();
1034 
1035         cpu_init();
1036 	s390_init_cpu_topology();
1037 
1038 	/*
1039 	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1040 	 */
1041 	setup_hwcaps();
1042 
1043 	/*
1044 	 * Create kernel page tables and switch to virtual addressing.
1045 	 */
1046         paging_init();
1047 
1048         /* Setup default console */
1049 	conmode_default();
1050 	set_preferred_console();
1051 
1052 	/* Setup zfcpdump support */
1053 	setup_zfcpdump(console_devno);
1054 }
1055