xref: /linux/arch/s390/kernel/setup.c (revision 9d14070f656addddce3d63fd483de46930b51850)
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/memblock.h>
25 #include <linux/mm.h>
26 #include <linux/stddef.h>
27 #include <linux/unistd.h>
28 #include <linux/ptrace.h>
29 #include <linux/user.h>
30 #include <linux/tty.h>
31 #include <linux/ioport.h>
32 #include <linux/delay.h>
33 #include <linux/init.h>
34 #include <linux/initrd.h>
35 #include <linux/bootmem.h>
36 #include <linux/root_dev.h>
37 #include <linux/console.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/device.h>
40 #include <linux/notifier.h>
41 #include <linux/pfn.h>
42 #include <linux/ctype.h>
43 #include <linux/reboot.h>
44 #include <linux/topology.h>
45 #include <linux/ftrace.h>
46 #include <linux/kexec.h>
47 #include <linux/crash_dump.h>
48 #include <linux/memory.h>
49 
50 #include <asm/ipl.h>
51 #include <asm/uaccess.h>
52 #include <asm/system.h>
53 #include <asm/smp.h>
54 #include <asm/mmu_context.h>
55 #include <asm/cpcmd.h>
56 #include <asm/lowcore.h>
57 #include <asm/irq.h>
58 #include <asm/page.h>
59 #include <asm/ptrace.h>
60 #include <asm/sections.h>
61 #include <asm/ebcdic.h>
62 #include <asm/compat.h>
63 #include <asm/kvm_virtio.h>
64 #include <asm/diag.h>
65 
66 long psw_kernel_bits	= PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
67 			  PSW_MASK_EA | PSW_MASK_BA;
68 long psw_user_bits	= PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
69 			  PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
70 			  PSW_MASK_PSTATE | PSW_ASC_HOME;
71 
72 /*
73  * User copy operations.
74  */
75 struct uaccess_ops uaccess;
76 EXPORT_SYMBOL(uaccess);
77 
78 /*
79  * Machine setup..
80  */
81 unsigned int console_mode = 0;
82 EXPORT_SYMBOL(console_mode);
83 
84 unsigned int console_devno = -1;
85 EXPORT_SYMBOL(console_devno);
86 
87 unsigned int console_irq = -1;
88 EXPORT_SYMBOL(console_irq);
89 
90 unsigned long elf_hwcap = 0;
91 char elf_platform[ELF_PLATFORM_SIZE];
92 
93 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
94 
95 int __initdata memory_end_set;
96 unsigned long __initdata memory_end;
97 
98 /* An array with a pointer to the lowcore of every CPU. */
99 struct _lowcore *lowcore_ptr[NR_CPUS];
100 EXPORT_SYMBOL(lowcore_ptr);
101 
102 /*
103  * This is set up by the setup-routine at boot-time
104  * for S390 need to find out, what we have to setup
105  * using address 0x10400 ...
106  */
107 
108 #include <asm/setup.h>
109 
110 /*
111  * condev= and conmode= setup parameter.
112  */
113 
114 static int __init condev_setup(char *str)
115 {
116 	int vdev;
117 
118 	vdev = simple_strtoul(str, &str, 0);
119 	if (vdev >= 0 && vdev < 65536) {
120 		console_devno = vdev;
121 		console_irq = -1;
122 	}
123 	return 1;
124 }
125 
126 __setup("condev=", condev_setup);
127 
128 static void __init set_preferred_console(void)
129 {
130 	if (MACHINE_IS_KVM)
131 		add_preferred_console("hvc", 0, NULL);
132 	else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
133 		add_preferred_console("ttyS", 0, NULL);
134 	else if (CONSOLE_IS_3270)
135 		add_preferred_console("tty3270", 0, NULL);
136 }
137 
138 static int __init conmode_setup(char *str)
139 {
140 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
141 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
142                 SET_CONSOLE_SCLP;
143 #endif
144 #if defined(CONFIG_TN3215_CONSOLE)
145 	if (strncmp(str, "3215", 5) == 0)
146 		SET_CONSOLE_3215;
147 #endif
148 #if defined(CONFIG_TN3270_CONSOLE)
149 	if (strncmp(str, "3270", 5) == 0)
150 		SET_CONSOLE_3270;
151 #endif
152 	set_preferred_console();
153         return 1;
154 }
155 
156 __setup("conmode=", conmode_setup);
157 
158 static void __init conmode_default(void)
159 {
160 	char query_buffer[1024];
161 	char *ptr;
162 
163         if (MACHINE_IS_VM) {
164 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
165 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
166 		ptr = strstr(query_buffer, "SUBCHANNEL =");
167 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
168 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
169 		ptr = strstr(query_buffer, "CONMODE");
170 		/*
171 		 * Set the conmode to 3215 so that the device recognition
172 		 * will set the cu_type of the console to 3215. If the
173 		 * conmode is 3270 and we don't set it back then both
174 		 * 3215 and the 3270 driver will try to access the console
175 		 * device (3215 as console and 3270 as normal tty).
176 		 */
177 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
178 		if (ptr == NULL) {
179 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
180 			SET_CONSOLE_SCLP;
181 #endif
182 			return;
183 		}
184 		if (strncmp(ptr + 8, "3270", 4) == 0) {
185 #if defined(CONFIG_TN3270_CONSOLE)
186 			SET_CONSOLE_3270;
187 #elif defined(CONFIG_TN3215_CONSOLE)
188 			SET_CONSOLE_3215;
189 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
190 			SET_CONSOLE_SCLP;
191 #endif
192 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
193 #if defined(CONFIG_TN3215_CONSOLE)
194 			SET_CONSOLE_3215;
195 #elif defined(CONFIG_TN3270_CONSOLE)
196 			SET_CONSOLE_3270;
197 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198 			SET_CONSOLE_SCLP;
199 #endif
200 		}
201 	} else {
202 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
203 		SET_CONSOLE_SCLP;
204 #endif
205 	}
206 }
207 
208 #ifdef CONFIG_ZFCPDUMP
209 static void __init setup_zfcpdump(unsigned int console_devno)
210 {
211 	static char str[41];
212 
213 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
214 		return;
215 	if (OLDMEM_BASE)
216 		return;
217 	if (console_devno != -1)
218 		sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
219 			ipl_info.data.fcp.dev_id.devno, console_devno);
220 	else
221 		sprintf(str, " cio_ignore=all,!0.0.%04x",
222 			ipl_info.data.fcp.dev_id.devno);
223 	strcat(boot_command_line, str);
224 	console_loglevel = 2;
225 }
226 #else
227 static inline void setup_zfcpdump(unsigned int console_devno) {}
228 #endif /* CONFIG_ZFCPDUMP */
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 
273 static int __init early_parse_mem(char *p)
274 {
275 	memory_end = memparse(p, &p);
276 	memory_end_set = 1;
277 	return 0;
278 }
279 early_param("mem", early_parse_mem);
280 
281 unsigned int user_mode = HOME_SPACE_MODE;
282 EXPORT_SYMBOL_GPL(user_mode);
283 
284 static int set_amode_primary(void)
285 {
286 	psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
287 	psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
288 #ifdef CONFIG_COMPAT
289 	psw32_user_bits =
290 		(psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
291 #endif
292 
293 	if (MACHINE_HAS_MVCOS) {
294 		memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
295 		return 1;
296 	} else {
297 		memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
298 		return 0;
299 	}
300 }
301 
302 /*
303  * Switch kernel/user addressing modes?
304  */
305 static int __init early_parse_switch_amode(char *p)
306 {
307 	user_mode = PRIMARY_SPACE_MODE;
308 	return 0;
309 }
310 early_param("switch_amode", early_parse_switch_amode);
311 
312 static int __init early_parse_user_mode(char *p)
313 {
314 	if (p && strcmp(p, "primary") == 0)
315 		user_mode = PRIMARY_SPACE_MODE;
316 	else if (!p || strcmp(p, "home") == 0)
317 		user_mode = HOME_SPACE_MODE;
318 	else
319 		return 1;
320 	return 0;
321 }
322 early_param("user_mode", early_parse_user_mode);
323 
324 static void setup_addressing_mode(void)
325 {
326 	if (user_mode == PRIMARY_SPACE_MODE) {
327 		if (set_amode_primary())
328 			pr_info("Address spaces switched, "
329 				"mvcos available\n");
330 		else
331 			pr_info("Address spaces switched, "
332 				"mvcos not available\n");
333 	}
334 }
335 
336 static void __init
337 setup_lowcore(void)
338 {
339 	struct _lowcore *lc;
340 
341 	/*
342 	 * Setup lowcore for boot cpu
343 	 */
344 	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
345 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
346 	lc->restart_psw.mask = psw_kernel_bits;
347 	lc->restart_psw.addr =
348 		PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
349 	lc->external_new_psw.mask = psw_kernel_bits |
350 		PSW_MASK_DAT | PSW_MASK_MCHECK;
351 	lc->external_new_psw.addr =
352 		PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
353 	lc->svc_new_psw.mask = psw_kernel_bits |
354 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
355 	lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
356 	lc->program_new_psw.mask = psw_kernel_bits |
357 		PSW_MASK_DAT | PSW_MASK_MCHECK;
358 	lc->program_new_psw.addr =
359 		PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
360 	lc->mcck_new_psw.mask = psw_kernel_bits;
361 	lc->mcck_new_psw.addr =
362 		PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
363 	lc->io_new_psw.mask = psw_kernel_bits |
364 		PSW_MASK_DAT | PSW_MASK_MCHECK;
365 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
366 	lc->clock_comparator = -1ULL;
367 	lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
368 	lc->async_stack = (unsigned long)
369 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
370 	lc->panic_stack = (unsigned long)
371 		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
372 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
373 	lc->thread_info = (unsigned long) &init_thread_union;
374 	lc->machine_flags = S390_lowcore.machine_flags;
375 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
376 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
377 	       MAX_FACILITY_BIT/8);
378 #ifndef CONFIG_64BIT
379 	if (MACHINE_HAS_IEEE) {
380 		lc->extended_save_area_addr = (__u32)
381 			__alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
382 		/* enable extended save area */
383 		__ctl_set_bit(14, 29);
384 	}
385 #else
386 	lc->cmf_hpp = -1ULL;
387 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
388 #endif
389 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
390 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
391 	lc->exit_timer = S390_lowcore.exit_timer;
392 	lc->user_timer = S390_lowcore.user_timer;
393 	lc->system_timer = S390_lowcore.system_timer;
394 	lc->steal_timer = S390_lowcore.steal_timer;
395 	lc->last_update_timer = S390_lowcore.last_update_timer;
396 	lc->last_update_clock = S390_lowcore.last_update_clock;
397 	lc->ftrace_func = S390_lowcore.ftrace_func;
398 	set_prefix((u32)(unsigned long) lc);
399 	lowcore_ptr[0] = lc;
400 }
401 
402 static struct resource code_resource = {
403 	.name  = "Kernel code",
404 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
405 };
406 
407 static struct resource data_resource = {
408 	.name = "Kernel data",
409 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
410 };
411 
412 static struct resource bss_resource = {
413 	.name = "Kernel bss",
414 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
415 };
416 
417 static struct resource __initdata *standard_resources[] = {
418 	&code_resource,
419 	&data_resource,
420 	&bss_resource,
421 };
422 
423 static void __init setup_resources(void)
424 {
425 	struct resource *res, *std_res, *sub_res;
426 	int i, j;
427 
428 	code_resource.start = (unsigned long) &_text;
429 	code_resource.end = (unsigned long) &_etext - 1;
430 	data_resource.start = (unsigned long) &_etext;
431 	data_resource.end = (unsigned long) &_edata - 1;
432 	bss_resource.start = (unsigned long) &__bss_start;
433 	bss_resource.end = (unsigned long) &__bss_stop - 1;
434 
435 	for (i = 0; i < MEMORY_CHUNKS; i++) {
436 		if (!memory_chunk[i].size)
437 			continue;
438 		if (memory_chunk[i].type == CHUNK_OLDMEM ||
439 		    memory_chunk[i].type == CHUNK_CRASHK)
440 			continue;
441 		res = alloc_bootmem_low(sizeof(*res));
442 		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
443 		switch (memory_chunk[i].type) {
444 		case CHUNK_READ_WRITE:
445 		case CHUNK_CRASHK:
446 			res->name = "System RAM";
447 			break;
448 		case CHUNK_READ_ONLY:
449 			res->name = "System ROM";
450 			res->flags |= IORESOURCE_READONLY;
451 			break;
452 		default:
453 			res->name = "reserved";
454 		}
455 		res->start = memory_chunk[i].addr;
456 		res->end = res->start + memory_chunk[i].size - 1;
457 		request_resource(&iomem_resource, res);
458 
459 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
460 			std_res = standard_resources[j];
461 			if (std_res->start < res->start ||
462 			    std_res->start > res->end)
463 				continue;
464 			if (std_res->end > res->end) {
465 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
466 				*sub_res = *std_res;
467 				sub_res->end = res->end;
468 				std_res->start = res->end + 1;
469 				request_resource(res, sub_res);
470 			} else {
471 				request_resource(res, std_res);
472 			}
473 		}
474 	}
475 }
476 
477 unsigned long real_memory_size;
478 EXPORT_SYMBOL_GPL(real_memory_size);
479 
480 static void __init setup_memory_end(void)
481 {
482 	unsigned long memory_size;
483 	unsigned long max_mem;
484 	int i;
485 
486 
487 #ifdef CONFIG_ZFCPDUMP
488 	if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
489 		memory_end = ZFCPDUMP_HSA_SIZE;
490 		memory_end_set = 1;
491 	}
492 #endif
493 	memory_size = 0;
494 	memory_end &= PAGE_MASK;
495 
496 	max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
497 	memory_end = min(max_mem, memory_end);
498 
499 	/*
500 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
501 	 * extra checks that HOLES_IN_ZONE would require.
502 	 */
503 	for (i = 0; i < MEMORY_CHUNKS; i++) {
504 		unsigned long start, end;
505 		struct mem_chunk *chunk;
506 		unsigned long align;
507 
508 		chunk = &memory_chunk[i];
509 		align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
510 		start = (chunk->addr + align - 1) & ~(align - 1);
511 		end = (chunk->addr + chunk->size) & ~(align - 1);
512 		if (start >= end)
513 			memset(chunk, 0, sizeof(*chunk));
514 		else {
515 			chunk->addr = start;
516 			chunk->size = end - start;
517 		}
518 	}
519 
520 	for (i = 0; i < MEMORY_CHUNKS; i++) {
521 		struct mem_chunk *chunk = &memory_chunk[i];
522 
523 		real_memory_size = max(real_memory_size,
524 				       chunk->addr + chunk->size);
525 		if (chunk->addr >= max_mem) {
526 			memset(chunk, 0, sizeof(*chunk));
527 			continue;
528 		}
529 		if (chunk->addr + chunk->size > max_mem)
530 			chunk->size = max_mem - chunk->addr;
531 		memory_size = max(memory_size, chunk->addr + chunk->size);
532 	}
533 	if (!memory_end)
534 		memory_end = memory_size;
535 }
536 
537 void *restart_stack __attribute__((__section__(".data")));
538 
539 /*
540  * Setup new PSW and allocate stack for PSW restart interrupt
541  */
542 static void __init setup_restart_psw(void)
543 {
544 	psw_t psw;
545 
546 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
547 	restart_stack += ASYNC_SIZE;
548 
549 	/*
550 	 * Setup restart PSW for absolute zero lowcore. This is necesary
551 	 * if PSW restart is done on an offline CPU that has lowcore zero
552 	 */
553 	psw.mask = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_EA | PSW_MASK_BA;
554 	psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
555 	copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
556 }
557 
558 static void __init setup_vmcoreinfo(void)
559 {
560 #ifdef CONFIG_KEXEC
561 	unsigned long ptr = paddr_vmcoreinfo_note();
562 
563 	copy_to_absolute_zero(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
564 #endif
565 }
566 
567 #ifdef CONFIG_CRASH_DUMP
568 
569 /*
570  * Find suitable location for crashkernel memory
571  */
572 static unsigned long __init find_crash_base(unsigned long crash_size,
573 					    char **msg)
574 {
575 	unsigned long crash_base;
576 	struct mem_chunk *chunk;
577 	int i;
578 
579 	if (memory_chunk[0].size < crash_size) {
580 		*msg = "first memory chunk must be at least crashkernel size";
581 		return 0;
582 	}
583 	if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
584 		return OLDMEM_BASE;
585 
586 	for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
587 		chunk = &memory_chunk[i];
588 		if (chunk->size == 0)
589 			continue;
590 		if (chunk->type != CHUNK_READ_WRITE)
591 			continue;
592 		if (chunk->size < crash_size)
593 			continue;
594 		crash_base = (chunk->addr + chunk->size) - crash_size;
595 		if (crash_base < crash_size)
596 			continue;
597 		if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
598 			continue;
599 		if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
600 			continue;
601 		return crash_base;
602 	}
603 	*msg = "no suitable area found";
604 	return 0;
605 }
606 
607 /*
608  * Check if crash_base and crash_size is valid
609  */
610 static int __init verify_crash_base(unsigned long crash_base,
611 				    unsigned long crash_size,
612 				    char **msg)
613 {
614 	struct mem_chunk *chunk;
615 	int i;
616 
617 	/*
618 	 * Because we do the swap to zero, we must have at least 'crash_size'
619 	 * bytes free space before crash_base
620 	 */
621 	if (crash_size > crash_base) {
622 		*msg = "crashkernel offset must be greater than size";
623 		return -EINVAL;
624 	}
625 
626 	/* First memory chunk must be at least crash_size */
627 	if (memory_chunk[0].size < crash_size) {
628 		*msg = "first memory chunk must be at least crashkernel size";
629 		return -EINVAL;
630 	}
631 	/* Check if we fit into the respective memory chunk */
632 	for (i = 0; i < MEMORY_CHUNKS; i++) {
633 		chunk = &memory_chunk[i];
634 		if (chunk->size == 0)
635 			continue;
636 		if (crash_base < chunk->addr)
637 			continue;
638 		if (crash_base >= chunk->addr + chunk->size)
639 			continue;
640 		/* we have found the memory chunk */
641 		if (crash_base + crash_size > chunk->addr + chunk->size) {
642 			*msg = "selected memory chunk is too small for "
643 				"crashkernel memory";
644 			return -EINVAL;
645 		}
646 		return 0;
647 	}
648 	*msg = "invalid memory range specified";
649 	return -EINVAL;
650 }
651 
652 /*
653  * Reserve kdump memory by creating a memory hole in the mem_chunk array
654  */
655 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
656 					 int type)
657 {
658 
659 	create_mem_hole(memory_chunk, addr, size, type);
660 }
661 
662 /*
663  * When kdump is enabled, we have to ensure that no memory from
664  * the area [0 - crashkernel memory size] and
665  * [crashk_res.start - crashk_res.end] is set offline.
666  */
667 static int kdump_mem_notifier(struct notifier_block *nb,
668 			      unsigned long action, void *data)
669 {
670 	struct memory_notify *arg = data;
671 
672 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
673 		return NOTIFY_BAD;
674 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
675 		return NOTIFY_OK;
676 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
677 		return NOTIFY_OK;
678 	return NOTIFY_BAD;
679 }
680 
681 static struct notifier_block kdump_mem_nb = {
682 	.notifier_call = kdump_mem_notifier,
683 };
684 
685 #endif
686 
687 /*
688  * Make sure that oldmem, where the dump is stored, is protected
689  */
690 static void reserve_oldmem(void)
691 {
692 #ifdef CONFIG_CRASH_DUMP
693 	if (!OLDMEM_BASE)
694 		return;
695 
696 	reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
697 	reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
698 			      CHUNK_OLDMEM);
699 	if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
700 		saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
701 	else
702 		saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
703 #endif
704 }
705 
706 /*
707  * Reserve memory for kdump kernel to be loaded with kexec
708  */
709 static void __init reserve_crashkernel(void)
710 {
711 #ifdef CONFIG_CRASH_DUMP
712 	unsigned long long crash_base, crash_size;
713 	char *msg;
714 	int rc;
715 
716 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
717 			       &crash_base);
718 	if (rc || crash_size == 0)
719 		return;
720 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
721 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
722 	if (register_memory_notifier(&kdump_mem_nb))
723 		return;
724 	if (!crash_base)
725 		crash_base = find_crash_base(crash_size, &msg);
726 	if (!crash_base) {
727 		pr_info("crashkernel reservation failed: %s\n", msg);
728 		unregister_memory_notifier(&kdump_mem_nb);
729 		return;
730 	}
731 	if (verify_crash_base(crash_base, crash_size, &msg)) {
732 		pr_info("crashkernel reservation failed: %s\n", msg);
733 		unregister_memory_notifier(&kdump_mem_nb);
734 		return;
735 	}
736 	if (!OLDMEM_BASE && MACHINE_IS_VM)
737 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
738 	crashk_res.start = crash_base;
739 	crashk_res.end = crash_base + crash_size - 1;
740 	insert_resource(&iomem_resource, &crashk_res);
741 	reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
742 	pr_info("Reserving %lluMB of memory at %lluMB "
743 		"for crashkernel (System RAM: %luMB)\n",
744 		crash_size >> 20, crash_base >> 20, memory_end >> 20);
745 #endif
746 }
747 
748 static void __init
749 setup_memory(void)
750 {
751         unsigned long bootmap_size;
752 	unsigned long start_pfn, end_pfn;
753 	int i;
754 
755 	/*
756 	 * partially used pages are not usable - thus
757 	 * we are rounding upwards:
758 	 */
759 	start_pfn = PFN_UP(__pa(&_end));
760 	end_pfn = max_pfn = PFN_DOWN(memory_end);
761 
762 #ifdef CONFIG_BLK_DEV_INITRD
763 	/*
764 	 * Move the initrd in case the bitmap of the bootmem allocater
765 	 * would overwrite it.
766 	 */
767 
768 	if (INITRD_START && INITRD_SIZE) {
769 		unsigned long bmap_size;
770 		unsigned long start;
771 
772 		bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
773 		bmap_size = PFN_PHYS(bmap_size);
774 
775 		if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
776 			start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
777 
778 #ifdef CONFIG_CRASH_DUMP
779 			if (OLDMEM_BASE) {
780 				/* Move initrd behind kdump oldmem */
781 				if (start + INITRD_SIZE > OLDMEM_BASE &&
782 				    start < OLDMEM_BASE + OLDMEM_SIZE)
783 					start = OLDMEM_BASE + OLDMEM_SIZE;
784 			}
785 #endif
786 			if (start + INITRD_SIZE > memory_end) {
787 				pr_err("initrd extends beyond end of "
788 				       "memory (0x%08lx > 0x%08lx) "
789 				       "disabling initrd\n",
790 				       start + INITRD_SIZE, memory_end);
791 				INITRD_START = INITRD_SIZE = 0;
792 			} else {
793 				pr_info("Moving initrd (0x%08lx -> "
794 					"0x%08lx, size: %ld)\n",
795 					INITRD_START, start, INITRD_SIZE);
796 				memmove((void *) start, (void *) INITRD_START,
797 					INITRD_SIZE);
798 				INITRD_START = start;
799 			}
800 		}
801 	}
802 #endif
803 
804 	/*
805 	 * Initialize the boot-time allocator
806 	 */
807 	bootmap_size = init_bootmem(start_pfn, end_pfn);
808 
809 	/*
810 	 * Register RAM areas with the bootmem allocator.
811 	 */
812 
813 	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
814 		unsigned long start_chunk, end_chunk, pfn;
815 
816 		if (memory_chunk[i].type != CHUNK_READ_WRITE &&
817 		    memory_chunk[i].type != CHUNK_CRASHK)
818 			continue;
819 		start_chunk = PFN_DOWN(memory_chunk[i].addr);
820 		end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
821 		end_chunk = min(end_chunk, end_pfn);
822 		if (start_chunk >= end_chunk)
823 			continue;
824 		memblock_add_node(PFN_PHYS(start_chunk),
825 				  PFN_PHYS(end_chunk - start_chunk), 0);
826 		pfn = max(start_chunk, start_pfn);
827 		for (; pfn < end_chunk; pfn++)
828 			page_set_storage_key(PFN_PHYS(pfn),
829 					     PAGE_DEFAULT_KEY, 0);
830 	}
831 
832 	psw_set_key(PAGE_DEFAULT_KEY);
833 
834 	free_bootmem_with_active_regions(0, max_pfn);
835 
836 	/*
837 	 * Reserve memory used for lowcore/command line/kernel image.
838 	 */
839 	reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
840 	reserve_bootmem((unsigned long)_stext,
841 			PFN_PHYS(start_pfn) - (unsigned long)_stext,
842 			BOOTMEM_DEFAULT);
843 	/*
844 	 * Reserve the bootmem bitmap itself as well. We do this in two
845 	 * steps (first step was init_bootmem()) because this catches
846 	 * the (very unlikely) case of us accidentally initializing the
847 	 * bootmem allocator with an invalid RAM area.
848 	 */
849 	reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
850 			BOOTMEM_DEFAULT);
851 
852 #ifdef CONFIG_CRASH_DUMP
853 	if (crashk_res.start)
854 		reserve_bootmem(crashk_res.start,
855 				crashk_res.end - crashk_res.start + 1,
856 				BOOTMEM_DEFAULT);
857 	if (is_kdump_kernel())
858 		reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
859 				PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
860 #endif
861 #ifdef CONFIG_BLK_DEV_INITRD
862 	if (INITRD_START && INITRD_SIZE) {
863 		if (INITRD_START + INITRD_SIZE <= memory_end) {
864 			reserve_bootmem(INITRD_START, INITRD_SIZE,
865 					BOOTMEM_DEFAULT);
866 			initrd_start = INITRD_START;
867 			initrd_end = initrd_start + INITRD_SIZE;
868 		} else {
869 			pr_err("initrd extends beyond end of "
870 			       "memory (0x%08lx > 0x%08lx) "
871 			       "disabling initrd\n",
872 			       initrd_start + INITRD_SIZE, memory_end);
873 			initrd_start = initrd_end = 0;
874 		}
875 	}
876 #endif
877 }
878 
879 /*
880  * Setup hardware capabilities.
881  */
882 static void __init setup_hwcaps(void)
883 {
884 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
885 	struct cpuid cpu_id;
886 	int i;
887 
888 	/*
889 	 * The store facility list bits numbers as found in the principles
890 	 * of operation are numbered with bit 1UL<<31 as number 0 to
891 	 * bit 1UL<<0 as number 31.
892 	 *   Bit 0: instructions named N3, "backported" to esa-mode
893 	 *   Bit 2: z/Architecture mode is active
894 	 *   Bit 7: the store-facility-list-extended facility is installed
895 	 *   Bit 17: the message-security assist is installed
896 	 *   Bit 19: the long-displacement facility is installed
897 	 *   Bit 21: the extended-immediate facility is installed
898 	 *   Bit 22: extended-translation facility 3 is installed
899 	 *   Bit 30: extended-translation facility 3 enhancement facility
900 	 * These get translated to:
901 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
902 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
903 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
904 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
905 	 */
906 	for (i = 0; i < 6; i++)
907 		if (test_facility(stfl_bits[i]))
908 			elf_hwcap |= 1UL << i;
909 
910 	if (test_facility(22) && test_facility(30))
911 		elf_hwcap |= HWCAP_S390_ETF3EH;
912 
913 	/*
914 	 * Check for additional facilities with store-facility-list-extended.
915 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
916 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
917 	 * as stored by stfl, bits 32-xxx contain additional facilities.
918 	 * How many facility words are stored depends on the number of
919 	 * doublewords passed to the instruction. The additional facilities
920 	 * are:
921 	 *   Bit 42: decimal floating point facility is installed
922 	 *   Bit 44: perform floating point operation facility is installed
923 	 * translated to:
924 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
925 	 */
926 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
927 		elf_hwcap |= HWCAP_S390_DFP;
928 
929 	/*
930 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
931 	 */
932 	if (MACHINE_HAS_HPAGE)
933 		elf_hwcap |= HWCAP_S390_HPAGE;
934 
935 	/*
936 	 * 64-bit register support for 31-bit processes
937 	 * HWCAP_S390_HIGH_GPRS is bit 9.
938 	 */
939 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
940 
941 	get_cpu_id(&cpu_id);
942 	switch (cpu_id.machine) {
943 	case 0x9672:
944 #if !defined(CONFIG_64BIT)
945 	default:	/* Use "g5" as default for 31 bit kernels. */
946 #endif
947 		strcpy(elf_platform, "g5");
948 		break;
949 	case 0x2064:
950 	case 0x2066:
951 #if defined(CONFIG_64BIT)
952 	default:	/* Use "z900" as default for 64 bit kernels. */
953 #endif
954 		strcpy(elf_platform, "z900");
955 		break;
956 	case 0x2084:
957 	case 0x2086:
958 		strcpy(elf_platform, "z990");
959 		break;
960 	case 0x2094:
961 	case 0x2096:
962 		strcpy(elf_platform, "z9-109");
963 		break;
964 	case 0x2097:
965 	case 0x2098:
966 		strcpy(elf_platform, "z10");
967 		break;
968 	case 0x2817:
969 	case 0x2818:
970 		strcpy(elf_platform, "z196");
971 		break;
972 	}
973 }
974 
975 /*
976  * Setup function called from init/main.c just after the banner
977  * was printed.
978  */
979 
980 void __init
981 setup_arch(char **cmdline_p)
982 {
983         /*
984          * print what head.S has found out about the machine
985          */
986 #ifndef CONFIG_64BIT
987 	if (MACHINE_IS_VM)
988 		pr_info("Linux is running as a z/VM "
989 			"guest operating system in 31-bit mode\n");
990 	else if (MACHINE_IS_LPAR)
991 		pr_info("Linux is running natively in 31-bit mode\n");
992 	if (MACHINE_HAS_IEEE)
993 		pr_info("The hardware system has IEEE compatible "
994 			"floating point units\n");
995 	else
996 		pr_info("The hardware system has no IEEE compatible "
997 			"floating point units\n");
998 #else /* CONFIG_64BIT */
999 	if (MACHINE_IS_VM)
1000 		pr_info("Linux is running as a z/VM "
1001 			"guest operating system in 64-bit mode\n");
1002 	else if (MACHINE_IS_KVM)
1003 		pr_info("Linux is running under KVM in 64-bit mode\n");
1004 	else if (MACHINE_IS_LPAR)
1005 		pr_info("Linux is running natively in 64-bit mode\n");
1006 #endif /* CONFIG_64BIT */
1007 
1008 	/* Have one command line that is parsed and saved in /proc/cmdline */
1009 	/* boot_command_line has been already set up in early.c */
1010 	*cmdline_p = boot_command_line;
1011 
1012         ROOT_DEV = Root_RAM0;
1013 
1014 	init_mm.start_code = PAGE_OFFSET;
1015 	init_mm.end_code = (unsigned long) &_etext;
1016 	init_mm.end_data = (unsigned long) &_edata;
1017 	init_mm.brk = (unsigned long) &_end;
1018 
1019 	if (MACHINE_HAS_MVCOS)
1020 		memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1021 	else
1022 		memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1023 
1024 	parse_early_param();
1025 
1026 	setup_ipl();
1027 	setup_memory_end();
1028 	setup_addressing_mode();
1029 	reserve_oldmem();
1030 	reserve_crashkernel();
1031 	setup_memory();
1032 	setup_resources();
1033 	setup_vmcoreinfo();
1034 	setup_restart_psw();
1035 	setup_lowcore();
1036 
1037         cpu_init();
1038 	s390_init_cpu_topology();
1039 
1040 	/*
1041 	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1042 	 */
1043 	setup_hwcaps();
1044 
1045 	/*
1046 	 * Create kernel page tables and switch to virtual addressing.
1047 	 */
1048         paging_init();
1049 
1050         /* Setup default console */
1051 	conmode_default();
1052 	set_preferred_console();
1053 
1054 	/* Setup zfcpdump support */
1055 	setup_zfcpdump(console_devno);
1056 }
1057