xref: /linux/arch/s390/kernel/setup.c (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
1 /*
2  *  arch/s390/kernel/setup.c
3  *
4  *  S390 version
5  *    Copyright (C) IBM Corp. 1999,2012
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 #include <linux/compat.h>
50 
51 #include <asm/ipl.h>
52 #include <asm/uaccess.h>
53 #include <asm/facility.h>
54 #include <asm/smp.h>
55 #include <asm/mmu_context.h>
56 #include <asm/cpcmd.h>
57 #include <asm/lowcore.h>
58 #include <asm/irq.h>
59 #include <asm/page.h>
60 #include <asm/ptrace.h>
61 #include <asm/sections.h>
62 #include <asm/ebcdic.h>
63 #include <asm/kvm_virtio.h>
64 #include <asm/diag.h>
65 #include <asm/os_info.h>
66 #include "entry.h"
67 
68 long psw_kernel_bits	= PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
69 			  PSW_MASK_EA | PSW_MASK_BA;
70 long psw_user_bits	= PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
71 			  PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
72 			  PSW_MASK_PSTATE | PSW_ASC_HOME;
73 
74 /*
75  * User copy operations.
76  */
77 struct uaccess_ops uaccess;
78 EXPORT_SYMBOL(uaccess);
79 
80 /*
81  * Machine setup..
82  */
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
85 
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
88 
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
91 
92 unsigned long elf_hwcap = 0;
93 char elf_platform[ELF_PLATFORM_SIZE];
94 
95 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
96 
97 int __initdata memory_end_set;
98 unsigned long __initdata memory_end;
99 
100 unsigned long VMALLOC_START;
101 EXPORT_SYMBOL(VMALLOC_START);
102 
103 unsigned long VMALLOC_END;
104 EXPORT_SYMBOL(VMALLOC_END);
105 
106 struct page *vmemmap;
107 EXPORT_SYMBOL(vmemmap);
108 
109 /* An array with a pointer to the lowcore of every CPU. */
110 struct _lowcore *lowcore_ptr[NR_CPUS];
111 EXPORT_SYMBOL(lowcore_ptr);
112 
113 /*
114  * This is set up by the setup-routine at boot-time
115  * for S390 need to find out, what we have to setup
116  * using address 0x10400 ...
117  */
118 
119 #include <asm/setup.h>
120 
121 /*
122  * condev= and conmode= setup parameter.
123  */
124 
125 static int __init condev_setup(char *str)
126 {
127 	int vdev;
128 
129 	vdev = simple_strtoul(str, &str, 0);
130 	if (vdev >= 0 && vdev < 65536) {
131 		console_devno = vdev;
132 		console_irq = -1;
133 	}
134 	return 1;
135 }
136 
137 __setup("condev=", condev_setup);
138 
139 static void __init set_preferred_console(void)
140 {
141 	if (MACHINE_IS_KVM)
142 		add_preferred_console("hvc", 0, NULL);
143 	else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
144 		add_preferred_console("ttyS", 0, NULL);
145 	else if (CONSOLE_IS_3270)
146 		add_preferred_console("tty3270", 0, NULL);
147 }
148 
149 static int __init conmode_setup(char *str)
150 {
151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
152 	if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
153                 SET_CONSOLE_SCLP;
154 #endif
155 #if defined(CONFIG_TN3215_CONSOLE)
156 	if (strncmp(str, "3215", 5) == 0)
157 		SET_CONSOLE_3215;
158 #endif
159 #if defined(CONFIG_TN3270_CONSOLE)
160 	if (strncmp(str, "3270", 5) == 0)
161 		SET_CONSOLE_3270;
162 #endif
163 	set_preferred_console();
164         return 1;
165 }
166 
167 __setup("conmode=", conmode_setup);
168 
169 static void __init conmode_default(void)
170 {
171 	char query_buffer[1024];
172 	char *ptr;
173 
174         if (MACHINE_IS_VM) {
175 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
176 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
177 		ptr = strstr(query_buffer, "SUBCHANNEL =");
178 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
179 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
180 		ptr = strstr(query_buffer, "CONMODE");
181 		/*
182 		 * Set the conmode to 3215 so that the device recognition
183 		 * will set the cu_type of the console to 3215. If the
184 		 * conmode is 3270 and we don't set it back then both
185 		 * 3215 and the 3270 driver will try to access the console
186 		 * device (3215 as console and 3270 as normal tty).
187 		 */
188 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
189 		if (ptr == NULL) {
190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
191 			SET_CONSOLE_SCLP;
192 #endif
193 			return;
194 		}
195 		if (strncmp(ptr + 8, "3270", 4) == 0) {
196 #if defined(CONFIG_TN3270_CONSOLE)
197 			SET_CONSOLE_3270;
198 #elif defined(CONFIG_TN3215_CONSOLE)
199 			SET_CONSOLE_3215;
200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
201 			SET_CONSOLE_SCLP;
202 #endif
203 		} else if (strncmp(ptr + 8, "3215", 4) == 0) {
204 #if defined(CONFIG_TN3215_CONSOLE)
205 			SET_CONSOLE_3215;
206 #elif defined(CONFIG_TN3270_CONSOLE)
207 			SET_CONSOLE_3270;
208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
209 			SET_CONSOLE_SCLP;
210 #endif
211 		}
212 	} else {
213 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
214 		SET_CONSOLE_SCLP;
215 #endif
216 	}
217 }
218 
219 #ifdef CONFIG_ZFCPDUMP
220 static void __init setup_zfcpdump(unsigned int console_devno)
221 {
222 	static char str[41];
223 
224 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
225 		return;
226 	if (OLDMEM_BASE)
227 		return;
228 	if (console_devno != -1)
229 		sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
230 			ipl_info.data.fcp.dev_id.devno, console_devno);
231 	else
232 		sprintf(str, " cio_ignore=all,!0.0.%04x",
233 			ipl_info.data.fcp.dev_id.devno);
234 	strcat(boot_command_line, str);
235 	console_loglevel = 2;
236 }
237 #else
238 static inline void setup_zfcpdump(unsigned int console_devno) {}
239 #endif /* CONFIG_ZFCPDUMP */
240 
241  /*
242  * Reboot, halt and power_off stubs. They just call _machine_restart,
243  * _machine_halt or _machine_power_off.
244  */
245 
246 void machine_restart(char *command)
247 {
248 	if ((!in_interrupt() && !in_atomic()) || 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_restart(command);
255 }
256 
257 void machine_halt(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_halt();
266 }
267 
268 void machine_power_off(void)
269 {
270 	if (!in_interrupt() || oops_in_progress)
271 		/*
272 		 * Only unblank the console if we are called in enabled
273 		 * context or a bust_spinlocks cleared the way for us.
274 		 */
275 		console_unblank();
276 	_machine_power_off();
277 }
278 
279 /*
280  * Dummy power off function.
281  */
282 void (*pm_power_off)(void) = machine_power_off;
283 
284 static int __init early_parse_mem(char *p)
285 {
286 	memory_end = memparse(p, &p);
287 	memory_end_set = 1;
288 	return 0;
289 }
290 early_param("mem", early_parse_mem);
291 
292 static int __init parse_vmalloc(char *arg)
293 {
294 	if (!arg)
295 		return -EINVAL;
296 	VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
297 	return 0;
298 }
299 early_param("vmalloc", parse_vmalloc);
300 
301 unsigned int user_mode = HOME_SPACE_MODE;
302 EXPORT_SYMBOL_GPL(user_mode);
303 
304 static int set_amode_primary(void)
305 {
306 	psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
307 	psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
308 #ifdef CONFIG_COMPAT
309 	psw32_user_bits =
310 		(psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
311 #endif
312 
313 	if (MACHINE_HAS_MVCOS) {
314 		memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
315 		return 1;
316 	} else {
317 		memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
318 		return 0;
319 	}
320 }
321 
322 /*
323  * Switch kernel/user addressing modes?
324  */
325 static int __init early_parse_switch_amode(char *p)
326 {
327 	user_mode = PRIMARY_SPACE_MODE;
328 	return 0;
329 }
330 early_param("switch_amode", early_parse_switch_amode);
331 
332 static int __init early_parse_user_mode(char *p)
333 {
334 	if (p && strcmp(p, "primary") == 0)
335 		user_mode = PRIMARY_SPACE_MODE;
336 	else if (!p || strcmp(p, "home") == 0)
337 		user_mode = HOME_SPACE_MODE;
338 	else
339 		return 1;
340 	return 0;
341 }
342 early_param("user_mode", early_parse_user_mode);
343 
344 static void setup_addressing_mode(void)
345 {
346 	if (user_mode == PRIMARY_SPACE_MODE) {
347 		if (set_amode_primary())
348 			pr_info("Address spaces switched, "
349 				"mvcos available\n");
350 		else
351 			pr_info("Address spaces switched, "
352 				"mvcos not available\n");
353 	}
354 }
355 
356 void *restart_stack __attribute__((__section__(".data")));
357 
358 static void __init setup_lowcore(void)
359 {
360 	struct _lowcore *lc;
361 
362 	/*
363 	 * Setup lowcore for boot cpu
364 	 */
365 	BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
366 	lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
367 	lc->restart_psw.mask = psw_kernel_bits;
368 	lc->restart_psw.addr =
369 		PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
370 	lc->external_new_psw.mask = psw_kernel_bits |
371 		PSW_MASK_DAT | PSW_MASK_MCHECK;
372 	lc->external_new_psw.addr =
373 		PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
374 	lc->svc_new_psw.mask = psw_kernel_bits |
375 		PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
376 	lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
377 	lc->program_new_psw.mask = psw_kernel_bits |
378 		PSW_MASK_DAT | PSW_MASK_MCHECK;
379 	lc->program_new_psw.addr =
380 		PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
381 	lc->mcck_new_psw.mask = psw_kernel_bits;
382 	lc->mcck_new_psw.addr =
383 		PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
384 	lc->io_new_psw.mask = psw_kernel_bits |
385 		PSW_MASK_DAT | PSW_MASK_MCHECK;
386 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
387 	lc->clock_comparator = -1ULL;
388 	lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
389 	lc->async_stack = (unsigned long)
390 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
391 	lc->panic_stack = (unsigned long)
392 		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
393 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
394 	lc->thread_info = (unsigned long) &init_thread_union;
395 	lc->machine_flags = S390_lowcore.machine_flags;
396 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
397 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
398 	       MAX_FACILITY_BIT/8);
399 #ifndef CONFIG_64BIT
400 	if (MACHINE_HAS_IEEE) {
401 		lc->extended_save_area_addr = (__u32)
402 			__alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
403 		/* enable extended save area */
404 		__ctl_set_bit(14, 29);
405 	}
406 #else
407 	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
408 #endif
409 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
410 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
411 	lc->exit_timer = S390_lowcore.exit_timer;
412 	lc->user_timer = S390_lowcore.user_timer;
413 	lc->system_timer = S390_lowcore.system_timer;
414 	lc->steal_timer = S390_lowcore.steal_timer;
415 	lc->last_update_timer = S390_lowcore.last_update_timer;
416 	lc->last_update_clock = S390_lowcore.last_update_clock;
417 	lc->ftrace_func = S390_lowcore.ftrace_func;
418 
419 	restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
420 	restart_stack += ASYNC_SIZE;
421 
422 	/*
423 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
424 	 * restart data to the absolute zero lowcore. This is necesary if
425 	 * PSW restart is done on an offline CPU that has lowcore zero.
426 	 */
427 	lc->restart_stack = (unsigned long) restart_stack;
428 	lc->restart_fn = (unsigned long) do_restart;
429 	lc->restart_data = 0;
430 	lc->restart_source = -1UL;
431 
432 	/* Setup absolute zero lowcore */
433 	memcpy_absolute(&S390_lowcore.restart_stack, &lc->restart_stack,
434 			4 * sizeof(unsigned long));
435 	memcpy_absolute(&S390_lowcore.restart_psw, &lc->restart_psw,
436 			sizeof(lc->restart_psw));
437 
438 	set_prefix((u32)(unsigned long) lc);
439 	lowcore_ptr[0] = lc;
440 }
441 
442 static struct resource code_resource = {
443 	.name  = "Kernel code",
444 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
445 };
446 
447 static struct resource data_resource = {
448 	.name = "Kernel data",
449 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
450 };
451 
452 static struct resource bss_resource = {
453 	.name = "Kernel bss",
454 	.flags = IORESOURCE_BUSY | IORESOURCE_MEM,
455 };
456 
457 static struct resource __initdata *standard_resources[] = {
458 	&code_resource,
459 	&data_resource,
460 	&bss_resource,
461 };
462 
463 static void __init setup_resources(void)
464 {
465 	struct resource *res, *std_res, *sub_res;
466 	int i, j;
467 
468 	code_resource.start = (unsigned long) &_text;
469 	code_resource.end = (unsigned long) &_etext - 1;
470 	data_resource.start = (unsigned long) &_etext;
471 	data_resource.end = (unsigned long) &_edata - 1;
472 	bss_resource.start = (unsigned long) &__bss_start;
473 	bss_resource.end = (unsigned long) &__bss_stop - 1;
474 
475 	for (i = 0; i < MEMORY_CHUNKS; i++) {
476 		if (!memory_chunk[i].size)
477 			continue;
478 		if (memory_chunk[i].type == CHUNK_OLDMEM ||
479 		    memory_chunk[i].type == CHUNK_CRASHK)
480 			continue;
481 		res = alloc_bootmem_low(sizeof(*res));
482 		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
483 		switch (memory_chunk[i].type) {
484 		case CHUNK_READ_WRITE:
485 		case CHUNK_CRASHK:
486 			res->name = "System RAM";
487 			break;
488 		case CHUNK_READ_ONLY:
489 			res->name = "System ROM";
490 			res->flags |= IORESOURCE_READONLY;
491 			break;
492 		default:
493 			res->name = "reserved";
494 		}
495 		res->start = memory_chunk[i].addr;
496 		res->end = res->start + memory_chunk[i].size - 1;
497 		request_resource(&iomem_resource, res);
498 
499 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
500 			std_res = standard_resources[j];
501 			if (std_res->start < res->start ||
502 			    std_res->start > res->end)
503 				continue;
504 			if (std_res->end > res->end) {
505 				sub_res = alloc_bootmem_low(sizeof(*sub_res));
506 				*sub_res = *std_res;
507 				sub_res->end = res->end;
508 				std_res->start = res->end + 1;
509 				request_resource(res, sub_res);
510 			} else {
511 				request_resource(res, std_res);
512 			}
513 		}
514 	}
515 }
516 
517 unsigned long real_memory_size;
518 EXPORT_SYMBOL_GPL(real_memory_size);
519 
520 static void __init setup_memory_end(void)
521 {
522 	unsigned long vmax, vmalloc_size, tmp;
523 	int i;
524 
525 
526 #ifdef CONFIG_ZFCPDUMP
527 	if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
528 		memory_end = ZFCPDUMP_HSA_SIZE;
529 		memory_end_set = 1;
530 	}
531 #endif
532 	real_memory_size = 0;
533 	memory_end &= PAGE_MASK;
534 
535 	/*
536 	 * Make sure all chunks are MAX_ORDER aligned so we don't need the
537 	 * extra checks that HOLES_IN_ZONE would require.
538 	 */
539 	for (i = 0; i < MEMORY_CHUNKS; i++) {
540 		unsigned long start, end;
541 		struct mem_chunk *chunk;
542 		unsigned long align;
543 
544 		chunk = &memory_chunk[i];
545 		align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
546 		start = (chunk->addr + align - 1) & ~(align - 1);
547 		end = (chunk->addr + chunk->size) & ~(align - 1);
548 		if (start >= end)
549 			memset(chunk, 0, sizeof(*chunk));
550 		else {
551 			chunk->addr = start;
552 			chunk->size = end - start;
553 		}
554 		real_memory_size = max(real_memory_size,
555 				       chunk->addr + chunk->size);
556 	}
557 
558 	/* Choose kernel address space layout: 2, 3, or 4 levels. */
559 #ifdef CONFIG_64BIT
560 	vmalloc_size = VMALLOC_END ?: 128UL << 30;
561 	tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
562 	tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
563 	if (tmp <= (1UL << 42))
564 		vmax = 1UL << 42;	/* 3-level kernel page table */
565 	else
566 		vmax = 1UL << 53;	/* 4-level kernel page table */
567 #else
568 	vmalloc_size = VMALLOC_END ?: 96UL << 20;
569 	vmax = 1UL << 31;		/* 2-level kernel page table */
570 #endif
571 	/* vmalloc area is at the end of the kernel address space. */
572 	VMALLOC_END = vmax;
573 	VMALLOC_START = vmax - vmalloc_size;
574 
575 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
576 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
577 	tmp = VMALLOC_START - tmp * sizeof(struct page);
578 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
579 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
580 	vmemmap = (struct page *) tmp;
581 
582 	/* Take care that memory_end is set and <= vmemmap */
583 	memory_end = min(memory_end ?: real_memory_size, tmp);
584 
585 	/* Fixup memory chunk array to fit into 0..memory_end */
586 	for (i = 0; i < MEMORY_CHUNKS; i++) {
587 		struct mem_chunk *chunk = &memory_chunk[i];
588 
589 		if (chunk->addr >= memory_end) {
590 			memset(chunk, 0, sizeof(*chunk));
591 			continue;
592 		}
593 		if (chunk->addr + chunk->size > memory_end)
594 			chunk->size = memory_end - chunk->addr;
595 	}
596 }
597 
598 static void __init setup_vmcoreinfo(void)
599 {
600 #ifdef CONFIG_KEXEC
601 	unsigned long ptr = paddr_vmcoreinfo_note();
602 
603 	memcpy_absolute(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
604 #endif
605 }
606 
607 #ifdef CONFIG_CRASH_DUMP
608 
609 /*
610  * Find suitable location for crashkernel memory
611  */
612 static unsigned long __init find_crash_base(unsigned long crash_size,
613 					    char **msg)
614 {
615 	unsigned long crash_base;
616 	struct mem_chunk *chunk;
617 	int i;
618 
619 	if (memory_chunk[0].size < crash_size) {
620 		*msg = "first memory chunk must be at least crashkernel size";
621 		return 0;
622 	}
623 	if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
624 		return OLDMEM_BASE;
625 
626 	for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
627 		chunk = &memory_chunk[i];
628 		if (chunk->size == 0)
629 			continue;
630 		if (chunk->type != CHUNK_READ_WRITE)
631 			continue;
632 		if (chunk->size < crash_size)
633 			continue;
634 		crash_base = (chunk->addr + chunk->size) - crash_size;
635 		if (crash_base < crash_size)
636 			continue;
637 		if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
638 			continue;
639 		if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
640 			continue;
641 		return crash_base;
642 	}
643 	*msg = "no suitable area found";
644 	return 0;
645 }
646 
647 /*
648  * Check if crash_base and crash_size is valid
649  */
650 static int __init verify_crash_base(unsigned long crash_base,
651 				    unsigned long crash_size,
652 				    char **msg)
653 {
654 	struct mem_chunk *chunk;
655 	int i;
656 
657 	/*
658 	 * Because we do the swap to zero, we must have at least 'crash_size'
659 	 * bytes free space before crash_base
660 	 */
661 	if (crash_size > crash_base) {
662 		*msg = "crashkernel offset must be greater than size";
663 		return -EINVAL;
664 	}
665 
666 	/* First memory chunk must be at least crash_size */
667 	if (memory_chunk[0].size < crash_size) {
668 		*msg = "first memory chunk must be at least crashkernel size";
669 		return -EINVAL;
670 	}
671 	/* Check if we fit into the respective memory chunk */
672 	for (i = 0; i < MEMORY_CHUNKS; i++) {
673 		chunk = &memory_chunk[i];
674 		if (chunk->size == 0)
675 			continue;
676 		if (crash_base < chunk->addr)
677 			continue;
678 		if (crash_base >= chunk->addr + chunk->size)
679 			continue;
680 		/* we have found the memory chunk */
681 		if (crash_base + crash_size > chunk->addr + chunk->size) {
682 			*msg = "selected memory chunk is too small for "
683 				"crashkernel memory";
684 			return -EINVAL;
685 		}
686 		return 0;
687 	}
688 	*msg = "invalid memory range specified";
689 	return -EINVAL;
690 }
691 
692 /*
693  * Reserve kdump memory by creating a memory hole in the mem_chunk array
694  */
695 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
696 					 int type)
697 {
698 	create_mem_hole(memory_chunk, addr, size, type);
699 }
700 
701 /*
702  * When kdump is enabled, we have to ensure that no memory from
703  * the area [0 - crashkernel memory size] and
704  * [crashk_res.start - crashk_res.end] is set offline.
705  */
706 static int kdump_mem_notifier(struct notifier_block *nb,
707 			      unsigned long action, void *data)
708 {
709 	struct memory_notify *arg = data;
710 
711 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
712 		return NOTIFY_BAD;
713 	if (arg->start_pfn > PFN_DOWN(crashk_res.end))
714 		return NOTIFY_OK;
715 	if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
716 		return NOTIFY_OK;
717 	return NOTIFY_BAD;
718 }
719 
720 static struct notifier_block kdump_mem_nb = {
721 	.notifier_call = kdump_mem_notifier,
722 };
723 
724 #endif
725 
726 /*
727  * Make sure that oldmem, where the dump is stored, is protected
728  */
729 static void reserve_oldmem(void)
730 {
731 #ifdef CONFIG_CRASH_DUMP
732 	if (!OLDMEM_BASE)
733 		return;
734 
735 	reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
736 	reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
737 			      CHUNK_OLDMEM);
738 	if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
739 		saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
740 	else
741 		saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
742 #endif
743 }
744 
745 /*
746  * Reserve memory for kdump kernel to be loaded with kexec
747  */
748 static void __init reserve_crashkernel(void)
749 {
750 #ifdef CONFIG_CRASH_DUMP
751 	unsigned long long crash_base, crash_size;
752 	char *msg = NULL;
753 	int rc;
754 
755 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
756 			       &crash_base);
757 	if (rc || crash_size == 0)
758 		return;
759 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
760 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
761 	if (register_memory_notifier(&kdump_mem_nb))
762 		return;
763 	if (!crash_base)
764 		crash_base = find_crash_base(crash_size, &msg);
765 	if (!crash_base) {
766 		pr_info("crashkernel reservation failed: %s\n", msg);
767 		unregister_memory_notifier(&kdump_mem_nb);
768 		return;
769 	}
770 	if (verify_crash_base(crash_base, crash_size, &msg)) {
771 		pr_info("crashkernel reservation failed: %s\n", msg);
772 		unregister_memory_notifier(&kdump_mem_nb);
773 		return;
774 	}
775 	if (!OLDMEM_BASE && MACHINE_IS_VM)
776 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
777 	crashk_res.start = crash_base;
778 	crashk_res.end = crash_base + crash_size - 1;
779 	insert_resource(&iomem_resource, &crashk_res);
780 	reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
781 	pr_info("Reserving %lluMB of memory at %lluMB "
782 		"for crashkernel (System RAM: %luMB)\n",
783 		crash_size >> 20, crash_base >> 20, memory_end >> 20);
784 	os_info_crashkernel_add(crash_base, crash_size);
785 #endif
786 }
787 
788 static void __init setup_memory(void)
789 {
790         unsigned long bootmap_size;
791 	unsigned long start_pfn, end_pfn;
792 	int i;
793 
794 	/*
795 	 * partially used pages are not usable - thus
796 	 * we are rounding upwards:
797 	 */
798 	start_pfn = PFN_UP(__pa(&_end));
799 	end_pfn = max_pfn = PFN_DOWN(memory_end);
800 
801 #ifdef CONFIG_BLK_DEV_INITRD
802 	/*
803 	 * Move the initrd in case the bitmap of the bootmem allocater
804 	 * would overwrite it.
805 	 */
806 
807 	if (INITRD_START && INITRD_SIZE) {
808 		unsigned long bmap_size;
809 		unsigned long start;
810 
811 		bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
812 		bmap_size = PFN_PHYS(bmap_size);
813 
814 		if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
815 			start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
816 
817 #ifdef CONFIG_CRASH_DUMP
818 			if (OLDMEM_BASE) {
819 				/* Move initrd behind kdump oldmem */
820 				if (start + INITRD_SIZE > OLDMEM_BASE &&
821 				    start < OLDMEM_BASE + OLDMEM_SIZE)
822 					start = OLDMEM_BASE + OLDMEM_SIZE;
823 			}
824 #endif
825 			if (start + INITRD_SIZE > memory_end) {
826 				pr_err("initrd extends beyond end of "
827 				       "memory (0x%08lx > 0x%08lx) "
828 				       "disabling initrd\n",
829 				       start + INITRD_SIZE, memory_end);
830 				INITRD_START = INITRD_SIZE = 0;
831 			} else {
832 				pr_info("Moving initrd (0x%08lx -> "
833 					"0x%08lx, size: %ld)\n",
834 					INITRD_START, start, INITRD_SIZE);
835 				memmove((void *) start, (void *) INITRD_START,
836 					INITRD_SIZE);
837 				INITRD_START = start;
838 			}
839 		}
840 	}
841 #endif
842 
843 	/*
844 	 * Initialize the boot-time allocator
845 	 */
846 	bootmap_size = init_bootmem(start_pfn, end_pfn);
847 
848 	/*
849 	 * Register RAM areas with the bootmem allocator.
850 	 */
851 
852 	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
853 		unsigned long start_chunk, end_chunk, pfn;
854 
855 		if (memory_chunk[i].type != CHUNK_READ_WRITE &&
856 		    memory_chunk[i].type != CHUNK_CRASHK)
857 			continue;
858 		start_chunk = PFN_DOWN(memory_chunk[i].addr);
859 		end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
860 		end_chunk = min(end_chunk, end_pfn);
861 		if (start_chunk >= end_chunk)
862 			continue;
863 		memblock_add_node(PFN_PHYS(start_chunk),
864 				  PFN_PHYS(end_chunk - start_chunk), 0);
865 		pfn = max(start_chunk, start_pfn);
866 		for (; pfn < end_chunk; pfn++)
867 			page_set_storage_key(PFN_PHYS(pfn),
868 					     PAGE_DEFAULT_KEY, 0);
869 	}
870 
871 	psw_set_key(PAGE_DEFAULT_KEY);
872 
873 	free_bootmem_with_active_regions(0, max_pfn);
874 
875 	/*
876 	 * Reserve memory used for lowcore/command line/kernel image.
877 	 */
878 	reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
879 	reserve_bootmem((unsigned long)_stext,
880 			PFN_PHYS(start_pfn) - (unsigned long)_stext,
881 			BOOTMEM_DEFAULT);
882 	/*
883 	 * Reserve the bootmem bitmap itself as well. We do this in two
884 	 * steps (first step was init_bootmem()) because this catches
885 	 * the (very unlikely) case of us accidentally initializing the
886 	 * bootmem allocator with an invalid RAM area.
887 	 */
888 	reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
889 			BOOTMEM_DEFAULT);
890 
891 #ifdef CONFIG_CRASH_DUMP
892 	if (crashk_res.start)
893 		reserve_bootmem(crashk_res.start,
894 				crashk_res.end - crashk_res.start + 1,
895 				BOOTMEM_DEFAULT);
896 	if (is_kdump_kernel())
897 		reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
898 				PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
899 #endif
900 #ifdef CONFIG_BLK_DEV_INITRD
901 	if (INITRD_START && INITRD_SIZE) {
902 		if (INITRD_START + INITRD_SIZE <= memory_end) {
903 			reserve_bootmem(INITRD_START, INITRD_SIZE,
904 					BOOTMEM_DEFAULT);
905 			initrd_start = INITRD_START;
906 			initrd_end = initrd_start + INITRD_SIZE;
907 		} else {
908 			pr_err("initrd extends beyond end of "
909 			       "memory (0x%08lx > 0x%08lx) "
910 			       "disabling initrd\n",
911 			       initrd_start + INITRD_SIZE, memory_end);
912 			initrd_start = initrd_end = 0;
913 		}
914 	}
915 #endif
916 }
917 
918 /*
919  * Setup hardware capabilities.
920  */
921 static void __init setup_hwcaps(void)
922 {
923 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
924 	struct cpuid cpu_id;
925 	int i;
926 
927 	/*
928 	 * The store facility list bits numbers as found in the principles
929 	 * of operation are numbered with bit 1UL<<31 as number 0 to
930 	 * bit 1UL<<0 as number 31.
931 	 *   Bit 0: instructions named N3, "backported" to esa-mode
932 	 *   Bit 2: z/Architecture mode is active
933 	 *   Bit 7: the store-facility-list-extended facility is installed
934 	 *   Bit 17: the message-security assist is installed
935 	 *   Bit 19: the long-displacement facility is installed
936 	 *   Bit 21: the extended-immediate facility is installed
937 	 *   Bit 22: extended-translation facility 3 is installed
938 	 *   Bit 30: extended-translation facility 3 enhancement facility
939 	 * These get translated to:
940 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
941 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
942 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
943 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
944 	 */
945 	for (i = 0; i < 6; i++)
946 		if (test_facility(stfl_bits[i]))
947 			elf_hwcap |= 1UL << i;
948 
949 	if (test_facility(22) && test_facility(30))
950 		elf_hwcap |= HWCAP_S390_ETF3EH;
951 
952 	/*
953 	 * Check for additional facilities with store-facility-list-extended.
954 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
955 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
956 	 * as stored by stfl, bits 32-xxx contain additional facilities.
957 	 * How many facility words are stored depends on the number of
958 	 * doublewords passed to the instruction. The additional facilities
959 	 * are:
960 	 *   Bit 42: decimal floating point facility is installed
961 	 *   Bit 44: perform floating point operation facility is installed
962 	 * translated to:
963 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
964 	 */
965 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
966 		elf_hwcap |= HWCAP_S390_DFP;
967 
968 	/*
969 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
970 	 */
971 	if (MACHINE_HAS_HPAGE)
972 		elf_hwcap |= HWCAP_S390_HPAGE;
973 
974 	/*
975 	 * 64-bit register support for 31-bit processes
976 	 * HWCAP_S390_HIGH_GPRS is bit 9.
977 	 */
978 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
979 
980 	get_cpu_id(&cpu_id);
981 	switch (cpu_id.machine) {
982 	case 0x9672:
983 #if !defined(CONFIG_64BIT)
984 	default:	/* Use "g5" as default for 31 bit kernels. */
985 #endif
986 		strcpy(elf_platform, "g5");
987 		break;
988 	case 0x2064:
989 	case 0x2066:
990 #if defined(CONFIG_64BIT)
991 	default:	/* Use "z900" as default for 64 bit kernels. */
992 #endif
993 		strcpy(elf_platform, "z900");
994 		break;
995 	case 0x2084:
996 	case 0x2086:
997 		strcpy(elf_platform, "z990");
998 		break;
999 	case 0x2094:
1000 	case 0x2096:
1001 		strcpy(elf_platform, "z9-109");
1002 		break;
1003 	case 0x2097:
1004 	case 0x2098:
1005 		strcpy(elf_platform, "z10");
1006 		break;
1007 	case 0x2817:
1008 	case 0x2818:
1009 		strcpy(elf_platform, "z196");
1010 		break;
1011 	}
1012 }
1013 
1014 /*
1015  * Setup function called from init/main.c just after the banner
1016  * was printed.
1017  */
1018 
1019 void __init setup_arch(char **cmdline_p)
1020 {
1021         /*
1022          * print what head.S has found out about the machine
1023          */
1024 #ifndef CONFIG_64BIT
1025 	if (MACHINE_IS_VM)
1026 		pr_info("Linux is running as a z/VM "
1027 			"guest operating system in 31-bit mode\n");
1028 	else if (MACHINE_IS_LPAR)
1029 		pr_info("Linux is running natively in 31-bit mode\n");
1030 	if (MACHINE_HAS_IEEE)
1031 		pr_info("The hardware system has IEEE compatible "
1032 			"floating point units\n");
1033 	else
1034 		pr_info("The hardware system has no IEEE compatible "
1035 			"floating point units\n");
1036 #else /* CONFIG_64BIT */
1037 	if (MACHINE_IS_VM)
1038 		pr_info("Linux is running as a z/VM "
1039 			"guest operating system in 64-bit mode\n");
1040 	else if (MACHINE_IS_KVM)
1041 		pr_info("Linux is running under KVM in 64-bit mode\n");
1042 	else if (MACHINE_IS_LPAR)
1043 		pr_info("Linux is running natively in 64-bit mode\n");
1044 #endif /* CONFIG_64BIT */
1045 
1046 	/* Have one command line that is parsed and saved in /proc/cmdline */
1047 	/* boot_command_line has been already set up in early.c */
1048 	*cmdline_p = boot_command_line;
1049 
1050         ROOT_DEV = Root_RAM0;
1051 
1052 	init_mm.start_code = PAGE_OFFSET;
1053 	init_mm.end_code = (unsigned long) &_etext;
1054 	init_mm.end_data = (unsigned long) &_edata;
1055 	init_mm.brk = (unsigned long) &_end;
1056 
1057 	if (MACHINE_HAS_MVCOS)
1058 		memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1059 	else
1060 		memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1061 
1062 	parse_early_param();
1063 
1064 	os_info_init();
1065 	setup_ipl();
1066 	setup_memory_end();
1067 	setup_addressing_mode();
1068 	reserve_oldmem();
1069 	reserve_crashkernel();
1070 	setup_memory();
1071 	setup_resources();
1072 	setup_vmcoreinfo();
1073 	setup_lowcore();
1074 
1075         cpu_init();
1076 	s390_init_cpu_topology();
1077 
1078 	/*
1079 	 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1080 	 */
1081 	setup_hwcaps();
1082 
1083 	/*
1084 	 * Create kernel page tables and switch to virtual addressing.
1085 	 */
1086         paging_init();
1087 
1088         /* Setup default console */
1089 	conmode_default();
1090 	set_preferred_console();
1091 
1092 	/* Setup zfcpdump support */
1093 	setup_zfcpdump(console_devno);
1094 }
1095