Loading...
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
46#include <asm/ipl.h>
47#include <asm/uaccess.h>
48#include <asm/system.h>
49#include <asm/smp.h>
50#include <asm/mmu_context.h>
51#include <asm/cpcmd.h>
52#include <asm/lowcore.h>
53#include <asm/irq.h>
54#include <asm/page.h>
55#include <asm/ptrace.h>
56#include <asm/sections.h>
57#include <asm/ebcdic.h>
58#include <asm/compat.h>
59#include <asm/kvm_virtio.h>
60
61long psw_kernel_bits = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
62 PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
63long psw_user_bits = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
64 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
65 PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
66
67/*
68 * User copy operations.
69 */
70struct uaccess_ops uaccess;
71EXPORT_SYMBOL(uaccess);
72
73/*
74 * Machine setup..
75 */
76unsigned int console_mode = 0;
77EXPORT_SYMBOL(console_mode);
78
79unsigned int console_devno = -1;
80EXPORT_SYMBOL(console_devno);
81
82unsigned int console_irq = -1;
83EXPORT_SYMBOL(console_irq);
84
85unsigned long elf_hwcap = 0;
86char elf_platform[ELF_PLATFORM_SIZE];
87
88struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
89
90int __initdata memory_end_set;
91unsigned long __initdata memory_end;
92
93/* An array with a pointer to the lowcore of every CPU. */
94struct _lowcore *lowcore_ptr[NR_CPUS];
95EXPORT_SYMBOL(lowcore_ptr);
96
97/*
98 * This is set up by the setup-routine at boot-time
99 * for S390 need to find out, what we have to setup
100 * using address 0x10400 ...
101 */
102
103#include <asm/setup.h>
104
105/*
106 * condev= and conmode= setup parameter.
107 */
108
109static int __init condev_setup(char *str)
110{
111 int vdev;
112
113 vdev = simple_strtoul(str, &str, 0);
114 if (vdev >= 0 && vdev < 65536) {
115 console_devno = vdev;
116 console_irq = -1;
117 }
118 return 1;
119}
120
121__setup("condev=", condev_setup);
122
123static void __init set_preferred_console(void)
124{
125 if (MACHINE_IS_KVM)
126 add_preferred_console("hvc", 0, NULL);
127 else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
128 add_preferred_console("ttyS", 0, NULL);
129 else if (CONSOLE_IS_3270)
130 add_preferred_console("tty3270", 0, NULL);
131}
132
133static int __init conmode_setup(char *str)
134{
135#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
136 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
137 SET_CONSOLE_SCLP;
138#endif
139#if defined(CONFIG_TN3215_CONSOLE)
140 if (strncmp(str, "3215", 5) == 0)
141 SET_CONSOLE_3215;
142#endif
143#if defined(CONFIG_TN3270_CONSOLE)
144 if (strncmp(str, "3270", 5) == 0)
145 SET_CONSOLE_3270;
146#endif
147 set_preferred_console();
148 return 1;
149}
150
151__setup("conmode=", conmode_setup);
152
153static void __init conmode_default(void)
154{
155 char query_buffer[1024];
156 char *ptr;
157
158 if (MACHINE_IS_VM) {
159 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
160 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
161 ptr = strstr(query_buffer, "SUBCHANNEL =");
162 console_irq = simple_strtoul(ptr + 13, NULL, 16);
163 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
164 ptr = strstr(query_buffer, "CONMODE");
165 /*
166 * Set the conmode to 3215 so that the device recognition
167 * will set the cu_type of the console to 3215. If the
168 * conmode is 3270 and we don't set it back then both
169 * 3215 and the 3270 driver will try to access the console
170 * device (3215 as console and 3270 as normal tty).
171 */
172 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
173 if (ptr == NULL) {
174#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
175 SET_CONSOLE_SCLP;
176#endif
177 return;
178 }
179 if (strncmp(ptr + 8, "3270", 4) == 0) {
180#if defined(CONFIG_TN3270_CONSOLE)
181 SET_CONSOLE_3270;
182#elif defined(CONFIG_TN3215_CONSOLE)
183 SET_CONSOLE_3215;
184#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
185 SET_CONSOLE_SCLP;
186#endif
187 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
188#if defined(CONFIG_TN3215_CONSOLE)
189 SET_CONSOLE_3215;
190#elif defined(CONFIG_TN3270_CONSOLE)
191 SET_CONSOLE_3270;
192#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
193 SET_CONSOLE_SCLP;
194#endif
195 }
196 } else {
197#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198 SET_CONSOLE_SCLP;
199#endif
200 }
201}
202
203#ifdef CONFIG_ZFCPDUMP
204static void __init setup_zfcpdump(unsigned int console_devno)
205{
206 static char str[41];
207
208 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
209 return;
210 if (console_devno != -1)
211 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
212 ipl_info.data.fcp.dev_id.devno, console_devno);
213 else
214 sprintf(str, " cio_ignore=all,!0.0.%04x",
215 ipl_info.data.fcp.dev_id.devno);
216 strcat(boot_command_line, str);
217 console_loglevel = 2;
218}
219#else
220static inline void setup_zfcpdump(unsigned int console_devno) {}
221#endif /* CONFIG_ZFCPDUMP */
222
223 /*
224 * Reboot, halt and power_off stubs. They just call _machine_restart,
225 * _machine_halt or _machine_power_off.
226 */
227
228void machine_restart(char *command)
229{
230 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
231 /*
232 * Only unblank the console if we are called in enabled
233 * context or a bust_spinlocks cleared the way for us.
234 */
235 console_unblank();
236 _machine_restart(command);
237}
238
239void machine_halt(void)
240{
241 if (!in_interrupt() || oops_in_progress)
242 /*
243 * Only unblank the console if we are called in enabled
244 * context or a bust_spinlocks cleared the way for us.
245 */
246 console_unblank();
247 _machine_halt();
248}
249
250void machine_power_off(void)
251{
252 if (!in_interrupt() || oops_in_progress)
253 /*
254 * Only unblank the console if we are called in enabled
255 * context or a bust_spinlocks cleared the way for us.
256 */
257 console_unblank();
258 _machine_power_off();
259}
260
261/*
262 * Dummy power off function.
263 */
264void (*pm_power_off)(void) = machine_power_off;
265
266static int __init early_parse_mem(char *p)
267{
268 memory_end = memparse(p, &p);
269 memory_end_set = 1;
270 return 0;
271}
272early_param("mem", early_parse_mem);
273
274unsigned int user_mode = HOME_SPACE_MODE;
275EXPORT_SYMBOL_GPL(user_mode);
276
277static int set_amode_and_uaccess(unsigned long user_amode,
278 unsigned long user32_amode)
279{
280 psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
281 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
282 PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
283#ifdef CONFIG_COMPAT
284 psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
285 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
286 PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
287 psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
288 PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
289 PSW32_MASK_PSTATE;
290#endif
291 psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
292 PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
293
294 if (MACHINE_HAS_MVCOS) {
295 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
296 return 1;
297 } else {
298 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
299 return 0;
300 }
301}
302
303/*
304 * Switch kernel/user addressing modes?
305 */
306static int __init early_parse_switch_amode(char *p)
307{
308 user_mode = PRIMARY_SPACE_MODE;
309 return 0;
310}
311early_param("switch_amode", early_parse_switch_amode);
312
313static int __init early_parse_user_mode(char *p)
314{
315 if (p && strcmp(p, "primary") == 0)
316 user_mode = PRIMARY_SPACE_MODE;
317 else if (!p || strcmp(p, "home") == 0)
318 user_mode = HOME_SPACE_MODE;
319 else
320 return 1;
321 return 0;
322}
323early_param("user_mode", early_parse_user_mode);
324
325static void setup_addressing_mode(void)
326{
327 if (user_mode == PRIMARY_SPACE_MODE) {
328 if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
329 pr_info("Address spaces switched, "
330 "mvcos available\n");
331 else
332 pr_info("Address spaces switched, "
333 "mvcos not available\n");
334 }
335}
336
337static void __init
338setup_lowcore(void)
339{
340 struct _lowcore *lc;
341
342 /*
343 * Setup lowcore for boot cpu
344 */
345 BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
346 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
347 lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
348 lc->restart_psw.addr =
349 PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
350 if (user_mode != HOME_SPACE_MODE)
351 lc->restart_psw.mask |= PSW_ASC_HOME;
352 lc->external_new_psw.mask = psw_kernel_bits;
353 lc->external_new_psw.addr =
354 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
355 lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
356 lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
357 lc->program_new_psw.mask = psw_kernel_bits;
358 lc->program_new_psw.addr =
359 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
360 lc->mcck_new_psw.mask =
361 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
362 lc->mcck_new_psw.addr =
363 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
364 lc->io_new_psw.mask = psw_kernel_bits;
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
402static struct resource code_resource = {
403 .name = "Kernel code",
404 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
405};
406
407static struct resource data_resource = {
408 .name = "Kernel data",
409 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
410};
411
412static struct resource bss_resource = {
413 .name = "Kernel bss",
414 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
415};
416
417static struct resource __initdata *standard_resources[] = {
418 &code_resource,
419 &data_resource,
420 &bss_resource,
421};
422
423static 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 res = alloc_bootmem_low(sizeof(*res));
439 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
440 switch (memory_chunk[i].type) {
441 case CHUNK_READ_WRITE:
442 res->name = "System RAM";
443 break;
444 case CHUNK_READ_ONLY:
445 res->name = "System ROM";
446 res->flags |= IORESOURCE_READONLY;
447 break;
448 default:
449 res->name = "reserved";
450 }
451 res->start = memory_chunk[i].addr;
452 res->end = res->start + memory_chunk[i].size - 1;
453 request_resource(&iomem_resource, res);
454
455 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
456 std_res = standard_resources[j];
457 if (std_res->start < res->start ||
458 std_res->start > res->end)
459 continue;
460 if (std_res->end > res->end) {
461 sub_res = alloc_bootmem_low(sizeof(*sub_res));
462 *sub_res = *std_res;
463 sub_res->end = res->end;
464 std_res->start = res->end + 1;
465 request_resource(res, sub_res);
466 } else {
467 request_resource(res, std_res);
468 }
469 }
470 }
471}
472
473unsigned long real_memory_size;
474EXPORT_SYMBOL_GPL(real_memory_size);
475
476static void __init setup_memory_end(void)
477{
478 unsigned long memory_size;
479 unsigned long max_mem;
480 int i;
481
482#ifdef CONFIG_ZFCPDUMP
483 if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
484 memory_end = ZFCPDUMP_HSA_SIZE;
485 memory_end_set = 1;
486 }
487#endif
488 memory_size = 0;
489 memory_end &= PAGE_MASK;
490
491 max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
492 memory_end = min(max_mem, memory_end);
493
494 /*
495 * Make sure all chunks are MAX_ORDER aligned so we don't need the
496 * extra checks that HOLES_IN_ZONE would require.
497 */
498 for (i = 0; i < MEMORY_CHUNKS; i++) {
499 unsigned long start, end;
500 struct mem_chunk *chunk;
501 unsigned long align;
502
503 chunk = &memory_chunk[i];
504 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
505 start = (chunk->addr + align - 1) & ~(align - 1);
506 end = (chunk->addr + chunk->size) & ~(align - 1);
507 if (start >= end)
508 memset(chunk, 0, sizeof(*chunk));
509 else {
510 chunk->addr = start;
511 chunk->size = end - start;
512 }
513 }
514
515 for (i = 0; i < MEMORY_CHUNKS; i++) {
516 struct mem_chunk *chunk = &memory_chunk[i];
517
518 real_memory_size = max(real_memory_size,
519 chunk->addr + chunk->size);
520 if (chunk->addr >= max_mem) {
521 memset(chunk, 0, sizeof(*chunk));
522 continue;
523 }
524 if (chunk->addr + chunk->size > max_mem)
525 chunk->size = max_mem - chunk->addr;
526 memory_size = max(memory_size, chunk->addr + chunk->size);
527 }
528 if (!memory_end)
529 memory_end = memory_size;
530}
531
532void *restart_stack __attribute__((__section__(".data")));
533
534/*
535 * Setup new PSW and allocate stack for PSW restart interrupt
536 */
537static void __init setup_restart_psw(void)
538{
539 psw_t psw;
540
541 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
542 restart_stack += ASYNC_SIZE;
543
544 /*
545 * Setup restart PSW for absolute zero lowcore. This is necesary
546 * if PSW restart is done on an offline CPU that has lowcore zero
547 */
548 psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
549 psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
550 copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
551}
552
553static void __init
554setup_memory(void)
555{
556 unsigned long bootmap_size;
557 unsigned long start_pfn, end_pfn;
558 int i;
559
560 /*
561 * partially used pages are not usable - thus
562 * we are rounding upwards:
563 */
564 start_pfn = PFN_UP(__pa(&_end));
565 end_pfn = max_pfn = PFN_DOWN(memory_end);
566
567#ifdef CONFIG_BLK_DEV_INITRD
568 /*
569 * Move the initrd in case the bitmap of the bootmem allocater
570 * would overwrite it.
571 */
572
573 if (INITRD_START && INITRD_SIZE) {
574 unsigned long bmap_size;
575 unsigned long start;
576
577 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
578 bmap_size = PFN_PHYS(bmap_size);
579
580 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
581 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
582
583 if (start + INITRD_SIZE > memory_end) {
584 pr_err("initrd extends beyond end of "
585 "memory (0x%08lx > 0x%08lx) "
586 "disabling initrd\n",
587 start + INITRD_SIZE, memory_end);
588 INITRD_START = INITRD_SIZE = 0;
589 } else {
590 pr_info("Moving initrd (0x%08lx -> "
591 "0x%08lx, size: %ld)\n",
592 INITRD_START, start, INITRD_SIZE);
593 memmove((void *) start, (void *) INITRD_START,
594 INITRD_SIZE);
595 INITRD_START = start;
596 }
597 }
598 }
599#endif
600
601 /*
602 * Initialize the boot-time allocator
603 */
604 bootmap_size = init_bootmem(start_pfn, end_pfn);
605
606 /*
607 * Register RAM areas with the bootmem allocator.
608 */
609
610 for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
611 unsigned long start_chunk, end_chunk, pfn;
612
613 if (memory_chunk[i].type != CHUNK_READ_WRITE)
614 continue;
615 start_chunk = PFN_DOWN(memory_chunk[i].addr);
616 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
617 end_chunk = min(end_chunk, end_pfn);
618 if (start_chunk >= end_chunk)
619 continue;
620 add_active_range(0, start_chunk, end_chunk);
621 pfn = max(start_chunk, start_pfn);
622 for (; pfn < end_chunk; pfn++)
623 page_set_storage_key(PFN_PHYS(pfn),
624 PAGE_DEFAULT_KEY, 0);
625 }
626
627 psw_set_key(PAGE_DEFAULT_KEY);
628
629 free_bootmem_with_active_regions(0, max_pfn);
630
631 /*
632 * Reserve memory used for lowcore/command line/kernel image.
633 */
634 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
635 reserve_bootmem((unsigned long)_stext,
636 PFN_PHYS(start_pfn) - (unsigned long)_stext,
637 BOOTMEM_DEFAULT);
638 /*
639 * Reserve the bootmem bitmap itself as well. We do this in two
640 * steps (first step was init_bootmem()) because this catches
641 * the (very unlikely) case of us accidentally initializing the
642 * bootmem allocator with an invalid RAM area.
643 */
644 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
645 BOOTMEM_DEFAULT);
646
647#ifdef CONFIG_BLK_DEV_INITRD
648 if (INITRD_START && INITRD_SIZE) {
649 if (INITRD_START + INITRD_SIZE <= memory_end) {
650 reserve_bootmem(INITRD_START, INITRD_SIZE,
651 BOOTMEM_DEFAULT);
652 initrd_start = INITRD_START;
653 initrd_end = initrd_start + INITRD_SIZE;
654 } else {
655 pr_err("initrd extends beyond end of "
656 "memory (0x%08lx > 0x%08lx) "
657 "disabling initrd\n",
658 initrd_start + INITRD_SIZE, memory_end);
659 initrd_start = initrd_end = 0;
660 }
661 }
662#endif
663}
664
665/*
666 * Setup hardware capabilities.
667 */
668static void __init setup_hwcaps(void)
669{
670 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
671 struct cpuid cpu_id;
672 int i;
673
674 /*
675 * The store facility list bits numbers as found in the principles
676 * of operation are numbered with bit 1UL<<31 as number 0 to
677 * bit 1UL<<0 as number 31.
678 * Bit 0: instructions named N3, "backported" to esa-mode
679 * Bit 2: z/Architecture mode is active
680 * Bit 7: the store-facility-list-extended facility is installed
681 * Bit 17: the message-security assist is installed
682 * Bit 19: the long-displacement facility is installed
683 * Bit 21: the extended-immediate facility is installed
684 * Bit 22: extended-translation facility 3 is installed
685 * Bit 30: extended-translation facility 3 enhancement facility
686 * These get translated to:
687 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
688 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
689 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
690 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
691 */
692 for (i = 0; i < 6; i++)
693 if (test_facility(stfl_bits[i]))
694 elf_hwcap |= 1UL << i;
695
696 if (test_facility(22) && test_facility(30))
697 elf_hwcap |= HWCAP_S390_ETF3EH;
698
699 /*
700 * Check for additional facilities with store-facility-list-extended.
701 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
702 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
703 * as stored by stfl, bits 32-xxx contain additional facilities.
704 * How many facility words are stored depends on the number of
705 * doublewords passed to the instruction. The additional facilities
706 * are:
707 * Bit 42: decimal floating point facility is installed
708 * Bit 44: perform floating point operation facility is installed
709 * translated to:
710 * HWCAP_S390_DFP bit 6 (42 && 44).
711 */
712 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
713 elf_hwcap |= HWCAP_S390_DFP;
714
715 /*
716 * Huge page support HWCAP_S390_HPAGE is bit 7.
717 */
718 if (MACHINE_HAS_HPAGE)
719 elf_hwcap |= HWCAP_S390_HPAGE;
720
721 /*
722 * 64-bit register support for 31-bit processes
723 * HWCAP_S390_HIGH_GPRS is bit 9.
724 */
725 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
726
727 get_cpu_id(&cpu_id);
728 switch (cpu_id.machine) {
729 case 0x9672:
730#if !defined(CONFIG_64BIT)
731 default: /* Use "g5" as default for 31 bit kernels. */
732#endif
733 strcpy(elf_platform, "g5");
734 break;
735 case 0x2064:
736 case 0x2066:
737#if defined(CONFIG_64BIT)
738 default: /* Use "z900" as default for 64 bit kernels. */
739#endif
740 strcpy(elf_platform, "z900");
741 break;
742 case 0x2084:
743 case 0x2086:
744 strcpy(elf_platform, "z990");
745 break;
746 case 0x2094:
747 case 0x2096:
748 strcpy(elf_platform, "z9-109");
749 break;
750 case 0x2097:
751 case 0x2098:
752 strcpy(elf_platform, "z10");
753 break;
754 case 0x2817:
755 case 0x2818:
756 strcpy(elf_platform, "z196");
757 break;
758 }
759}
760
761/*
762 * Setup function called from init/main.c just after the banner
763 * was printed.
764 */
765
766void __init
767setup_arch(char **cmdline_p)
768{
769 /*
770 * print what head.S has found out about the machine
771 */
772#ifndef CONFIG_64BIT
773 if (MACHINE_IS_VM)
774 pr_info("Linux is running as a z/VM "
775 "guest operating system in 31-bit mode\n");
776 else if (MACHINE_IS_LPAR)
777 pr_info("Linux is running natively in 31-bit mode\n");
778 if (MACHINE_HAS_IEEE)
779 pr_info("The hardware system has IEEE compatible "
780 "floating point units\n");
781 else
782 pr_info("The hardware system has no IEEE compatible "
783 "floating point units\n");
784#else /* CONFIG_64BIT */
785 if (MACHINE_IS_VM)
786 pr_info("Linux is running as a z/VM "
787 "guest operating system in 64-bit mode\n");
788 else if (MACHINE_IS_KVM)
789 pr_info("Linux is running under KVM in 64-bit mode\n");
790 else if (MACHINE_IS_LPAR)
791 pr_info("Linux is running natively in 64-bit mode\n");
792#endif /* CONFIG_64BIT */
793
794 /* Have one command line that is parsed and saved in /proc/cmdline */
795 /* boot_command_line has been already set up in early.c */
796 *cmdline_p = boot_command_line;
797
798 ROOT_DEV = Root_RAM0;
799
800 init_mm.start_code = PAGE_OFFSET;
801 init_mm.end_code = (unsigned long) &_etext;
802 init_mm.end_data = (unsigned long) &_edata;
803 init_mm.brk = (unsigned long) &_end;
804
805 if (MACHINE_HAS_MVCOS)
806 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
807 else
808 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
809
810 parse_early_param();
811
812 setup_ipl();
813 setup_memory_end();
814 setup_addressing_mode();
815 setup_memory();
816 setup_resources();
817 setup_restart_psw();
818 setup_lowcore();
819
820 cpu_init();
821 s390_init_cpu_topology();
822
823 /*
824 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
825 */
826 setup_hwcaps();
827
828 /*
829 * Create kernel page tables and switch to virtual addressing.
830 */
831 paging_init();
832
833 /* Setup default console */
834 conmode_default();
835 set_preferred_console();
836
837 /* Setup zfcpdump support */
838 setup_zfcpdump(console_devno);
839}
1/*
2 * S390 version
3 * Copyright IBM Corp. 1999, 2012
4 * Author(s): Hartmut Penner (hp@de.ibm.com),
5 * Martin Schwidefsky (schwidefsky@de.ibm.com)
6 *
7 * Derived from "arch/i386/kernel/setup.c"
8 * Copyright (C) 1995, Linus Torvalds
9 */
10
11/*
12 * This file handles the architecture-dependent parts of initialization
13 */
14
15#define KMSG_COMPONENT "setup"
16#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17
18#include <linux/errno.h>
19#include <linux/export.h>
20#include <linux/sched.h>
21#include <linux/kernel.h>
22#include <linux/memblock.h>
23#include <linux/mm.h>
24#include <linux/stddef.h>
25#include <linux/unistd.h>
26#include <linux/ptrace.h>
27#include <linux/random.h>
28#include <linux/user.h>
29#include <linux/tty.h>
30#include <linux/ioport.h>
31#include <linux/delay.h>
32#include <linux/init.h>
33#include <linux/initrd.h>
34#include <linux/bootmem.h>
35#include <linux/root_dev.h>
36#include <linux/console.h>
37#include <linux/kernel_stat.h>
38#include <linux/device.h>
39#include <linux/notifier.h>
40#include <linux/pfn.h>
41#include <linux/ctype.h>
42#include <linux/reboot.h>
43#include <linux/topology.h>
44#include <linux/kexec.h>
45#include <linux/crash_dump.h>
46#include <linux/memory.h>
47#include <linux/compat.h>
48
49#include <asm/ipl.h>
50#include <asm/facility.h>
51#include <asm/smp.h>
52#include <asm/mmu_context.h>
53#include <asm/cpcmd.h>
54#include <asm/lowcore.h>
55#include <asm/irq.h>
56#include <asm/page.h>
57#include <asm/ptrace.h>
58#include <asm/sections.h>
59#include <asm/ebcdic.h>
60#include <asm/kvm_virtio.h>
61#include <asm/diag.h>
62#include <asm/os_info.h>
63#include <asm/sclp.h>
64#include <asm/sysinfo.h>
65#include <asm/numa.h>
66#include "entry.h"
67
68/*
69 * Machine setup..
70 */
71unsigned int console_mode = 0;
72EXPORT_SYMBOL(console_mode);
73
74unsigned int console_devno = -1;
75EXPORT_SYMBOL(console_devno);
76
77unsigned int console_irq = -1;
78EXPORT_SYMBOL(console_irq);
79
80unsigned long elf_hwcap __read_mostly = 0;
81char elf_platform[ELF_PLATFORM_SIZE];
82
83unsigned long int_hwcap = 0;
84
85int __initdata memory_end_set;
86unsigned long __initdata memory_end;
87unsigned long __initdata max_physmem_end;
88
89unsigned long VMALLOC_START;
90EXPORT_SYMBOL(VMALLOC_START);
91
92unsigned long VMALLOC_END;
93EXPORT_SYMBOL(VMALLOC_END);
94
95struct page *vmemmap;
96EXPORT_SYMBOL(vmemmap);
97
98unsigned long MODULES_VADDR;
99unsigned long MODULES_END;
100
101/* An array with a pointer to the lowcore of every CPU. */
102struct lowcore *lowcore_ptr[NR_CPUS];
103EXPORT_SYMBOL(lowcore_ptr);
104
105/*
106 * This is set up by the setup-routine at boot-time
107 * for S390 need to find out, what we have to setup
108 * using address 0x10400 ...
109 */
110
111#include <asm/setup.h>
112
113/*
114 * condev= and conmode= setup parameter.
115 */
116
117static int __init condev_setup(char *str)
118{
119 int vdev;
120
121 vdev = simple_strtoul(str, &str, 0);
122 if (vdev >= 0 && vdev < 65536) {
123 console_devno = vdev;
124 console_irq = -1;
125 }
126 return 1;
127}
128
129__setup("condev=", condev_setup);
130
131static void __init set_preferred_console(void)
132{
133 if (MACHINE_IS_KVM) {
134 if (sclp.has_vt220)
135 add_preferred_console("ttyS", 1, NULL);
136 else if (sclp.has_linemode)
137 add_preferred_console("ttyS", 0, NULL);
138 else
139 add_preferred_console("hvc", 0, NULL);
140 } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
141 add_preferred_console("ttyS", 0, NULL);
142 else if (CONSOLE_IS_3270)
143 add_preferred_console("tty3270", 0, NULL);
144}
145
146static int __init conmode_setup(char *str)
147{
148#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
149 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
150 SET_CONSOLE_SCLP;
151#endif
152#if defined(CONFIG_TN3215_CONSOLE)
153 if (strncmp(str, "3215", 5) == 0)
154 SET_CONSOLE_3215;
155#endif
156#if defined(CONFIG_TN3270_CONSOLE)
157 if (strncmp(str, "3270", 5) == 0)
158 SET_CONSOLE_3270;
159#endif
160 set_preferred_console();
161 return 1;
162}
163
164__setup("conmode=", conmode_setup);
165
166static void __init conmode_default(void)
167{
168 char query_buffer[1024];
169 char *ptr;
170
171 if (MACHINE_IS_VM) {
172 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
173 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
174 ptr = strstr(query_buffer, "SUBCHANNEL =");
175 console_irq = simple_strtoul(ptr + 13, NULL, 16);
176 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
177 ptr = strstr(query_buffer, "CONMODE");
178 /*
179 * Set the conmode to 3215 so that the device recognition
180 * will set the cu_type of the console to 3215. If the
181 * conmode is 3270 and we don't set it back then both
182 * 3215 and the 3270 driver will try to access the console
183 * device (3215 as console and 3270 as normal tty).
184 */
185 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
186 if (ptr == NULL) {
187#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
188 SET_CONSOLE_SCLP;
189#endif
190 return;
191 }
192 if (strncmp(ptr + 8, "3270", 4) == 0) {
193#if defined(CONFIG_TN3270_CONSOLE)
194 SET_CONSOLE_3270;
195#elif defined(CONFIG_TN3215_CONSOLE)
196 SET_CONSOLE_3215;
197#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198 SET_CONSOLE_SCLP;
199#endif
200 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
201#if defined(CONFIG_TN3215_CONSOLE)
202 SET_CONSOLE_3215;
203#elif defined(CONFIG_TN3270_CONSOLE)
204 SET_CONSOLE_3270;
205#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
206 SET_CONSOLE_SCLP;
207#endif
208 }
209 } else {
210#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
211 SET_CONSOLE_SCLP;
212#endif
213 }
214}
215
216#ifdef CONFIG_CRASH_DUMP
217static void __init setup_zfcpdump(void)
218{
219 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
220 return;
221 if (OLDMEM_BASE)
222 return;
223 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
224 console_loglevel = 2;
225}
226#else
227static inline void setup_zfcpdump(void) {}
228#endif /* CONFIG_CRASH_DUMP */
229
230 /*
231 * Reboot, halt and power_off stubs. They just call _machine_restart,
232 * _machine_halt or _machine_power_off.
233 */
234
235void 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
246void 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
257void 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 */
271void (*pm_power_off)(void) = machine_power_off;
272EXPORT_SYMBOL_GPL(pm_power_off);
273
274static int __init early_parse_mem(char *p)
275{
276 memory_end = memparse(p, &p);
277 memory_end &= PAGE_MASK;
278 memory_end_set = 1;
279 return 0;
280}
281early_param("mem", early_parse_mem);
282
283static int __init parse_vmalloc(char *arg)
284{
285 if (!arg)
286 return -EINVAL;
287 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
288 return 0;
289}
290early_param("vmalloc", parse_vmalloc);
291
292void *restart_stack __attribute__((__section__(".data")));
293
294static void __init setup_lowcore(void)
295{
296 struct lowcore *lc;
297
298 /*
299 * Setup lowcore for boot cpu
300 */
301 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
302 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
303 lc->restart_psw.mask = PSW_KERNEL_BITS;
304 lc->restart_psw.addr = (unsigned long) restart_int_handler;
305 lc->external_new_psw.mask = PSW_KERNEL_BITS |
306 PSW_MASK_DAT | PSW_MASK_MCHECK;
307 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
308 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
309 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
310 lc->svc_new_psw.addr = (unsigned long) system_call;
311 lc->program_new_psw.mask = PSW_KERNEL_BITS |
312 PSW_MASK_DAT | PSW_MASK_MCHECK;
313 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
314 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
315 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
316 lc->io_new_psw.mask = PSW_KERNEL_BITS |
317 PSW_MASK_DAT | PSW_MASK_MCHECK;
318 lc->io_new_psw.addr = (unsigned long) io_int_handler;
319 lc->clock_comparator = -1ULL;
320 lc->kernel_stack = ((unsigned long) &init_thread_union)
321 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
322 lc->async_stack = (unsigned long)
323 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
324 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
325 lc->panic_stack = (unsigned long)
326 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
327 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
328 lc->current_task = (unsigned long) init_thread_union.thread_info.task;
329 lc->thread_info = (unsigned long) &init_thread_union;
330 lc->lpp = LPP_MAGIC;
331 lc->machine_flags = S390_lowcore.machine_flags;
332 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
333 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
334 MAX_FACILITY_BIT/8);
335 if (MACHINE_HAS_VX)
336 lc->vector_save_area_addr =
337 (unsigned long) &lc->vector_save_area;
338 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
339 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
340 lc->async_enter_timer = S390_lowcore.async_enter_timer;
341 lc->exit_timer = S390_lowcore.exit_timer;
342 lc->user_timer = S390_lowcore.user_timer;
343 lc->system_timer = S390_lowcore.system_timer;
344 lc->steal_timer = S390_lowcore.steal_timer;
345 lc->last_update_timer = S390_lowcore.last_update_timer;
346 lc->last_update_clock = S390_lowcore.last_update_clock;
347
348 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
349 restart_stack += ASYNC_SIZE;
350
351 /*
352 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
353 * restart data to the absolute zero lowcore. This is necessary if
354 * PSW restart is done on an offline CPU that has lowcore zero.
355 */
356 lc->restart_stack = (unsigned long) restart_stack;
357 lc->restart_fn = (unsigned long) do_restart;
358 lc->restart_data = 0;
359 lc->restart_source = -1UL;
360
361 /* Setup absolute zero lowcore */
362 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
363 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
364 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
365 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
366 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
367
368#ifdef CONFIG_SMP
369 lc->spinlock_lockval = arch_spin_lockval(0);
370#endif
371
372 set_prefix((u32)(unsigned long) lc);
373 lowcore_ptr[0] = lc;
374}
375
376static struct resource code_resource = {
377 .name = "Kernel code",
378 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
379};
380
381static struct resource data_resource = {
382 .name = "Kernel data",
383 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
384};
385
386static struct resource bss_resource = {
387 .name = "Kernel bss",
388 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
389};
390
391static struct resource __initdata *standard_resources[] = {
392 &code_resource,
393 &data_resource,
394 &bss_resource,
395};
396
397static void __init setup_resources(void)
398{
399 struct resource *res, *std_res, *sub_res;
400 struct memblock_region *reg;
401 int j;
402
403 code_resource.start = (unsigned long) &_text;
404 code_resource.end = (unsigned long) &_etext - 1;
405 data_resource.start = (unsigned long) &_etext;
406 data_resource.end = (unsigned long) &_edata - 1;
407 bss_resource.start = (unsigned long) &__bss_start;
408 bss_resource.end = (unsigned long) &__bss_stop - 1;
409
410 for_each_memblock(memory, reg) {
411 res = alloc_bootmem_low(sizeof(*res));
412 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
413
414 res->name = "System RAM";
415 res->start = reg->base;
416 res->end = reg->base + reg->size - 1;
417 request_resource(&iomem_resource, res);
418
419 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
420 std_res = standard_resources[j];
421 if (std_res->start < res->start ||
422 std_res->start > res->end)
423 continue;
424 if (std_res->end > res->end) {
425 sub_res = alloc_bootmem_low(sizeof(*sub_res));
426 *sub_res = *std_res;
427 sub_res->end = res->end;
428 std_res->start = res->end + 1;
429 request_resource(res, sub_res);
430 } else {
431 request_resource(res, std_res);
432 }
433 }
434 }
435}
436
437static void __init setup_memory_end(void)
438{
439 unsigned long vmax, vmalloc_size, tmp;
440
441 /* Choose kernel address space layout: 2, 3, or 4 levels. */
442 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
443 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
444 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
445 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
446 vmax = 1UL << 42; /* 3-level kernel page table */
447 else
448 vmax = 1UL << 53; /* 4-level kernel page table */
449 /* module area is at the end of the kernel address space. */
450 MODULES_END = vmax;
451 MODULES_VADDR = MODULES_END - MODULES_LEN;
452 VMALLOC_END = MODULES_VADDR;
453 VMALLOC_START = vmax - vmalloc_size;
454
455 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
456 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
457 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
458 tmp = SECTION_ALIGN_UP(tmp);
459 tmp = VMALLOC_START - tmp * sizeof(struct page);
460 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
461 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
462 vmemmap = (struct page *) tmp;
463
464 /* Take care that memory_end is set and <= vmemmap */
465 memory_end = min(memory_end ?: max_physmem_end, tmp);
466 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
467 memblock_remove(memory_end, ULONG_MAX);
468
469 pr_notice("Max memory size: %luMB\n", memory_end >> 20);
470}
471
472static void __init setup_vmcoreinfo(void)
473{
474 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
475}
476
477#ifdef CONFIG_CRASH_DUMP
478
479/*
480 * When kdump is enabled, we have to ensure that no memory from
481 * the area [0 - crashkernel memory size] and
482 * [crashk_res.start - crashk_res.end] is set offline.
483 */
484static int kdump_mem_notifier(struct notifier_block *nb,
485 unsigned long action, void *data)
486{
487 struct memory_notify *arg = data;
488
489 if (action != MEM_GOING_OFFLINE)
490 return NOTIFY_OK;
491 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
492 return NOTIFY_BAD;
493 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
494 return NOTIFY_OK;
495 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
496 return NOTIFY_OK;
497 return NOTIFY_BAD;
498}
499
500static struct notifier_block kdump_mem_nb = {
501 .notifier_call = kdump_mem_notifier,
502};
503
504#endif
505
506/*
507 * Make sure that the area behind memory_end is protected
508 */
509static void reserve_memory_end(void)
510{
511#ifdef CONFIG_CRASH_DUMP
512 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
513 !OLDMEM_BASE && sclp.hsa_size) {
514 memory_end = sclp.hsa_size;
515 memory_end &= PAGE_MASK;
516 memory_end_set = 1;
517 }
518#endif
519 if (!memory_end_set)
520 return;
521 memblock_reserve(memory_end, ULONG_MAX);
522}
523
524/*
525 * Make sure that oldmem, where the dump is stored, is protected
526 */
527static void reserve_oldmem(void)
528{
529#ifdef CONFIG_CRASH_DUMP
530 if (OLDMEM_BASE)
531 /* Forget all memory above the running kdump system */
532 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
533#endif
534}
535
536/*
537 * Make sure that oldmem, where the dump is stored, is protected
538 */
539static void remove_oldmem(void)
540{
541#ifdef CONFIG_CRASH_DUMP
542 if (OLDMEM_BASE)
543 /* Forget all memory above the running kdump system */
544 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
545#endif
546}
547
548/*
549 * Reserve memory for kdump kernel to be loaded with kexec
550 */
551static void __init reserve_crashkernel(void)
552{
553#ifdef CONFIG_CRASH_DUMP
554 unsigned long long crash_base, crash_size;
555 phys_addr_t low, high;
556 int rc;
557
558 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
559 &crash_base);
560
561 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
562 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
563 if (rc || crash_size == 0)
564 return;
565
566 if (memblock.memory.regions[0].size < crash_size) {
567 pr_info("crashkernel reservation failed: %s\n",
568 "first memory chunk must be at least crashkernel size");
569 return;
570 }
571
572 low = crash_base ?: OLDMEM_BASE;
573 high = low + crash_size;
574 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
575 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
576 crash_base = low;
577 } else {
578 /* Find suitable area in free memory */
579 low = max_t(unsigned long, crash_size, sclp.hsa_size);
580 high = crash_base ? crash_base + crash_size : ULONG_MAX;
581
582 if (crash_base && crash_base < low) {
583 pr_info("crashkernel reservation failed: %s\n",
584 "crash_base too low");
585 return;
586 }
587 low = crash_base ?: low;
588 crash_base = memblock_find_in_range(low, high, crash_size,
589 KEXEC_CRASH_MEM_ALIGN);
590 }
591
592 if (!crash_base) {
593 pr_info("crashkernel reservation failed: %s\n",
594 "no suitable area found");
595 return;
596 }
597
598 if (register_memory_notifier(&kdump_mem_nb))
599 return;
600
601 if (!OLDMEM_BASE && MACHINE_IS_VM)
602 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
603 crashk_res.start = crash_base;
604 crashk_res.end = crash_base + crash_size - 1;
605 insert_resource(&iomem_resource, &crashk_res);
606 memblock_remove(crash_base, crash_size);
607 pr_info("Reserving %lluMB of memory at %lluMB "
608 "for crashkernel (System RAM: %luMB)\n",
609 crash_size >> 20, crash_base >> 20,
610 (unsigned long)memblock.memory.total_size >> 20);
611 os_info_crashkernel_add(crash_base, crash_size);
612#endif
613}
614
615/*
616 * Reserve the initrd from being used by memblock
617 */
618static void __init reserve_initrd(void)
619{
620#ifdef CONFIG_BLK_DEV_INITRD
621 initrd_start = INITRD_START;
622 initrd_end = initrd_start + INITRD_SIZE;
623 memblock_reserve(INITRD_START, INITRD_SIZE);
624#endif
625}
626
627/*
628 * Check for initrd being in usable memory
629 */
630static void __init check_initrd(void)
631{
632#ifdef CONFIG_BLK_DEV_INITRD
633 if (INITRD_START && INITRD_SIZE &&
634 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
635 pr_err("initrd does not fit memory.\n");
636 memblock_free(INITRD_START, INITRD_SIZE);
637 initrd_start = initrd_end = 0;
638 }
639#endif
640}
641
642/*
643 * Reserve memory used for lowcore/command line/kernel image.
644 */
645static void __init reserve_kernel(void)
646{
647 unsigned long start_pfn = PFN_UP(__pa(&_end));
648
649#ifdef CONFIG_DMA_API_DEBUG
650 /*
651 * DMA_API_DEBUG code stumbles over addresses from the
652 * range [_ehead, _stext]. Mark the memory as reserved
653 * so it is not used for CONFIG_DMA_API_DEBUG=y.
654 */
655 memblock_reserve(0, PFN_PHYS(start_pfn));
656#else
657 memblock_reserve(0, (unsigned long)_ehead);
658 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
659 - (unsigned long)_stext);
660#endif
661}
662
663static void __init setup_memory(void)
664{
665 struct memblock_region *reg;
666
667 /*
668 * Init storage key for present memory
669 */
670 for_each_memblock(memory, reg) {
671 storage_key_init_range(reg->base, reg->base + reg->size);
672 }
673 psw_set_key(PAGE_DEFAULT_KEY);
674
675 /* Only cosmetics */
676 memblock_enforce_memory_limit(memblock_end_of_DRAM());
677}
678
679/*
680 * Setup hardware capabilities.
681 */
682static int __init setup_hwcaps(void)
683{
684 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
685 struct cpuid cpu_id;
686 int i;
687
688 /*
689 * The store facility list bits numbers as found in the principles
690 * of operation are numbered with bit 1UL<<31 as number 0 to
691 * bit 1UL<<0 as number 31.
692 * Bit 0: instructions named N3, "backported" to esa-mode
693 * Bit 2: z/Architecture mode is active
694 * Bit 7: the store-facility-list-extended facility is installed
695 * Bit 17: the message-security assist is installed
696 * Bit 19: the long-displacement facility is installed
697 * Bit 21: the extended-immediate facility is installed
698 * Bit 22: extended-translation facility 3 is installed
699 * Bit 30: extended-translation facility 3 enhancement facility
700 * These get translated to:
701 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
702 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
703 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
704 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
705 */
706 for (i = 0; i < 6; i++)
707 if (test_facility(stfl_bits[i]))
708 elf_hwcap |= 1UL << i;
709
710 if (test_facility(22) && test_facility(30))
711 elf_hwcap |= HWCAP_S390_ETF3EH;
712
713 /*
714 * Check for additional facilities with store-facility-list-extended.
715 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
716 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
717 * as stored by stfl, bits 32-xxx contain additional facilities.
718 * How many facility words are stored depends on the number of
719 * doublewords passed to the instruction. The additional facilities
720 * are:
721 * Bit 42: decimal floating point facility is installed
722 * Bit 44: perform floating point operation facility is installed
723 * translated to:
724 * HWCAP_S390_DFP bit 6 (42 && 44).
725 */
726 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
727 elf_hwcap |= HWCAP_S390_DFP;
728
729 /*
730 * Huge page support HWCAP_S390_HPAGE is bit 7.
731 */
732 if (MACHINE_HAS_HPAGE)
733 elf_hwcap |= HWCAP_S390_HPAGE;
734
735 /*
736 * 64-bit register support for 31-bit processes
737 * HWCAP_S390_HIGH_GPRS is bit 9.
738 */
739 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
740
741 /*
742 * Transactional execution support HWCAP_S390_TE is bit 10.
743 */
744 if (test_facility(50) && test_facility(73))
745 elf_hwcap |= HWCAP_S390_TE;
746
747 /*
748 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
749 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
750 * instead of facility bit 129.
751 */
752 if (MACHINE_HAS_VX)
753 elf_hwcap |= HWCAP_S390_VXRS;
754 get_cpu_id(&cpu_id);
755 add_device_randomness(&cpu_id, sizeof(cpu_id));
756 switch (cpu_id.machine) {
757 case 0x2064:
758 case 0x2066:
759 default: /* Use "z900" as default for 64 bit kernels. */
760 strcpy(elf_platform, "z900");
761 break;
762 case 0x2084:
763 case 0x2086:
764 strcpy(elf_platform, "z990");
765 break;
766 case 0x2094:
767 case 0x2096:
768 strcpy(elf_platform, "z9-109");
769 break;
770 case 0x2097:
771 case 0x2098:
772 strcpy(elf_platform, "z10");
773 break;
774 case 0x2817:
775 case 0x2818:
776 strcpy(elf_platform, "z196");
777 break;
778 case 0x2827:
779 case 0x2828:
780 strcpy(elf_platform, "zEC12");
781 break;
782 case 0x2964:
783 case 0x2965:
784 strcpy(elf_platform, "z13");
785 break;
786 }
787
788 /*
789 * Virtualization support HWCAP_INT_SIE is bit 0.
790 */
791 if (sclp.has_sief2)
792 int_hwcap |= HWCAP_INT_SIE;
793
794 return 0;
795}
796arch_initcall(setup_hwcaps);
797
798/*
799 * Add system information as device randomness
800 */
801static void __init setup_randomness(void)
802{
803 struct sysinfo_3_2_2 *vmms;
804
805 vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
806 if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
807 add_device_randomness(&vmms, vmms->count);
808 free_page((unsigned long) vmms);
809}
810
811/*
812 * Setup function called from init/main.c just after the banner
813 * was printed.
814 */
815
816void __init setup_arch(char **cmdline_p)
817{
818 /*
819 * print what head.S has found out about the machine
820 */
821 if (MACHINE_IS_VM)
822 pr_info("Linux is running as a z/VM "
823 "guest operating system in 64-bit mode\n");
824 else if (MACHINE_IS_KVM)
825 pr_info("Linux is running under KVM in 64-bit mode\n");
826 else if (MACHINE_IS_LPAR)
827 pr_info("Linux is running natively in 64-bit mode\n");
828
829 /* Have one command line that is parsed and saved in /proc/cmdline */
830 /* boot_command_line has been already set up in early.c */
831 *cmdline_p = boot_command_line;
832
833 ROOT_DEV = Root_RAM0;
834
835 /* Is init_mm really needed? */
836 init_mm.start_code = PAGE_OFFSET;
837 init_mm.end_code = (unsigned long) &_etext;
838 init_mm.end_data = (unsigned long) &_edata;
839 init_mm.brk = (unsigned long) &_end;
840
841 parse_early_param();
842#ifdef CONFIG_CRASH_DUMP
843 /* Deactivate elfcorehdr= kernel parameter */
844 elfcorehdr_addr = ELFCORE_ADDR_MAX;
845#endif
846
847 os_info_init();
848 setup_ipl();
849
850 /* Do some memory reservations *before* memory is added to memblock */
851 reserve_memory_end();
852 reserve_oldmem();
853 reserve_kernel();
854 reserve_initrd();
855 memblock_allow_resize();
856
857 /* Get information about *all* installed memory */
858 detect_memory_memblock();
859
860 remove_oldmem();
861
862 /*
863 * Make sure all chunks are MAX_ORDER aligned so we don't need the
864 * extra checks that HOLES_IN_ZONE would require.
865 *
866 * Is this still required?
867 */
868 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
869
870 setup_memory_end();
871 setup_memory();
872
873 check_initrd();
874 reserve_crashkernel();
875#ifdef CONFIG_CRASH_DUMP
876 /*
877 * Be aware that smp_save_dump_cpus() triggers a system reset.
878 * Therefore CPU and device initialization should be done afterwards.
879 */
880 smp_save_dump_cpus();
881#endif
882
883 setup_resources();
884 setup_vmcoreinfo();
885 setup_lowcore();
886 smp_fill_possible_mask();
887 cpu_init();
888 numa_setup();
889
890 /*
891 * Create kernel page tables and switch to virtual addressing.
892 */
893 paging_init();
894
895 /* Setup default console */
896 conmode_default();
897 set_preferred_console();
898
899 /* Setup zfcpdump support */
900 setup_zfcpdump();
901
902 /* Add system specific data to the random pool */
903 setup_randomness();
904}