[S390] improve irq tracing code in entry[64].S
[linux-2.6.git] / arch / s390 / kernel / setup.c
1 /*
2  *  arch/s390/kernel/setup.c
3  *
4  *  S390 version
5  *    Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
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
61 long psw_kernel_bits    = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
62                            PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
63 long 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  */
70 struct uaccess_ops uaccess;
71 EXPORT_SYMBOL(uaccess);
72
73 /*
74  * Machine setup..
75  */
76 unsigned int console_mode = 0;
77 EXPORT_SYMBOL(console_mode);
78
79 unsigned int console_devno = -1;
80 EXPORT_SYMBOL(console_devno);
81
82 unsigned int console_irq = -1;
83 EXPORT_SYMBOL(console_irq);
84
85 unsigned long elf_hwcap = 0;
86 char elf_platform[ELF_PLATFORM_SIZE];
87
88 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
89
90 int __initdata memory_end_set;
91 unsigned long __initdata memory_end;
92
93 /* An array with a pointer to the lowcore of every CPU. */
94 struct _lowcore *lowcore_ptr[NR_CPUS];
95 EXPORT_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 static struct resource code_resource = {
106         .name  = "Kernel code",
107         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
108 };
109
110 static struct resource data_resource = {
111         .name = "Kernel data",
112         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
113 };
114
115 /*
116  * condev= and conmode= setup parameter.
117  */
118
119 static int __init condev_setup(char *str)
120 {
121         int vdev;
122
123         vdev = simple_strtoul(str, &str, 0);
124         if (vdev >= 0 && vdev < 65536) {
125                 console_devno = vdev;
126                 console_irq = -1;
127         }
128         return 1;
129 }
130
131 __setup("condev=", condev_setup);
132
133 static void __init set_preferred_console(void)
134 {
135         if (MACHINE_IS_KVM)
136                 add_preferred_console("hvc", 0, NULL);
137         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
138                 add_preferred_console("ttyS", 0, NULL);
139         else if (CONSOLE_IS_3270)
140                 add_preferred_console("tty3270", 0, NULL);
141 }
142
143 static int __init conmode_setup(char *str)
144 {
145 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
146         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
147                 SET_CONSOLE_SCLP;
148 #endif
149 #if defined(CONFIG_TN3215_CONSOLE)
150         if (strncmp(str, "3215", 5) == 0)
151                 SET_CONSOLE_3215;
152 #endif
153 #if defined(CONFIG_TN3270_CONSOLE)
154         if (strncmp(str, "3270", 5) == 0)
155                 SET_CONSOLE_3270;
156 #endif
157         set_preferred_console();
158         return 1;
159 }
160
161 __setup("conmode=", conmode_setup);
162
163 static void __init conmode_default(void)
164 {
165         char query_buffer[1024];
166         char *ptr;
167
168         if (MACHINE_IS_VM) {
169                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
170                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
171                 ptr = strstr(query_buffer, "SUBCHANNEL =");
172                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
173                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
174                 ptr = strstr(query_buffer, "CONMODE");
175                 /*
176                  * Set the conmode to 3215 so that the device recognition 
177                  * will set the cu_type of the console to 3215. If the
178                  * conmode is 3270 and we don't set it back then both
179                  * 3215 and the 3270 driver will try to access the console
180                  * device (3215 as console and 3270 as normal tty).
181                  */
182                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
183                 if (ptr == NULL) {
184 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
185                         SET_CONSOLE_SCLP;
186 #endif
187                         return;
188                 }
189                 if (strncmp(ptr + 8, "3270", 4) == 0) {
190 #if defined(CONFIG_TN3270_CONSOLE)
191                         SET_CONSOLE_3270;
192 #elif defined(CONFIG_TN3215_CONSOLE)
193                         SET_CONSOLE_3215;
194 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
195                         SET_CONSOLE_SCLP;
196 #endif
197                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
198 #if defined(CONFIG_TN3215_CONSOLE)
199                         SET_CONSOLE_3215;
200 #elif defined(CONFIG_TN3270_CONSOLE)
201                         SET_CONSOLE_3270;
202 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
203                         SET_CONSOLE_SCLP;
204 #endif
205                 }
206         } else {
207 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
208                 SET_CONSOLE_SCLP;
209 #endif
210         }
211 }
212
213 #ifdef CONFIG_ZFCPDUMP
214 static void __init setup_zfcpdump(unsigned int console_devno)
215 {
216         static char str[41];
217
218         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
219                 return;
220         if (console_devno != -1)
221                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
222                         ipl_info.data.fcp.dev_id.devno, console_devno);
223         else
224                 sprintf(str, " cio_ignore=all,!0.0.%04x",
225                         ipl_info.data.fcp.dev_id.devno);
226         strcat(boot_command_line, str);
227         console_loglevel = 2;
228 }
229 #else
230 static inline void setup_zfcpdump(unsigned int console_devno) {}
231 #endif /* CONFIG_ZFCPDUMP */
232
233  /*
234  * Reboot, halt and power_off stubs. They just call _machine_restart,
235  * _machine_halt or _machine_power_off. 
236  */
237
238 void machine_restart(char *command)
239 {
240         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
241                 /*
242                  * Only unblank the console if we are called in enabled
243                  * context or a bust_spinlocks cleared the way for us.
244                  */
245                 console_unblank();
246         _machine_restart(command);
247 }
248
249 void machine_halt(void)
250 {
251         if (!in_interrupt() || oops_in_progress)
252                 /*
253                  * Only unblank the console if we are called in enabled
254                  * context or a bust_spinlocks cleared the way for us.
255                  */
256                 console_unblank();
257         _machine_halt();
258 }
259
260 void machine_power_off(void)
261 {
262         if (!in_interrupt() || oops_in_progress)
263                 /*
264                  * Only unblank the console if we are called in enabled
265                  * context or a bust_spinlocks cleared the way for us.
266                  */
267                 console_unblank();
268         _machine_power_off();
269 }
270
271 /*
272  * Dummy power off function.
273  */
274 void (*pm_power_off)(void) = machine_power_off;
275
276 static int __init early_parse_mem(char *p)
277 {
278         memory_end = memparse(p, &p);
279         memory_end_set = 1;
280         return 0;
281 }
282 early_param("mem", early_parse_mem);
283
284 unsigned int user_mode = HOME_SPACE_MODE;
285 EXPORT_SYMBOL_GPL(user_mode);
286
287 static int set_amode_and_uaccess(unsigned long user_amode,
288                                  unsigned long user32_amode)
289 {
290         psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
291                         PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
292                         PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
293 #ifdef CONFIG_COMPAT
294         psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
295                           PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
296                           PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
297         psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
298                           PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
299                           PSW32_MASK_PSTATE;
300 #endif
301         psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
302                           PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
303
304         if (MACHINE_HAS_MVCOS) {
305                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
306                 return 1;
307         } else {
308                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
309                 return 0;
310         }
311 }
312
313 /*
314  * Switch kernel/user addressing modes?
315  */
316 static int __init early_parse_switch_amode(char *p)
317 {
318         if (user_mode != SECONDARY_SPACE_MODE)
319                 user_mode = PRIMARY_SPACE_MODE;
320         return 0;
321 }
322 early_param("switch_amode", early_parse_switch_amode);
323
324 static int __init early_parse_user_mode(char *p)
325 {
326         if (p && strcmp(p, "primary") == 0)
327                 user_mode = PRIMARY_SPACE_MODE;
328 #ifdef CONFIG_S390_EXEC_PROTECT
329         else if (p && strcmp(p, "secondary") == 0)
330                 user_mode = SECONDARY_SPACE_MODE;
331 #endif
332         else if (!p || strcmp(p, "home") == 0)
333                 user_mode = HOME_SPACE_MODE;
334         else
335                 return 1;
336         return 0;
337 }
338 early_param("user_mode", early_parse_user_mode);
339
340 #ifdef CONFIG_S390_EXEC_PROTECT
341 /*
342  * Enable execute protection?
343  */
344 static int __init early_parse_noexec(char *p)
345 {
346         if (!strncmp(p, "off", 3))
347                 return 0;
348         user_mode = SECONDARY_SPACE_MODE;
349         return 0;
350 }
351 early_param("noexec", early_parse_noexec);
352 #endif /* CONFIG_S390_EXEC_PROTECT */
353
354 static void setup_addressing_mode(void)
355 {
356         if (user_mode == SECONDARY_SPACE_MODE) {
357                 if (set_amode_and_uaccess(PSW_ASC_SECONDARY,
358                                           PSW32_ASC_SECONDARY))
359                         pr_info("Execute protection active, "
360                                 "mvcos available\n");
361                 else
362                         pr_info("Execute protection active, "
363                                 "mvcos not available\n");
364         } else if (user_mode == PRIMARY_SPACE_MODE) {
365                 if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
366                         pr_info("Address spaces switched, "
367                                 "mvcos available\n");
368                 else
369                         pr_info("Address spaces switched, "
370                                 "mvcos not available\n");
371         }
372 }
373
374 static void __init
375 setup_lowcore(void)
376 {
377         struct _lowcore *lc;
378
379         /*
380          * Setup lowcore for boot cpu
381          */
382         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
383         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
384         lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
385         lc->restart_psw.addr =
386                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
387         if (user_mode != HOME_SPACE_MODE)
388                 lc->restart_psw.mask |= PSW_ASC_HOME;
389         lc->external_new_psw.mask = psw_kernel_bits;
390         lc->external_new_psw.addr =
391                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
392         lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
393         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
394         lc->program_new_psw.mask = psw_kernel_bits;
395         lc->program_new_psw.addr =
396                 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
397         lc->mcck_new_psw.mask =
398                 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
399         lc->mcck_new_psw.addr =
400                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
401         lc->io_new_psw.mask = psw_kernel_bits;
402         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
403         lc->clock_comparator = -1ULL;
404         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
405         lc->async_stack = (unsigned long)
406                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
407         lc->panic_stack = (unsigned long)
408                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
409         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
410         lc->thread_info = (unsigned long) &init_thread_union;
411         lc->machine_flags = S390_lowcore.machine_flags;
412 #ifndef CONFIG_64BIT
413         if (MACHINE_HAS_IEEE) {
414                 lc->extended_save_area_addr = (__u32)
415                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
416                 /* enable extended save area */
417                 __ctl_set_bit(14, 29);
418         }
419 #else
420         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
421 #endif
422         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
423         lc->async_enter_timer = S390_lowcore.async_enter_timer;
424         lc->exit_timer = S390_lowcore.exit_timer;
425         lc->user_timer = S390_lowcore.user_timer;
426         lc->system_timer = S390_lowcore.system_timer;
427         lc->steal_timer = S390_lowcore.steal_timer;
428         lc->last_update_timer = S390_lowcore.last_update_timer;
429         lc->last_update_clock = S390_lowcore.last_update_clock;
430         lc->ftrace_func = S390_lowcore.ftrace_func;
431         set_prefix((u32)(unsigned long) lc);
432         lowcore_ptr[0] = lc;
433 }
434
435 static void __init
436 setup_resources(void)
437 {
438         struct resource *res, *sub_res;
439         int i;
440
441         code_resource.start = (unsigned long) &_text;
442         code_resource.end = (unsigned long) &_etext - 1;
443         data_resource.start = (unsigned long) &_etext;
444         data_resource.end = (unsigned long) &_edata - 1;
445
446         for (i = 0; i < MEMORY_CHUNKS; i++) {
447                 if (!memory_chunk[i].size)
448                         continue;
449                 res = alloc_bootmem_low(sizeof(struct resource));
450                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
451                 switch (memory_chunk[i].type) {
452                 case CHUNK_READ_WRITE:
453                         res->name = "System RAM";
454                         break;
455                 case CHUNK_READ_ONLY:
456                         res->name = "System ROM";
457                         res->flags |= IORESOURCE_READONLY;
458                         break;
459                 default:
460                         res->name = "reserved";
461                 }
462                 res->start = memory_chunk[i].addr;
463                 res->end = memory_chunk[i].addr +  memory_chunk[i].size - 1;
464                 request_resource(&iomem_resource, res);
465
466                 if (code_resource.start >= res->start  &&
467                         code_resource.start <= res->end &&
468                         code_resource.end > res->end) {
469                         sub_res = alloc_bootmem_low(sizeof(struct resource));
470                         memcpy(sub_res, &code_resource,
471                                 sizeof(struct resource));
472                         sub_res->end = res->end;
473                         code_resource.start = res->end + 1;
474                         request_resource(res, sub_res);
475                 }
476
477                 if (code_resource.start >= res->start &&
478                         code_resource.start <= res->end &&
479                         code_resource.end <= res->end)
480                         request_resource(res, &code_resource);
481
482                 if (data_resource.start >= res->start &&
483                         data_resource.start <= res->end &&
484                         data_resource.end > res->end) {
485                         sub_res = alloc_bootmem_low(sizeof(struct resource));
486                         memcpy(sub_res, &data_resource,
487                                 sizeof(struct resource));
488                         sub_res->end = res->end;
489                         data_resource.start = res->end + 1;
490                         request_resource(res, sub_res);
491                 }
492
493                 if (data_resource.start >= res->start &&
494                         data_resource.start <= res->end &&
495                         data_resource.end <= res->end)
496                         request_resource(res, &data_resource);
497         }
498 }
499
500 unsigned long real_memory_size;
501 EXPORT_SYMBOL_GPL(real_memory_size);
502
503 static void __init setup_memory_end(void)
504 {
505         unsigned long memory_size;
506         unsigned long max_mem;
507         int i;
508
509 #ifdef CONFIG_ZFCPDUMP
510         if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
511                 memory_end = ZFCPDUMP_HSA_SIZE;
512                 memory_end_set = 1;
513         }
514 #endif
515         memory_size = 0;
516         memory_end &= PAGE_MASK;
517
518         max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
519         memory_end = min(max_mem, memory_end);
520
521         /*
522          * Make sure all chunks are MAX_ORDER aligned so we don't need the
523          * extra checks that HOLES_IN_ZONE would require.
524          */
525         for (i = 0; i < MEMORY_CHUNKS; i++) {
526                 unsigned long start, end;
527                 struct mem_chunk *chunk;
528                 unsigned long align;
529
530                 chunk = &memory_chunk[i];
531                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
532                 start = (chunk->addr + align - 1) & ~(align - 1);
533                 end = (chunk->addr + chunk->size) & ~(align - 1);
534                 if (start >= end)
535                         memset(chunk, 0, sizeof(*chunk));
536                 else {
537                         chunk->addr = start;
538                         chunk->size = end - start;
539                 }
540         }
541
542         for (i = 0; i < MEMORY_CHUNKS; i++) {
543                 struct mem_chunk *chunk = &memory_chunk[i];
544
545                 real_memory_size = max(real_memory_size,
546                                        chunk->addr + chunk->size);
547                 if (chunk->addr >= max_mem) {
548                         memset(chunk, 0, sizeof(*chunk));
549                         continue;
550                 }
551                 if (chunk->addr + chunk->size > max_mem)
552                         chunk->size = max_mem - chunk->addr;
553                 memory_size = max(memory_size, chunk->addr + chunk->size);
554         }
555         if (!memory_end)
556                 memory_end = memory_size;
557 }
558
559 static void __init
560 setup_memory(void)
561 {
562         unsigned long bootmap_size;
563         unsigned long start_pfn, end_pfn;
564         int i;
565
566         /*
567          * partially used pages are not usable - thus
568          * we are rounding upwards:
569          */
570         start_pfn = PFN_UP(__pa(&_end));
571         end_pfn = max_pfn = PFN_DOWN(memory_end);
572
573 #ifdef CONFIG_BLK_DEV_INITRD
574         /*
575          * Move the initrd in case the bitmap of the bootmem allocater
576          * would overwrite it.
577          */
578
579         if (INITRD_START && INITRD_SIZE) {
580                 unsigned long bmap_size;
581                 unsigned long start;
582
583                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
584                 bmap_size = PFN_PHYS(bmap_size);
585
586                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
587                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
588
589                         if (start + INITRD_SIZE > memory_end) {
590                                 pr_err("initrd extends beyond end of "
591                                        "memory (0x%08lx > 0x%08lx) "
592                                        "disabling initrd\n",
593                                        start + INITRD_SIZE, memory_end);
594                                 INITRD_START = INITRD_SIZE = 0;
595                         } else {
596                                 pr_info("Moving initrd (0x%08lx -> "
597                                         "0x%08lx, size: %ld)\n",
598                                         INITRD_START, start, INITRD_SIZE);
599                                 memmove((void *) start, (void *) INITRD_START,
600                                         INITRD_SIZE);
601                                 INITRD_START = start;
602                         }
603                 }
604         }
605 #endif
606
607         /*
608          * Initialize the boot-time allocator
609          */
610         bootmap_size = init_bootmem(start_pfn, end_pfn);
611
612         /*
613          * Register RAM areas with the bootmem allocator.
614          */
615
616         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
617                 unsigned long start_chunk, end_chunk, pfn;
618
619                 if (memory_chunk[i].type != CHUNK_READ_WRITE)
620                         continue;
621                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
622                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
623                 end_chunk = min(end_chunk, end_pfn);
624                 if (start_chunk >= end_chunk)
625                         continue;
626                 add_active_range(0, start_chunk, end_chunk);
627                 pfn = max(start_chunk, start_pfn);
628                 for (; pfn < end_chunk; pfn++)
629                         page_set_storage_key(PFN_PHYS(pfn), PAGE_DEFAULT_KEY);
630         }
631
632         psw_set_key(PAGE_DEFAULT_KEY);
633
634         free_bootmem_with_active_regions(0, max_pfn);
635
636         /*
637          * Reserve memory used for lowcore/command line/kernel image.
638          */
639         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
640         reserve_bootmem((unsigned long)_stext,
641                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
642                         BOOTMEM_DEFAULT);
643         /*
644          * Reserve the bootmem bitmap itself as well. We do this in two
645          * steps (first step was init_bootmem()) because this catches
646          * the (very unlikely) case of us accidentally initializing the
647          * bootmem allocator with an invalid RAM area.
648          */
649         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
650                         BOOTMEM_DEFAULT);
651
652 #ifdef CONFIG_BLK_DEV_INITRD
653         if (INITRD_START && INITRD_SIZE) {
654                 if (INITRD_START + INITRD_SIZE <= memory_end) {
655                         reserve_bootmem(INITRD_START, INITRD_SIZE,
656                                         BOOTMEM_DEFAULT);
657                         initrd_start = INITRD_START;
658                         initrd_end = initrd_start + INITRD_SIZE;
659                 } else {
660                         pr_err("initrd extends beyond end of "
661                                "memory (0x%08lx > 0x%08lx) "
662                                "disabling initrd\n",
663                                initrd_start + INITRD_SIZE, memory_end);
664                         initrd_start = initrd_end = 0;
665                 }
666         }
667 #endif
668 }
669
670 /*
671  * Setup hardware capabilities.
672  */
673 static void __init setup_hwcaps(void)
674 {
675         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
676         unsigned long long facility_list_extended;
677         unsigned int facility_list;
678         struct cpuid cpu_id;
679         int i;
680
681         facility_list = stfl();
682         /*
683          * The store facility list bits numbers as found in the principles
684          * of operation are numbered with bit 1UL<<31 as number 0 to
685          * bit 1UL<<0 as number 31.
686          *   Bit 0: instructions named N3, "backported" to esa-mode
687          *   Bit 2: z/Architecture mode is active
688          *   Bit 7: the store-facility-list-extended facility is installed
689          *   Bit 17: the message-security assist is installed
690          *   Bit 19: the long-displacement facility is installed
691          *   Bit 21: the extended-immediate facility is installed
692          *   Bit 22: extended-translation facility 3 is installed
693          *   Bit 30: extended-translation facility 3 enhancement facility
694          * These get translated to:
695          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
696          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
697          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
698          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
699          */
700         for (i = 0; i < 6; i++)
701                 if (facility_list & (1UL << (31 - stfl_bits[i])))
702                         elf_hwcap |= 1UL << i;
703
704         if ((facility_list & (1UL << (31 - 22)))
705             && (facility_list & (1UL << (31 - 30))))
706                 elf_hwcap |= HWCAP_S390_ETF3EH;
707
708         /*
709          * Check for additional facilities with store-facility-list-extended.
710          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
711          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
712          * as stored by stfl, bits 32-xxx contain additional facilities.
713          * How many facility words are stored depends on the number of
714          * doublewords passed to the instruction. The additional facilites
715          * are:
716          *   Bit 42: decimal floating point facility is installed
717          *   Bit 44: perform floating point operation facility is installed
718          * translated to:
719          *   HWCAP_S390_DFP bit 6 (42 && 44).
720          */
721         if ((elf_hwcap & (1UL << 2)) &&
722             __stfle(&facility_list_extended, 1) > 0) {
723                 if ((facility_list_extended & (1ULL << (63 - 42)))
724                     && (facility_list_extended & (1ULL << (63 - 44))))
725                         elf_hwcap |= HWCAP_S390_DFP;
726         }
727
728         /*
729          * Huge page support HWCAP_S390_HPAGE is bit 7.
730          */
731         if (MACHINE_HAS_HPAGE)
732                 elf_hwcap |= HWCAP_S390_HPAGE;
733
734         /*
735          * 64-bit register support for 31-bit processes
736          * HWCAP_S390_HIGH_GPRS is bit 9.
737          */
738         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
739
740         get_cpu_id(&cpu_id);
741         switch (cpu_id.machine) {
742         case 0x9672:
743 #if !defined(CONFIG_64BIT)
744         default:        /* Use "g5" as default for 31 bit kernels. */
745 #endif
746                 strcpy(elf_platform, "g5");
747                 break;
748         case 0x2064:
749         case 0x2066:
750 #if defined(CONFIG_64BIT)
751         default:        /* Use "z900" as default for 64 bit kernels. */
752 #endif
753                 strcpy(elf_platform, "z900");
754                 break;
755         case 0x2084:
756         case 0x2086:
757                 strcpy(elf_platform, "z990");
758                 break;
759         case 0x2094:
760         case 0x2096:
761                 strcpy(elf_platform, "z9-109");
762                 break;
763         case 0x2097:
764         case 0x2098:
765                 strcpy(elf_platform, "z10");
766                 break;
767         }
768 }
769
770 /*
771  * Setup function called from init/main.c just after the banner
772  * was printed.
773  */
774
775 void __init
776 setup_arch(char **cmdline_p)
777 {
778         /*
779          * print what head.S has found out about the machine
780          */
781 #ifndef CONFIG_64BIT
782         if (MACHINE_IS_VM)
783                 pr_info("Linux is running as a z/VM "
784                         "guest operating system in 31-bit mode\n");
785         else if (MACHINE_IS_LPAR)
786                 pr_info("Linux is running natively in 31-bit mode\n");
787         if (MACHINE_HAS_IEEE)
788                 pr_info("The hardware system has IEEE compatible "
789                         "floating point units\n");
790         else
791                 pr_info("The hardware system has no IEEE compatible "
792                         "floating point units\n");
793 #else /* CONFIG_64BIT */
794         if (MACHINE_IS_VM)
795                 pr_info("Linux is running as a z/VM "
796                         "guest operating system in 64-bit mode\n");
797         else if (MACHINE_IS_KVM)
798                 pr_info("Linux is running under KVM in 64-bit mode\n");
799         else if (MACHINE_IS_LPAR)
800                 pr_info("Linux is running natively in 64-bit mode\n");
801 #endif /* CONFIG_64BIT */
802
803         /* Have one command line that is parsed and saved in /proc/cmdline */
804         /* boot_command_line has been already set up in early.c */
805         *cmdline_p = boot_command_line;
806
807         ROOT_DEV = Root_RAM0;
808
809         init_mm.start_code = PAGE_OFFSET;
810         init_mm.end_code = (unsigned long) &_etext;
811         init_mm.end_data = (unsigned long) &_edata;
812         init_mm.brk = (unsigned long) &_end;
813
814         if (MACHINE_HAS_MVCOS)
815                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
816         else
817                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
818
819         parse_early_param();
820
821         setup_ipl();
822         setup_memory_end();
823         setup_addressing_mode();
824         setup_memory();
825         setup_resources();
826         setup_lowcore();
827
828         cpu_init();
829         s390_init_cpu_topology();
830
831         /*
832          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
833          */
834         setup_hwcaps();
835
836         /*
837          * Create kernel page tables and switch to virtual addressing.
838          */
839         paging_init();
840
841         /* Setup default console */
842         conmode_default();
843         set_preferred_console();
844
845         /* Setup zfcpdump support */
846         setup_zfcpdump(console_devno);
847 }