[S390] Standby cpu activation/deactivation.
[safe/jmp/linux-2.6] / 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 #include <linux/errno.h>
18 #include <linux/module.h>
19 #include <linux/sched.h>
20 #include <linux/kernel.h>
21 #include <linux/mm.h>
22 #include <linux/stddef.h>
23 #include <linux/unistd.h>
24 #include <linux/ptrace.h>
25 #include <linux/slab.h>
26 #include <linux/user.h>
27 #include <linux/a.out.h>
28 #include <linux/tty.h>
29 #include <linux/ioport.h>
30 #include <linux/delay.h>
31 #include <linux/init.h>
32 #include <linux/initrd.h>
33 #include <linux/bootmem.h>
34 #include <linux/root_dev.h>
35 #include <linux/console.h>
36 #include <linux/seq_file.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
44 #include <asm/ipl.h>
45 #include <asm/uaccess.h>
46 #include <asm/system.h>
47 #include <asm/smp.h>
48 #include <asm/mmu_context.h>
49 #include <asm/cpcmd.h>
50 #include <asm/lowcore.h>
51 #include <asm/irq.h>
52 #include <asm/page.h>
53 #include <asm/ptrace.h>
54 #include <asm/sections.h>
55 #include <asm/ebcdic.h>
56 #include <asm/compat.h>
57
58 long psw_kernel_bits    = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
59                            PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
60 long psw_user_bits      = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
61                            PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
62                            PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
63
64 /*
65  * User copy operations.
66  */
67 struct uaccess_ops uaccess;
68 EXPORT_SYMBOL(uaccess);
69
70 /*
71  * Machine setup..
72  */
73 unsigned int console_mode = 0;
74 unsigned int console_devno = -1;
75 unsigned int console_irq = -1;
76 unsigned long machine_flags = 0;
77 unsigned long elf_hwcap = 0;
78 char elf_platform[ELF_PLATFORM_SIZE];
79
80 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
81 volatile int __cpu_logical_map[NR_CPUS]; /* logical cpu to cpu address */
82 static unsigned long __initdata memory_end;
83
84 /*
85  * This is set up by the setup-routine at boot-time
86  * for S390 need to find out, what we have to setup
87  * using address 0x10400 ...
88  */
89
90 #include <asm/setup.h>
91
92 static struct resource code_resource = {
93         .name  = "Kernel code",
94         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
95 };
96
97 static struct resource data_resource = {
98         .name = "Kernel data",
99         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
100 };
101
102 /*
103  * cpu_init() initializes state that is per-CPU.
104  */
105 void __cpuinit cpu_init(void)
106 {
107         int addr = hard_smp_processor_id();
108
109         /*
110          * Store processor id in lowcore (used e.g. in timer_interrupt)
111          */
112         get_cpu_id(&S390_lowcore.cpu_data.cpu_id);
113         S390_lowcore.cpu_data.cpu_addr = addr;
114
115         /*
116          * Force FPU initialization:
117          */
118         clear_thread_flag(TIF_USEDFPU);
119         clear_used_math();
120
121         atomic_inc(&init_mm.mm_count);
122         current->active_mm = &init_mm;
123         if (current->mm)
124                 BUG();
125         enter_lazy_tlb(&init_mm, current);
126 }
127
128 /*
129  * VM halt and poweroff setup routines
130  */
131 char vmhalt_cmd[128] = "";
132 char vmpoff_cmd[128] = "";
133 static char vmpanic_cmd[128] = "";
134
135 static void strncpy_skip_quote(char *dst, char *src, int n)
136 {
137         int sx, dx;
138
139         dx = 0;
140         for (sx = 0; src[sx] != 0; sx++) {
141                 if (src[sx] == '"') continue;
142                 dst[dx++] = src[sx];
143                 if (dx >= n) break;
144         }
145 }
146
147 static int __init vmhalt_setup(char *str)
148 {
149         strncpy_skip_quote(vmhalt_cmd, str, 127);
150         vmhalt_cmd[127] = 0;
151         return 1;
152 }
153
154 __setup("vmhalt=", vmhalt_setup);
155
156 static int __init vmpoff_setup(char *str)
157 {
158         strncpy_skip_quote(vmpoff_cmd, str, 127);
159         vmpoff_cmd[127] = 0;
160         return 1;
161 }
162
163 __setup("vmpoff=", vmpoff_setup);
164
165 static int vmpanic_notify(struct notifier_block *self, unsigned long event,
166                           void *data)
167 {
168         if (MACHINE_IS_VM && strlen(vmpanic_cmd) > 0)
169                 cpcmd(vmpanic_cmd, NULL, 0, NULL);
170
171         return NOTIFY_OK;
172 }
173
174 #define PANIC_PRI_VMPANIC       0
175
176 static struct notifier_block vmpanic_nb = {
177         .notifier_call = vmpanic_notify,
178         .priority = PANIC_PRI_VMPANIC
179 };
180
181 static int __init vmpanic_setup(char *str)
182 {
183         static int register_done __initdata = 0;
184
185         strncpy_skip_quote(vmpanic_cmd, str, 127);
186         vmpanic_cmd[127] = 0;
187         if (!register_done) {
188                 register_done = 1;
189                 atomic_notifier_chain_register(&panic_notifier_list,
190                                                &vmpanic_nb);
191         }
192         return 1;
193 }
194
195 __setup("vmpanic=", vmpanic_setup);
196
197 /*
198  * condev= and conmode= setup parameter.
199  */
200
201 static int __init condev_setup(char *str)
202 {
203         int vdev;
204
205         vdev = simple_strtoul(str, &str, 0);
206         if (vdev >= 0 && vdev < 65536) {
207                 console_devno = vdev;
208                 console_irq = -1;
209         }
210         return 1;
211 }
212
213 __setup("condev=", condev_setup);
214
215 static int __init conmode_setup(char *str)
216 {
217 #if defined(CONFIG_SCLP_CONSOLE)
218         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
219                 SET_CONSOLE_SCLP;
220 #endif
221 #if defined(CONFIG_TN3215_CONSOLE)
222         if (strncmp(str, "3215", 5) == 0)
223                 SET_CONSOLE_3215;
224 #endif
225 #if defined(CONFIG_TN3270_CONSOLE)
226         if (strncmp(str, "3270", 5) == 0)
227                 SET_CONSOLE_3270;
228 #endif
229         return 1;
230 }
231
232 __setup("conmode=", conmode_setup);
233
234 static void __init conmode_default(void)
235 {
236         char query_buffer[1024];
237         char *ptr;
238
239         if (MACHINE_IS_VM) {
240                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
241                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
242                 ptr = strstr(query_buffer, "SUBCHANNEL =");
243                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
244                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
245                 ptr = strstr(query_buffer, "CONMODE");
246                 /*
247                  * Set the conmode to 3215 so that the device recognition 
248                  * will set the cu_type of the console to 3215. If the
249                  * conmode is 3270 and we don't set it back then both
250                  * 3215 and the 3270 driver will try to access the console
251                  * device (3215 as console and 3270 as normal tty).
252                  */
253                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
254                 if (ptr == NULL) {
255 #if defined(CONFIG_SCLP_CONSOLE)
256                         SET_CONSOLE_SCLP;
257 #endif
258                         return;
259                 }
260                 if (strncmp(ptr + 8, "3270", 4) == 0) {
261 #if defined(CONFIG_TN3270_CONSOLE)
262                         SET_CONSOLE_3270;
263 #elif defined(CONFIG_TN3215_CONSOLE)
264                         SET_CONSOLE_3215;
265 #elif defined(CONFIG_SCLP_CONSOLE)
266                         SET_CONSOLE_SCLP;
267 #endif
268                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
269 #if defined(CONFIG_TN3215_CONSOLE)
270                         SET_CONSOLE_3215;
271 #elif defined(CONFIG_TN3270_CONSOLE)
272                         SET_CONSOLE_3270;
273 #elif defined(CONFIG_SCLP_CONSOLE)
274                         SET_CONSOLE_SCLP;
275 #endif
276                 }
277         } else if (MACHINE_IS_P390) {
278 #if defined(CONFIG_TN3215_CONSOLE)
279                 SET_CONSOLE_3215;
280 #elif defined(CONFIG_TN3270_CONSOLE)
281                 SET_CONSOLE_3270;
282 #endif
283         } else {
284 #if defined(CONFIG_SCLP_CONSOLE)
285                 SET_CONSOLE_SCLP;
286 #endif
287         }
288 }
289
290 #if defined(CONFIG_ZFCPDUMP) || defined(CONFIG_ZFCPDUMP_MODULE)
291 static void __init setup_zfcpdump(unsigned int console_devno)
292 {
293         static char str[64];
294
295         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
296                 return;
297         if (console_devno != -1)
298                 sprintf(str, "cio_ignore=all,!0.0.%04x,!0.0.%04x",
299                         ipl_info.data.fcp.dev_id.devno, console_devno);
300         else
301                 sprintf(str, "cio_ignore=all,!0.0.%04x",
302                         ipl_info.data.fcp.dev_id.devno);
303         strcat(COMMAND_LINE, " ");
304         strcat(COMMAND_LINE, str);
305         console_loglevel = 2;
306 }
307 #else
308 static inline void setup_zfcpdump(unsigned int console_devno) {}
309 #endif /* CONFIG_ZFCPDUMP */
310
311 #ifdef CONFIG_SMP
312 void (*_machine_restart)(char *command) = machine_restart_smp;
313 void (*_machine_halt)(void) = machine_halt_smp;
314 void (*_machine_power_off)(void) = machine_power_off_smp;
315 #else
316 /*
317  * Reboot, halt and power_off routines for non SMP.
318  */
319 static void do_machine_restart_nonsmp(char * __unused)
320 {
321         do_reipl();
322 }
323
324 static void do_machine_halt_nonsmp(void)
325 {
326         if (MACHINE_IS_VM && strlen(vmhalt_cmd) > 0)
327                 __cpcmd(vmhalt_cmd, NULL, 0, NULL);
328         signal_processor(smp_processor_id(), sigp_stop_and_store_status);
329 }
330
331 static void do_machine_power_off_nonsmp(void)
332 {
333         if (MACHINE_IS_VM && strlen(vmpoff_cmd) > 0)
334                 __cpcmd(vmpoff_cmd, NULL, 0, NULL);
335         signal_processor(smp_processor_id(), sigp_stop_and_store_status);
336 }
337
338 void (*_machine_restart)(char *command) = do_machine_restart_nonsmp;
339 void (*_machine_halt)(void) = do_machine_halt_nonsmp;
340 void (*_machine_power_off)(void) = do_machine_power_off_nonsmp;
341 #endif
342
343  /*
344  * Reboot, halt and power_off stubs. They just call _machine_restart,
345  * _machine_halt or _machine_power_off. 
346  */
347
348 void machine_restart(char *command)
349 {
350         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
351                 /*
352                  * Only unblank the console if we are called in enabled
353                  * context or a bust_spinlocks cleared the way for us.
354                  */
355                 console_unblank();
356         _machine_restart(command);
357 }
358
359 void machine_halt(void)
360 {
361         if (!in_interrupt() || oops_in_progress)
362                 /*
363                  * Only unblank the console if we are called in enabled
364                  * context or a bust_spinlocks cleared the way for us.
365                  */
366                 console_unblank();
367         _machine_halt();
368 }
369
370 void machine_power_off(void)
371 {
372         if (!in_interrupt() || oops_in_progress)
373                 /*
374                  * Only unblank the console if we are called in enabled
375                  * context or a bust_spinlocks cleared the way for us.
376                  */
377                 console_unblank();
378         _machine_power_off();
379 }
380
381 /*
382  * Dummy power off function.
383  */
384 void (*pm_power_off)(void) = machine_power_off;
385
386 static int __init early_parse_mem(char *p)
387 {
388         memory_end = memparse(p, &p);
389         return 0;
390 }
391 early_param("mem", early_parse_mem);
392
393 /*
394  * "ipldelay=XXX[sm]" sets ipl delay in seconds or minutes
395  */
396 static int __init early_parse_ipldelay(char *p)
397 {
398         unsigned long delay = 0;
399
400         delay = simple_strtoul(p, &p, 0);
401
402         switch (*p) {
403         case 's':
404         case 'S':
405                 delay *= 1000000;
406                 break;
407         case 'm':
408         case 'M':
409                 delay *= 60 * 1000000;
410         }
411
412         /* now wait for the requested amount of time */
413         udelay(delay);
414
415         return 0;
416 }
417 early_param("ipldelay", early_parse_ipldelay);
418
419 #ifdef CONFIG_S390_SWITCH_AMODE
420 unsigned int switch_amode = 0;
421 EXPORT_SYMBOL_GPL(switch_amode);
422
423 static void set_amode_and_uaccess(unsigned long user_amode,
424                                   unsigned long user32_amode)
425 {
426         psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
427                         PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
428                         PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
429 #ifdef CONFIG_COMPAT
430         psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
431                           PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
432                           PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
433         psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
434                           PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
435                           PSW32_MASK_PSTATE;
436 #endif
437         psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
438                           PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
439
440         if (MACHINE_HAS_MVCOS) {
441                 printk("mvcos available.\n");
442                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
443         } else {
444                 printk("mvcos not available.\n");
445                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
446         }
447 }
448
449 /*
450  * Switch kernel/user addressing modes?
451  */
452 static int __init early_parse_switch_amode(char *p)
453 {
454         switch_amode = 1;
455         return 0;
456 }
457 early_param("switch_amode", early_parse_switch_amode);
458
459 #else /* CONFIG_S390_SWITCH_AMODE */
460 static inline void set_amode_and_uaccess(unsigned long user_amode,
461                                          unsigned long user32_amode)
462 {
463 }
464 #endif /* CONFIG_S390_SWITCH_AMODE */
465
466 #ifdef CONFIG_S390_EXEC_PROTECT
467 unsigned int s390_noexec = 0;
468 EXPORT_SYMBOL_GPL(s390_noexec);
469
470 /*
471  * Enable execute protection?
472  */
473 static int __init early_parse_noexec(char *p)
474 {
475         if (!strncmp(p, "off", 3))
476                 return 0;
477         switch_amode = 1;
478         s390_noexec = 1;
479         return 0;
480 }
481 early_param("noexec", early_parse_noexec);
482 #endif /* CONFIG_S390_EXEC_PROTECT */
483
484 static void setup_addressing_mode(void)
485 {
486         if (s390_noexec) {
487                 printk("S390 execute protection active, ");
488                 set_amode_and_uaccess(PSW_ASC_SECONDARY, PSW32_ASC_SECONDARY);
489         } else if (switch_amode) {
490                 printk("S390 address spaces switched, ");
491                 set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY);
492         }
493 #ifdef CONFIG_TRACE_IRQFLAGS
494         sysc_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
495         io_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
496 #endif
497 }
498
499 static void __init
500 setup_lowcore(void)
501 {
502         struct _lowcore *lc;
503         int lc_pages;
504
505         /*
506          * Setup lowcore for boot cpu
507          */
508         lc_pages = sizeof(void *) == 8 ? 2 : 1;
509         lc = (struct _lowcore *)
510                 __alloc_bootmem(lc_pages * PAGE_SIZE, lc_pages * PAGE_SIZE, 0);
511         memset(lc, 0, lc_pages * PAGE_SIZE);
512         lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
513         lc->restart_psw.addr =
514                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
515         if (switch_amode)
516                 lc->restart_psw.mask |= PSW_ASC_HOME;
517         lc->external_new_psw.mask = psw_kernel_bits;
518         lc->external_new_psw.addr =
519                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
520         lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
521         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
522         lc->program_new_psw.mask = psw_kernel_bits;
523         lc->program_new_psw.addr =
524                 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
525         lc->mcck_new_psw.mask =
526                 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
527         lc->mcck_new_psw.addr =
528                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
529         lc->io_new_psw.mask = psw_kernel_bits;
530         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
531         lc->ipl_device = S390_lowcore.ipl_device;
532         lc->jiffy_timer = -1LL;
533         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
534         lc->async_stack = (unsigned long)
535                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
536         lc->panic_stack = (unsigned long)
537                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
538         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
539         lc->thread_info = (unsigned long) &init_thread_union;
540 #ifndef CONFIG_64BIT
541         if (MACHINE_HAS_IEEE) {
542                 lc->extended_save_area_addr = (__u32)
543                         __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0);
544                 /* enable extended save area */
545                 __ctl_set_bit(14, 29);
546         }
547 #endif
548         set_prefix((u32)(unsigned long) lc);
549 }
550
551 static void __init
552 setup_resources(void)
553 {
554         struct resource *res, *sub_res;
555         int i;
556
557         code_resource.start = (unsigned long) &_text;
558         code_resource.end = (unsigned long) &_etext - 1;
559         data_resource.start = (unsigned long) &_etext;
560         data_resource.end = (unsigned long) &_edata - 1;
561
562         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
563                 res = alloc_bootmem_low(sizeof(struct resource));
564                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
565                 switch (memory_chunk[i].type) {
566                 case CHUNK_READ_WRITE:
567                         res->name = "System RAM";
568                         break;
569                 case CHUNK_READ_ONLY:
570                         res->name = "System ROM";
571                         res->flags |= IORESOURCE_READONLY;
572                         break;
573                 default:
574                         res->name = "reserved";
575                 }
576                 res->start = memory_chunk[i].addr;
577                 res->end = memory_chunk[i].addr +  memory_chunk[i].size - 1;
578                 request_resource(&iomem_resource, res);
579
580                 if (code_resource.start >= res->start  &&
581                         code_resource.start <= res->end &&
582                         code_resource.end > res->end) {
583                         sub_res = alloc_bootmem_low(sizeof(struct resource));
584                         memcpy(sub_res, &code_resource,
585                                 sizeof(struct resource));
586                         sub_res->end = res->end;
587                         code_resource.start = res->end + 1;
588                         request_resource(res, sub_res);
589                 }
590
591                 if (code_resource.start >= res->start &&
592                         code_resource.start <= res->end &&
593                         code_resource.end <= res->end)
594                         request_resource(res, &code_resource);
595
596                 if (data_resource.start >= res->start &&
597                         data_resource.start <= res->end &&
598                         data_resource.end > res->end) {
599                         sub_res = alloc_bootmem_low(sizeof(struct resource));
600                         memcpy(sub_res, &data_resource,
601                                 sizeof(struct resource));
602                         sub_res->end = res->end;
603                         data_resource.start = res->end + 1;
604                         request_resource(res, sub_res);
605                 }
606
607                 if (data_resource.start >= res->start &&
608                         data_resource.start <= res->end &&
609                         data_resource.end <= res->end)
610                         request_resource(res, &data_resource);
611         }
612 }
613
614 unsigned long real_memory_size;
615 EXPORT_SYMBOL_GPL(real_memory_size);
616
617 static void __init setup_memory_end(void)
618 {
619         unsigned long memory_size;
620         unsigned long max_mem, max_phys;
621         int i;
622
623 #if defined(CONFIG_ZFCPDUMP) || defined(CONFIG_ZFCPDUMP_MODULE)
624         if (ipl_info.type == IPL_TYPE_FCP_DUMP)
625                 memory_end = ZFCPDUMP_HSA_SIZE;
626 #endif
627         memory_size = 0;
628         max_phys = VMALLOC_END_INIT - VMALLOC_MIN_SIZE;
629         memory_end &= PAGE_MASK;
630
631         max_mem = memory_end ? min(max_phys, memory_end) : max_phys;
632
633         for (i = 0; i < MEMORY_CHUNKS; i++) {
634                 struct mem_chunk *chunk = &memory_chunk[i];
635
636                 real_memory_size = max(real_memory_size,
637                                        chunk->addr + chunk->size);
638                 if (chunk->addr >= max_mem) {
639                         memset(chunk, 0, sizeof(*chunk));
640                         continue;
641                 }
642                 if (chunk->addr + chunk->size > max_mem)
643                         chunk->size = max_mem - chunk->addr;
644                 memory_size = max(memory_size, chunk->addr + chunk->size);
645         }
646         if (!memory_end)
647                 memory_end = memory_size;
648 }
649
650 static void __init
651 setup_memory(void)
652 {
653         unsigned long bootmap_size;
654         unsigned long start_pfn, end_pfn;
655         int i;
656
657         /*
658          * partially used pages are not usable - thus
659          * we are rounding upwards:
660          */
661         start_pfn = PFN_UP(__pa(&_end));
662         end_pfn = max_pfn = PFN_DOWN(memory_end);
663
664 #ifdef CONFIG_BLK_DEV_INITRD
665         /*
666          * Move the initrd in case the bitmap of the bootmem allocater
667          * would overwrite it.
668          */
669
670         if (INITRD_START && INITRD_SIZE) {
671                 unsigned long bmap_size;
672                 unsigned long start;
673
674                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
675                 bmap_size = PFN_PHYS(bmap_size);
676
677                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
678                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
679
680                         if (start + INITRD_SIZE > memory_end) {
681                                 printk("initrd extends beyond end of memory "
682                                        "(0x%08lx > 0x%08lx)\n"
683                                        "disabling initrd\n",
684                                        start + INITRD_SIZE, memory_end);
685                                 INITRD_START = INITRD_SIZE = 0;
686                         } else {
687                                 printk("Moving initrd (0x%08lx -> 0x%08lx, "
688                                        "size: %ld)\n",
689                                        INITRD_START, start, INITRD_SIZE);
690                                 memmove((void *) start, (void *) INITRD_START,
691                                         INITRD_SIZE);
692                                 INITRD_START = start;
693                         }
694                 }
695         }
696 #endif
697
698         /*
699          * Initialize the boot-time allocator
700          */
701         bootmap_size = init_bootmem(start_pfn, end_pfn);
702
703         /*
704          * Register RAM areas with the bootmem allocator.
705          */
706
707         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
708                 unsigned long start_chunk, end_chunk, pfn;
709
710                 if (memory_chunk[i].type != CHUNK_READ_WRITE)
711                         continue;
712                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
713                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size) - 1;
714                 end_chunk = min(end_chunk, end_pfn);
715                 if (start_chunk >= end_chunk)
716                         continue;
717                 add_active_range(0, start_chunk, end_chunk);
718                 pfn = max(start_chunk, start_pfn);
719                 for (; pfn <= end_chunk; pfn++)
720                         page_set_storage_key(PFN_PHYS(pfn), PAGE_DEFAULT_KEY);
721         }
722
723         psw_set_key(PAGE_DEFAULT_KEY);
724
725         free_bootmem_with_active_regions(0, max_pfn);
726
727         /*
728          * Reserve memory used for lowcore/command line/kernel image.
729          */
730         reserve_bootmem(0, (unsigned long)_ehead);
731         reserve_bootmem((unsigned long)_stext,
732                         PFN_PHYS(start_pfn) - (unsigned long)_stext);
733         /*
734          * Reserve the bootmem bitmap itself as well. We do this in two
735          * steps (first step was init_bootmem()) because this catches
736          * the (very unlikely) case of us accidentally initializing the
737          * bootmem allocator with an invalid RAM area.
738          */
739         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size);
740
741 #ifdef CONFIG_BLK_DEV_INITRD
742         if (INITRD_START && INITRD_SIZE) {
743                 if (INITRD_START + INITRD_SIZE <= memory_end) {
744                         reserve_bootmem(INITRD_START, INITRD_SIZE);
745                         initrd_start = INITRD_START;
746                         initrd_end = initrd_start + INITRD_SIZE;
747                 } else {
748                         printk("initrd extends beyond end of memory "
749                                "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
750                                initrd_start + INITRD_SIZE, memory_end);
751                         initrd_start = initrd_end = 0;
752                 }
753         }
754 #endif
755 }
756
757 static __init unsigned int stfl(void)
758 {
759         asm volatile(
760                 "       .insn   s,0xb2b10000,0(0)\n" /* stfl */
761                 "0:\n"
762                 EX_TABLE(0b,0b));
763         return S390_lowcore.stfl_fac_list;
764 }
765
766 static __init int stfle(unsigned long long *list, int doublewords)
767 {
768         typedef struct { unsigned long long _[doublewords]; } addrtype;
769         register unsigned long __nr asm("0") = doublewords - 1;
770
771         asm volatile(".insn s,0xb2b00000,%0" /* stfle */
772                      : "=m" (*(addrtype *) list), "+d" (__nr) : : "cc");
773         return __nr + 1;
774 }
775
776 /*
777  * Setup hardware capabilities.
778  */
779 static void __init setup_hwcaps(void)
780 {
781         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
782         struct cpuinfo_S390 *cpuinfo = &S390_lowcore.cpu_data;
783         unsigned long long facility_list_extended;
784         unsigned int facility_list;
785         int i;
786
787         facility_list = stfl();
788         /*
789          * The store facility list bits numbers as found in the principles
790          * of operation are numbered with bit 1UL<<31 as number 0 to
791          * bit 1UL<<0 as number 31.
792          *   Bit 0: instructions named N3, "backported" to esa-mode
793          *   Bit 2: z/Architecture mode is active
794          *   Bit 7: the store-facility-list-extended facility is installed
795          *   Bit 17: the message-security assist is installed
796          *   Bit 19: the long-displacement facility is installed
797          *   Bit 21: the extended-immediate facility is installed
798          * These get translated to:
799          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
800          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
801          *   HWCAP_S390_LDISP bit 4, and HWCAP_S390_EIMM bit 5.
802          */
803         for (i = 0; i < 6; i++)
804                 if (facility_list & (1UL << (31 - stfl_bits[i])))
805                         elf_hwcap |= 1UL << i;
806
807         /*
808          * Check for additional facilities with store-facility-list-extended.
809          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
810          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
811          * as stored by stfl, bits 32-xxx contain additional facilities.
812          * How many facility words are stored depends on the number of
813          * doublewords passed to the instruction. The additional facilites
814          * are:
815          *   Bit 43: decimal floating point facility is installed
816          * translated to:
817          *   HWCAP_S390_DFP bit 6.
818          */
819         if ((elf_hwcap & (1UL << 2)) &&
820             stfle(&facility_list_extended, 1) > 0) {
821                 if (facility_list_extended & (1ULL << (64 - 43)))
822                         elf_hwcap |= 1UL << 6;
823         }
824
825         switch (cpuinfo->cpu_id.machine) {
826         case 0x9672:
827 #if !defined(CONFIG_64BIT)
828         default:        /* Use "g5" as default for 31 bit kernels. */
829 #endif
830                 strcpy(elf_platform, "g5");
831                 break;
832         case 0x2064:
833         case 0x2066:
834 #if defined(CONFIG_64BIT)
835         default:        /* Use "z900" as default for 64 bit kernels. */
836 #endif
837                 strcpy(elf_platform, "z900");
838                 break;
839         case 0x2084:
840         case 0x2086:
841                 strcpy(elf_platform, "z990");
842                 break;
843         case 0x2094:
844                 strcpy(elf_platform, "z9-109");
845                 break;
846         }
847 }
848
849 /*
850  * Setup function called from init/main.c just after the banner
851  * was printed.
852  */
853
854 void __init
855 setup_arch(char **cmdline_p)
856 {
857         /*
858          * print what head.S has found out about the machine
859          */
860 #ifndef CONFIG_64BIT
861         printk((MACHINE_IS_VM) ?
862                "We are running under VM (31 bit mode)\n" :
863                "We are running native (31 bit mode)\n");
864         printk((MACHINE_HAS_IEEE) ?
865                "This machine has an IEEE fpu\n" :
866                "This machine has no IEEE fpu\n");
867 #else /* CONFIG_64BIT */
868         printk((MACHINE_IS_VM) ?
869                "We are running under VM (64 bit mode)\n" :
870                "We are running native (64 bit mode)\n");
871 #endif /* CONFIG_64BIT */
872
873         /* Save unparsed command line copy for /proc/cmdline */
874         strlcpy(boot_command_line, COMMAND_LINE, COMMAND_LINE_SIZE);
875
876         *cmdline_p = COMMAND_LINE;
877         *(*cmdline_p + COMMAND_LINE_SIZE - 1) = '\0';
878
879         ROOT_DEV = Root_RAM0;
880
881         init_mm.start_code = PAGE_OFFSET;
882         init_mm.end_code = (unsigned long) &_etext;
883         init_mm.end_data = (unsigned long) &_edata;
884         init_mm.brk = (unsigned long) &_end;
885
886         if (MACHINE_HAS_MVCOS)
887                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
888         else
889                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
890
891         parse_early_param();
892
893         setup_ipl_info();
894         setup_memory_end();
895         setup_addressing_mode();
896         setup_memory();
897         setup_resources();
898         setup_lowcore();
899
900         cpu_init();
901         __cpu_logical_map[0] = S390_lowcore.cpu_data.cpu_addr;
902         smp_setup_cpu_possible_map();
903
904         /*
905          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
906          */
907         setup_hwcaps();
908
909         /*
910          * Create kernel page tables and switch to virtual addressing.
911          */
912         paging_init();
913
914         /* Setup default console */
915         conmode_default();
916
917         /* Setup zfcpdump support */
918         setup_zfcpdump(console_devno);
919 }
920
921 void __cpuinit print_cpu_info(struct cpuinfo_S390 *cpuinfo)
922 {
923    printk(KERN_INFO "cpu %d "
924 #ifdef CONFIG_SMP
925            "phys_idx=%d "
926 #endif
927            "vers=%02X ident=%06X machine=%04X unused=%04X\n",
928            cpuinfo->cpu_nr,
929 #ifdef CONFIG_SMP
930            cpuinfo->cpu_addr,
931 #endif
932            cpuinfo->cpu_id.version,
933            cpuinfo->cpu_id.ident,
934            cpuinfo->cpu_id.machine,
935            cpuinfo->cpu_id.unused);
936 }
937
938 /*
939  * show_cpuinfo - Get information on one CPU for use by procfs.
940  */
941
942 static int show_cpuinfo(struct seq_file *m, void *v)
943 {
944         static const char *hwcap_str[7] = {
945                 "esan3", "zarch", "stfle", "msa", "ldisp", "eimm", "dfp"
946         };
947         struct cpuinfo_S390 *cpuinfo;
948         unsigned long n = (unsigned long) v - 1;
949         int i;
950
951         s390_adjust_jiffies();
952         preempt_disable();
953         if (!n) {
954                 seq_printf(m, "vendor_id       : IBM/S390\n"
955                                "# processors    : %i\n"
956                                "bogomips per cpu: %lu.%02lu\n",
957                                num_online_cpus(), loops_per_jiffy/(500000/HZ),
958                                (loops_per_jiffy/(5000/HZ))%100);
959                 seq_puts(m, "features\t: ");
960                 for (i = 0; i < 7; i++)
961                         if (hwcap_str[i] && (elf_hwcap & (1UL << i)))
962                                 seq_printf(m, "%s ", hwcap_str[i]);
963                 seq_puts(m, "\n");
964         }
965
966         if (cpu_online(n)) {
967 #ifdef CONFIG_SMP
968                 if (smp_processor_id() == n)
969                         cpuinfo = &S390_lowcore.cpu_data;
970                 else
971                         cpuinfo = &lowcore_ptr[n]->cpu_data;
972 #else
973                 cpuinfo = &S390_lowcore.cpu_data;
974 #endif
975                 seq_printf(m, "processor %li: "
976                                "version = %02X,  "
977                                "identification = %06X,  "
978                                "machine = %04X\n",
979                                n, cpuinfo->cpu_id.version,
980                                cpuinfo->cpu_id.ident,
981                                cpuinfo->cpu_id.machine);
982         }
983         preempt_enable();
984         return 0;
985 }
986
987 static void *c_start(struct seq_file *m, loff_t *pos)
988 {
989         return *pos < NR_CPUS ? (void *)((unsigned long) *pos + 1) : NULL;
990 }
991 static void *c_next(struct seq_file *m, void *v, loff_t *pos)
992 {
993         ++*pos;
994         return c_start(m, pos);
995 }
996 static void c_stop(struct seq_file *m, void *v)
997 {
998 }
999 struct seq_operations cpuinfo_op = {
1000         .start  = c_start,
1001         .next   = c_next,
1002         .stop   = c_stop,
1003         .show   = show_cpuinfo,
1004 };
1005