[S390] smp: fix lowcore allocation
[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 #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/slab.h>
29 #include <linux/user.h>
30 #include <linux/tty.h>
31 #include <linux/ioport.h>
32 #include <linux/delay.h>
33 #include <linux/init.h>
34 #include <linux/initrd.h>
35 #include <linux/bootmem.h>
36 #include <linux/root_dev.h>
37 #include <linux/console.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/device.h>
40 #include <linux/notifier.h>
41 #include <linux/pfn.h>
42 #include <linux/ctype.h>
43 #include <linux/reboot.h>
44 #include <linux/topology.h>
45 #include <linux/ftrace.h>
46
47 #include <asm/ipl.h>
48 #include <asm/uaccess.h>
49 #include <asm/system.h>
50 #include <asm/smp.h>
51 #include <asm/mmu_context.h>
52 #include <asm/cpcmd.h>
53 #include <asm/lowcore.h>
54 #include <asm/irq.h>
55 #include <asm/page.h>
56 #include <asm/ptrace.h>
57 #include <asm/sections.h>
58 #include <asm/ebcdic.h>
59 #include <asm/compat.h>
60 #include <asm/kvm_virtio.h>
61
62 long psw_kernel_bits    = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
63                            PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
64 long psw_user_bits      = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
65                            PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
66                            PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
67
68 /*
69  * User copy operations.
70  */
71 struct uaccess_ops uaccess;
72 EXPORT_SYMBOL(uaccess);
73
74 /*
75  * Machine setup..
76  */
77 unsigned int console_mode = 0;
78 EXPORT_SYMBOL(console_mode);
79
80 unsigned int console_devno = -1;
81 EXPORT_SYMBOL(console_devno);
82
83 unsigned int console_irq = -1;
84 EXPORT_SYMBOL(console_irq);
85
86 unsigned long elf_hwcap = 0;
87 char elf_platform[ELF_PLATFORM_SIZE];
88
89 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
90
91 int __initdata memory_end_set;
92 unsigned long __initdata memory_end;
93
94 /* An array with a pointer to the lowcore of every CPU. */
95 struct _lowcore *lowcore_ptr[NR_CPUS];
96 EXPORT_SYMBOL(lowcore_ptr);
97
98 /*
99  * This is set up by the setup-routine at boot-time
100  * for S390 need to find out, what we have to setup
101  * using address 0x10400 ...
102  */
103
104 #include <asm/setup.h>
105
106 static struct resource code_resource = {
107         .name  = "Kernel code",
108         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
109 };
110
111 static struct resource data_resource = {
112         .name = "Kernel data",
113         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
114 };
115
116 /*
117  * cpu_init() initializes state that is per-CPU.
118  */
119 void __cpuinit cpu_init(void)
120 {
121         /*
122          * Store processor id in lowcore (used e.g. in timer_interrupt)
123          */
124         get_cpu_id(&S390_lowcore.cpu_id);
125
126         atomic_inc(&init_mm.mm_count);
127         current->active_mm = &init_mm;
128         BUG_ON(current->mm);
129         enter_lazy_tlb(&init_mm, current);
130 }
131
132 /*
133  * condev= and conmode= setup parameter.
134  */
135
136 static int __init condev_setup(char *str)
137 {
138         int vdev;
139
140         vdev = simple_strtoul(str, &str, 0);
141         if (vdev >= 0 && vdev < 65536) {
142                 console_devno = vdev;
143                 console_irq = -1;
144         }
145         return 1;
146 }
147
148 __setup("condev=", condev_setup);
149
150 static void __init set_preferred_console(void)
151 {
152         if (MACHINE_IS_KVM)
153                 add_preferred_console("hvc", 0, NULL);
154         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
155                 add_preferred_console("ttyS", 0, NULL);
156         else if (CONSOLE_IS_3270)
157                 add_preferred_console("tty3270", 0, NULL);
158 }
159
160 static int __init conmode_setup(char *str)
161 {
162 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
163         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
164                 SET_CONSOLE_SCLP;
165 #endif
166 #if defined(CONFIG_TN3215_CONSOLE)
167         if (strncmp(str, "3215", 5) == 0)
168                 SET_CONSOLE_3215;
169 #endif
170 #if defined(CONFIG_TN3270_CONSOLE)
171         if (strncmp(str, "3270", 5) == 0)
172                 SET_CONSOLE_3270;
173 #endif
174         set_preferred_console();
175         return 1;
176 }
177
178 __setup("conmode=", conmode_setup);
179
180 static void __init conmode_default(void)
181 {
182         char query_buffer[1024];
183         char *ptr;
184
185         if (MACHINE_IS_VM) {
186                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
187                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
188                 ptr = strstr(query_buffer, "SUBCHANNEL =");
189                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
190                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
191                 ptr = strstr(query_buffer, "CONMODE");
192                 /*
193                  * Set the conmode to 3215 so that the device recognition 
194                  * will set the cu_type of the console to 3215. If the
195                  * conmode is 3270 and we don't set it back then both
196                  * 3215 and the 3270 driver will try to access the console
197                  * device (3215 as console and 3270 as normal tty).
198                  */
199                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
200                 if (ptr == NULL) {
201 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
202                         SET_CONSOLE_SCLP;
203 #endif
204                         return;
205                 }
206                 if (strncmp(ptr + 8, "3270", 4) == 0) {
207 #if defined(CONFIG_TN3270_CONSOLE)
208                         SET_CONSOLE_3270;
209 #elif defined(CONFIG_TN3215_CONSOLE)
210                         SET_CONSOLE_3215;
211 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
212                         SET_CONSOLE_SCLP;
213 #endif
214                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
215 #if defined(CONFIG_TN3215_CONSOLE)
216                         SET_CONSOLE_3215;
217 #elif defined(CONFIG_TN3270_CONSOLE)
218                         SET_CONSOLE_3270;
219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
220                         SET_CONSOLE_SCLP;
221 #endif
222                 }
223         } else {
224 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
225                 SET_CONSOLE_SCLP;
226 #endif
227         }
228 }
229
230 #ifdef CONFIG_ZFCPDUMP
231 static void __init setup_zfcpdump(unsigned int console_devno)
232 {
233         static char str[41];
234
235         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
236                 return;
237         if (console_devno != -1)
238                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
239                         ipl_info.data.fcp.dev_id.devno, console_devno);
240         else
241                 sprintf(str, " cio_ignore=all,!0.0.%04x",
242                         ipl_info.data.fcp.dev_id.devno);
243         strcat(boot_command_line, str);
244         console_loglevel = 2;
245 }
246 #else
247 static inline void setup_zfcpdump(unsigned int console_devno) {}
248 #endif /* CONFIG_ZFCPDUMP */
249
250  /*
251  * Reboot, halt and power_off stubs. They just call _machine_restart,
252  * _machine_halt or _machine_power_off. 
253  */
254
255 void machine_restart(char *command)
256 {
257         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
258                 /*
259                  * Only unblank the console if we are called in enabled
260                  * context or a bust_spinlocks cleared the way for us.
261                  */
262                 console_unblank();
263         _machine_restart(command);
264 }
265
266 void machine_halt(void)
267 {
268         if (!in_interrupt() || oops_in_progress)
269                 /*
270                  * Only unblank the console if we are called in enabled
271                  * context or a bust_spinlocks cleared the way for us.
272                  */
273                 console_unblank();
274         _machine_halt();
275 }
276
277 void machine_power_off(void)
278 {
279         if (!in_interrupt() || oops_in_progress)
280                 /*
281                  * Only unblank the console if we are called in enabled
282                  * context or a bust_spinlocks cleared the way for us.
283                  */
284                 console_unblank();
285         _machine_power_off();
286 }
287
288 /*
289  * Dummy power off function.
290  */
291 void (*pm_power_off)(void) = machine_power_off;
292
293 static int __init early_parse_mem(char *p)
294 {
295         memory_end = memparse(p, &p);
296         memory_end_set = 1;
297         return 0;
298 }
299 early_param("mem", early_parse_mem);
300
301 unsigned int user_mode = HOME_SPACE_MODE;
302 EXPORT_SYMBOL_GPL(user_mode);
303
304 static int set_amode_and_uaccess(unsigned long user_amode,
305                                  unsigned long user32_amode)
306 {
307         psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
308                         PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
309                         PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
310 #ifdef CONFIG_COMPAT
311         psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
312                           PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
313                           PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
314         psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
315                           PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
316                           PSW32_MASK_PSTATE;
317 #endif
318         psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
319                           PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
320
321         if (MACHINE_HAS_MVCOS) {
322                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
323                 return 1;
324         } else {
325                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
326                 return 0;
327         }
328 }
329
330 /*
331  * Switch kernel/user addressing modes?
332  */
333 static int __init early_parse_switch_amode(char *p)
334 {
335         if (user_mode != SECONDARY_SPACE_MODE)
336                 user_mode = PRIMARY_SPACE_MODE;
337         return 0;
338 }
339 early_param("switch_amode", early_parse_switch_amode);
340
341 static int __init early_parse_user_mode(char *p)
342 {
343         if (p && strcmp(p, "primary") == 0)
344                 user_mode = PRIMARY_SPACE_MODE;
345 #ifdef CONFIG_S390_EXEC_PROTECT
346         else if (p && strcmp(p, "secondary") == 0)
347                 user_mode = SECONDARY_SPACE_MODE;
348 #endif
349         else if (!p || strcmp(p, "home") == 0)
350                 user_mode = HOME_SPACE_MODE;
351         else
352                 return 1;
353         return 0;
354 }
355 early_param("user_mode", early_parse_user_mode);
356
357 #ifdef CONFIG_S390_EXEC_PROTECT
358 /*
359  * Enable execute protection?
360  */
361 static int __init early_parse_noexec(char *p)
362 {
363         if (!strncmp(p, "off", 3))
364                 return 0;
365         user_mode = SECONDARY_SPACE_MODE;
366         return 0;
367 }
368 early_param("noexec", early_parse_noexec);
369 #endif /* CONFIG_S390_EXEC_PROTECT */
370
371 static void setup_addressing_mode(void)
372 {
373         if (user_mode == SECONDARY_SPACE_MODE) {
374                 if (set_amode_and_uaccess(PSW_ASC_SECONDARY,
375                                           PSW32_ASC_SECONDARY))
376                         pr_info("Execute protection active, "
377                                 "mvcos available\n");
378                 else
379                         pr_info("Execute protection active, "
380                                 "mvcos not available\n");
381         } else if (user_mode == PRIMARY_SPACE_MODE) {
382                 if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
383                         pr_info("Address spaces switched, "
384                                 "mvcos available\n");
385                 else
386                         pr_info("Address spaces switched, "
387                                 "mvcos not available\n");
388         }
389 #ifdef CONFIG_TRACE_IRQFLAGS
390         sysc_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
391         io_restore_trace_psw.mask = psw_kernel_bits & ~PSW_MASK_MCHECK;
392 #endif
393 }
394
395 static void __init
396 setup_lowcore(void)
397 {
398         struct _lowcore *lc;
399
400         /*
401          * Setup lowcore for boot cpu
402          */
403         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
404         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
405         lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
406         lc->restart_psw.addr =
407                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
408         if (user_mode != HOME_SPACE_MODE)
409                 lc->restart_psw.mask |= PSW_ASC_HOME;
410         lc->external_new_psw.mask = psw_kernel_bits;
411         lc->external_new_psw.addr =
412                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
413         lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
414         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
415         lc->program_new_psw.mask = psw_kernel_bits;
416         lc->program_new_psw.addr =
417                 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
418         lc->mcck_new_psw.mask =
419                 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
420         lc->mcck_new_psw.addr =
421                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
422         lc->io_new_psw.mask = psw_kernel_bits;
423         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
424         lc->clock_comparator = -1ULL;
425         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
426         lc->async_stack = (unsigned long)
427                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
428         lc->panic_stack = (unsigned long)
429                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
430         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
431         lc->thread_info = (unsigned long) &init_thread_union;
432         lc->machine_flags = S390_lowcore.machine_flags;
433 #ifndef CONFIG_64BIT
434         if (MACHINE_HAS_IEEE) {
435                 lc->extended_save_area_addr = (__u32)
436                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
437                 /* enable extended save area */
438                 __ctl_set_bit(14, 29);
439         }
440 #else
441         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
442 #endif
443         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
444         lc->async_enter_timer = S390_lowcore.async_enter_timer;
445         lc->exit_timer = S390_lowcore.exit_timer;
446         lc->user_timer = S390_lowcore.user_timer;
447         lc->system_timer = S390_lowcore.system_timer;
448         lc->steal_timer = S390_lowcore.steal_timer;
449         lc->last_update_timer = S390_lowcore.last_update_timer;
450         lc->last_update_clock = S390_lowcore.last_update_clock;
451         lc->ftrace_func = S390_lowcore.ftrace_func;
452         set_prefix((u32)(unsigned long) lc);
453         lowcore_ptr[0] = lc;
454 }
455
456 static void __init
457 setup_resources(void)
458 {
459         struct resource *res, *sub_res;
460         int i;
461
462         code_resource.start = (unsigned long) &_text;
463         code_resource.end = (unsigned long) &_etext - 1;
464         data_resource.start = (unsigned long) &_etext;
465         data_resource.end = (unsigned long) &_edata - 1;
466
467         for (i = 0; i < MEMORY_CHUNKS; i++) {
468                 if (!memory_chunk[i].size)
469                         continue;
470                 res = alloc_bootmem_low(sizeof(struct resource));
471                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
472                 switch (memory_chunk[i].type) {
473                 case CHUNK_READ_WRITE:
474                         res->name = "System RAM";
475                         break;
476                 case CHUNK_READ_ONLY:
477                         res->name = "System ROM";
478                         res->flags |= IORESOURCE_READONLY;
479                         break;
480                 default:
481                         res->name = "reserved";
482                 }
483                 res->start = memory_chunk[i].addr;
484                 res->end = memory_chunk[i].addr +  memory_chunk[i].size - 1;
485                 request_resource(&iomem_resource, res);
486
487                 if (code_resource.start >= res->start  &&
488                         code_resource.start <= res->end &&
489                         code_resource.end > res->end) {
490                         sub_res = alloc_bootmem_low(sizeof(struct resource));
491                         memcpy(sub_res, &code_resource,
492                                 sizeof(struct resource));
493                         sub_res->end = res->end;
494                         code_resource.start = res->end + 1;
495                         request_resource(res, sub_res);
496                 }
497
498                 if (code_resource.start >= res->start &&
499                         code_resource.start <= res->end &&
500                         code_resource.end <= res->end)
501                         request_resource(res, &code_resource);
502
503                 if (data_resource.start >= res->start &&
504                         data_resource.start <= res->end &&
505                         data_resource.end > res->end) {
506                         sub_res = alloc_bootmem_low(sizeof(struct resource));
507                         memcpy(sub_res, &data_resource,
508                                 sizeof(struct resource));
509                         sub_res->end = res->end;
510                         data_resource.start = res->end + 1;
511                         request_resource(res, sub_res);
512                 }
513
514                 if (data_resource.start >= res->start &&
515                         data_resource.start <= res->end &&
516                         data_resource.end <= res->end)
517                         request_resource(res, &data_resource);
518         }
519 }
520
521 unsigned long real_memory_size;
522 EXPORT_SYMBOL_GPL(real_memory_size);
523
524 static void __init setup_memory_end(void)
525 {
526         unsigned long memory_size;
527         unsigned long max_mem;
528         int i;
529
530 #ifdef CONFIG_ZFCPDUMP
531         if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
532                 memory_end = ZFCPDUMP_HSA_SIZE;
533                 memory_end_set = 1;
534         }
535 #endif
536         memory_size = 0;
537         memory_end &= PAGE_MASK;
538
539         max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
540         memory_end = min(max_mem, memory_end);
541
542         /*
543          * Make sure all chunks are MAX_ORDER aligned so we don't need the
544          * extra checks that HOLES_IN_ZONE would require.
545          */
546         for (i = 0; i < MEMORY_CHUNKS; i++) {
547                 unsigned long start, end;
548                 struct mem_chunk *chunk;
549                 unsigned long align;
550
551                 chunk = &memory_chunk[i];
552                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
553                 start = (chunk->addr + align - 1) & ~(align - 1);
554                 end = (chunk->addr + chunk->size) & ~(align - 1);
555                 if (start >= end)
556                         memset(chunk, 0, sizeof(*chunk));
557                 else {
558                         chunk->addr = start;
559                         chunk->size = end - start;
560                 }
561         }
562
563         for (i = 0; i < MEMORY_CHUNKS; i++) {
564                 struct mem_chunk *chunk = &memory_chunk[i];
565
566                 real_memory_size = max(real_memory_size,
567                                        chunk->addr + chunk->size);
568                 if (chunk->addr >= max_mem) {
569                         memset(chunk, 0, sizeof(*chunk));
570                         continue;
571                 }
572                 if (chunk->addr + chunk->size > max_mem)
573                         chunk->size = max_mem - chunk->addr;
574                 memory_size = max(memory_size, chunk->addr + chunk->size);
575         }
576         if (!memory_end)
577                 memory_end = memory_size;
578 }
579
580 static void __init
581 setup_memory(void)
582 {
583         unsigned long bootmap_size;
584         unsigned long start_pfn, end_pfn;
585         int i;
586
587         /*
588          * partially used pages are not usable - thus
589          * we are rounding upwards:
590          */
591         start_pfn = PFN_UP(__pa(&_end));
592         end_pfn = max_pfn = PFN_DOWN(memory_end);
593
594 #ifdef CONFIG_BLK_DEV_INITRD
595         /*
596          * Move the initrd in case the bitmap of the bootmem allocater
597          * would overwrite it.
598          */
599
600         if (INITRD_START && INITRD_SIZE) {
601                 unsigned long bmap_size;
602                 unsigned long start;
603
604                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
605                 bmap_size = PFN_PHYS(bmap_size);
606
607                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
608                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
609
610                         if (start + INITRD_SIZE > memory_end) {
611                                 pr_err("initrd extends beyond end of "
612                                        "memory (0x%08lx > 0x%08lx) "
613                                        "disabling initrd\n",
614                                        start + INITRD_SIZE, memory_end);
615                                 INITRD_START = INITRD_SIZE = 0;
616                         } else {
617                                 pr_info("Moving initrd (0x%08lx -> "
618                                         "0x%08lx, size: %ld)\n",
619                                         INITRD_START, start, INITRD_SIZE);
620                                 memmove((void *) start, (void *) INITRD_START,
621                                         INITRD_SIZE);
622                                 INITRD_START = start;
623                         }
624                 }
625         }
626 #endif
627
628         /*
629          * Initialize the boot-time allocator
630          */
631         bootmap_size = init_bootmem(start_pfn, end_pfn);
632
633         /*
634          * Register RAM areas with the bootmem allocator.
635          */
636
637         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
638                 unsigned long start_chunk, end_chunk, pfn;
639
640                 if (memory_chunk[i].type != CHUNK_READ_WRITE)
641                         continue;
642                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
643                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
644                 end_chunk = min(end_chunk, end_pfn);
645                 if (start_chunk >= end_chunk)
646                         continue;
647                 add_active_range(0, start_chunk, end_chunk);
648                 pfn = max(start_chunk, start_pfn);
649                 for (; pfn < end_chunk; pfn++)
650                         page_set_storage_key(PFN_PHYS(pfn), PAGE_DEFAULT_KEY);
651         }
652
653         psw_set_key(PAGE_DEFAULT_KEY);
654
655         free_bootmem_with_active_regions(0, max_pfn);
656
657         /*
658          * Reserve memory used for lowcore/command line/kernel image.
659          */
660         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
661         reserve_bootmem((unsigned long)_stext,
662                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
663                         BOOTMEM_DEFAULT);
664         /*
665          * Reserve the bootmem bitmap itself as well. We do this in two
666          * steps (first step was init_bootmem()) because this catches
667          * the (very unlikely) case of us accidentally initializing the
668          * bootmem allocator with an invalid RAM area.
669          */
670         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
671                         BOOTMEM_DEFAULT);
672
673 #ifdef CONFIG_BLK_DEV_INITRD
674         if (INITRD_START && INITRD_SIZE) {
675                 if (INITRD_START + INITRD_SIZE <= memory_end) {
676                         reserve_bootmem(INITRD_START, INITRD_SIZE,
677                                         BOOTMEM_DEFAULT);
678                         initrd_start = INITRD_START;
679                         initrd_end = initrd_start + INITRD_SIZE;
680                 } else {
681                         pr_err("initrd extends beyond end of "
682                                "memory (0x%08lx > 0x%08lx) "
683                                "disabling initrd\n",
684                                initrd_start + INITRD_SIZE, memory_end);
685                         initrd_start = initrd_end = 0;
686                 }
687         }
688 #endif
689 }
690
691 /*
692  * Setup hardware capabilities.
693  */
694 static void __init setup_hwcaps(void)
695 {
696         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
697         unsigned long long facility_list_extended;
698         unsigned int facility_list;
699         int i;
700
701         facility_list = stfl();
702         /*
703          * The store facility list bits numbers as found in the principles
704          * of operation are numbered with bit 1UL<<31 as number 0 to
705          * bit 1UL<<0 as number 31.
706          *   Bit 0: instructions named N3, "backported" to esa-mode
707          *   Bit 2: z/Architecture mode is active
708          *   Bit 7: the store-facility-list-extended facility is installed
709          *   Bit 17: the message-security assist is installed
710          *   Bit 19: the long-displacement facility is installed
711          *   Bit 21: the extended-immediate facility is installed
712          *   Bit 22: extended-translation facility 3 is installed
713          *   Bit 30: extended-translation facility 3 enhancement facility
714          * These get translated to:
715          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
716          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
717          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
718          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
719          */
720         for (i = 0; i < 6; i++)
721                 if (facility_list & (1UL << (31 - stfl_bits[i])))
722                         elf_hwcap |= 1UL << i;
723
724         if ((facility_list & (1UL << (31 - 22)))
725             && (facility_list & (1UL << (31 - 30))))
726                 elf_hwcap |= HWCAP_S390_ETF3EH;
727
728         /*
729          * Check for additional facilities with store-facility-list-extended.
730          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
731          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
732          * as stored by stfl, bits 32-xxx contain additional facilities.
733          * How many facility words are stored depends on the number of
734          * doublewords passed to the instruction. The additional facilites
735          * are:
736          *   Bit 42: decimal floating point facility is installed
737          *   Bit 44: perform floating point operation facility is installed
738          * translated to:
739          *   HWCAP_S390_DFP bit 6 (42 && 44).
740          */
741         if ((elf_hwcap & (1UL << 2)) &&
742             __stfle(&facility_list_extended, 1) > 0) {
743                 if ((facility_list_extended & (1ULL << (63 - 42)))
744                     && (facility_list_extended & (1ULL << (63 - 44))))
745                         elf_hwcap |= HWCAP_S390_DFP;
746         }
747
748         /*
749          * Huge page support HWCAP_S390_HPAGE is bit 7.
750          */
751         if (MACHINE_HAS_HPAGE)
752                 elf_hwcap |= HWCAP_S390_HPAGE;
753
754         /*
755          * 64-bit register support for 31-bit processes
756          * HWCAP_S390_HIGH_GPRS is bit 9.
757          */
758         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
759
760         switch (S390_lowcore.cpu_id.machine) {
761         case 0x9672:
762 #if !defined(CONFIG_64BIT)
763         default:        /* Use "g5" as default for 31 bit kernels. */
764 #endif
765                 strcpy(elf_platform, "g5");
766                 break;
767         case 0x2064:
768         case 0x2066:
769 #if defined(CONFIG_64BIT)
770         default:        /* Use "z900" as default for 64 bit kernels. */
771 #endif
772                 strcpy(elf_platform, "z900");
773                 break;
774         case 0x2084:
775         case 0x2086:
776                 strcpy(elf_platform, "z990");
777                 break;
778         case 0x2094:
779         case 0x2096:
780                 strcpy(elf_platform, "z9-109");
781                 break;
782         case 0x2097:
783         case 0x2098:
784                 strcpy(elf_platform, "z10");
785                 break;
786         }
787 }
788
789 /*
790  * Setup function called from init/main.c just after the banner
791  * was printed.
792  */
793
794 void __init
795 setup_arch(char **cmdline_p)
796 {
797         /*
798          * print what head.S has found out about the machine
799          */
800 #ifndef CONFIG_64BIT
801         if (MACHINE_IS_VM)
802                 pr_info("Linux is running as a z/VM "
803                         "guest operating system in 31-bit mode\n");
804         else if (MACHINE_IS_LPAR)
805                 pr_info("Linux is running natively in 31-bit mode\n");
806         if (MACHINE_HAS_IEEE)
807                 pr_info("The hardware system has IEEE compatible "
808                         "floating point units\n");
809         else
810                 pr_info("The hardware system has no IEEE compatible "
811                         "floating point units\n");
812 #else /* CONFIG_64BIT */
813         if (MACHINE_IS_VM)
814                 pr_info("Linux is running as a z/VM "
815                         "guest operating system in 64-bit mode\n");
816         else if (MACHINE_IS_KVM)
817                 pr_info("Linux is running under KVM in 64-bit mode\n");
818         else if (MACHINE_IS_LPAR)
819                 pr_info("Linux is running natively in 64-bit mode\n");
820 #endif /* CONFIG_64BIT */
821
822         /* Have one command line that is parsed and saved in /proc/cmdline */
823         /* boot_command_line has been already set up in early.c */
824         *cmdline_p = boot_command_line;
825
826         ROOT_DEV = Root_RAM0;
827
828         init_mm.start_code = PAGE_OFFSET;
829         init_mm.end_code = (unsigned long) &_etext;
830         init_mm.end_data = (unsigned long) &_edata;
831         init_mm.brk = (unsigned long) &_end;
832
833         if (MACHINE_HAS_MVCOS)
834                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
835         else
836                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
837
838         parse_early_param();
839
840         setup_ipl();
841         setup_memory_end();
842         setup_addressing_mode();
843         setup_memory();
844         setup_resources();
845         setup_lowcore();
846
847         cpu_init();
848         s390_init_cpu_topology();
849
850         /*
851          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
852          */
853         setup_hwcaps();
854
855         /*
856          * Create kernel page tables and switch to virtual addressing.
857          */
858         paging_init();
859
860         /* Setup default console */
861         conmode_default();
862         set_preferred_console();
863
864         /* Setup zfcpdump support */
865         setup_zfcpdump(console_devno);
866 }