[S390] Fix build errors (fallout from system.h disintegration)
[linux-2.6.git] / arch / s390 / kernel / setup.c
1 /*
2  *  arch/s390/kernel/setup.c
3  *
4  *  S390 version
5  *    Copyright (C) IBM Corp. 1999,2012
6  *    Author(s): Hartmut Penner (hp@de.ibm.com),
7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
8  *
9  *  Derived from "arch/i386/kernel/setup.c"
10  *    Copyright (C) 1995, Linus Torvalds
11  */
12
13 /*
14  * This file handles the architecture-dependent parts of initialization
15  */
16
17 #define KMSG_COMPONENT "setup"
18 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/sched.h>
23 #include <linux/kernel.h>
24 #include <linux/memblock.h>
25 #include <linux/mm.h>
26 #include <linux/stddef.h>
27 #include <linux/unistd.h>
28 #include <linux/ptrace.h>
29 #include <linux/user.h>
30 #include <linux/tty.h>
31 #include <linux/ioport.h>
32 #include <linux/delay.h>
33 #include <linux/init.h>
34 #include <linux/initrd.h>
35 #include <linux/bootmem.h>
36 #include <linux/root_dev.h>
37 #include <linux/console.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/device.h>
40 #include <linux/notifier.h>
41 #include <linux/pfn.h>
42 #include <linux/ctype.h>
43 #include <linux/reboot.h>
44 #include <linux/topology.h>
45 #include <linux/ftrace.h>
46 #include <linux/kexec.h>
47 #include <linux/crash_dump.h>
48 #include <linux/memory.h>
49 #include <linux/compat.h>
50
51 #include <asm/ipl.h>
52 #include <asm/uaccess.h>
53 #include <asm/facility.h>
54 #include <asm/smp.h>
55 #include <asm/mmu_context.h>
56 #include <asm/cpcmd.h>
57 #include <asm/lowcore.h>
58 #include <asm/irq.h>
59 #include <asm/page.h>
60 #include <asm/ptrace.h>
61 #include <asm/sections.h>
62 #include <asm/ebcdic.h>
63 #include <asm/kvm_virtio.h>
64 #include <asm/diag.h>
65 #include <asm/os_info.h>
66 #include "entry.h"
67
68 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
69                           PSW_MASK_EA | PSW_MASK_BA;
70 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
71                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
72                           PSW_MASK_PSTATE | PSW_ASC_HOME;
73
74 /*
75  * User copy operations.
76  */
77 struct uaccess_ops uaccess;
78 EXPORT_SYMBOL(uaccess);
79
80 /*
81  * Machine setup..
82  */
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
85
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
88
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
91
92 unsigned long elf_hwcap = 0;
93 char elf_platform[ELF_PLATFORM_SIZE];
94
95 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
96
97 int __initdata memory_end_set;
98 unsigned long __initdata memory_end;
99
100 unsigned long VMALLOC_START;
101 EXPORT_SYMBOL(VMALLOC_START);
102
103 unsigned long VMALLOC_END;
104 EXPORT_SYMBOL(VMALLOC_END);
105
106 struct page *vmemmap;
107 EXPORT_SYMBOL(vmemmap);
108
109 /* An array with a pointer to the lowcore of every CPU. */
110 struct _lowcore *lowcore_ptr[NR_CPUS];
111 EXPORT_SYMBOL(lowcore_ptr);
112
113 /*
114  * This is set up by the setup-routine at boot-time
115  * for S390 need to find out, what we have to setup
116  * using address 0x10400 ...
117  */
118
119 #include <asm/setup.h>
120
121 /*
122  * condev= and conmode= setup parameter.
123  */
124
125 static int __init condev_setup(char *str)
126 {
127         int vdev;
128
129         vdev = simple_strtoul(str, &str, 0);
130         if (vdev >= 0 && vdev < 65536) {
131                 console_devno = vdev;
132                 console_irq = -1;
133         }
134         return 1;
135 }
136
137 __setup("condev=", condev_setup);
138
139 static void __init set_preferred_console(void)
140 {
141         if (MACHINE_IS_KVM)
142                 add_preferred_console("hvc", 0, NULL);
143         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
144                 add_preferred_console("ttyS", 0, NULL);
145         else if (CONSOLE_IS_3270)
146                 add_preferred_console("tty3270", 0, NULL);
147 }
148
149 static int __init conmode_setup(char *str)
150 {
151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
152         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
153                 SET_CONSOLE_SCLP;
154 #endif
155 #if defined(CONFIG_TN3215_CONSOLE)
156         if (strncmp(str, "3215", 5) == 0)
157                 SET_CONSOLE_3215;
158 #endif
159 #if defined(CONFIG_TN3270_CONSOLE)
160         if (strncmp(str, "3270", 5) == 0)
161                 SET_CONSOLE_3270;
162 #endif
163         set_preferred_console();
164         return 1;
165 }
166
167 __setup("conmode=", conmode_setup);
168
169 static void __init conmode_default(void)
170 {
171         char query_buffer[1024];
172         char *ptr;
173
174         if (MACHINE_IS_VM) {
175                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
176                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
177                 ptr = strstr(query_buffer, "SUBCHANNEL =");
178                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
179                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
180                 ptr = strstr(query_buffer, "CONMODE");
181                 /*
182                  * Set the conmode to 3215 so that the device recognition 
183                  * will set the cu_type of the console to 3215. If the
184                  * conmode is 3270 and we don't set it back then both
185                  * 3215 and the 3270 driver will try to access the console
186                  * device (3215 as console and 3270 as normal tty).
187                  */
188                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
189                 if (ptr == NULL) {
190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
191                         SET_CONSOLE_SCLP;
192 #endif
193                         return;
194                 }
195                 if (strncmp(ptr + 8, "3270", 4) == 0) {
196 #if defined(CONFIG_TN3270_CONSOLE)
197                         SET_CONSOLE_3270;
198 #elif defined(CONFIG_TN3215_CONSOLE)
199                         SET_CONSOLE_3215;
200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
201                         SET_CONSOLE_SCLP;
202 #endif
203                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
204 #if defined(CONFIG_TN3215_CONSOLE)
205                         SET_CONSOLE_3215;
206 #elif defined(CONFIG_TN3270_CONSOLE)
207                         SET_CONSOLE_3270;
208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
209                         SET_CONSOLE_SCLP;
210 #endif
211                 }
212         } else {
213 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
214                 SET_CONSOLE_SCLP;
215 #endif
216         }
217 }
218
219 #ifdef CONFIG_ZFCPDUMP
220 static void __init setup_zfcpdump(unsigned int console_devno)
221 {
222         static char str[41];
223
224         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
225                 return;
226         if (OLDMEM_BASE)
227                 return;
228         if (console_devno != -1)
229                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
230                         ipl_info.data.fcp.dev_id.devno, console_devno);
231         else
232                 sprintf(str, " cio_ignore=all,!0.0.%04x",
233                         ipl_info.data.fcp.dev_id.devno);
234         strcat(boot_command_line, str);
235         console_loglevel = 2;
236 }
237 #else
238 static inline void setup_zfcpdump(unsigned int console_devno) {}
239 #endif /* CONFIG_ZFCPDUMP */
240
241  /*
242  * Reboot, halt and power_off stubs. They just call _machine_restart,
243  * _machine_halt or _machine_power_off. 
244  */
245
246 void machine_restart(char *command)
247 {
248         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
249                 /*
250                  * Only unblank the console if we are called in enabled
251                  * context or a bust_spinlocks cleared the way for us.
252                  */
253                 console_unblank();
254         _machine_restart(command);
255 }
256
257 void machine_halt(void)
258 {
259         if (!in_interrupt() || oops_in_progress)
260                 /*
261                  * Only unblank the console if we are called in enabled
262                  * context or a bust_spinlocks cleared the way for us.
263                  */
264                 console_unblank();
265         _machine_halt();
266 }
267
268 void machine_power_off(void)
269 {
270         if (!in_interrupt() || oops_in_progress)
271                 /*
272                  * Only unblank the console if we are called in enabled
273                  * context or a bust_spinlocks cleared the way for us.
274                  */
275                 console_unblank();
276         _machine_power_off();
277 }
278
279 /*
280  * Dummy power off function.
281  */
282 void (*pm_power_off)(void) = machine_power_off;
283
284 static int __init early_parse_mem(char *p)
285 {
286         memory_end = memparse(p, &p);
287         memory_end_set = 1;
288         return 0;
289 }
290 early_param("mem", early_parse_mem);
291
292 static int __init parse_vmalloc(char *arg)
293 {
294         if (!arg)
295                 return -EINVAL;
296         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
297         return 0;
298 }
299 early_param("vmalloc", parse_vmalloc);
300
301 unsigned int user_mode = HOME_SPACE_MODE;
302 EXPORT_SYMBOL_GPL(user_mode);
303
304 static int set_amode_primary(void)
305 {
306         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
307         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
308 #ifdef CONFIG_COMPAT
309         psw32_user_bits =
310                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
311 #endif
312
313         if (MACHINE_HAS_MVCOS) {
314                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
315                 return 1;
316         } else {
317                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
318                 return 0;
319         }
320 }
321
322 /*
323  * Switch kernel/user addressing modes?
324  */
325 static int __init early_parse_switch_amode(char *p)
326 {
327         user_mode = PRIMARY_SPACE_MODE;
328         return 0;
329 }
330 early_param("switch_amode", early_parse_switch_amode);
331
332 static int __init early_parse_user_mode(char *p)
333 {
334         if (p && strcmp(p, "primary") == 0)
335                 user_mode = PRIMARY_SPACE_MODE;
336         else if (!p || strcmp(p, "home") == 0)
337                 user_mode = HOME_SPACE_MODE;
338         else
339                 return 1;
340         return 0;
341 }
342 early_param("user_mode", early_parse_user_mode);
343
344 static void setup_addressing_mode(void)
345 {
346         if (user_mode == PRIMARY_SPACE_MODE) {
347                 if (set_amode_primary())
348                         pr_info("Address spaces switched, "
349                                 "mvcos available\n");
350                 else
351                         pr_info("Address spaces switched, "
352                                 "mvcos not available\n");
353         }
354 }
355
356 void *restart_stack __attribute__((__section__(".data")));
357
358 static void __init setup_lowcore(void)
359 {
360         struct _lowcore *lc;
361
362         /*
363          * Setup lowcore for boot cpu
364          */
365         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
366         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
367         lc->restart_psw.mask = psw_kernel_bits;
368         lc->restart_psw.addr =
369                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
370         lc->external_new_psw.mask = psw_kernel_bits |
371                 PSW_MASK_DAT | PSW_MASK_MCHECK;
372         lc->external_new_psw.addr =
373                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
374         lc->svc_new_psw.mask = psw_kernel_bits |
375                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
376         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
377         lc->program_new_psw.mask = psw_kernel_bits |
378                 PSW_MASK_DAT | PSW_MASK_MCHECK;
379         lc->program_new_psw.addr =
380                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
381         lc->mcck_new_psw.mask = psw_kernel_bits;
382         lc->mcck_new_psw.addr =
383                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
384         lc->io_new_psw.mask = psw_kernel_bits |
385                 PSW_MASK_DAT | PSW_MASK_MCHECK;
386         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
387         lc->clock_comparator = -1ULL;
388         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
389         lc->async_stack = (unsigned long)
390                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
391         lc->panic_stack = (unsigned long)
392                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
393         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
394         lc->thread_info = (unsigned long) &init_thread_union;
395         lc->machine_flags = S390_lowcore.machine_flags;
396         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
397         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
398                MAX_FACILITY_BIT/8);
399 #ifndef CONFIG_64BIT
400         if (MACHINE_HAS_IEEE) {
401                 lc->extended_save_area_addr = (__u32)
402                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
403                 /* enable extended save area */
404                 __ctl_set_bit(14, 29);
405         }
406 #else
407         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
408 #endif
409         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
410         lc->async_enter_timer = S390_lowcore.async_enter_timer;
411         lc->exit_timer = S390_lowcore.exit_timer;
412         lc->user_timer = S390_lowcore.user_timer;
413         lc->system_timer = S390_lowcore.system_timer;
414         lc->steal_timer = S390_lowcore.steal_timer;
415         lc->last_update_timer = S390_lowcore.last_update_timer;
416         lc->last_update_clock = S390_lowcore.last_update_clock;
417         lc->ftrace_func = S390_lowcore.ftrace_func;
418
419         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
420         restart_stack += ASYNC_SIZE;
421
422         /*
423          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
424          * restart data to the absolute zero lowcore. This is necesary if
425          * PSW restart is done on an offline CPU that has lowcore zero.
426          */
427         lc->restart_stack = (unsigned long) restart_stack;
428         lc->restart_fn = (unsigned long) do_restart;
429         lc->restart_data = 0;
430         lc->restart_source = -1UL;
431         memcpy(&S390_lowcore.restart_stack, &lc->restart_stack,
432                4*sizeof(unsigned long));
433         copy_to_absolute_zero(&S390_lowcore.restart_psw,
434                               &lc->restart_psw, sizeof(psw_t));
435
436         set_prefix((u32)(unsigned long) lc);
437         lowcore_ptr[0] = lc;
438 }
439
440 static struct resource code_resource = {
441         .name  = "Kernel code",
442         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
443 };
444
445 static struct resource data_resource = {
446         .name = "Kernel data",
447         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
448 };
449
450 static struct resource bss_resource = {
451         .name = "Kernel bss",
452         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
453 };
454
455 static struct resource __initdata *standard_resources[] = {
456         &code_resource,
457         &data_resource,
458         &bss_resource,
459 };
460
461 static void __init setup_resources(void)
462 {
463         struct resource *res, *std_res, *sub_res;
464         int i, j;
465
466         code_resource.start = (unsigned long) &_text;
467         code_resource.end = (unsigned long) &_etext - 1;
468         data_resource.start = (unsigned long) &_etext;
469         data_resource.end = (unsigned long) &_edata - 1;
470         bss_resource.start = (unsigned long) &__bss_start;
471         bss_resource.end = (unsigned long) &__bss_stop - 1;
472
473         for (i = 0; i < MEMORY_CHUNKS; i++) {
474                 if (!memory_chunk[i].size)
475                         continue;
476                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
477                     memory_chunk[i].type == CHUNK_CRASHK)
478                         continue;
479                 res = alloc_bootmem_low(sizeof(*res));
480                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
481                 switch (memory_chunk[i].type) {
482                 case CHUNK_READ_WRITE:
483                 case CHUNK_CRASHK:
484                         res->name = "System RAM";
485                         break;
486                 case CHUNK_READ_ONLY:
487                         res->name = "System ROM";
488                         res->flags |= IORESOURCE_READONLY;
489                         break;
490                 default:
491                         res->name = "reserved";
492                 }
493                 res->start = memory_chunk[i].addr;
494                 res->end = res->start + memory_chunk[i].size - 1;
495                 request_resource(&iomem_resource, res);
496
497                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
498                         std_res = standard_resources[j];
499                         if (std_res->start < res->start ||
500                             std_res->start > res->end)
501                                 continue;
502                         if (std_res->end > res->end) {
503                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
504                                 *sub_res = *std_res;
505                                 sub_res->end = res->end;
506                                 std_res->start = res->end + 1;
507                                 request_resource(res, sub_res);
508                         } else {
509                                 request_resource(res, std_res);
510                         }
511                 }
512         }
513 }
514
515 unsigned long real_memory_size;
516 EXPORT_SYMBOL_GPL(real_memory_size);
517
518 static void __init setup_memory_end(void)
519 {
520         unsigned long vmax, vmalloc_size, tmp;
521         int i;
522
523
524 #ifdef CONFIG_ZFCPDUMP
525         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
526                 memory_end = ZFCPDUMP_HSA_SIZE;
527                 memory_end_set = 1;
528         }
529 #endif
530         real_memory_size = 0;
531         memory_end &= PAGE_MASK;
532
533         /*
534          * Make sure all chunks are MAX_ORDER aligned so we don't need the
535          * extra checks that HOLES_IN_ZONE would require.
536          */
537         for (i = 0; i < MEMORY_CHUNKS; i++) {
538                 unsigned long start, end;
539                 struct mem_chunk *chunk;
540                 unsigned long align;
541
542                 chunk = &memory_chunk[i];
543                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
544                 start = (chunk->addr + align - 1) & ~(align - 1);
545                 end = (chunk->addr + chunk->size) & ~(align - 1);
546                 if (start >= end)
547                         memset(chunk, 0, sizeof(*chunk));
548                 else {
549                         chunk->addr = start;
550                         chunk->size = end - start;
551                 }
552                 real_memory_size = max(real_memory_size,
553                                        chunk->addr + chunk->size);
554         }
555
556         /* Choose kernel address space layout: 2, 3, or 4 levels. */
557 #ifdef CONFIG_64BIT
558         vmalloc_size = VMALLOC_END ?: 128UL << 30;
559         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
560         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
561         if (tmp <= (1UL << 42))
562                 vmax = 1UL << 42;       /* 3-level kernel page table */
563         else
564                 vmax = 1UL << 53;       /* 4-level kernel page table */
565 #else
566         vmalloc_size = VMALLOC_END ?: 96UL << 20;
567         vmax = 1UL << 31;               /* 2-level kernel page table */
568 #endif
569         /* vmalloc area is at the end of the kernel address space. */
570         VMALLOC_END = vmax;
571         VMALLOC_START = vmax - vmalloc_size;
572
573         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
574         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
575         tmp = VMALLOC_START - tmp * sizeof(struct page);
576         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
577         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
578         vmemmap = (struct page *) tmp;
579
580         /* Take care that memory_end is set and <= vmemmap */
581         memory_end = min(memory_end ?: real_memory_size, tmp);
582
583         /* Fixup memory chunk array to fit into 0..memory_end */
584         for (i = 0; i < MEMORY_CHUNKS; i++) {
585                 struct mem_chunk *chunk = &memory_chunk[i];
586
587                 if (chunk->addr >= memory_end) {
588                         memset(chunk, 0, sizeof(*chunk));
589                         continue;
590                 }
591                 if (chunk->addr + chunk->size > memory_end)
592                         chunk->size = memory_end - chunk->addr;
593         }
594 }
595
596 static void __init setup_vmcoreinfo(void)
597 {
598 #ifdef CONFIG_KEXEC
599         unsigned long ptr = paddr_vmcoreinfo_note();
600
601         copy_to_absolute_zero(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
602 #endif
603 }
604
605 #ifdef CONFIG_CRASH_DUMP
606
607 /*
608  * Find suitable location for crashkernel memory
609  */
610 static unsigned long __init find_crash_base(unsigned long crash_size,
611                                             char **msg)
612 {
613         unsigned long crash_base;
614         struct mem_chunk *chunk;
615         int i;
616
617         if (memory_chunk[0].size < crash_size) {
618                 *msg = "first memory chunk must be at least crashkernel size";
619                 return 0;
620         }
621         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
622                 return OLDMEM_BASE;
623
624         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
625                 chunk = &memory_chunk[i];
626                 if (chunk->size == 0)
627                         continue;
628                 if (chunk->type != CHUNK_READ_WRITE)
629                         continue;
630                 if (chunk->size < crash_size)
631                         continue;
632                 crash_base = (chunk->addr + chunk->size) - crash_size;
633                 if (crash_base < crash_size)
634                         continue;
635                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
636                         continue;
637                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
638                         continue;
639                 return crash_base;
640         }
641         *msg = "no suitable area found";
642         return 0;
643 }
644
645 /*
646  * Check if crash_base and crash_size is valid
647  */
648 static int __init verify_crash_base(unsigned long crash_base,
649                                     unsigned long crash_size,
650                                     char **msg)
651 {
652         struct mem_chunk *chunk;
653         int i;
654
655         /*
656          * Because we do the swap to zero, we must have at least 'crash_size'
657          * bytes free space before crash_base
658          */
659         if (crash_size > crash_base) {
660                 *msg = "crashkernel offset must be greater than size";
661                 return -EINVAL;
662         }
663
664         /* First memory chunk must be at least crash_size */
665         if (memory_chunk[0].size < crash_size) {
666                 *msg = "first memory chunk must be at least crashkernel size";
667                 return -EINVAL;
668         }
669         /* Check if we fit into the respective memory chunk */
670         for (i = 0; i < MEMORY_CHUNKS; i++) {
671                 chunk = &memory_chunk[i];
672                 if (chunk->size == 0)
673                         continue;
674                 if (crash_base < chunk->addr)
675                         continue;
676                 if (crash_base >= chunk->addr + chunk->size)
677                         continue;
678                 /* we have found the memory chunk */
679                 if (crash_base + crash_size > chunk->addr + chunk->size) {
680                         *msg = "selected memory chunk is too small for "
681                                 "crashkernel memory";
682                         return -EINVAL;
683                 }
684                 return 0;
685         }
686         *msg = "invalid memory range specified";
687         return -EINVAL;
688 }
689
690 /*
691  * Reserve kdump memory by creating a memory hole in the mem_chunk array
692  */
693 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
694                                          int type)
695 {
696         create_mem_hole(memory_chunk, addr, size, type);
697 }
698
699 /*
700  * When kdump is enabled, we have to ensure that no memory from
701  * the area [0 - crashkernel memory size] and
702  * [crashk_res.start - crashk_res.end] is set offline.
703  */
704 static int kdump_mem_notifier(struct notifier_block *nb,
705                               unsigned long action, void *data)
706 {
707         struct memory_notify *arg = data;
708
709         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
710                 return NOTIFY_BAD;
711         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
712                 return NOTIFY_OK;
713         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
714                 return NOTIFY_OK;
715         return NOTIFY_BAD;
716 }
717
718 static struct notifier_block kdump_mem_nb = {
719         .notifier_call = kdump_mem_notifier,
720 };
721
722 #endif
723
724 /*
725  * Make sure that oldmem, where the dump is stored, is protected
726  */
727 static void reserve_oldmem(void)
728 {
729 #ifdef CONFIG_CRASH_DUMP
730         if (!OLDMEM_BASE)
731                 return;
732
733         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
734         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
735                               CHUNK_OLDMEM);
736         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
737                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
738         else
739                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
740 #endif
741 }
742
743 /*
744  * Reserve memory for kdump kernel to be loaded with kexec
745  */
746 static void __init reserve_crashkernel(void)
747 {
748 #ifdef CONFIG_CRASH_DUMP
749         unsigned long long crash_base, crash_size;
750         char *msg = NULL;
751         int rc;
752
753         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
754                                &crash_base);
755         if (rc || crash_size == 0)
756                 return;
757         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
758         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
759         if (register_memory_notifier(&kdump_mem_nb))
760                 return;
761         if (!crash_base)
762                 crash_base = find_crash_base(crash_size, &msg);
763         if (!crash_base) {
764                 pr_info("crashkernel reservation failed: %s\n", msg);
765                 unregister_memory_notifier(&kdump_mem_nb);
766                 return;
767         }
768         if (verify_crash_base(crash_base, crash_size, &msg)) {
769                 pr_info("crashkernel reservation failed: %s\n", msg);
770                 unregister_memory_notifier(&kdump_mem_nb);
771                 return;
772         }
773         if (!OLDMEM_BASE && MACHINE_IS_VM)
774                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
775         crashk_res.start = crash_base;
776         crashk_res.end = crash_base + crash_size - 1;
777         insert_resource(&iomem_resource, &crashk_res);
778         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
779         pr_info("Reserving %lluMB of memory at %lluMB "
780                 "for crashkernel (System RAM: %luMB)\n",
781                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
782         os_info_crashkernel_add(crash_base, crash_size);
783 #endif
784 }
785
786 static void __init setup_memory(void)
787 {
788         unsigned long bootmap_size;
789         unsigned long start_pfn, end_pfn;
790         int i;
791
792         /*
793          * partially used pages are not usable - thus
794          * we are rounding upwards:
795          */
796         start_pfn = PFN_UP(__pa(&_end));
797         end_pfn = max_pfn = PFN_DOWN(memory_end);
798
799 #ifdef CONFIG_BLK_DEV_INITRD
800         /*
801          * Move the initrd in case the bitmap of the bootmem allocater
802          * would overwrite it.
803          */
804
805         if (INITRD_START && INITRD_SIZE) {
806                 unsigned long bmap_size;
807                 unsigned long start;
808
809                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
810                 bmap_size = PFN_PHYS(bmap_size);
811
812                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
813                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
814
815 #ifdef CONFIG_CRASH_DUMP
816                         if (OLDMEM_BASE) {
817                                 /* Move initrd behind kdump oldmem */
818                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
819                                     start < OLDMEM_BASE + OLDMEM_SIZE)
820                                         start = OLDMEM_BASE + OLDMEM_SIZE;
821                         }
822 #endif
823                         if (start + INITRD_SIZE > memory_end) {
824                                 pr_err("initrd extends beyond end of "
825                                        "memory (0x%08lx > 0x%08lx) "
826                                        "disabling initrd\n",
827                                        start + INITRD_SIZE, memory_end);
828                                 INITRD_START = INITRD_SIZE = 0;
829                         } else {
830                                 pr_info("Moving initrd (0x%08lx -> "
831                                         "0x%08lx, size: %ld)\n",
832                                         INITRD_START, start, INITRD_SIZE);
833                                 memmove((void *) start, (void *) INITRD_START,
834                                         INITRD_SIZE);
835                                 INITRD_START = start;
836                         }
837                 }
838         }
839 #endif
840
841         /*
842          * Initialize the boot-time allocator
843          */
844         bootmap_size = init_bootmem(start_pfn, end_pfn);
845
846         /*
847          * Register RAM areas with the bootmem allocator.
848          */
849
850         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
851                 unsigned long start_chunk, end_chunk, pfn;
852
853                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
854                     memory_chunk[i].type != CHUNK_CRASHK)
855                         continue;
856                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
857                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
858                 end_chunk = min(end_chunk, end_pfn);
859                 if (start_chunk >= end_chunk)
860                         continue;
861                 memblock_add_node(PFN_PHYS(start_chunk),
862                                   PFN_PHYS(end_chunk - start_chunk), 0);
863                 pfn = max(start_chunk, start_pfn);
864                 for (; pfn < end_chunk; pfn++)
865                         page_set_storage_key(PFN_PHYS(pfn),
866                                              PAGE_DEFAULT_KEY, 0);
867         }
868
869         psw_set_key(PAGE_DEFAULT_KEY);
870
871         free_bootmem_with_active_regions(0, max_pfn);
872
873         /*
874          * Reserve memory used for lowcore/command line/kernel image.
875          */
876         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
877         reserve_bootmem((unsigned long)_stext,
878                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
879                         BOOTMEM_DEFAULT);
880         /*
881          * Reserve the bootmem bitmap itself as well. We do this in two
882          * steps (first step was init_bootmem()) because this catches
883          * the (very unlikely) case of us accidentally initializing the
884          * bootmem allocator with an invalid RAM area.
885          */
886         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
887                         BOOTMEM_DEFAULT);
888
889 #ifdef CONFIG_CRASH_DUMP
890         if (crashk_res.start)
891                 reserve_bootmem(crashk_res.start,
892                                 crashk_res.end - crashk_res.start + 1,
893                                 BOOTMEM_DEFAULT);
894         if (is_kdump_kernel())
895                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
896                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
897 #endif
898 #ifdef CONFIG_BLK_DEV_INITRD
899         if (INITRD_START && INITRD_SIZE) {
900                 if (INITRD_START + INITRD_SIZE <= memory_end) {
901                         reserve_bootmem(INITRD_START, INITRD_SIZE,
902                                         BOOTMEM_DEFAULT);
903                         initrd_start = INITRD_START;
904                         initrd_end = initrd_start + INITRD_SIZE;
905                 } else {
906                         pr_err("initrd extends beyond end of "
907                                "memory (0x%08lx > 0x%08lx) "
908                                "disabling initrd\n",
909                                initrd_start + INITRD_SIZE, memory_end);
910                         initrd_start = initrd_end = 0;
911                 }
912         }
913 #endif
914 }
915
916 /*
917  * Setup hardware capabilities.
918  */
919 static void __init setup_hwcaps(void)
920 {
921         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
922         struct cpuid cpu_id;
923         int i;
924
925         /*
926          * The store facility list bits numbers as found in the principles
927          * of operation are numbered with bit 1UL<<31 as number 0 to
928          * bit 1UL<<0 as number 31.
929          *   Bit 0: instructions named N3, "backported" to esa-mode
930          *   Bit 2: z/Architecture mode is active
931          *   Bit 7: the store-facility-list-extended facility is installed
932          *   Bit 17: the message-security assist is installed
933          *   Bit 19: the long-displacement facility is installed
934          *   Bit 21: the extended-immediate facility is installed
935          *   Bit 22: extended-translation facility 3 is installed
936          *   Bit 30: extended-translation facility 3 enhancement facility
937          * These get translated to:
938          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
939          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
940          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
941          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
942          */
943         for (i = 0; i < 6; i++)
944                 if (test_facility(stfl_bits[i]))
945                         elf_hwcap |= 1UL << i;
946
947         if (test_facility(22) && test_facility(30))
948                 elf_hwcap |= HWCAP_S390_ETF3EH;
949
950         /*
951          * Check for additional facilities with store-facility-list-extended.
952          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
953          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
954          * as stored by stfl, bits 32-xxx contain additional facilities.
955          * How many facility words are stored depends on the number of
956          * doublewords passed to the instruction. The additional facilities
957          * are:
958          *   Bit 42: decimal floating point facility is installed
959          *   Bit 44: perform floating point operation facility is installed
960          * translated to:
961          *   HWCAP_S390_DFP bit 6 (42 && 44).
962          */
963         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
964                 elf_hwcap |= HWCAP_S390_DFP;
965
966         /*
967          * Huge page support HWCAP_S390_HPAGE is bit 7.
968          */
969         if (MACHINE_HAS_HPAGE)
970                 elf_hwcap |= HWCAP_S390_HPAGE;
971
972         /*
973          * 64-bit register support for 31-bit processes
974          * HWCAP_S390_HIGH_GPRS is bit 9.
975          */
976         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
977
978         get_cpu_id(&cpu_id);
979         switch (cpu_id.machine) {
980         case 0x9672:
981 #if !defined(CONFIG_64BIT)
982         default:        /* Use "g5" as default for 31 bit kernels. */
983 #endif
984                 strcpy(elf_platform, "g5");
985                 break;
986         case 0x2064:
987         case 0x2066:
988 #if defined(CONFIG_64BIT)
989         default:        /* Use "z900" as default for 64 bit kernels. */
990 #endif
991                 strcpy(elf_platform, "z900");
992                 break;
993         case 0x2084:
994         case 0x2086:
995                 strcpy(elf_platform, "z990");
996                 break;
997         case 0x2094:
998         case 0x2096:
999                 strcpy(elf_platform, "z9-109");
1000                 break;
1001         case 0x2097:
1002         case 0x2098:
1003                 strcpy(elf_platform, "z10");
1004                 break;
1005         case 0x2817:
1006         case 0x2818:
1007                 strcpy(elf_platform, "z196");
1008                 break;
1009         }
1010 }
1011
1012 /*
1013  * Setup function called from init/main.c just after the banner
1014  * was printed.
1015  */
1016
1017 void __init setup_arch(char **cmdline_p)
1018 {
1019         /*
1020          * print what head.S has found out about the machine
1021          */
1022 #ifndef CONFIG_64BIT
1023         if (MACHINE_IS_VM)
1024                 pr_info("Linux is running as a z/VM "
1025                         "guest operating system in 31-bit mode\n");
1026         else if (MACHINE_IS_LPAR)
1027                 pr_info("Linux is running natively in 31-bit mode\n");
1028         if (MACHINE_HAS_IEEE)
1029                 pr_info("The hardware system has IEEE compatible "
1030                         "floating point units\n");
1031         else
1032                 pr_info("The hardware system has no IEEE compatible "
1033                         "floating point units\n");
1034 #else /* CONFIG_64BIT */
1035         if (MACHINE_IS_VM)
1036                 pr_info("Linux is running as a z/VM "
1037                         "guest operating system in 64-bit mode\n");
1038         else if (MACHINE_IS_KVM)
1039                 pr_info("Linux is running under KVM in 64-bit mode\n");
1040         else if (MACHINE_IS_LPAR)
1041                 pr_info("Linux is running natively in 64-bit mode\n");
1042 #endif /* CONFIG_64BIT */
1043
1044         /* Have one command line that is parsed and saved in /proc/cmdline */
1045         /* boot_command_line has been already set up in early.c */
1046         *cmdline_p = boot_command_line;
1047
1048         ROOT_DEV = Root_RAM0;
1049
1050         init_mm.start_code = PAGE_OFFSET;
1051         init_mm.end_code = (unsigned long) &_etext;
1052         init_mm.end_data = (unsigned long) &_edata;
1053         init_mm.brk = (unsigned long) &_end;
1054
1055         if (MACHINE_HAS_MVCOS)
1056                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1057         else
1058                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1059
1060         parse_early_param();
1061
1062         os_info_init();
1063         setup_ipl();
1064         setup_memory_end();
1065         setup_addressing_mode();
1066         reserve_oldmem();
1067         reserve_crashkernel();
1068         setup_memory();
1069         setup_resources();
1070         setup_vmcoreinfo();
1071         setup_lowcore();
1072
1073         cpu_init();
1074         s390_init_cpu_topology();
1075
1076         /*
1077          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1078          */
1079         setup_hwcaps();
1080
1081         /*
1082          * Create kernel page tables and switch to virtual addressing.
1083          */
1084         paging_init();
1085
1086         /* Setup default console */
1087         conmode_default();
1088         set_preferred_console();
1089
1090         /* Setup zfcpdump support */
1091         setup_zfcpdump(console_devno);
1092 }