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