Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * S390 version
4 * Copyright IBM Corp. 1999, 2012
5 * Author(s): Hartmut Penner (hp@de.ibm.com),
6 * Martin Schwidefsky (schwidefsky@de.ibm.com)
7 *
8 * Derived from "arch/i386/kernel/setup.c"
9 * Copyright (C) 1995, Linus Torvalds
10 */
11
12/*
13 * This file handles the architecture-dependent parts of initialization
14 */
15
16#define KMSG_COMPONENT "setup"
17#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18
19#include <linux/errno.h>
20#include <linux/export.h>
21#include <linux/sched.h>
22#include <linux/sched/task.h>
23#include <linux/cpu.h>
24#include <linux/kernel.h>
25#include <linux/memblock.h>
26#include <linux/mm.h>
27#include <linux/stddef.h>
28#include <linux/unistd.h>
29#include <linux/ptrace.h>
30#include <linux/random.h>
31#include <linux/user.h>
32#include <linux/tty.h>
33#include <linux/ioport.h>
34#include <linux/delay.h>
35#include <linux/init.h>
36#include <linux/initrd.h>
37#include <linux/root_dev.h>
38#include <linux/console.h>
39#include <linux/kernel_stat.h>
40#include <linux/dma-map-ops.h>
41#include <linux/device.h>
42#include <linux/notifier.h>
43#include <linux/pfn.h>
44#include <linux/ctype.h>
45#include <linux/reboot.h>
46#include <linux/topology.h>
47#include <linux/kexec.h>
48#include <linux/crash_dump.h>
49#include <linux/memory.h>
50#include <linux/compat.h>
51#include <linux/start_kernel.h>
52#include <linux/hugetlb.h>
53#include <linux/kmemleak.h>
54
55#include <asm/archrandom.h>
56#include <asm/boot_data.h>
57#include <asm/ipl.h>
58#include <asm/facility.h>
59#include <asm/smp.h>
60#include <asm/mmu_context.h>
61#include <asm/cpcmd.h>
62#include <asm/abs_lowcore.h>
63#include <asm/nmi.h>
64#include <asm/irq.h>
65#include <asm/page.h>
66#include <asm/ptrace.h>
67#include <asm/sections.h>
68#include <asm/ebcdic.h>
69#include <asm/diag.h>
70#include <asm/os_info.h>
71#include <asm/sclp.h>
72#include <asm/stacktrace.h>
73#include <asm/sysinfo.h>
74#include <asm/numa.h>
75#include <asm/alternative.h>
76#include <asm/nospec-branch.h>
77#include <asm/physmem_info.h>
78#include <asm/maccess.h>
79#include <asm/uv.h>
80#include <asm/asm-offsets.h>
81#include "entry.h"
82
83/*
84 * Machine setup..
85 */
86unsigned int console_mode = 0;
87EXPORT_SYMBOL(console_mode);
88
89unsigned int console_devno = -1;
90EXPORT_SYMBOL(console_devno);
91
92unsigned int console_irq = -1;
93EXPORT_SYMBOL(console_irq);
94
95/*
96 * Some code and data needs to stay below 2 GB, even when the kernel would be
97 * relocated above 2 GB, because it has to use 31 bit addresses.
98 * Such code and data is part of the .amode31 section.
99 */
100char __amode31_ref *__samode31 = _samode31;
101char __amode31_ref *__eamode31 = _eamode31;
102char __amode31_ref *__stext_amode31 = _stext_amode31;
103char __amode31_ref *__etext_amode31 = _etext_amode31;
104struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
105struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
106
107/*
108 * Control registers CR2, CR5 and CR15 are initialized with addresses
109 * of tables that must be placed below 2G which is handled by the AMODE31
110 * sections.
111 * Because the AMODE31 sections are relocated below 2G at startup,
112 * the content of control registers CR2, CR5 and CR15 must be updated
113 * with new addresses after the relocation. The initial initialization of
114 * control registers occurs in head64.S and then gets updated again after AMODE31
115 * relocation. We must access the relevant AMODE31 tables indirectly via
116 * pointers placed in the .amode31.refs linker section. Those pointers get
117 * updated automatically during AMODE31 relocation and always contain a valid
118 * address within AMODE31 sections.
119 */
120
121static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
122
123static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
124 [1] = 0xffffffffffffffff
125};
126
127static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
128 0x80000000, 0, 0, 0,
129 0x80000000, 0, 0, 0,
130 0x80000000, 0, 0, 0,
131 0x80000000, 0, 0, 0,
132 0x80000000, 0, 0, 0,
133 0x80000000, 0, 0, 0,
134 0x80000000, 0, 0, 0,
135 0x80000000, 0, 0, 0
136};
137
138static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
139 0, 0, 0x89000000, 0,
140 0, 0, 0x8a000000, 0
141};
142
143static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
144static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
145static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
146static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
147
148unsigned long __bootdata_preserved(max_mappable);
149unsigned long __bootdata(ident_map_size);
150struct physmem_info __bootdata(physmem_info);
151
152unsigned long __bootdata_preserved(__kaslr_offset);
153int __bootdata_preserved(__kaslr_enabled);
154unsigned int __bootdata_preserved(zlib_dfltcc_support);
155EXPORT_SYMBOL(zlib_dfltcc_support);
156u64 __bootdata_preserved(stfle_fac_list[16]);
157EXPORT_SYMBOL(stfle_fac_list);
158u64 __bootdata_preserved(alt_stfle_fac_list[16]);
159struct oldmem_data __bootdata_preserved(oldmem_data);
160
161unsigned long VMALLOC_START;
162EXPORT_SYMBOL(VMALLOC_START);
163
164unsigned long VMALLOC_END;
165EXPORT_SYMBOL(VMALLOC_END);
166
167struct page *vmemmap;
168EXPORT_SYMBOL(vmemmap);
169unsigned long vmemmap_size;
170
171unsigned long MODULES_VADDR;
172unsigned long MODULES_END;
173
174/* An array with a pointer to the lowcore of every CPU. */
175struct lowcore *lowcore_ptr[NR_CPUS];
176EXPORT_SYMBOL(lowcore_ptr);
177
178DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
179
180/*
181 * The Write Back bit position in the physaddr is given by the SLPC PCI.
182 * Leaving the mask zero always uses write through which is safe
183 */
184unsigned long mio_wb_bit_mask __ro_after_init;
185
186/*
187 * This is set up by the setup-routine at boot-time
188 * for S390 need to find out, what we have to setup
189 * using address 0x10400 ...
190 */
191
192#include <asm/setup.h>
193
194/*
195 * condev= and conmode= setup parameter.
196 */
197
198static int __init condev_setup(char *str)
199{
200 int vdev;
201
202 vdev = simple_strtoul(str, &str, 0);
203 if (vdev >= 0 && vdev < 65536) {
204 console_devno = vdev;
205 console_irq = -1;
206 }
207 return 1;
208}
209
210__setup("condev=", condev_setup);
211
212static void __init set_preferred_console(void)
213{
214 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
215 add_preferred_console("ttyS", 0, NULL);
216 else if (CONSOLE_IS_3270)
217 add_preferred_console("tty3270", 0, NULL);
218 else if (CONSOLE_IS_VT220)
219 add_preferred_console("ttysclp", 0, NULL);
220 else if (CONSOLE_IS_HVC)
221 add_preferred_console("hvc", 0, NULL);
222}
223
224static int __init conmode_setup(char *str)
225{
226#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
227 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
228 SET_CONSOLE_SCLP;
229#endif
230#if defined(CONFIG_TN3215_CONSOLE)
231 if (!strcmp(str, "3215"))
232 SET_CONSOLE_3215;
233#endif
234#if defined(CONFIG_TN3270_CONSOLE)
235 if (!strcmp(str, "3270"))
236 SET_CONSOLE_3270;
237#endif
238 set_preferred_console();
239 return 1;
240}
241
242__setup("conmode=", conmode_setup);
243
244static void __init conmode_default(void)
245{
246 char query_buffer[1024];
247 char *ptr;
248
249 if (MACHINE_IS_VM) {
250 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
251 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
252 ptr = strstr(query_buffer, "SUBCHANNEL =");
253 console_irq = simple_strtoul(ptr + 13, NULL, 16);
254 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
255 ptr = strstr(query_buffer, "CONMODE");
256 /*
257 * Set the conmode to 3215 so that the device recognition
258 * will set the cu_type of the console to 3215. If the
259 * conmode is 3270 and we don't set it back then both
260 * 3215 and the 3270 driver will try to access the console
261 * device (3215 as console and 3270 as normal tty).
262 */
263 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
264 if (ptr == NULL) {
265#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
266 SET_CONSOLE_SCLP;
267#endif
268 return;
269 }
270 if (str_has_prefix(ptr + 8, "3270")) {
271#if defined(CONFIG_TN3270_CONSOLE)
272 SET_CONSOLE_3270;
273#elif defined(CONFIG_TN3215_CONSOLE)
274 SET_CONSOLE_3215;
275#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
276 SET_CONSOLE_SCLP;
277#endif
278 } else if (str_has_prefix(ptr + 8, "3215")) {
279#if defined(CONFIG_TN3215_CONSOLE)
280 SET_CONSOLE_3215;
281#elif defined(CONFIG_TN3270_CONSOLE)
282 SET_CONSOLE_3270;
283#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
284 SET_CONSOLE_SCLP;
285#endif
286 }
287 } else if (MACHINE_IS_KVM) {
288 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
289 SET_CONSOLE_VT220;
290 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
291 SET_CONSOLE_SCLP;
292 else
293 SET_CONSOLE_HVC;
294 } else {
295#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
296 SET_CONSOLE_SCLP;
297#endif
298 }
299}
300
301#ifdef CONFIG_CRASH_DUMP
302static void __init setup_zfcpdump(void)
303{
304 if (!is_ipl_type_dump())
305 return;
306 if (oldmem_data.start)
307 return;
308 strlcat(boot_command_line, " cio_ignore=all,!ipldev,!condev", COMMAND_LINE_SIZE);
309 console_loglevel = 2;
310}
311#else
312static inline void setup_zfcpdump(void) {}
313#endif /* CONFIG_CRASH_DUMP */
314
315 /*
316 * Reboot, halt and power_off stubs. They just call _machine_restart,
317 * _machine_halt or _machine_power_off.
318 */
319
320void machine_restart(char *command)
321{
322 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
323 /*
324 * Only unblank the console if we are called in enabled
325 * context or a bust_spinlocks cleared the way for us.
326 */
327 console_unblank();
328 _machine_restart(command);
329}
330
331void machine_halt(void)
332{
333 if (!in_interrupt() || oops_in_progress)
334 /*
335 * Only unblank the console if we are called in enabled
336 * context or a bust_spinlocks cleared the way for us.
337 */
338 console_unblank();
339 _machine_halt();
340}
341
342void machine_power_off(void)
343{
344 if (!in_interrupt() || oops_in_progress)
345 /*
346 * Only unblank the console if we are called in enabled
347 * context or a bust_spinlocks cleared the way for us.
348 */
349 console_unblank();
350 _machine_power_off();
351}
352
353/*
354 * Dummy power off function.
355 */
356void (*pm_power_off)(void) = machine_power_off;
357EXPORT_SYMBOL_GPL(pm_power_off);
358
359void *restart_stack;
360
361unsigned long stack_alloc(void)
362{
363#ifdef CONFIG_VMAP_STACK
364 void *ret;
365
366 ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
367 NUMA_NO_NODE, __builtin_return_address(0));
368 kmemleak_not_leak(ret);
369 return (unsigned long)ret;
370#else
371 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
372#endif
373}
374
375void stack_free(unsigned long stack)
376{
377#ifdef CONFIG_VMAP_STACK
378 vfree((void *) stack);
379#else
380 free_pages(stack, THREAD_SIZE_ORDER);
381#endif
382}
383
384static unsigned long __init stack_alloc_early(void)
385{
386 unsigned long stack;
387
388 stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
389 if (!stack) {
390 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
391 __func__, THREAD_SIZE, THREAD_SIZE);
392 }
393 return stack;
394}
395
396static void __init setup_lowcore(void)
397{
398 struct lowcore *lc, *abs_lc;
399
400 /*
401 * Setup lowcore for boot cpu
402 */
403 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
404 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
405 if (!lc)
406 panic("%s: Failed to allocate %zu bytes align=%zx\n",
407 __func__, sizeof(*lc), sizeof(*lc));
408
409 lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;
410 lc->restart_psw.addr = __pa(restart_int_handler);
411 lc->external_new_psw.mask = PSW_KERNEL_BITS;
412 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
413 lc->svc_new_psw.mask = PSW_KERNEL_BITS;
414 lc->svc_new_psw.addr = (unsigned long) system_call;
415 lc->program_new_psw.mask = PSW_KERNEL_BITS;
416 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
417 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
418 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
419 lc->io_new_psw.mask = PSW_KERNEL_BITS;
420 lc->io_new_psw.addr = (unsigned long) io_int_handler;
421 lc->clock_comparator = clock_comparator_max;
422 lc->current_task = (unsigned long)&init_task;
423 lc->lpp = LPP_MAGIC;
424 lc->machine_flags = S390_lowcore.machine_flags;
425 lc->preempt_count = S390_lowcore.preempt_count;
426 nmi_alloc_mcesa_early(&lc->mcesad);
427 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
428 lc->exit_timer = S390_lowcore.exit_timer;
429 lc->user_timer = S390_lowcore.user_timer;
430 lc->system_timer = S390_lowcore.system_timer;
431 lc->steal_timer = S390_lowcore.steal_timer;
432 lc->last_update_timer = S390_lowcore.last_update_timer;
433 lc->last_update_clock = S390_lowcore.last_update_clock;
434 /*
435 * Allocate the global restart stack which is the same for
436 * all CPUs in case *one* of them does a PSW restart.
437 */
438 restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET);
439 lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET;
440 lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET;
441 lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET;
442 lc->kernel_stack = S390_lowcore.kernel_stack;
443 /*
444 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
445 * restart data to the absolute zero lowcore. This is necessary if
446 * PSW restart is done on an offline CPU that has lowcore zero.
447 */
448 lc->restart_stack = (unsigned long) restart_stack;
449 lc->restart_fn = (unsigned long) do_restart;
450 lc->restart_data = 0;
451 lc->restart_source = -1U;
452 lc->spinlock_lockval = arch_spin_lockval(0);
453 lc->spinlock_index = 0;
454 arch_spin_lock_setup(0);
455 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
456 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
457 lc->preempt_count = PREEMPT_DISABLED;
458 lc->kernel_asce = S390_lowcore.kernel_asce;
459 lc->user_asce = S390_lowcore.user_asce;
460
461 system_ctlreg_init_save_area(lc);
462 abs_lc = get_abs_lowcore();
463 abs_lc->restart_stack = lc->restart_stack;
464 abs_lc->restart_fn = lc->restart_fn;
465 abs_lc->restart_data = lc->restart_data;
466 abs_lc->restart_source = lc->restart_source;
467 abs_lc->restart_psw = lc->restart_psw;
468 abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
469 abs_lc->program_new_psw = lc->program_new_psw;
470 abs_lc->mcesad = lc->mcesad;
471 put_abs_lowcore(abs_lc);
472
473 set_prefix(__pa(lc));
474 lowcore_ptr[0] = lc;
475 if (abs_lowcore_map(0, lowcore_ptr[0], false))
476 panic("Couldn't setup absolute lowcore");
477}
478
479static struct resource code_resource = {
480 .name = "Kernel code",
481 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
482};
483
484static struct resource data_resource = {
485 .name = "Kernel data",
486 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
487};
488
489static struct resource bss_resource = {
490 .name = "Kernel bss",
491 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
492};
493
494static struct resource __initdata *standard_resources[] = {
495 &code_resource,
496 &data_resource,
497 &bss_resource,
498};
499
500static void __init setup_resources(void)
501{
502 struct resource *res, *std_res, *sub_res;
503 phys_addr_t start, end;
504 int j;
505 u64 i;
506
507 code_resource.start = (unsigned long) _text;
508 code_resource.end = (unsigned long) _etext - 1;
509 data_resource.start = (unsigned long) _etext;
510 data_resource.end = (unsigned long) _edata - 1;
511 bss_resource.start = (unsigned long) __bss_start;
512 bss_resource.end = (unsigned long) __bss_stop - 1;
513
514 for_each_mem_range(i, &start, &end) {
515 res = memblock_alloc(sizeof(*res), 8);
516 if (!res)
517 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
518 __func__, sizeof(*res), 8);
519 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
520
521 res->name = "System RAM";
522 res->start = start;
523 /*
524 * In memblock, end points to the first byte after the
525 * range while in resources, end points to the last byte in
526 * the range.
527 */
528 res->end = end - 1;
529 request_resource(&iomem_resource, res);
530
531 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
532 std_res = standard_resources[j];
533 if (std_res->start < res->start ||
534 std_res->start > res->end)
535 continue;
536 if (std_res->end > res->end) {
537 sub_res = memblock_alloc(sizeof(*sub_res), 8);
538 if (!sub_res)
539 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
540 __func__, sizeof(*sub_res), 8);
541 *sub_res = *std_res;
542 sub_res->end = res->end;
543 std_res->start = res->end + 1;
544 request_resource(res, sub_res);
545 } else {
546 request_resource(res, std_res);
547 }
548 }
549 }
550#ifdef CONFIG_CRASH_DUMP
551 /*
552 * Re-add removed crash kernel memory as reserved memory. This makes
553 * sure it will be mapped with the identity mapping and struct pages
554 * will be created, so it can be resized later on.
555 * However add it later since the crash kernel resource should not be
556 * part of the System RAM resource.
557 */
558 if (crashk_res.end) {
559 memblock_add_node(crashk_res.start, resource_size(&crashk_res),
560 0, MEMBLOCK_NONE);
561 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
562 insert_resource(&iomem_resource, &crashk_res);
563 }
564#endif
565}
566
567static void __init setup_memory_end(void)
568{
569 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
570 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
571}
572
573#ifdef CONFIG_CRASH_DUMP
574
575/*
576 * When kdump is enabled, we have to ensure that no memory from the area
577 * [0 - crashkernel memory size] is set offline - it will be exchanged with
578 * the crashkernel memory region when kdump is triggered. The crashkernel
579 * memory region can never get offlined (pages are unmovable).
580 */
581static int kdump_mem_notifier(struct notifier_block *nb,
582 unsigned long action, void *data)
583{
584 struct memory_notify *arg = data;
585
586 if (action != MEM_GOING_OFFLINE)
587 return NOTIFY_OK;
588 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
589 return NOTIFY_BAD;
590 return NOTIFY_OK;
591}
592
593static struct notifier_block kdump_mem_nb = {
594 .notifier_call = kdump_mem_notifier,
595};
596
597#endif
598
599/*
600 * Reserve page tables created by decompressor
601 */
602static void __init reserve_pgtables(void)
603{
604 unsigned long start, end;
605 struct reserved_range *range;
606
607 for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end)
608 memblock_reserve(start, end - start);
609}
610
611/*
612 * Reserve memory for kdump kernel to be loaded with kexec
613 */
614static void __init reserve_crashkernel(void)
615{
616#ifdef CONFIG_CRASH_DUMP
617 unsigned long long crash_base, crash_size;
618 phys_addr_t low, high;
619 int rc;
620
621 rc = parse_crashkernel(boot_command_line, ident_map_size,
622 &crash_size, &crash_base, NULL, NULL);
623
624 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
625 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
626 if (rc || crash_size == 0)
627 return;
628
629 if (memblock.memory.regions[0].size < crash_size) {
630 pr_info("crashkernel reservation failed: %s\n",
631 "first memory chunk must be at least crashkernel size");
632 return;
633 }
634
635 low = crash_base ?: oldmem_data.start;
636 high = low + crash_size;
637 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
638 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
639 crash_base = low;
640 } else {
641 /* Find suitable area in free memory */
642 low = max_t(unsigned long, crash_size, sclp.hsa_size);
643 high = crash_base ? crash_base + crash_size : ULONG_MAX;
644
645 if (crash_base && crash_base < low) {
646 pr_info("crashkernel reservation failed: %s\n",
647 "crash_base too low");
648 return;
649 }
650 low = crash_base ?: low;
651 crash_base = memblock_phys_alloc_range(crash_size,
652 KEXEC_CRASH_MEM_ALIGN,
653 low, high);
654 }
655
656 if (!crash_base) {
657 pr_info("crashkernel reservation failed: %s\n",
658 "no suitable area found");
659 return;
660 }
661
662 if (register_memory_notifier(&kdump_mem_nb)) {
663 memblock_phys_free(crash_base, crash_size);
664 return;
665 }
666
667 if (!oldmem_data.start && MACHINE_IS_VM)
668 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
669 crashk_res.start = crash_base;
670 crashk_res.end = crash_base + crash_size - 1;
671 memblock_remove(crash_base, crash_size);
672 pr_info("Reserving %lluMB of memory at %lluMB "
673 "for crashkernel (System RAM: %luMB)\n",
674 crash_size >> 20, crash_base >> 20,
675 (unsigned long)memblock.memory.total_size >> 20);
676 os_info_crashkernel_add(crash_base, crash_size);
677#endif
678}
679
680/*
681 * Reserve the initrd from being used by memblock
682 */
683static void __init reserve_initrd(void)
684{
685 unsigned long addr, size;
686
687 if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size))
688 return;
689 initrd_start = (unsigned long)__va(addr);
690 initrd_end = initrd_start + size;
691 memblock_reserve(addr, size);
692}
693
694/*
695 * Reserve the memory area used to pass the certificate lists
696 */
697static void __init reserve_certificate_list(void)
698{
699 if (ipl_cert_list_addr)
700 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
701}
702
703static void __init reserve_physmem_info(void)
704{
705 unsigned long addr, size;
706
707 if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
708 memblock_reserve(addr, size);
709}
710
711static void __init free_physmem_info(void)
712{
713 unsigned long addr, size;
714
715 if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
716 memblock_phys_free(addr, size);
717}
718
719static void __init memblock_add_physmem_info(void)
720{
721 unsigned long start, end;
722 int i;
723
724 pr_debug("physmem info source: %s (%hhd)\n",
725 get_physmem_info_source(), physmem_info.info_source);
726 /* keep memblock lists close to the kernel */
727 memblock_set_bottom_up(true);
728 for_each_physmem_usable_range(i, &start, &end)
729 memblock_add(start, end - start);
730 for_each_physmem_online_range(i, &start, &end)
731 memblock_physmem_add(start, end - start);
732 memblock_set_bottom_up(false);
733 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
734}
735
736/*
737 * Reserve memory used for lowcore/command line/kernel image.
738 */
739static void __init reserve_kernel(void)
740{
741 memblock_reserve(0, STARTUP_NORMAL_OFFSET);
742 memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
743 memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
744 memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31);
745 memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
746 memblock_reserve(__pa(_stext), _end - _stext);
747}
748
749static void __init setup_memory(void)
750{
751 phys_addr_t start, end;
752 u64 i;
753
754 /*
755 * Init storage key for present memory
756 */
757 for_each_mem_range(i, &start, &end)
758 storage_key_init_range(start, end);
759
760 psw_set_key(PAGE_DEFAULT_KEY);
761}
762
763static void __init relocate_amode31_section(void)
764{
765 unsigned long amode31_size = __eamode31 - __samode31;
766 long amode31_offset, *ptr;
767
768 amode31_offset = physmem_info.reserved[RR_AMODE31].start - (unsigned long)__samode31;
769 pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
770
771 /* Move original AMODE31 section to the new one */
772 memmove((void *)physmem_info.reserved[RR_AMODE31].start, __samode31, amode31_size);
773 /* Zero out the old AMODE31 section to catch invalid accesses within it */
774 memset(__samode31, 0, amode31_size);
775
776 /* Update all AMODE31 region references */
777 for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
778 *ptr += amode31_offset;
779}
780
781/* This must be called after AMODE31 relocation */
782static void __init setup_cr(void)
783{
784 union ctlreg2 cr2;
785 union ctlreg5 cr5;
786 union ctlreg15 cr15;
787
788 __ctl_duct[1] = (unsigned long)__ctl_aste;
789 __ctl_duct[2] = (unsigned long)__ctl_aste;
790 __ctl_duct[4] = (unsigned long)__ctl_duald;
791
792 /* Update control registers CR2, CR5 and CR15 */
793 local_ctl_store(2, &cr2.reg);
794 local_ctl_store(5, &cr5.reg);
795 local_ctl_store(15, &cr15.reg);
796 cr2.ducto = (unsigned long)__ctl_duct >> 6;
797 cr5.pasteo = (unsigned long)__ctl_duct >> 6;
798 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
799 system_ctl_load(2, &cr2.reg);
800 system_ctl_load(5, &cr5.reg);
801 system_ctl_load(15, &cr15.reg);
802}
803
804/*
805 * Add system information as device randomness
806 */
807static void __init setup_randomness(void)
808{
809 struct sysinfo_3_2_2 *vmms;
810
811 vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
812 if (!vmms)
813 panic("Failed to allocate memory for sysinfo structure\n");
814 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
815 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
816 memblock_free(vmms, PAGE_SIZE);
817
818 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
819 static_branch_enable(&s390_arch_random_available);
820}
821
822/*
823 * Issue diagnose 318 to set the control program name and
824 * version codes.
825 */
826static void __init setup_control_program_code(void)
827{
828 union diag318_info diag318_info = {
829 .cpnc = CPNC_LINUX,
830 .cpvc = 0,
831 };
832
833 if (!sclp.has_diag318)
834 return;
835
836 diag_stat_inc(DIAG_STAT_X318);
837 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
838}
839
840/*
841 * Print the component list from the IPL report
842 */
843static void __init log_component_list(void)
844{
845 struct ipl_rb_component_entry *ptr, *end;
846 char *str;
847
848 if (!early_ipl_comp_list_addr)
849 return;
850 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
851 pr_info("Linux is running with Secure-IPL enabled\n");
852 else
853 pr_info("Linux is running with Secure-IPL disabled\n");
854 ptr = __va(early_ipl_comp_list_addr);
855 end = (void *) ptr + early_ipl_comp_list_size;
856 pr_info("The IPL report contains the following components:\n");
857 while (ptr < end) {
858 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
859 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
860 str = "signed, verified";
861 else
862 str = "signed, verification failed";
863 } else {
864 str = "not signed";
865 }
866 pr_info("%016llx - %016llx (%s)\n",
867 ptr->addr, ptr->addr + ptr->len, str);
868 ptr++;
869 }
870}
871
872/*
873 * Setup function called from init/main.c just after the banner
874 * was printed.
875 */
876
877void __init setup_arch(char **cmdline_p)
878{
879 /*
880 * print what head.S has found out about the machine
881 */
882 if (MACHINE_IS_VM)
883 pr_info("Linux is running as a z/VM "
884 "guest operating system in 64-bit mode\n");
885 else if (MACHINE_IS_KVM)
886 pr_info("Linux is running under KVM in 64-bit mode\n");
887 else if (MACHINE_IS_LPAR)
888 pr_info("Linux is running natively in 64-bit mode\n");
889 else
890 pr_info("Linux is running as a guest in 64-bit mode\n");
891
892 log_component_list();
893
894 /* Have one command line that is parsed and saved in /proc/cmdline */
895 /* boot_command_line has been already set up in early.c */
896 *cmdline_p = boot_command_line;
897
898 ROOT_DEV = Root_RAM0;
899
900 setup_initial_init_mm(_text, _etext, _edata, _end);
901
902 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
903 nospec_auto_detect();
904
905 jump_label_init();
906 parse_early_param();
907#ifdef CONFIG_CRASH_DUMP
908 /* Deactivate elfcorehdr= kernel parameter */
909 elfcorehdr_addr = ELFCORE_ADDR_MAX;
910#endif
911
912 os_info_init();
913 setup_ipl();
914 setup_control_program_code();
915
916 /* Do some memory reservations *before* memory is added to memblock */
917 reserve_pgtables();
918 reserve_kernel();
919 reserve_initrd();
920 reserve_certificate_list();
921 reserve_physmem_info();
922 memblock_set_current_limit(ident_map_size);
923 memblock_allow_resize();
924
925 /* Get information about *all* installed memory */
926 memblock_add_physmem_info();
927
928 free_physmem_info();
929 setup_memory_end();
930 memblock_dump_all();
931 setup_memory();
932
933 relocate_amode31_section();
934 setup_cr();
935 setup_uv();
936 dma_contiguous_reserve(ident_map_size);
937 vmcp_cma_reserve();
938 if (MACHINE_HAS_EDAT2)
939 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
940
941 reserve_crashkernel();
942#ifdef CONFIG_CRASH_DUMP
943 /*
944 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
945 * Therefore CPU and device initialization should be done afterwards.
946 */
947 smp_save_dump_secondary_cpus();
948#endif
949
950 setup_resources();
951 setup_lowcore();
952 smp_fill_possible_mask();
953 cpu_detect_mhz_feature();
954 cpu_init();
955 numa_setup();
956 smp_detect_cpus();
957 topology_init_early();
958
959 if (test_facility(193))
960 static_branch_enable(&cpu_has_bear);
961
962 /*
963 * Create kernel page tables.
964 */
965 paging_init();
966
967 /*
968 * After paging_init created the kernel page table, the new PSWs
969 * in lowcore can now run with DAT enabled.
970 */
971#ifdef CONFIG_CRASH_DUMP
972 smp_save_dump_ipl_cpu();
973#endif
974
975 /* Setup default console */
976 conmode_default();
977 set_preferred_console();
978
979 apply_alternative_instructions();
980 if (IS_ENABLED(CONFIG_EXPOLINE))
981 nospec_init_branches();
982
983 /* Setup zfcp/nvme dump support */
984 setup_zfcpdump();
985
986 /* Add system specific data to the random pool */
987 setup_randomness();
988}
1/*
2 * arch/s390/kernel/setup.c
3 *
4 * S390 version
5 * Copyright (C) IBM Corp. 1999,2010
6 * Author(s): Hartmut Penner (hp@de.ibm.com),
7 * Martin Schwidefsky (schwidefsky@de.ibm.com)
8 *
9 * Derived from "arch/i386/kernel/setup.c"
10 * Copyright (C) 1995, Linus Torvalds
11 */
12
13/*
14 * This file handles the architecture-dependent parts of initialization
15 */
16
17#define KMSG_COMPONENT "setup"
18#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19
20#include <linux/errno.h>
21#include <linux/module.h>
22#include <linux/sched.h>
23#include <linux/kernel.h>
24#include <linux/mm.h>
25#include <linux/stddef.h>
26#include <linux/unistd.h>
27#include <linux/ptrace.h>
28#include <linux/user.h>
29#include <linux/tty.h>
30#include <linux/ioport.h>
31#include <linux/delay.h>
32#include <linux/init.h>
33#include <linux/initrd.h>
34#include <linux/bootmem.h>
35#include <linux/root_dev.h>
36#include <linux/console.h>
37#include <linux/kernel_stat.h>
38#include <linux/device.h>
39#include <linux/notifier.h>
40#include <linux/pfn.h>
41#include <linux/ctype.h>
42#include <linux/reboot.h>
43#include <linux/topology.h>
44#include <linux/ftrace.h>
45
46#include <asm/ipl.h>
47#include <asm/uaccess.h>
48#include <asm/system.h>
49#include <asm/smp.h>
50#include <asm/mmu_context.h>
51#include <asm/cpcmd.h>
52#include <asm/lowcore.h>
53#include <asm/irq.h>
54#include <asm/page.h>
55#include <asm/ptrace.h>
56#include <asm/sections.h>
57#include <asm/ebcdic.h>
58#include <asm/compat.h>
59#include <asm/kvm_virtio.h>
60
61long psw_kernel_bits = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
62 PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
63long psw_user_bits = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
64 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
65 PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
66
67/*
68 * User copy operations.
69 */
70struct uaccess_ops uaccess;
71EXPORT_SYMBOL(uaccess);
72
73/*
74 * Machine setup..
75 */
76unsigned int console_mode = 0;
77EXPORT_SYMBOL(console_mode);
78
79unsigned int console_devno = -1;
80EXPORT_SYMBOL(console_devno);
81
82unsigned int console_irq = -1;
83EXPORT_SYMBOL(console_irq);
84
85unsigned long elf_hwcap = 0;
86char elf_platform[ELF_PLATFORM_SIZE];
87
88struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
89
90int __initdata memory_end_set;
91unsigned long __initdata memory_end;
92
93/* An array with a pointer to the lowcore of every CPU. */
94struct _lowcore *lowcore_ptr[NR_CPUS];
95EXPORT_SYMBOL(lowcore_ptr);
96
97/*
98 * This is set up by the setup-routine at boot-time
99 * for S390 need to find out, what we have to setup
100 * using address 0x10400 ...
101 */
102
103#include <asm/setup.h>
104
105/*
106 * condev= and conmode= setup parameter.
107 */
108
109static int __init condev_setup(char *str)
110{
111 int vdev;
112
113 vdev = simple_strtoul(str, &str, 0);
114 if (vdev >= 0 && vdev < 65536) {
115 console_devno = vdev;
116 console_irq = -1;
117 }
118 return 1;
119}
120
121__setup("condev=", condev_setup);
122
123static void __init set_preferred_console(void)
124{
125 if (MACHINE_IS_KVM)
126 add_preferred_console("hvc", 0, NULL);
127 else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
128 add_preferred_console("ttyS", 0, NULL);
129 else if (CONSOLE_IS_3270)
130 add_preferred_console("tty3270", 0, NULL);
131}
132
133static int __init conmode_setup(char *str)
134{
135#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
136 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
137 SET_CONSOLE_SCLP;
138#endif
139#if defined(CONFIG_TN3215_CONSOLE)
140 if (strncmp(str, "3215", 5) == 0)
141 SET_CONSOLE_3215;
142#endif
143#if defined(CONFIG_TN3270_CONSOLE)
144 if (strncmp(str, "3270", 5) == 0)
145 SET_CONSOLE_3270;
146#endif
147 set_preferred_console();
148 return 1;
149}
150
151__setup("conmode=", conmode_setup);
152
153static void __init conmode_default(void)
154{
155 char query_buffer[1024];
156 char *ptr;
157
158 if (MACHINE_IS_VM) {
159 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
160 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
161 ptr = strstr(query_buffer, "SUBCHANNEL =");
162 console_irq = simple_strtoul(ptr + 13, NULL, 16);
163 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
164 ptr = strstr(query_buffer, "CONMODE");
165 /*
166 * Set the conmode to 3215 so that the device recognition
167 * will set the cu_type of the console to 3215. If the
168 * conmode is 3270 and we don't set it back then both
169 * 3215 and the 3270 driver will try to access the console
170 * device (3215 as console and 3270 as normal tty).
171 */
172 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
173 if (ptr == NULL) {
174#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
175 SET_CONSOLE_SCLP;
176#endif
177 return;
178 }
179 if (strncmp(ptr + 8, "3270", 4) == 0) {
180#if defined(CONFIG_TN3270_CONSOLE)
181 SET_CONSOLE_3270;
182#elif defined(CONFIG_TN3215_CONSOLE)
183 SET_CONSOLE_3215;
184#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
185 SET_CONSOLE_SCLP;
186#endif
187 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
188#if defined(CONFIG_TN3215_CONSOLE)
189 SET_CONSOLE_3215;
190#elif defined(CONFIG_TN3270_CONSOLE)
191 SET_CONSOLE_3270;
192#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
193 SET_CONSOLE_SCLP;
194#endif
195 }
196 } else {
197#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
198 SET_CONSOLE_SCLP;
199#endif
200 }
201}
202
203#ifdef CONFIG_ZFCPDUMP
204static void __init setup_zfcpdump(unsigned int console_devno)
205{
206 static char str[41];
207
208 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
209 return;
210 if (console_devno != -1)
211 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
212 ipl_info.data.fcp.dev_id.devno, console_devno);
213 else
214 sprintf(str, " cio_ignore=all,!0.0.%04x",
215 ipl_info.data.fcp.dev_id.devno);
216 strcat(boot_command_line, str);
217 console_loglevel = 2;
218}
219#else
220static inline void setup_zfcpdump(unsigned int console_devno) {}
221#endif /* CONFIG_ZFCPDUMP */
222
223 /*
224 * Reboot, halt and power_off stubs. They just call _machine_restart,
225 * _machine_halt or _machine_power_off.
226 */
227
228void machine_restart(char *command)
229{
230 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
231 /*
232 * Only unblank the console if we are called in enabled
233 * context or a bust_spinlocks cleared the way for us.
234 */
235 console_unblank();
236 _machine_restart(command);
237}
238
239void machine_halt(void)
240{
241 if (!in_interrupt() || oops_in_progress)
242 /*
243 * Only unblank the console if we are called in enabled
244 * context or a bust_spinlocks cleared the way for us.
245 */
246 console_unblank();
247 _machine_halt();
248}
249
250void machine_power_off(void)
251{
252 if (!in_interrupt() || oops_in_progress)
253 /*
254 * Only unblank the console if we are called in enabled
255 * context or a bust_spinlocks cleared the way for us.
256 */
257 console_unblank();
258 _machine_power_off();
259}
260
261/*
262 * Dummy power off function.
263 */
264void (*pm_power_off)(void) = machine_power_off;
265
266static int __init early_parse_mem(char *p)
267{
268 memory_end = memparse(p, &p);
269 memory_end_set = 1;
270 return 0;
271}
272early_param("mem", early_parse_mem);
273
274unsigned int user_mode = HOME_SPACE_MODE;
275EXPORT_SYMBOL_GPL(user_mode);
276
277static int set_amode_and_uaccess(unsigned long user_amode,
278 unsigned long user32_amode)
279{
280 psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
281 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
282 PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
283#ifdef CONFIG_COMPAT
284 psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
285 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
286 PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
287 psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
288 PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
289 PSW32_MASK_PSTATE;
290#endif
291 psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
292 PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
293
294 if (MACHINE_HAS_MVCOS) {
295 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
296 return 1;
297 } else {
298 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
299 return 0;
300 }
301}
302
303/*
304 * Switch kernel/user addressing modes?
305 */
306static int __init early_parse_switch_amode(char *p)
307{
308 user_mode = PRIMARY_SPACE_MODE;
309 return 0;
310}
311early_param("switch_amode", early_parse_switch_amode);
312
313static int __init early_parse_user_mode(char *p)
314{
315 if (p && strcmp(p, "primary") == 0)
316 user_mode = PRIMARY_SPACE_MODE;
317 else if (!p || strcmp(p, "home") == 0)
318 user_mode = HOME_SPACE_MODE;
319 else
320 return 1;
321 return 0;
322}
323early_param("user_mode", early_parse_user_mode);
324
325static void setup_addressing_mode(void)
326{
327 if (user_mode == PRIMARY_SPACE_MODE) {
328 if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
329 pr_info("Address spaces switched, "
330 "mvcos available\n");
331 else
332 pr_info("Address spaces switched, "
333 "mvcos not available\n");
334 }
335}
336
337static void __init
338setup_lowcore(void)
339{
340 struct _lowcore *lc;
341
342 /*
343 * Setup lowcore for boot cpu
344 */
345 BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
346 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
347 lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
348 lc->restart_psw.addr =
349 PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
350 if (user_mode != HOME_SPACE_MODE)
351 lc->restart_psw.mask |= PSW_ASC_HOME;
352 lc->external_new_psw.mask = psw_kernel_bits;
353 lc->external_new_psw.addr =
354 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
355 lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
356 lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
357 lc->program_new_psw.mask = psw_kernel_bits;
358 lc->program_new_psw.addr =
359 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
360 lc->mcck_new_psw.mask =
361 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
362 lc->mcck_new_psw.addr =
363 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
364 lc->io_new_psw.mask = psw_kernel_bits;
365 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
366 lc->clock_comparator = -1ULL;
367 lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
368 lc->async_stack = (unsigned long)
369 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
370 lc->panic_stack = (unsigned long)
371 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
372 lc->current_task = (unsigned long) init_thread_union.thread_info.task;
373 lc->thread_info = (unsigned long) &init_thread_union;
374 lc->machine_flags = S390_lowcore.machine_flags;
375 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
376 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
377 MAX_FACILITY_BIT/8);
378#ifndef CONFIG_64BIT
379 if (MACHINE_HAS_IEEE) {
380 lc->extended_save_area_addr = (__u32)
381 __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
382 /* enable extended save area */
383 __ctl_set_bit(14, 29);
384 }
385#else
386 lc->cmf_hpp = -1ULL;
387 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
388#endif
389 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
390 lc->async_enter_timer = S390_lowcore.async_enter_timer;
391 lc->exit_timer = S390_lowcore.exit_timer;
392 lc->user_timer = S390_lowcore.user_timer;
393 lc->system_timer = S390_lowcore.system_timer;
394 lc->steal_timer = S390_lowcore.steal_timer;
395 lc->last_update_timer = S390_lowcore.last_update_timer;
396 lc->last_update_clock = S390_lowcore.last_update_clock;
397 lc->ftrace_func = S390_lowcore.ftrace_func;
398 set_prefix((u32)(unsigned long) lc);
399 lowcore_ptr[0] = lc;
400}
401
402static struct resource code_resource = {
403 .name = "Kernel code",
404 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
405};
406
407static struct resource data_resource = {
408 .name = "Kernel data",
409 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
410};
411
412static struct resource bss_resource = {
413 .name = "Kernel bss",
414 .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
415};
416
417static struct resource __initdata *standard_resources[] = {
418 &code_resource,
419 &data_resource,
420 &bss_resource,
421};
422
423static void __init setup_resources(void)
424{
425 struct resource *res, *std_res, *sub_res;
426 int i, j;
427
428 code_resource.start = (unsigned long) &_text;
429 code_resource.end = (unsigned long) &_etext - 1;
430 data_resource.start = (unsigned long) &_etext;
431 data_resource.end = (unsigned long) &_edata - 1;
432 bss_resource.start = (unsigned long) &__bss_start;
433 bss_resource.end = (unsigned long) &__bss_stop - 1;
434
435 for (i = 0; i < MEMORY_CHUNKS; i++) {
436 if (!memory_chunk[i].size)
437 continue;
438 res = alloc_bootmem_low(sizeof(*res));
439 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
440 switch (memory_chunk[i].type) {
441 case CHUNK_READ_WRITE:
442 res->name = "System RAM";
443 break;
444 case CHUNK_READ_ONLY:
445 res->name = "System ROM";
446 res->flags |= IORESOURCE_READONLY;
447 break;
448 default:
449 res->name = "reserved";
450 }
451 res->start = memory_chunk[i].addr;
452 res->end = res->start + memory_chunk[i].size - 1;
453 request_resource(&iomem_resource, res);
454
455 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
456 std_res = standard_resources[j];
457 if (std_res->start < res->start ||
458 std_res->start > res->end)
459 continue;
460 if (std_res->end > res->end) {
461 sub_res = alloc_bootmem_low(sizeof(*sub_res));
462 *sub_res = *std_res;
463 sub_res->end = res->end;
464 std_res->start = res->end + 1;
465 request_resource(res, sub_res);
466 } else {
467 request_resource(res, std_res);
468 }
469 }
470 }
471}
472
473unsigned long real_memory_size;
474EXPORT_SYMBOL_GPL(real_memory_size);
475
476static void __init setup_memory_end(void)
477{
478 unsigned long memory_size;
479 unsigned long max_mem;
480 int i;
481
482#ifdef CONFIG_ZFCPDUMP
483 if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
484 memory_end = ZFCPDUMP_HSA_SIZE;
485 memory_end_set = 1;
486 }
487#endif
488 memory_size = 0;
489 memory_end &= PAGE_MASK;
490
491 max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
492 memory_end = min(max_mem, memory_end);
493
494 /*
495 * Make sure all chunks are MAX_ORDER aligned so we don't need the
496 * extra checks that HOLES_IN_ZONE would require.
497 */
498 for (i = 0; i < MEMORY_CHUNKS; i++) {
499 unsigned long start, end;
500 struct mem_chunk *chunk;
501 unsigned long align;
502
503 chunk = &memory_chunk[i];
504 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
505 start = (chunk->addr + align - 1) & ~(align - 1);
506 end = (chunk->addr + chunk->size) & ~(align - 1);
507 if (start >= end)
508 memset(chunk, 0, sizeof(*chunk));
509 else {
510 chunk->addr = start;
511 chunk->size = end - start;
512 }
513 }
514
515 for (i = 0; i < MEMORY_CHUNKS; i++) {
516 struct mem_chunk *chunk = &memory_chunk[i];
517
518 real_memory_size = max(real_memory_size,
519 chunk->addr + chunk->size);
520 if (chunk->addr >= max_mem) {
521 memset(chunk, 0, sizeof(*chunk));
522 continue;
523 }
524 if (chunk->addr + chunk->size > max_mem)
525 chunk->size = max_mem - chunk->addr;
526 memory_size = max(memory_size, chunk->addr + chunk->size);
527 }
528 if (!memory_end)
529 memory_end = memory_size;
530}
531
532void *restart_stack __attribute__((__section__(".data")));
533
534/*
535 * Setup new PSW and allocate stack for PSW restart interrupt
536 */
537static void __init setup_restart_psw(void)
538{
539 psw_t psw;
540
541 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
542 restart_stack += ASYNC_SIZE;
543
544 /*
545 * Setup restart PSW for absolute zero lowcore. This is necesary
546 * if PSW restart is done on an offline CPU that has lowcore zero
547 */
548 psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
549 psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
550 copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
551}
552
553static void __init
554setup_memory(void)
555{
556 unsigned long bootmap_size;
557 unsigned long start_pfn, end_pfn;
558 int i;
559
560 /*
561 * partially used pages are not usable - thus
562 * we are rounding upwards:
563 */
564 start_pfn = PFN_UP(__pa(&_end));
565 end_pfn = max_pfn = PFN_DOWN(memory_end);
566
567#ifdef CONFIG_BLK_DEV_INITRD
568 /*
569 * Move the initrd in case the bitmap of the bootmem allocater
570 * would overwrite it.
571 */
572
573 if (INITRD_START && INITRD_SIZE) {
574 unsigned long bmap_size;
575 unsigned long start;
576
577 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
578 bmap_size = PFN_PHYS(bmap_size);
579
580 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
581 start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
582
583 if (start + INITRD_SIZE > memory_end) {
584 pr_err("initrd extends beyond end of "
585 "memory (0x%08lx > 0x%08lx) "
586 "disabling initrd\n",
587 start + INITRD_SIZE, memory_end);
588 INITRD_START = INITRD_SIZE = 0;
589 } else {
590 pr_info("Moving initrd (0x%08lx -> "
591 "0x%08lx, size: %ld)\n",
592 INITRD_START, start, INITRD_SIZE);
593 memmove((void *) start, (void *) INITRD_START,
594 INITRD_SIZE);
595 INITRD_START = start;
596 }
597 }
598 }
599#endif
600
601 /*
602 * Initialize the boot-time allocator
603 */
604 bootmap_size = init_bootmem(start_pfn, end_pfn);
605
606 /*
607 * Register RAM areas with the bootmem allocator.
608 */
609
610 for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
611 unsigned long start_chunk, end_chunk, pfn;
612
613 if (memory_chunk[i].type != CHUNK_READ_WRITE)
614 continue;
615 start_chunk = PFN_DOWN(memory_chunk[i].addr);
616 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
617 end_chunk = min(end_chunk, end_pfn);
618 if (start_chunk >= end_chunk)
619 continue;
620 add_active_range(0, start_chunk, end_chunk);
621 pfn = max(start_chunk, start_pfn);
622 for (; pfn < end_chunk; pfn++)
623 page_set_storage_key(PFN_PHYS(pfn),
624 PAGE_DEFAULT_KEY, 0);
625 }
626
627 psw_set_key(PAGE_DEFAULT_KEY);
628
629 free_bootmem_with_active_regions(0, max_pfn);
630
631 /*
632 * Reserve memory used for lowcore/command line/kernel image.
633 */
634 reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
635 reserve_bootmem((unsigned long)_stext,
636 PFN_PHYS(start_pfn) - (unsigned long)_stext,
637 BOOTMEM_DEFAULT);
638 /*
639 * Reserve the bootmem bitmap itself as well. We do this in two
640 * steps (first step was init_bootmem()) because this catches
641 * the (very unlikely) case of us accidentally initializing the
642 * bootmem allocator with an invalid RAM area.
643 */
644 reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
645 BOOTMEM_DEFAULT);
646
647#ifdef CONFIG_BLK_DEV_INITRD
648 if (INITRD_START && INITRD_SIZE) {
649 if (INITRD_START + INITRD_SIZE <= memory_end) {
650 reserve_bootmem(INITRD_START, INITRD_SIZE,
651 BOOTMEM_DEFAULT);
652 initrd_start = INITRD_START;
653 initrd_end = initrd_start + INITRD_SIZE;
654 } else {
655 pr_err("initrd extends beyond end of "
656 "memory (0x%08lx > 0x%08lx) "
657 "disabling initrd\n",
658 initrd_start + INITRD_SIZE, memory_end);
659 initrd_start = initrd_end = 0;
660 }
661 }
662#endif
663}
664
665/*
666 * Setup hardware capabilities.
667 */
668static void __init setup_hwcaps(void)
669{
670 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
671 struct cpuid cpu_id;
672 int i;
673
674 /*
675 * The store facility list bits numbers as found in the principles
676 * of operation are numbered with bit 1UL<<31 as number 0 to
677 * bit 1UL<<0 as number 31.
678 * Bit 0: instructions named N3, "backported" to esa-mode
679 * Bit 2: z/Architecture mode is active
680 * Bit 7: the store-facility-list-extended facility is installed
681 * Bit 17: the message-security assist is installed
682 * Bit 19: the long-displacement facility is installed
683 * Bit 21: the extended-immediate facility is installed
684 * Bit 22: extended-translation facility 3 is installed
685 * Bit 30: extended-translation facility 3 enhancement facility
686 * These get translated to:
687 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
688 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
689 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
690 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
691 */
692 for (i = 0; i < 6; i++)
693 if (test_facility(stfl_bits[i]))
694 elf_hwcap |= 1UL << i;
695
696 if (test_facility(22) && test_facility(30))
697 elf_hwcap |= HWCAP_S390_ETF3EH;
698
699 /*
700 * Check for additional facilities with store-facility-list-extended.
701 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
702 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
703 * as stored by stfl, bits 32-xxx contain additional facilities.
704 * How many facility words are stored depends on the number of
705 * doublewords passed to the instruction. The additional facilities
706 * are:
707 * Bit 42: decimal floating point facility is installed
708 * Bit 44: perform floating point operation facility is installed
709 * translated to:
710 * HWCAP_S390_DFP bit 6 (42 && 44).
711 */
712 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
713 elf_hwcap |= HWCAP_S390_DFP;
714
715 /*
716 * Huge page support HWCAP_S390_HPAGE is bit 7.
717 */
718 if (MACHINE_HAS_HPAGE)
719 elf_hwcap |= HWCAP_S390_HPAGE;
720
721 /*
722 * 64-bit register support for 31-bit processes
723 * HWCAP_S390_HIGH_GPRS is bit 9.
724 */
725 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
726
727 get_cpu_id(&cpu_id);
728 switch (cpu_id.machine) {
729 case 0x9672:
730#if !defined(CONFIG_64BIT)
731 default: /* Use "g5" as default for 31 bit kernels. */
732#endif
733 strcpy(elf_platform, "g5");
734 break;
735 case 0x2064:
736 case 0x2066:
737#if defined(CONFIG_64BIT)
738 default: /* Use "z900" as default for 64 bit kernels. */
739#endif
740 strcpy(elf_platform, "z900");
741 break;
742 case 0x2084:
743 case 0x2086:
744 strcpy(elf_platform, "z990");
745 break;
746 case 0x2094:
747 case 0x2096:
748 strcpy(elf_platform, "z9-109");
749 break;
750 case 0x2097:
751 case 0x2098:
752 strcpy(elf_platform, "z10");
753 break;
754 case 0x2817:
755 case 0x2818:
756 strcpy(elf_platform, "z196");
757 break;
758 }
759}
760
761/*
762 * Setup function called from init/main.c just after the banner
763 * was printed.
764 */
765
766void __init
767setup_arch(char **cmdline_p)
768{
769 /*
770 * print what head.S has found out about the machine
771 */
772#ifndef CONFIG_64BIT
773 if (MACHINE_IS_VM)
774 pr_info("Linux is running as a z/VM "
775 "guest operating system in 31-bit mode\n");
776 else if (MACHINE_IS_LPAR)
777 pr_info("Linux is running natively in 31-bit mode\n");
778 if (MACHINE_HAS_IEEE)
779 pr_info("The hardware system has IEEE compatible "
780 "floating point units\n");
781 else
782 pr_info("The hardware system has no IEEE compatible "
783 "floating point units\n");
784#else /* CONFIG_64BIT */
785 if (MACHINE_IS_VM)
786 pr_info("Linux is running as a z/VM "
787 "guest operating system in 64-bit mode\n");
788 else if (MACHINE_IS_KVM)
789 pr_info("Linux is running under KVM in 64-bit mode\n");
790 else if (MACHINE_IS_LPAR)
791 pr_info("Linux is running natively in 64-bit mode\n");
792#endif /* CONFIG_64BIT */
793
794 /* Have one command line that is parsed and saved in /proc/cmdline */
795 /* boot_command_line has been already set up in early.c */
796 *cmdline_p = boot_command_line;
797
798 ROOT_DEV = Root_RAM0;
799
800 init_mm.start_code = PAGE_OFFSET;
801 init_mm.end_code = (unsigned long) &_etext;
802 init_mm.end_data = (unsigned long) &_edata;
803 init_mm.brk = (unsigned long) &_end;
804
805 if (MACHINE_HAS_MVCOS)
806 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
807 else
808 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
809
810 parse_early_param();
811
812 setup_ipl();
813 setup_memory_end();
814 setup_addressing_mode();
815 setup_memory();
816 setup_resources();
817 setup_restart_psw();
818 setup_lowcore();
819
820 cpu_init();
821 s390_init_cpu_topology();
822
823 /*
824 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
825 */
826 setup_hwcaps();
827
828 /*
829 * Create kernel page tables and switch to virtual addressing.
830 */
831 paging_init();
832
833 /* Setup default console */
834 conmode_default();
835 set_preferred_console();
836
837 /* Setup zfcpdump support */
838 setup_zfcpdump(console_devno);
839}