Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Architecture-specific setup.
4 *
5 * Copyright (C) 1998-2003 Hewlett-Packard Co
6 * David Mosberger-Tang <davidm@hpl.hp.com>
7 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
8 *
9 * 2005-10-07 Keith Owens <kaos@sgi.com>
10 * Add notify_die() hooks.
11 */
12#include <linux/cpu.h>
13#include <linux/pm.h>
14#include <linux/elf.h>
15#include <linux/errno.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/slab.h>
19#include <linux/module.h>
20#include <linux/notifier.h>
21#include <linux/personality.h>
22#include <linux/sched.h>
23#include <linux/sched/debug.h>
24#include <linux/sched/hotplug.h>
25#include <linux/sched/task.h>
26#include <linux/sched/task_stack.h>
27#include <linux/stddef.h>
28#include <linux/thread_info.h>
29#include <linux/unistd.h>
30#include <linux/efi.h>
31#include <linux/interrupt.h>
32#include <linux/delay.h>
33#include <linux/kdebug.h>
34#include <linux/utsname.h>
35#include <linux/tracehook.h>
36#include <linux/rcupdate.h>
37
38#include <asm/cpu.h>
39#include <asm/delay.h>
40#include <asm/elf.h>
41#include <asm/irq.h>
42#include <asm/kexec.h>
43#include <asm/processor.h>
44#include <asm/sal.h>
45#include <asm/switch_to.h>
46#include <asm/tlbflush.h>
47#include <linux/uaccess.h>
48#include <asm/unwind.h>
49#include <asm/user.h>
50#include <asm/xtp.h>
51
52#include "entry.h"
53
54#ifdef CONFIG_PERFMON
55# include <asm/perfmon.h>
56#endif
57
58#include "sigframe.h"
59
60void (*ia64_mark_idle)(int);
61
62unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
63EXPORT_SYMBOL(boot_option_idle_override);
64void (*pm_power_off) (void);
65EXPORT_SYMBOL(pm_power_off);
66
67static void
68ia64_do_show_stack (struct unw_frame_info *info, void *arg)
69{
70 unsigned long ip, sp, bsp;
71 const char *loglvl = arg;
72
73 printk("%s\nCall Trace:\n", loglvl);
74 do {
75 unw_get_ip(info, &ip);
76 if (ip == 0)
77 break;
78
79 unw_get_sp(info, &sp);
80 unw_get_bsp(info, &bsp);
81 printk("%s [<%016lx>] %pS\n"
82 " sp=%016lx bsp=%016lx\n",
83 loglvl, ip, (void *)ip, sp, bsp);
84 } while (unw_unwind(info) >= 0);
85}
86
87void
88show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
89{
90 if (!task)
91 unw_init_running(ia64_do_show_stack, (void *)loglvl);
92 else {
93 struct unw_frame_info info;
94
95 unw_init_from_blocked_task(&info, task);
96 ia64_do_show_stack(&info, (void *)loglvl);
97 }
98}
99
100void
101show_regs (struct pt_regs *regs)
102{
103 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
104
105 print_modules();
106 printk("\n");
107 show_regs_print_info(KERN_DEFAULT);
108 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
109 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
110 init_utsname()->release);
111 printk("ip is at %pS\n", (void *)ip);
112 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
113 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
114 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
115 regs->ar_rnat, regs->ar_bspstore, regs->pr);
116 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
117 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
118 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
119 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
120 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
121 regs->f6.u.bits[1], regs->f6.u.bits[0],
122 regs->f7.u.bits[1], regs->f7.u.bits[0]);
123 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
124 regs->f8.u.bits[1], regs->f8.u.bits[0],
125 regs->f9.u.bits[1], regs->f9.u.bits[0]);
126 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
127 regs->f10.u.bits[1], regs->f10.u.bits[0],
128 regs->f11.u.bits[1], regs->f11.u.bits[0]);
129
130 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
131 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
132 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
133 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
134 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
135 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
136 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
137 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
138 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
139
140 if (user_mode(regs)) {
141 /* print the stacked registers */
142 unsigned long val, *bsp, ndirty;
143 int i, sof, is_nat = 0;
144
145 sof = regs->cr_ifs & 0x7f; /* size of frame */
146 ndirty = (regs->loadrs >> 19);
147 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
148 for (i = 0; i < sof; ++i) {
149 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
150 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
151 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
152 }
153 } else
154 show_stack(NULL, NULL, KERN_DEFAULT);
155}
156
157/* local support for deprecated console_print */
158void
159console_print(const char *s)
160{
161 printk(KERN_EMERG "%s", s);
162}
163
164void
165do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
166{
167 if (fsys_mode(current, &scr->pt)) {
168 /*
169 * defer signal-handling etc. until we return to
170 * privilege-level 0.
171 */
172 if (!ia64_psr(&scr->pt)->lp)
173 ia64_psr(&scr->pt)->lp = 1;
174 return;
175 }
176
177#ifdef CONFIG_PERFMON
178 if (current->thread.pfm_needs_checking)
179 /*
180 * Note: pfm_handle_work() allow us to call it with interrupts
181 * disabled, and may enable interrupts within the function.
182 */
183 pfm_handle_work();
184#endif
185
186 /* deal with pending signal delivery */
187 if (test_thread_flag(TIF_SIGPENDING)) {
188 local_irq_enable(); /* force interrupt enable */
189 ia64_do_signal(scr, in_syscall);
190 }
191
192 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) {
193 local_irq_enable(); /* force interrupt enable */
194 tracehook_notify_resume(&scr->pt);
195 }
196
197 /* copy user rbs to kernel rbs */
198 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
199 local_irq_enable(); /* force interrupt enable */
200 ia64_sync_krbs();
201 }
202
203 local_irq_disable(); /* force interrupt disable */
204}
205
206static int __init nohalt_setup(char * str)
207{
208 cpu_idle_poll_ctrl(true);
209 return 1;
210}
211__setup("nohalt", nohalt_setup);
212
213#ifdef CONFIG_HOTPLUG_CPU
214/* We don't actually take CPU down, just spin without interrupts. */
215static inline void play_dead(void)
216{
217 unsigned int this_cpu = smp_processor_id();
218
219 /* Ack it */
220 __this_cpu_write(cpu_state, CPU_DEAD);
221
222 max_xtp();
223 local_irq_disable();
224 idle_task_exit();
225 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
226 /*
227 * The above is a point of no-return, the processor is
228 * expected to be in SAL loop now.
229 */
230 BUG();
231}
232#else
233static inline void play_dead(void)
234{
235 BUG();
236}
237#endif /* CONFIG_HOTPLUG_CPU */
238
239void arch_cpu_idle_dead(void)
240{
241 play_dead();
242}
243
244void arch_cpu_idle(void)
245{
246 void (*mark_idle)(int) = ia64_mark_idle;
247
248#ifdef CONFIG_SMP
249 min_xtp();
250#endif
251 rmb();
252 if (mark_idle)
253 (*mark_idle)(1);
254
255 safe_halt();
256
257 if (mark_idle)
258 (*mark_idle)(0);
259#ifdef CONFIG_SMP
260 normal_xtp();
261#endif
262}
263
264void
265ia64_save_extra (struct task_struct *task)
266{
267#ifdef CONFIG_PERFMON
268 unsigned long info;
269#endif
270
271 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
272 ia64_save_debug_regs(&task->thread.dbr[0]);
273
274#ifdef CONFIG_PERFMON
275 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
276 pfm_save_regs(task);
277
278 info = __this_cpu_read(pfm_syst_info);
279 if (info & PFM_CPUINFO_SYST_WIDE)
280 pfm_syst_wide_update_task(task, info, 0);
281#endif
282}
283
284void
285ia64_load_extra (struct task_struct *task)
286{
287#ifdef CONFIG_PERFMON
288 unsigned long info;
289#endif
290
291 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
292 ia64_load_debug_regs(&task->thread.dbr[0]);
293
294#ifdef CONFIG_PERFMON
295 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
296 pfm_load_regs(task);
297
298 info = __this_cpu_read(pfm_syst_info);
299 if (info & PFM_CPUINFO_SYST_WIDE)
300 pfm_syst_wide_update_task(task, info, 1);
301#endif
302}
303
304/*
305 * Copy the state of an ia-64 thread.
306 *
307 * We get here through the following call chain:
308 *
309 * from user-level: from kernel:
310 *
311 * <clone syscall> <some kernel call frames>
312 * sys_clone :
313 * _do_fork _do_fork
314 * copy_thread copy_thread
315 *
316 * This means that the stack layout is as follows:
317 *
318 * +---------------------+ (highest addr)
319 * | struct pt_regs |
320 * +---------------------+
321 * | struct switch_stack |
322 * +---------------------+
323 * | |
324 * | memory stack |
325 * | | <-- sp (lowest addr)
326 * +---------------------+
327 *
328 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
329 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
330 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
331 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
332 * the stack is page aligned and the page size is at least 4KB, this is always the case,
333 * so there is nothing to worry about.
334 */
335int
336copy_thread(unsigned long clone_flags, unsigned long user_stack_base,
337 unsigned long user_stack_size, struct task_struct *p, unsigned long tls)
338{
339 extern char ia64_ret_from_clone;
340 struct switch_stack *child_stack, *stack;
341 unsigned long rbs, child_rbs, rbs_size;
342 struct pt_regs *child_ptregs;
343 struct pt_regs *regs = current_pt_regs();
344 int retval = 0;
345
346 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
347 child_stack = (struct switch_stack *) child_ptregs - 1;
348
349 rbs = (unsigned long) current + IA64_RBS_OFFSET;
350 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
351
352 /* copy parts of thread_struct: */
353 p->thread.ksp = (unsigned long) child_stack - 16;
354
355 /*
356 * NOTE: The calling convention considers all floating point
357 * registers in the high partition (fph) to be scratch. Since
358 * the only way to get to this point is through a system call,
359 * we know that the values in fph are all dead. Hence, there
360 * is no need to inherit the fph state from the parent to the
361 * child and all we have to do is to make sure that
362 * IA64_THREAD_FPH_VALID is cleared in the child.
363 *
364 * XXX We could push this optimization a bit further by
365 * clearing IA64_THREAD_FPH_VALID on ANY system call.
366 * However, it's not clear this is worth doing. Also, it
367 * would be a slight deviation from the normal Linux system
368 * call behavior where scratch registers are preserved across
369 * system calls (unless used by the system call itself).
370 */
371# define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
372 | IA64_THREAD_PM_VALID)
373# define THREAD_FLAGS_TO_SET 0
374 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
375 | THREAD_FLAGS_TO_SET);
376
377 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
378
379 if (unlikely(p->flags & PF_KTHREAD)) {
380 if (unlikely(!user_stack_base)) {
381 /* fork_idle() called us */
382 return 0;
383 }
384 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
385 child_stack->r4 = user_stack_base; /* payload */
386 child_stack->r5 = user_stack_size; /* argument */
387 /*
388 * Preserve PSR bits, except for bits 32-34 and 37-45,
389 * which we can't read.
390 */
391 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
392 /* mark as valid, empty frame */
393 child_ptregs->cr_ifs = 1UL << 63;
394 child_stack->ar_fpsr = child_ptregs->ar_fpsr
395 = ia64_getreg(_IA64_REG_AR_FPSR);
396 child_stack->pr = (1 << PRED_KERNEL_STACK);
397 child_stack->ar_bspstore = child_rbs;
398 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
399
400 /* stop some PSR bits from being inherited.
401 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
402 * therefore we must specify them explicitly here and not include them in
403 * IA64_PSR_BITS_TO_CLEAR.
404 */
405 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
406 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
407
408 return 0;
409 }
410 stack = ((struct switch_stack *) regs) - 1;
411 /* copy parent's switch_stack & pt_regs to child: */
412 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
413
414 /* copy the parent's register backing store to the child: */
415 rbs_size = stack->ar_bspstore - rbs;
416 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
417 if (clone_flags & CLONE_SETTLS)
418 child_ptregs->r13 = tls;
419 if (user_stack_base) {
420 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
421 child_ptregs->ar_bspstore = user_stack_base;
422 child_ptregs->ar_rnat = 0;
423 child_ptregs->loadrs = 0;
424 }
425 child_stack->ar_bspstore = child_rbs + rbs_size;
426 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
427
428 /* stop some PSR bits from being inherited.
429 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
430 * therefore we must specify them explicitly here and not include them in
431 * IA64_PSR_BITS_TO_CLEAR.
432 */
433 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
434 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
435
436#ifdef CONFIG_PERFMON
437 if (current->thread.pfm_context)
438 pfm_inherit(p, child_ptregs);
439#endif
440 return retval;
441}
442
443asmlinkage long ia64_clone(unsigned long clone_flags, unsigned long stack_start,
444 unsigned long stack_size, unsigned long parent_tidptr,
445 unsigned long child_tidptr, unsigned long tls)
446{
447 struct kernel_clone_args args = {
448 .flags = (lower_32_bits(clone_flags) & ~CSIGNAL),
449 .pidfd = (int __user *)parent_tidptr,
450 .child_tid = (int __user *)child_tidptr,
451 .parent_tid = (int __user *)parent_tidptr,
452 .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL),
453 .stack = stack_start,
454 .stack_size = stack_size,
455 .tls = tls,
456 };
457
458 return _do_fork(&args);
459}
460
461static void
462do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
463{
464 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
465 unsigned long ip;
466 elf_greg_t *dst = arg;
467 struct pt_regs *pt;
468 char nat;
469 int i;
470
471 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
472
473 if (unw_unwind_to_user(info) < 0)
474 return;
475
476 unw_get_sp(info, &sp);
477 pt = (struct pt_regs *) (sp + 16);
478
479 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
480
481 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
482 return;
483
484 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
485 &ar_rnat);
486
487 /*
488 * coredump format:
489 * r0-r31
490 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
491 * predicate registers (p0-p63)
492 * b0-b7
493 * ip cfm user-mask
494 * ar.rsc ar.bsp ar.bspstore ar.rnat
495 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
496 */
497
498 /* r0 is zero */
499 for (i = 1, mask = (1UL << i); i < 32; ++i) {
500 unw_get_gr(info, i, &dst[i], &nat);
501 if (nat)
502 nat_bits |= mask;
503 mask <<= 1;
504 }
505 dst[32] = nat_bits;
506 unw_get_pr(info, &dst[33]);
507
508 for (i = 0; i < 8; ++i)
509 unw_get_br(info, i, &dst[34 + i]);
510
511 unw_get_rp(info, &ip);
512 dst[42] = ip + ia64_psr(pt)->ri;
513 dst[43] = cfm;
514 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
515
516 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
517 /*
518 * For bsp and bspstore, unw_get_ar() would return the kernel
519 * addresses, but we need the user-level addresses instead:
520 */
521 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
522 dst[47] = pt->ar_bspstore;
523 dst[48] = ar_rnat;
524 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
525 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
526 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
527 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
528 unw_get_ar(info, UNW_AR_LC, &dst[53]);
529 unw_get_ar(info, UNW_AR_EC, &dst[54]);
530 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
531 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
532}
533
534void
535do_copy_regs (struct unw_frame_info *info, void *arg)
536{
537 do_copy_task_regs(current, info, arg);
538}
539
540void
541ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
542{
543 unw_init_running(do_copy_regs, dst);
544}
545
546/*
547 * Flush thread state. This is called when a thread does an execve().
548 */
549void
550flush_thread (void)
551{
552 /* drop floating-point and debug-register state if it exists: */
553 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
554 ia64_drop_fpu(current);
555}
556
557/*
558 * Clean up state associated with a thread. This is called when
559 * the thread calls exit().
560 */
561void
562exit_thread (struct task_struct *tsk)
563{
564
565 ia64_drop_fpu(tsk);
566#ifdef CONFIG_PERFMON
567 /* if needed, stop monitoring and flush state to perfmon context */
568 if (tsk->thread.pfm_context)
569 pfm_exit_thread(tsk);
570
571 /* free debug register resources */
572 if (tsk->thread.flags & IA64_THREAD_DBG_VALID)
573 pfm_release_debug_registers(tsk);
574#endif
575}
576
577unsigned long
578get_wchan (struct task_struct *p)
579{
580 struct unw_frame_info info;
581 unsigned long ip;
582 int count = 0;
583
584 if (!p || p == current || p->state == TASK_RUNNING)
585 return 0;
586
587 /*
588 * Note: p may not be a blocked task (it could be current or
589 * another process running on some other CPU. Rather than
590 * trying to determine if p is really blocked, we just assume
591 * it's blocked and rely on the unwind routines to fail
592 * gracefully if the process wasn't really blocked after all.
593 * --davidm 99/12/15
594 */
595 unw_init_from_blocked_task(&info, p);
596 do {
597 if (p->state == TASK_RUNNING)
598 return 0;
599 if (unw_unwind(&info) < 0)
600 return 0;
601 unw_get_ip(&info, &ip);
602 if (!in_sched_functions(ip))
603 return ip;
604 } while (count++ < 16);
605 return 0;
606}
607
608void
609cpu_halt (void)
610{
611 pal_power_mgmt_info_u_t power_info[8];
612 unsigned long min_power;
613 int i, min_power_state;
614
615 if (ia64_pal_halt_info(power_info) != 0)
616 return;
617
618 min_power_state = 0;
619 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
620 for (i = 1; i < 8; ++i)
621 if (power_info[i].pal_power_mgmt_info_s.im
622 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
623 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
624 min_power_state = i;
625 }
626
627 while (1)
628 ia64_pal_halt(min_power_state);
629}
630
631void machine_shutdown(void)
632{
633 smp_shutdown_nonboot_cpus(reboot_cpu);
634
635#ifdef CONFIG_KEXEC
636 kexec_disable_iosapic();
637#endif
638}
639
640void
641machine_restart (char *restart_cmd)
642{
643 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
644 efi_reboot(REBOOT_WARM, NULL);
645}
646
647void
648machine_halt (void)
649{
650 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
651 cpu_halt();
652}
653
654void
655machine_power_off (void)
656{
657 if (pm_power_off)
658 pm_power_off();
659 machine_halt();
660}
661
662EXPORT_SYMBOL(ia64_delay_loop);
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Architecture-specific setup.
4 *
5 * Copyright (C) 1998-2003 Hewlett-Packard Co
6 * David Mosberger-Tang <davidm@hpl.hp.com>
7 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
8 *
9 * 2005-10-07 Keith Owens <kaos@sgi.com>
10 * Add notify_die() hooks.
11 */
12#include <linux/cpu.h>
13#include <linux/pm.h>
14#include <linux/elf.h>
15#include <linux/errno.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/slab.h>
19#include <linux/module.h>
20#include <linux/notifier.h>
21#include <linux/personality.h>
22#include <linux/reboot.h>
23#include <linux/sched.h>
24#include <linux/sched/debug.h>
25#include <linux/sched/hotplug.h>
26#include <linux/sched/task.h>
27#include <linux/sched/task_stack.h>
28#include <linux/stddef.h>
29#include <linux/thread_info.h>
30#include <linux/unistd.h>
31#include <linux/efi.h>
32#include <linux/interrupt.h>
33#include <linux/delay.h>
34#include <linux/kdebug.h>
35#include <linux/utsname.h>
36#include <linux/resume_user_mode.h>
37#include <linux/rcupdate.h>
38
39#include <asm/cpu.h>
40#include <asm/delay.h>
41#include <asm/elf.h>
42#include <asm/irq.h>
43#include <asm/kexec.h>
44#include <asm/processor.h>
45#include <asm/sal.h>
46#include <asm/switch_to.h>
47#include <asm/tlbflush.h>
48#include <linux/uaccess.h>
49#include <asm/unwind.h>
50#include <asm/user.h>
51#include <asm/xtp.h>
52
53#include "entry.h"
54
55#include "sigframe.h"
56
57void (*ia64_mark_idle)(int);
58
59unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
60EXPORT_SYMBOL(boot_option_idle_override);
61void (*pm_power_off) (void);
62EXPORT_SYMBOL(pm_power_off);
63
64static void
65ia64_do_show_stack (struct unw_frame_info *info, void *arg)
66{
67 unsigned long ip, sp, bsp;
68 const char *loglvl = arg;
69
70 printk("%s\nCall Trace:\n", loglvl);
71 do {
72 unw_get_ip(info, &ip);
73 if (ip == 0)
74 break;
75
76 unw_get_sp(info, &sp);
77 unw_get_bsp(info, &bsp);
78 printk("%s [<%016lx>] %pS\n"
79 " sp=%016lx bsp=%016lx\n",
80 loglvl, ip, (void *)ip, sp, bsp);
81 } while (unw_unwind(info) >= 0);
82}
83
84void
85show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
86{
87 if (!task)
88 unw_init_running(ia64_do_show_stack, (void *)loglvl);
89 else {
90 struct unw_frame_info info;
91
92 unw_init_from_blocked_task(&info, task);
93 ia64_do_show_stack(&info, (void *)loglvl);
94 }
95}
96
97void
98show_regs (struct pt_regs *regs)
99{
100 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
101
102 print_modules();
103 printk("\n");
104 show_regs_print_info(KERN_DEFAULT);
105 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
106 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
107 init_utsname()->release);
108 printk("ip is at %pS\n", (void *)ip);
109 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
110 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
111 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
112 regs->ar_rnat, regs->ar_bspstore, regs->pr);
113 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
114 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
115 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
116 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
117 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
118 regs->f6.u.bits[1], regs->f6.u.bits[0],
119 regs->f7.u.bits[1], regs->f7.u.bits[0]);
120 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
121 regs->f8.u.bits[1], regs->f8.u.bits[0],
122 regs->f9.u.bits[1], regs->f9.u.bits[0]);
123 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
124 regs->f10.u.bits[1], regs->f10.u.bits[0],
125 regs->f11.u.bits[1], regs->f11.u.bits[0]);
126
127 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
128 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
129 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
130 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
131 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
132 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
133 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
134 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
135 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
136
137 if (user_mode(regs)) {
138 /* print the stacked registers */
139 unsigned long val, *bsp, ndirty;
140 int i, sof, is_nat = 0;
141
142 sof = regs->cr_ifs & 0x7f; /* size of frame */
143 ndirty = (regs->loadrs >> 19);
144 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
145 for (i = 0; i < sof; ++i) {
146 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
147 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
148 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
149 }
150 } else
151 show_stack(NULL, NULL, KERN_DEFAULT);
152}
153
154/* local support for deprecated console_print */
155void
156console_print(const char *s)
157{
158 printk(KERN_EMERG "%s", s);
159}
160
161void
162do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
163{
164 if (fsys_mode(current, &scr->pt)) {
165 /*
166 * defer signal-handling etc. until we return to
167 * privilege-level 0.
168 */
169 if (!ia64_psr(&scr->pt)->lp)
170 ia64_psr(&scr->pt)->lp = 1;
171 return;
172 }
173
174 /* deal with pending signal delivery */
175 if (test_thread_flag(TIF_SIGPENDING) ||
176 test_thread_flag(TIF_NOTIFY_SIGNAL)) {
177 local_irq_enable(); /* force interrupt enable */
178 ia64_do_signal(scr, in_syscall);
179 }
180
181 if (test_thread_flag(TIF_NOTIFY_RESUME)) {
182 local_irq_enable(); /* force interrupt enable */
183 resume_user_mode_work(&scr->pt);
184 }
185
186 /* copy user rbs to kernel rbs */
187 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
188 local_irq_enable(); /* force interrupt enable */
189 ia64_sync_krbs();
190 }
191
192 local_irq_disable(); /* force interrupt disable */
193}
194
195static int __init nohalt_setup(char * str)
196{
197 cpu_idle_poll_ctrl(true);
198 return 1;
199}
200__setup("nohalt", nohalt_setup);
201
202#ifdef CONFIG_HOTPLUG_CPU
203/* We don't actually take CPU down, just spin without interrupts. */
204static inline void play_dead(void)
205{
206 unsigned int this_cpu = smp_processor_id();
207
208 /* Ack it */
209 __this_cpu_write(cpu_state, CPU_DEAD);
210
211 max_xtp();
212 local_irq_disable();
213 idle_task_exit();
214 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
215 /*
216 * The above is a point of no-return, the processor is
217 * expected to be in SAL loop now.
218 */
219 BUG();
220}
221#else
222static inline void play_dead(void)
223{
224 BUG();
225}
226#endif /* CONFIG_HOTPLUG_CPU */
227
228void arch_cpu_idle_dead(void)
229{
230 play_dead();
231}
232
233void arch_cpu_idle(void)
234{
235 void (*mark_idle)(int) = ia64_mark_idle;
236
237#ifdef CONFIG_SMP
238 min_xtp();
239#endif
240 rmb();
241 if (mark_idle)
242 (*mark_idle)(1);
243
244 raw_safe_halt();
245
246 if (mark_idle)
247 (*mark_idle)(0);
248#ifdef CONFIG_SMP
249 normal_xtp();
250#endif
251}
252
253void
254ia64_save_extra (struct task_struct *task)
255{
256 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
257 ia64_save_debug_regs(&task->thread.dbr[0]);
258}
259
260void
261ia64_load_extra (struct task_struct *task)
262{
263 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
264 ia64_load_debug_regs(&task->thread.dbr[0]);
265}
266
267/*
268 * Copy the state of an ia-64 thread.
269 *
270 * We get here through the following call chain:
271 *
272 * from user-level: from kernel:
273 *
274 * <clone syscall> <some kernel call frames>
275 * sys_clone :
276 * kernel_clone kernel_clone
277 * copy_thread copy_thread
278 *
279 * This means that the stack layout is as follows:
280 *
281 * +---------------------+ (highest addr)
282 * | struct pt_regs |
283 * +---------------------+
284 * | struct switch_stack |
285 * +---------------------+
286 * | |
287 * | memory stack |
288 * | | <-- sp (lowest addr)
289 * +---------------------+
290 *
291 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
292 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
293 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
294 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
295 * the stack is page aligned and the page size is at least 4KB, this is always the case,
296 * so there is nothing to worry about.
297 */
298int
299copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
300{
301 unsigned long clone_flags = args->flags;
302 unsigned long user_stack_base = args->stack;
303 unsigned long user_stack_size = args->stack_size;
304 unsigned long tls = args->tls;
305 extern char ia64_ret_from_clone;
306 struct switch_stack *child_stack, *stack;
307 unsigned long rbs, child_rbs, rbs_size;
308 struct pt_regs *child_ptregs;
309 struct pt_regs *regs = current_pt_regs();
310 int retval = 0;
311
312 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
313 child_stack = (struct switch_stack *) child_ptregs - 1;
314
315 rbs = (unsigned long) current + IA64_RBS_OFFSET;
316 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
317
318 /* copy parts of thread_struct: */
319 p->thread.ksp = (unsigned long) child_stack - 16;
320
321 /*
322 * NOTE: The calling convention considers all floating point
323 * registers in the high partition (fph) to be scratch. Since
324 * the only way to get to this point is through a system call,
325 * we know that the values in fph are all dead. Hence, there
326 * is no need to inherit the fph state from the parent to the
327 * child and all we have to do is to make sure that
328 * IA64_THREAD_FPH_VALID is cleared in the child.
329 *
330 * XXX We could push this optimization a bit further by
331 * clearing IA64_THREAD_FPH_VALID on ANY system call.
332 * However, it's not clear this is worth doing. Also, it
333 * would be a slight deviation from the normal Linux system
334 * call behavior where scratch registers are preserved across
335 * system calls (unless used by the system call itself).
336 */
337# define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
338 | IA64_THREAD_PM_VALID)
339# define THREAD_FLAGS_TO_SET 0
340 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
341 | THREAD_FLAGS_TO_SET);
342
343 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
344
345 if (unlikely(args->fn)) {
346 if (unlikely(args->idle)) {
347 /* fork_idle() called us */
348 return 0;
349 }
350 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
351 child_stack->r4 = (unsigned long) args->fn;
352 child_stack->r5 = (unsigned long) args->fn_arg;
353 /*
354 * Preserve PSR bits, except for bits 32-34 and 37-45,
355 * which we can't read.
356 */
357 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
358 /* mark as valid, empty frame */
359 child_ptregs->cr_ifs = 1UL << 63;
360 child_stack->ar_fpsr = child_ptregs->ar_fpsr
361 = ia64_getreg(_IA64_REG_AR_FPSR);
362 child_stack->pr = (1 << PRED_KERNEL_STACK);
363 child_stack->ar_bspstore = child_rbs;
364 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
365
366 /* stop some PSR bits from being inherited.
367 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
368 * therefore we must specify them explicitly here and not include them in
369 * IA64_PSR_BITS_TO_CLEAR.
370 */
371 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
372 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
373
374 return 0;
375 }
376 stack = ((struct switch_stack *) regs) - 1;
377 /* copy parent's switch_stack & pt_regs to child: */
378 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
379
380 /* copy the parent's register backing store to the child: */
381 rbs_size = stack->ar_bspstore - rbs;
382 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
383 if (clone_flags & CLONE_SETTLS)
384 child_ptregs->r13 = tls;
385 if (user_stack_base) {
386 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
387 child_ptregs->ar_bspstore = user_stack_base;
388 child_ptregs->ar_rnat = 0;
389 child_ptregs->loadrs = 0;
390 }
391 child_stack->ar_bspstore = child_rbs + rbs_size;
392 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
393
394 /* stop some PSR bits from being inherited.
395 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
396 * therefore we must specify them explicitly here and not include them in
397 * IA64_PSR_BITS_TO_CLEAR.
398 */
399 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
400 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
401 return retval;
402}
403
404asmlinkage long ia64_clone(unsigned long clone_flags, unsigned long stack_start,
405 unsigned long stack_size, unsigned long parent_tidptr,
406 unsigned long child_tidptr, unsigned long tls)
407{
408 struct kernel_clone_args args = {
409 .flags = (lower_32_bits(clone_flags) & ~CSIGNAL),
410 .pidfd = (int __user *)parent_tidptr,
411 .child_tid = (int __user *)child_tidptr,
412 .parent_tid = (int __user *)parent_tidptr,
413 .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL),
414 .stack = stack_start,
415 .stack_size = stack_size,
416 .tls = tls,
417 };
418
419 return kernel_clone(&args);
420}
421
422static void
423do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
424{
425 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
426 unsigned long ip;
427 elf_greg_t *dst = arg;
428 struct pt_regs *pt;
429 char nat;
430 int i;
431
432 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
433
434 if (unw_unwind_to_user(info) < 0)
435 return;
436
437 unw_get_sp(info, &sp);
438 pt = (struct pt_regs *) (sp + 16);
439
440 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
441
442 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
443 return;
444
445 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
446 &ar_rnat);
447
448 /*
449 * coredump format:
450 * r0-r31
451 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
452 * predicate registers (p0-p63)
453 * b0-b7
454 * ip cfm user-mask
455 * ar.rsc ar.bsp ar.bspstore ar.rnat
456 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
457 */
458
459 /* r0 is zero */
460 for (i = 1, mask = (1UL << i); i < 32; ++i) {
461 unw_get_gr(info, i, &dst[i], &nat);
462 if (nat)
463 nat_bits |= mask;
464 mask <<= 1;
465 }
466 dst[32] = nat_bits;
467 unw_get_pr(info, &dst[33]);
468
469 for (i = 0; i < 8; ++i)
470 unw_get_br(info, i, &dst[34 + i]);
471
472 unw_get_rp(info, &ip);
473 dst[42] = ip + ia64_psr(pt)->ri;
474 dst[43] = cfm;
475 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
476
477 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
478 /*
479 * For bsp and bspstore, unw_get_ar() would return the kernel
480 * addresses, but we need the user-level addresses instead:
481 */
482 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
483 dst[47] = pt->ar_bspstore;
484 dst[48] = ar_rnat;
485 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
486 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
487 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
488 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
489 unw_get_ar(info, UNW_AR_LC, &dst[53]);
490 unw_get_ar(info, UNW_AR_EC, &dst[54]);
491 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
492 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
493}
494
495static void
496do_copy_regs (struct unw_frame_info *info, void *arg)
497{
498 do_copy_task_regs(current, info, arg);
499}
500
501void
502ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
503{
504 unw_init_running(do_copy_regs, dst);
505}
506
507/*
508 * Flush thread state. This is called when a thread does an execve().
509 */
510void
511flush_thread (void)
512{
513 /* drop floating-point and debug-register state if it exists: */
514 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
515 ia64_drop_fpu(current);
516}
517
518/*
519 * Clean up state associated with a thread. This is called when
520 * the thread calls exit().
521 */
522void
523exit_thread (struct task_struct *tsk)
524{
525
526 ia64_drop_fpu(tsk);
527}
528
529unsigned long
530__get_wchan (struct task_struct *p)
531{
532 struct unw_frame_info info;
533 unsigned long ip;
534 int count = 0;
535
536 /*
537 * Note: p may not be a blocked task (it could be current or
538 * another process running on some other CPU. Rather than
539 * trying to determine if p is really blocked, we just assume
540 * it's blocked and rely on the unwind routines to fail
541 * gracefully if the process wasn't really blocked after all.
542 * --davidm 99/12/15
543 */
544 unw_init_from_blocked_task(&info, p);
545 do {
546 if (task_is_running(p))
547 return 0;
548 if (unw_unwind(&info) < 0)
549 return 0;
550 unw_get_ip(&info, &ip);
551 if (!in_sched_functions(ip))
552 return ip;
553 } while (count++ < 16);
554 return 0;
555}
556
557void
558cpu_halt (void)
559{
560 pal_power_mgmt_info_u_t power_info[8];
561 unsigned long min_power;
562 int i, min_power_state;
563
564 if (ia64_pal_halt_info(power_info) != 0)
565 return;
566
567 min_power_state = 0;
568 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
569 for (i = 1; i < 8; ++i)
570 if (power_info[i].pal_power_mgmt_info_s.im
571 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
572 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
573 min_power_state = i;
574 }
575
576 while (1)
577 ia64_pal_halt(min_power_state);
578}
579
580void machine_shutdown(void)
581{
582 smp_shutdown_nonboot_cpus(reboot_cpu);
583
584#ifdef CONFIG_KEXEC
585 kexec_disable_iosapic();
586#endif
587}
588
589void
590machine_restart (char *restart_cmd)
591{
592 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
593 efi_reboot(REBOOT_WARM, NULL);
594}
595
596void
597machine_halt (void)
598{
599 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
600 cpu_halt();
601}
602
603void
604machine_power_off (void)
605{
606 do_kernel_power_off();
607 machine_halt();
608}
609
610EXPORT_SYMBOL(ia64_delay_loop);