1 // SPDX-License-Identifier: GPL-2.0
3 * Architecture-specific setup.
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
9 * 2005-10-07 Keith Owens <kaos@sgi.com>
10 * Add notify_die() hooks.
12 #include <linux/cpu.h>
14 #include <linux/elf.h>
15 #include <linux/errno.h>
16 #include <linux/kernel.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>
39 #include <asm/delay.h>
42 #include <asm/kexec.h>
43 #include <asm/pgalloc.h>
44 #include <asm/processor.h>
46 #include <asm/switch_to.h>
47 #include <asm/tlbflush.h>
48 #include <linux/uaccess.h>
49 #include <asm/unwind.h>
55 # include <asm/perfmon.h>
60 void (*ia64_mark_idle)(int);
62 unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
63 EXPORT_SYMBOL(boot_option_idle_override);
64 void (*pm_power_off) (void);
65 EXPORT_SYMBOL(pm_power_off);
68 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
70 unsigned long ip, sp, bsp;
71 const char *loglvl = arg;
73 printk("%s\nCall Trace:\n", loglvl);
75 unw_get_ip(info, &ip);
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);
88 show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
91 unw_init_running(ia64_do_show_stack, (void *)loglvl);
93 struct unw_frame_info info;
95 unw_init_from_blocked_task(&info, task);
96 ia64_do_show_stack(&info, (void *)loglvl);
101 show_regs (struct pt_regs *regs)
103 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
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]);
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);
140 if (user_mode(regs)) {
141 /* print the stacked registers */
142 unsigned long val, *bsp, ndirty;
143 int i, sof, is_nat = 0;
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" : " ");
154 show_stack(NULL, NULL, KERN_DEFAULT);
157 /* local support for deprecated console_print */
159 console_print(const char *s)
161 printk(KERN_EMERG "%s", s);
165 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
167 if (fsys_mode(current, &scr->pt)) {
169 * defer signal-handling etc. until we return to
172 if (!ia64_psr(&scr->pt)->lp)
173 ia64_psr(&scr->pt)->lp = 1;
177 #ifdef CONFIG_PERFMON
178 if (current->thread.pfm_needs_checking)
180 * Note: pfm_handle_work() allow us to call it with interrupts
181 * disabled, and may enable interrupts within the function.
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);
192 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) {
193 local_irq_enable(); /* force interrupt enable */
194 tracehook_notify_resume(&scr->pt);
197 /* copy user rbs to kernel rbs */
198 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
199 local_irq_enable(); /* force interrupt enable */
203 local_irq_disable(); /* force interrupt disable */
206 static int __init nohalt_setup(char * str)
208 cpu_idle_poll_ctrl(true);
211 __setup("nohalt", nohalt_setup);
213 #ifdef CONFIG_HOTPLUG_CPU
214 /* We don't actually take CPU down, just spin without interrupts. */
215 static inline void play_dead(void)
217 unsigned int this_cpu = smp_processor_id();
220 __this_cpu_write(cpu_state, CPU_DEAD);
225 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
227 * The above is a point of no-return, the processor is
228 * expected to be in SAL loop now.
233 static inline void play_dead(void)
237 #endif /* CONFIG_HOTPLUG_CPU */
239 void arch_cpu_idle_dead(void)
244 void arch_cpu_idle(void)
246 void (*mark_idle)(int) = ia64_mark_idle;
265 ia64_save_extra (struct task_struct *task)
267 #ifdef CONFIG_PERFMON
271 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
272 ia64_save_debug_regs(&task->thread.dbr[0]);
274 #ifdef CONFIG_PERFMON
275 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
278 info = __this_cpu_read(pfm_syst_info);
279 if (info & PFM_CPUINFO_SYST_WIDE)
280 pfm_syst_wide_update_task(task, info, 0);
285 ia64_load_extra (struct task_struct *task)
287 #ifdef CONFIG_PERFMON
291 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
292 ia64_load_debug_regs(&task->thread.dbr[0]);
294 #ifdef CONFIG_PERFMON
295 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
298 info = __this_cpu_read(pfm_syst_info);
299 if (info & PFM_CPUINFO_SYST_WIDE)
300 pfm_syst_wide_update_task(task, info, 1);
305 * Copy the state of an ia-64 thread.
307 * We get here through the following call chain:
309 * from user-level: from kernel:
311 * <clone syscall> <some kernel call frames>
314 * copy_thread copy_thread
316 * This means that the stack layout is as follows:
318 * +---------------------+ (highest addr)
320 * +---------------------+
321 * | struct switch_stack |
322 * +---------------------+
325 * | | <-- sp (lowest addr)
326 * +---------------------+
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.
336 copy_thread(unsigned long clone_flags, unsigned long user_stack_base,
337 unsigned long user_stack_size, struct task_struct *p, unsigned long tls)
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();
346 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
347 child_stack = (struct switch_stack *) child_ptregs - 1;
349 rbs = (unsigned long) current + IA64_RBS_OFFSET;
350 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
352 /* copy parts of thread_struct: */
353 p->thread.ksp = (unsigned long) child_stack - 16;
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.
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).
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);
377 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
379 if (unlikely(p->flags & PF_KTHREAD)) {
380 if (unlikely(!user_stack_base)) {
381 /* fork_idle() called us */
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 */
388 * Preserve PSR bits, except for bits 32-34 and 37-45,
389 * which we can't read.
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;
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.
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));
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));
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;
425 child_stack->ar_bspstore = child_rbs + rbs_size;
426 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
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.
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));
436 #ifdef CONFIG_PERFMON
437 if (current->thread.pfm_context)
438 pfm_inherit(p, child_ptregs);
443 asmlinkage 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)
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,
458 return _do_fork(&args);
462 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
464 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
466 elf_greg_t *dst = arg;
471 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
473 if (unw_unwind_to_user(info) < 0)
476 unw_get_sp(info, &sp);
477 pt = (struct pt_regs *) (sp + 16);
479 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
481 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
484 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
490 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
491 * predicate registers (p0-p63)
494 * ar.rsc ar.bsp ar.bspstore ar.rnat
495 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
499 for (i = 1, mask = (1UL << i); i < 32; ++i) {
500 unw_get_gr(info, i, &dst[i], &nat);
506 unw_get_pr(info, &dst[33]);
508 for (i = 0; i < 8; ++i)
509 unw_get_br(info, i, &dst[34 + i]);
511 unw_get_rp(info, &ip);
512 dst[42] = ip + ia64_psr(pt)->ri;
514 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
516 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
518 * For bsp and bspstore, unw_get_ar() would return the kernel
519 * addresses, but we need the user-level addresses instead:
521 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
522 dst[47] = pt->ar_bspstore;
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]);
535 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
537 elf_fpreg_t *dst = arg;
540 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
542 if (unw_unwind_to_user(info) < 0)
545 /* f0 is 0.0, f1 is 1.0 */
547 for (i = 2; i < 32; ++i)
548 unw_get_fr(info, i, dst + i);
550 ia64_flush_fph(task);
551 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
552 memcpy(dst + 32, task->thread.fph, 96*16);
556 do_copy_regs (struct unw_frame_info *info, void *arg)
558 do_copy_task_regs(current, info, arg);
562 do_dump_fpu (struct unw_frame_info *info, void *arg)
564 do_dump_task_fpu(current, info, arg);
568 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
570 unw_init_running(do_copy_regs, dst);
574 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
576 unw_init_running(do_dump_fpu, dst);
577 return 1; /* f0-f31 are always valid so we always return 1 */
581 * Flush thread state. This is called when a thread does an execve().
586 /* drop floating-point and debug-register state if it exists: */
587 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
588 ia64_drop_fpu(current);
592 * Clean up state associated with a thread. This is called when
593 * the thread calls exit().
596 exit_thread (struct task_struct *tsk)
600 #ifdef CONFIG_PERFMON
601 /* if needed, stop monitoring and flush state to perfmon context */
602 if (tsk->thread.pfm_context)
603 pfm_exit_thread(tsk);
605 /* free debug register resources */
606 if (tsk->thread.flags & IA64_THREAD_DBG_VALID)
607 pfm_release_debug_registers(tsk);
612 get_wchan (struct task_struct *p)
614 struct unw_frame_info info;
618 if (!p || p == current || p->state == TASK_RUNNING)
622 * Note: p may not be a blocked task (it could be current or
623 * another process running on some other CPU. Rather than
624 * trying to determine if p is really blocked, we just assume
625 * it's blocked and rely on the unwind routines to fail
626 * gracefully if the process wasn't really blocked after all.
629 unw_init_from_blocked_task(&info, p);
631 if (p->state == TASK_RUNNING)
633 if (unw_unwind(&info) < 0)
635 unw_get_ip(&info, &ip);
636 if (!in_sched_functions(ip))
638 } while (count++ < 16);
645 pal_power_mgmt_info_u_t power_info[8];
646 unsigned long min_power;
647 int i, min_power_state;
649 if (ia64_pal_halt_info(power_info) != 0)
653 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
654 for (i = 1; i < 8; ++i)
655 if (power_info[i].pal_power_mgmt_info_s.im
656 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
657 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
662 ia64_pal_halt(min_power_state);
665 void machine_shutdown(void)
667 smp_shutdown_nonboot_cpus(reboot_cpu);
670 kexec_disable_iosapic();
675 machine_restart (char *restart_cmd)
677 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
678 efi_reboot(REBOOT_WARM, NULL);
684 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
689 machine_power_off (void)
696 EXPORT_SYMBOL(ia64_delay_loop);