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>
56 # include <asm/perfmon.h>
61 void (*ia64_mark_idle)(int);
63 unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
64 EXPORT_SYMBOL(boot_option_idle_override);
65 void (*pm_power_off) (void);
66 EXPORT_SYMBOL(pm_power_off);
69 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
71 unsigned long ip, sp, bsp;
72 const char *loglvl = arg;
74 printk("%s\nCall Trace:\n", loglvl);
76 unw_get_ip(info, &ip);
80 unw_get_sp(info, &sp);
81 unw_get_bsp(info, &bsp);
82 printk("%s [<%016lx>] %pS\n"
83 " sp=%016lx bsp=%016lx\n",
84 loglvl, ip, (void *)ip, sp, bsp);
85 } while (unw_unwind(info) >= 0);
89 show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
92 unw_init_running(ia64_do_show_stack, (void *)loglvl);
94 struct unw_frame_info info;
96 unw_init_from_blocked_task(&info, task);
97 ia64_do_show_stack(&info, (void *)loglvl);
102 show_regs (struct pt_regs *regs)
104 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
108 show_regs_print_info(KERN_DEFAULT);
109 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
110 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
111 init_utsname()->release);
112 printk("ip is at %pS\n", (void *)ip);
113 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
114 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
115 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
116 regs->ar_rnat, regs->ar_bspstore, regs->pr);
117 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
118 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
119 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
120 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
121 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
122 regs->f6.u.bits[1], regs->f6.u.bits[0],
123 regs->f7.u.bits[1], regs->f7.u.bits[0]);
124 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
125 regs->f8.u.bits[1], regs->f8.u.bits[0],
126 regs->f9.u.bits[1], regs->f9.u.bits[0]);
127 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
128 regs->f10.u.bits[1], regs->f10.u.bits[0],
129 regs->f11.u.bits[1], regs->f11.u.bits[0]);
131 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
132 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
133 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
134 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
135 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
136 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
137 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
138 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
139 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
141 if (user_mode(regs)) {
142 /* print the stacked registers */
143 unsigned long val, *bsp, ndirty;
144 int i, sof, is_nat = 0;
146 sof = regs->cr_ifs & 0x7f; /* size of frame */
147 ndirty = (regs->loadrs >> 19);
148 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
149 for (i = 0; i < sof; ++i) {
150 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
151 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
152 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
155 show_stack(NULL, NULL, KERN_DEFAULT);
158 /* local support for deprecated console_print */
160 console_print(const char *s)
162 printk(KERN_EMERG "%s", s);
166 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
168 if (fsys_mode(current, &scr->pt)) {
170 * defer signal-handling etc. until we return to
173 if (!ia64_psr(&scr->pt)->lp)
174 ia64_psr(&scr->pt)->lp = 1;
178 #ifdef CONFIG_PERFMON
179 if (current->thread.pfm_needs_checking)
181 * Note: pfm_handle_work() allow us to call it with interrupts
182 * disabled, and may enable interrupts within the function.
187 /* deal with pending signal delivery */
188 if (test_thread_flag(TIF_SIGPENDING)) {
189 local_irq_enable(); /* force interrupt enable */
190 ia64_do_signal(scr, in_syscall);
193 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) {
194 local_irq_enable(); /* force interrupt enable */
195 tracehook_notify_resume(&scr->pt);
198 /* copy user rbs to kernel rbs */
199 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
200 local_irq_enable(); /* force interrupt enable */
204 local_irq_disable(); /* force interrupt disable */
207 static int __init nohalt_setup(char * str)
209 cpu_idle_poll_ctrl(true);
212 __setup("nohalt", nohalt_setup);
214 #ifdef CONFIG_HOTPLUG_CPU
215 /* We don't actually take CPU down, just spin without interrupts. */
216 static inline void play_dead(void)
218 unsigned int this_cpu = smp_processor_id();
221 __this_cpu_write(cpu_state, CPU_DEAD);
226 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
228 * The above is a point of no-return, the processor is
229 * expected to be in SAL loop now.
234 static inline void play_dead(void)
238 #endif /* CONFIG_HOTPLUG_CPU */
240 void arch_cpu_idle_dead(void)
245 void arch_cpu_idle(void)
247 void (*mark_idle)(int) = ia64_mark_idle;
266 ia64_save_extra (struct task_struct *task)
268 #ifdef CONFIG_PERFMON
272 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
273 ia64_save_debug_regs(&task->thread.dbr[0]);
275 #ifdef CONFIG_PERFMON
276 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
279 info = __this_cpu_read(pfm_syst_info);
280 if (info & PFM_CPUINFO_SYST_WIDE)
281 pfm_syst_wide_update_task(task, info, 0);
286 ia64_load_extra (struct task_struct *task)
288 #ifdef CONFIG_PERFMON
292 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
293 ia64_load_debug_regs(&task->thread.dbr[0]);
295 #ifdef CONFIG_PERFMON
296 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
299 info = __this_cpu_read(pfm_syst_info);
300 if (info & PFM_CPUINFO_SYST_WIDE)
301 pfm_syst_wide_update_task(task, info, 1);
306 * Copy the state of an ia-64 thread.
308 * We get here through the following call chain:
310 * from user-level: from kernel:
312 * <clone syscall> <some kernel call frames>
315 * copy_thread copy_thread
317 * This means that the stack layout is as follows:
319 * +---------------------+ (highest addr)
321 * +---------------------+
322 * | struct switch_stack |
323 * +---------------------+
326 * | | <-- sp (lowest addr)
327 * +---------------------+
329 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
330 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
331 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
332 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
333 * the stack is page aligned and the page size is at least 4KB, this is always the case,
334 * so there is nothing to worry about.
337 copy_thread(unsigned long clone_flags,
338 unsigned long user_stack_base, unsigned long user_stack_size,
339 struct task_struct *p)
341 extern char ia64_ret_from_clone;
342 struct switch_stack *child_stack, *stack;
343 unsigned long rbs, child_rbs, rbs_size;
344 struct pt_regs *child_ptregs;
345 struct pt_regs *regs = current_pt_regs();
348 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
349 child_stack = (struct switch_stack *) child_ptregs - 1;
351 rbs = (unsigned long) current + IA64_RBS_OFFSET;
352 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
354 /* copy parts of thread_struct: */
355 p->thread.ksp = (unsigned long) child_stack - 16;
358 * NOTE: The calling convention considers all floating point
359 * registers in the high partition (fph) to be scratch. Since
360 * the only way to get to this point is through a system call,
361 * we know that the values in fph are all dead. Hence, there
362 * is no need to inherit the fph state from the parent to the
363 * child and all we have to do is to make sure that
364 * IA64_THREAD_FPH_VALID is cleared in the child.
366 * XXX We could push this optimization a bit further by
367 * clearing IA64_THREAD_FPH_VALID on ANY system call.
368 * However, it's not clear this is worth doing. Also, it
369 * would be a slight deviation from the normal Linux system
370 * call behavior where scratch registers are preserved across
371 * system calls (unless used by the system call itself).
373 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
374 | IA64_THREAD_PM_VALID)
375 # define THREAD_FLAGS_TO_SET 0
376 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
377 | THREAD_FLAGS_TO_SET);
379 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
381 if (unlikely(p->flags & PF_KTHREAD)) {
382 if (unlikely(!user_stack_base)) {
383 /* fork_idle() called us */
386 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
387 child_stack->r4 = user_stack_base; /* payload */
388 child_stack->r5 = user_stack_size; /* argument */
390 * Preserve PSR bits, except for bits 32-34 and 37-45,
391 * which we can't read.
393 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
394 /* mark as valid, empty frame */
395 child_ptregs->cr_ifs = 1UL << 63;
396 child_stack->ar_fpsr = child_ptregs->ar_fpsr
397 = ia64_getreg(_IA64_REG_AR_FPSR);
398 child_stack->pr = (1 << PRED_KERNEL_STACK);
399 child_stack->ar_bspstore = child_rbs;
400 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
402 /* stop some PSR bits from being inherited.
403 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
404 * therefore we must specify them explicitly here and not include them in
405 * IA64_PSR_BITS_TO_CLEAR.
407 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
408 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
412 stack = ((struct switch_stack *) regs) - 1;
413 /* copy parent's switch_stack & pt_regs to child: */
414 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
416 /* copy the parent's register backing store to the child: */
417 rbs_size = stack->ar_bspstore - rbs;
418 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
419 if (clone_flags & CLONE_SETTLS)
420 child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */
421 if (user_stack_base) {
422 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
423 child_ptregs->ar_bspstore = user_stack_base;
424 child_ptregs->ar_rnat = 0;
425 child_ptregs->loadrs = 0;
427 child_stack->ar_bspstore = child_rbs + rbs_size;
428 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
430 /* stop some PSR bits from being inherited.
431 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
432 * therefore we must specify them explicitly here and not include them in
433 * IA64_PSR_BITS_TO_CLEAR.
435 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
436 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
438 #ifdef CONFIG_PERFMON
439 if (current->thread.pfm_context)
440 pfm_inherit(p, child_ptregs);
446 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
448 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
449 unsigned long uninitialized_var(ip); /* GCC be quiet */
450 elf_greg_t *dst = arg;
455 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
457 if (unw_unwind_to_user(info) < 0)
460 unw_get_sp(info, &sp);
461 pt = (struct pt_regs *) (sp + 16);
463 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
465 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
468 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
474 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
475 * predicate registers (p0-p63)
478 * ar.rsc ar.bsp ar.bspstore ar.rnat
479 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
483 for (i = 1, mask = (1UL << i); i < 32; ++i) {
484 unw_get_gr(info, i, &dst[i], &nat);
490 unw_get_pr(info, &dst[33]);
492 for (i = 0; i < 8; ++i)
493 unw_get_br(info, i, &dst[34 + i]);
495 unw_get_rp(info, &ip);
496 dst[42] = ip + ia64_psr(pt)->ri;
498 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
500 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
502 * For bsp and bspstore, unw_get_ar() would return the kernel
503 * addresses, but we need the user-level addresses instead:
505 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
506 dst[47] = pt->ar_bspstore;
508 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
509 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
510 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
511 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
512 unw_get_ar(info, UNW_AR_LC, &dst[53]);
513 unw_get_ar(info, UNW_AR_EC, &dst[54]);
514 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
515 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
519 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
521 elf_fpreg_t *dst = arg;
524 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
526 if (unw_unwind_to_user(info) < 0)
529 /* f0 is 0.0, f1 is 1.0 */
531 for (i = 2; i < 32; ++i)
532 unw_get_fr(info, i, dst + i);
534 ia64_flush_fph(task);
535 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
536 memcpy(dst + 32, task->thread.fph, 96*16);
540 do_copy_regs (struct unw_frame_info *info, void *arg)
542 do_copy_task_regs(current, info, arg);
546 do_dump_fpu (struct unw_frame_info *info, void *arg)
548 do_dump_task_fpu(current, info, arg);
552 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
554 unw_init_running(do_copy_regs, dst);
558 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
560 unw_init_running(do_dump_fpu, dst);
561 return 1; /* f0-f31 are always valid so we always return 1 */
565 * Flush thread state. This is called when a thread does an execve().
570 /* drop floating-point and debug-register state if it exists: */
571 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
572 ia64_drop_fpu(current);
576 * Clean up state associated with a thread. This is called when
577 * the thread calls exit().
580 exit_thread (struct task_struct *tsk)
584 #ifdef CONFIG_PERFMON
585 /* if needed, stop monitoring and flush state to perfmon context */
586 if (tsk->thread.pfm_context)
587 pfm_exit_thread(tsk);
589 /* free debug register resources */
590 if (tsk->thread.flags & IA64_THREAD_DBG_VALID)
591 pfm_release_debug_registers(tsk);
596 get_wchan (struct task_struct *p)
598 struct unw_frame_info info;
602 if (!p || p == current || p->state == TASK_RUNNING)
606 * Note: p may not be a blocked task (it could be current or
607 * another process running on some other CPU. Rather than
608 * trying to determine if p is really blocked, we just assume
609 * it's blocked and rely on the unwind routines to fail
610 * gracefully if the process wasn't really blocked after all.
613 unw_init_from_blocked_task(&info, p);
615 if (p->state == TASK_RUNNING)
617 if (unw_unwind(&info) < 0)
619 unw_get_ip(&info, &ip);
620 if (!in_sched_functions(ip))
622 } while (count++ < 16);
629 pal_power_mgmt_info_u_t power_info[8];
630 unsigned long min_power;
631 int i, min_power_state;
633 if (ia64_pal_halt_info(power_info) != 0)
637 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
638 for (i = 1; i < 8; ++i)
639 if (power_info[i].pal_power_mgmt_info_s.im
640 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
641 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
646 ia64_pal_halt(min_power_state);
649 void machine_shutdown(void)
651 smp_shutdown_nonboot_cpus(reboot_cpu);
654 kexec_disable_iosapic();
659 machine_restart (char *restart_cmd)
661 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
662 efi_reboot(REBOOT_WARM, NULL);
668 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
673 machine_power_off (void)
680 EXPORT_SYMBOL(ia64_delay_loop);