2 * Architecture-specific setup.
4 * Copyright (C) 1998-2003 Hewlett-Packard Co
5 * David Mosberger-Tang <davidm@hpl.hp.com>
6 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
8 * 2005-10-07 Keith Owens <kaos@sgi.com>
9 * Add notify_die() hooks.
11 #include <linux/cpu.h>
13 #include <linux/elf.h>
14 #include <linux/errno.h>
15 #include <linux/kallsyms.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/stddef.h>
24 #include <linux/thread_info.h>
25 #include <linux/unistd.h>
26 #include <linux/efi.h>
27 #include <linux/interrupt.h>
28 #include <linux/delay.h>
29 #include <linux/kdebug.h>
30 #include <linux/utsname.h>
31 #include <linux/tracehook.h>
32 #include <linux/rcupdate.h>
35 #include <asm/delay.h>
38 #include <asm/kexec.h>
39 #include <asm/pgalloc.h>
40 #include <asm/processor.h>
42 #include <asm/switch_to.h>
43 #include <asm/tlbflush.h>
44 #include <asm/uaccess.h>
45 #include <asm/unwind.h>
51 # include <asm/perfmon.h>
56 void (*ia64_mark_idle)(int);
58 unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
59 EXPORT_SYMBOL(boot_option_idle_override);
60 void (*pm_idle) (void);
61 EXPORT_SYMBOL(pm_idle);
62 void (*pm_power_off) (void);
63 EXPORT_SYMBOL(pm_power_off);
66 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
68 unsigned long ip, sp, bsp;
69 char buf[128]; /* don't make it so big that it overflows the stack! */
71 printk("\nCall Trace:\n");
73 unw_get_ip(info, &ip);
77 unw_get_sp(info, &sp);
78 unw_get_bsp(info, &bsp);
79 snprintf(buf, sizeof(buf),
81 " sp=%016lx bsp=%016lx\n",
83 print_symbol(buf, ip);
84 } while (unw_unwind(info) >= 0);
88 show_stack (struct task_struct *task, unsigned long *sp)
91 unw_init_running(ia64_do_show_stack, NULL);
93 struct unw_frame_info info;
95 unw_init_from_blocked_task(&info, task);
96 ia64_do_show_stack(&info, NULL);
103 show_stack(NULL, NULL);
106 EXPORT_SYMBOL(dump_stack);
109 show_regs (struct pt_regs *regs)
111 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
114 printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
115 smp_processor_id(), current->comm);
116 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
117 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
118 init_utsname()->release);
119 print_symbol("ip is at %s\n", ip);
120 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
121 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
122 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
123 regs->ar_rnat, regs->ar_bspstore, regs->pr);
124 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
125 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
126 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
127 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
128 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
129 regs->f6.u.bits[1], regs->f6.u.bits[0],
130 regs->f7.u.bits[1], regs->f7.u.bits[0]);
131 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
132 regs->f8.u.bits[1], regs->f8.u.bits[0],
133 regs->f9.u.bits[1], regs->f9.u.bits[0]);
134 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
135 regs->f10.u.bits[1], regs->f10.u.bits[0],
136 regs->f11.u.bits[1], regs->f11.u.bits[0]);
138 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
139 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
140 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
141 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
142 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
143 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
144 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
145 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
146 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
148 if (user_mode(regs)) {
149 /* print the stacked registers */
150 unsigned long val, *bsp, ndirty;
151 int i, sof, is_nat = 0;
153 sof = regs->cr_ifs & 0x7f; /* size of frame */
154 ndirty = (regs->loadrs >> 19);
155 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
156 for (i = 0; i < sof; ++i) {
157 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
158 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
159 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
162 show_stack(NULL, NULL);
165 /* local support for deprecated console_print */
167 console_print(const char *s)
169 printk(KERN_EMERG "%s", s);
173 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
175 if (fsys_mode(current, &scr->pt)) {
177 * defer signal-handling etc. until we return to
180 if (!ia64_psr(&scr->pt)->lp)
181 ia64_psr(&scr->pt)->lp = 1;
185 #ifdef CONFIG_PERFMON
186 if (current->thread.pfm_needs_checking)
188 * Note: pfm_handle_work() allow us to call it with interrupts
189 * disabled, and may enable interrupts within the function.
194 /* deal with pending signal delivery */
195 if (test_thread_flag(TIF_SIGPENDING)) {
196 local_irq_enable(); /* force interrupt enable */
197 ia64_do_signal(scr, in_syscall);
200 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) {
201 local_irq_enable(); /* force interrupt enable */
202 tracehook_notify_resume(&scr->pt);
205 /* copy user rbs to kernel rbs */
206 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
207 local_irq_enable(); /* force interrupt enable */
211 local_irq_disable(); /* force interrupt disable */
214 static int pal_halt = 1;
215 static int can_do_pal_halt = 1;
217 static int __init nohalt_setup(char * str)
219 pal_halt = can_do_pal_halt = 0;
222 __setup("nohalt", nohalt_setup);
225 update_pal_halt_status(int status)
227 can_do_pal_halt = pal_halt && status;
231 * We use this if we don't have any better idle routine..
237 while (!need_resched()) {
238 if (can_do_pal_halt) {
240 if (!need_resched()) {
249 #ifdef CONFIG_HOTPLUG_CPU
250 /* We don't actually take CPU down, just spin without interrupts. */
251 static inline void play_dead(void)
253 unsigned int this_cpu = smp_processor_id();
256 __get_cpu_var(cpu_state) = CPU_DEAD;
261 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
263 * The above is a point of no-return, the processor is
264 * expected to be in SAL loop now.
269 static inline void play_dead(void)
273 #endif /* CONFIG_HOTPLUG_CPU */
275 void __attribute__((noreturn))
278 void (*mark_idle)(int) = ia64_mark_idle;
279 int cpu = smp_processor_id();
281 /* endless idle loop with no priority at all */
284 if (can_do_pal_halt) {
285 current_thread_info()->status &= ~TS_POLLING;
287 * TS_POLLING-cleared state must be visible before we
292 current_thread_info()->status |= TS_POLLING;
295 if (!need_resched()) {
315 schedule_preempt_disabled();
317 if (cpu_is_offline(cpu))
323 ia64_save_extra (struct task_struct *task)
325 #ifdef CONFIG_PERFMON
329 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
330 ia64_save_debug_regs(&task->thread.dbr[0]);
332 #ifdef CONFIG_PERFMON
333 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
336 info = __get_cpu_var(pfm_syst_info);
337 if (info & PFM_CPUINFO_SYST_WIDE)
338 pfm_syst_wide_update_task(task, info, 0);
343 ia64_load_extra (struct task_struct *task)
345 #ifdef CONFIG_PERFMON
349 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
350 ia64_load_debug_regs(&task->thread.dbr[0]);
352 #ifdef CONFIG_PERFMON
353 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
356 info = __get_cpu_var(pfm_syst_info);
357 if (info & PFM_CPUINFO_SYST_WIDE)
358 pfm_syst_wide_update_task(task, info, 1);
363 * Copy the state of an ia-64 thread.
365 * We get here through the following call chain:
367 * from user-level: from kernel:
369 * <clone syscall> <some kernel call frames>
372 * copy_thread copy_thread
374 * This means that the stack layout is as follows:
376 * +---------------------+ (highest addr)
378 * +---------------------+
379 * | struct switch_stack |
380 * +---------------------+
383 * | | <-- sp (lowest addr)
384 * +---------------------+
386 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
387 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
388 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
389 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
390 * the stack is page aligned and the page size is at least 4KB, this is always the case,
391 * so there is nothing to worry about.
394 copy_thread(unsigned long clone_flags,
395 unsigned long user_stack_base, unsigned long user_stack_size,
396 struct task_struct *p, struct pt_regs *regs)
398 extern char ia64_ret_from_clone;
399 struct switch_stack *child_stack, *stack;
400 unsigned long rbs, child_rbs, rbs_size;
401 struct pt_regs *child_ptregs;
404 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
405 child_stack = (struct switch_stack *) child_ptregs - 1;
407 rbs = (unsigned long) current + IA64_RBS_OFFSET;
408 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
410 /* copy parts of thread_struct: */
411 p->thread.ksp = (unsigned long) child_stack - 16;
414 * NOTE: The calling convention considers all floating point
415 * registers in the high partition (fph) to be scratch. Since
416 * the only way to get to this point is through a system call,
417 * we know that the values in fph are all dead. Hence, there
418 * is no need to inherit the fph state from the parent to the
419 * child and all we have to do is to make sure that
420 * IA64_THREAD_FPH_VALID is cleared in the child.
422 * XXX We could push this optimization a bit further by
423 * clearing IA64_THREAD_FPH_VALID on ANY system call.
424 * However, it's not clear this is worth doing. Also, it
425 * would be a slight deviation from the normal Linux system
426 * call behavior where scratch registers are preserved across
427 * system calls (unless used by the system call itself).
429 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
430 | IA64_THREAD_PM_VALID)
431 # define THREAD_FLAGS_TO_SET 0
432 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
433 | THREAD_FLAGS_TO_SET);
435 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
437 if (unlikely(p->flags & PF_KTHREAD)) {
438 if (unlikely(!user_stack_base)) {
439 /* fork_idle() called us */
442 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
443 child_stack->r4 = user_stack_base; /* payload */
444 child_stack->r5 = user_stack_size; /* argument */
446 * Preserve PSR bits, except for bits 32-34 and 37-45,
447 * which we can't read.
449 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
450 /* mark as valid, empty frame */
451 child_ptregs->cr_ifs = 1UL << 63;
452 child_stack->ar_fpsr = child_ptregs->ar_fpsr
453 = ia64_getreg(_IA64_REG_AR_FPSR);
454 child_stack->pr = (1 << PRED_KERNEL_STACK);
455 child_stack->ar_bspstore = child_rbs;
456 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
458 /* stop some PSR bits from being inherited.
459 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
460 * therefore we must specify them explicitly here and not include them in
461 * IA64_PSR_BITS_TO_CLEAR.
463 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
464 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
468 stack = ((struct switch_stack *) regs) - 1;
469 /* copy parent's switch_stack & pt_regs to child: */
470 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
472 /* copy the parent's register backing store to the child: */
473 rbs_size = stack->ar_bspstore - rbs;
474 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
475 if (clone_flags & CLONE_SETTLS)
476 child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */
477 if (user_stack_base) {
478 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
479 child_ptregs->ar_bspstore = user_stack_base;
480 child_ptregs->ar_rnat = 0;
481 child_ptregs->loadrs = 0;
483 child_stack->ar_bspstore = child_rbs + rbs_size;
484 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
486 /* stop some PSR bits from being inherited.
487 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
488 * therefore we must specify them explicitly here and not include them in
489 * IA64_PSR_BITS_TO_CLEAR.
491 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
492 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
494 #ifdef CONFIG_PERFMON
495 if (current->thread.pfm_context)
496 pfm_inherit(p, child_ptregs);
502 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
504 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
505 unsigned long uninitialized_var(ip); /* GCC be quiet */
506 elf_greg_t *dst = arg;
511 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
513 if (unw_unwind_to_user(info) < 0)
516 unw_get_sp(info, &sp);
517 pt = (struct pt_regs *) (sp + 16);
519 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
521 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
524 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
530 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
531 * predicate registers (p0-p63)
534 * ar.rsc ar.bsp ar.bspstore ar.rnat
535 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
539 for (i = 1, mask = (1UL << i); i < 32; ++i) {
540 unw_get_gr(info, i, &dst[i], &nat);
546 unw_get_pr(info, &dst[33]);
548 for (i = 0; i < 8; ++i)
549 unw_get_br(info, i, &dst[34 + i]);
551 unw_get_rp(info, &ip);
552 dst[42] = ip + ia64_psr(pt)->ri;
554 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
556 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
558 * For bsp and bspstore, unw_get_ar() would return the kernel
559 * addresses, but we need the user-level addresses instead:
561 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
562 dst[47] = pt->ar_bspstore;
564 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
565 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
566 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
567 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
568 unw_get_ar(info, UNW_AR_LC, &dst[53]);
569 unw_get_ar(info, UNW_AR_EC, &dst[54]);
570 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
571 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
575 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
577 elf_fpreg_t *dst = arg;
580 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
582 if (unw_unwind_to_user(info) < 0)
585 /* f0 is 0.0, f1 is 1.0 */
587 for (i = 2; i < 32; ++i)
588 unw_get_fr(info, i, dst + i);
590 ia64_flush_fph(task);
591 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
592 memcpy(dst + 32, task->thread.fph, 96*16);
596 do_copy_regs (struct unw_frame_info *info, void *arg)
598 do_copy_task_regs(current, info, arg);
602 do_dump_fpu (struct unw_frame_info *info, void *arg)
604 do_dump_task_fpu(current, info, arg);
608 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
610 unw_init_running(do_copy_regs, dst);
614 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
616 unw_init_running(do_dump_fpu, dst);
617 return 1; /* f0-f31 are always valid so we always return 1 */
621 * Flush thread state. This is called when a thread does an execve().
626 /* drop floating-point and debug-register state if it exists: */
627 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
628 ia64_drop_fpu(current);
632 * Clean up state associated with current thread. This is called when
633 * the thread calls exit().
639 ia64_drop_fpu(current);
640 #ifdef CONFIG_PERFMON
641 /* if needed, stop monitoring and flush state to perfmon context */
642 if (current->thread.pfm_context)
643 pfm_exit_thread(current);
645 /* free debug register resources */
646 if (current->thread.flags & IA64_THREAD_DBG_VALID)
647 pfm_release_debug_registers(current);
652 get_wchan (struct task_struct *p)
654 struct unw_frame_info info;
658 if (!p || p == current || p->state == TASK_RUNNING)
662 * Note: p may not be a blocked task (it could be current or
663 * another process running on some other CPU. Rather than
664 * trying to determine if p is really blocked, we just assume
665 * it's blocked and rely on the unwind routines to fail
666 * gracefully if the process wasn't really blocked after all.
669 unw_init_from_blocked_task(&info, p);
671 if (p->state == TASK_RUNNING)
673 if (unw_unwind(&info) < 0)
675 unw_get_ip(&info, &ip);
676 if (!in_sched_functions(ip))
678 } while (count++ < 16);
685 pal_power_mgmt_info_u_t power_info[8];
686 unsigned long min_power;
687 int i, min_power_state;
689 if (ia64_pal_halt_info(power_info) != 0)
693 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
694 for (i = 1; i < 8; ++i)
695 if (power_info[i].pal_power_mgmt_info_s.im
696 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
697 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
702 ia64_pal_halt(min_power_state);
705 void machine_shutdown(void)
707 #ifdef CONFIG_HOTPLUG_CPU
710 for_each_online_cpu(cpu) {
711 if (cpu != smp_processor_id())
716 kexec_disable_iosapic();
721 machine_restart (char *restart_cmd)
723 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
724 (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
730 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
735 machine_power_off (void)