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>
34 #include <asm/delay.h>
37 #include <asm/kexec.h>
38 #include <asm/pgalloc.h>
39 #include <asm/processor.h>
41 #include <asm/tlbflush.h>
42 #include <asm/uaccess.h>
43 #include <asm/unwind.h>
49 # include <asm/perfmon.h>
54 void (*ia64_mark_idle)(int);
56 unsigned long boot_option_idle_override = 0;
57 EXPORT_SYMBOL(boot_option_idle_override);
58 unsigned long idle_halt;
59 EXPORT_SYMBOL(idle_halt);
60 unsigned long idle_nomwait;
61 EXPORT_SYMBOL(idle_nomwait);
62 void (*pm_idle) (void);
63 EXPORT_SYMBOL(pm_idle);
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 char buf[128]; /* don't make it so big that it overflows the stack! */
73 printk("\nCall Trace:\n");
75 unw_get_ip(info, &ip);
79 unw_get_sp(info, &sp);
80 unw_get_bsp(info, &bsp);
81 snprintf(buf, sizeof(buf),
83 " sp=%016lx bsp=%016lx\n",
85 print_symbol(buf, ip);
86 } while (unw_unwind(info) >= 0);
90 show_stack (struct task_struct *task, unsigned long *sp)
93 unw_init_running(ia64_do_show_stack, NULL);
95 struct unw_frame_info info;
97 unw_init_from_blocked_task(&info, task);
98 ia64_do_show_stack(&info, NULL);
105 show_stack(NULL, NULL);
108 EXPORT_SYMBOL(dump_stack);
111 show_regs (struct pt_regs *regs)
113 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
116 printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
117 smp_processor_id(), current->comm);
118 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
119 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
120 init_utsname()->release);
121 print_symbol("ip is at %s\n", ip);
122 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
123 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
124 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
125 regs->ar_rnat, regs->ar_bspstore, regs->pr);
126 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
127 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
128 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
129 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
130 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
131 regs->f6.u.bits[1], regs->f6.u.bits[0],
132 regs->f7.u.bits[1], regs->f7.u.bits[0]);
133 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
134 regs->f8.u.bits[1], regs->f8.u.bits[0],
135 regs->f9.u.bits[1], regs->f9.u.bits[0]);
136 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
137 regs->f10.u.bits[1], regs->f10.u.bits[0],
138 regs->f11.u.bits[1], regs->f11.u.bits[0]);
140 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
141 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
142 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
143 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
144 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
145 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
146 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
147 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
148 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
150 if (user_mode(regs)) {
151 /* print the stacked registers */
152 unsigned long val, *bsp, ndirty;
153 int i, sof, is_nat = 0;
155 sof = regs->cr_ifs & 0x7f; /* size of frame */
156 ndirty = (regs->loadrs >> 19);
157 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
158 for (i = 0; i < sof; ++i) {
159 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
160 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
161 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
164 show_stack(NULL, NULL);
167 /* local support for deprecated console_print */
169 console_print(const char *s)
171 printk(KERN_EMERG "%s", s);
175 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
177 if (fsys_mode(current, &scr->pt)) {
179 * defer signal-handling etc. until we return to
182 if (!ia64_psr(&scr->pt)->lp)
183 ia64_psr(&scr->pt)->lp = 1;
187 #ifdef CONFIG_PERFMON
188 if (current->thread.pfm_needs_checking)
190 * Note: pfm_handle_work() allow us to call it with interrupts
191 * disabled, and may enable interrupts within the function.
196 /* deal with pending signal delivery */
197 if (test_thread_flag(TIF_SIGPENDING)) {
198 local_irq_enable(); /* force interrupt enable */
199 ia64_do_signal(scr, in_syscall);
202 if (test_thread_flag(TIF_NOTIFY_RESUME)) {
203 clear_thread_flag(TIF_NOTIFY_RESUME);
204 tracehook_notify_resume(&scr->pt);
205 if (current->replacement_session_keyring)
206 key_replace_session_keyring();
209 /* copy user rbs to kernel rbs */
210 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
211 local_irq_enable(); /* force interrupt enable */
215 local_irq_disable(); /* force interrupt disable */
218 static int pal_halt = 1;
219 static int can_do_pal_halt = 1;
221 static int __init nohalt_setup(char * str)
223 pal_halt = can_do_pal_halt = 0;
226 __setup("nohalt", nohalt_setup);
229 update_pal_halt_status(int status)
231 can_do_pal_halt = pal_halt && status;
235 * We use this if we don't have any better idle routine..
241 while (!need_resched()) {
242 if (can_do_pal_halt) {
244 if (!need_resched()) {
253 #ifdef CONFIG_HOTPLUG_CPU
254 /* We don't actually take CPU down, just spin without interrupts. */
255 static inline void play_dead(void)
257 unsigned int this_cpu = smp_processor_id();
260 __get_cpu_var(cpu_state) = CPU_DEAD;
265 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
267 * The above is a point of no-return, the processor is
268 * expected to be in SAL loop now.
273 static inline void play_dead(void)
277 #endif /* CONFIG_HOTPLUG_CPU */
279 static void do_nothing(void *unused)
284 * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
285 * pm_idle and update to new pm_idle value. Required while changing pm_idle
286 * handler on SMP systems.
288 * Caller must have changed pm_idle to the new value before the call. Old
289 * pm_idle value will not be used by any CPU after the return of this function.
291 void cpu_idle_wait(void)
294 /* kick all the CPUs so that they exit out of pm_idle */
295 smp_call_function(do_nothing, NULL, 1);
297 EXPORT_SYMBOL_GPL(cpu_idle_wait);
299 void __attribute__((noreturn))
302 void (*mark_idle)(int) = ia64_mark_idle;
303 int cpu = smp_processor_id();
305 /* endless idle loop with no priority at all */
307 if (can_do_pal_halt) {
308 current_thread_info()->status &= ~TS_POLLING;
310 * TS_POLLING-cleared state must be visible before we
315 current_thread_info()->status |= TS_POLLING;
318 if (!need_resched()) {
337 preempt_enable_no_resched();
341 if (cpu_is_offline(cpu))
347 ia64_save_extra (struct task_struct *task)
349 #ifdef CONFIG_PERFMON
353 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
354 ia64_save_debug_regs(&task->thread.dbr[0]);
356 #ifdef CONFIG_PERFMON
357 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
360 info = __get_cpu_var(pfm_syst_info);
361 if (info & PFM_CPUINFO_SYST_WIDE)
362 pfm_syst_wide_update_task(task, info, 0);
367 ia64_load_extra (struct task_struct *task)
369 #ifdef CONFIG_PERFMON
373 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
374 ia64_load_debug_regs(&task->thread.dbr[0]);
376 #ifdef CONFIG_PERFMON
377 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
380 info = __get_cpu_var(pfm_syst_info);
381 if (info & PFM_CPUINFO_SYST_WIDE)
382 pfm_syst_wide_update_task(task, info, 1);
387 * Copy the state of an ia-64 thread.
389 * We get here through the following call chain:
391 * from user-level: from kernel:
393 * <clone syscall> <some kernel call frames>
396 * copy_thread copy_thread
398 * This means that the stack layout is as follows:
400 * +---------------------+ (highest addr)
402 * +---------------------+
403 * | struct switch_stack |
404 * +---------------------+
407 * | | <-- sp (lowest addr)
408 * +---------------------+
410 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
411 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
412 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
413 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
414 * the stack is page aligned and the page size is at least 4KB, this is always the case,
415 * so there is nothing to worry about.
418 copy_thread(unsigned long clone_flags,
419 unsigned long user_stack_base, unsigned long user_stack_size,
420 struct task_struct *p, struct pt_regs *regs)
422 extern char ia64_ret_from_clone;
423 struct switch_stack *child_stack, *stack;
424 unsigned long rbs, child_rbs, rbs_size;
425 struct pt_regs *child_ptregs;
430 * For SMP idle threads, fork_by_hand() calls do_fork with
437 stack = ((struct switch_stack *) regs) - 1;
439 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
440 child_stack = (struct switch_stack *) child_ptregs - 1;
442 /* copy parent's switch_stack & pt_regs to child: */
443 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
445 rbs = (unsigned long) current + IA64_RBS_OFFSET;
446 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
447 rbs_size = stack->ar_bspstore - rbs;
449 /* copy the parent's register backing store to the child: */
450 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
452 if (likely(user_mode(child_ptregs))) {
453 if (clone_flags & CLONE_SETTLS)
454 child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */
455 if (user_stack_base) {
456 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
457 child_ptregs->ar_bspstore = user_stack_base;
458 child_ptregs->ar_rnat = 0;
459 child_ptregs->loadrs = 0;
463 * Note: we simply preserve the relative position of
464 * the stack pointer here. There is no need to
465 * allocate a scratch area here, since that will have
466 * been taken care of by the caller of sys_clone()
469 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
470 child_ptregs->r13 = (unsigned long) p; /* set `current' pointer */
472 child_stack->ar_bspstore = child_rbs + rbs_size;
473 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
475 /* copy parts of thread_struct: */
476 p->thread.ksp = (unsigned long) child_stack - 16;
478 /* stop some PSR bits from being inherited.
479 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
480 * therefore we must specify them explicitly here and not include them in
481 * IA64_PSR_BITS_TO_CLEAR.
483 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
484 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
487 * NOTE: The calling convention considers all floating point
488 * registers in the high partition (fph) to be scratch. Since
489 * the only way to get to this point is through a system call,
490 * we know that the values in fph are all dead. Hence, there
491 * is no need to inherit the fph state from the parent to the
492 * child and all we have to do is to make sure that
493 * IA64_THREAD_FPH_VALID is cleared in the child.
495 * XXX We could push this optimization a bit further by
496 * clearing IA64_THREAD_FPH_VALID on ANY system call.
497 * However, it's not clear this is worth doing. Also, it
498 * would be a slight deviation from the normal Linux system
499 * call behavior where scratch registers are preserved across
500 * system calls (unless used by the system call itself).
502 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
503 | IA64_THREAD_PM_VALID)
504 # define THREAD_FLAGS_TO_SET 0
505 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
506 | THREAD_FLAGS_TO_SET);
507 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
509 #ifdef CONFIG_PERFMON
510 if (current->thread.pfm_context)
511 pfm_inherit(p, child_ptregs);
517 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
519 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
520 unsigned long uninitialized_var(ip); /* GCC be quiet */
521 elf_greg_t *dst = arg;
526 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
528 if (unw_unwind_to_user(info) < 0)
531 unw_get_sp(info, &sp);
532 pt = (struct pt_regs *) (sp + 16);
534 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
536 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
539 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
545 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
546 * predicate registers (p0-p63)
549 * ar.rsc ar.bsp ar.bspstore ar.rnat
550 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
554 for (i = 1, mask = (1UL << i); i < 32; ++i) {
555 unw_get_gr(info, i, &dst[i], &nat);
561 unw_get_pr(info, &dst[33]);
563 for (i = 0; i < 8; ++i)
564 unw_get_br(info, i, &dst[34 + i]);
566 unw_get_rp(info, &ip);
567 dst[42] = ip + ia64_psr(pt)->ri;
569 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
571 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
573 * For bsp and bspstore, unw_get_ar() would return the kernel
574 * addresses, but we need the user-level addresses instead:
576 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
577 dst[47] = pt->ar_bspstore;
579 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
580 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
581 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
582 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
583 unw_get_ar(info, UNW_AR_LC, &dst[53]);
584 unw_get_ar(info, UNW_AR_EC, &dst[54]);
585 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
586 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
590 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
592 elf_fpreg_t *dst = arg;
595 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
597 if (unw_unwind_to_user(info) < 0)
600 /* f0 is 0.0, f1 is 1.0 */
602 for (i = 2; i < 32; ++i)
603 unw_get_fr(info, i, dst + i);
605 ia64_flush_fph(task);
606 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
607 memcpy(dst + 32, task->thread.fph, 96*16);
611 do_copy_regs (struct unw_frame_info *info, void *arg)
613 do_copy_task_regs(current, info, arg);
617 do_dump_fpu (struct unw_frame_info *info, void *arg)
619 do_dump_task_fpu(current, info, arg);
623 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
625 unw_init_running(do_copy_regs, dst);
629 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
631 unw_init_running(do_dump_fpu, dst);
632 return 1; /* f0-f31 are always valid so we always return 1 */
636 sys_execve (const char __user *filename, char __user * __user *argv, char __user * __user *envp,
637 struct pt_regs *regs)
642 fname = getname(filename);
643 error = PTR_ERR(fname);
646 error = do_execve(fname, argv, envp, regs);
653 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
655 extern void start_kernel_thread (void);
656 unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
658 struct switch_stack sw;
662 memset(®s, 0, sizeof(regs));
663 regs.pt.cr_iip = helper_fptr[0]; /* set entry point (IP) */
664 regs.pt.r1 = helper_fptr[1]; /* set GP */
665 regs.pt.r9 = (unsigned long) fn; /* 1st argument */
666 regs.pt.r11 = (unsigned long) arg; /* 2nd argument */
667 /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read. */
668 regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
669 regs.pt.cr_ifs = 1UL << 63; /* mark as valid, empty frame */
670 regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
671 regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
672 regs.sw.pr = (1 << PRED_KERNEL_STACK);
673 return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s.pt, 0, NULL, NULL);
675 EXPORT_SYMBOL(kernel_thread);
677 /* This gets called from kernel_thread() via ia64_invoke_thread_helper(). */
679 kernel_thread_helper (int (*fn)(void *), void *arg)
685 * Flush thread state. This is called when a thread does an execve().
690 /* drop floating-point and debug-register state if it exists: */
691 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
692 ia64_drop_fpu(current);
696 * Clean up state associated with current thread. This is called when
697 * the thread calls exit().
703 ia64_drop_fpu(current);
704 #ifdef CONFIG_PERFMON
705 /* if needed, stop monitoring and flush state to perfmon context */
706 if (current->thread.pfm_context)
707 pfm_exit_thread(current);
709 /* free debug register resources */
710 if (current->thread.flags & IA64_THREAD_DBG_VALID)
711 pfm_release_debug_registers(current);
716 get_wchan (struct task_struct *p)
718 struct unw_frame_info info;
722 if (!p || p == current || p->state == TASK_RUNNING)
726 * Note: p may not be a blocked task (it could be current or
727 * another process running on some other CPU. Rather than
728 * trying to determine if p is really blocked, we just assume
729 * it's blocked and rely on the unwind routines to fail
730 * gracefully if the process wasn't really blocked after all.
733 unw_init_from_blocked_task(&info, p);
735 if (p->state == TASK_RUNNING)
737 if (unw_unwind(&info) < 0)
739 unw_get_ip(&info, &ip);
740 if (!in_sched_functions(ip))
742 } while (count++ < 16);
749 pal_power_mgmt_info_u_t power_info[8];
750 unsigned long min_power;
751 int i, min_power_state;
753 if (ia64_pal_halt_info(power_info) != 0)
757 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
758 for (i = 1; i < 8; ++i)
759 if (power_info[i].pal_power_mgmt_info_s.im
760 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
761 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
766 ia64_pal_halt(min_power_state);
769 void machine_shutdown(void)
771 #ifdef CONFIG_HOTPLUG_CPU
774 for_each_online_cpu(cpu) {
775 if (cpu != smp_processor_id())
780 kexec_disable_iosapic();
785 machine_restart (char *restart_cmd)
787 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
788 (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
794 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
799 machine_power_off (void)