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/resume_user_mode.h>
36 #include <linux/rcupdate.h>
39 #include <asm/delay.h>
42 #include <asm/kexec.h>
43 #include <asm/processor.h>
45 #include <asm/switch_to.h>
46 #include <asm/tlbflush.h>
47 #include <linux/uaccess.h>
48 #include <asm/unwind.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_power_off) (void);
61 EXPORT_SYMBOL(pm_power_off);
64 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
66 unsigned long ip, sp, bsp;
67 const char *loglvl = arg;
69 printk("%s\nCall Trace:\n", loglvl);
71 unw_get_ip(info, &ip);
75 unw_get_sp(info, &sp);
76 unw_get_bsp(info, &bsp);
77 printk("%s [<%016lx>] %pS\n"
78 " sp=%016lx bsp=%016lx\n",
79 loglvl, ip, (void *)ip, sp, bsp);
80 } while (unw_unwind(info) >= 0);
84 show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
87 unw_init_running(ia64_do_show_stack, (void *)loglvl);
89 struct unw_frame_info info;
91 unw_init_from_blocked_task(&info, task);
92 ia64_do_show_stack(&info, (void *)loglvl);
97 show_regs (struct pt_regs *regs)
99 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
103 show_regs_print_info(KERN_DEFAULT);
104 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
105 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
106 init_utsname()->release);
107 printk("ip is at %pS\n", (void *)ip);
108 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
109 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
110 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
111 regs->ar_rnat, regs->ar_bspstore, regs->pr);
112 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
113 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
114 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
115 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
116 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
117 regs->f6.u.bits[1], regs->f6.u.bits[0],
118 regs->f7.u.bits[1], regs->f7.u.bits[0]);
119 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
120 regs->f8.u.bits[1], regs->f8.u.bits[0],
121 regs->f9.u.bits[1], regs->f9.u.bits[0]);
122 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
123 regs->f10.u.bits[1], regs->f10.u.bits[0],
124 regs->f11.u.bits[1], regs->f11.u.bits[0]);
126 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
127 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
128 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
129 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
130 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
131 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
132 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
133 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
134 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
136 if (user_mode(regs)) {
137 /* print the stacked registers */
138 unsigned long val, *bsp, ndirty;
139 int i, sof, is_nat = 0;
141 sof = regs->cr_ifs & 0x7f; /* size of frame */
142 ndirty = (regs->loadrs >> 19);
143 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
144 for (i = 0; i < sof; ++i) {
145 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
146 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
147 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
150 show_stack(NULL, NULL, KERN_DEFAULT);
153 /* local support for deprecated console_print */
155 console_print(const char *s)
157 printk(KERN_EMERG "%s", s);
161 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
163 if (fsys_mode(current, &scr->pt)) {
165 * defer signal-handling etc. until we return to
168 if (!ia64_psr(&scr->pt)->lp)
169 ia64_psr(&scr->pt)->lp = 1;
173 /* deal with pending signal delivery */
174 if (test_thread_flag(TIF_SIGPENDING) ||
175 test_thread_flag(TIF_NOTIFY_SIGNAL)) {
176 local_irq_enable(); /* force interrupt enable */
177 ia64_do_signal(scr, in_syscall);
180 if (test_thread_flag(TIF_NOTIFY_RESUME)) {
181 local_irq_enable(); /* force interrupt enable */
182 resume_user_mode_work(&scr->pt);
185 /* copy user rbs to kernel rbs */
186 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
187 local_irq_enable(); /* force interrupt enable */
191 local_irq_disable(); /* force interrupt disable */
194 static int __init nohalt_setup(char * str)
196 cpu_idle_poll_ctrl(true);
199 __setup("nohalt", nohalt_setup);
201 #ifdef CONFIG_HOTPLUG_CPU
202 /* We don't actually take CPU down, just spin without interrupts. */
203 static inline void play_dead(void)
205 unsigned int this_cpu = smp_processor_id();
208 __this_cpu_write(cpu_state, CPU_DEAD);
213 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
215 * The above is a point of no-return, the processor is
216 * expected to be in SAL loop now.
221 static inline void play_dead(void)
225 #endif /* CONFIG_HOTPLUG_CPU */
227 void arch_cpu_idle_dead(void)
232 void arch_cpu_idle(void)
234 void (*mark_idle)(int) = ia64_mark_idle;
253 ia64_save_extra (struct task_struct *task)
255 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
256 ia64_save_debug_regs(&task->thread.dbr[0]);
260 ia64_load_extra (struct task_struct *task)
262 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
263 ia64_load_debug_regs(&task->thread.dbr[0]);
267 * Copy the state of an ia-64 thread.
269 * We get here through the following call chain:
271 * from user-level: from kernel:
273 * <clone syscall> <some kernel call frames>
275 * kernel_clone kernel_clone
276 * copy_thread copy_thread
278 * This means that the stack layout is as follows:
280 * +---------------------+ (highest addr)
282 * +---------------------+
283 * | struct switch_stack |
284 * +---------------------+
287 * | | <-- sp (lowest addr)
288 * +---------------------+
290 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
291 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
292 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
293 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
294 * the stack is page aligned and the page size is at least 4KB, this is always the case,
295 * so there is nothing to worry about.
298 copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
300 unsigned long clone_flags = args->flags;
301 unsigned long user_stack_base = args->stack;
302 unsigned long user_stack_size = args->stack_size;
303 unsigned long tls = args->tls;
304 extern char ia64_ret_from_clone;
305 struct switch_stack *child_stack, *stack;
306 unsigned long rbs, child_rbs, rbs_size;
307 struct pt_regs *child_ptregs;
308 struct pt_regs *regs = current_pt_regs();
311 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
312 child_stack = (struct switch_stack *) child_ptregs - 1;
314 rbs = (unsigned long) current + IA64_RBS_OFFSET;
315 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
317 /* copy parts of thread_struct: */
318 p->thread.ksp = (unsigned long) child_stack - 16;
321 * NOTE: The calling convention considers all floating point
322 * registers in the high partition (fph) to be scratch. Since
323 * the only way to get to this point is through a system call,
324 * we know that the values in fph are all dead. Hence, there
325 * is no need to inherit the fph state from the parent to the
326 * child and all we have to do is to make sure that
327 * IA64_THREAD_FPH_VALID is cleared in the child.
329 * XXX We could push this optimization a bit further by
330 * clearing IA64_THREAD_FPH_VALID on ANY system call.
331 * However, it's not clear this is worth doing. Also, it
332 * would be a slight deviation from the normal Linux system
333 * call behavior where scratch registers are preserved across
334 * system calls (unless used by the system call itself).
336 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
337 | IA64_THREAD_PM_VALID)
338 # define THREAD_FLAGS_TO_SET 0
339 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
340 | THREAD_FLAGS_TO_SET);
342 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
344 if (unlikely(args->fn)) {
345 if (unlikely(args->idle)) {
346 /* fork_idle() called us */
349 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
350 child_stack->r4 = (unsigned long) args->fn;
351 child_stack->r5 = (unsigned long) args->fn_arg;
353 * Preserve PSR bits, except for bits 32-34 and 37-45,
354 * which we can't read.
356 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
357 /* mark as valid, empty frame */
358 child_ptregs->cr_ifs = 1UL << 63;
359 child_stack->ar_fpsr = child_ptregs->ar_fpsr
360 = ia64_getreg(_IA64_REG_AR_FPSR);
361 child_stack->pr = (1 << PRED_KERNEL_STACK);
362 child_stack->ar_bspstore = child_rbs;
363 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
365 /* stop some PSR bits from being inherited.
366 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
367 * therefore we must specify them explicitly here and not include them in
368 * IA64_PSR_BITS_TO_CLEAR.
370 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
371 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
375 stack = ((struct switch_stack *) regs) - 1;
376 /* copy parent's switch_stack & pt_regs to child: */
377 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
379 /* copy the parent's register backing store to the child: */
380 rbs_size = stack->ar_bspstore - rbs;
381 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
382 if (clone_flags & CLONE_SETTLS)
383 child_ptregs->r13 = tls;
384 if (user_stack_base) {
385 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
386 child_ptregs->ar_bspstore = user_stack_base;
387 child_ptregs->ar_rnat = 0;
388 child_ptregs->loadrs = 0;
390 child_stack->ar_bspstore = child_rbs + rbs_size;
391 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
393 /* stop some PSR bits from being inherited.
394 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
395 * therefore we must specify them explicitly here and not include them in
396 * IA64_PSR_BITS_TO_CLEAR.
398 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
399 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
403 asmlinkage long ia64_clone(unsigned long clone_flags, unsigned long stack_start,
404 unsigned long stack_size, unsigned long parent_tidptr,
405 unsigned long child_tidptr, unsigned long tls)
407 struct kernel_clone_args args = {
408 .flags = (lower_32_bits(clone_flags) & ~CSIGNAL),
409 .pidfd = (int __user *)parent_tidptr,
410 .child_tid = (int __user *)child_tidptr,
411 .parent_tid = (int __user *)parent_tidptr,
412 .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL),
413 .stack = stack_start,
414 .stack_size = stack_size,
418 return kernel_clone(&args);
422 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
424 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
426 elf_greg_t *dst = arg;
431 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
433 if (unw_unwind_to_user(info) < 0)
436 unw_get_sp(info, &sp);
437 pt = (struct pt_regs *) (sp + 16);
439 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
441 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
444 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
450 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
451 * predicate registers (p0-p63)
454 * ar.rsc ar.bsp ar.bspstore ar.rnat
455 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
459 for (i = 1, mask = (1UL << i); i < 32; ++i) {
460 unw_get_gr(info, i, &dst[i], &nat);
466 unw_get_pr(info, &dst[33]);
468 for (i = 0; i < 8; ++i)
469 unw_get_br(info, i, &dst[34 + i]);
471 unw_get_rp(info, &ip);
472 dst[42] = ip + ia64_psr(pt)->ri;
474 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
476 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
478 * For bsp and bspstore, unw_get_ar() would return the kernel
479 * addresses, but we need the user-level addresses instead:
481 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
482 dst[47] = pt->ar_bspstore;
484 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
485 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
486 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
487 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
488 unw_get_ar(info, UNW_AR_LC, &dst[53]);
489 unw_get_ar(info, UNW_AR_EC, &dst[54]);
490 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
491 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
495 do_copy_regs (struct unw_frame_info *info, void *arg)
497 do_copy_task_regs(current, info, arg);
501 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
503 unw_init_running(do_copy_regs, dst);
507 * Flush thread state. This is called when a thread does an execve().
512 /* drop floating-point and debug-register state if it exists: */
513 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
514 ia64_drop_fpu(current);
518 * Clean up state associated with a thread. This is called when
519 * the thread calls exit().
522 exit_thread (struct task_struct *tsk)
529 __get_wchan (struct task_struct *p)
531 struct unw_frame_info info;
536 * Note: p may not be a blocked task (it could be current or
537 * another process running on some other CPU. Rather than
538 * trying to determine if p is really blocked, we just assume
539 * it's blocked and rely on the unwind routines to fail
540 * gracefully if the process wasn't really blocked after all.
543 unw_init_from_blocked_task(&info, p);
545 if (task_is_running(p))
547 if (unw_unwind(&info) < 0)
549 unw_get_ip(&info, &ip);
550 if (!in_sched_functions(ip))
552 } while (count++ < 16);
559 pal_power_mgmt_info_u_t power_info[8];
560 unsigned long min_power;
561 int i, min_power_state;
563 if (ia64_pal_halt_info(power_info) != 0)
567 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
568 for (i = 1; i < 8; ++i)
569 if (power_info[i].pal_power_mgmt_info_s.im
570 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
571 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
576 ia64_pal_halt(min_power_state);
579 void machine_shutdown(void)
581 smp_shutdown_nonboot_cpus(reboot_cpu);
584 kexec_disable_iosapic();
589 machine_restart (char *restart_cmd)
591 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
592 efi_reboot(REBOOT_WARM, NULL);
598 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
603 machine_power_off (void)
610 EXPORT_SYMBOL(ia64_delay_loop);