xref: /freebsd/sys/powerpc/powerpc/machdep.c (revision a0409676120c1e558d0ade943019934e0f15118d)
1 /*-
2  * Copyright (C) 1995, 1996 Wolfgang Solfrank.
3  * Copyright (C) 1995, 1996 TooLs GmbH.
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *      This product includes software developed by TooLs GmbH.
17  * 4. The name of TooLs GmbH may not be used to endorse or promote products
18  *    derived from this software without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
21  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23  * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
25  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
26  * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
27  * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
28  * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
29  * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30  */
31 /*-
32  * Copyright (C) 2001 Benno Rice
33  * All rights reserved.
34  *
35  * Redistribution and use in source and binary forms, with or without
36  * modification, are permitted provided that the following conditions
37  * are met:
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  *
44  * THIS SOFTWARE IS PROVIDED BY Benno Rice ``AS IS'' AND ANY EXPRESS OR
45  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
46  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
47  * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
48  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
49  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
50  * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
51  * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
52  * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
53  * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
54  *	$NetBSD: machdep.c,v 1.74.2.1 2000/11/01 16:13:48 tv Exp $
55  */
56 
57 #include <sys/cdefs.h>
58 __FBSDID("$FreeBSD$");
59 
60 #include "opt_ddb.h"
61 #include "opt_kstack_pages.h"
62 #include "opt_platform.h"
63 
64 #include <sys/param.h>
65 #include <sys/proc.h>
66 #include <sys/systm.h>
67 #include <sys/bio.h>
68 #include <sys/buf.h>
69 #include <sys/bus.h>
70 #include <sys/cons.h>
71 #include <sys/cpu.h>
72 #include <sys/eventhandler.h>
73 #include <sys/exec.h>
74 #include <sys/imgact.h>
75 #include <sys/kdb.h>
76 #include <sys/kernel.h>
77 #include <sys/ktr.h>
78 #include <sys/linker.h>
79 #include <sys/lock.h>
80 #include <sys/malloc.h>
81 #include <sys/mbuf.h>
82 #include <sys/msgbuf.h>
83 #include <sys/mutex.h>
84 #include <sys/ptrace.h>
85 #include <sys/reboot.h>
86 #include <sys/rwlock.h>
87 #include <sys/signalvar.h>
88 #include <sys/syscallsubr.h>
89 #include <sys/sysctl.h>
90 #include <sys/sysent.h>
91 #include <sys/sysproto.h>
92 #include <sys/ucontext.h>
93 #include <sys/uio.h>
94 #include <sys/vmmeter.h>
95 #include <sys/vnode.h>
96 
97 #include <net/netisr.h>
98 
99 #include <vm/vm.h>
100 #include <vm/vm_extern.h>
101 #include <vm/vm_kern.h>
102 #include <vm/vm_page.h>
103 #include <vm/vm_phys.h>
104 #include <vm/vm_map.h>
105 #include <vm/vm_object.h>
106 #include <vm/vm_pager.h>
107 
108 #include <machine/altivec.h>
109 #ifndef __powerpc64__
110 #include <machine/bat.h>
111 #endif
112 #include <machine/cpu.h>
113 #include <machine/elf.h>
114 #include <machine/fpu.h>
115 #include <machine/hid.h>
116 #include <machine/ifunc.h>
117 #include <machine/kdb.h>
118 #include <machine/md_var.h>
119 #include <machine/metadata.h>
120 #include <machine/mmuvar.h>
121 #include <machine/pcb.h>
122 #include <machine/reg.h>
123 #include <machine/sigframe.h>
124 #include <machine/spr.h>
125 #include <machine/trap.h>
126 #include <machine/vmparam.h>
127 #include <machine/ofw_machdep.h>
128 
129 #include <ddb/ddb.h>
130 
131 #include <dev/ofw/openfirm.h>
132 #include <dev/ofw/ofw_subr.h>
133 
134 int cold = 1;
135 #ifdef __powerpc64__
136 int cacheline_size = 128;
137 #else
138 int cacheline_size = 32;
139 #endif
140 int hw_direct_map = 1;
141 
142 #ifdef BOOKE
143 extern vm_paddr_t kernload;
144 #endif
145 
146 extern void *ap_pcpu;
147 
148 struct pcpu __pcpu[MAXCPU] __aligned(PAGE_SIZE);
149 static char init_kenv[2048];
150 
151 static struct trapframe frame0;
152 
153 char		machine[] = "powerpc";
154 SYSCTL_STRING(_hw, HW_MACHINE, machine, CTLFLAG_RD, machine, 0, "");
155 
156 static void	cpu_startup(void *);
157 SYSINIT(cpu, SI_SUB_CPU, SI_ORDER_FIRST, cpu_startup, NULL);
158 
159 SYSCTL_INT(_machdep, CPU_CACHELINE, cacheline_size,
160 	   CTLFLAG_RD, &cacheline_size, 0, "");
161 
162 uintptr_t	powerpc_init(vm_offset_t, vm_offset_t, vm_offset_t, void *,
163 		    uint32_t);
164 
165 static void	fake_preload_metadata(void);
166 
167 long		Maxmem = 0;
168 long		realmem = 0;
169 
170 /* Default MSR values set in the AIM/Book-E early startup code */
171 register_t	psl_kernset;
172 register_t	psl_userset;
173 register_t	psl_userstatic;
174 #ifdef __powerpc64__
175 register_t	psl_userset32;
176 #endif
177 
178 struct kva_md_info kmi;
179 
180 static void
181 cpu_startup(void *dummy)
182 {
183 
184 	/*
185 	 * Initialise the decrementer-based clock.
186 	 */
187 	decr_init();
188 
189 	/*
190 	 * Good {morning,afternoon,evening,night}.
191 	 */
192 	cpu_setup(PCPU_GET(cpuid));
193 
194 #ifdef PERFMON
195 	perfmon_init();
196 #endif
197 	printf("real memory  = %ju (%ju MB)\n", ptoa((uintmax_t)physmem),
198 	    ptoa((uintmax_t)physmem) / 1048576);
199 	realmem = physmem;
200 
201 	if (bootverbose)
202 		printf("available KVA = %zu (%zu MB)\n",
203 		    virtual_end - virtual_avail,
204 		    (virtual_end - virtual_avail) / 1048576);
205 
206 	/*
207 	 * Display any holes after the first chunk of extended memory.
208 	 */
209 	if (bootverbose) {
210 		int indx;
211 
212 		printf("Physical memory chunk(s):\n");
213 		for (indx = 0; phys_avail[indx + 1] != 0; indx += 2) {
214 			vm_paddr_t size1 =
215 			    phys_avail[indx + 1] - phys_avail[indx];
216 
217 			#ifdef __powerpc64__
218 			printf("0x%016jx - 0x%016jx, %ju bytes (%ju pages)\n",
219 			#else
220 			printf("0x%09jx - 0x%09jx, %ju bytes (%ju pages)\n",
221 			#endif
222 			    (uintmax_t)phys_avail[indx],
223 			    (uintmax_t)phys_avail[indx + 1] - 1,
224 			    (uintmax_t)size1, (uintmax_t)size1 / PAGE_SIZE);
225 		}
226 	}
227 
228 	vm_ksubmap_init(&kmi);
229 
230 	printf("avail memory = %ju (%ju MB)\n",
231 	    ptoa((uintmax_t)vm_free_count()),
232 	    ptoa((uintmax_t)vm_free_count()) / 1048576);
233 
234 	/*
235 	 * Set up buffers, so they can be used to read disk labels.
236 	 */
237 	bufinit();
238 	vm_pager_bufferinit();
239 }
240 
241 extern vm_offset_t	__startkernel, __endkernel;
242 extern unsigned char	__bss_start[];
243 extern unsigned char	__sbss_start[];
244 extern unsigned char	__sbss_end[];
245 extern unsigned char	_end[];
246 
247 void aim_early_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry,
248     void *mdp, uint32_t mdp_cookie);
249 void aim_cpu_init(vm_offset_t toc);
250 void booke_cpu_init(void);
251 
252 #ifdef DDB
253 static void	load_external_symtab(void);
254 #endif
255 
256 uintptr_t
257 powerpc_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry, void *mdp,
258     uint32_t mdp_cookie)
259 {
260 	struct		pcpu *pc;
261 	struct cpuref	bsp;
262 	vm_offset_t	startkernel, endkernel;
263 	char		*env;
264 	void		*kmdp = NULL;
265         bool		ofw_bootargs = false;
266 #ifdef DDB
267 	bool		symbols_provided = false;
268 	vm_offset_t ksym_start;
269 	vm_offset_t ksym_end;
270 	vm_offset_t ksym_sz;
271 #endif
272 
273 	/* First guess at start/end kernel positions */
274 	startkernel = __startkernel;
275 	endkernel = __endkernel;
276 
277 	/*
278 	 * If the metadata pointer cookie is not set to the magic value,
279 	 * the number in mdp should be treated as nonsense.
280 	 */
281 	if (mdp_cookie != 0xfb5d104d)
282 		mdp = NULL;
283 
284 #if !defined(BOOKE)
285 	/*
286 	 * On BOOKE the BSS is already cleared and some variables
287 	 * initialized.  Do not wipe them out.
288 	 */
289 	bzero(__sbss_start, __sbss_end - __sbss_start);
290 	bzero(__bss_start, _end - __bss_start);
291 #endif
292 
293 	cpu_feature_setup();
294 
295 #ifdef AIM
296 	aim_early_init(fdt, toc, ofentry, mdp, mdp_cookie);
297 #endif
298 
299 	/*
300 	 * At this point, we are executing in our correct memory space.
301 	 * Book-E started there, and AIM has done an rfi and restarted
302 	 * execution from _start.
303 	 *
304 	 * We may still be in real mode, however. If we are running out of
305 	 * the direct map on 64 bit, this is possible to do.
306 	 */
307 
308 	/*
309 	 * Parse metadata if present and fetch parameters.  Must be done
310 	 * before console is inited so cninit gets the right value of
311 	 * boothowto.
312 	 */
313 	if (mdp != NULL) {
314 		/*
315 		 * Starting up from loader.
316 		 *
317 		 * Full metadata has been provided, but we need to figure
318 		 * out the correct address to relocate it to.
319 		 */
320 		char *envp = NULL;
321 		uintptr_t md_offset = 0;
322 		vm_paddr_t kernelstartphys, kernelendphys;
323 
324 #ifdef AIM
325 		if ((uintptr_t)&powerpc_init > DMAP_BASE_ADDRESS)
326 			md_offset = DMAP_BASE_ADDRESS;
327 #else /* BOOKE */
328 		md_offset = VM_MIN_KERNEL_ADDRESS - kernload;
329 #endif
330 
331 		preload_metadata = mdp;
332 		if (md_offset > 0) {
333 			/* Translate phys offset into DMAP offset. */
334 			preload_metadata += md_offset;
335 			preload_bootstrap_relocate(md_offset);
336 		}
337 		kmdp = preload_search_by_type("elf kernel");
338 		if (kmdp != NULL) {
339 			boothowto = MD_FETCH(kmdp, MODINFOMD_HOWTO, int);
340 			envp = MD_FETCH(kmdp, MODINFOMD_ENVP, char *);
341 			if (envp != NULL)
342 				envp += md_offset;
343 			init_static_kenv(envp, 0);
344 			if (fdt == 0) {
345 				fdt = MD_FETCH(kmdp, MODINFOMD_DTBP, uintptr_t);
346 				if (fdt != 0)
347 					fdt += md_offset;
348 			}
349 			kernelstartphys = MD_FETCH(kmdp, MODINFO_ADDR,
350 			    vm_offset_t);
351 			/* kernelstartphys is already relocated. */
352 			kernelendphys = MD_FETCH(kmdp, MODINFOMD_KERNEND,
353 			    vm_offset_t);
354 			if (kernelendphys != 0)
355 				kernelendphys += md_offset;
356 			endkernel = ulmax(endkernel, kernelendphys);
357 #ifdef DDB
358 			ksym_start = MD_FETCH(kmdp, MODINFOMD_SSYM, uintptr_t);
359 			ksym_end = MD_FETCH(kmdp, MODINFOMD_ESYM, uintptr_t);
360 			ksym_sz = *(Elf_Size*)ksym_start;
361 
362 			db_fetch_ksymtab(ksym_start, ksym_end, md_offset);
363 			/* Symbols provided by loader. */
364 			symbols_provided = true;
365 #endif
366 		}
367 	} else {
368 		/*
369 		 * Self-loading kernel, we have to fake up metadata.
370 		 *
371 		 * Since we are creating the metadata from the final
372 		 * memory space, we don't need to call
373 		 * preload_boostrap_relocate().
374 		 */
375 		fake_preload_metadata();
376 		kmdp = preload_search_by_type("elf kernel");
377 		init_static_kenv(init_kenv, sizeof(init_kenv));
378 		ofw_bootargs = true;
379 	}
380 
381 	/* Store boot environment state */
382 	OF_initial_setup((void *)fdt, NULL, (int (*)(void *))ofentry);
383 
384 	/*
385 	 * Init params/tunables that can be overridden by the loader
386 	 */
387 	init_param1();
388 
389 	/*
390 	 * Start initializing proc0 and thread0.
391 	 */
392 	proc_linkup0(&proc0, &thread0);
393 	thread0.td_frame = &frame0;
394 #ifdef __powerpc64__
395 	__asm __volatile("mr 13,%0" :: "r"(&thread0));
396 #else
397 	__asm __volatile("mr 2,%0" :: "r"(&thread0));
398 #endif
399 
400 	/*
401 	 * Init mutexes, which we use heavily in PMAP
402 	 */
403 	mutex_init();
404 
405 	/*
406 	 * Install the OF client interface
407 	 */
408 	OF_bootstrap();
409 
410 #ifdef DDB
411 	if (!symbols_provided && hw_direct_map)
412 		load_external_symtab();
413 #endif
414 
415 	if (ofw_bootargs)
416 		ofw_parse_bootargs();
417 
418 	/*
419 	 * Initialize the console before printing anything.
420 	 */
421 	cninit();
422 
423 #ifdef AIM
424 	aim_cpu_init(toc);
425 #else /* BOOKE */
426 	booke_cpu_init();
427 
428 	/* Make sure the kernel icache is valid before we go too much further */
429 	__syncicache((caddr_t)startkernel, endkernel - startkernel);
430 #endif
431 
432 	/*
433 	 * Choose a platform module so we can get the physical memory map.
434 	 */
435 
436 	platform_probe_and_attach();
437 
438 	/*
439 	 * Set up per-cpu data for the BSP now that the platform can tell
440 	 * us which that is.
441 	 */
442 	if (platform_smp_get_bsp(&bsp) != 0)
443 		bsp.cr_cpuid = 0;
444 	pc = &__pcpu[bsp.cr_cpuid];
445 	__asm __volatile("mtsprg 0, %0" :: "r"(pc));
446 	pcpu_init(pc, bsp.cr_cpuid, sizeof(struct pcpu));
447 	pc->pc_curthread = &thread0;
448 	thread0.td_oncpu = bsp.cr_cpuid;
449 	pc->pc_cpuid = bsp.cr_cpuid;
450 	pc->pc_hwref = bsp.cr_hwref;
451 
452 	/*
453 	 * Init KDB
454 	 */
455 	kdb_init();
456 
457 	/*
458 	 * Bring up MMU
459 	 */
460 	pmap_mmu_init();
461 	link_elf_ireloc(kmdp);
462 	pmap_bootstrap(startkernel, endkernel);
463 	mtmsr(psl_kernset & ~PSL_EE);
464 
465 	/*
466 	 * Initialize params/tunables that are derived from memsize
467 	 */
468 	init_param2(physmem);
469 
470 	/*
471 	 * Grab booted kernel's name
472 	 */
473         env = kern_getenv("kernelname");
474         if (env != NULL) {
475 		strlcpy(kernelname, env, sizeof(kernelname));
476 		freeenv(env);
477 	}
478 
479 	/*
480 	 * Finish setting up thread0.
481 	 */
482 	thread0.td_pcb = (struct pcb *)
483 	    ((thread0.td_kstack + thread0.td_kstack_pages * PAGE_SIZE -
484 	    sizeof(struct pcb)) & ~15UL);
485 	bzero((void *)thread0.td_pcb, sizeof(struct pcb));
486 	pc->pc_curpcb = thread0.td_pcb;
487 
488 	/* Initialise the message buffer. */
489 	msgbufinit(msgbufp, msgbufsize);
490 
491 #ifdef KDB
492 	if (boothowto & RB_KDB)
493 		kdb_enter(KDB_WHY_BOOTFLAGS,
494 		    "Boot flags requested debugger");
495 #endif
496 
497 	return (((uintptr_t)thread0.td_pcb -
498 	    (sizeof(struct callframe) - 3*sizeof(register_t))) & ~15UL);
499 }
500 
501 #ifdef DDB
502 /*
503  * On powernv and some booke systems, we might not have symbols loaded via
504  * loader. However, if the user passed the kernel in as the initrd as well,
505  * we can manually load it via reinterpreting the initrd copy of the kernel.
506  *
507  * In the BOOKE case, we don't actually have a DMAP yet, so we have to use
508  * temporary maps to inspect the memory, but write DMAP addresses to the
509  * configuration variables.
510  */
511 static void
512 load_external_symtab(void) {
513 	phandle_t chosen;
514 	vm_paddr_t start, end;
515 	pcell_t cell[2];
516 	ssize_t size;
517 	u_char *kernelimg;		/* Temporary map */
518 	u_char *kernelimg_final;	/* Final location */
519 
520 	int i;
521 
522 	Elf_Ehdr *ehdr;
523 	Elf_Phdr *phdr;
524 	Elf_Shdr *shdr;
525 
526 	vm_offset_t ksym_start, ksym_sz, kstr_start, kstr_sz,
527 	    ksym_start_final, kstr_start_final;
528 
529 	if (!hw_direct_map)
530 		return;
531 
532 	chosen = OF_finddevice("/chosen");
533 	if (chosen <= 0)
534 		return;
535 
536 	if (!OF_hasprop(chosen, "linux,initrd-start") ||
537 	    !OF_hasprop(chosen, "linux,initrd-end"))
538 		return;
539 
540 	size = OF_getencprop(chosen, "linux,initrd-start", cell, sizeof(cell));
541 	if (size == 4)
542 		start = cell[0];
543 	else if (size == 8)
544 		start = (uint64_t)cell[0] << 32 | cell[1];
545 	else
546 		return;
547 
548 	size = OF_getencprop(chosen, "linux,initrd-end", cell, sizeof(cell));
549 	if (size == 4)
550 		end = cell[0];
551 	else if (size == 8)
552 		end = (uint64_t)cell[0] << 32 | cell[1];
553 	else
554 		return;
555 
556 	if (!(end - start > 0))
557 		return;
558 
559 	kernelimg_final = (u_char *) PHYS_TO_DMAP(start);
560 #ifdef	AIM
561 	kernelimg = kernelimg_final;
562 #else	/* BOOKE */
563 	kernelimg = (u_char *)pmap_early_io_map(start, PAGE_SIZE);
564 #endif
565 	ehdr = (Elf_Ehdr *)kernelimg;
566 
567 	if (!IS_ELF(*ehdr)) {
568 #ifdef	BOOKE
569 		pmap_early_io_unmap(start, PAGE_SIZE);
570 #endif
571 		return;
572 	}
573 
574 #ifdef	BOOKE
575 	pmap_early_io_unmap(start, PAGE_SIZE);
576 	kernelimg = (u_char *)pmap_early_io_map(start, (end - start));
577 #endif
578 
579 	phdr = (Elf_Phdr *)(kernelimg + ehdr->e_phoff);
580 	shdr = (Elf_Shdr *)(kernelimg + ehdr->e_shoff);
581 
582 	ksym_start = 0;
583 	ksym_sz = 0;
584 	ksym_start_final = 0;
585 	kstr_start = 0;
586 	kstr_sz = 0;
587 	kstr_start_final = 0;
588 	for (i = 0; i < ehdr->e_shnum; i++) {
589 		if (shdr[i].sh_type == SHT_SYMTAB) {
590 			ksym_start = (vm_offset_t)(kernelimg +
591 			    shdr[i].sh_offset);
592 			ksym_start_final = (vm_offset_t)
593 			    (kernelimg_final + shdr[i].sh_offset);
594 			ksym_sz = (vm_offset_t)(shdr[i].sh_size);
595 			kstr_start = (vm_offset_t)(kernelimg +
596 			    shdr[shdr[i].sh_link].sh_offset);
597 			kstr_start_final = (vm_offset_t)
598 			    (kernelimg_final +
599 			    shdr[shdr[i].sh_link].sh_offset);
600 
601 			kstr_sz = (vm_offset_t)
602 			    (shdr[shdr[i].sh_link].sh_size);
603 		}
604 	}
605 
606 	if (ksym_start != 0 && kstr_start != 0 && ksym_sz != 0 &&
607 	    kstr_sz != 0 && ksym_start < kstr_start) {
608 		/*
609 		 * We can't use db_fetch_ksymtab() here, because we need to
610 		 * feed in DMAP addresses that are not mapped yet on booke.
611 		 *
612 		 * Write the variables directly, where db_init() will pick
613 		 * them up later, after the DMAP is up.
614 		 */
615 		ksymtab = ksym_start_final;
616 		ksymtab_size = ksym_sz;
617 		kstrtab = kstr_start_final;
618 		ksymtab_relbase = (__startkernel - KERNBASE);
619 	}
620 
621 #ifdef	BOOKE
622 	pmap_early_io_unmap(start, (end - start));
623 #endif
624 
625 };
626 #endif
627 
628 /*
629  * When not being loaded from loader, we need to create our own metadata
630  * so we can interact with the kernel linker.
631  */
632 static void
633 fake_preload_metadata(void) {
634 	/* We depend on dword alignment here. */
635 	static uint32_t fake_preload[36] __aligned(8);
636 	int i = 0;
637 
638 	fake_preload[i++] = MODINFO_NAME;
639 	fake_preload[i++] = strlen("kernel") + 1;
640 	strcpy((char*)&fake_preload[i], "kernel");
641 	/* ['k' 'e' 'r' 'n'] ['e' 'l' '\0' ..] */
642 	i += 2;
643 
644 	fake_preload[i++] = MODINFO_TYPE;
645 	fake_preload[i++] = strlen("elf kernel") + 1;
646 	strcpy((char*)&fake_preload[i], "elf kernel");
647 	/* ['e' 'l' 'f' ' '] ['k' 'e' 'r' 'n'] ['e' 'l' '\0' ..] */
648 	i += 3;
649 
650 #ifdef __powerpc64__
651 	/* Padding -- Fields start on u_long boundaries */
652 	fake_preload[i++] = 0;
653 #endif
654 
655 	fake_preload[i++] = MODINFO_ADDR;
656 	fake_preload[i++] = sizeof(vm_offset_t);
657 	*(vm_offset_t *)&fake_preload[i] =
658 	    (vm_offset_t)(__startkernel);
659 	i += (sizeof(vm_offset_t) / 4);
660 
661 	fake_preload[i++] = MODINFO_SIZE;
662 	fake_preload[i++] = sizeof(vm_offset_t);
663 	*(vm_offset_t *)&fake_preload[i] =
664 	    (vm_offset_t)(__endkernel) - (vm_offset_t)(__startkernel);
665 	i += (sizeof(vm_offset_t) / 4);
666 
667 	/*
668 	 * MODINFOMD_SSYM and MODINFOMD_ESYM cannot be provided here,
669 	 * as the memory comes from outside the loaded ELF sections.
670 	 *
671 	 * If the symbols are being provided by other means (MFS), the
672 	 * tables will be loaded into the debugger directly.
673 	 */
674 
675 	/* Null field at end to mark end of data. */
676 	fake_preload[i++] = 0;
677 	fake_preload[i] = 0;
678 	preload_metadata = (void*)fake_preload;
679 }
680 
681 /*
682  * Flush the D-cache for non-DMA I/O so that the I-cache can
683  * be made coherent later.
684  */
685 void
686 cpu_flush_dcache(void *ptr, size_t len)
687 {
688 	register_t addr, off;
689 
690 	/*
691 	 * Align the address to a cacheline and adjust the length
692 	 * accordingly. Then round the length to a multiple of the
693 	 * cacheline for easy looping.
694 	 */
695 	addr = (uintptr_t)ptr;
696 	off = addr & (cacheline_size - 1);
697 	addr -= off;
698 	len = roundup2(len + off, cacheline_size);
699 
700 	while (len > 0) {
701 		__asm __volatile ("dcbf 0,%0" :: "r"(addr));
702 		__asm __volatile ("sync");
703 		addr += cacheline_size;
704 		len -= cacheline_size;
705 	}
706 }
707 
708 int
709 ptrace_set_pc(struct thread *td, unsigned long addr)
710 {
711 	struct trapframe *tf;
712 
713 	tf = td->td_frame;
714 	tf->srr0 = (register_t)addr;
715 
716 	return (0);
717 }
718 
719 void
720 spinlock_enter(void)
721 {
722 	struct thread *td;
723 	register_t msr;
724 
725 	td = curthread;
726 	if (td->td_md.md_spinlock_count == 0) {
727 		nop_prio_mhigh();
728 		msr = intr_disable();
729 		td->td_md.md_spinlock_count = 1;
730 		td->td_md.md_saved_msr = msr;
731 		critical_enter();
732 	} else
733 		td->td_md.md_spinlock_count++;
734 }
735 
736 void
737 spinlock_exit(void)
738 {
739 	struct thread *td;
740 	register_t msr;
741 
742 	td = curthread;
743 	msr = td->td_md.md_saved_msr;
744 	td->td_md.md_spinlock_count--;
745 	if (td->td_md.md_spinlock_count == 0) {
746 		critical_exit();
747 		intr_restore(msr);
748 		nop_prio_medium();
749 	}
750 }
751 
752 /*
753  * Simple ddb(4) command/hack to view any SPR on the running CPU.
754  * Uses a trivial asm function to perform the mfspr, and rewrites the mfspr
755  * instruction each time.
756  * XXX: Since it uses code modification, it won't work if the kernel code pages
757  * are marked RO.
758  */
759 extern register_t get_spr(int);
760 
761 #ifdef DDB
762 DB_SHOW_COMMAND(spr, db_show_spr)
763 {
764 	register_t spr;
765 	volatile uint32_t *p;
766 	int sprno, saved_sprno;
767 
768 	if (!have_addr)
769 		return;
770 
771 	saved_sprno = sprno = (intptr_t) addr;
772 	sprno = ((sprno & 0x3e0) >> 5) | ((sprno & 0x1f) << 5);
773 	p = (uint32_t *)(void *)&get_spr;
774 #ifdef __powerpc64__
775 #if defined(_CALL_ELF) && _CALL_ELF == 2
776 	/* Account for ELFv2 function prologue. */
777 	p += 2;
778 #else
779 	p = *(volatile uint32_t * volatile *)p;
780 #endif
781 #endif
782 	*p = (*p & ~0x001ff800) | (sprno << 11);
783 	__syncicache(__DEVOLATILE(uint32_t *, p), cacheline_size);
784 	spr = get_spr(sprno);
785 
786 	db_printf("SPR %d(%x): %lx\n", saved_sprno, saved_sprno,
787 	    (unsigned long)spr);
788 }
789 
790 DB_SHOW_COMMAND(frame, db_show_frame)
791 {
792 	struct trapframe *tf;
793 	long reg;
794 	int i;
795 
796 	tf = have_addr ? (struct trapframe *)addr : curthread->td_frame;
797 
798 	/*
799 	 * Everything casts through long to simplify the printing.
800 	 * 'long' is native register size anyway.
801 	 */
802 	db_printf("trap frame %p\n", tf);
803 	for (i = 0; i < nitems(tf->fixreg); i++) {
804 		reg = tf->fixreg[i];
805 		db_printf("  r%d:\t%#lx (%ld)\n", i, reg, reg);
806 	}
807 	reg = tf->lr;
808 	db_printf("  lr:\t%#lx\n", reg);
809 	reg = tf->cr;
810 	db_printf("  cr:\t%#lx\n", reg);
811 	reg = tf->xer;
812 	db_printf("  xer:\t%#lx\n", reg);
813 	reg = tf->ctr;
814 	db_printf("  ctr:\t%#lx (%ld)\n", reg, reg);
815 	reg = tf->srr0;
816 	db_printf("  srr0:\t%#lx\n", reg);
817 	reg = tf->srr1;
818 	db_printf("  srr1:\t%#lx\n", reg);
819 	reg = tf->exc;
820 	db_printf("  exc:\t%#lx\n", reg);
821 	reg = tf->dar;
822 	db_printf("  dar:\t%#lx\n", reg);
823 #ifdef AIM
824 	reg = tf->cpu.aim.dsisr;
825 	db_printf("  dsisr:\t%#lx\n", reg);
826 #else
827 	reg = tf->cpu.booke.esr;
828 	db_printf("  esr:\t%#lx\n", reg);
829 	reg = tf->cpu.booke.dbcr0;
830 	db_printf("  dbcr0:\t%#lx\n", reg);
831 #endif
832 }
833 #endif
834 
835 #undef bzero
836 void
837 bzero(void *buf, size_t len)
838 {
839 	caddr_t	p;
840 
841 	p = buf;
842 
843 	while (((vm_offset_t) p & (sizeof(u_long) - 1)) && len) {
844 		*p++ = 0;
845 		len--;
846 	}
847 
848 	while (len >= sizeof(u_long) * 8) {
849 		*(u_long*) p = 0;
850 		*((u_long*) p + 1) = 0;
851 		*((u_long*) p + 2) = 0;
852 		*((u_long*) p + 3) = 0;
853 		len -= sizeof(u_long) * 8;
854 		*((u_long*) p + 4) = 0;
855 		*((u_long*) p + 5) = 0;
856 		*((u_long*) p + 6) = 0;
857 		*((u_long*) p + 7) = 0;
858 		p += sizeof(u_long) * 8;
859 	}
860 
861 	while (len >= sizeof(u_long)) {
862 		*(u_long*) p = 0;
863 		len -= sizeof(u_long);
864 		p += sizeof(u_long);
865 	}
866 
867 	while (len) {
868 		*p++ = 0;
869 		len--;
870 	}
871 }
872 
873 /* __stack_chk_fail_local() is called in secure-plt (32-bit). */
874 #if !defined(__powerpc64__)
875 extern void __stack_chk_fail(void);
876 void __stack_chk_fail_local(void);
877 
878 void
879 __stack_chk_fail_local(void)
880 {
881 
882 	__stack_chk_fail();
883 }
884 #endif
885