elf_trampoline.c revision 194609
1/*-
2 * Copyright (c) 2005 Olivier Houchard.  All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 *    notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 *    notice, this list of conditions and the following disclaimer in the
11 *    documentation and/or other materials provided with the distribution.
12 *
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
14 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
15 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
16 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
17 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
18 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
19 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
20 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
21 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
22 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
23 */
24
25#include <sys/cdefs.h>
26__FBSDID("$FreeBSD: head/sys/arm/arm/elf_trampoline.c 194609 2009-06-21 21:38:12Z cognet $");
27#include <machine/asm.h>
28#include <sys/param.h>
29#include <sys/elf32.h>
30#include <sys/inflate.h>
31#include <machine/elf.h>
32#include <machine/pte.h>
33#include <machine/cpufunc.h>
34#include <machine/armreg.h>
35
36/*
37 * Since we are compiled outside of the normal kernel build process, we
38 * need to include opt_global.h manually.
39 */
40#include "opt_global.h"
41#include "opt_kernname.h"
42
43extern char kernel_start[];
44extern char kernel_end[];
45
46extern void *_end;
47
48void _start(void);
49void __start(void);
50void __startC(void);
51
52#define GZ_HEAD	0xa
53
54#ifdef CPU_ARM7TDMI
55#define cpu_idcache_wbinv_all	arm7tdmi_cache_flushID
56#elif defined(CPU_ARM8)
57#define cpu_idcache_wbinv_all	arm8_cache_purgeID
58#elif defined(CPU_ARM9)
59#define cpu_idcache_wbinv_all	arm9_idcache_wbinv_all
60#elif defined(CPU_ARM9E)
61#define cpu_idcache_wbinv_all	armv5_ec_idcache_wbinv_all
62#elif defined(CPU_ARM10)
63#define cpu_idcache_wbinv_all	arm10_idcache_wbinv_all
64#elif defined(CPU_SA110) || defined(CPU_SA1110) || defined(CPU_SA1100) || \
65    defined(CPU_IXP12X0)
66#define cpu_idcache_wbinv_all	sa1_cache_purgeID
67#elif defined(CPU_XSCALE_80200) || defined(CPU_XSCALE_80321) || \
68  defined(CPU_XSCALE_PXA2X0) || defined(CPU_XSCALE_IXP425) ||	\
69  defined(CPU_XSCALE_80219)
70#define cpu_idcache_wbinv_all	xscale_cache_purgeID
71#elif defined(CPU_XSCALE_81342)
72#define cpu_idcache_wbinv_all	xscalec3_cache_purgeID
73#endif
74#ifdef CPU_XSCALE_81342
75#define cpu_l2cache_wbinv_all	xscalec3_l2cache_purge
76#elif defined(SOC_MV_KIRKWOOD) || defined(SOC_MV_DISCOVERY)
77#define cpu_l2cache_wbinv_all	sheeva_l2cache_wbinv_all
78#else
79#define cpu_l2cache_wbinv_all()
80#endif
81
82
83int     arm_picache_size;
84int     arm_picache_line_size;
85int     arm_picache_ways;
86
87int     arm_pdcache_size;       /* and unified */
88int     arm_pdcache_line_size = 32;
89int     arm_pdcache_ways;
90
91int     arm_pcache_type;
92int     arm_pcache_unified;
93
94int     arm_dcache_align;
95int     arm_dcache_align_mask;
96
97/* Additional cache information local to this file.  Log2 of some of the
98      above numbers.  */
99static int      arm_dcache_l2_nsets;
100static int      arm_dcache_l2_assoc;
101static int      arm_dcache_l2_linesize;
102
103
104int block_userspace_access = 0;
105extern int arm9_dcache_sets_inc;
106extern int arm9_dcache_sets_max;
107extern int arm9_dcache_index_max;
108extern int arm9_dcache_index_inc;
109
110static __inline void *
111memcpy(void *dst, const void *src, int len)
112{
113	const char *s = src;
114    	char *d = dst;
115
116	while (len) {
117		if (0 && len >= 4 && !((vm_offset_t)d & 3) &&
118		    !((vm_offset_t)s & 3)) {
119			*(uint32_t *)d = *(uint32_t *)s;
120			s += 4;
121			d += 4;
122			len -= 4;
123		} else {
124			*d++ = *s++;
125			len--;
126		}
127	}
128	return (dst);
129}
130
131static __inline void
132bzero(void *addr, int count)
133{
134	char *tmp = (char *)addr;
135
136	while (count > 0) {
137		if (count >= 4 && !((vm_offset_t)tmp & 3)) {
138			*(uint32_t *)tmp = 0;
139			tmp += 4;
140			count -= 4;
141		} else {
142			*tmp = 0;
143			tmp++;
144			count--;
145		}
146	}
147}
148
149static void arm9_setup(void);
150
151void
152_startC(void)
153{
154	int physaddr = KERNPHYSADDR;
155	int tmp1;
156	unsigned int sp = ((unsigned int)&_end & ~3) + 4;
157#if defined(FLASHADDR) && defined(LOADERRAMADDR)
158	unsigned int pc;
159
160	__asm __volatile("adr %0, _start\n"
161	    : "=r" (pc));
162	if ((FLASHADDR > LOADERRAMADDR && pc >= FLASHADDR) ||
163	    (FLASHADDR < LOADERRAMADDR && pc < LOADERRAMADDR)) {
164		/*
165		 * We're running from flash, so just copy the whole thing
166		 * from flash to memory.
167		 * This is far from optimal, we could do the relocation or
168		 * the unzipping directly from flash to memory to avoid this
169		 * needless copy, but it would require to know the flash
170		 * physical address.
171		 */
172		unsigned int target_addr;
173		unsigned int tmp_sp;
174
175		target_addr = (unsigned int)&_start - PHYSADDR + LOADERRAMADDR;
176		tmp_sp = target_addr + 0x100000 +
177		    (unsigned int)&_end - (unsigned int)&_start;
178		memcpy((char *)target_addr, (char *)pc,
179		    (unsigned int)&_end - (unsigned int)&_start);
180		/* Temporary set the sp and jump to the new location. */
181		__asm __volatile(
182		    "mov sp, %1\n"
183		    "mov pc, %0\n"
184		    : : "r" (target_addr), "r" (tmp_sp));
185
186	}
187#endif
188#ifdef KZIP
189	sp += KERNSIZE + 0x100;
190	sp &= ~(L1_TABLE_SIZE - 1);
191	sp += 2 * L1_TABLE_SIZE;
192#endif
193	sp += 1024 * 1024; /* Should be enough for a stack */
194
195	__asm __volatile("adr %0, 2f\n"
196	    		 "bic %0, %0, #0xff000000\n"
197			 "and %1, %1, #0xff000000\n"
198			 "orr %0, %0, %1\n"
199			 "mrc p15, 0, %1, c1, c0, 0\n"
200			 "bic %1, %1, #1\n" /* Disable MMU */
201			 "orr %1, %1, #(4 | 8)\n" /* Add DC enable,
202						     WBUF enable */
203			 "orr %1, %1, #0x1000\n" /* Add IC enable */
204			 "orr %1, %1, #(0x800)\n" /* BPRD enable */
205
206			 "mcr p15, 0, %1, c1, c0, 0\n"
207			 "nop\n"
208			 "nop\n"
209			 "nop\n"
210			 "mov pc, %0\n"
211			 "2: nop\n"
212			 "mov sp, %2\n"
213			 : "=r" (tmp1), "+r" (physaddr), "+r" (sp));
214#ifndef KZIP
215#ifdef CPU_ARM9
216	/* So that idcache_wbinv works; */
217	if ((cpufunc_id() & 0x0000f000) == 0x00009000)
218		arm9_setup();
219#endif
220	cpu_idcache_wbinv_all();
221	cpu_l2cache_wbinv_all();
222#endif
223	__start();
224}
225
226static void
227get_cachetype_cp15()
228{
229	u_int ctype, isize, dsize;
230	u_int multiplier;
231
232	__asm __volatile("mrc p15, 0, %0, c0, c0, 1"
233	    : "=r" (ctype));
234
235	/*
236	 * ...and thus spake the ARM ARM:
237	 *
238 	 * If an <opcode2> value corresponding to an unimplemented or
239	 * reserved ID register is encountered, the System Control
240	 * processor returns the value of the main ID register.
241	 */
242	if (ctype == cpufunc_id())
243		goto out;
244
245	if ((ctype & CPU_CT_S) == 0)
246		arm_pcache_unified = 1;
247
248	/*
249	 * If you want to know how this code works, go read the ARM ARM.
250	 */
251
252	arm_pcache_type = CPU_CT_CTYPE(ctype);
253        if (arm_pcache_unified == 0) {
254		isize = CPU_CT_ISIZE(ctype);
255	    	multiplier = (isize & CPU_CT_xSIZE_M) ? 3 : 2;
256		arm_picache_line_size = 1U << (CPU_CT_xSIZE_LEN(isize) + 3);
257		if (CPU_CT_xSIZE_ASSOC(isize) == 0) {
258			if (isize & CPU_CT_xSIZE_M)
259				arm_picache_line_size = 0; /* not present */
260			else
261				arm_picache_ways = 1;
262		} else {
263			arm_picache_ways = multiplier <<
264			    (CPU_CT_xSIZE_ASSOC(isize) - 1);
265		}
266		arm_picache_size = multiplier << (CPU_CT_xSIZE_SIZE(isize) + 8);
267	}
268
269	dsize = CPU_CT_DSIZE(ctype);
270	multiplier = (dsize & CPU_CT_xSIZE_M) ? 3 : 2;
271	arm_pdcache_line_size = 1U << (CPU_CT_xSIZE_LEN(dsize) + 3);
272	if (CPU_CT_xSIZE_ASSOC(dsize) == 0) {
273		if (dsize & CPU_CT_xSIZE_M)
274			arm_pdcache_line_size = 0; /* not present */
275		else
276			arm_pdcache_ways = 1;
277	} else {
278		arm_pdcache_ways = multiplier <<
279		    (CPU_CT_xSIZE_ASSOC(dsize) - 1);
280	}
281	arm_pdcache_size = multiplier << (CPU_CT_xSIZE_SIZE(dsize) + 8);
282
283	arm_dcache_align = arm_pdcache_line_size;
284
285	arm_dcache_l2_assoc = CPU_CT_xSIZE_ASSOC(dsize) + multiplier - 2;
286	arm_dcache_l2_linesize = CPU_CT_xSIZE_LEN(dsize) + 3;
287	arm_dcache_l2_nsets = 6 + CPU_CT_xSIZE_SIZE(dsize) -
288	    CPU_CT_xSIZE_ASSOC(dsize) - CPU_CT_xSIZE_LEN(dsize);
289 out:
290	arm_dcache_align_mask = arm_dcache_align - 1;
291}
292
293static void
294arm9_setup(void)
295{
296
297	get_cachetype_cp15();
298	arm9_dcache_sets_inc = 1U << arm_dcache_l2_linesize;
299	arm9_dcache_sets_max = (1U << (arm_dcache_l2_linesize +
300	    arm_dcache_l2_nsets)) - arm9_dcache_sets_inc;
301	arm9_dcache_index_inc = 1U << (32 - arm_dcache_l2_assoc);
302	arm9_dcache_index_max = 0U - arm9_dcache_index_inc;
303}
304
305
306#ifdef KZIP
307static  unsigned char *orig_input, *i_input, *i_output;
308
309
310static u_int memcnt;		/* Memory allocated: blocks */
311static size_t memtot;		/* Memory allocated: bytes */
312/*
313 * Library functions required by inflate().
314 */
315
316#define MEMSIZ 0x8000
317
318/*
319 * Allocate memory block.
320 */
321unsigned char *
322kzipmalloc(int size)
323{
324	void *ptr;
325	static u_char mem[MEMSIZ];
326
327	if (memtot + size > MEMSIZ)
328		return NULL;
329	ptr = mem + memtot;
330	memtot += size;
331	memcnt++;
332	return ptr;
333}
334
335/*
336 * Free allocated memory block.
337 */
338void
339kzipfree(void *ptr)
340{
341	memcnt--;
342	if (!memcnt)
343		memtot = 0;
344}
345
346void
347putstr(char *dummy)
348{
349}
350
351static int
352input(void *dummy)
353{
354	if ((size_t)(i_input - orig_input) >= KERNCOMPSIZE) {
355		return (GZ_EOF);
356	}
357	return *i_input++;
358}
359
360static int
361output(void *dummy, unsigned char *ptr, unsigned long len)
362{
363
364
365	memcpy(i_output, ptr, len);
366	i_output += len;
367	return (0);
368}
369
370static void *
371inflate_kernel(void *kernel, void *startaddr)
372{
373	struct inflate infl;
374	char slide[GZ_WSIZE];
375
376	orig_input = kernel;
377	memcnt = memtot = 0;
378	i_input = (char *)kernel + GZ_HEAD;
379	if (((char *)kernel)[3] & 0x18) {
380		while (*i_input)
381			i_input++;
382		i_input++;
383	}
384	i_output = startaddr;
385	bzero(&infl, sizeof(infl));
386	infl.gz_input = input;
387	infl.gz_output = output;
388	infl.gz_slide = slide;
389	inflate(&infl);
390	return ((char *)(((vm_offset_t)i_output & ~3) + 4));
391}
392
393#endif
394
395void *
396load_kernel(unsigned int kstart, unsigned int curaddr,unsigned int func_end,
397    int d)
398{
399	Elf32_Ehdr *eh;
400	Elf32_Phdr phdr[64] /* XXX */, *php;
401	Elf32_Shdr shdr[64] /* XXX */;
402	int i,j;
403	void *entry_point;
404	int symtabindex = -1;
405	int symstrindex = -1;
406	vm_offset_t lastaddr = 0;
407	Elf_Addr ssym = 0;
408	Elf_Dyn *dp;
409
410	eh = (Elf32_Ehdr *)kstart;
411	ssym = 0;
412	entry_point = (void*)eh->e_entry;
413	memcpy(phdr, (void *)(kstart + eh->e_phoff ),
414	    eh->e_phnum * sizeof(phdr[0]));
415
416	/* Determine lastaddr. */
417	for (i = 0; i < eh->e_phnum; i++) {
418		if (lastaddr < (phdr[i].p_vaddr - KERNVIRTADDR + curaddr
419		    + phdr[i].p_memsz))
420			lastaddr = phdr[i].p_vaddr - KERNVIRTADDR +
421			    curaddr + phdr[i].p_memsz;
422	}
423
424	/* Save the symbol tables, as there're about to be scratched. */
425	memcpy(shdr, (void *)(kstart + eh->e_shoff),
426	    sizeof(*shdr) * eh->e_shnum);
427	if (eh->e_shnum * eh->e_shentsize != 0 &&
428	    eh->e_shoff != 0) {
429		for (i = 0; i < eh->e_shnum; i++) {
430			if (shdr[i].sh_type == SHT_SYMTAB) {
431				for (j = 0; j < eh->e_phnum; j++) {
432					if (phdr[j].p_type == PT_LOAD &&
433					    shdr[i].sh_offset >=
434					    phdr[j].p_offset &&
435					    (shdr[i].sh_offset +
436					     shdr[i].sh_size <=
437					     phdr[j].p_offset +
438					     phdr[j].p_filesz)) {
439						shdr[i].sh_offset = 0;
440						shdr[i].sh_size = 0;
441						j = eh->e_phnum;
442					}
443				}
444				if (shdr[i].sh_offset != 0 &&
445				    shdr[i].sh_size != 0) {
446					symtabindex = i;
447					symstrindex = shdr[i].sh_link;
448				}
449			}
450		}
451		func_end = roundup(func_end, sizeof(long));
452		if (symtabindex >= 0 && symstrindex >= 0) {
453			ssym = lastaddr;
454			if (d) {
455				memcpy((void *)func_end, (void *)(
456				    shdr[symtabindex].sh_offset + kstart),
457				    shdr[symtabindex].sh_size);
458				memcpy((void *)(func_end +
459				    shdr[symtabindex].sh_size),
460				    (void *)(shdr[symstrindex].sh_offset +
461				    kstart), shdr[symstrindex].sh_size);
462			} else {
463				lastaddr += shdr[symtabindex].sh_size;
464				lastaddr = roundup(lastaddr,
465				    sizeof(shdr[symtabindex].sh_size));
466				lastaddr += sizeof(shdr[symstrindex].sh_size);
467				lastaddr += shdr[symstrindex].sh_size;
468				lastaddr = roundup(lastaddr,
469				    sizeof(shdr[symstrindex].sh_size));
470			}
471
472		}
473	}
474	if (!d)
475		return ((void *)lastaddr);
476
477	j = eh->e_phnum;
478	for (i = 0; i < j; i++) {
479		volatile char c;
480
481		if (phdr[i].p_type != PT_LOAD)
482			continue;
483		memcpy((void *)(phdr[i].p_vaddr - KERNVIRTADDR + curaddr),
484		    (void*)(kstart + phdr[i].p_offset), phdr[i].p_filesz);
485		/* Clean space from oversized segments, eg: bss. */
486		if (phdr[i].p_filesz < phdr[i].p_memsz)
487			bzero((void *)(phdr[i].p_vaddr - KERNVIRTADDR +
488			    curaddr + phdr[i].p_filesz), phdr[i].p_memsz -
489			    phdr[i].p_filesz);
490	}
491	/* Now grab the symbol tables. */
492	if (symtabindex >= 0 && symstrindex >= 0) {
493		*(Elf_Size *)lastaddr =
494		    shdr[symtabindex].sh_size;
495		lastaddr += sizeof(shdr[symtabindex].sh_size);
496		memcpy((void*)lastaddr,
497		    (void *)func_end,
498		    shdr[symtabindex].sh_size);
499		lastaddr += shdr[symtabindex].sh_size;
500		lastaddr = roundup(lastaddr,
501		    sizeof(shdr[symtabindex].sh_size));
502		*(Elf_Size *)lastaddr =
503		    shdr[symstrindex].sh_size;
504		lastaddr += sizeof(shdr[symstrindex].sh_size);
505		memcpy((void*)lastaddr,
506		    (void*)(func_end +
507			    shdr[symtabindex].sh_size),
508		    shdr[symstrindex].sh_size);
509		lastaddr += shdr[symstrindex].sh_size;
510		lastaddr = roundup(lastaddr,
511   		    sizeof(shdr[symstrindex].sh_size));
512		*(Elf_Addr *)curaddr = MAGIC_TRAMP_NUMBER;
513		*((Elf_Addr *)curaddr + 1) = ssym - curaddr + KERNVIRTADDR;
514		*((Elf_Addr *)curaddr + 2) = lastaddr - curaddr + KERNVIRTADDR;
515	} else
516		*(Elf_Addr *)curaddr = 0;
517	/* Invalidate the instruction cache. */
518	__asm __volatile("mcr p15, 0, %0, c7, c5, 0\n"
519	    		 "mcr p15, 0, %0, c7, c10, 4\n"
520			 : : "r" (curaddr));
521	__asm __volatile("mrc p15, 0, %0, c1, c0, 0\n"
522	    "bic %0, %0, #1\n" /* MMU_ENABLE */
523	    "mcr p15, 0, %0, c1, c0, 0\n"
524	    : "=r" (ssym));
525	/* Jump to the entry point. */
526	((void(*)(void))(entry_point - KERNVIRTADDR + curaddr))();
527	__asm __volatile(".globl func_end\n"
528	    "func_end:");
529
530}
531
532extern char func_end[];
533
534
535#define PMAP_DOMAIN_KERNEL	0 /*
536				    * Just define it instead of including the
537				    * whole VM headers set.
538				    */
539int __hack;
540static __inline void
541setup_pagetables(unsigned int pt_addr, vm_paddr_t physstart, vm_paddr_t physend,
542    int write_back)
543{
544	unsigned int *pd = (unsigned int *)pt_addr;
545	vm_paddr_t addr;
546	int domain = (DOMAIN_CLIENT << (PMAP_DOMAIN_KERNEL * 2)) | DOMAIN_CLIENT;
547	int tmp;
548
549	bzero(pd, L1_TABLE_SIZE);
550	for (addr = physstart; addr < physend; addr += L1_S_SIZE) {
551		pd[addr >> L1_S_SHIFT] = L1_TYPE_S|L1_S_C|L1_S_AP(AP_KRW)|
552		    L1_S_DOM(PMAP_DOMAIN_KERNEL) | addr;
553		if (write_back && 0)
554			pd[addr >> L1_S_SHIFT] |= L1_S_B;
555	}
556	/* XXX: See below */
557	if (0xfff00000 < physstart || 0xfff00000 > physend)
558		pd[0xfff00000 >> L1_S_SHIFT] = L1_TYPE_S|L1_S_AP(AP_KRW)|
559		    L1_S_DOM(PMAP_DOMAIN_KERNEL)|physstart;
560	__asm __volatile("mcr p15, 0, %1, c2, c0, 0\n" /* set TTB */
561	    		 "mcr p15, 0, %1, c8, c7, 0\n" /* Flush TTB */
562			 "mcr p15, 0, %2, c3, c0, 0\n" /* Set DAR */
563			 "mrc p15, 0, %0, c1, c0, 0\n"
564			 "orr %0, %0, #1\n" /* MMU_ENABLE */
565			 "mcr p15, 0, %0, c1, c0, 0\n"
566			 "mrc p15, 0, %0, c2, c0, 0\n" /* CPWAIT */
567			 "mov r0, r0\n"
568			 "sub pc, pc, #4\n" :
569			 "=r" (tmp) : "r" (pd), "r" (domain));
570
571	/*
572	 * XXX: This is the most stupid workaround I've ever wrote.
573	 * For some reason, the KB9202 won't boot the kernel unless
574	 * we access an address which is not in the
575	 * 0x20000000 - 0x20ffffff range. I hope I'll understand
576	 * what's going on later.
577	 */
578	__hack = *(volatile int *)0xfffff21c;
579}
580
581void
582__start(void)
583{
584	void *curaddr;
585	void *dst, *altdst;
586	char *kernel = (char *)&kernel_start;
587	int sp;
588	int pt_addr;
589
590	__asm __volatile("mov %0, pc"  :
591	    "=r" (curaddr));
592	curaddr = (void*)((unsigned int)curaddr & 0xfff00000);
593#ifdef KZIP
594	if (*kernel == 0x1f && kernel[1] == 0x8b) {
595		pt_addr = (((int)&_end + KERNSIZE + 0x100) &
596		    ~(L1_TABLE_SIZE - 1)) + L1_TABLE_SIZE;
597
598#ifdef CPU_ARM9
599		/* So that idcache_wbinv works; */
600		if ((cpufunc_id() & 0x0000f000) == 0x00009000)
601			arm9_setup();
602#endif
603		setup_pagetables(pt_addr, (vm_paddr_t)curaddr,
604		    (vm_paddr_t)curaddr + 0x10000000, 1);
605		/* Gzipped kernel */
606		dst = inflate_kernel(kernel, &_end);
607		kernel = (char *)&_end;
608		altdst = 4 + load_kernel((unsigned int)kernel,
609		    (unsigned int)curaddr,
610		    (unsigned int)&func_end + 800 , 0);
611		if (altdst > dst)
612			dst = altdst;
613	} else
614#endif
615		dst = 4 + load_kernel((unsigned int)&kernel_start,
616	    (unsigned int)curaddr,
617	    (unsigned int)&func_end, 0);
618	dst = (void *)(((vm_offset_t)dst & ~3));
619	pt_addr = ((unsigned int)dst &~(L1_TABLE_SIZE - 1)) + L1_TABLE_SIZE;
620	setup_pagetables(pt_addr, (vm_paddr_t)curaddr,
621	    (vm_paddr_t)curaddr + 0x10000000, 0);
622	sp = pt_addr + L1_TABLE_SIZE + 8192;
623	sp = sp &~3;
624	dst = (void *)(sp + 4);
625	memcpy((void *)dst, (void *)&load_kernel, (unsigned int)&func_end -
626	    (unsigned int)&load_kernel + 800);
627	do_call(dst, kernel, dst + (unsigned int)(&func_end) -
628	    (unsigned int)(&load_kernel) + 800, sp);
629}
630