machine_kexec_32.c 6.78 KB
Newer Older
1
/*
Dave Jones's avatar
Dave Jones committed
2
 * handle transition of Linux booting another kernel
3 4 5 6 7 8 9 10 11
 * Copyright (C) 2002-2005 Eric Biederman  <ebiederm@xmission.com>
 *
 * This source code is licensed under the GNU General Public License,
 * Version 2.  See the file COPYING for more details.
 */

#include <linux/mm.h>
#include <linux/kexec.h>
#include <linux/delay.h>
Ken'ichi Ohmichi's avatar
Ken'ichi Ohmichi committed
12
#include <linux/numa.h>
Ingo Molnar's avatar
Ingo Molnar committed
13
#include <linux/ftrace.h>
Huang Ying's avatar
Huang Ying committed
14
#include <linux/suspend.h>
15
#include <linux/gfp.h>
16
#include <linux/io.h>
Ingo Molnar's avatar
Ingo Molnar committed
17

18 19 20 21 22
#include <asm/pgtable.h>
#include <asm/pgalloc.h>
#include <asm/tlbflush.h>
#include <asm/mmu_context.h>
#include <asm/apic.h>
23
#include <asm/io_apic.h>
24
#include <asm/cpufeature.h>
25
#include <asm/desc.h>
Laura Abbott's avatar
Laura Abbott committed
26
#include <asm/set_memory.h>
27
#include <asm/debugreg.h>
28 29 30

static void set_gdt(void *newgdt, __u16 limit)
{
31
	struct desc_ptr curgdt;
32 33

	/* ia32 supports unaligned loads & stores */
34 35
	curgdt.size    = limit;
	curgdt.address = (unsigned long)newgdt;
36

37
	load_gdt(&curgdt);
38
}
39 40 41 42 43 44 45 46 47

static void load_segments(void)
{
#define __STR(X) #X
#define STR(X) __STR(X)

	__asm__ __volatile__ (
		"\tljmp $"STR(__KERNEL_CS)",$1f\n"
		"\t1:\n"
Michael Matz's avatar
Michael Matz committed
48 49 50 51
		"\tmovl $"STR(__KERNEL_DS)",%%eax\n"
		"\tmovl %%eax,%%ds\n"
		"\tmovl %%eax,%%es\n"
		"\tmovl %%eax,%%ss\n"
52
		: : : "eax", "memory");
53 54 55 56
#undef STR
#undef __STR
}

57 58 59
static void machine_kexec_free_page_tables(struct kimage *image)
{
	free_page((unsigned long)image->arch.pgd);
60
	image->arch.pgd = NULL;
61 62
#ifdef CONFIG_X86_PAE
	free_page((unsigned long)image->arch.pmd0);
63
	image->arch.pmd0 = NULL;
64
	free_page((unsigned long)image->arch.pmd1);
65
	image->arch.pmd1 = NULL;
66 67
#endif
	free_page((unsigned long)image->arch.pte0);
68
	image->arch.pte0 = NULL;
69
	free_page((unsigned long)image->arch.pte1);
70
	image->arch.pte1 = NULL;
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
}

static int machine_kexec_alloc_page_tables(struct kimage *image)
{
	image->arch.pgd = (pgd_t *)get_zeroed_page(GFP_KERNEL);
#ifdef CONFIG_X86_PAE
	image->arch.pmd0 = (pmd_t *)get_zeroed_page(GFP_KERNEL);
	image->arch.pmd1 = (pmd_t *)get_zeroed_page(GFP_KERNEL);
#endif
	image->arch.pte0 = (pte_t *)get_zeroed_page(GFP_KERNEL);
	image->arch.pte1 = (pte_t *)get_zeroed_page(GFP_KERNEL);
	if (!image->arch.pgd ||
#ifdef CONFIG_X86_PAE
	    !image->arch.pmd0 || !image->arch.pmd1 ||
#endif
	    !image->arch.pte0 || !image->arch.pte1) {
		return -ENOMEM;
	}
	return 0;
}

92 93 94 95
static void machine_kexec_page_table_set_one(
	pgd_t *pgd, pmd_t *pmd, pte_t *pte,
	unsigned long vaddr, unsigned long paddr)
{
96
	p4d_t *p4d;
97 98 99 100 101 102 103
	pud_t *pud;

	pgd += pgd_index(vaddr);
#ifdef CONFIG_X86_PAE
	if (!(pgd_val(*pgd) & _PAGE_PRESENT))
		set_pgd(pgd, __pgd(__pa(pmd) | _PAGE_PRESENT));
#endif
104 105
	p4d = p4d_offset(pgd, vaddr);
	pud = pud_offset(p4d, vaddr);
106 107 108 109 110 111 112 113 114 115
	pmd = pmd_offset(pud, vaddr);
	if (!(pmd_val(*pmd) & _PAGE_PRESENT))
		set_pmd(pmd, __pmd(__pa(pte) | _PAGE_TABLE));
	pte = pte_offset_kernel(pmd, vaddr);
	set_pte(pte, pfn_pte(paddr >> PAGE_SHIFT, PAGE_KERNEL_EXEC));
}

static void machine_kexec_prepare_page_tables(struct kimage *image)
{
	void *control_page;
116
	pmd_t *pmd = NULL;
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132

	control_page = page_address(image->control_code_page);
#ifdef CONFIG_X86_PAE
	pmd = image->arch.pmd0;
#endif
	machine_kexec_page_table_set_one(
		image->arch.pgd, pmd, image->arch.pte0,
		(unsigned long)control_page, __pa(control_page));
#ifdef CONFIG_X86_PAE
	pmd = image->arch.pmd1;
#endif
	machine_kexec_page_table_set_one(
		image->arch.pgd, pmd, image->arch.pte1,
		__pa(control_page), __pa(control_page));
}

133 134 135
/*
 * A architecture hook called to validate the
 * proposed image and prepare the control pages
136
 * as needed.  The pages for KEXEC_CONTROL_PAGE_SIZE
137 138 139 140 141 142 143
 * have been allocated, but the segments have yet
 * been copied into the kernel.
 *
 * Do what every setup is needed on image and the
 * reboot code buffer to allow us to avoid allocations
 * later.
 *
144 145
 * - Make control page executable.
 * - Allocate page tables
146
 * - Setup page tables
147 148 149
 */
int machine_kexec_prepare(struct kimage *image)
{
150 151
	int error;

152
	set_pages_x(image->control_code_page, 1);
153 154 155 156 157
	error = machine_kexec_alloc_page_tables(image);
	if (error)
		return error;
	machine_kexec_prepare_page_tables(image);
	return 0;
158 159 160 161 162 163 164 165
}

/*
 * Undo anything leftover by machine_kexec_prepare
 * when an image is freed.
 */
void machine_kexec_cleanup(struct kimage *image)
{
166
	set_pages_nx(image->control_code_page, 1);
167
	machine_kexec_free_page_tables(image);
168 169 170 171 172 173
}

/*
 * Do not allocate memory (or fail in any way) in machine_kexec().
 * We are past the point of no return, committed to rebooting now.
 */
Huang Ying's avatar
Huang Ying committed
174
void machine_kexec(struct kimage *image)
175
{
176 177
	unsigned long page_list[PAGES_NR];
	void *control_page;
Huang Ying's avatar
Huang Ying committed
178
	int save_ftrace_enabled;
Huang Ying's avatar
Huang Ying committed
179 180 181 182 183 184
	asmlinkage unsigned long
		(*relocate_kernel_ptr)(unsigned long indirection_page,
				       unsigned long control_page,
				       unsigned long start_address,
				       unsigned int has_pae,
				       unsigned int preserve_context);
185

Huang Ying's avatar
Huang Ying committed
186
#ifdef CONFIG_KEXEC_JUMP
187
	if (image->preserve_context)
Huang Ying's avatar
Huang Ying committed
188 189 190 191
		save_processor_state();
#endif

	save_ftrace_enabled = __ftrace_enabled_save();
Ingo Molnar's avatar
Ingo Molnar committed
192

193 194
	/* Interrupts aren't acceptable while we reboot */
	local_irq_disable();
195
	hw_breakpoint_disable();
196

197 198
	if (image->preserve_context) {
#ifdef CONFIG_X86_IO_APIC
199 200
		/*
		 * We need to put APICs in legacy mode so that we can
201 202 203 204 205 206 207 208 209
		 * get timer interrupts in second kernel. kexec/kdump
		 * paths already have calls to disable_IO_APIC() in
		 * one form or other. kexec jump path also need
		 * one.
		 */
		disable_IO_APIC();
#endif
	}

210
	control_page = page_address(image->control_code_page);
211
	memcpy(control_page, relocate_kernel, KEXEC_CONTROL_CODE_MAX_SIZE);
212

Huang Ying's avatar
Huang Ying committed
213
	relocate_kernel_ptr = control_page;
214
	page_list[PA_CONTROL_PAGE] = __pa(control_page);
Huang Ying's avatar
Huang Ying committed
215
	page_list[VA_CONTROL_PAGE] = (unsigned long)control_page;
216
	page_list[PA_PGD] = __pa(image->arch.pgd);
217 218 219 220

	if (image->type == KEXEC_TYPE_DEFAULT)
		page_list[PA_SWAP_PAGE] = (page_to_pfn(image->swap_page)
						<< PAGE_SHIFT);
221

222 223
	/*
	 * The segment registers are funny things, they have both a
224 225 226 227
	 * visible and an invisible part.  Whenever the visible part is
	 * set to a specific selector, the invisible part is loaded
	 * with from a table in memory.  At no other time is the
	 * descriptor table in memory accessed.
228 229 230 231 232
	 *
	 * I take advantage of this here by force loading the
	 * segments, before I zap the gdt with an invalid value.
	 */
	load_segments();
233 234
	/*
	 * The gdt & idt are now invalid.
235 236
	 * If you want to load them you must set up your own idt & gdt.
	 */
237
	idt_invalidate(phys_to_virt(0));
238
	set_gdt(phys_to_virt(0), 0);
239 240

	/* now call it */
Huang Ying's avatar
Huang Ying committed
241 242
	image->start = relocate_kernel_ptr((unsigned long)image->head,
					   (unsigned long)page_list,
243 244
					   image->start,
					   boot_cpu_has(X86_FEATURE_PAE),
Huang Ying's avatar
Huang Ying committed
245
					   image->preserve_context);
Huang Ying's avatar
Huang Ying committed
246 247

#ifdef CONFIG_KEXEC_JUMP
248
	if (image->preserve_context)
Huang Ying's avatar
Huang Ying committed
249 250 251 252
		restore_processor_state();
#endif

	__ftrace_enabled_restore(save_ftrace_enabled);
253
}
254

Ken'ichi Ohmichi's avatar
Ken'ichi Ohmichi committed
255 256
void arch_crash_save_vmcoreinfo(void)
{
257
#ifdef CONFIG_NUMA
258 259
	VMCOREINFO_SYMBOL(node_data);
	VMCOREINFO_LENGTH(node_data, MAX_NUMNODES);
Ken'ichi Ohmichi's avatar
Ken'ichi Ohmichi committed
260 261
#endif
#ifdef CONFIG_X86_PAE
262
	VMCOREINFO_CONFIG(X86_PAE);
Ken'ichi Ohmichi's avatar
Ken'ichi Ohmichi committed
263 264 265
#endif
}