2 * handle transition of Linux booting another kernel
3 * Copyright (C) 2002-2005 Eric Biederman <ebiederm@xmission.com>
5 * This source code is licensed under the GNU General Public License,
6 * Version 2. See the file COPYING for more details.
10 #include <linux/kexec.h>
11 #include <linux/delay.h>
12 #include <linux/numa.h>
13 #include <linux/ftrace.h>
14 #include <linux/suspend.h>
15 #include <linux/gfp.h>
18 #include <asm/pgtable.h>
19 #include <asm/pgalloc.h>
20 #include <asm/tlbflush.h>
21 #include <asm/mmu_context.h>
23 #include <asm/cpufeature.h>
25 #include <asm/cacheflush.h>
26 #include <asm/debugreg.h>
28 static void set_idt(void *newidt
, __u16 limit
)
30 struct desc_ptr curidt
;
32 /* ia32 supports unaliged loads & stores */
34 curidt
.address
= (unsigned long)newidt
;
40 static void set_gdt(void *newgdt
, __u16 limit
)
42 struct desc_ptr curgdt
;
44 /* ia32 supports unaligned loads & stores */
46 curgdt
.address
= (unsigned long)newgdt
;
51 static void load_segments(void)
54 #define STR(X) __STR(X)
56 __asm__
__volatile__ (
57 "\tljmp $"STR(__KERNEL_CS
)",$1f\n"
59 "\tmovl $"STR(__KERNEL_DS
)",%%eax\n"
65 : : : "eax", "memory");
70 static void machine_kexec_free_page_tables(struct kimage
*image
)
72 free_page((unsigned long)image
->arch
.pgd
);
74 free_page((unsigned long)image
->arch
.pmd0
);
75 free_page((unsigned long)image
->arch
.pmd1
);
77 free_page((unsigned long)image
->arch
.pte0
);
78 free_page((unsigned long)image
->arch
.pte1
);
81 static int machine_kexec_alloc_page_tables(struct kimage
*image
)
83 image
->arch
.pgd
= (pgd_t
*)get_zeroed_page(GFP_KERNEL
);
85 image
->arch
.pmd0
= (pmd_t
*)get_zeroed_page(GFP_KERNEL
);
86 image
->arch
.pmd1
= (pmd_t
*)get_zeroed_page(GFP_KERNEL
);
88 image
->arch
.pte0
= (pte_t
*)get_zeroed_page(GFP_KERNEL
);
89 image
->arch
.pte1
= (pte_t
*)get_zeroed_page(GFP_KERNEL
);
90 if (!image
->arch
.pgd
||
92 !image
->arch
.pmd0
|| !image
->arch
.pmd1
||
94 !image
->arch
.pte0
|| !image
->arch
.pte1
) {
95 machine_kexec_free_page_tables(image
);
101 static void machine_kexec_page_table_set_one(
102 pgd_t
*pgd
, pmd_t
*pmd
, pte_t
*pte
,
103 unsigned long vaddr
, unsigned long paddr
)
107 pgd
+= pgd_index(vaddr
);
108 #ifdef CONFIG_X86_PAE
109 if (!(pgd_val(*pgd
) & _PAGE_PRESENT
))
110 set_pgd(pgd
, __pgd(__pa(pmd
) | _PAGE_PRESENT
));
112 pud
= pud_offset(pgd
, vaddr
);
113 pmd
= pmd_offset(pud
, vaddr
);
114 if (!(pmd_val(*pmd
) & _PAGE_PRESENT
))
115 set_pmd(pmd
, __pmd(__pa(pte
) | _PAGE_TABLE
));
116 pte
= pte_offset_kernel(pmd
, vaddr
);
117 set_pte(pte
, pfn_pte(paddr
>> PAGE_SHIFT
, PAGE_KERNEL_EXEC
));
120 static void machine_kexec_prepare_page_tables(struct kimage
*image
)
125 control_page
= page_address(image
->control_code_page
);
126 #ifdef CONFIG_X86_PAE
127 pmd
= image
->arch
.pmd0
;
129 machine_kexec_page_table_set_one(
130 image
->arch
.pgd
, pmd
, image
->arch
.pte0
,
131 (unsigned long)control_page
, __pa(control_page
));
132 #ifdef CONFIG_X86_PAE
133 pmd
= image
->arch
.pmd1
;
135 machine_kexec_page_table_set_one(
136 image
->arch
.pgd
, pmd
, image
->arch
.pte1
,
137 __pa(control_page
), __pa(control_page
));
141 * A architecture hook called to validate the
142 * proposed image and prepare the control pages
143 * as needed. The pages for KEXEC_CONTROL_PAGE_SIZE
144 * have been allocated, but the segments have yet
145 * been copied into the kernel.
147 * Do what every setup is needed on image and the
148 * reboot code buffer to allow us to avoid allocations
151 * - Make control page executable.
152 * - Allocate page tables
153 * - Setup page tables
155 int machine_kexec_prepare(struct kimage
*image
)
159 set_pages_x(image
->control_code_page
, 1);
160 error
= machine_kexec_alloc_page_tables(image
);
163 machine_kexec_prepare_page_tables(image
);
168 * Undo anything leftover by machine_kexec_prepare
169 * when an image is freed.
171 void machine_kexec_cleanup(struct kimage
*image
)
173 set_pages_nx(image
->control_code_page
, 1);
174 machine_kexec_free_page_tables(image
);
178 * Do not allocate memory (or fail in any way) in machine_kexec().
179 * We are past the point of no return, committed to rebooting now.
181 void machine_kexec(struct kimage
*image
)
183 unsigned long page_list
[PAGES_NR
];
185 int save_ftrace_enabled
;
186 asmlinkage
unsigned long
187 (*relocate_kernel_ptr
)(unsigned long indirection_page
,
188 unsigned long control_page
,
189 unsigned long start_address
,
190 unsigned int has_pae
,
191 unsigned int preserve_context
);
193 #ifdef CONFIG_KEXEC_JUMP
194 if (image
->preserve_context
)
195 save_processor_state();
198 save_ftrace_enabled
= __ftrace_enabled_save();
200 /* Interrupts aren't acceptable while we reboot */
202 hw_breakpoint_disable();
204 if (image
->preserve_context
) {
205 #ifdef CONFIG_X86_IO_APIC
207 * We need to put APICs in legacy mode so that we can
208 * get timer interrupts in second kernel. kexec/kdump
209 * paths already have calls to disable_IO_APIC() in
210 * one form or other. kexec jump path also need
217 control_page
= page_address(image
->control_code_page
);
218 memcpy(control_page
, relocate_kernel
, KEXEC_CONTROL_CODE_MAX_SIZE
);
220 relocate_kernel_ptr
= control_page
;
221 page_list
[PA_CONTROL_PAGE
] = __pa(control_page
);
222 page_list
[VA_CONTROL_PAGE
] = (unsigned long)control_page
;
223 page_list
[PA_PGD
] = __pa(image
->arch
.pgd
);
225 if (image
->type
== KEXEC_TYPE_DEFAULT
)
226 page_list
[PA_SWAP_PAGE
] = (page_to_pfn(image
->swap_page
)
230 * The segment registers are funny things, they have both a
231 * visible and an invisible part. Whenever the visible part is
232 * set to a specific selector, the invisible part is loaded
233 * with from a table in memory. At no other time is the
234 * descriptor table in memory accessed.
236 * I take advantage of this here by force loading the
237 * segments, before I zap the gdt with an invalid value.
241 * The gdt & idt are now invalid.
242 * If you want to load them you must set up your own idt & gdt.
244 set_gdt(phys_to_virt(0), 0);
245 set_idt(phys_to_virt(0), 0);
248 image
->start
= relocate_kernel_ptr((unsigned long)image
->head
,
249 (unsigned long)page_list
,
251 boot_cpu_has(X86_FEATURE_PAE
),
252 image
->preserve_context
);
254 #ifdef CONFIG_KEXEC_JUMP
255 if (image
->preserve_context
)
256 restore_processor_state();
259 __ftrace_enabled_restore(save_ftrace_enabled
);
262 void arch_crash_save_vmcoreinfo(void)
265 VMCOREINFO_SYMBOL(node_data
);
266 VMCOREINFO_LENGTH(node_data
, MAX_NUMNODES
);
268 #ifdef CONFIG_X86_PAE
269 VMCOREINFO_CONFIG(X86_PAE
);