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/string.h>
12 #include <linux/reboot.h>
13 #include <linux/numa.h>
14 #include <linux/ftrace.h>
16 #include <asm/pgtable.h>
17 #include <asm/tlbflush.h>
18 #include <asm/mmu_context.h>
21 static void init_level2_page(pmd_t
*level2p
, unsigned long addr
)
23 unsigned long end_addr
;
26 end_addr
= addr
+ PUD_SIZE
;
27 while (addr
< end_addr
) {
28 set_pmd(level2p
++, __pmd(addr
| __PAGE_KERNEL_LARGE_EXEC
));
33 static int init_level3_page(struct kimage
*image
, pud_t
*level3p
,
34 unsigned long addr
, unsigned long last_addr
)
36 unsigned long end_addr
;
41 end_addr
= addr
+ PGDIR_SIZE
;
42 while ((addr
< last_addr
) && (addr
< end_addr
)) {
46 page
= kimage_alloc_control_pages(image
, 0);
51 level2p
= (pmd_t
*)page_address(page
);
52 init_level2_page(level2p
, addr
);
53 set_pud(level3p
++, __pud(__pa(level2p
) | _KERNPG_TABLE
));
56 /* clear the unused entries */
57 while (addr
< end_addr
) {
66 static int init_level4_page(struct kimage
*image
, pgd_t
*level4p
,
67 unsigned long addr
, unsigned long last_addr
)
69 unsigned long end_addr
;
74 end_addr
= addr
+ (PTRS_PER_PGD
* PGDIR_SIZE
);
75 while ((addr
< last_addr
) && (addr
< end_addr
)) {
79 page
= kimage_alloc_control_pages(image
, 0);
84 level3p
= (pud_t
*)page_address(page
);
85 result
= init_level3_page(image
, level3p
, addr
, last_addr
);
89 set_pgd(level4p
++, __pgd(__pa(level3p
) | _KERNPG_TABLE
));
92 /* clear the unused entries */
93 while (addr
< end_addr
) {
101 static void free_transition_pgtable(struct kimage
*image
)
103 free_page((unsigned long)image
->arch
.pud
);
104 free_page((unsigned long)image
->arch
.pmd
);
105 free_page((unsigned long)image
->arch
.pte
);
108 static int init_transition_pgtable(struct kimage
*image
, pgd_t
*pgd
)
113 unsigned long vaddr
, paddr
;
114 int result
= -ENOMEM
;
116 vaddr
= (unsigned long)relocate_kernel
;
117 paddr
= __pa(page_address(image
->control_code_page
)+PAGE_SIZE
);
118 pgd
+= pgd_index(vaddr
);
119 if (!pgd_present(*pgd
)) {
120 pud
= (pud_t
*)get_zeroed_page(GFP_KERNEL
);
123 image
->arch
.pud
= pud
;
124 set_pgd(pgd
, __pgd(__pa(pud
) | _KERNPG_TABLE
));
126 pud
= pud_offset(pgd
, vaddr
);
127 if (!pud_present(*pud
)) {
128 pmd
= (pmd_t
*)get_zeroed_page(GFP_KERNEL
);
131 image
->arch
.pmd
= pmd
;
132 set_pud(pud
, __pud(__pa(pmd
) | _KERNPG_TABLE
));
134 pmd
= pmd_offset(pud
, vaddr
);
135 if (!pmd_present(*pmd
)) {
136 pte
= (pte_t
*)get_zeroed_page(GFP_KERNEL
);
139 image
->arch
.pte
= pte
;
140 set_pmd(pmd
, __pmd(__pa(pte
) | _KERNPG_TABLE
));
142 pte
= pte_offset_kernel(pmd
, vaddr
);
143 set_pte(pte
, pfn_pte(paddr
>> PAGE_SHIFT
, PAGE_KERNEL_EXEC
));
146 free_transition_pgtable(image
);
151 static int init_pgtable(struct kimage
*image
, unsigned long start_pgtable
)
155 level4p
= (pgd_t
*)__va(start_pgtable
);
156 result
= init_level4_page(image
, level4p
, 0, max_pfn
<< PAGE_SHIFT
);
159 return init_transition_pgtable(image
, level4p
);
162 static void set_idt(void *newidt
, u16 limit
)
164 struct desc_ptr curidt
;
166 /* x86-64 supports unaliged loads & stores */
168 curidt
.address
= (unsigned long)newidt
;
170 __asm__
__volatile__ (
177 static void set_gdt(void *newgdt
, u16 limit
)
179 struct desc_ptr curgdt
;
181 /* x86-64 supports unaligned loads & stores */
183 curgdt
.address
= (unsigned long)newgdt
;
185 __asm__
__volatile__ (
191 static void load_segments(void)
193 __asm__
__volatile__ (
199 : : "a" (__KERNEL_DS
) : "memory"
203 int machine_kexec_prepare(struct kimage
*image
)
205 unsigned long start_pgtable
;
208 /* Calculate the offsets */
209 start_pgtable
= page_to_pfn(image
->control_code_page
) << PAGE_SHIFT
;
211 /* Setup the identity mapped 64bit page table */
212 result
= init_pgtable(image
, start_pgtable
);
219 void machine_kexec_cleanup(struct kimage
*image
)
221 free_transition_pgtable(image
);
225 * Do not allocate memory (or fail in any way) in machine_kexec().
226 * We are past the point of no return, committed to rebooting now.
228 void machine_kexec(struct kimage
*image
)
230 unsigned long page_list
[PAGES_NR
];
235 /* Interrupts aren't acceptable while we reboot */
238 control_page
= page_address(image
->control_code_page
) + PAGE_SIZE
;
239 memcpy(control_page
, relocate_kernel
, PAGE_SIZE
);
241 page_list
[PA_CONTROL_PAGE
] = virt_to_phys(control_page
);
242 page_list
[PA_TABLE_PAGE
] =
243 (unsigned long)__pa(page_address(image
->control_code_page
));
245 /* The segment registers are funny things, they have both a
246 * visible and an invisible part. Whenever the visible part is
247 * set to a specific selector, the invisible part is loaded
248 * with from a table in memory. At no other time is the
249 * descriptor table in memory accessed.
251 * I take advantage of this here by force loading the
252 * segments, before I zap the gdt with an invalid value.
255 /* The gdt & idt are now invalid.
256 * If you want to load them you must set up your own idt & gdt.
258 set_gdt(phys_to_virt(0),0);
259 set_idt(phys_to_virt(0),0);
262 relocate_kernel((unsigned long)image
->head
, (unsigned long)page_list
,
266 void arch_crash_save_vmcoreinfo(void)
268 VMCOREINFO_SYMBOL(phys_base
);
269 VMCOREINFO_SYMBOL(init_level4_pgt
);
272 VMCOREINFO_SYMBOL(node_data
);
273 VMCOREINFO_LENGTH(node_data
, MAX_NUMNODES
);