arch/x86/kernel/machine_kexec_32.c

   1 /*
   2  * handle transition of Linux booting another kernel
   3  * Copyright (C) 2002-2005 Eric Biederman  <ebiederm@xmission.com>
   4  *
   5  * This source code is licensed under the GNU General Public License,
   6  * Version 2.  See the file COPYING for more details.
   7  */
   8
   9 #include <linux/mm.h>
  10 #include <linux/kexec.h>
  11 #include <linux/delay.h>
  12 #include <linux/init.h>
  13 #include <linux/numa.h>
  14 #include <linux/ftrace.h>
  15 #include <linux/suspend.h>
  16
  17 #include <asm/pgtable.h>
  18 #include <asm/pgalloc.h>
  19 #include <asm/tlbflush.h>
  20 #include <asm/mmu_context.h>
  21 #include <asm/io.h>
  22 #include <asm/apic.h>
  23 #include <asm/cpufeature.h>
  24 #include <asm/desc.h>
  25 #include <asm/system.h>
  26 #include <asm/cacheflush.h>
  27
  28 #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
  29 static u32 kexec_pgd[1024] PAGE_ALIGNED;
  30 #ifdef CONFIG_X86_PAE
  31 static u32 kexec_pmd0[1024] PAGE_ALIGNED;
  32 static u32 kexec_pmd1[1024] PAGE_ALIGNED;
  33 #endif
  34 static u32 kexec_pte0[1024] PAGE_ALIGNED;
  35 static u32 kexec_pte1[1024] PAGE_ALIGNED;
  36
  37 static void set_idt(void *newidt, __u16 limit)
  38 {
  39         struct desc_ptr curidt;
  40
  41         /* ia32 supports unaliged loads & stores */
  42         curidt.size    = limit;
  43         curidt.address = (unsigned long)newidt;
  44
  45         load_idt(&curidt);
  46 }
  47
  48
  49 static void set_gdt(void *newgdt, __u16 limit)
  50 {
  51         struct desc_ptr curgdt;
  52
  53         /* ia32 supports unaligned loads & stores */
  54         curgdt.size    = limit;
  55         curgdt.address = (unsigned long)newgdt;
  56
  57         load_gdt(&curgdt);
  58 }
  59
  60 static void load_segments(void)
  61 {
  62 #define __STR(X) #X
  63 #define STR(X) __STR(X)
  64
  65         __asm__ __volatile__ (
  66                 "\tljmp $"STR(__KERNEL_CS)",$1f\n"
  67                 "\t1:\n"
  68                 "\tmovl $"STR(__KERNEL_DS)",%%eax\n"
  69                 "\tmovl %%eax,%%ds\n"
  70                 "\tmovl %%eax,%%es\n"
  71                 "\tmovl %%eax,%%fs\n"
  72                 "\tmovl %%eax,%%gs\n"
  73                 "\tmovl %%eax,%%ss\n"
  74                 ::: "eax", "memory");
  75 #undef STR
  76 #undef __STR
  77 }
  78
  79 /*
  80  * A architecture hook called to validate the
  81  * proposed image and prepare the control pages
  82  * as needed.  The pages for KEXEC_CONTROL_PAGE_SIZE
  83  * have been allocated, but the segments have yet
  84  * been copied into the kernel.
  85  *
  86  * Do what every setup is needed on image and the
  87  * reboot code buffer to allow us to avoid allocations
  88  * later.
  89  *
  90  * Make control page executable.
  91  */
  92 int machine_kexec_prepare(struct kimage *image)
  93 {
  94         if (nx_enabled)
  95                 set_pages_x(image->control_code_page, 1);
  96         return 0;
  97 }
  98
  99 /*
 100  * Undo anything leftover by machine_kexec_prepare
 101  * when an image is freed.
 102  */
 103 void machine_kexec_cleanup(struct kimage *image)
 104 {
 105         if (nx_enabled)
 106                 set_pages_nx(image->control_code_page, 1);
 107 }
 108
 109 /*
 110  * Do not allocate memory (or fail in any way) in machine_kexec().
 111  * We are past the point of no return, committed to rebooting now.
 112  */
 113 void machine_kexec(struct kimage *image)
 114 {
 115         unsigned long page_list[PAGES_NR];
 116         void *control_page;
 117         int save_ftrace_enabled;
 118         asmlinkage unsigned long
 119                 (*relocate_kernel_ptr)(unsigned long indirection_page,
 120                                        unsigned long control_page,
 121                                        unsigned long start_address,
 122                                        unsigned int has_pae,
 123                                        unsigned int preserve_context);
 124
 125 #ifdef CONFIG_KEXEC_JUMP
 126         if (kexec_image->preserve_context)
 127                 save_processor_state();
 128 #endif
 129
 130         save_ftrace_enabled = __ftrace_enabled_save();
 131
 132         /* Interrupts aren't acceptable while we reboot */
 133         local_irq_disable();
 134
 135         if (image->preserve_context) {
 136 #ifdef CONFIG_X86_IO_APIC
 137                 /* We need to put APICs in legacy mode so that we can
 138                  * get timer interrupts in second kernel. kexec/kdump
 139                  * paths already have calls to disable_IO_APIC() in
 140                  * one form or other. kexec jump path also need
 141                  * one.
 142                  */
 143                 disable_IO_APIC();
 144 #endif
 145         }
 146
 147         control_page = page_address(image->control_code_page);
 148         memcpy(control_page, relocate_kernel, KEXEC_CONTROL_CODE_MAX_SIZE);
 149
 150         relocate_kernel_ptr = control_page;
 151         page_list[PA_CONTROL_PAGE] = __pa(control_page);
 152         page_list[VA_CONTROL_PAGE] = (unsigned long)control_page;
 153         page_list[PA_PGD] = __pa(kexec_pgd);
 154         page_list[VA_PGD] = (unsigned long)kexec_pgd;
 155 #ifdef CONFIG_X86_PAE
 156         page_list[PA_PMD_0] = __pa(kexec_pmd0);
 157         page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
 158         page_list[PA_PMD_1] = __pa(kexec_pmd1);
 159         page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
 160 #endif
 161         page_list[PA_PTE_0] = __pa(kexec_pte0);
 162         page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
 163         page_list[PA_PTE_1] = __pa(kexec_pte1);
 164         page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
 165         page_list[PA_SWAP_PAGE] = (page_to_pfn(image->swap_page) << PAGE_SHIFT);
 166
 167         /* The segment registers are funny things, they have both a
 168          * visible and an invisible part.  Whenever the visible part is
 169          * set to a specific selector, the invisible part is loaded
 170          * with from a table in memory.  At no other time is the
 171          * descriptor table in memory accessed.
 172          *
 173          * I take advantage of this here by force loading the
 174          * segments, before I zap the gdt with an invalid value.
 175          */
 176         load_segments();
 177         /* The gdt & idt are now invalid.
 178          * If you want to load them you must set up your own idt & gdt.
 179          */
 180         set_gdt(phys_to_virt(0),0);
 181         set_idt(phys_to_virt(0),0);
 182
 183         /* now call it */
 184         image->start = relocate_kernel_ptr((unsigned long)image->head,
 185                                            (unsigned long)page_list,
 186                                            image->start, cpu_has_pae,
 187                                            image->preserve_context);
 188
 189 #ifdef CONFIG_KEXEC_JUMP
 190         if (kexec_image->preserve_context)
 191                 restore_processor_state();
 192 #endif
 193
 194         __ftrace_enabled_restore(save_ftrace_enabled);
 195 }
 196
 197 void arch_crash_save_vmcoreinfo(void)
 198 {
 199 #ifdef CONFIG_NUMA
 200         VMCOREINFO_SYMBOL(node_data);
 201         VMCOREINFO_LENGTH(node_data, MAX_NUMNODES);
 202 #endif
 203 #ifdef CONFIG_X86_PAE
 204         VMCOREINFO_CONFIG(X86_PAE);
 205 #endif
 206 }
 207