arch/i386/kernel/machine_kexec.c

   1 /*
   2  * machine_kexec.c - 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 <asm/pgtable.h>
  14 #include <asm/pgalloc.h>
  15 #include <asm/tlbflush.h>
  16 #include <asm/mmu_context.h>
  17 #include <asm/io.h>
  18 #include <asm/apic.h>
  19 #include <asm/cpufeature.h>
  20 #include <asm/desc.h>
  21 #include <asm/system.h>
  22
  23 #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
  24 static u32 kexec_pgd[1024] PAGE_ALIGNED;
  25 #ifdef CONFIG_X86_PAE
  26 static u32 kexec_pmd0[1024] PAGE_ALIGNED;
  27 static u32 kexec_pmd1[1024] PAGE_ALIGNED;
  28 #endif
  29 static u32 kexec_pte0[1024] PAGE_ALIGNED;
  30 static u32 kexec_pte1[1024] PAGE_ALIGNED;
  31
  32 static void set_idt(void *newidt, __u16 limit)
  33 {
  34         struct Xgt_desc_struct curidt;
  35
  36         /* ia32 supports unaliged loads & stores */
  37         curidt.size    = limit;
  38         curidt.address = (unsigned long)newidt;
  39
  40         load_idt(&curidt);
  41 };
  42
  43
  44 static void set_gdt(void *newgdt, __u16 limit)
  45 {
  46         struct Xgt_desc_struct curgdt;
  47
  48         /* ia32 supports unaligned loads & stores */
  49         curgdt.size    = limit;
  50         curgdt.address = (unsigned long)newgdt;
  51
  52         load_gdt(&curgdt);
  53 };
  54
  55 static void load_segments(void)
  56 {
  57 #define __STR(X) #X
  58 #define STR(X) __STR(X)
  59
  60         __asm__ __volatile__ (
  61                 "\tljmp $"STR(__KERNEL_CS)",$1f\n"
  62                 "\t1:\n"
  63                 "\tmovl $"STR(__KERNEL_DS)",%%eax\n"
  64                 "\tmovl %%eax,%%ds\n"
  65                 "\tmovl %%eax,%%es\n"
  66                 "\tmovl %%eax,%%fs\n"
  67                 "\tmovl %%eax,%%gs\n"
  68                 "\tmovl %%eax,%%ss\n"
  69                 ::: "eax", "memory");
  70 #undef STR
  71 #undef __STR
  72 }
  73
  74 /*
  75  * A architecture hook called to validate the
  76  * proposed image and prepare the control pages
  77  * as needed.  The pages for KEXEC_CONTROL_CODE_SIZE
  78  * have been allocated, but the segments have yet
  79  * been copied into the kernel.
  80  *
  81  * Do what every setup is needed on image and the
  82  * reboot code buffer to allow us to avoid allocations
  83  * later.
  84  *
  85  * Currently nothing.
  86  */
  87 int machine_kexec_prepare(struct kimage *image)
  88 {
  89         return 0;
  90 }
  91
  92 /*
  93  * Undo anything leftover by machine_kexec_prepare
  94  * when an image is freed.
  95  */
  96 void machine_kexec_cleanup(struct kimage *image)
  97 {
  98 }
  99
 100 /*
 101  * Do not allocate memory (or fail in any way) in machine_kexec().
 102  * We are past the point of no return, committed to rebooting now.
 103  */
 104 NORET_TYPE void machine_kexec(struct kimage *image)
 105 {
 106         unsigned long page_list[PAGES_NR];
 107         void *control_page;
 108
 109         /* Interrupts aren't acceptable while we reboot */
 110         local_irq_disable();
 111
 112         control_page = page_address(image->control_code_page);
 113         memcpy(control_page, relocate_kernel, PAGE_SIZE);
 114
 115         page_list[PA_CONTROL_PAGE] = __pa(control_page);
 116         page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
 117         page_list[PA_PGD] = __pa(kexec_pgd);
 118         page_list[VA_PGD] = (unsigned long)kexec_pgd;
 119 #ifdef CONFIG_X86_PAE
 120         page_list[PA_PMD_0] = __pa(kexec_pmd0);
 121         page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
 122         page_list[PA_PMD_1] = __pa(kexec_pmd1);
 123         page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
 124 #endif
 125         page_list[PA_PTE_0] = __pa(kexec_pte0);
 126         page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
 127         page_list[PA_PTE_1] = __pa(kexec_pte1);
 128         page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
 129
 130         /* The segment registers are funny things, they have both a
 131          * visible and an invisible part.  Whenever the visible part is
 132          * set to a specific selector, the invisible part is loaded
 133          * with from a table in memory.  At no other time is the
 134          * descriptor table in memory accessed.
 135          *
 136          * I take advantage of this here by force loading the
 137          * segments, before I zap the gdt with an invalid value.
 138          */
 139         load_segments();
 140         /* The gdt & idt are now invalid.
 141          * If you want to load them you must set up your own idt & gdt.
 142          */
 143         set_gdt(phys_to_virt(0),0);
 144         set_idt(phys_to_virt(0),0);
 145
 146         /* now call it */
 147         relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
 148                         image->start, cpu_has_pae);
 149 }
 150
 151 /* crashkernel=size@addr specifies the location to reserve for
 152  * a crash kernel.  By reserving this memory we guarantee
 153  * that linux never sets it up as a DMA target.
 154  * Useful for holding code to do something appropriate
 155  * after a kernel panic.
 156  */
 157 static int __init parse_crashkernel(char *arg)
 158 {
 159         unsigned long size, base;
 160         size = memparse(arg, &arg);
 161         if (*arg == '@') {
 162                 base = memparse(arg+1, &arg);
 163                 /* FIXME: Do I want a sanity check
 164                  * to validate the memory range?
 165                  */
 166                 crashk_res.start = base;
 167                 crashk_res.end   = base + size - 1;
 168         }
 169         return 0;
 170 }
 171 early_param("crashkernel", parse_crashkernel);