kernel/kexec.c

   1 /*
   2  * kexec.c - kexec_load system call
   3  * Copyright (C) 2002-2004 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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  10
  11 #include <linux/capability.h>
  12 #include <linux/mm.h>
  13 #include <linux/file.h>
  14 #include <linux/kexec.h>
  15 #include <linux/mutex.h>
  16 #include <linux/list.h>
  17 #include <linux/syscalls.h>
  18 #include <linux/vmalloc.h>
  19 #include <linux/slab.h>
  20
  21 #include "kexec_internal.h"
  22
  23 static int copy_user_segment_list(struct kimage *image,
  24                                   unsigned long nr_segments,
  25                                   struct kexec_segment __user *segments)
  26 {
  27         int ret;
  28         size_t segment_bytes;
  29
  30         /* Read in the segments */
  31         image->nr_segments = nr_segments;
  32         segment_bytes = nr_segments * sizeof(*segments);
  33         ret = copy_from_user(image->segment, segments, segment_bytes);
  34         if (ret)
  35                 ret = -EFAULT;
  36
  37         return ret;
  38 }
  39
  40 static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
  41                              unsigned long nr_segments,
  42                              struct kexec_segment __user *segments,
  43                              unsigned long flags)
  44 {
  45         int ret;
  46         struct kimage *image;
  47         bool kexec_on_panic = flags & KEXEC_ON_CRASH;
  48
  49         if (kexec_on_panic) {
  50                 /* Verify we have a valid entry point */
  51                 if ((entry < crashk_res.start) || (entry > crashk_res.end))
  52                         return -EADDRNOTAVAIL;
  53         }
  54
  55         /* Allocate and initialize a controlling structure */
  56         image = do_kimage_alloc_init();
  57         if (!image)
  58                 return -ENOMEM;
  59
  60         image->start = entry;
  61
  62         ret = copy_user_segment_list(image, nr_segments, segments);
  63         if (ret)
  64                 goto out_free_image;
  65
  66         if (kexec_on_panic) {
  67                 /* Enable special crash kernel control page alloc policy. */
  68                 image->control_page = crashk_res.start;
  69                 image->type = KEXEC_TYPE_CRASH;
  70         }
  71
  72         ret = sanity_check_segment_list(image);
  73         if (ret)
  74                 goto out_free_image;
  75
  76         /*
  77          * Find a location for the control code buffer, and add it
  78          * the vector of segments so that it's pages will also be
  79          * counted as destination pages.
  80          */
  81         ret = -ENOMEM;
  82         image->control_code_page = kimage_alloc_control_pages(image,
  83                                            get_order(KEXEC_CONTROL_PAGE_SIZE));
  84         if (!image->control_code_page) {
  85                 pr_err("Could not allocate control_code_buffer\n");
  86                 goto out_free_image;
  87         }
  88
  89         if (!kexec_on_panic) {
  90                 image->swap_page = kimage_alloc_control_pages(image, 0);
  91                 if (!image->swap_page) {
  92                         pr_err("Could not allocate swap buffer\n");
  93                         goto out_free_control_pages;
  94                 }
  95         }
  96
  97         *rimage = image;
  98         return 0;
  99 out_free_control_pages:
 100         kimage_free_page_list(&image->control_pages);
 101 out_free_image:
 102         kfree(image);
 103         return ret;
 104 }
 105
 106 static int do_kexec_load(unsigned long entry, unsigned long nr_segments,
 107                 struct kexec_segment __user *segments, unsigned long flags)
 108 {
 109         struct kimage **dest_image, *image;
 110         unsigned long i;
 111         int ret;
 112
 113         if (flags & KEXEC_ON_CRASH) {
 114                 dest_image = &kexec_crash_image;
 115                 if (kexec_crash_image)
 116                         arch_kexec_unprotect_crashkres();
 117         } else {
 118                 dest_image = &kexec_image;
 119         }
 120
 121         if (nr_segments == 0) {
 122                 /* Uninstall image */
 123                 kimage_free(xchg(dest_image, NULL));
 124                 return 0;
 125         }
 126         if (flags & KEXEC_ON_CRASH) {
 127                 /*
 128                  * Loading another kernel to switch to if this one
 129                  * crashes.  Free any current crash dump kernel before
 130                  * we corrupt it.
 131                  */
 132                 kimage_free(xchg(&kexec_crash_image, NULL));
 133         }
 134
 135         ret = kimage_alloc_init(&image, entry, nr_segments, segments, flags);
 136         if (ret)
 137                 return ret;
 138
 139         if (flags & KEXEC_PRESERVE_CONTEXT)
 140                 image->preserve_context = 1;
 141
 142         ret = machine_kexec_prepare(image);
 143         if (ret)
 144                 goto out;
 145
 146         for (i = 0; i < nr_segments; i++) {
 147                 ret = kimage_load_segment(image, &image->segment[i]);
 148                 if (ret)
 149                         goto out;
 150         }
 151
 152         kimage_terminate(image);
 153
 154         /* Install the new kernel and uninstall the old */
 155         image = xchg(dest_image, image);
 156
 157 out:
 158         if ((flags & KEXEC_ON_CRASH) && kexec_crash_image)
 159                 arch_kexec_protect_crashkres();
 160
 161         kimage_free(image);
 162         return ret;
 163 }
 164
 165 /*
 166  * Exec Kernel system call: for obvious reasons only root may call it.
 167  *
 168  * This call breaks up into three pieces.
 169  * - A generic part which loads the new kernel from the current
 170  *   address space, and very carefully places the data in the
 171  *   allocated pages.
 172  *
 173  * - A generic part that interacts with the kernel and tells all of
 174  *   the devices to shut down.  Preventing on-going dmas, and placing
 175  *   the devices in a consistent state so a later kernel can
 176  *   reinitialize them.
 177  *
 178  * - A machine specific part that includes the syscall number
 179  *   and then copies the image to it's final destination.  And
 180  *   jumps into the image at entry.
 181  *
 182  * kexec does not sync, or unmount filesystems so if you need
 183  * that to happen you need to do that yourself.
 184  */
 185
 186 SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
 187                 struct kexec_segment __user *, segments, unsigned long, flags)
 188 {
 189         int result;
 190
 191         /* We only trust the superuser with rebooting the system. */
 192         if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
 193                 return -EPERM;
 194
 195         /*
 196          * Verify we have a legal set of flags
 197          * This leaves us room for future extensions.
 198          */
 199         if ((flags & KEXEC_FLAGS) != (flags & ~KEXEC_ARCH_MASK))
 200                 return -EINVAL;
 201
 202         /* Verify we are on the appropriate architecture */
 203         if (((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH) &&
 204                 ((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH_DEFAULT))
 205                 return -EINVAL;
 206
 207         /* Put an artificial cap on the number
 208          * of segments passed to kexec_load.
 209          */
 210         if (nr_segments > KEXEC_SEGMENT_MAX)
 211                 return -EINVAL;
 212
 213         /* Because we write directly to the reserved memory
 214          * region when loading crash kernels we need a mutex here to
 215          * prevent multiple crash  kernels from attempting to load
 216          * simultaneously, and to prevent a crash kernel from loading
 217          * over the top of a in use crash kernel.
 218          *
 219          * KISS: always take the mutex.
 220          */
 221         if (!mutex_trylock(&kexec_mutex))
 222                 return -EBUSY;
 223
 224         result = do_kexec_load(entry, nr_segments, segments, flags);
 225
 226         mutex_unlock(&kexec_mutex);
 227
 228         return result;
 229 }
 230
 231 #ifdef CONFIG_COMPAT
 232 COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
 233                        compat_ulong_t, nr_segments,
 234                        struct compat_kexec_segment __user *, segments,
 235                        compat_ulong_t, flags)
 236 {
 237         struct compat_kexec_segment in;
 238         struct kexec_segment out, __user *ksegments;
 239         unsigned long i, result;
 240
 241         /* Don't allow clients that don't understand the native
 242          * architecture to do anything.
 243          */
 244         if ((flags & KEXEC_ARCH_MASK) == KEXEC_ARCH_DEFAULT)
 245                 return -EINVAL;
 246
 247         if (nr_segments > KEXEC_SEGMENT_MAX)
 248                 return -EINVAL;
 249
 250         ksegments = compat_alloc_user_space(nr_segments * sizeof(out));
 251         for (i = 0; i < nr_segments; i++) {
 252                 result = copy_from_user(&in, &segments[i], sizeof(in));
 253                 if (result)
 254                         return -EFAULT;
 255
 256                 out.buf   = compat_ptr(in.buf);
 257                 out.bufsz = in.bufsz;
 258                 out.mem   = in.mem;
 259                 out.memsz = in.memsz;
 260
 261                 result = copy_to_user(&ksegments[i], &out, sizeof(out));
 262                 if (result)
 263                         return -EFAULT;
 264         }
 265
 266         return sys_kexec_load(entry, nr_segments, ksegments, flags);
 267 }
 268 #endif