2 * kexec.c - kexec_load system call
3 * Copyright (C) 2002-2004 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.
9 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11 #include <linux/capability.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>
21 #include "kexec_internal.h"
23 static int copy_user_segment_list(struct kimage *image,
24 unsigned long nr_segments,
25 struct kexec_segment __user *segments)
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);
40 static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
41 unsigned long nr_segments,
42 struct kexec_segment __user *segments,
47 bool kexec_on_panic = flags & KEXEC_ON_CRASH;
50 /* Verify we have a valid entry point */
51 if ((entry < phys_to_boot_phys(crashk_res.start)) ||
52 (entry > phys_to_boot_phys(crashk_res.end)))
53 return -EADDRNOTAVAIL;
56 /* Allocate and initialize a controlling structure */
57 image = do_kimage_alloc_init();
63 ret = copy_user_segment_list(image, nr_segments, segments);
68 /* Enable special crash kernel control page alloc policy. */
69 image->control_page = crashk_res.start;
70 image->type = KEXEC_TYPE_CRASH;
73 ret = sanity_check_segment_list(image);
78 * Find a location for the control code buffer, and add it
79 * the vector of segments so that it's pages will also be
80 * counted as destination pages.
83 image->control_code_page = kimage_alloc_control_pages(image,
84 get_order(KEXEC_CONTROL_PAGE_SIZE));
85 if (!image->control_code_page) {
86 pr_err("Could not allocate control_code_buffer\n");
90 if (!kexec_on_panic) {
91 image->swap_page = kimage_alloc_control_pages(image, 0);
92 if (!image->swap_page) {
93 pr_err("Could not allocate swap buffer\n");
94 goto out_free_control_pages;
100 out_free_control_pages:
101 kimage_free_page_list(&image->control_pages);
107 static int do_kexec_load(unsigned long entry, unsigned long nr_segments,
108 struct kexec_segment __user *segments, unsigned long flags)
110 struct kimage **dest_image, *image;
114 if (flags & KEXEC_ON_CRASH) {
115 dest_image = &kexec_crash_image;
116 if (kexec_crash_image)
117 arch_kexec_unprotect_crashkres();
119 dest_image = &kexec_image;
122 if (nr_segments == 0) {
123 /* Uninstall image */
124 kimage_free(xchg(dest_image, NULL));
127 if (flags & KEXEC_ON_CRASH) {
129 * Loading another kernel to switch to if this one
130 * crashes. Free any current crash dump kernel before
133 kimage_free(xchg(&kexec_crash_image, NULL));
136 ret = kimage_alloc_init(&image, entry, nr_segments, segments, flags);
140 if (flags & KEXEC_PRESERVE_CONTEXT)
141 image->preserve_context = 1;
143 ret = machine_kexec_prepare(image);
148 * Some architecture(like S390) may touch the crash memory before
149 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
151 ret = kimage_crash_copy_vmcoreinfo(image);
155 for (i = 0; i < nr_segments; i++) {
156 ret = kimage_load_segment(image, &image->segment[i]);
161 kimage_terminate(image);
163 /* Install the new kernel and uninstall the old */
164 image = xchg(dest_image, image);
167 if ((flags & KEXEC_ON_CRASH) && kexec_crash_image)
168 arch_kexec_protect_crashkres();
175 * Exec Kernel system call: for obvious reasons only root may call it.
177 * This call breaks up into three pieces.
178 * - A generic part which loads the new kernel from the current
179 * address space, and very carefully places the data in the
182 * - A generic part that interacts with the kernel and tells all of
183 * the devices to shut down. Preventing on-going dmas, and placing
184 * the devices in a consistent state so a later kernel can
187 * - A machine specific part that includes the syscall number
188 * and then copies the image to it's final destination. And
189 * jumps into the image at entry.
191 * kexec does not sync, or unmount filesystems so if you need
192 * that to happen you need to do that yourself.
195 static inline int kexec_load_check(unsigned long nr_segments,
198 /* We only trust the superuser with rebooting the system. */
199 if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
203 * Verify we have a legal set of flags
204 * This leaves us room for future extensions.
206 if ((flags & KEXEC_FLAGS) != (flags & ~KEXEC_ARCH_MASK))
209 /* Put an artificial cap on the number
210 * of segments passed to kexec_load.
212 if (nr_segments > KEXEC_SEGMENT_MAX)
218 SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
219 struct kexec_segment __user *, segments, unsigned long, flags)
223 result = kexec_load_check(nr_segments, flags);
227 /* Verify we are on the appropriate architecture */
228 if (((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH) &&
229 ((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH_DEFAULT))
232 /* Because we write directly to the reserved memory
233 * region when loading crash kernels we need a mutex here to
234 * prevent multiple crash kernels from attempting to load
235 * simultaneously, and to prevent a crash kernel from loading
236 * over the top of a in use crash kernel.
238 * KISS: always take the mutex.
240 if (!mutex_trylock(&kexec_mutex))
243 result = do_kexec_load(entry, nr_segments, segments, flags);
245 mutex_unlock(&kexec_mutex);
251 COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
252 compat_ulong_t, nr_segments,
253 struct compat_kexec_segment __user *, segments,
254 compat_ulong_t, flags)
256 struct compat_kexec_segment in;
257 struct kexec_segment out, __user *ksegments;
258 unsigned long i, result;
260 result = kexec_load_check(nr_segments, flags);
264 /* Don't allow clients that don't understand the native
265 * architecture to do anything.
267 if ((flags & KEXEC_ARCH_MASK) == KEXEC_ARCH_DEFAULT)
270 ksegments = compat_alloc_user_space(nr_segments * sizeof(out));
271 for (i = 0; i < nr_segments; i++) {
272 result = copy_from_user(&in, &segments[i], sizeof(in));
276 out.buf = compat_ptr(in.buf);
277 out.bufsz = in.bufsz;
279 out.memsz = in.memsz;
281 result = copy_to_user(&ksegments[i], &out, sizeof(out));
286 /* Because we write directly to the reserved memory
287 * region when loading crash kernels we need a mutex here to
288 * prevent multiple crash kernels from attempting to load
289 * simultaneously, and to prevent a crash kernel from loading
290 * over the top of a in use crash kernel.
292 * KISS: always take the mutex.
294 if (!mutex_trylock(&kexec_mutex))
297 result = do_kexec_load(entry, nr_segments, ksegments, flags);
299 mutex_unlock(&kexec_mutex);