blob: ee70aef5cd81dfb7ee5c4397b91810b0109a58b8 [file] [log] [blame]
Eric W. Biedermandc009d92005-06-25 14:57:52 -07001/*
Dave Young2965faa2015-09-09 15:38:55 -07002 * kexec.c - kexec_load system call
Eric W. Biedermandc009d92005-06-25 14:57:52 -07003 * 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
Minfei Huangde90a6b2015-11-06 16:32:45 -08009#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10
Randy.Dunlapc59ede72006-01-11 12:17:46 -080011#include <linux/capability.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070012#include <linux/mm.h>
13#include <linux/file.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070014#include <linux/kexec.h>
Andrew Morton8c5a1cf2008-08-15 00:40:27 -070015#include <linux/mutex.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070016#include <linux/list.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070017#include <linux/syscalls.h>
Dave Younga43cac02015-09-09 15:38:51 -070018#include <linux/vmalloc.h>
Dave Young2965faa2015-09-09 15:38:55 -070019#include <linux/slab.h>
Alexander Nyberg6e274d12005-06-25 14:58:26 -070020
Dave Younga43cac02015-09-09 15:38:51 -070021#include "kexec_internal.h"
22
Vivek Goyaldabe7862014-08-08 14:25:45 -070023static int copy_user_segment_list(struct kimage *image,
24 unsigned long nr_segments,
25 struct kexec_segment __user *segments)
Eric W. Biedermandc009d92005-06-25 14:57:52 -070026{
Vivek Goyaldabe7862014-08-08 14:25:45 -070027 int ret;
Eric W. Biedermandc009d92005-06-25 14:57:52 -070028 size_t segment_bytes;
Eric W. Biedermandc009d92005-06-25 14:57:52 -070029
30 /* Read in the segments */
31 image->nr_segments = nr_segments;
32 segment_bytes = nr_segments * sizeof(*segments);
Vivek Goyaldabe7862014-08-08 14:25:45 -070033 ret = copy_from_user(image->segment, segments, segment_bytes);
34 if (ret)
35 ret = -EFAULT;
36
37 return ret;
38}
39
Vivek Goyal255aedd2014-08-08 14:25:48 -070040static 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)
Eric W. Biedermandc009d92005-06-25 14:57:52 -070044{
Vivek Goyal255aedd2014-08-08 14:25:48 -070045 int ret;
Eric W. Biedermandc009d92005-06-25 14:57:52 -070046 struct kimage *image;
Vivek Goyal255aedd2014-08-08 14:25:48 -070047 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 }
Eric W. Biedermandc009d92005-06-25 14:57:52 -070054
55 /* Allocate and initialize a controlling structure */
Vivek Goyaldabe7862014-08-08 14:25:45 -070056 image = do_kimage_alloc_init();
57 if (!image)
58 return -ENOMEM;
59
60 image->start = entry;
61
Vivek Goyal255aedd2014-08-08 14:25:48 -070062 ret = copy_user_segment_list(image, nr_segments, segments);
63 if (ret)
Vivek Goyaldabe7862014-08-08 14:25:45 -070064 goto out_free_image;
65
Vivek Goyal255aedd2014-08-08 14:25:48 -070066 if (kexec_on_panic) {
Xunlei Pangcdf4b3f2016-01-20 15:00:31 -080067 /* Enable special crash kernel control page alloc policy. */
Vivek Goyal255aedd2014-08-08 14:25:48 -070068 image->control_page = crashk_res.start;
69 image->type = KEXEC_TYPE_CRASH;
70 }
71
Xunlei Pangcdf4b3f2016-01-20 15:00:31 -080072 ret = sanity_check_segment_list(image);
73 if (ret)
74 goto out_free_image;
75
Eric W. Biedermandc009d92005-06-25 14:57:52 -070076 /*
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 */
Vivek Goyal255aedd2014-08-08 14:25:48 -070081 ret = -ENOMEM;
Eric W. Biedermandc009d92005-06-25 14:57:52 -070082 image->control_code_page = kimage_alloc_control_pages(image,
Huang Ying163f6872008-08-15 00:40:22 -070083 get_order(KEXEC_CONTROL_PAGE_SIZE));
Eric W. Biedermandc009d92005-06-25 14:57:52 -070084 if (!image->control_code_page) {
Fabian Fredericke1bebcf2014-06-06 14:37:09 -070085 pr_err("Could not allocate control_code_buffer\n");
Vivek Goyaldabe7862014-08-08 14:25:45 -070086 goto out_free_image;
Eric W. Biedermandc009d92005-06-25 14:57:52 -070087 }
88
Vivek Goyal255aedd2014-08-08 14:25:48 -070089 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 }
Huang Ying3ab83522008-07-25 19:45:07 -070095 }
96
Zhang Yanfeib92e7e02013-02-27 17:03:29 -080097 *rimage = image;
98 return 0;
Vivek Goyaldabe7862014-08-08 14:25:45 -070099out_free_control_pages:
Zhang Yanfeib92e7e02013-02-27 17:03:29 -0800100 kimage_free_page_list(&image->control_pages);
Vivek Goyaldabe7862014-08-08 14:25:45 -0700101out_free_image:
Zhang Yanfeib92e7e02013-02-27 17:03:29 -0800102 kfree(image);
Vivek Goyal255aedd2014-08-08 14:25:48 -0700103 return ret;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700104}
105
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700106/*
107 * Exec Kernel system call: for obvious reasons only root may call it.
108 *
109 * This call breaks up into three pieces.
110 * - A generic part which loads the new kernel from the current
111 * address space, and very carefully places the data in the
112 * allocated pages.
113 *
114 * - A generic part that interacts with the kernel and tells all of
115 * the devices to shut down. Preventing on-going dmas, and placing
116 * the devices in a consistent state so a later kernel can
117 * reinitialize them.
118 *
119 * - A machine specific part that includes the syscall number
Geert Uytterhoeven002ace72013-09-15 11:35:37 +0200120 * and then copies the image to it's final destination. And
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700121 * jumps into the image at entry.
122 *
123 * kexec does not sync, or unmount filesystems so if you need
124 * that to happen you need to do that yourself.
125 */
Andrew Morton8c5a1cf2008-08-15 00:40:27 -0700126
Heiko Carstens754fe8d2009-01-14 14:14:09 +0100127SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
128 struct kexec_segment __user *, segments, unsigned long, flags)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700129{
130 struct kimage **dest_image, *image;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700131 int result;
132
133 /* We only trust the superuser with rebooting the system. */
Kees Cook79847542014-01-23 15:55:59 -0800134 if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700135 return -EPERM;
136
137 /*
138 * Verify we have a legal set of flags
139 * This leaves us room for future extensions.
140 */
141 if ((flags & KEXEC_FLAGS) != (flags & ~KEXEC_ARCH_MASK))
142 return -EINVAL;
143
144 /* Verify we are on the appropriate architecture */
145 if (((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH) &&
146 ((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH_DEFAULT))
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700147 return -EINVAL;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700148
149 /* Put an artificial cap on the number
150 * of segments passed to kexec_load.
151 */
152 if (nr_segments > KEXEC_SEGMENT_MAX)
153 return -EINVAL;
154
155 image = NULL;
156 result = 0;
157
158 /* Because we write directly to the reserved memory
159 * region when loading crash kernels we need a mutex here to
160 * prevent multiple crash kernels from attempting to load
161 * simultaneously, and to prevent a crash kernel from loading
162 * over the top of a in use crash kernel.
163 *
164 * KISS: always take the mutex.
165 */
Andrew Morton8c5a1cf2008-08-15 00:40:27 -0700166 if (!mutex_trylock(&kexec_mutex))
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700167 return -EBUSY;
Maneesh Soni72414d32005-06-25 14:58:28 -0700168
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700169 dest_image = &kexec_image;
Maneesh Soni72414d32005-06-25 14:58:28 -0700170 if (flags & KEXEC_ON_CRASH)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700171 dest_image = &kexec_crash_image;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700172 if (nr_segments > 0) {
173 unsigned long i;
Maneesh Soni72414d32005-06-25 14:58:28 -0700174
Geoff Levand518a0c72015-02-17 13:45:53 -0800175 if (flags & KEXEC_ON_CRASH) {
176 /*
177 * Loading another kernel to switch to if this one
178 * crashes. Free any current crash dump kernel before
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700179 * we corrupt it.
180 */
Geoff Levand518a0c72015-02-17 13:45:53 -0800181
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700182 kimage_free(xchg(&kexec_crash_image, NULL));
Vivek Goyal255aedd2014-08-08 14:25:48 -0700183 result = kimage_alloc_init(&image, entry, nr_segments,
184 segments, flags);
Michael Holzheu558df722011-10-30 15:16:43 +0100185 crash_map_reserved_pages();
Geoff Levand518a0c72015-02-17 13:45:53 -0800186 } else {
187 /* Loading another kernel to reboot into. */
188
189 result = kimage_alloc_init(&image, entry, nr_segments,
190 segments, flags);
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700191 }
Maneesh Soni72414d32005-06-25 14:58:28 -0700192 if (result)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700193 goto out;
Maneesh Soni72414d32005-06-25 14:58:28 -0700194
Huang Ying3ab83522008-07-25 19:45:07 -0700195 if (flags & KEXEC_PRESERVE_CONTEXT)
196 image->preserve_context = 1;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700197 result = machine_kexec_prepare(image);
Maneesh Soni72414d32005-06-25 14:58:28 -0700198 if (result)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700199 goto out;
Maneesh Soni72414d32005-06-25 14:58:28 -0700200
201 for (i = 0; i < nr_segments; i++) {
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700202 result = kimage_load_segment(image, &image->segment[i]);
Maneesh Soni72414d32005-06-25 14:58:28 -0700203 if (result)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700204 goto out;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700205 }
WANG Cong7fccf032008-07-25 19:45:02 -0700206 kimage_terminate(image);
Michael Holzheu558df722011-10-30 15:16:43 +0100207 if (flags & KEXEC_ON_CRASH)
208 crash_unmap_reserved_pages();
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700209 }
210 /* Install the new kernel, and Uninstall the old */
211 image = xchg(dest_image, image);
212
Maneesh Soni72414d32005-06-25 14:58:28 -0700213out:
Andrew Morton8c5a1cf2008-08-15 00:40:27 -0700214 mutex_unlock(&kexec_mutex);
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700215 kimage_free(image);
Maneesh Soni72414d32005-06-25 14:58:28 -0700216
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700217 return result;
218}
219
220#ifdef CONFIG_COMPAT
Heiko Carstensca2c4052014-03-04 17:13:42 +0100221COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
222 compat_ulong_t, nr_segments,
223 struct compat_kexec_segment __user *, segments,
224 compat_ulong_t, flags)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700225{
226 struct compat_kexec_segment in;
227 struct kexec_segment out, __user *ksegments;
228 unsigned long i, result;
229
230 /* Don't allow clients that don't understand the native
231 * architecture to do anything.
232 */
Maneesh Soni72414d32005-06-25 14:58:28 -0700233 if ((flags & KEXEC_ARCH_MASK) == KEXEC_ARCH_DEFAULT)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700234 return -EINVAL;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700235
Maneesh Soni72414d32005-06-25 14:58:28 -0700236 if (nr_segments > KEXEC_SEGMENT_MAX)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700237 return -EINVAL;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700238
239 ksegments = compat_alloc_user_space(nr_segments * sizeof(out));
Fabian Fredericke1bebcf2014-06-06 14:37:09 -0700240 for (i = 0; i < nr_segments; i++) {
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700241 result = copy_from_user(&in, &segments[i], sizeof(in));
Maneesh Soni72414d32005-06-25 14:58:28 -0700242 if (result)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700243 return -EFAULT;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700244
245 out.buf = compat_ptr(in.buf);
246 out.bufsz = in.bufsz;
247 out.mem = in.mem;
248 out.memsz = in.memsz;
249
250 result = copy_to_user(&ksegments[i], &out, sizeof(out));
Maneesh Soni72414d32005-06-25 14:58:28 -0700251 if (result)
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700252 return -EFAULT;
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700253 }
254
255 return sys_kexec_load(entry, nr_segments, ksegments, flags);
256}
257#endif