1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * kexec.c - kexec_load system call
4 * Copyright (C) 2002-2004 Eric Biederman <ebiederm@xmission.com>
5 */
6
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8
9 #include <linux/capability.h>
10 #include <linux/mm.h>
11 #include <linux/file.h>
12 #include <linux/security.h>
13 #include <linux/kexec.h>
14 #include <linux/mutex.h>
15 #include <linux/list.h>
16 #include <linux/syscalls.h>
17 #include <linux/vmalloc.h>
18 #include <linux/slab.h>
19
20 #include "kexec_internal.h"
21
kimage_alloc_init(struct kimage ** rimage,unsigned long entry,unsigned long nr_segments,struct kexec_segment * segments,unsigned long flags)22 static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
23 unsigned long nr_segments,
24 struct kexec_segment *segments,
25 unsigned long flags)
26 {
27 int ret;
28 struct kimage *image;
29 bool kexec_on_panic = flags & KEXEC_ON_CRASH;
30
31 if (kexec_on_panic) {
32 /* Verify we have a valid entry point */
33 if ((entry < phys_to_boot_phys(crashk_res.start)) ||
34 (entry > phys_to_boot_phys(crashk_res.end)))
35 return -EADDRNOTAVAIL;
36 }
37
38 /* Allocate and initialize a controlling structure */
39 image = do_kimage_alloc_init();
40 if (!image)
41 return -ENOMEM;
42
43 image->start = entry;
44 image->nr_segments = nr_segments;
45 memcpy(image->segment, segments, nr_segments * sizeof(*segments));
46
47 if (kexec_on_panic) {
48 /* Enable special crash kernel control page alloc policy. */
49 image->control_page = crashk_res.start;
50 image->type = KEXEC_TYPE_CRASH;
51 }
52
53 ret = sanity_check_segment_list(image);
54 if (ret)
55 goto out_free_image;
56
57 /*
58 * Find a location for the control code buffer, and add it
59 * the vector of segments so that it's pages will also be
60 * counted as destination pages.
61 */
62 ret = -ENOMEM;
63 image->control_code_page = kimage_alloc_control_pages(image,
64 get_order(KEXEC_CONTROL_PAGE_SIZE));
65 if (!image->control_code_page) {
66 pr_err("Could not allocate control_code_buffer\n");
67 goto out_free_image;
68 }
69
70 if (!kexec_on_panic) {
71 image->swap_page = kimage_alloc_control_pages(image, 0);
72 if (!image->swap_page) {
73 pr_err("Could not allocate swap buffer\n");
74 goto out_free_control_pages;
75 }
76 }
77
78 *rimage = image;
79 return 0;
80 out_free_control_pages:
81 kimage_free_page_list(&image->control_pages);
82 out_free_image:
83 kfree(image);
84 return ret;
85 }
86
do_kexec_load(unsigned long entry,unsigned long nr_segments,struct kexec_segment * segments,unsigned long flags)87 static int do_kexec_load(unsigned long entry, unsigned long nr_segments,
88 struct kexec_segment *segments, unsigned long flags)
89 {
90 struct kimage **dest_image, *image;
91 unsigned long i;
92 int ret;
93
94 /*
95 * Because we write directly to the reserved memory region when loading
96 * crash kernels we need a mutex here to prevent multiple crash kernels
97 * from attempting to load simultaneously, and to prevent a crash kernel
98 * from loading over the top of a in use crash kernel.
99 *
100 * KISS: always take the mutex.
101 */
102 if (!mutex_trylock(&kexec_mutex))
103 return -EBUSY;
104
105 if (flags & KEXEC_ON_CRASH) {
106 dest_image = &kexec_crash_image;
107 if (kexec_crash_image)
108 arch_kexec_unprotect_crashkres();
109 } else {
110 dest_image = &kexec_image;
111 }
112
113 if (nr_segments == 0) {
114 /* Uninstall image */
115 kimage_free(xchg(dest_image, NULL));
116 ret = 0;
117 goto out_unlock;
118 }
119 if (flags & KEXEC_ON_CRASH) {
120 /*
121 * Loading another kernel to switch to if this one
122 * crashes. Free any current crash dump kernel before
123 * we corrupt it.
124 */
125 kimage_free(xchg(&kexec_crash_image, NULL));
126 }
127
128 ret = kimage_alloc_init(&image, entry, nr_segments, segments, flags);
129 if (ret)
130 goto out_unlock;
131
132 if (flags & KEXEC_PRESERVE_CONTEXT)
133 image->preserve_context = 1;
134
135 ret = machine_kexec_prepare(image);
136 if (ret)
137 goto out;
138
139 /*
140 * Some architecture(like S390) may touch the crash memory before
141 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
142 */
143 ret = kimage_crash_copy_vmcoreinfo(image);
144 if (ret)
145 goto out;
146
147 for (i = 0; i < nr_segments; i++) {
148 ret = kimage_load_segment(image, &image->segment[i]);
149 if (ret)
150 goto out;
151 }
152
153 kimage_terminate(image);
154
155 ret = machine_kexec_post_load(image);
156 if (ret)
157 goto out;
158
159 /* Install the new kernel and uninstall the old */
160 image = xchg(dest_image, image);
161
162 out:
163 if ((flags & KEXEC_ON_CRASH) && kexec_crash_image)
164 arch_kexec_protect_crashkres();
165
166 kimage_free(image);
167 out_unlock:
168 mutex_unlock(&kexec_mutex);
169 return ret;
170 }
171
172 /*
173 * Exec Kernel system call: for obvious reasons only root may call it.
174 *
175 * This call breaks up into three pieces.
176 * - A generic part which loads the new kernel from the current
177 * address space, and very carefully places the data in the
178 * allocated pages.
179 *
180 * - A generic part that interacts with the kernel and tells all of
181 * the devices to shut down. Preventing on-going dmas, and placing
182 * the devices in a consistent state so a later kernel can
183 * reinitialize them.
184 *
185 * - A machine specific part that includes the syscall number
186 * and then copies the image to it's final destination. And
187 * jumps into the image at entry.
188 *
189 * kexec does not sync, or unmount filesystems so if you need
190 * that to happen you need to do that yourself.
191 */
192
kexec_load_check(unsigned long nr_segments,unsigned long flags)193 static inline int kexec_load_check(unsigned long nr_segments,
194 unsigned long flags)
195 {
196 int result;
197
198 /* We only trust the superuser with rebooting the system. */
199 if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
200 return -EPERM;
201
202 /* Permit LSMs and IMA to fail the kexec */
203 result = security_kernel_load_data(LOADING_KEXEC_IMAGE, false);
204 if (result < 0)
205 return result;
206
207 /*
208 * kexec can be used to circumvent module loading restrictions, so
209 * prevent loading in that case
210 */
211 result = security_locked_down(LOCKDOWN_KEXEC);
212 if (result)
213 return result;
214
215 /*
216 * Verify we have a legal set of flags
217 * This leaves us room for future extensions.
218 */
219 if ((flags & KEXEC_FLAGS) != (flags & ~KEXEC_ARCH_MASK))
220 return -EINVAL;
221
222 /* Put an artificial cap on the number
223 * of segments passed to kexec_load.
224 */
225 if (nr_segments > KEXEC_SEGMENT_MAX)
226 return -EINVAL;
227
228 return 0;
229 }
230
SYSCALL_DEFINE4(kexec_load,unsigned long,entry,unsigned long,nr_segments,struct kexec_segment __user *,segments,unsigned long,flags)231 SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
232 struct kexec_segment __user *, segments, unsigned long, flags)
233 {
234 struct kexec_segment *ksegments;
235 unsigned long result;
236
237 result = kexec_load_check(nr_segments, flags);
238 if (result)
239 return result;
240
241 /* Verify we are on the appropriate architecture */
242 if (((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH) &&
243 ((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH_DEFAULT))
244 return -EINVAL;
245
246 ksegments = memdup_user(segments, nr_segments * sizeof(ksegments[0]));
247 if (IS_ERR(ksegments))
248 return PTR_ERR(ksegments);
249
250 result = do_kexec_load(entry, nr_segments, ksegments, flags);
251 kfree(ksegments);
252
253 return result;
254 }
255
256 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(kexec_load,compat_ulong_t,entry,compat_ulong_t,nr_segments,struct compat_kexec_segment __user *,segments,compat_ulong_t,flags)257 COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
258 compat_ulong_t, nr_segments,
259 struct compat_kexec_segment __user *, segments,
260 compat_ulong_t, flags)
261 {
262 struct compat_kexec_segment in;
263 struct kexec_segment *ksegments;
264 unsigned long i, result;
265
266 result = kexec_load_check(nr_segments, flags);
267 if (result)
268 return result;
269
270 /* Don't allow clients that don't understand the native
271 * architecture to do anything.
272 */
273 if ((flags & KEXEC_ARCH_MASK) == KEXEC_ARCH_DEFAULT)
274 return -EINVAL;
275
276 ksegments = kmalloc_array(nr_segments, sizeof(ksegments[0]),
277 GFP_KERNEL);
278 if (!ksegments)
279 return -ENOMEM;
280
281 for (i = 0; i < nr_segments; i++) {
282 result = copy_from_user(&in, &segments[i], sizeof(in));
283 if (result)
284 goto fail;
285
286 ksegments[i].buf = compat_ptr(in.buf);
287 ksegments[i].bufsz = in.bufsz;
288 ksegments[i].mem = in.mem;
289 ksegments[i].memsz = in.memsz;
290 }
291
292 result = do_kexec_load(entry, nr_segments, ksegments, flags);
293
294 fail:
295 kfree(ksegments);
296 return result;
297 }
298 #endif
299
