]>
Commit | Line | Data |
---|---|---|
dc009d92 | 1 | /* |
2965faa5 | 2 | * kexec.c - kexec_load system call |
dc009d92 EB |
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 | ||
de90a6bc MH |
9 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
10 | ||
c59ede7b | 11 | #include <linux/capability.h> |
dc009d92 EB |
12 | #include <linux/mm.h> |
13 | #include <linux/file.h> | |
dc009d92 | 14 | #include <linux/kexec.h> |
8c5a1cf0 | 15 | #include <linux/mutex.h> |
dc009d92 | 16 | #include <linux/list.h> |
dc009d92 | 17 | #include <linux/syscalls.h> |
a43cac0d | 18 | #include <linux/vmalloc.h> |
2965faa5 | 19 | #include <linux/slab.h> |
17c953d3 | 20 | #include <linux/module.h> |
dc009d92 | 21 | |
a43cac0d DY |
22 | #include "kexec_internal.h" |
23 | ||
dabe7862 VG |
24 | static int copy_user_segment_list(struct kimage *image, |
25 | unsigned long nr_segments, | |
26 | struct kexec_segment __user *segments) | |
dc009d92 | 27 | { |
dabe7862 | 28 | int ret; |
dc009d92 | 29 | size_t segment_bytes; |
dc009d92 EB |
30 | |
31 | /* Read in the segments */ | |
32 | image->nr_segments = nr_segments; | |
33 | segment_bytes = nr_segments * sizeof(*segments); | |
dabe7862 VG |
34 | ret = copy_from_user(image->segment, segments, segment_bytes); |
35 | if (ret) | |
36 | ret = -EFAULT; | |
37 | ||
38 | return ret; | |
39 | } | |
40 | ||
255aedd9 VG |
41 | static int kimage_alloc_init(struct kimage **rimage, unsigned long entry, |
42 | unsigned long nr_segments, | |
43 | struct kexec_segment __user *segments, | |
44 | unsigned long flags) | |
dc009d92 | 45 | { |
255aedd9 | 46 | int ret; |
dc009d92 | 47 | struct kimage *image; |
255aedd9 VG |
48 | bool kexec_on_panic = flags & KEXEC_ON_CRASH; |
49 | ||
50 | if (kexec_on_panic) { | |
51 | /* Verify we have a valid entry point */ | |
52 | if ((entry < crashk_res.start) || (entry > crashk_res.end)) | |
53 | return -EADDRNOTAVAIL; | |
54 | } | |
dc009d92 EB |
55 | |
56 | /* Allocate and initialize a controlling structure */ | |
dabe7862 VG |
57 | image = do_kimage_alloc_init(); |
58 | if (!image) | |
59 | return -ENOMEM; | |
60 | ||
61 | image->start = entry; | |
62 | ||
255aedd9 VG |
63 | ret = copy_user_segment_list(image, nr_segments, segments); |
64 | if (ret) | |
dabe7862 VG |
65 | goto out_free_image; |
66 | ||
255aedd9 VG |
67 | ret = sanity_check_segment_list(image); |
68 | if (ret) | |
dabe7862 | 69 | goto out_free_image; |
72414d3f | 70 | |
255aedd9 VG |
71 | /* Enable the special crash kernel control page allocation policy. */ |
72 | if (kexec_on_panic) { | |
73 | image->control_page = crashk_res.start; | |
74 | image->type = KEXEC_TYPE_CRASH; | |
75 | } | |
76 | ||
dc009d92 EB |
77 | /* |
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. | |
81 | */ | |
255aedd9 | 82 | ret = -ENOMEM; |
dc009d92 | 83 | image->control_code_page = kimage_alloc_control_pages(image, |
163f6876 | 84 | get_order(KEXEC_CONTROL_PAGE_SIZE)); |
dc009d92 | 85 | if (!image->control_code_page) { |
e1bebcf4 | 86 | pr_err("Could not allocate control_code_buffer\n"); |
dabe7862 | 87 | goto out_free_image; |
dc009d92 EB |
88 | } |
89 | ||
255aedd9 VG |
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; | |
95 | } | |
3ab83521 HY |
96 | } |
97 | ||
b92e7e0d ZY |
98 | *rimage = image; |
99 | return 0; | |
dabe7862 | 100 | out_free_control_pages: |
b92e7e0d | 101 | kimage_free_page_list(&image->control_pages); |
dabe7862 | 102 | out_free_image: |
b92e7e0d | 103 | kfree(image); |
255aedd9 | 104 | return ret; |
dc009d92 EB |
105 | } |
106 | ||
dc009d92 EB |
107 | /* |
108 | * Exec Kernel system call: for obvious reasons only root may call it. | |
109 | * | |
110 | * This call breaks up into three pieces. | |
111 | * - A generic part which loads the new kernel from the current | |
112 | * address space, and very carefully places the data in the | |
113 | * allocated pages. | |
114 | * | |
115 | * - A generic part that interacts with the kernel and tells all of | |
116 | * the devices to shut down. Preventing on-going dmas, and placing | |
117 | * the devices in a consistent state so a later kernel can | |
118 | * reinitialize them. | |
119 | * | |
120 | * - A machine specific part that includes the syscall number | |
002ace78 | 121 | * and then copies the image to it's final destination. And |
dc009d92 EB |
122 | * jumps into the image at entry. |
123 | * | |
124 | * kexec does not sync, or unmount filesystems so if you need | |
125 | * that to happen you need to do that yourself. | |
126 | */ | |
8c5a1cf0 | 127 | |
754fe8d2 HC |
128 | SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments, |
129 | struct kexec_segment __user *, segments, unsigned long, flags) | |
dc009d92 EB |
130 | { |
131 | struct kimage **dest_image, *image; | |
dc009d92 EB |
132 | int result; |
133 | ||
134 | /* We only trust the superuser with rebooting the system. */ | |
17c953d3 | 135 | if (!capable(CAP_SYS_BOOT) || kexec_load_disabled || secure_modules()) |
dc009d92 EB |
136 | return -EPERM; |
137 | ||
138 | /* | |
139 | * Verify we have a legal set of flags | |
140 | * This leaves us room for future extensions. | |
141 | */ | |
142 | if ((flags & KEXEC_FLAGS) != (flags & ~KEXEC_ARCH_MASK)) | |
143 | return -EINVAL; | |
144 | ||
145 | /* Verify we are on the appropriate architecture */ | |
146 | if (((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH) && | |
147 | ((flags & KEXEC_ARCH_MASK) != KEXEC_ARCH_DEFAULT)) | |
dc009d92 | 148 | return -EINVAL; |
dc009d92 EB |
149 | |
150 | /* Put an artificial cap on the number | |
151 | * of segments passed to kexec_load. | |
152 | */ | |
153 | if (nr_segments > KEXEC_SEGMENT_MAX) | |
154 | return -EINVAL; | |
155 | ||
156 | image = NULL; | |
157 | result = 0; | |
158 | ||
159 | /* Because we write directly to the reserved memory | |
160 | * region when loading crash kernels we need a mutex here to | |
161 | * prevent multiple crash kernels from attempting to load | |
162 | * simultaneously, and to prevent a crash kernel from loading | |
163 | * over the top of a in use crash kernel. | |
164 | * | |
165 | * KISS: always take the mutex. | |
166 | */ | |
8c5a1cf0 | 167 | if (!mutex_trylock(&kexec_mutex)) |
dc009d92 | 168 | return -EBUSY; |
72414d3f | 169 | |
dc009d92 | 170 | dest_image = &kexec_image; |
72414d3f | 171 | if (flags & KEXEC_ON_CRASH) |
dc009d92 | 172 | dest_image = &kexec_crash_image; |
dc009d92 EB |
173 | if (nr_segments > 0) { |
174 | unsigned long i; | |
72414d3f | 175 | |
518a0c71 GL |
176 | if (flags & KEXEC_ON_CRASH) { |
177 | /* | |
178 | * Loading another kernel to switch to if this one | |
179 | * crashes. Free any current crash dump kernel before | |
dc009d92 EB |
180 | * we corrupt it. |
181 | */ | |
518a0c71 | 182 | |
dc009d92 | 183 | kimage_free(xchg(&kexec_crash_image, NULL)); |
255aedd9 VG |
184 | result = kimage_alloc_init(&image, entry, nr_segments, |
185 | segments, flags); | |
558df720 | 186 | crash_map_reserved_pages(); |
518a0c71 GL |
187 | } else { |
188 | /* Loading another kernel to reboot into. */ | |
189 | ||
190 | result = kimage_alloc_init(&image, entry, nr_segments, | |
191 | segments, flags); | |
dc009d92 | 192 | } |
72414d3f | 193 | if (result) |
dc009d92 | 194 | goto out; |
72414d3f | 195 | |
3ab83521 HY |
196 | if (flags & KEXEC_PRESERVE_CONTEXT) |
197 | image->preserve_context = 1; | |
dc009d92 | 198 | result = machine_kexec_prepare(image); |
72414d3f | 199 | if (result) |
dc009d92 | 200 | goto out; |
72414d3f MS |
201 | |
202 | for (i = 0; i < nr_segments; i++) { | |
dc009d92 | 203 | result = kimage_load_segment(image, &image->segment[i]); |
72414d3f | 204 | if (result) |
dc009d92 | 205 | goto out; |
dc009d92 | 206 | } |
7fccf032 | 207 | kimage_terminate(image); |
558df720 MH |
208 | if (flags & KEXEC_ON_CRASH) |
209 | crash_unmap_reserved_pages(); | |
dc009d92 EB |
210 | } |
211 | /* Install the new kernel, and Uninstall the old */ | |
212 | image = xchg(dest_image, image); | |
213 | ||
72414d3f | 214 | out: |
8c5a1cf0 | 215 | mutex_unlock(&kexec_mutex); |
dc009d92 | 216 | kimage_free(image); |
72414d3f | 217 | |
dc009d92 EB |
218 | return result; |
219 | } | |
220 | ||
221 | #ifdef CONFIG_COMPAT | |
ca2c405a HC |
222 | COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry, |
223 | compat_ulong_t, nr_segments, | |
224 | struct compat_kexec_segment __user *, segments, | |
225 | compat_ulong_t, flags) | |
dc009d92 EB |
226 | { |
227 | struct compat_kexec_segment in; | |
228 | struct kexec_segment out, __user *ksegments; | |
229 | unsigned long i, result; | |
230 | ||
231 | /* Don't allow clients that don't understand the native | |
232 | * architecture to do anything. | |
233 | */ | |
72414d3f | 234 | if ((flags & KEXEC_ARCH_MASK) == KEXEC_ARCH_DEFAULT) |
dc009d92 | 235 | return -EINVAL; |
dc009d92 | 236 | |
72414d3f | 237 | if (nr_segments > KEXEC_SEGMENT_MAX) |
dc009d92 | 238 | return -EINVAL; |
dc009d92 EB |
239 | |
240 | ksegments = compat_alloc_user_space(nr_segments * sizeof(out)); | |
e1bebcf4 | 241 | for (i = 0; i < nr_segments; i++) { |
dc009d92 | 242 | result = copy_from_user(&in, &segments[i], sizeof(in)); |
72414d3f | 243 | if (result) |
dc009d92 | 244 | return -EFAULT; |
dc009d92 EB |
245 | |
246 | out.buf = compat_ptr(in.buf); | |
247 | out.bufsz = in.bufsz; | |
248 | out.mem = in.mem; | |
249 | out.memsz = in.memsz; | |
250 | ||
251 | result = copy_to_user(&ksegments[i], &out, sizeof(out)); | |
72414d3f | 252 | if (result) |
dc009d92 | 253 | return -EFAULT; |
dc009d92 EB |
254 | } |
255 | ||
256 | return sys_kexec_load(entry, nr_segments, ksegments, flags); | |
257 | } | |
258 | #endif |