1 // SPDX-License-Identifier: GPL-2.0-only <<
2 /* 1 /*
3 * kexec.c - kexec system call core code. 2 * kexec.c - kexec system call core code.
4 * Copyright (C) 2002-2004 Eric Biederman <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.
5 */ 7 */
6 8
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8 10
9 #include <linux/btf.h> <<
10 #include <linux/capability.h> 11 #include <linux/capability.h>
11 #include <linux/mm.h> 12 #include <linux/mm.h>
12 #include <linux/file.h> 13 #include <linux/file.h>
13 #include <linux/slab.h> 14 #include <linux/slab.h>
14 #include <linux/fs.h> 15 #include <linux/fs.h>
15 #include <linux/kexec.h> 16 #include <linux/kexec.h>
16 #include <linux/mutex.h> 17 #include <linux/mutex.h>
17 #include <linux/list.h> 18 #include <linux/list.h>
18 #include <linux/highmem.h> 19 #include <linux/highmem.h>
19 #include <linux/syscalls.h> 20 #include <linux/syscalls.h>
20 #include <linux/reboot.h> 21 #include <linux/reboot.h>
21 #include <linux/ioport.h> 22 #include <linux/ioport.h>
22 #include <linux/hardirq.h> 23 #include <linux/hardirq.h>
23 #include <linux/elf.h> 24 #include <linux/elf.h>
24 #include <linux/elfcore.h> 25 #include <linux/elfcore.h>
25 #include <linux/utsname.h> 26 #include <linux/utsname.h>
26 #include <linux/numa.h> 27 #include <linux/numa.h>
27 #include <linux/suspend.h> 28 #include <linux/suspend.h>
28 #include <linux/device.h> 29 #include <linux/device.h>
29 #include <linux/freezer.h> 30 #include <linux/freezer.h>
30 #include <linux/panic_notifier.h> <<
31 #include <linux/pm.h> 31 #include <linux/pm.h>
32 #include <linux/cpu.h> 32 #include <linux/cpu.h>
33 #include <linux/uaccess.h> 33 #include <linux/uaccess.h>
34 #include <linux/io.h> 34 #include <linux/io.h>
35 #include <linux/console.h> 35 #include <linux/console.h>
36 #include <linux/vmalloc.h> 36 #include <linux/vmalloc.h>
37 #include <linux/swap.h> 37 #include <linux/swap.h>
38 #include <linux/syscore_ops.h> 38 #include <linux/syscore_ops.h>
39 #include <linux/compiler.h> 39 #include <linux/compiler.h>
40 #include <linux/hugetlb.h> 40 #include <linux/hugetlb.h>
41 #include <linux/objtool.h> <<
42 #include <linux/kmsg_dump.h> <<
43 41
44 #include <asm/page.h> 42 #include <asm/page.h>
45 #include <asm/sections.h> 43 #include <asm/sections.h>
46 44
47 #include <crypto/hash.h> 45 #include <crypto/hash.h>
>> 46 #include <crypto/sha.h>
48 #include "kexec_internal.h" 47 #include "kexec_internal.h"
49 48
50 atomic_t __kexec_lock = ATOMIC_INIT(0); !! 49 DEFINE_MUTEX(kexec_mutex);
>> 50
>> 51 /* Per cpu memory for storing cpu states in case of system crash. */
>> 52 note_buf_t __percpu *crash_notes;
>> 53
>> 54 /* vmcoreinfo stuff */
>> 55 static unsigned char vmcoreinfo_data[VMCOREINFO_BYTES];
>> 56 u32 vmcoreinfo_note[VMCOREINFO_NOTE_SIZE/4];
>> 57 size_t vmcoreinfo_size;
>> 58 size_t vmcoreinfo_max_size = sizeof(vmcoreinfo_data);
51 59
52 /* Flag to indicate we are going to kexec a ne 60 /* Flag to indicate we are going to kexec a new kernel */
53 bool kexec_in_progress = false; 61 bool kexec_in_progress = false;
54 62
55 bool kexec_file_dbg_print; !! 63
>> 64 /* Location of the reserved area for the crash kernel */
>> 65 struct resource crashk_res = {
>> 66 .name = "Crash kernel",
>> 67 .start = 0,
>> 68 .end = 0,
>> 69 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
>> 70 };
>> 71 struct resource crashk_low_res = {
>> 72 .name = "Crash kernel",
>> 73 .start = 0,
>> 74 .end = 0,
>> 75 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
>> 76 };
>> 77
>> 78 int kexec_should_crash(struct task_struct *p)
>> 79 {
>> 80 /*
>> 81 * If crash_kexec_post_notifiers is enabled, don't run
>> 82 * crash_kexec() here yet, which must be run after panic
>> 83 * notifiers in panic().
>> 84 */
>> 85 if (crash_kexec_post_notifiers)
>> 86 return 0;
>> 87 /*
>> 88 * There are 4 panic() calls in do_exit() path, each of which
>> 89 * corresponds to each of these 4 conditions.
>> 90 */
>> 91 if (in_interrupt() || !p->pid || is_global_init(p) || panic_on_oops)
>> 92 return 1;
>> 93 return 0;
>> 94 }
56 95
57 /* 96 /*
58 * When kexec transitions to the new kernel th 97 * When kexec transitions to the new kernel there is a one-to-one
59 * mapping between physical and virtual addres 98 * mapping between physical and virtual addresses. On processors
60 * where you can disable the MMU this is trivi 99 * where you can disable the MMU this is trivial, and easy. For
61 * others it is still a simple predictable pag 100 * others it is still a simple predictable page table to setup.
62 * 101 *
63 * In that environment kexec copies the new ke 102 * In that environment kexec copies the new kernel to its final
64 * resting place. This means I can only suppo 103 * resting place. This means I can only support memory whose
65 * physical address can fit in an unsigned lon 104 * physical address can fit in an unsigned long. In particular
66 * addresses where (pfn << PAGE_SHIFT) > ULONG 105 * addresses where (pfn << PAGE_SHIFT) > ULONG_MAX cannot be handled.
67 * If the assembly stub has more restrictive r 106 * If the assembly stub has more restrictive requirements
68 * KEXEC_SOURCE_MEMORY_LIMIT and KEXEC_DEST_ME 107 * KEXEC_SOURCE_MEMORY_LIMIT and KEXEC_DEST_MEMORY_LIMIT can be
69 * defined more restrictively in <asm/kexec.h> 108 * defined more restrictively in <asm/kexec.h>.
70 * 109 *
71 * The code for the transition from the curren 110 * The code for the transition from the current kernel to the
72 * new kernel is placed in the control_code_bu !! 111 * the new kernel is placed in the control_code_buffer, whose size
73 * is given by KEXEC_CONTROL_PAGE_SIZE. In th 112 * is given by KEXEC_CONTROL_PAGE_SIZE. In the best case only a single
74 * page of memory is necessary, but some archi 113 * page of memory is necessary, but some architectures require more.
75 * Because this memory must be identity mapped 114 * Because this memory must be identity mapped in the transition from
76 * virtual to physical addresses it must live 115 * virtual to physical addresses it must live in the range
77 * 0 - TASK_SIZE, as only the user space mappi 116 * 0 - TASK_SIZE, as only the user space mappings are arbitrarily
78 * modifiable. 117 * modifiable.
79 * 118 *
80 * The assembly stub in the control code buffe 119 * The assembly stub in the control code buffer is passed a linked list
81 * of descriptor pages detailing the source pa 120 * of descriptor pages detailing the source pages of the new kernel,
82 * and the destination addresses of those sour 121 * and the destination addresses of those source pages. As this data
83 * structure is not used in the context of the 122 * structure is not used in the context of the current OS, it must
84 * be self-contained. 123 * be self-contained.
85 * 124 *
86 * The code has been made to work with highmem 125 * The code has been made to work with highmem pages and will use a
87 * destination page in its final resting place 126 * destination page in its final resting place (if it happens
88 * to allocate it). The end product of this i 127 * to allocate it). The end product of this is that most of the
89 * physical address space, and most of RAM can 128 * physical address space, and most of RAM can be used.
90 * 129 *
91 * Future directions include: 130 * Future directions include:
92 * - allocating a page table with the control 131 * - allocating a page table with the control code buffer identity
93 * mapped, to simplify machine_kexec and ma 132 * mapped, to simplify machine_kexec and make kexec_on_panic more
94 * reliable. 133 * reliable.
95 */ 134 */
96 135
97 /* 136 /*
98 * KIMAGE_NO_DEST is an impossible destination 137 * KIMAGE_NO_DEST is an impossible destination address..., for
99 * allocating pages whose destination address 138 * allocating pages whose destination address we do not care about.
100 */ 139 */
101 #define KIMAGE_NO_DEST (-1UL) 140 #define KIMAGE_NO_DEST (-1UL)
102 #define PAGE_COUNT(x) (((x) + PAGE_SIZE - 1) > <<
103 141
104 static struct page *kimage_alloc_page(struct k 142 static struct page *kimage_alloc_page(struct kimage *image,
105 gfp_t g 143 gfp_t gfp_mask,
106 unsigne 144 unsigned long dest);
107 145
108 int sanity_check_segment_list(struct kimage *i 146 int sanity_check_segment_list(struct kimage *image)
109 { 147 {
110 int i; !! 148 int result, i;
111 unsigned long nr_segments = image->nr_ 149 unsigned long nr_segments = image->nr_segments;
112 unsigned long total_pages = 0; <<
113 unsigned long nr_pages = totalram_page <<
114 150
115 /* 151 /*
116 * Verify we have good destination add 152 * Verify we have good destination addresses. The caller is
117 * responsible for making certain we d 153 * responsible for making certain we don't attempt to load
118 * the new image into invalid or reser 154 * the new image into invalid or reserved areas of RAM. This
119 * just verifies it is an address we c 155 * just verifies it is an address we can use.
120 * 156 *
121 * Since the kernel does everything in 157 * Since the kernel does everything in page size chunks ensure
122 * the destination addresses are page 158 * the destination addresses are page aligned. Too many
123 * special cases crop of when we don't 159 * special cases crop of when we don't do this. The most
124 * insidious is getting overlapping de 160 * insidious is getting overlapping destination addresses
125 * simply because addresses are change 161 * simply because addresses are changed to page size
126 * granularity. 162 * granularity.
127 */ 163 */
>> 164 result = -EADDRNOTAVAIL;
128 for (i = 0; i < nr_segments; i++) { 165 for (i = 0; i < nr_segments; i++) {
129 unsigned long mstart, mend; 166 unsigned long mstart, mend;
130 167
131 mstart = image->segment[i].mem 168 mstart = image->segment[i].mem;
132 mend = mstart + image->segme 169 mend = mstart + image->segment[i].memsz;
133 if (mstart > mend) <<
134 return -EADDRNOTAVAIL; <<
135 if ((mstart & ~PAGE_MASK) || ( 170 if ((mstart & ~PAGE_MASK) || (mend & ~PAGE_MASK))
136 return -EADDRNOTAVAIL; !! 171 return result;
137 if (mend >= KEXEC_DESTINATION_ 172 if (mend >= KEXEC_DESTINATION_MEMORY_LIMIT)
138 return -EADDRNOTAVAIL; !! 173 return result;
139 } 174 }
140 175
141 /* Verify our destination addresses do 176 /* Verify our destination addresses do not overlap.
142 * If we alloed overlapping destinatio 177 * If we alloed overlapping destination addresses
143 * through very weird things can happe 178 * through very weird things can happen with no
144 * easy explanation as one segment sto 179 * easy explanation as one segment stops on another.
145 */ 180 */
>> 181 result = -EINVAL;
146 for (i = 0; i < nr_segments; i++) { 182 for (i = 0; i < nr_segments; i++) {
147 unsigned long mstart, mend; 183 unsigned long mstart, mend;
148 unsigned long j; 184 unsigned long j;
149 185
150 mstart = image->segment[i].mem 186 mstart = image->segment[i].mem;
151 mend = mstart + image->segme 187 mend = mstart + image->segment[i].memsz;
152 for (j = 0; j < i; j++) { 188 for (j = 0; j < i; j++) {
153 unsigned long pstart, 189 unsigned long pstart, pend;
154 190
155 pstart = image->segmen 191 pstart = image->segment[j].mem;
156 pend = pstart + imag 192 pend = pstart + image->segment[j].memsz;
157 /* Do the segments ove 193 /* Do the segments overlap ? */
158 if ((mend > pstart) && 194 if ((mend > pstart) && (mstart < pend))
159 return -EINVAL !! 195 return result;
160 } 196 }
161 } 197 }
162 198
163 /* Ensure our buffer sizes are strictl 199 /* Ensure our buffer sizes are strictly less than
164 * our memory sizes. This should alwa 200 * our memory sizes. This should always be the case,
165 * and it is easier to check up front 201 * and it is easier to check up front than to be surprised
166 * later on. 202 * later on.
167 */ 203 */
>> 204 result = -EINVAL;
168 for (i = 0; i < nr_segments; i++) { 205 for (i = 0; i < nr_segments; i++) {
169 if (image->segment[i].bufsz > 206 if (image->segment[i].bufsz > image->segment[i].memsz)
170 return -EINVAL; !! 207 return result;
171 } 208 }
172 209
173 /* 210 /*
174 * Verify that no more than half of me <<
175 * request from userspace is too large <<
176 * wasted allocating pages, which can <<
177 */ <<
178 for (i = 0; i < nr_segments; i++) { <<
179 if (PAGE_COUNT(image->segment[ <<
180 return -EINVAL; <<
181 <<
182 total_pages += PAGE_COUNT(imag <<
183 } <<
184 <<
185 if (total_pages > nr_pages / 2) <<
186 return -EINVAL; <<
187 <<
188 #ifdef CONFIG_CRASH_DUMP <<
189 /* <<
190 * Verify we have good destination add 211 * Verify we have good destination addresses. Normally
191 * the caller is responsible for makin 212 * the caller is responsible for making certain we don't
192 * attempt to load the new image into 213 * attempt to load the new image into invalid or reserved
193 * areas of RAM. But crash kernels ar 214 * areas of RAM. But crash kernels are preloaded into a
194 * reserved area of ram. We must ensu 215 * reserved area of ram. We must ensure the addresses
195 * are in the reserved area otherwise 216 * are in the reserved area otherwise preloading the
196 * kernel could corrupt things. 217 * kernel could corrupt things.
197 */ 218 */
198 219
199 if (image->type == KEXEC_TYPE_CRASH) { 220 if (image->type == KEXEC_TYPE_CRASH) {
>> 221 result = -EADDRNOTAVAIL;
200 for (i = 0; i < nr_segments; i 222 for (i = 0; i < nr_segments; i++) {
201 unsigned long mstart, 223 unsigned long mstart, mend;
202 224
203 mstart = image->segmen 225 mstart = image->segment[i].mem;
204 mend = mstart + image- 226 mend = mstart + image->segment[i].memsz - 1;
205 /* Ensure we are withi 227 /* Ensure we are within the crash kernel limits */
206 if ((mstart < phys_to_ !! 228 if ((mstart < crashk_res.start) ||
207 (mend > phys_to_bo !! 229 (mend > crashk_res.end))
208 return -EADDRN !! 230 return result;
209 } 231 }
210 } 232 }
211 #endif <<
212 233
213 return 0; 234 return 0;
214 } 235 }
215 236
216 struct kimage *do_kimage_alloc_init(void) 237 struct kimage *do_kimage_alloc_init(void)
217 { 238 {
218 struct kimage *image; 239 struct kimage *image;
219 240
220 /* Allocate a controlling structure */ 241 /* Allocate a controlling structure */
221 image = kzalloc(sizeof(*image), GFP_KE 242 image = kzalloc(sizeof(*image), GFP_KERNEL);
222 if (!image) 243 if (!image)
223 return NULL; 244 return NULL;
224 245
225 image->head = 0; 246 image->head = 0;
226 image->entry = &image->head; 247 image->entry = &image->head;
227 image->last_entry = &image->head; 248 image->last_entry = &image->head;
228 image->control_page = ~0; /* By defaul 249 image->control_page = ~0; /* By default this does not apply */
229 image->type = KEXEC_TYPE_DEFAULT; 250 image->type = KEXEC_TYPE_DEFAULT;
230 251
231 /* Initialize the list of control page 252 /* Initialize the list of control pages */
232 INIT_LIST_HEAD(&image->control_pages); 253 INIT_LIST_HEAD(&image->control_pages);
233 254
234 /* Initialize the list of destination 255 /* Initialize the list of destination pages */
235 INIT_LIST_HEAD(&image->dest_pages); 256 INIT_LIST_HEAD(&image->dest_pages);
236 257
237 /* Initialize the list of unusable pag 258 /* Initialize the list of unusable pages */
238 INIT_LIST_HEAD(&image->unusable_pages) 259 INIT_LIST_HEAD(&image->unusable_pages);
239 260
240 #ifdef CONFIG_CRASH_HOTPLUG <<
241 image->hp_action = KEXEC_CRASH_HP_NONE <<
242 image->elfcorehdr_index = -1; <<
243 image->elfcorehdr_updated = false; <<
244 #endif <<
245 <<
246 return image; 261 return image;
247 } 262 }
248 263
249 int kimage_is_destination_range(struct kimage 264 int kimage_is_destination_range(struct kimage *image,
250 unsign 265 unsigned long start,
251 unsign 266 unsigned long end)
252 { 267 {
253 unsigned long i; 268 unsigned long i;
254 269
255 for (i = 0; i < image->nr_segments; i+ 270 for (i = 0; i < image->nr_segments; i++) {
256 unsigned long mstart, mend; 271 unsigned long mstart, mend;
257 272
258 mstart = image->segment[i].mem 273 mstart = image->segment[i].mem;
259 mend = mstart + image->segment !! 274 mend = mstart + image->segment[i].memsz;
260 if ((end >= mstart) && (start !! 275 if ((end > mstart) && (start < mend))
261 return 1; 276 return 1;
262 } 277 }
263 278
264 return 0; 279 return 0;
265 } 280 }
266 281
267 static struct page *kimage_alloc_pages(gfp_t g 282 static struct page *kimage_alloc_pages(gfp_t gfp_mask, unsigned int order)
268 { 283 {
269 struct page *pages; 284 struct page *pages;
270 285
271 if (fatal_signal_pending(current)) !! 286 pages = alloc_pages(gfp_mask, order);
272 return NULL; <<
273 pages = alloc_pages(gfp_mask & ~__GFP_ <<
274 if (pages) { 287 if (pages) {
275 unsigned int count, i; 288 unsigned int count, i;
276 289
277 pages->mapping = NULL; 290 pages->mapping = NULL;
278 set_page_private(pages, order) 291 set_page_private(pages, order);
279 count = 1 << order; 292 count = 1 << order;
280 for (i = 0; i < count; i++) 293 for (i = 0; i < count; i++)
281 SetPageReserved(pages 294 SetPageReserved(pages + i);
282 <<
283 arch_kexec_post_alloc_pages(pa <<
284 gf <<
285 <<
286 if (gfp_mask & __GFP_ZERO) <<
287 for (i = 0; i < count; <<
288 clear_highpage <<
289 } 295 }
290 296
291 return pages; 297 return pages;
292 } 298 }
293 299
294 static void kimage_free_pages(struct page *pag 300 static void kimage_free_pages(struct page *page)
295 { 301 {
296 unsigned int order, count, i; 302 unsigned int order, count, i;
297 303
298 order = page_private(page); 304 order = page_private(page);
299 count = 1 << order; 305 count = 1 << order;
300 <<
301 arch_kexec_pre_free_pages(page_address <<
302 <<
303 for (i = 0; i < count; i++) 306 for (i = 0; i < count; i++)
304 ClearPageReserved(page + i); 307 ClearPageReserved(page + i);
305 __free_pages(page, order); 308 __free_pages(page, order);
306 } 309 }
307 310
308 void kimage_free_page_list(struct list_head *l 311 void kimage_free_page_list(struct list_head *list)
309 { 312 {
310 struct page *page, *next; !! 313 struct list_head *pos, *next;
>> 314
>> 315 list_for_each_safe(pos, next, list) {
>> 316 struct page *page;
311 317
312 list_for_each_entry_safe(page, next, l !! 318 page = list_entry(pos, struct page, lru);
313 list_del(&page->lru); 319 list_del(&page->lru);
314 kimage_free_pages(page); 320 kimage_free_pages(page);
315 } 321 }
316 } 322 }
317 323
318 static struct page *kimage_alloc_normal_contro 324 static struct page *kimage_alloc_normal_control_pages(struct kimage *image,
319 325 unsigned int order)
320 { 326 {
321 /* Control pages are special, they are 327 /* Control pages are special, they are the intermediaries
322 * that are needed while we copy the r 328 * that are needed while we copy the rest of the pages
323 * to their final resting place. As s 329 * to their final resting place. As such they must
324 * not conflict with either the destin 330 * not conflict with either the destination addresses
325 * or memory the kernel is already usi 331 * or memory the kernel is already using.
326 * 332 *
327 * The only case where we really need 333 * The only case where we really need more than one of
328 * these are for architectures where w 334 * these are for architectures where we cannot disable
329 * the MMU and must instead generate a 335 * the MMU and must instead generate an identity mapped
330 * page table for all of the memory. 336 * page table for all of the memory.
331 * 337 *
332 * At worst this runs in O(N) of the i 338 * At worst this runs in O(N) of the image size.
333 */ 339 */
334 struct list_head extra_pages; 340 struct list_head extra_pages;
335 struct page *pages; 341 struct page *pages;
336 unsigned int count; 342 unsigned int count;
337 343
338 count = 1 << order; 344 count = 1 << order;
339 INIT_LIST_HEAD(&extra_pages); 345 INIT_LIST_HEAD(&extra_pages);
340 346
341 /* Loop while I can allocate a page an 347 /* Loop while I can allocate a page and the page allocated
342 * is a destination page. 348 * is a destination page.
343 */ 349 */
344 do { 350 do {
345 unsigned long pfn, epfn, addr, 351 unsigned long pfn, epfn, addr, eaddr;
346 352
347 pages = kimage_alloc_pages(KEX 353 pages = kimage_alloc_pages(KEXEC_CONTROL_MEMORY_GFP, order);
348 if (!pages) 354 if (!pages)
349 break; 355 break;
350 pfn = page_to_boot_pfn(pages !! 356 pfn = page_to_pfn(pages);
351 epfn = pfn + count; 357 epfn = pfn + count;
352 addr = pfn << PAGE_SHIFT; 358 addr = pfn << PAGE_SHIFT;
353 eaddr = (epfn << PAGE_SHIFT) - !! 359 eaddr = epfn << PAGE_SHIFT;
354 if ((epfn >= (KEXEC_CONTROL_ME 360 if ((epfn >= (KEXEC_CONTROL_MEMORY_LIMIT >> PAGE_SHIFT)) ||
355 kimage_is_destin 361 kimage_is_destination_range(image, addr, eaddr)) {
356 list_add(&pages->lru, 362 list_add(&pages->lru, &extra_pages);
357 pages = NULL; 363 pages = NULL;
358 } 364 }
359 } while (!pages); 365 } while (!pages);
360 366
361 if (pages) { 367 if (pages) {
362 /* Remember the allocated page 368 /* Remember the allocated page... */
363 list_add(&pages->lru, &image-> 369 list_add(&pages->lru, &image->control_pages);
364 370
365 /* Because the page is already 371 /* Because the page is already in it's destination
366 * location we will never allo 372 * location we will never allocate another page at
367 * that address. Therefore ki 373 * that address. Therefore kimage_alloc_pages
368 * will not return it (again) 374 * will not return it (again) and we don't need
369 * to give it an entry in imag 375 * to give it an entry in image->segment[].
370 */ 376 */
371 } 377 }
372 /* Deal with the destination pages I h 378 /* Deal with the destination pages I have inadvertently allocated.
373 * 379 *
374 * Ideally I would convert multi-page 380 * Ideally I would convert multi-page allocations into single
375 * page allocations, and add everythin 381 * page allocations, and add everything to image->dest_pages.
376 * 382 *
377 * For now it is simpler to just free 383 * For now it is simpler to just free the pages.
378 */ 384 */
379 kimage_free_page_list(&extra_pages); 385 kimage_free_page_list(&extra_pages);
380 386
381 return pages; 387 return pages;
382 } 388 }
383 389
384 #ifdef CONFIG_CRASH_DUMP <<
385 static struct page *kimage_alloc_crash_control 390 static struct page *kimage_alloc_crash_control_pages(struct kimage *image,
386 391 unsigned int order)
387 { 392 {
388 /* Control pages are special, they are 393 /* Control pages are special, they are the intermediaries
389 * that are needed while we copy the r 394 * that are needed while we copy the rest of the pages
390 * to their final resting place. As s 395 * to their final resting place. As such they must
391 * not conflict with either the destin 396 * not conflict with either the destination addresses
392 * or memory the kernel is already usi 397 * or memory the kernel is already using.
393 * 398 *
394 * Control pages are also the only pag 399 * Control pages are also the only pags we must allocate
395 * when loading a crash kernel. All o 400 * when loading a crash kernel. All of the other pages
396 * are specified by the segments and w 401 * are specified by the segments and we just memcpy
397 * into them directly. 402 * into them directly.
398 * 403 *
399 * The only case where we really need 404 * The only case where we really need more than one of
400 * these are for architectures where w 405 * these are for architectures where we cannot disable
401 * the MMU and must instead generate a 406 * the MMU and must instead generate an identity mapped
402 * page table for all of the memory. 407 * page table for all of the memory.
403 * 408 *
404 * Given the low demand this implement 409 * Given the low demand this implements a very simple
405 * allocator that finds the first hole 410 * allocator that finds the first hole of the appropriate
406 * size in the reserved memory region, 411 * size in the reserved memory region, and allocates all
407 * of the memory up to and including t 412 * of the memory up to and including the hole.
408 */ 413 */
409 unsigned long hole_start, hole_end, si 414 unsigned long hole_start, hole_end, size;
410 struct page *pages; 415 struct page *pages;
411 416
412 pages = NULL; 417 pages = NULL;
413 size = (1 << order) << PAGE_SHIFT; 418 size = (1 << order) << PAGE_SHIFT;
414 hole_start = ALIGN(image->control_page !! 419 hole_start = (image->control_page + (size - 1)) & ~(size - 1);
415 hole_end = hole_start + size - 1; 420 hole_end = hole_start + size - 1;
416 while (hole_end <= crashk_res.end) { 421 while (hole_end <= crashk_res.end) {
417 unsigned long i; 422 unsigned long i;
418 423
419 cond_resched(); <<
420 <<
421 if (hole_end > KEXEC_CRASH_CON 424 if (hole_end > KEXEC_CRASH_CONTROL_MEMORY_LIMIT)
422 break; 425 break;
423 /* See if I overlap any of the 426 /* See if I overlap any of the segments */
424 for (i = 0; i < image->nr_segm 427 for (i = 0; i < image->nr_segments; i++) {
425 unsigned long mstart, 428 unsigned long mstart, mend;
426 429
427 mstart = image->segmen 430 mstart = image->segment[i].mem;
428 mend = mstart + imag 431 mend = mstart + image->segment[i].memsz - 1;
429 if ((hole_end >= mstar 432 if ((hole_end >= mstart) && (hole_start <= mend)) {
430 /* Advance the 433 /* Advance the hole to the end of the segment */
431 hole_start = A !! 434 hole_start = (mend + (size - 1)) & ~(size - 1);
432 hole_end = h 435 hole_end = hole_start + size - 1;
433 break; 436 break;
434 } 437 }
435 } 438 }
436 /* If I don't overlap any segm 439 /* If I don't overlap any segments I have found my hole! */
437 if (i == image->nr_segments) { 440 if (i == image->nr_segments) {
438 pages = pfn_to_page(ho 441 pages = pfn_to_page(hole_start >> PAGE_SHIFT);
439 image->control_page = !! 442 image->control_page = hole_end;
440 break; 443 break;
441 } 444 }
442 } 445 }
443 446
444 /* Ensure that these pages are decrypt <<
445 if (pages) <<
446 arch_kexec_post_alloc_pages(pa <<
447 <<
448 return pages; 447 return pages;
449 } 448 }
450 #endif <<
451 449
452 450
453 struct page *kimage_alloc_control_pages(struct 451 struct page *kimage_alloc_control_pages(struct kimage *image,
454 unsig 452 unsigned int order)
455 { 453 {
456 struct page *pages = NULL; 454 struct page *pages = NULL;
457 455
458 switch (image->type) { 456 switch (image->type) {
459 case KEXEC_TYPE_DEFAULT: 457 case KEXEC_TYPE_DEFAULT:
460 pages = kimage_alloc_normal_co 458 pages = kimage_alloc_normal_control_pages(image, order);
461 break; 459 break;
462 #ifdef CONFIG_CRASH_DUMP <<
463 case KEXEC_TYPE_CRASH: 460 case KEXEC_TYPE_CRASH:
464 pages = kimage_alloc_crash_con 461 pages = kimage_alloc_crash_control_pages(image, order);
465 break; 462 break;
466 #endif <<
467 } 463 }
468 464
469 return pages; 465 return pages;
470 } 466 }
471 467
472 static int kimage_add_entry(struct kimage *ima 468 static int kimage_add_entry(struct kimage *image, kimage_entry_t entry)
473 { 469 {
474 if (*image->entry != 0) 470 if (*image->entry != 0)
475 image->entry++; 471 image->entry++;
476 472
477 if (image->entry == image->last_entry) 473 if (image->entry == image->last_entry) {
478 kimage_entry_t *ind_page; 474 kimage_entry_t *ind_page;
479 struct page *page; 475 struct page *page;
480 476
481 page = kimage_alloc_page(image 477 page = kimage_alloc_page(image, GFP_KERNEL, KIMAGE_NO_DEST);
482 if (!page) 478 if (!page)
483 return -ENOMEM; 479 return -ENOMEM;
484 480
485 ind_page = page_address(page); 481 ind_page = page_address(page);
486 *image->entry = virt_to_boot_p !! 482 *image->entry = virt_to_phys(ind_page) | IND_INDIRECTION;
487 image->entry = ind_page; 483 image->entry = ind_page;
488 image->last_entry = ind_page + 484 image->last_entry = ind_page +
489 ((PAGE_S 485 ((PAGE_SIZE/sizeof(kimage_entry_t)) - 1);
490 } 486 }
491 *image->entry = entry; 487 *image->entry = entry;
492 image->entry++; 488 image->entry++;
493 *image->entry = 0; 489 *image->entry = 0;
494 490
495 return 0; 491 return 0;
496 } 492 }
497 493
498 static int kimage_set_destination(struct kimag 494 static int kimage_set_destination(struct kimage *image,
499 unsigned lo 495 unsigned long destination)
500 { 496 {
>> 497 int result;
>> 498
501 destination &= PAGE_MASK; 499 destination &= PAGE_MASK;
>> 500 result = kimage_add_entry(image, destination | IND_DESTINATION);
502 501
503 return kimage_add_entry(image, destina !! 502 return result;
504 } 503 }
505 504
506 505
507 static int kimage_add_page(struct kimage *imag 506 static int kimage_add_page(struct kimage *image, unsigned long page)
508 { 507 {
>> 508 int result;
>> 509
509 page &= PAGE_MASK; 510 page &= PAGE_MASK;
>> 511 result = kimage_add_entry(image, page | IND_SOURCE);
510 512
511 return kimage_add_entry(image, page | !! 513 return result;
512 } 514 }
513 515
514 516
515 static void kimage_free_extra_pages(struct kim 517 static void kimage_free_extra_pages(struct kimage *image)
516 { 518 {
517 /* Walk through and free any extra des 519 /* Walk through and free any extra destination pages I may have */
518 kimage_free_page_list(&image->dest_pag 520 kimage_free_page_list(&image->dest_pages);
519 521
520 /* Walk through and free any unusable 522 /* Walk through and free any unusable pages I have cached */
521 kimage_free_page_list(&image->unusable 523 kimage_free_page_list(&image->unusable_pages);
522 524
523 } 525 }
524 <<
525 void kimage_terminate(struct kimage *image) 526 void kimage_terminate(struct kimage *image)
526 { 527 {
527 if (*image->entry != 0) 528 if (*image->entry != 0)
528 image->entry++; 529 image->entry++;
529 530
530 *image->entry = IND_DONE; 531 *image->entry = IND_DONE;
531 } 532 }
532 533
533 #define for_each_kimage_entry(image, ptr, entr 534 #define for_each_kimage_entry(image, ptr, entry) \
534 for (ptr = &image->head; (entry = *ptr 535 for (ptr = &image->head; (entry = *ptr) && !(entry & IND_DONE); \
535 ptr = (entry & IND_INDIRECTION 536 ptr = (entry & IND_INDIRECTION) ? \
536 boot_phys_to_virt((ent !! 537 phys_to_virt((entry & PAGE_MASK)) : ptr + 1)
537 538
538 static void kimage_free_entry(kimage_entry_t e 539 static void kimage_free_entry(kimage_entry_t entry)
539 { 540 {
540 struct page *page; 541 struct page *page;
541 542
542 page = boot_pfn_to_page(entry >> PAGE_ !! 543 page = pfn_to_page(entry >> PAGE_SHIFT);
543 kimage_free_pages(page); 544 kimage_free_pages(page);
544 } 545 }
545 546
546 void kimage_free(struct kimage *image) 547 void kimage_free(struct kimage *image)
547 { 548 {
548 kimage_entry_t *ptr, entry; 549 kimage_entry_t *ptr, entry;
549 kimage_entry_t ind = 0; 550 kimage_entry_t ind = 0;
550 551
551 if (!image) 552 if (!image)
552 return; 553 return;
553 554
554 #ifdef CONFIG_CRASH_DUMP <<
555 if (image->vmcoreinfo_data_copy) { <<
556 crash_update_vmcoreinfo_safeco <<
557 vunmap(image->vmcoreinfo_data_ <<
558 } <<
559 #endif <<
560 <<
561 kimage_free_extra_pages(image); 555 kimage_free_extra_pages(image);
562 for_each_kimage_entry(image, ptr, entr 556 for_each_kimage_entry(image, ptr, entry) {
563 if (entry & IND_INDIRECTION) { 557 if (entry & IND_INDIRECTION) {
564 /* Free the previous i 558 /* Free the previous indirection page */
565 if (ind & IND_INDIRECT 559 if (ind & IND_INDIRECTION)
566 kimage_free_en 560 kimage_free_entry(ind);
567 /* Save this indirecti 561 /* Save this indirection page until we are
568 * done with it. 562 * done with it.
569 */ 563 */
570 ind = entry; 564 ind = entry;
571 } else if (entry & IND_SOURCE) 565 } else if (entry & IND_SOURCE)
572 kimage_free_entry(entr 566 kimage_free_entry(entry);
573 } 567 }
574 /* Free the final indirection page */ 568 /* Free the final indirection page */
575 if (ind & IND_INDIRECTION) 569 if (ind & IND_INDIRECTION)
576 kimage_free_entry(ind); 570 kimage_free_entry(ind);
577 571
578 /* Handle any machine specific cleanup 572 /* Handle any machine specific cleanup */
579 machine_kexec_cleanup(image); 573 machine_kexec_cleanup(image);
580 574
581 /* Free the kexec control pages... */ 575 /* Free the kexec control pages... */
582 kimage_free_page_list(&image->control_ 576 kimage_free_page_list(&image->control_pages);
583 577
584 /* 578 /*
585 * Free up any temporary buffers alloc 579 * Free up any temporary buffers allocated. This might hit if
586 * error occurred much later after buf 580 * error occurred much later after buffer allocation.
587 */ 581 */
588 if (image->file_mode) 582 if (image->file_mode)
589 kimage_file_post_load_cleanup( 583 kimage_file_post_load_cleanup(image);
590 584
591 kfree(image); 585 kfree(image);
592 } 586 }
593 587
594 static kimage_entry_t *kimage_dst_used(struct 588 static kimage_entry_t *kimage_dst_used(struct kimage *image,
595 unsign 589 unsigned long page)
596 { 590 {
597 kimage_entry_t *ptr, entry; 591 kimage_entry_t *ptr, entry;
598 unsigned long destination = 0; 592 unsigned long destination = 0;
599 593
600 for_each_kimage_entry(image, ptr, entr 594 for_each_kimage_entry(image, ptr, entry) {
601 if (entry & IND_DESTINATION) 595 if (entry & IND_DESTINATION)
602 destination = entry & 596 destination = entry & PAGE_MASK;
603 else if (entry & IND_SOURCE) { 597 else if (entry & IND_SOURCE) {
604 if (page == destinatio 598 if (page == destination)
605 return ptr; 599 return ptr;
606 destination += PAGE_SI 600 destination += PAGE_SIZE;
607 } 601 }
608 } 602 }
609 603
610 return NULL; 604 return NULL;
611 } 605 }
612 606
613 static struct page *kimage_alloc_page(struct k 607 static struct page *kimage_alloc_page(struct kimage *image,
614 gfp_t 608 gfp_t gfp_mask,
615 unsign 609 unsigned long destination)
616 { 610 {
617 /* 611 /*
618 * Here we implement safeguards to ens 612 * Here we implement safeguards to ensure that a source page
619 * is not copied to its destination pa 613 * is not copied to its destination page before the data on
620 * the destination page is no longer u 614 * the destination page is no longer useful.
621 * 615 *
622 * To do this we maintain the invarian 616 * To do this we maintain the invariant that a source page is
623 * either its own destination page, or 617 * either its own destination page, or it is not a
624 * destination page at all. 618 * destination page at all.
625 * 619 *
626 * That is slightly stronger than requ 620 * That is slightly stronger than required, but the proof
627 * that no problems will not occur is 621 * that no problems will not occur is trivial, and the
628 * implementation is simply to verify. 622 * implementation is simply to verify.
629 * 623 *
630 * When allocating all pages normally 624 * When allocating all pages normally this algorithm will run
631 * in O(N) time, but in the worst case 625 * in O(N) time, but in the worst case it will run in O(N^2)
632 * time. If the runtime is a problem 626 * time. If the runtime is a problem the data structures can
633 * be fixed. 627 * be fixed.
634 */ 628 */
635 struct page *page; 629 struct page *page;
636 unsigned long addr; 630 unsigned long addr;
637 631
638 /* 632 /*
639 * Walk through the list of destinatio 633 * Walk through the list of destination pages, and see if I
640 * have a match. 634 * have a match.
641 */ 635 */
642 list_for_each_entry(page, &image->dest 636 list_for_each_entry(page, &image->dest_pages, lru) {
643 addr = page_to_boot_pfn(page) !! 637 addr = page_to_pfn(page) << PAGE_SHIFT;
644 if (addr == destination) { 638 if (addr == destination) {
645 list_del(&page->lru); 639 list_del(&page->lru);
646 return page; 640 return page;
647 } 641 }
648 } 642 }
649 page = NULL; 643 page = NULL;
650 while (1) { 644 while (1) {
651 kimage_entry_t *old; 645 kimage_entry_t *old;
652 646
653 /* Allocate a page, if we run 647 /* Allocate a page, if we run out of memory give up */
654 page = kimage_alloc_pages(gfp_ 648 page = kimage_alloc_pages(gfp_mask, 0);
655 if (!page) 649 if (!page)
656 return NULL; 650 return NULL;
657 /* If the page cannot be used 651 /* If the page cannot be used file it away */
658 if (page_to_boot_pfn(page) > !! 652 if (page_to_pfn(page) >
659 (KEXEC_SOURCE_ 653 (KEXEC_SOURCE_MEMORY_LIMIT >> PAGE_SHIFT)) {
660 list_add(&page->lru, & 654 list_add(&page->lru, &image->unusable_pages);
661 continue; 655 continue;
662 } 656 }
663 addr = page_to_boot_pfn(page) !! 657 addr = page_to_pfn(page) << PAGE_SHIFT;
664 658
665 /* If it is the destination pa 659 /* If it is the destination page we want use it */
666 if (addr == destination) 660 if (addr == destination)
667 break; 661 break;
668 662
669 /* If the page is not a destin 663 /* If the page is not a destination page use it */
670 if (!kimage_is_destination_ran 664 if (!kimage_is_destination_range(image, addr,
671 !! 665 addr + PAGE_SIZE))
672 break; 666 break;
673 667
674 /* 668 /*
675 * I know that the page is som 669 * I know that the page is someones destination page.
676 * See if there is already a s 670 * See if there is already a source page for this
677 * destination page. And if s 671 * destination page. And if so swap the source pages.
678 */ 672 */
679 old = kimage_dst_used(image, a 673 old = kimage_dst_used(image, addr);
680 if (old) { 674 if (old) {
681 /* If so move it */ 675 /* If so move it */
682 unsigned long old_addr 676 unsigned long old_addr;
683 struct page *old_page; 677 struct page *old_page;
684 678
685 old_addr = *old & PAGE 679 old_addr = *old & PAGE_MASK;
686 old_page = boot_pfn_to !! 680 old_page = pfn_to_page(old_addr >> PAGE_SHIFT);
687 copy_highpage(page, ol 681 copy_highpage(page, old_page);
688 *old = addr | (*old & 682 *old = addr | (*old & ~PAGE_MASK);
689 683
690 /* The old page I have 684 /* The old page I have found cannot be a
691 * destination page, s 685 * destination page, so return it if it's
692 * gfp_flags honor the 686 * gfp_flags honor the ones passed in.
693 */ 687 */
694 if (!(gfp_mask & __GFP 688 if (!(gfp_mask & __GFP_HIGHMEM) &&
695 PageHighMem(old_pa 689 PageHighMem(old_page)) {
696 kimage_free_pa 690 kimage_free_pages(old_page);
697 continue; 691 continue;
698 } 692 }
>> 693 addr = old_addr;
699 page = old_page; 694 page = old_page;
700 break; 695 break;
701 } 696 }
702 /* Place the page on the desti 697 /* Place the page on the destination list, to be used later */
703 list_add(&page->lru, &image->d 698 list_add(&page->lru, &image->dest_pages);
704 } 699 }
705 700
706 return page; 701 return page;
707 } 702 }
708 703
709 static int kimage_load_normal_segment(struct k 704 static int kimage_load_normal_segment(struct kimage *image,
710 struc 705 struct kexec_segment *segment)
711 { 706 {
712 unsigned long maddr; 707 unsigned long maddr;
713 size_t ubytes, mbytes; 708 size_t ubytes, mbytes;
714 int result; 709 int result;
715 unsigned char __user *buf = NULL; 710 unsigned char __user *buf = NULL;
716 unsigned char *kbuf = NULL; 711 unsigned char *kbuf = NULL;
717 712
>> 713 result = 0;
718 if (image->file_mode) 714 if (image->file_mode)
719 kbuf = segment->kbuf; 715 kbuf = segment->kbuf;
720 else 716 else
721 buf = segment->buf; 717 buf = segment->buf;
722 ubytes = segment->bufsz; 718 ubytes = segment->bufsz;
723 mbytes = segment->memsz; 719 mbytes = segment->memsz;
724 maddr = segment->mem; 720 maddr = segment->mem;
725 721
726 result = kimage_set_destination(image, 722 result = kimage_set_destination(image, maddr);
727 if (result < 0) 723 if (result < 0)
728 goto out; 724 goto out;
729 725
730 while (mbytes) { 726 while (mbytes) {
731 struct page *page; 727 struct page *page;
732 char *ptr; 728 char *ptr;
733 size_t uchunk, mchunk; 729 size_t uchunk, mchunk;
734 730
735 page = kimage_alloc_page(image 731 page = kimage_alloc_page(image, GFP_HIGHUSER, maddr);
736 if (!page) { 732 if (!page) {
737 result = -ENOMEM; 733 result = -ENOMEM;
738 goto out; 734 goto out;
739 } 735 }
740 result = kimage_add_page(image !! 736 result = kimage_add_page(image, page_to_pfn(page)
741 737 << PAGE_SHIFT);
742 if (result < 0) 738 if (result < 0)
743 goto out; 739 goto out;
744 740
745 ptr = kmap_local_page(page); !! 741 ptr = kmap(page);
746 /* Start with a clear page */ 742 /* Start with a clear page */
747 clear_page(ptr); 743 clear_page(ptr);
748 ptr += maddr & ~PAGE_MASK; 744 ptr += maddr & ~PAGE_MASK;
749 mchunk = min_t(size_t, mbytes, 745 mchunk = min_t(size_t, mbytes,
750 PAGE_SIZE - (m 746 PAGE_SIZE - (maddr & ~PAGE_MASK));
751 uchunk = min(ubytes, mchunk); 747 uchunk = min(ubytes, mchunk);
752 748
753 if (uchunk) { !! 749 /* For file based kexec, source pages are in kernel memory */
754 /* For file based kexe !! 750 if (image->file_mode)
755 if (image->file_mode) !! 751 memcpy(ptr, kbuf, uchunk);
756 memcpy(ptr, kb !! 752 else
757 else !! 753 result = copy_from_user(ptr, buf, uchunk);
758 result = copy_ !! 754 kunmap(page);
759 ubytes -= uchunk; <<
760 if (image->file_mode) <<
761 kbuf += uchunk <<
762 else <<
763 buf += uchunk; <<
764 } <<
765 kunmap_local(ptr); <<
766 if (result) { 755 if (result) {
767 result = -EFAULT; 756 result = -EFAULT;
768 goto out; 757 goto out;
769 } 758 }
>> 759 ubytes -= uchunk;
770 maddr += mchunk; 760 maddr += mchunk;
>> 761 if (image->file_mode)
>> 762 kbuf += mchunk;
>> 763 else
>> 764 buf += mchunk;
771 mbytes -= mchunk; 765 mbytes -= mchunk;
772 <<
773 cond_resched(); <<
774 } 766 }
775 out: 767 out:
776 return result; 768 return result;
777 } 769 }
778 770
779 #ifdef CONFIG_CRASH_DUMP <<
780 static int kimage_load_crash_segment(struct ki 771 static int kimage_load_crash_segment(struct kimage *image,
781 struct 772 struct kexec_segment *segment)
782 { 773 {
783 /* For crash dumps kernels we simply c 774 /* For crash dumps kernels we simply copy the data from
784 * user space to it's destination. 775 * user space to it's destination.
785 * We do things a page at a time for t 776 * We do things a page at a time for the sake of kmap.
786 */ 777 */
787 unsigned long maddr; 778 unsigned long maddr;
788 size_t ubytes, mbytes; 779 size_t ubytes, mbytes;
789 int result; 780 int result;
790 unsigned char __user *buf = NULL; 781 unsigned char __user *buf = NULL;
791 unsigned char *kbuf = NULL; 782 unsigned char *kbuf = NULL;
792 783
793 result = 0; 784 result = 0;
794 if (image->file_mode) 785 if (image->file_mode)
795 kbuf = segment->kbuf; 786 kbuf = segment->kbuf;
796 else 787 else
797 buf = segment->buf; 788 buf = segment->buf;
798 ubytes = segment->bufsz; 789 ubytes = segment->bufsz;
799 mbytes = segment->memsz; 790 mbytes = segment->memsz;
800 maddr = segment->mem; 791 maddr = segment->mem;
801 while (mbytes) { 792 while (mbytes) {
802 struct page *page; 793 struct page *page;
803 char *ptr; 794 char *ptr;
804 size_t uchunk, mchunk; 795 size_t uchunk, mchunk;
805 796
806 page = boot_pfn_to_page(maddr !! 797 page = pfn_to_page(maddr >> PAGE_SHIFT);
807 if (!page) { 798 if (!page) {
808 result = -ENOMEM; 799 result = -ENOMEM;
809 goto out; 800 goto out;
810 } 801 }
811 arch_kexec_post_alloc_pages(pa !! 802 ptr = kmap(page);
812 ptr = kmap_local_page(page); <<
813 ptr += maddr & ~PAGE_MASK; 803 ptr += maddr & ~PAGE_MASK;
814 mchunk = min_t(size_t, mbytes, 804 mchunk = min_t(size_t, mbytes,
815 PAGE_SIZE - (m 805 PAGE_SIZE - (maddr & ~PAGE_MASK));
816 uchunk = min(ubytes, mchunk); 806 uchunk = min(ubytes, mchunk);
817 if (mchunk > uchunk) { 807 if (mchunk > uchunk) {
818 /* Zero the trailing p 808 /* Zero the trailing part of the page */
819 memset(ptr + uchunk, 0 809 memset(ptr + uchunk, 0, mchunk - uchunk);
820 } 810 }
821 811
822 if (uchunk) { !! 812 /* For file based kexec, source pages are in kernel memory */
823 /* For file based kexe !! 813 if (image->file_mode)
824 if (image->file_mode) !! 814 memcpy(ptr, kbuf, uchunk);
825 memcpy(ptr, kb !! 815 else
826 else !! 816 result = copy_from_user(ptr, buf, uchunk);
827 result = copy_ <<
828 ubytes -= uchunk; <<
829 if (image->file_mode) <<
830 kbuf += uchunk <<
831 else <<
832 buf += uchunk; <<
833 } <<
834 kexec_flush_icache_page(page); 817 kexec_flush_icache_page(page);
835 kunmap_local(ptr); !! 818 kunmap(page);
836 arch_kexec_pre_free_pages(page <<
837 if (result) { 819 if (result) {
838 result = -EFAULT; 820 result = -EFAULT;
839 goto out; 821 goto out;
840 } 822 }
>> 823 ubytes -= uchunk;
841 maddr += mchunk; 824 maddr += mchunk;
>> 825 if (image->file_mode)
>> 826 kbuf += mchunk;
>> 827 else
>> 828 buf += mchunk;
842 mbytes -= mchunk; 829 mbytes -= mchunk;
843 <<
844 cond_resched(); <<
845 } 830 }
846 out: 831 out:
847 return result; 832 return result;
848 } 833 }
849 #endif <<
850 834
851 int kimage_load_segment(struct kimage *image, 835 int kimage_load_segment(struct kimage *image,
852 struct kexec_s 836 struct kexec_segment *segment)
853 { 837 {
854 int result = -ENOMEM; 838 int result = -ENOMEM;
855 839
856 switch (image->type) { 840 switch (image->type) {
857 case KEXEC_TYPE_DEFAULT: 841 case KEXEC_TYPE_DEFAULT:
858 result = kimage_load_normal_se 842 result = kimage_load_normal_segment(image, segment);
859 break; 843 break;
860 #ifdef CONFIG_CRASH_DUMP <<
861 case KEXEC_TYPE_CRASH: 844 case KEXEC_TYPE_CRASH:
862 result = kimage_load_crash_seg 845 result = kimage_load_crash_segment(image, segment);
863 break; 846 break;
864 #endif <<
865 } 847 }
866 848
867 return result; 849 return result;
868 } 850 }
869 851
870 struct kexec_load_limit { !! 852 struct kimage *kexec_image;
871 /* Mutex protects the limit count. */ !! 853 struct kimage *kexec_crash_image;
872 struct mutex mutex; !! 854 int kexec_load_disabled;
873 int limit; <<
874 }; <<
875 855
876 static struct kexec_load_limit load_limit_rebo !! 856 void crash_kexec(struct pt_regs *regs)
877 .mutex = __MUTEX_INITIALIZER(load_limi !! 857 {
878 .limit = -1, !! 858 /* Take the kexec_mutex here to prevent sys_kexec_load
879 }; !! 859 * running on one cpu from replacing the crash kernel
>> 860 * we are using after a panic on a different cpu.
>> 861 *
>> 862 * If the crash kernel was not located in a fixed area
>> 863 * of memory the xchg(&kexec_crash_image) would be
>> 864 * sufficient. But since I reuse the memory...
>> 865 */
>> 866 if (mutex_trylock(&kexec_mutex)) {
>> 867 if (kexec_crash_image) {
>> 868 struct pt_regs fixed_regs;
>> 869
>> 870 crash_setup_regs(&fixed_regs, regs);
>> 871 crash_save_vmcoreinfo();
>> 872 machine_crash_shutdown(&fixed_regs);
>> 873 machine_kexec(kexec_crash_image);
>> 874 }
>> 875 mutex_unlock(&kexec_mutex);
>> 876 }
>> 877 }
880 878
881 static struct kexec_load_limit load_limit_pani !! 879 size_t crash_get_memory_size(void)
882 .mutex = __MUTEX_INITIALIZER(load_limi !! 880 {
883 .limit = -1, !! 881 size_t size = 0;
884 }; <<
885 882
886 struct kimage *kexec_image; !! 883 mutex_lock(&kexec_mutex);
887 struct kimage *kexec_crash_image; !! 884 if (crashk_res.end != crashk_res.start)
888 static int kexec_load_disabled; !! 885 size = resource_size(&crashk_res);
>> 886 mutex_unlock(&kexec_mutex);
>> 887 return size;
>> 888 }
>> 889
>> 890 void __weak crash_free_reserved_phys_range(unsigned long begin,
>> 891 unsigned long end)
>> 892 {
>> 893 unsigned long addr;
>> 894
>> 895 for (addr = begin; addr < end; addr += PAGE_SIZE)
>> 896 free_reserved_page(pfn_to_page(addr >> PAGE_SHIFT));
>> 897 }
>> 898
>> 899 int crash_shrink_memory(unsigned long new_size)
>> 900 {
>> 901 int ret = 0;
>> 902 unsigned long start, end;
>> 903 unsigned long old_size;
>> 904 struct resource *ram_res;
>> 905
>> 906 mutex_lock(&kexec_mutex);
>> 907
>> 908 if (kexec_crash_image) {
>> 909 ret = -ENOENT;
>> 910 goto unlock;
>> 911 }
>> 912 start = crashk_res.start;
>> 913 end = crashk_res.end;
>> 914 old_size = (end == 0) ? 0 : end - start + 1;
>> 915 if (new_size >= old_size) {
>> 916 ret = (new_size == old_size) ? 0 : -EINVAL;
>> 917 goto unlock;
>> 918 }
>> 919
>> 920 ram_res = kzalloc(sizeof(*ram_res), GFP_KERNEL);
>> 921 if (!ram_res) {
>> 922 ret = -ENOMEM;
>> 923 goto unlock;
>> 924 }
>> 925
>> 926 start = roundup(start, KEXEC_CRASH_MEM_ALIGN);
>> 927 end = roundup(start + new_size, KEXEC_CRASH_MEM_ALIGN);
>> 928
>> 929 crash_map_reserved_pages();
>> 930 crash_free_reserved_phys_range(end, crashk_res.end);
>> 931
>> 932 if ((start == end) && (crashk_res.parent != NULL))
>> 933 release_resource(&crashk_res);
>> 934
>> 935 ram_res->start = end;
>> 936 ram_res->end = crashk_res.end;
>> 937 ram_res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
>> 938 ram_res->name = "System RAM";
>> 939
>> 940 crashk_res.end = end - 1;
>> 941
>> 942 insert_resource(&iomem_resource, ram_res);
>> 943 crash_unmap_reserved_pages();
>> 944
>> 945 unlock:
>> 946 mutex_unlock(&kexec_mutex);
>> 947 return ret;
>> 948 }
>> 949
>> 950 static u32 *append_elf_note(u32 *buf, char *name, unsigned type, void *data,
>> 951 size_t data_len)
>> 952 {
>> 953 struct elf_note note;
>> 954
>> 955 note.n_namesz = strlen(name) + 1;
>> 956 note.n_descsz = data_len;
>> 957 note.n_type = type;
>> 958 memcpy(buf, ¬e, sizeof(note));
>> 959 buf += (sizeof(note) + 3)/4;
>> 960 memcpy(buf, name, note.n_namesz);
>> 961 buf += (note.n_namesz + 3)/4;
>> 962 memcpy(buf, data, note.n_descsz);
>> 963 buf += (note.n_descsz + 3)/4;
>> 964
>> 965 return buf;
>> 966 }
>> 967
>> 968 static void final_note(u32 *buf)
>> 969 {
>> 970 struct elf_note note;
>> 971
>> 972 note.n_namesz = 0;
>> 973 note.n_descsz = 0;
>> 974 note.n_type = 0;
>> 975 memcpy(buf, ¬e, sizeof(note));
>> 976 }
>> 977
>> 978 void crash_save_cpu(struct pt_regs *regs, int cpu)
>> 979 {
>> 980 struct elf_prstatus prstatus;
>> 981 u32 *buf;
>> 982
>> 983 if ((cpu < 0) || (cpu >= nr_cpu_ids))
>> 984 return;
>> 985
>> 986 /* Using ELF notes here is opportunistic.
>> 987 * I need a well defined structure format
>> 988 * for the data I pass, and I need tags
>> 989 * on the data to indicate what information I have
>> 990 * squirrelled away. ELF notes happen to provide
>> 991 * all of that, so there is no need to invent something new.
>> 992 */
>> 993 buf = (u32 *)per_cpu_ptr(crash_notes, cpu);
>> 994 if (!buf)
>> 995 return;
>> 996 memset(&prstatus, 0, sizeof(prstatus));
>> 997 prstatus.pr_pid = current->pid;
>> 998 elf_core_copy_kernel_regs(&prstatus.pr_reg, regs);
>> 999 buf = append_elf_note(buf, KEXEC_CORE_NOTE_NAME, NT_PRSTATUS,
>> 1000 &prstatus, sizeof(prstatus));
>> 1001 final_note(buf);
>> 1002 }
>> 1003
>> 1004 static int __init crash_notes_memory_init(void)
>> 1005 {
>> 1006 /* Allocate memory for saving cpu registers. */
>> 1007 size_t size, align;
>> 1008
>> 1009 /*
>> 1010 * crash_notes could be allocated across 2 vmalloc pages when percpu
>> 1011 * is vmalloc based . vmalloc doesn't guarantee 2 continuous vmalloc
>> 1012 * pages are also on 2 continuous physical pages. In this case the
>> 1013 * 2nd part of crash_notes in 2nd page could be lost since only the
>> 1014 * starting address and size of crash_notes are exported through sysfs.
>> 1015 * Here round up the size of crash_notes to the nearest power of two
>> 1016 * and pass it to __alloc_percpu as align value. This can make sure
>> 1017 * crash_notes is allocated inside one physical page.
>> 1018 */
>> 1019 size = sizeof(note_buf_t);
>> 1020 align = min(roundup_pow_of_two(sizeof(note_buf_t)), PAGE_SIZE);
>> 1021
>> 1022 /*
>> 1023 * Break compile if size is bigger than PAGE_SIZE since crash_notes
>> 1024 * definitely will be in 2 pages with that.
>> 1025 */
>> 1026 BUILD_BUG_ON(size > PAGE_SIZE);
>> 1027
>> 1028 crash_notes = __alloc_percpu(size, align);
>> 1029 if (!crash_notes) {
>> 1030 pr_warn("Memory allocation for saving cpu register states failed\n");
>> 1031 return -ENOMEM;
>> 1032 }
>> 1033 return 0;
>> 1034 }
>> 1035 subsys_initcall(crash_notes_memory_init);
>> 1036
>> 1037
>> 1038 /*
>> 1039 * parsing the "crashkernel" commandline
>> 1040 *
>> 1041 * this code is intended to be called from architecture specific code
>> 1042 */
>> 1043
>> 1044
>> 1045 /*
>> 1046 * This function parses command lines in the format
>> 1047 *
>> 1048 * crashkernel=ramsize-range:size[,...][@offset]
>> 1049 *
>> 1050 * The function returns 0 on success and -EINVAL on failure.
>> 1051 */
>> 1052 static int __init parse_crashkernel_mem(char *cmdline,
>> 1053 unsigned long long system_ram,
>> 1054 unsigned long long *crash_size,
>> 1055 unsigned long long *crash_base)
>> 1056 {
>> 1057 char *cur = cmdline, *tmp;
889 1058
890 #ifdef CONFIG_SYSCTL !! 1059 /* for each entry of the comma-separated list */
891 static int kexec_limit_handler(const struct ct !! 1060 do {
892 void *buffer, s !! 1061 unsigned long long start, end = ULLONG_MAX, size;
893 { <<
894 struct kexec_load_limit *limit = table <<
895 int val; <<
896 struct ctl_table tmp = { <<
897 .data = &val, <<
898 .maxlen = sizeof(val), <<
899 .mode = table->mode, <<
900 }; <<
901 int ret; <<
902 <<
903 if (write) { <<
904 ret = proc_dointvec(&tmp, writ <<
905 if (ret) <<
906 return ret; <<
907 1062
908 if (val < 0) !! 1063 /* get the start of the range */
>> 1064 start = memparse(cur, &tmp);
>> 1065 if (cur == tmp) {
>> 1066 pr_warn("crashkernel: Memory value expected\n");
>> 1067 return -EINVAL;
>> 1068 }
>> 1069 cur = tmp;
>> 1070 if (*cur != '-') {
>> 1071 pr_warn("crashkernel: '-' expected\n");
909 return -EINVAL; 1072 return -EINVAL;
>> 1073 }
>> 1074 cur++;
910 1075
911 mutex_lock(&limit->mutex); !! 1076 /* if no ':' is here, than we read the end */
912 if (limit->limit != -1 && val !! 1077 if (*cur != ':') {
913 ret = -EINVAL; !! 1078 end = memparse(cur, &tmp);
914 else !! 1079 if (cur == tmp) {
915 limit->limit = val; !! 1080 pr_warn("crashkernel: Memory value expected\n");
916 mutex_unlock(&limit->mutex); !! 1081 return -EINVAL;
>> 1082 }
>> 1083 cur = tmp;
>> 1084 if (end <= start) {
>> 1085 pr_warn("crashkernel: end <= start\n");
>> 1086 return -EINVAL;
>> 1087 }
>> 1088 }
>> 1089
>> 1090 if (*cur != ':') {
>> 1091 pr_warn("crashkernel: ':' expected\n");
>> 1092 return -EINVAL;
>> 1093 }
>> 1094 cur++;
>> 1095
>> 1096 size = memparse(cur, &tmp);
>> 1097 if (cur == tmp) {
>> 1098 pr_warn("Memory value expected\n");
>> 1099 return -EINVAL;
>> 1100 }
>> 1101 cur = tmp;
>> 1102 if (size >= system_ram) {
>> 1103 pr_warn("crashkernel: invalid size\n");
>> 1104 return -EINVAL;
>> 1105 }
>> 1106
>> 1107 /* match ? */
>> 1108 if (system_ram >= start && system_ram < end) {
>> 1109 *crash_size = size;
>> 1110 break;
>> 1111 }
>> 1112 } while (*cur++ == ',');
>> 1113
>> 1114 if (*crash_size > 0) {
>> 1115 while (*cur && *cur != ' ' && *cur != '@')
>> 1116 cur++;
>> 1117 if (*cur == '@') {
>> 1118 cur++;
>> 1119 *crash_base = memparse(cur, &tmp);
>> 1120 if (cur == tmp) {
>> 1121 pr_warn("Memory value expected after '@'\n");
>> 1122 return -EINVAL;
>> 1123 }
>> 1124 }
>> 1125 }
>> 1126
>> 1127 return 0;
>> 1128 }
917 1129
918 return ret; !! 1130 /*
>> 1131 * That function parses "simple" (old) crashkernel command lines like
>> 1132 *
>> 1133 * crashkernel=size[@offset]
>> 1134 *
>> 1135 * It returns 0 on success and -EINVAL on failure.
>> 1136 */
>> 1137 static int __init parse_crashkernel_simple(char *cmdline,
>> 1138 unsigned long long *crash_size,
>> 1139 unsigned long long *crash_base)
>> 1140 {
>> 1141 char *cur = cmdline;
>> 1142
>> 1143 *crash_size = memparse(cmdline, &cur);
>> 1144 if (cmdline == cur) {
>> 1145 pr_warn("crashkernel: memory value expected\n");
>> 1146 return -EINVAL;
919 } 1147 }
920 1148
921 mutex_lock(&limit->mutex); !! 1149 if (*cur == '@')
922 val = limit->limit; !! 1150 *crash_base = memparse(cur+1, &cur);
923 mutex_unlock(&limit->mutex); !! 1151 else if (*cur != ' ' && *cur != '\0') {
>> 1152 pr_warn("crashkernel: unrecognized char: %c\n", *cur);
>> 1153 return -EINVAL;
>> 1154 }
924 1155
925 return proc_dointvec(&tmp, write, buff !! 1156 return 0;
926 } 1157 }
927 1158
928 static struct ctl_table kexec_core_sysctls[] = !! 1159 #define SUFFIX_HIGH 0
929 { !! 1160 #define SUFFIX_LOW 1
930 .procname = "kexec_load_ !! 1161 #define SUFFIX_NULL 2
931 .data = &kexec_load_ !! 1162 static __initdata char *suffix_tbl[] = {
932 .maxlen = sizeof(int), !! 1163 [SUFFIX_HIGH] = ",high",
933 .mode = 0644, !! 1164 [SUFFIX_LOW] = ",low",
934 /* only handle a transition fr !! 1165 [SUFFIX_NULL] = NULL,
935 .proc_handler = proc_dointve <<
936 .extra1 = SYSCTL_ONE, <<
937 .extra2 = SYSCTL_ONE, <<
938 }, <<
939 { <<
940 .procname = "kexec_load_ <<
941 .data = &load_limit_ <<
942 .mode = 0644, <<
943 .proc_handler = kexec_limit_ <<
944 }, <<
945 { <<
946 .procname = "kexec_load_ <<
947 .data = &load_limit_ <<
948 .mode = 0644, <<
949 .proc_handler = kexec_limit_ <<
950 }, <<
951 }; 1166 };
952 1167
953 static int __init kexec_core_sysctl_init(void) !! 1168 /*
954 { !! 1169 * That function parses "suffix" crashkernel command lines like
955 register_sysctl_init("kernel", kexec_c !! 1170 *
>> 1171 * crashkernel=size,[high|low]
>> 1172 *
>> 1173 * It returns 0 on success and -EINVAL on failure.
>> 1174 */
>> 1175 static int __init parse_crashkernel_suffix(char *cmdline,
>> 1176 unsigned long long *crash_size,
>> 1177 const char *suffix)
>> 1178 {
>> 1179 char *cur = cmdline;
>> 1180
>> 1181 *crash_size = memparse(cmdline, &cur);
>> 1182 if (cmdline == cur) {
>> 1183 pr_warn("crashkernel: memory value expected\n");
>> 1184 return -EINVAL;
>> 1185 }
>> 1186
>> 1187 /* check with suffix */
>> 1188 if (strncmp(cur, suffix, strlen(suffix))) {
>> 1189 pr_warn("crashkernel: unrecognized char: %c\n", *cur);
>> 1190 return -EINVAL;
>> 1191 }
>> 1192 cur += strlen(suffix);
>> 1193 if (*cur != ' ' && *cur != '\0') {
>> 1194 pr_warn("crashkernel: unrecognized char: %c\n", *cur);
>> 1195 return -EINVAL;
>> 1196 }
>> 1197
956 return 0; 1198 return 0;
957 } 1199 }
958 late_initcall(kexec_core_sysctl_init); <<
959 #endif <<
960 1200
961 bool kexec_load_permitted(int kexec_image_type !! 1201 static __init char *get_last_crashkernel(char *cmdline,
>> 1202 const char *name,
>> 1203 const char *suffix)
>> 1204 {
>> 1205 char *p = cmdline, *ck_cmdline = NULL;
>> 1206
>> 1207 /* find crashkernel and use the last one if there are more */
>> 1208 p = strstr(p, name);
>> 1209 while (p) {
>> 1210 char *end_p = strchr(p, ' ');
>> 1211 char *q;
>> 1212
>> 1213 if (!end_p)
>> 1214 end_p = p + strlen(p);
>> 1215
>> 1216 if (!suffix) {
>> 1217 int i;
>> 1218
>> 1219 /* skip the one with any known suffix */
>> 1220 for (i = 0; suffix_tbl[i]; i++) {
>> 1221 q = end_p - strlen(suffix_tbl[i]);
>> 1222 if (!strncmp(q, suffix_tbl[i],
>> 1223 strlen(suffix_tbl[i])))
>> 1224 goto next;
>> 1225 }
>> 1226 ck_cmdline = p;
>> 1227 } else {
>> 1228 q = end_p - strlen(suffix);
>> 1229 if (!strncmp(q, suffix, strlen(suffix)))
>> 1230 ck_cmdline = p;
>> 1231 }
>> 1232 next:
>> 1233 p = strstr(p+1, name);
>> 1234 }
>> 1235
>> 1236 if (!ck_cmdline)
>> 1237 return NULL;
>> 1238
>> 1239 return ck_cmdline;
>> 1240 }
>> 1241
>> 1242 static int __init __parse_crashkernel(char *cmdline,
>> 1243 unsigned long long system_ram,
>> 1244 unsigned long long *crash_size,
>> 1245 unsigned long long *crash_base,
>> 1246 const char *name,
>> 1247 const char *suffix)
962 { 1248 {
963 struct kexec_load_limit *limit; !! 1249 char *first_colon, *first_space;
>> 1250 char *ck_cmdline;
>> 1251
>> 1252 BUG_ON(!crash_size || !crash_base);
>> 1253 *crash_size = 0;
>> 1254 *crash_base = 0;
964 1255
>> 1256 ck_cmdline = get_last_crashkernel(cmdline, name, suffix);
>> 1257
>> 1258 if (!ck_cmdline)
>> 1259 return -EINVAL;
>> 1260
>> 1261 ck_cmdline += strlen(name);
>> 1262
>> 1263 if (suffix)
>> 1264 return parse_crashkernel_suffix(ck_cmdline, crash_size,
>> 1265 suffix);
965 /* 1266 /*
966 * Only the superuser can use the kexe !! 1267 * if the commandline contains a ':', then that's the extended
967 * been disabled. !! 1268 * syntax -- if not, it must be the classic syntax
968 */ 1269 */
969 if (!capable(CAP_SYS_BOOT) || kexec_lo !! 1270 first_colon = strchr(ck_cmdline, ':');
970 return false; !! 1271 first_space = strchr(ck_cmdline, ' ');
>> 1272 if (first_colon && (!first_space || first_colon < first_space))
>> 1273 return parse_crashkernel_mem(ck_cmdline, system_ram,
>> 1274 crash_size, crash_base);
>> 1275
>> 1276 return parse_crashkernel_simple(ck_cmdline, crash_size, crash_base);
>> 1277 }
>> 1278
>> 1279 /*
>> 1280 * That function is the entry point for command line parsing and should be
>> 1281 * called from the arch-specific code.
>> 1282 */
>> 1283 int __init parse_crashkernel(char *cmdline,
>> 1284 unsigned long long system_ram,
>> 1285 unsigned long long *crash_size,
>> 1286 unsigned long long *crash_base)
>> 1287 {
>> 1288 return __parse_crashkernel(cmdline, system_ram, crash_size, crash_base,
>> 1289 "crashkernel=", NULL);
>> 1290 }
>> 1291
>> 1292 int __init parse_crashkernel_high(char *cmdline,
>> 1293 unsigned long long system_ram,
>> 1294 unsigned long long *crash_size,
>> 1295 unsigned long long *crash_base)
>> 1296 {
>> 1297 return __parse_crashkernel(cmdline, system_ram, crash_size, crash_base,
>> 1298 "crashkernel=", suffix_tbl[SUFFIX_HIGH]);
>> 1299 }
>> 1300
>> 1301 int __init parse_crashkernel_low(char *cmdline,
>> 1302 unsigned long long system_ram,
>> 1303 unsigned long long *crash_size,
>> 1304 unsigned long long *crash_base)
>> 1305 {
>> 1306 return __parse_crashkernel(cmdline, system_ram, crash_size, crash_base,
>> 1307 "crashkernel=", suffix_tbl[SUFFIX_LOW]);
>> 1308 }
>> 1309
>> 1310 static void update_vmcoreinfo_note(void)
>> 1311 {
>> 1312 u32 *buf = vmcoreinfo_note;
>> 1313
>> 1314 if (!vmcoreinfo_size)
>> 1315 return;
>> 1316 buf = append_elf_note(buf, VMCOREINFO_NOTE_NAME, 0, vmcoreinfo_data,
>> 1317 vmcoreinfo_size);
>> 1318 final_note(buf);
>> 1319 }
>> 1320
>> 1321 void crash_save_vmcoreinfo(void)
>> 1322 {
>> 1323 vmcoreinfo_append_str("CRASHTIME=%ld\n", get_seconds());
>> 1324 update_vmcoreinfo_note();
>> 1325 }
>> 1326
>> 1327 void vmcoreinfo_append_str(const char *fmt, ...)
>> 1328 {
>> 1329 va_list args;
>> 1330 char buf[0x50];
>> 1331 size_t r;
>> 1332
>> 1333 va_start(args, fmt);
>> 1334 r = vscnprintf(buf, sizeof(buf), fmt, args);
>> 1335 va_end(args);
>> 1336
>> 1337 r = min(r, vmcoreinfo_max_size - vmcoreinfo_size);
>> 1338
>> 1339 memcpy(&vmcoreinfo_data[vmcoreinfo_size], buf, r);
>> 1340
>> 1341 vmcoreinfo_size += r;
>> 1342 }
>> 1343
>> 1344 /*
>> 1345 * provide an empty default implementation here -- architecture
>> 1346 * code may override this
>> 1347 */
>> 1348 void __weak arch_crash_save_vmcoreinfo(void)
>> 1349 {}
>> 1350
>> 1351 unsigned long __weak paddr_vmcoreinfo_note(void)
>> 1352 {
>> 1353 return __pa((unsigned long)(char *)&vmcoreinfo_note);
>> 1354 }
>> 1355
>> 1356 static int __init crash_save_vmcoreinfo_init(void)
>> 1357 {
>> 1358 VMCOREINFO_OSRELEASE(init_uts_ns.name.release);
>> 1359 VMCOREINFO_PAGESIZE(PAGE_SIZE);
971 1360
972 /* Check limit counter and decrease it !! 1361 VMCOREINFO_SYMBOL(init_uts_ns);
973 limit = (kexec_image_type == KEXEC_TYP !! 1362 VMCOREINFO_SYMBOL(node_online_map);
974 &load_limit_panic : &load_limi !! 1363 #ifdef CONFIG_MMU
975 mutex_lock(&limit->mutex); !! 1364 VMCOREINFO_SYMBOL(swapper_pg_dir);
976 if (!limit->limit) { !! 1365 #endif
977 mutex_unlock(&limit->mutex); !! 1366 VMCOREINFO_SYMBOL(_stext);
978 return false; !! 1367 VMCOREINFO_SYMBOL(vmap_area_list);
979 } !! 1368
980 if (limit->limit != -1) !! 1369 #ifndef CONFIG_NEED_MULTIPLE_NODES
981 limit->limit--; !! 1370 VMCOREINFO_SYMBOL(mem_map);
982 mutex_unlock(&limit->mutex); !! 1371 VMCOREINFO_SYMBOL(contig_page_data);
>> 1372 #endif
>> 1373 #ifdef CONFIG_SPARSEMEM
>> 1374 VMCOREINFO_SYMBOL(mem_section);
>> 1375 VMCOREINFO_LENGTH(mem_section, NR_SECTION_ROOTS);
>> 1376 VMCOREINFO_STRUCT_SIZE(mem_section);
>> 1377 VMCOREINFO_OFFSET(mem_section, section_mem_map);
>> 1378 #endif
>> 1379 VMCOREINFO_STRUCT_SIZE(page);
>> 1380 VMCOREINFO_STRUCT_SIZE(pglist_data);
>> 1381 VMCOREINFO_STRUCT_SIZE(zone);
>> 1382 VMCOREINFO_STRUCT_SIZE(free_area);
>> 1383 VMCOREINFO_STRUCT_SIZE(list_head);
>> 1384 VMCOREINFO_SIZE(nodemask_t);
>> 1385 VMCOREINFO_OFFSET(page, flags);
>> 1386 VMCOREINFO_OFFSET(page, _count);
>> 1387 VMCOREINFO_OFFSET(page, mapping);
>> 1388 VMCOREINFO_OFFSET(page, lru);
>> 1389 VMCOREINFO_OFFSET(page, _mapcount);
>> 1390 VMCOREINFO_OFFSET(page, private);
>> 1391 VMCOREINFO_OFFSET(pglist_data, node_zones);
>> 1392 VMCOREINFO_OFFSET(pglist_data, nr_zones);
>> 1393 #ifdef CONFIG_FLAT_NODE_MEM_MAP
>> 1394 VMCOREINFO_OFFSET(pglist_data, node_mem_map);
>> 1395 #endif
>> 1396 VMCOREINFO_OFFSET(pglist_data, node_start_pfn);
>> 1397 VMCOREINFO_OFFSET(pglist_data, node_spanned_pages);
>> 1398 VMCOREINFO_OFFSET(pglist_data, node_id);
>> 1399 VMCOREINFO_OFFSET(zone, free_area);
>> 1400 VMCOREINFO_OFFSET(zone, vm_stat);
>> 1401 VMCOREINFO_OFFSET(zone, spanned_pages);
>> 1402 VMCOREINFO_OFFSET(free_area, free_list);
>> 1403 VMCOREINFO_OFFSET(list_head, next);
>> 1404 VMCOREINFO_OFFSET(list_head, prev);
>> 1405 VMCOREINFO_OFFSET(vmap_area, va_start);
>> 1406 VMCOREINFO_OFFSET(vmap_area, list);
>> 1407 VMCOREINFO_LENGTH(zone.free_area, MAX_ORDER);
>> 1408 log_buf_kexec_setup();
>> 1409 VMCOREINFO_LENGTH(free_area.free_list, MIGRATE_TYPES);
>> 1410 VMCOREINFO_NUMBER(NR_FREE_PAGES);
>> 1411 VMCOREINFO_NUMBER(PG_lru);
>> 1412 VMCOREINFO_NUMBER(PG_private);
>> 1413 VMCOREINFO_NUMBER(PG_swapcache);
>> 1414 VMCOREINFO_NUMBER(PG_slab);
>> 1415 #ifdef CONFIG_MEMORY_FAILURE
>> 1416 VMCOREINFO_NUMBER(PG_hwpoison);
>> 1417 #endif
>> 1418 VMCOREINFO_NUMBER(PG_head_mask);
>> 1419 VMCOREINFO_NUMBER(PAGE_BUDDY_MAPCOUNT_VALUE);
>> 1420 #ifdef CONFIG_X86
>> 1421 VMCOREINFO_NUMBER(KERNEL_IMAGE_SIZE);
>> 1422 #endif
>> 1423 #ifdef CONFIG_HUGETLBFS
>> 1424 VMCOREINFO_SYMBOL(free_huge_page);
>> 1425 #endif
983 1426
984 return true; !! 1427 arch_crash_save_vmcoreinfo();
>> 1428 update_vmcoreinfo_note();
>> 1429
>> 1430 return 0;
985 } 1431 }
986 1432
>> 1433 subsys_initcall(crash_save_vmcoreinfo_init);
>> 1434
987 /* 1435 /*
988 * Move into place and start executing a prelo 1436 * Move into place and start executing a preloaded standalone
989 * executable. If nothing was preloaded retur 1437 * executable. If nothing was preloaded return an error.
990 */ 1438 */
991 int kernel_kexec(void) 1439 int kernel_kexec(void)
992 { 1440 {
993 int error = 0; 1441 int error = 0;
994 1442
995 if (!kexec_trylock()) !! 1443 if (!mutex_trylock(&kexec_mutex))
996 return -EBUSY; 1444 return -EBUSY;
997 if (!kexec_image) { 1445 if (!kexec_image) {
998 error = -EINVAL; 1446 error = -EINVAL;
999 goto Unlock; 1447 goto Unlock;
1000 } 1448 }
1001 1449
1002 #ifdef CONFIG_KEXEC_JUMP 1450 #ifdef CONFIG_KEXEC_JUMP
1003 if (kexec_image->preserve_context) { 1451 if (kexec_image->preserve_context) {
>> 1452 lock_system_sleep();
1004 pm_prepare_console(); 1453 pm_prepare_console();
1005 error = freeze_processes(); 1454 error = freeze_processes();
1006 if (error) { 1455 if (error) {
1007 error = -EBUSY; 1456 error = -EBUSY;
1008 goto Restore_console; 1457 goto Restore_console;
1009 } 1458 }
1010 suspend_console(); 1459 suspend_console();
1011 error = dpm_suspend_start(PMS 1460 error = dpm_suspend_start(PMSG_FREEZE);
1012 if (error) 1461 if (error)
1013 goto Resume_console; 1462 goto Resume_console;
1014 /* At this point, dpm_suspend 1463 /* At this point, dpm_suspend_start() has been called,
1015 * but *not* dpm_suspend_end( 1464 * but *not* dpm_suspend_end(). We *must* call
1016 * dpm_suspend_end() now. Ot 1465 * dpm_suspend_end() now. Otherwise, drivers for
1017 * some devices (e.g. interru 1466 * some devices (e.g. interrupt controllers) become
1018 * desynchronized with the ac 1467 * desynchronized with the actual state of the
1019 * hardware at resume time, a 1468 * hardware at resume time, and evil weirdness ensues.
1020 */ 1469 */
1021 error = dpm_suspend_end(PMSG_ 1470 error = dpm_suspend_end(PMSG_FREEZE);
1022 if (error) 1471 if (error)
1023 goto Resume_devices; 1472 goto Resume_devices;
1024 error = suspend_disable_secon !! 1473 error = disable_nonboot_cpus();
1025 if (error) 1474 if (error)
1026 goto Enable_cpus; 1475 goto Enable_cpus;
1027 local_irq_disable(); 1476 local_irq_disable();
1028 error = syscore_suspend(); 1477 error = syscore_suspend();
1029 if (error) 1478 if (error)
1030 goto Enable_irqs; 1479 goto Enable_irqs;
1031 } else 1480 } else
1032 #endif 1481 #endif
1033 { 1482 {
1034 kexec_in_progress = true; 1483 kexec_in_progress = true;
1035 kernel_restart_prepare("kexec !! 1484 kernel_restart_prepare(NULL);
1036 migrate_to_reboot_cpu(); 1485 migrate_to_reboot_cpu();
1037 syscore_shutdown(); <<
1038 1486
1039 /* 1487 /*
1040 * migrate_to_reboot_cpu() di 1488 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
1041 * no further code needs to u 1489 * no further code needs to use CPU hotplug (which is true in
1042 * the reboot case). However, 1490 * the reboot case). However, the kexec path depends on using
1043 * CPU hotplug again; so re-e 1491 * CPU hotplug again; so re-enable it here.
1044 */ 1492 */
1045 cpu_hotplug_enable(); 1493 cpu_hotplug_enable();
1046 pr_notice("Starting new kerne !! 1494 pr_emerg("Starting new kernel\n");
1047 machine_shutdown(); 1495 machine_shutdown();
1048 } 1496 }
1049 1497
1050 kmsg_dump(KMSG_DUMP_SHUTDOWN); <<
1051 machine_kexec(kexec_image); 1498 machine_kexec(kexec_image);
1052 1499
1053 #ifdef CONFIG_KEXEC_JUMP 1500 #ifdef CONFIG_KEXEC_JUMP
1054 if (kexec_image->preserve_context) { 1501 if (kexec_image->preserve_context) {
1055 syscore_resume(); 1502 syscore_resume();
1056 Enable_irqs: 1503 Enable_irqs:
1057 local_irq_enable(); 1504 local_irq_enable();
1058 Enable_cpus: 1505 Enable_cpus:
1059 suspend_enable_secondary_cpus !! 1506 enable_nonboot_cpus();
1060 dpm_resume_start(PMSG_RESTORE 1507 dpm_resume_start(PMSG_RESTORE);
1061 Resume_devices: 1508 Resume_devices:
1062 dpm_resume_end(PMSG_RESTORE); 1509 dpm_resume_end(PMSG_RESTORE);
1063 Resume_console: 1510 Resume_console:
1064 resume_console(); 1511 resume_console();
1065 thaw_processes(); 1512 thaw_processes();
1066 Restore_console: 1513 Restore_console:
1067 pm_restore_console(); 1514 pm_restore_console();
>> 1515 unlock_system_sleep();
1068 } 1516 }
1069 #endif 1517 #endif
1070 1518
1071 Unlock: 1519 Unlock:
1072 kexec_unlock(); !! 1520 mutex_unlock(&kexec_mutex);
1073 return error; 1521 return error;
1074 } 1522 }
>> 1523
>> 1524 /*
>> 1525 * Add and remove page tables for crashkernel memory
>> 1526 *
>> 1527 * Provide an empty default implementation here -- architecture
>> 1528 * code may override this
>> 1529 */
>> 1530 void __weak crash_map_reserved_pages(void)
>> 1531 {}
>> 1532
>> 1533 void __weak crash_unmap_reserved_pages(void)
>> 1534 {}
1075 1535
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.