1 // SPDX-License-Identifier: GPL-2.0-only <<
2 /* 1 /*
3 * fs/proc/vmcore.c Interface for accessi 2 * fs/proc/vmcore.c Interface for accessing the crash
4 * dump from the 3 * dump from the system's previous life.
5 * Heavily borrowed from fs/proc/kcore.c 4 * Heavily borrowed from fs/proc/kcore.c
6 * Created by: Hariprasad Nellitheertha ( 5 * Created by: Hariprasad Nellitheertha (hari@in.ibm.com)
7 * Copyright (C) IBM Corporation, 2004. A 6 * Copyright (C) IBM Corporation, 2004. All rights reserved
8 * 7 *
9 */ 8 */
10 9
11 #include <linux/mm.h> 10 #include <linux/mm.h>
12 #include <linux/kcore.h> 11 #include <linux/kcore.h>
13 #include <linux/user.h> 12 #include <linux/user.h>
14 #include <linux/elf.h> 13 #include <linux/elf.h>
15 #include <linux/elfcore.h> 14 #include <linux/elfcore.h>
16 #include <linux/export.h> 15 #include <linux/export.h>
17 #include <linux/slab.h> 16 #include <linux/slab.h>
18 #include <linux/highmem.h> 17 #include <linux/highmem.h>
19 #include <linux/printk.h> 18 #include <linux/printk.h>
20 #include <linux/memblock.h> !! 19 #include <linux/bootmem.h>
21 #include <linux/init.h> 20 #include <linux/init.h>
22 #include <linux/crash_dump.h> 21 #include <linux/crash_dump.h>
23 #include <linux/list.h> 22 #include <linux/list.h>
24 #include <linux/moduleparam.h> <<
25 #include <linux/mutex.h> <<
26 #include <linux/vmalloc.h> 23 #include <linux/vmalloc.h>
27 #include <linux/pagemap.h> 24 #include <linux/pagemap.h>
28 #include <linux/uio.h> !! 25 #include <linux/uaccess.h>
29 #include <linux/cc_platform.h> <<
30 #include <asm/io.h> 26 #include <asm/io.h>
31 #include "internal.h" 27 #include "internal.h"
32 28
33 /* List representing chunks of contiguous memo 29 /* List representing chunks of contiguous memory areas and their offsets in
34 * vmcore file. 30 * vmcore file.
35 */ 31 */
36 static LIST_HEAD(vmcore_list); 32 static LIST_HEAD(vmcore_list);
37 33
38 /* Stores the pointer to the buffer containing 34 /* Stores the pointer to the buffer containing kernel elf core headers. */
39 static char *elfcorebuf; 35 static char *elfcorebuf;
40 static size_t elfcorebuf_sz; 36 static size_t elfcorebuf_sz;
41 static size_t elfcorebuf_sz_orig; 37 static size_t elfcorebuf_sz_orig;
42 38
43 static char *elfnotes_buf; 39 static char *elfnotes_buf;
44 static size_t elfnotes_sz; 40 static size_t elfnotes_sz;
45 /* Size of all notes minus the device dump not <<
46 static size_t elfnotes_orig_sz; <<
47 41
48 /* Total size of vmcore file. */ 42 /* Total size of vmcore file. */
49 static u64 vmcore_size; 43 static u64 vmcore_size;
50 44
51 static struct proc_dir_entry *proc_vmcore; 45 static struct proc_dir_entry *proc_vmcore;
52 46
53 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP !! 47 /*
54 /* Device Dump list and mutex to synchronize a !! 48 * Returns > 0 for RAM pages, 0 for non-RAM pages, < 0 on error
55 static LIST_HEAD(vmcoredd_list); !! 49 * The called function has to take care of module refcounting.
56 static DEFINE_MUTEX(vmcoredd_mutex); !! 50 */
57 !! 51 static int (*oldmem_pfn_is_ram)(unsigned long pfn);
58 static bool vmcoredd_disabled; <<
59 core_param(novmcoredd, vmcoredd_disabled, bool <<
60 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
61 <<
62 /* Device Dump Size */ <<
63 static size_t vmcoredd_orig_sz; <<
64 <<
65 static DEFINE_SPINLOCK(vmcore_cb_lock); <<
66 DEFINE_STATIC_SRCU(vmcore_cb_srcu); <<
67 /* List of registered vmcore callbacks. */ <<
68 static LIST_HEAD(vmcore_cb_list); <<
69 /* Whether the vmcore has been opened once. */ <<
70 static bool vmcore_opened; <<
71 <<
72 void register_vmcore_cb(struct vmcore_cb *cb) <<
73 { <<
74 INIT_LIST_HEAD(&cb->next); <<
75 spin_lock(&vmcore_cb_lock); <<
76 list_add_tail(&cb->next, &vmcore_cb_li <<
77 /* <<
78 * Registering a vmcore callback after <<
79 * very unusual (e.g., manual driver l <<
80 */ <<
81 if (vmcore_opened) <<
82 pr_warn_once("Unexpected vmcor <<
83 spin_unlock(&vmcore_cb_lock); <<
84 } <<
85 EXPORT_SYMBOL_GPL(register_vmcore_cb); <<
86 52
87 void unregister_vmcore_cb(struct vmcore_cb *cb !! 53 int register_oldmem_pfn_is_ram(int (*fn)(unsigned long pfn))
88 { 54 {
89 spin_lock(&vmcore_cb_lock); !! 55 if (oldmem_pfn_is_ram)
90 list_del_rcu(&cb->next); !! 56 return -EBUSY;
91 /* !! 57 oldmem_pfn_is_ram = fn;
92 * Unregistering a vmcore callback aft !! 58 return 0;
93 * very unusual (e.g., forced driver r <<
94 * unregistering. <<
95 */ <<
96 if (vmcore_opened) <<
97 pr_warn_once("Unexpected vmcor <<
98 spin_unlock(&vmcore_cb_lock); <<
99 <<
100 synchronize_srcu(&vmcore_cb_srcu); <<
101 } 59 }
102 EXPORT_SYMBOL_GPL(unregister_vmcore_cb); !! 60 EXPORT_SYMBOL_GPL(register_oldmem_pfn_is_ram);
103 61
104 static bool pfn_is_ram(unsigned long pfn) !! 62 void unregister_oldmem_pfn_is_ram(void)
105 { 63 {
106 struct vmcore_cb *cb; !! 64 oldmem_pfn_is_ram = NULL;
107 bool ret = true; !! 65 wmb();
108 <<
109 list_for_each_entry_srcu(cb, &vmcore_c <<
110 srcu_read_loc <<
111 if (unlikely(!cb->pfn_is_ram)) <<
112 continue; <<
113 ret = cb->pfn_is_ram(cb, pfn); <<
114 if (!ret) <<
115 break; <<
116 } <<
117 <<
118 return ret; <<
119 } 66 }
>> 67 EXPORT_SYMBOL_GPL(unregister_oldmem_pfn_is_ram);
120 68
121 static int open_vmcore(struct inode *inode, st !! 69 static int pfn_is_ram(unsigned long pfn)
122 { 70 {
123 spin_lock(&vmcore_cb_lock); !! 71 int (*fn)(unsigned long pfn);
124 vmcore_opened = true; !! 72 /* pfn is ram unless fn() checks pagetype */
125 spin_unlock(&vmcore_cb_lock); !! 73 int ret = 1;
126 74
127 return 0; !! 75 /*
>> 76 * Ask hypervisor if the pfn is really ram.
>> 77 * A ballooned page contains no data and reading from such a page
>> 78 * will cause high load in the hypervisor.
>> 79 */
>> 80 fn = oldmem_pfn_is_ram;
>> 81 if (fn)
>> 82 ret = fn(pfn);
>> 83
>> 84 return ret;
128 } 85 }
129 86
130 /* Reads a page from the oldmem device from gi 87 /* Reads a page from the oldmem device from given offset. */
131 ssize_t read_from_oldmem(struct iov_iter *iter !! 88 static ssize_t read_from_oldmem(char *buf, size_t count,
132 u64 *ppos, bool encry !! 89 u64 *ppos, int userbuf)
133 { 90 {
134 unsigned long pfn, offset; 91 unsigned long pfn, offset;
135 ssize_t nr_bytes; !! 92 size_t nr_bytes;
136 ssize_t read = 0, tmp; 93 ssize_t read = 0, tmp;
137 int idx; <<
138 94
139 if (!count) 95 if (!count)
140 return 0; 96 return 0;
141 97
142 offset = (unsigned long)(*ppos % PAGE_ 98 offset = (unsigned long)(*ppos % PAGE_SIZE);
143 pfn = (unsigned long)(*ppos / PAGE_SIZ 99 pfn = (unsigned long)(*ppos / PAGE_SIZE);
144 100
145 idx = srcu_read_lock(&vmcore_cb_srcu); <<
146 do { 101 do {
147 if (count > (PAGE_SIZE - offse 102 if (count > (PAGE_SIZE - offset))
148 nr_bytes = PAGE_SIZE - 103 nr_bytes = PAGE_SIZE - offset;
149 else 104 else
150 nr_bytes = count; 105 nr_bytes = count;
151 106
152 /* If pfn is not ram, return z 107 /* If pfn is not ram, return zeros for sparse dump files */
153 if (!pfn_is_ram(pfn)) { !! 108 if (pfn_is_ram(pfn) == 0)
154 tmp = iov_iter_zero(nr !! 109 memset(buf, 0, nr_bytes);
155 } else { !! 110 else {
156 if (encrypted) !! 111 tmp = copy_oldmem_page(pfn, buf, nr_bytes,
157 tmp = copy_old !! 112 offset, userbuf);
158 !! 113 if (tmp < 0)
159 !! 114 return tmp;
160 else <<
161 tmp = copy_old <<
162 <<
163 } <<
164 if (tmp < nr_bytes) { <<
165 srcu_read_unlock(&vmco <<
166 return -EFAULT; <<
167 } 115 }
168 <<
169 *ppos += nr_bytes; 116 *ppos += nr_bytes;
170 count -= nr_bytes; 117 count -= nr_bytes;
>> 118 buf += nr_bytes;
171 read += nr_bytes; 119 read += nr_bytes;
172 ++pfn; 120 ++pfn;
173 offset = 0; 121 offset = 0;
174 } while (count); 122 } while (count);
175 srcu_read_unlock(&vmcore_cb_srcu, idx) <<
176 123
177 return read; 124 return read;
178 } 125 }
179 126
180 /* 127 /*
181 * Architectures may override this function to 128 * Architectures may override this function to allocate ELF header in 2nd kernel
182 */ 129 */
183 int __weak elfcorehdr_alloc(unsigned long long 130 int __weak elfcorehdr_alloc(unsigned long long *addr, unsigned long long *size)
184 { 131 {
185 return 0; 132 return 0;
186 } 133 }
187 134
188 /* 135 /*
189 * Architectures may override this function to 136 * Architectures may override this function to free header
190 */ 137 */
191 void __weak elfcorehdr_free(unsigned long long 138 void __weak elfcorehdr_free(unsigned long long addr)
192 {} 139 {}
193 140
194 /* 141 /*
195 * Architectures may override this function to 142 * Architectures may override this function to read from ELF header
196 */ 143 */
197 ssize_t __weak elfcorehdr_read(char *buf, size 144 ssize_t __weak elfcorehdr_read(char *buf, size_t count, u64 *ppos)
198 { 145 {
199 struct kvec kvec = { .iov_base = buf, !! 146 return read_from_oldmem(buf, count, ppos, 0);
200 struct iov_iter iter; <<
201 <<
202 iov_iter_kvec(&iter, ITER_DEST, &kvec, <<
203 <<
204 return read_from_oldmem(&iter, count, <<
205 } 147 }
206 148
207 /* 149 /*
208 * Architectures may override this function to 150 * Architectures may override this function to read from notes sections
209 */ 151 */
210 ssize_t __weak elfcorehdr_read_notes(char *buf 152 ssize_t __weak elfcorehdr_read_notes(char *buf, size_t count, u64 *ppos)
211 { 153 {
212 struct kvec kvec = { .iov_base = buf, !! 154 return read_from_oldmem(buf, count, ppos, 0);
213 struct iov_iter iter; <<
214 <<
215 iov_iter_kvec(&iter, ITER_DEST, &kvec, <<
216 <<
217 return read_from_oldmem(&iter, count, <<
218 cc_platform_has(CC_ATT <<
219 } 155 }
220 156
221 /* 157 /*
222 * Architectures may override this function to 158 * Architectures may override this function to map oldmem
223 */ 159 */
224 int __weak remap_oldmem_pfn_range(struct vm_ar 160 int __weak remap_oldmem_pfn_range(struct vm_area_struct *vma,
225 unsigned lon 161 unsigned long from, unsigned long pfn,
226 unsigned lon 162 unsigned long size, pgprot_t prot)
227 { 163 {
228 prot = pgprot_encrypted(prot); <<
229 return remap_pfn_range(vma, from, pfn, 164 return remap_pfn_range(vma, from, pfn, size, prot);
230 } 165 }
231 166
232 /* 167 /*
233 * Architectures which support memory encrypti !! 168 * Copy to either kernel or user space
234 */ 169 */
235 ssize_t __weak copy_oldmem_page_encrypted(stru !! 170 static int copy_to(void *target, void *src, size_t size, int userbuf)
236 unsigned long pfn, size_t csiz <<
237 { 171 {
238 return copy_oldmem_page(iter, pfn, csi !! 172 if (userbuf) {
239 } !! 173 if (copy_to_user((char __user *) target, src, size))
240 !! 174 return -EFAULT;
241 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP !! 175 } else {
242 static int vmcoredd_copy_dumps(struct iov_iter !! 176 memcpy(target, src, size);
243 { <<
244 struct vmcoredd_node *dump; <<
245 u64 offset = 0; <<
246 int ret = 0; <<
247 size_t tsz; <<
248 char *buf; <<
249 <<
250 mutex_lock(&vmcoredd_mutex); <<
251 list_for_each_entry(dump, &vmcoredd_li <<
252 if (start < offset + dump->siz <<
253 tsz = min(offset + (u6 <<
254 buf = dump->buf + star <<
255 if (copy_to_iter(buf, <<
256 ret = -EFAULT; <<
257 goto out_unloc <<
258 } <<
259 <<
260 size -= tsz; <<
261 start += tsz; <<
262 <<
263 /* Leave now if buffer <<
264 if (!size) <<
265 goto out_unloc <<
266 } <<
267 offset += dump->size; <<
268 } <<
269 <<
270 out_unlock: <<
271 mutex_unlock(&vmcoredd_mutex); <<
272 return ret; <<
273 } <<
274 <<
275 #ifdef CONFIG_MMU <<
276 static int vmcoredd_mmap_dumps(struct vm_area_ <<
277 u64 start, size <<
278 { <<
279 struct vmcoredd_node *dump; <<
280 u64 offset = 0; <<
281 int ret = 0; <<
282 size_t tsz; <<
283 char *buf; <<
284 <<
285 mutex_lock(&vmcoredd_mutex); <<
286 list_for_each_entry(dump, &vmcoredd_li <<
287 if (start < offset + dump->siz <<
288 tsz = min(offset + (u6 <<
289 buf = dump->buf + star <<
290 if (remap_vmalloc_rang <<
291 <<
292 ret = -EFAULT; <<
293 goto out_unloc <<
294 } <<
295 <<
296 size -= tsz; <<
297 start += tsz; <<
298 dst += tsz; <<
299 <<
300 /* Leave now if buffer <<
301 if (!size) <<
302 goto out_unloc <<
303 } <<
304 offset += dump->size; <<
305 } 177 }
306 !! 178 return 0;
307 out_unlock: <<
308 mutex_unlock(&vmcoredd_mutex); <<
309 return ret; <<
310 } 179 }
311 #endif /* CONFIG_MMU */ <<
312 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
313 180
314 /* Read from the ELF header and then the crash 181 /* Read from the ELF header and then the crash dump. On error, negative value is
315 * returned otherwise number of bytes read are 182 * returned otherwise number of bytes read are returned.
316 */ 183 */
317 static ssize_t __read_vmcore(struct iov_iter * !! 184 static ssize_t __read_vmcore(char *buffer, size_t buflen, loff_t *fpos,
>> 185 int userbuf)
318 { 186 {
319 ssize_t acc = 0, tmp; 187 ssize_t acc = 0, tmp;
320 size_t tsz; 188 size_t tsz;
321 u64 start; 189 u64 start;
322 struct vmcore *m = NULL; 190 struct vmcore *m = NULL;
323 191
324 if (!iov_iter_count(iter) || *fpos >= !! 192 if (buflen == 0 || *fpos >= vmcore_size)
325 return 0; 193 return 0;
326 194
327 iov_iter_truncate(iter, vmcore_size - !! 195 /* trim buflen to not go beyond EOF */
>> 196 if (buflen > vmcore_size - *fpos)
>> 197 buflen = vmcore_size - *fpos;
328 198
329 /* Read ELF core header */ 199 /* Read ELF core header */
330 if (*fpos < elfcorebuf_sz) { 200 if (*fpos < elfcorebuf_sz) {
331 tsz = min(elfcorebuf_sz - (siz !! 201 tsz = min(elfcorebuf_sz - (size_t)*fpos, buflen);
332 if (copy_to_iter(elfcorebuf + !! 202 if (copy_to(buffer, elfcorebuf + *fpos, tsz, userbuf))
333 return -EFAULT; 203 return -EFAULT;
>> 204 buflen -= tsz;
334 *fpos += tsz; 205 *fpos += tsz;
>> 206 buffer += tsz;
335 acc += tsz; 207 acc += tsz;
336 208
337 /* leave now if filled buffer 209 /* leave now if filled buffer already */
338 if (!iov_iter_count(iter)) !! 210 if (buflen == 0)
339 return acc; 211 return acc;
340 } 212 }
341 213
342 /* Read ELF note segment */ !! 214 /* Read Elf note segment */
343 if (*fpos < elfcorebuf_sz + elfnotes_s 215 if (*fpos < elfcorebuf_sz + elfnotes_sz) {
344 void *kaddr; 216 void *kaddr;
345 217
346 /* We add device dumps before !! 218 tsz = min(elfcorebuf_sz + elfnotes_sz - (size_t)*fpos, buflen);
347 * other elf notes may not fil !! 219 kaddr = elfnotes_buf + *fpos - elfcorebuf_sz;
348 * completely and we will end !! 220 if (copy_to(buffer, kaddr, tsz, userbuf))
349 * between the elf notes and t <<
350 * then try to decode this zer <<
351 * and we don't want that. Hen <<
352 * the other elf notes ensure <<
353 * avoided. <<
354 */ <<
355 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP <<
356 /* Read device dumps */ <<
357 if (*fpos < elfcorebuf_sz + vm <<
358 tsz = min(elfcorebuf_s <<
359 (size_t)*fpo <<
360 start = *fpos - elfcor <<
361 if (vmcoredd_copy_dump <<
362 return -EFAULT <<
363 <<
364 *fpos += tsz; <<
365 acc += tsz; <<
366 <<
367 /* leave now if filled <<
368 if (!iov_iter_count(it <<
369 return acc; <<
370 } <<
371 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
372 <<
373 /* Read remaining elf notes */ <<
374 tsz = min(elfcorebuf_sz + elfn <<
375 iov_iter_count(iter) <<
376 kaddr = elfnotes_buf + *fpos - <<
377 if (copy_to_iter(kaddr, tsz, i <<
378 return -EFAULT; 221 return -EFAULT;
379 !! 222 buflen -= tsz;
380 *fpos += tsz; 223 *fpos += tsz;
>> 224 buffer += tsz;
381 acc += tsz; 225 acc += tsz;
382 226
383 /* leave now if filled buffer 227 /* leave now if filled buffer already */
384 if (!iov_iter_count(iter)) !! 228 if (buflen == 0)
385 return acc; 229 return acc;
386 <<
387 cond_resched(); <<
388 } 230 }
389 231
390 list_for_each_entry(m, &vmcore_list, l 232 list_for_each_entry(m, &vmcore_list, list) {
391 if (*fpos < m->offset + m->siz 233 if (*fpos < m->offset + m->size) {
392 tsz = (size_t)min_t(un 234 tsz = (size_t)min_t(unsigned long long,
393 m- 235 m->offset + m->size - *fpos,
394 io !! 236 buflen);
395 start = m->paddr + *fp 237 start = m->paddr + *fpos - m->offset;
396 tmp = read_from_oldmem !! 238 tmp = read_from_oldmem(buffer, tsz, &start, userbuf);
397 cc_pla <<
398 if (tmp < 0) 239 if (tmp < 0)
399 return tmp; 240 return tmp;
>> 241 buflen -= tsz;
400 *fpos += tsz; 242 *fpos += tsz;
>> 243 buffer += tsz;
401 acc += tsz; 244 acc += tsz;
402 245
403 /* leave now if filled 246 /* leave now if filled buffer already */
404 if (!iov_iter_count(it !! 247 if (buflen == 0)
405 return acc; 248 return acc;
406 } 249 }
407 } 250 }
408 251
409 return acc; 252 return acc;
410 } 253 }
411 254
412 static ssize_t read_vmcore(struct kiocb *iocb, !! 255 static ssize_t read_vmcore(struct file *file, char __user *buffer,
>> 256 size_t buflen, loff_t *fpos)
413 { 257 {
414 return __read_vmcore(iter, &iocb->ki_p !! 258 return __read_vmcore((__force char *) buffer, buflen, fpos, 1);
415 } 259 }
416 260
417 /* 261 /*
418 * The vmcore fault handler uses the page cach 262 * The vmcore fault handler uses the page cache and fills data using the
419 * standard __read_vmcore() function. !! 263 * standard __vmcore_read() function.
420 * 264 *
421 * On s390 the fault handler is used for memor 265 * On s390 the fault handler is used for memory regions that can't be mapped
422 * directly with remap_pfn_range(). 266 * directly with remap_pfn_range().
423 */ 267 */
424 static vm_fault_t mmap_vmcore_fault(struct vm_ !! 268 static int mmap_vmcore_fault(struct vm_fault *vmf)
425 { 269 {
426 #ifdef CONFIG_S390 270 #ifdef CONFIG_S390
427 struct address_space *mapping = vmf->v 271 struct address_space *mapping = vmf->vma->vm_file->f_mapping;
428 pgoff_t index = vmf->pgoff; 272 pgoff_t index = vmf->pgoff;
429 struct iov_iter iter; <<
430 struct kvec kvec; <<
431 struct page *page; 273 struct page *page;
432 loff_t offset; 274 loff_t offset;
>> 275 char *buf;
433 int rc; 276 int rc;
434 277
435 page = find_or_create_page(mapping, in 278 page = find_or_create_page(mapping, index, GFP_KERNEL);
436 if (!page) 279 if (!page)
437 return VM_FAULT_OOM; 280 return VM_FAULT_OOM;
438 if (!PageUptodate(page)) { 281 if (!PageUptodate(page)) {
439 offset = (loff_t) index << PAG 282 offset = (loff_t) index << PAGE_SHIFT;
440 kvec.iov_base = page_address(p !! 283 buf = __va((page_to_pfn(page) << PAGE_SHIFT));
441 kvec.iov_len = PAGE_SIZE; !! 284 rc = __read_vmcore(buf, PAGE_SIZE, &offset, 0);
442 iov_iter_kvec(&iter, ITER_DEST <<
443 <<
444 rc = __read_vmcore(&iter, &off <<
445 if (rc < 0) { 285 if (rc < 0) {
446 unlock_page(page); 286 unlock_page(page);
447 put_page(page); 287 put_page(page);
448 return vmf_error(rc); !! 288 return (rc == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS;
449 } 289 }
450 SetPageUptodate(page); 290 SetPageUptodate(page);
451 } 291 }
452 unlock_page(page); 292 unlock_page(page);
453 vmf->page = page; 293 vmf->page = page;
454 return 0; 294 return 0;
455 #else 295 #else
456 return VM_FAULT_SIGBUS; 296 return VM_FAULT_SIGBUS;
457 #endif 297 #endif
458 } 298 }
459 299
>> 300 static const struct vm_operations_struct vmcore_mmap_ops = {
>> 301 .fault = mmap_vmcore_fault,
>> 302 };
>> 303
460 /** 304 /**
461 * vmcore_alloc_buf - allocate buffer in vmall !! 305 * alloc_elfnotes_buf - allocate buffer for ELF note segment in
462 * @size: size of buffer !! 306 * vmalloc memory
>> 307 *
>> 308 * @notes_sz: size of buffer
463 * 309 *
464 * If CONFIG_MMU is defined, use vmalloc_user( 310 * If CONFIG_MMU is defined, use vmalloc_user() to allow users to mmap
465 * the buffer to user-space by means of remap_ 311 * the buffer to user-space by means of remap_vmalloc_range().
466 * 312 *
467 * If CONFIG_MMU is not defined, use vzalloc() 313 * If CONFIG_MMU is not defined, use vzalloc() since mmap_vmcore() is
468 * disabled and there's no need to allow users 314 * disabled and there's no need to allow users to mmap the buffer.
469 */ 315 */
470 static inline char *vmcore_alloc_buf(size_t si !! 316 static inline char *alloc_elfnotes_buf(size_t notes_sz)
471 { 317 {
472 #ifdef CONFIG_MMU 318 #ifdef CONFIG_MMU
473 return vmalloc_user(size); !! 319 return vmalloc_user(notes_sz);
474 #else 320 #else
475 return vzalloc(size); !! 321 return vzalloc(notes_sz);
476 #endif 322 #endif
477 } 323 }
478 324
479 /* 325 /*
480 * Disable mmap_vmcore() if CONFIG_MMU is not 326 * Disable mmap_vmcore() if CONFIG_MMU is not defined. MMU is
481 * essential for mmap_vmcore() in order to map 327 * essential for mmap_vmcore() in order to map physically
482 * non-contiguous objects (ELF header, ELF not 328 * non-contiguous objects (ELF header, ELF note segment and memory
483 * regions in the 1st kernel pointed to by PT_ 329 * regions in the 1st kernel pointed to by PT_LOAD entries) into
484 * virtually contiguous user-space in ELF layo 330 * virtually contiguous user-space in ELF layout.
485 */ 331 */
486 #ifdef CONFIG_MMU 332 #ifdef CONFIG_MMU
487 <<
488 static const struct vm_operations_struct vmcor <<
489 .fault = mmap_vmcore_fault, <<
490 }; <<
491 <<
492 /* 333 /*
493 * remap_oldmem_pfn_checked - do remap_oldmem_ 334 * remap_oldmem_pfn_checked - do remap_oldmem_pfn_range replacing all pages
494 * reported as not being ram with the zero pag 335 * reported as not being ram with the zero page.
495 * 336 *
496 * @vma: vm_area_struct describing requested m 337 * @vma: vm_area_struct describing requested mapping
497 * @from: start remapping from 338 * @from: start remapping from
498 * @pfn: page frame number to start remapping 339 * @pfn: page frame number to start remapping to
499 * @size: remapping size 340 * @size: remapping size
500 * @prot: protection bits 341 * @prot: protection bits
501 * 342 *
502 * Returns zero on success, -EAGAIN on failure 343 * Returns zero on success, -EAGAIN on failure.
503 */ 344 */
504 static int remap_oldmem_pfn_checked(struct vm_ 345 static int remap_oldmem_pfn_checked(struct vm_area_struct *vma,
505 unsigned l 346 unsigned long from, unsigned long pfn,
506 unsigned l 347 unsigned long size, pgprot_t prot)
507 { 348 {
508 unsigned long map_size; 349 unsigned long map_size;
509 unsigned long pos_start, pos_end, pos; 350 unsigned long pos_start, pos_end, pos;
510 unsigned long zeropage_pfn = my_zero_p 351 unsigned long zeropage_pfn = my_zero_pfn(0);
511 size_t len = 0; 352 size_t len = 0;
512 353
513 pos_start = pfn; 354 pos_start = pfn;
514 pos_end = pfn + (size >> PAGE_SHIFT); 355 pos_end = pfn + (size >> PAGE_SHIFT);
515 356
516 for (pos = pos_start; pos < pos_end; + 357 for (pos = pos_start; pos < pos_end; ++pos) {
517 if (!pfn_is_ram(pos)) { 358 if (!pfn_is_ram(pos)) {
518 /* 359 /*
519 * We hit a page which 360 * We hit a page which is not ram. Remap the continuous
520 * region between pos_ 361 * region between pos_start and pos-1 and replace
521 * the non-ram page at 362 * the non-ram page at pos with the zero page.
522 */ 363 */
523 if (pos > pos_start) { 364 if (pos > pos_start) {
524 /* Remap conti 365 /* Remap continuous region */
525 map_size = (po 366 map_size = (pos - pos_start) << PAGE_SHIFT;
526 if (remap_oldm 367 if (remap_oldmem_pfn_range(vma, from + len,
527 368 pos_start, map_size,
528 369 prot))
529 goto f 370 goto fail;
530 len += map_siz 371 len += map_size;
531 } 372 }
532 /* Remap the zero page 373 /* Remap the zero page */
533 if (remap_oldmem_pfn_r 374 if (remap_oldmem_pfn_range(vma, from + len,
534 375 zeropage_pfn,
535 376 PAGE_SIZE, prot))
536 goto fail; 377 goto fail;
537 len += PAGE_SIZE; 378 len += PAGE_SIZE;
538 pos_start = pos + 1; 379 pos_start = pos + 1;
539 } 380 }
540 } 381 }
541 if (pos > pos_start) { 382 if (pos > pos_start) {
542 /* Remap the rest */ 383 /* Remap the rest */
543 map_size = (pos - pos_start) < 384 map_size = (pos - pos_start) << PAGE_SHIFT;
544 if (remap_oldmem_pfn_range(vma 385 if (remap_oldmem_pfn_range(vma, from + len, pos_start,
545 map 386 map_size, prot))
546 goto fail; 387 goto fail;
547 } 388 }
548 return 0; 389 return 0;
549 fail: 390 fail:
550 do_munmap(vma->vm_mm, from, len, NULL) 391 do_munmap(vma->vm_mm, from, len, NULL);
551 return -EAGAIN; 392 return -EAGAIN;
552 } 393 }
553 394
554 static int vmcore_remap_oldmem_pfn(struct vm_a 395 static int vmcore_remap_oldmem_pfn(struct vm_area_struct *vma,
555 unsigned long from 396 unsigned long from, unsigned long pfn,
556 unsigned long size 397 unsigned long size, pgprot_t prot)
557 { 398 {
558 int ret, idx; <<
559 <<
560 /* 399 /*
561 * Check if a callback was registered !! 400 * Check if oldmem_pfn_is_ram was registered to avoid
562 * pages without a reason. !! 401 * looping over all pages without a reason.
563 */ 402 */
564 idx = srcu_read_lock(&vmcore_cb_srcu); !! 403 if (oldmem_pfn_is_ram)
565 if (!list_empty(&vmcore_cb_list)) !! 404 return remap_oldmem_pfn_checked(vma, from, pfn, size, prot);
566 ret = remap_oldmem_pfn_checked <<
567 else 405 else
568 ret = remap_oldmem_pfn_range(v !! 406 return remap_oldmem_pfn_range(vma, from, pfn, size, prot);
569 srcu_read_unlock(&vmcore_cb_srcu, idx) <<
570 return ret; <<
571 } 407 }
572 408
573 static int mmap_vmcore(struct file *file, stru 409 static int mmap_vmcore(struct file *file, struct vm_area_struct *vma)
574 { 410 {
575 size_t size = vma->vm_end - vma->vm_st 411 size_t size = vma->vm_end - vma->vm_start;
576 u64 start, end, len, tsz; 412 u64 start, end, len, tsz;
577 struct vmcore *m; 413 struct vmcore *m;
578 414
579 start = (u64)vma->vm_pgoff << PAGE_SHI 415 start = (u64)vma->vm_pgoff << PAGE_SHIFT;
580 end = start + size; 416 end = start + size;
581 417
582 if (size > vmcore_size || end > vmcore 418 if (size > vmcore_size || end > vmcore_size)
583 return -EINVAL; 419 return -EINVAL;
584 420
585 if (vma->vm_flags & (VM_WRITE | VM_EXE 421 if (vma->vm_flags & (VM_WRITE | VM_EXEC))
586 return -EPERM; 422 return -EPERM;
587 423
588 vm_flags_mod(vma, VM_MIXEDMAP, VM_MAYW !! 424 vma->vm_flags &= ~(VM_MAYWRITE | VM_MAYEXEC);
>> 425 vma->vm_flags |= VM_MIXEDMAP;
589 vma->vm_ops = &vmcore_mmap_ops; 426 vma->vm_ops = &vmcore_mmap_ops;
590 427
591 len = 0; 428 len = 0;
592 429
593 if (start < elfcorebuf_sz) { 430 if (start < elfcorebuf_sz) {
594 u64 pfn; 431 u64 pfn;
595 432
596 tsz = min(elfcorebuf_sz - (siz 433 tsz = min(elfcorebuf_sz - (size_t)start, size);
597 pfn = __pa(elfcorebuf + start) 434 pfn = __pa(elfcorebuf + start) >> PAGE_SHIFT;
598 if (remap_pfn_range(vma, vma-> 435 if (remap_pfn_range(vma, vma->vm_start, pfn, tsz,
599 vma->vm_pa 436 vma->vm_page_prot))
600 return -EAGAIN; 437 return -EAGAIN;
601 size -= tsz; 438 size -= tsz;
602 start += tsz; 439 start += tsz;
603 len += tsz; 440 len += tsz;
604 441
605 if (size == 0) 442 if (size == 0)
606 return 0; 443 return 0;
607 } 444 }
608 445
609 if (start < elfcorebuf_sz + elfnotes_s 446 if (start < elfcorebuf_sz + elfnotes_sz) {
610 void *kaddr; 447 void *kaddr;
611 448
612 /* We add device dumps before <<
613 * other elf notes may not fil <<
614 * completely and we will end <<
615 * between the elf notes and t <<
616 * then try to decode this zer <<
617 * and we don't want that. Hen <<
618 * the other elf notes ensure <<
619 * avoided. This also ensures <<
620 * other elf notes can be prop <<
621 * address. <<
622 */ <<
623 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP <<
624 /* Read device dumps */ <<
625 if (start < elfcorebuf_sz + vm <<
626 u64 start_off; <<
627 <<
628 tsz = min(elfcorebuf_s <<
629 (size_t)star <<
630 start_off = start - el <<
631 if (vmcoredd_mmap_dump <<
632 <<
633 goto fail; <<
634 <<
635 size -= tsz; <<
636 start += tsz; <<
637 len += tsz; <<
638 <<
639 /* leave now if filled <<
640 if (!size) <<
641 return 0; <<
642 } <<
643 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
644 <<
645 /* Read remaining elf notes */ <<
646 tsz = min(elfcorebuf_sz + elfn 449 tsz = min(elfcorebuf_sz + elfnotes_sz - (size_t)start, size);
647 kaddr = elfnotes_buf + start - !! 450 kaddr = elfnotes_buf + start - elfcorebuf_sz;
648 if (remap_vmalloc_range_partia 451 if (remap_vmalloc_range_partial(vma, vma->vm_start + len,
649 !! 452 kaddr, tsz))
650 goto fail; 453 goto fail;
651 <<
652 size -= tsz; 454 size -= tsz;
653 start += tsz; 455 start += tsz;
654 len += tsz; 456 len += tsz;
655 457
656 if (size == 0) 458 if (size == 0)
657 return 0; 459 return 0;
658 } 460 }
659 461
660 list_for_each_entry(m, &vmcore_list, l 462 list_for_each_entry(m, &vmcore_list, list) {
661 if (start < m->offset + m->siz 463 if (start < m->offset + m->size) {
662 u64 paddr = 0; 464 u64 paddr = 0;
663 465
664 tsz = (size_t)min_t(un 466 tsz = (size_t)min_t(unsigned long long,
665 m- 467 m->offset + m->size - start, size);
666 paddr = m->paddr + sta 468 paddr = m->paddr + start - m->offset;
667 if (vmcore_remap_oldme 469 if (vmcore_remap_oldmem_pfn(vma, vma->vm_start + len,
668 470 paddr >> PAGE_SHIFT, tsz,
669 471 vma->vm_page_prot))
670 goto fail; 472 goto fail;
671 size -= tsz; 473 size -= tsz;
672 start += tsz; 474 start += tsz;
673 len += tsz; 475 len += tsz;
674 476
675 if (size == 0) 477 if (size == 0)
676 return 0; 478 return 0;
677 } 479 }
678 } 480 }
679 481
680 return 0; 482 return 0;
681 fail: 483 fail:
682 do_munmap(vma->vm_mm, vma->vm_start, l 484 do_munmap(vma->vm_mm, vma->vm_start, len, NULL);
683 return -EAGAIN; 485 return -EAGAIN;
684 } 486 }
685 #else 487 #else
686 static int mmap_vmcore(struct file *file, stru 488 static int mmap_vmcore(struct file *file, struct vm_area_struct *vma)
687 { 489 {
688 return -ENOSYS; 490 return -ENOSYS;
689 } 491 }
690 #endif 492 #endif
691 493
692 static const struct proc_ops vmcore_proc_ops = !! 494 static const struct file_operations proc_vmcore_operations = {
693 .proc_open = open_vmcore, !! 495 .read = read_vmcore,
694 .proc_read_iter = read_vmcore, !! 496 .llseek = default_llseek,
695 .proc_lseek = default_llseek, !! 497 .mmap = mmap_vmcore,
696 .proc_mmap = mmap_vmcore, <<
697 }; 498 };
698 499
699 static struct vmcore* __init get_new_element(v 500 static struct vmcore* __init get_new_element(void)
700 { 501 {
701 return kzalloc(sizeof(struct vmcore), 502 return kzalloc(sizeof(struct vmcore), GFP_KERNEL);
702 } 503 }
703 504
704 static u64 get_vmcore_size(size_t elfsz, size_ !! 505 static u64 __init get_vmcore_size(size_t elfsz, size_t elfnotesegsz,
705 struct list_head *v !! 506 struct list_head *vc_list)
706 { 507 {
707 u64 size; 508 u64 size;
708 struct vmcore *m; 509 struct vmcore *m;
709 510
710 size = elfsz + elfnotesegsz; 511 size = elfsz + elfnotesegsz;
711 list_for_each_entry(m, vc_list, list) 512 list_for_each_entry(m, vc_list, list) {
712 size += m->size; 513 size += m->size;
713 } 514 }
714 return size; 515 return size;
715 } 516 }
716 517
717 /** 518 /**
718 * update_note_header_size_elf64 - update p_me 519 * update_note_header_size_elf64 - update p_memsz member of each PT_NOTE entry
719 * 520 *
720 * @ehdr_ptr: ELF header 521 * @ehdr_ptr: ELF header
721 * 522 *
722 * This function updates p_memsz member of eac 523 * This function updates p_memsz member of each PT_NOTE entry in the
723 * program header table pointed to by @ehdr_pt 524 * program header table pointed to by @ehdr_ptr to real size of ELF
724 * note segment. 525 * note segment.
725 */ 526 */
726 static int __init update_note_header_size_elf6 527 static int __init update_note_header_size_elf64(const Elf64_Ehdr *ehdr_ptr)
727 { 528 {
728 int i, rc=0; 529 int i, rc=0;
729 Elf64_Phdr *phdr_ptr; 530 Elf64_Phdr *phdr_ptr;
730 Elf64_Nhdr *nhdr_ptr; 531 Elf64_Nhdr *nhdr_ptr;
731 532
732 phdr_ptr = (Elf64_Phdr *)(ehdr_ptr + 1 533 phdr_ptr = (Elf64_Phdr *)(ehdr_ptr + 1);
733 for (i = 0; i < ehdr_ptr->e_phnum; i++ 534 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
734 void *notes_section; 535 void *notes_section;
735 u64 offset, max_sz, sz, real_s 536 u64 offset, max_sz, sz, real_sz = 0;
736 if (phdr_ptr->p_type != PT_NOT 537 if (phdr_ptr->p_type != PT_NOTE)
737 continue; 538 continue;
738 max_sz = phdr_ptr->p_memsz; 539 max_sz = phdr_ptr->p_memsz;
739 offset = phdr_ptr->p_offset; 540 offset = phdr_ptr->p_offset;
740 notes_section = kmalloc(max_sz 541 notes_section = kmalloc(max_sz, GFP_KERNEL);
741 if (!notes_section) 542 if (!notes_section)
742 return -ENOMEM; 543 return -ENOMEM;
743 rc = elfcorehdr_read_notes(not 544 rc = elfcorehdr_read_notes(notes_section, max_sz, &offset);
744 if (rc < 0) { 545 if (rc < 0) {
745 kfree(notes_section); 546 kfree(notes_section);
746 return rc; 547 return rc;
747 } 548 }
748 nhdr_ptr = notes_section; 549 nhdr_ptr = notes_section;
749 while (nhdr_ptr->n_namesz != 0 550 while (nhdr_ptr->n_namesz != 0) {
750 sz = sizeof(Elf64_Nhdr 551 sz = sizeof(Elf64_Nhdr) +
751 (((u64)nhdr_pt 552 (((u64)nhdr_ptr->n_namesz + 3) & ~3) +
752 (((u64)nhdr_pt 553 (((u64)nhdr_ptr->n_descsz + 3) & ~3);
753 if ((real_sz + sz) > m 554 if ((real_sz + sz) > max_sz) {
754 pr_warn("Warni 555 pr_warn("Warning: Exceeded p_memsz, dropping PT_NOTE entry n_namesz=0x%x, n_descsz=0x%x\n",
755 nhdr_p 556 nhdr_ptr->n_namesz, nhdr_ptr->n_descsz);
756 break; 557 break;
757 } 558 }
758 real_sz += sz; 559 real_sz += sz;
759 nhdr_ptr = (Elf64_Nhdr 560 nhdr_ptr = (Elf64_Nhdr*)((char*)nhdr_ptr + sz);
760 } 561 }
761 kfree(notes_section); 562 kfree(notes_section);
762 phdr_ptr->p_memsz = real_sz; 563 phdr_ptr->p_memsz = real_sz;
763 if (real_sz == 0) { 564 if (real_sz == 0) {
764 pr_warn("Warning: Zero 565 pr_warn("Warning: Zero PT_NOTE entries found\n");
765 } 566 }
766 } 567 }
767 568
768 return 0; 569 return 0;
769 } 570 }
770 571
771 /** 572 /**
772 * get_note_number_and_size_elf64 - get the nu 573 * get_note_number_and_size_elf64 - get the number of PT_NOTE program
773 * headers and sum of real size of their ELF n 574 * headers and sum of real size of their ELF note segment headers and
774 * data. 575 * data.
775 * 576 *
776 * @ehdr_ptr: ELF header 577 * @ehdr_ptr: ELF header
777 * @nr_ptnote: buffer for the number of PT_NOT 578 * @nr_ptnote: buffer for the number of PT_NOTE program headers
778 * @sz_ptnote: buffer for size of unique PT_NO 579 * @sz_ptnote: buffer for size of unique PT_NOTE program header
779 * 580 *
780 * This function is used to merge multiple PT_ 581 * This function is used to merge multiple PT_NOTE program headers
781 * into a unique single one. The resulting uni 582 * into a unique single one. The resulting unique entry will have
782 * @sz_ptnote in its phdr->p_mem. 583 * @sz_ptnote in its phdr->p_mem.
783 * 584 *
784 * It is assumed that program headers with PT_ 585 * It is assumed that program headers with PT_NOTE type pointed to by
785 * @ehdr_ptr has already been updated by updat 586 * @ehdr_ptr has already been updated by update_note_header_size_elf64
786 * and each of PT_NOTE program headers has act 587 * and each of PT_NOTE program headers has actual ELF note segment
787 * size in its p_memsz member. 588 * size in its p_memsz member.
788 */ 589 */
789 static int __init get_note_number_and_size_elf 590 static int __init get_note_number_and_size_elf64(const Elf64_Ehdr *ehdr_ptr,
790 591 int *nr_ptnote, u64 *sz_ptnote)
791 { 592 {
792 int i; 593 int i;
793 Elf64_Phdr *phdr_ptr; 594 Elf64_Phdr *phdr_ptr;
794 595
795 *nr_ptnote = *sz_ptnote = 0; 596 *nr_ptnote = *sz_ptnote = 0;
796 597
797 phdr_ptr = (Elf64_Phdr *)(ehdr_ptr + 1 598 phdr_ptr = (Elf64_Phdr *)(ehdr_ptr + 1);
798 for (i = 0; i < ehdr_ptr->e_phnum; i++ 599 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
799 if (phdr_ptr->p_type != PT_NOT 600 if (phdr_ptr->p_type != PT_NOTE)
800 continue; 601 continue;
801 *nr_ptnote += 1; 602 *nr_ptnote += 1;
802 *sz_ptnote += phdr_ptr->p_mems 603 *sz_ptnote += phdr_ptr->p_memsz;
803 } 604 }
804 605
805 return 0; 606 return 0;
806 } 607 }
807 608
808 /** 609 /**
809 * copy_notes_elf64 - copy ELF note segments i 610 * copy_notes_elf64 - copy ELF note segments in a given buffer
810 * 611 *
811 * @ehdr_ptr: ELF header 612 * @ehdr_ptr: ELF header
812 * @notes_buf: buffer into which ELF note segm 613 * @notes_buf: buffer into which ELF note segments are copied
813 * 614 *
814 * This function is used to copy ELF note segm 615 * This function is used to copy ELF note segment in the 1st kernel
815 * into the buffer @notes_buf in the 2nd kerne 616 * into the buffer @notes_buf in the 2nd kernel. It is assumed that
816 * size of the buffer @notes_buf is equal to o 617 * size of the buffer @notes_buf is equal to or larger than sum of the
817 * real ELF note segment headers and data. 618 * real ELF note segment headers and data.
818 * 619 *
819 * It is assumed that program headers with PT_ 620 * It is assumed that program headers with PT_NOTE type pointed to by
820 * @ehdr_ptr has already been updated by updat 621 * @ehdr_ptr has already been updated by update_note_header_size_elf64
821 * and each of PT_NOTE program headers has act 622 * and each of PT_NOTE program headers has actual ELF note segment
822 * size in its p_memsz member. 623 * size in its p_memsz member.
823 */ 624 */
824 static int __init copy_notes_elf64(const Elf64 625 static int __init copy_notes_elf64(const Elf64_Ehdr *ehdr_ptr, char *notes_buf)
825 { 626 {
826 int i, rc=0; 627 int i, rc=0;
827 Elf64_Phdr *phdr_ptr; 628 Elf64_Phdr *phdr_ptr;
828 629
829 phdr_ptr = (Elf64_Phdr*)(ehdr_ptr + 1) 630 phdr_ptr = (Elf64_Phdr*)(ehdr_ptr + 1);
830 631
831 for (i = 0; i < ehdr_ptr->e_phnum; i++ 632 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
832 u64 offset; 633 u64 offset;
833 if (phdr_ptr->p_type != PT_NOT 634 if (phdr_ptr->p_type != PT_NOTE)
834 continue; 635 continue;
835 offset = phdr_ptr->p_offset; 636 offset = phdr_ptr->p_offset;
836 rc = elfcorehdr_read_notes(not 637 rc = elfcorehdr_read_notes(notes_buf, phdr_ptr->p_memsz,
837 &of 638 &offset);
838 if (rc < 0) 639 if (rc < 0)
839 return rc; 640 return rc;
840 notes_buf += phdr_ptr->p_memsz 641 notes_buf += phdr_ptr->p_memsz;
841 } 642 }
842 643
843 return 0; 644 return 0;
844 } 645 }
845 646
846 /* Merges all the PT_NOTE headers into one. */ 647 /* Merges all the PT_NOTE headers into one. */
847 static int __init merge_note_headers_elf64(cha 648 static int __init merge_note_headers_elf64(char *elfptr, size_t *elfsz,
848 cha 649 char **notes_buf, size_t *notes_sz)
849 { 650 {
850 int i, nr_ptnote=0, rc=0; 651 int i, nr_ptnote=0, rc=0;
851 char *tmp; 652 char *tmp;
852 Elf64_Ehdr *ehdr_ptr; 653 Elf64_Ehdr *ehdr_ptr;
853 Elf64_Phdr phdr; 654 Elf64_Phdr phdr;
854 u64 phdr_sz = 0, note_off; 655 u64 phdr_sz = 0, note_off;
855 656
856 ehdr_ptr = (Elf64_Ehdr *)elfptr; 657 ehdr_ptr = (Elf64_Ehdr *)elfptr;
857 658
858 rc = update_note_header_size_elf64(ehd 659 rc = update_note_header_size_elf64(ehdr_ptr);
859 if (rc < 0) 660 if (rc < 0)
860 return rc; 661 return rc;
861 662
862 rc = get_note_number_and_size_elf64(eh 663 rc = get_note_number_and_size_elf64(ehdr_ptr, &nr_ptnote, &phdr_sz);
863 if (rc < 0) 664 if (rc < 0)
864 return rc; 665 return rc;
865 666
866 *notes_sz = roundup(phdr_sz, PAGE_SIZE 667 *notes_sz = roundup(phdr_sz, PAGE_SIZE);
867 *notes_buf = vmcore_alloc_buf(*notes_s !! 668 *notes_buf = alloc_elfnotes_buf(*notes_sz);
868 if (!*notes_buf) 669 if (!*notes_buf)
869 return -ENOMEM; 670 return -ENOMEM;
870 671
871 rc = copy_notes_elf64(ehdr_ptr, *notes 672 rc = copy_notes_elf64(ehdr_ptr, *notes_buf);
872 if (rc < 0) 673 if (rc < 0)
873 return rc; 674 return rc;
874 675
875 /* Prepare merged PT_NOTE program head 676 /* Prepare merged PT_NOTE program header. */
876 phdr.p_type = PT_NOTE; 677 phdr.p_type = PT_NOTE;
877 phdr.p_flags = 0; 678 phdr.p_flags = 0;
878 note_off = sizeof(Elf64_Ehdr) + 679 note_off = sizeof(Elf64_Ehdr) +
879 (ehdr_ptr->e_phnum - n 680 (ehdr_ptr->e_phnum - nr_ptnote +1) * sizeof(Elf64_Phdr);
880 phdr.p_offset = roundup(note_off, PAG 681 phdr.p_offset = roundup(note_off, PAGE_SIZE);
881 phdr.p_vaddr = phdr.p_paddr = 0; 682 phdr.p_vaddr = phdr.p_paddr = 0;
882 phdr.p_filesz = phdr.p_memsz = phdr_s 683 phdr.p_filesz = phdr.p_memsz = phdr_sz;
883 phdr.p_align = 4; !! 684 phdr.p_align = 0;
884 685
885 /* Add merged PT_NOTE program header*/ 686 /* Add merged PT_NOTE program header*/
886 tmp = elfptr + sizeof(Elf64_Ehdr); 687 tmp = elfptr + sizeof(Elf64_Ehdr);
887 memcpy(tmp, &phdr, sizeof(phdr)); 688 memcpy(tmp, &phdr, sizeof(phdr));
888 tmp += sizeof(phdr); 689 tmp += sizeof(phdr);
889 690
890 /* Remove unwanted PT_NOTE program hea 691 /* Remove unwanted PT_NOTE program headers. */
891 i = (nr_ptnote - 1) * sizeof(Elf64_Phd 692 i = (nr_ptnote - 1) * sizeof(Elf64_Phdr);
892 *elfsz = *elfsz - i; 693 *elfsz = *elfsz - i;
893 memmove(tmp, tmp+i, ((*elfsz)-sizeof(E 694 memmove(tmp, tmp+i, ((*elfsz)-sizeof(Elf64_Ehdr)-sizeof(Elf64_Phdr)));
894 memset(elfptr + *elfsz, 0, i); 695 memset(elfptr + *elfsz, 0, i);
895 *elfsz = roundup(*elfsz, PAGE_SIZE); 696 *elfsz = roundup(*elfsz, PAGE_SIZE);
896 697
897 /* Modify e_phnum to reflect merged he 698 /* Modify e_phnum to reflect merged headers. */
898 ehdr_ptr->e_phnum = ehdr_ptr->e_phnum 699 ehdr_ptr->e_phnum = ehdr_ptr->e_phnum - nr_ptnote + 1;
899 700
900 /* Store the size of all notes. We ne <<
901 * header when the device dumps will b <<
902 */ <<
903 elfnotes_orig_sz = phdr.p_memsz; <<
904 <<
905 return 0; 701 return 0;
906 } 702 }
907 703
908 /** 704 /**
909 * update_note_header_size_elf32 - update p_me 705 * update_note_header_size_elf32 - update p_memsz member of each PT_NOTE entry
910 * 706 *
911 * @ehdr_ptr: ELF header 707 * @ehdr_ptr: ELF header
912 * 708 *
913 * This function updates p_memsz member of eac 709 * This function updates p_memsz member of each PT_NOTE entry in the
914 * program header table pointed to by @ehdr_pt 710 * program header table pointed to by @ehdr_ptr to real size of ELF
915 * note segment. 711 * note segment.
916 */ 712 */
917 static int __init update_note_header_size_elf3 713 static int __init update_note_header_size_elf32(const Elf32_Ehdr *ehdr_ptr)
918 { 714 {
919 int i, rc=0; 715 int i, rc=0;
920 Elf32_Phdr *phdr_ptr; 716 Elf32_Phdr *phdr_ptr;
921 Elf32_Nhdr *nhdr_ptr; 717 Elf32_Nhdr *nhdr_ptr;
922 718
923 phdr_ptr = (Elf32_Phdr *)(ehdr_ptr + 1 719 phdr_ptr = (Elf32_Phdr *)(ehdr_ptr + 1);
924 for (i = 0; i < ehdr_ptr->e_phnum; i++ 720 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
925 void *notes_section; 721 void *notes_section;
926 u64 offset, max_sz, sz, real_s 722 u64 offset, max_sz, sz, real_sz = 0;
927 if (phdr_ptr->p_type != PT_NOT 723 if (phdr_ptr->p_type != PT_NOTE)
928 continue; 724 continue;
929 max_sz = phdr_ptr->p_memsz; 725 max_sz = phdr_ptr->p_memsz;
930 offset = phdr_ptr->p_offset; 726 offset = phdr_ptr->p_offset;
931 notes_section = kmalloc(max_sz 727 notes_section = kmalloc(max_sz, GFP_KERNEL);
932 if (!notes_section) 728 if (!notes_section)
933 return -ENOMEM; 729 return -ENOMEM;
934 rc = elfcorehdr_read_notes(not 730 rc = elfcorehdr_read_notes(notes_section, max_sz, &offset);
935 if (rc < 0) { 731 if (rc < 0) {
936 kfree(notes_section); 732 kfree(notes_section);
937 return rc; 733 return rc;
938 } 734 }
939 nhdr_ptr = notes_section; 735 nhdr_ptr = notes_section;
940 while (nhdr_ptr->n_namesz != 0 736 while (nhdr_ptr->n_namesz != 0) {
941 sz = sizeof(Elf32_Nhdr 737 sz = sizeof(Elf32_Nhdr) +
942 (((u64)nhdr_pt 738 (((u64)nhdr_ptr->n_namesz + 3) & ~3) +
943 (((u64)nhdr_pt 739 (((u64)nhdr_ptr->n_descsz + 3) & ~3);
944 if ((real_sz + sz) > m 740 if ((real_sz + sz) > max_sz) {
945 pr_warn("Warni 741 pr_warn("Warning: Exceeded p_memsz, dropping PT_NOTE entry n_namesz=0x%x, n_descsz=0x%x\n",
946 nhdr_p 742 nhdr_ptr->n_namesz, nhdr_ptr->n_descsz);
947 break; 743 break;
948 } 744 }
949 real_sz += sz; 745 real_sz += sz;
950 nhdr_ptr = (Elf32_Nhdr 746 nhdr_ptr = (Elf32_Nhdr*)((char*)nhdr_ptr + sz);
951 } 747 }
952 kfree(notes_section); 748 kfree(notes_section);
953 phdr_ptr->p_memsz = real_sz; 749 phdr_ptr->p_memsz = real_sz;
954 if (real_sz == 0) { 750 if (real_sz == 0) {
955 pr_warn("Warning: Zero 751 pr_warn("Warning: Zero PT_NOTE entries found\n");
956 } 752 }
957 } 753 }
958 754
959 return 0; 755 return 0;
960 } 756 }
961 757
962 /** 758 /**
963 * get_note_number_and_size_elf32 - get the nu 759 * get_note_number_and_size_elf32 - get the number of PT_NOTE program
964 * headers and sum of real size of their ELF n 760 * headers and sum of real size of their ELF note segment headers and
965 * data. 761 * data.
966 * 762 *
967 * @ehdr_ptr: ELF header 763 * @ehdr_ptr: ELF header
968 * @nr_ptnote: buffer for the number of PT_NOT 764 * @nr_ptnote: buffer for the number of PT_NOTE program headers
969 * @sz_ptnote: buffer for size of unique PT_NO 765 * @sz_ptnote: buffer for size of unique PT_NOTE program header
970 * 766 *
971 * This function is used to merge multiple PT_ 767 * This function is used to merge multiple PT_NOTE program headers
972 * into a unique single one. The resulting uni 768 * into a unique single one. The resulting unique entry will have
973 * @sz_ptnote in its phdr->p_mem. 769 * @sz_ptnote in its phdr->p_mem.
974 * 770 *
975 * It is assumed that program headers with PT_ 771 * It is assumed that program headers with PT_NOTE type pointed to by
976 * @ehdr_ptr has already been updated by updat 772 * @ehdr_ptr has already been updated by update_note_header_size_elf32
977 * and each of PT_NOTE program headers has act 773 * and each of PT_NOTE program headers has actual ELF note segment
978 * size in its p_memsz member. 774 * size in its p_memsz member.
979 */ 775 */
980 static int __init get_note_number_and_size_elf 776 static int __init get_note_number_and_size_elf32(const Elf32_Ehdr *ehdr_ptr,
981 777 int *nr_ptnote, u64 *sz_ptnote)
982 { 778 {
983 int i; 779 int i;
984 Elf32_Phdr *phdr_ptr; 780 Elf32_Phdr *phdr_ptr;
985 781
986 *nr_ptnote = *sz_ptnote = 0; 782 *nr_ptnote = *sz_ptnote = 0;
987 783
988 phdr_ptr = (Elf32_Phdr *)(ehdr_ptr + 1 784 phdr_ptr = (Elf32_Phdr *)(ehdr_ptr + 1);
989 for (i = 0; i < ehdr_ptr->e_phnum; i++ 785 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
990 if (phdr_ptr->p_type != PT_NOT 786 if (phdr_ptr->p_type != PT_NOTE)
991 continue; 787 continue;
992 *nr_ptnote += 1; 788 *nr_ptnote += 1;
993 *sz_ptnote += phdr_ptr->p_mems 789 *sz_ptnote += phdr_ptr->p_memsz;
994 } 790 }
995 791
996 return 0; 792 return 0;
997 } 793 }
998 794
999 /** 795 /**
1000 * copy_notes_elf32 - copy ELF note segments 796 * copy_notes_elf32 - copy ELF note segments in a given buffer
1001 * 797 *
1002 * @ehdr_ptr: ELF header 798 * @ehdr_ptr: ELF header
1003 * @notes_buf: buffer into which ELF note seg 799 * @notes_buf: buffer into which ELF note segments are copied
1004 * 800 *
1005 * This function is used to copy ELF note seg 801 * This function is used to copy ELF note segment in the 1st kernel
1006 * into the buffer @notes_buf in the 2nd kern 802 * into the buffer @notes_buf in the 2nd kernel. It is assumed that
1007 * size of the buffer @notes_buf is equal to 803 * size of the buffer @notes_buf is equal to or larger than sum of the
1008 * real ELF note segment headers and data. 804 * real ELF note segment headers and data.
1009 * 805 *
1010 * It is assumed that program headers with PT 806 * It is assumed that program headers with PT_NOTE type pointed to by
1011 * @ehdr_ptr has already been updated by upda 807 * @ehdr_ptr has already been updated by update_note_header_size_elf32
1012 * and each of PT_NOTE program headers has ac 808 * and each of PT_NOTE program headers has actual ELF note segment
1013 * size in its p_memsz member. 809 * size in its p_memsz member.
1014 */ 810 */
1015 static int __init copy_notes_elf32(const Elf3 811 static int __init copy_notes_elf32(const Elf32_Ehdr *ehdr_ptr, char *notes_buf)
1016 { 812 {
1017 int i, rc=0; 813 int i, rc=0;
1018 Elf32_Phdr *phdr_ptr; 814 Elf32_Phdr *phdr_ptr;
1019 815
1020 phdr_ptr = (Elf32_Phdr*)(ehdr_ptr + 1 816 phdr_ptr = (Elf32_Phdr*)(ehdr_ptr + 1);
1021 817
1022 for (i = 0; i < ehdr_ptr->e_phnum; i+ 818 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
1023 u64 offset; 819 u64 offset;
1024 if (phdr_ptr->p_type != PT_NO 820 if (phdr_ptr->p_type != PT_NOTE)
1025 continue; 821 continue;
1026 offset = phdr_ptr->p_offset; 822 offset = phdr_ptr->p_offset;
1027 rc = elfcorehdr_read_notes(no 823 rc = elfcorehdr_read_notes(notes_buf, phdr_ptr->p_memsz,
1028 &o 824 &offset);
1029 if (rc < 0) 825 if (rc < 0)
1030 return rc; 826 return rc;
1031 notes_buf += phdr_ptr->p_mems 827 notes_buf += phdr_ptr->p_memsz;
1032 } 828 }
1033 829
1034 return 0; 830 return 0;
1035 } 831 }
1036 832
1037 /* Merges all the PT_NOTE headers into one. * 833 /* Merges all the PT_NOTE headers into one. */
1038 static int __init merge_note_headers_elf32(ch 834 static int __init merge_note_headers_elf32(char *elfptr, size_t *elfsz,
1039 ch 835 char **notes_buf, size_t *notes_sz)
1040 { 836 {
1041 int i, nr_ptnote=0, rc=0; 837 int i, nr_ptnote=0, rc=0;
1042 char *tmp; 838 char *tmp;
1043 Elf32_Ehdr *ehdr_ptr; 839 Elf32_Ehdr *ehdr_ptr;
1044 Elf32_Phdr phdr; 840 Elf32_Phdr phdr;
1045 u64 phdr_sz = 0, note_off; 841 u64 phdr_sz = 0, note_off;
1046 842
1047 ehdr_ptr = (Elf32_Ehdr *)elfptr; 843 ehdr_ptr = (Elf32_Ehdr *)elfptr;
1048 844
1049 rc = update_note_header_size_elf32(eh 845 rc = update_note_header_size_elf32(ehdr_ptr);
1050 if (rc < 0) 846 if (rc < 0)
1051 return rc; 847 return rc;
1052 848
1053 rc = get_note_number_and_size_elf32(e 849 rc = get_note_number_and_size_elf32(ehdr_ptr, &nr_ptnote, &phdr_sz);
1054 if (rc < 0) 850 if (rc < 0)
1055 return rc; 851 return rc;
1056 852
1057 *notes_sz = roundup(phdr_sz, PAGE_SIZ 853 *notes_sz = roundup(phdr_sz, PAGE_SIZE);
1058 *notes_buf = vmcore_alloc_buf(*notes_ !! 854 *notes_buf = alloc_elfnotes_buf(*notes_sz);
1059 if (!*notes_buf) 855 if (!*notes_buf)
1060 return -ENOMEM; 856 return -ENOMEM;
1061 857
1062 rc = copy_notes_elf32(ehdr_ptr, *note 858 rc = copy_notes_elf32(ehdr_ptr, *notes_buf);
1063 if (rc < 0) 859 if (rc < 0)
1064 return rc; 860 return rc;
1065 861
1066 /* Prepare merged PT_NOTE program hea 862 /* Prepare merged PT_NOTE program header. */
1067 phdr.p_type = PT_NOTE; 863 phdr.p_type = PT_NOTE;
1068 phdr.p_flags = 0; 864 phdr.p_flags = 0;
1069 note_off = sizeof(Elf32_Ehdr) + 865 note_off = sizeof(Elf32_Ehdr) +
1070 (ehdr_ptr->e_phnum - 866 (ehdr_ptr->e_phnum - nr_ptnote +1) * sizeof(Elf32_Phdr);
1071 phdr.p_offset = roundup(note_off, PA 867 phdr.p_offset = roundup(note_off, PAGE_SIZE);
1072 phdr.p_vaddr = phdr.p_paddr = 0; 868 phdr.p_vaddr = phdr.p_paddr = 0;
1073 phdr.p_filesz = phdr.p_memsz = phdr_ 869 phdr.p_filesz = phdr.p_memsz = phdr_sz;
1074 phdr.p_align = 4; !! 870 phdr.p_align = 0;
1075 871
1076 /* Add merged PT_NOTE program header* 872 /* Add merged PT_NOTE program header*/
1077 tmp = elfptr + sizeof(Elf32_Ehdr); 873 tmp = elfptr + sizeof(Elf32_Ehdr);
1078 memcpy(tmp, &phdr, sizeof(phdr)); 874 memcpy(tmp, &phdr, sizeof(phdr));
1079 tmp += sizeof(phdr); 875 tmp += sizeof(phdr);
1080 876
1081 /* Remove unwanted PT_NOTE program he 877 /* Remove unwanted PT_NOTE program headers. */
1082 i = (nr_ptnote - 1) * sizeof(Elf32_Ph 878 i = (nr_ptnote - 1) * sizeof(Elf32_Phdr);
1083 *elfsz = *elfsz - i; 879 *elfsz = *elfsz - i;
1084 memmove(tmp, tmp+i, ((*elfsz)-sizeof( 880 memmove(tmp, tmp+i, ((*elfsz)-sizeof(Elf32_Ehdr)-sizeof(Elf32_Phdr)));
1085 memset(elfptr + *elfsz, 0, i); 881 memset(elfptr + *elfsz, 0, i);
1086 *elfsz = roundup(*elfsz, PAGE_SIZE); 882 *elfsz = roundup(*elfsz, PAGE_SIZE);
1087 883
1088 /* Modify e_phnum to reflect merged h 884 /* Modify e_phnum to reflect merged headers. */
1089 ehdr_ptr->e_phnum = ehdr_ptr->e_phnum 885 ehdr_ptr->e_phnum = ehdr_ptr->e_phnum - nr_ptnote + 1;
1090 886
1091 /* Store the size of all notes. We n <<
1092 * header when the device dumps will <<
1093 */ <<
1094 elfnotes_orig_sz = phdr.p_memsz; <<
1095 <<
1096 return 0; 887 return 0;
1097 } 888 }
1098 889
1099 /* Add memory chunks represented by program h 890 /* Add memory chunks represented by program headers to vmcore list. Also update
1100 * the new offset fields of exported program 891 * the new offset fields of exported program headers. */
1101 static int __init process_ptload_program_head 892 static int __init process_ptload_program_headers_elf64(char *elfptr,
1102 893 size_t elfsz,
1103 894 size_t elfnotes_sz,
1104 895 struct list_head *vc_list)
1105 { 896 {
1106 int i; 897 int i;
1107 Elf64_Ehdr *ehdr_ptr; 898 Elf64_Ehdr *ehdr_ptr;
1108 Elf64_Phdr *phdr_ptr; 899 Elf64_Phdr *phdr_ptr;
1109 loff_t vmcore_off; 900 loff_t vmcore_off;
1110 struct vmcore *new; 901 struct vmcore *new;
1111 902
1112 ehdr_ptr = (Elf64_Ehdr *)elfptr; 903 ehdr_ptr = (Elf64_Ehdr *)elfptr;
1113 phdr_ptr = (Elf64_Phdr*)(elfptr + siz 904 phdr_ptr = (Elf64_Phdr*)(elfptr + sizeof(Elf64_Ehdr)); /* PT_NOTE hdr */
1114 905
1115 /* Skip ELF header, program headers a !! 906 /* Skip Elf header, program headers and Elf note segment. */
1116 vmcore_off = elfsz + elfnotes_sz; 907 vmcore_off = elfsz + elfnotes_sz;
1117 908
1118 for (i = 0; i < ehdr_ptr->e_phnum; i+ 909 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
1119 u64 paddr, start, end, size; 910 u64 paddr, start, end, size;
1120 911
1121 if (phdr_ptr->p_type != PT_LO 912 if (phdr_ptr->p_type != PT_LOAD)
1122 continue; 913 continue;
1123 914
1124 paddr = phdr_ptr->p_offset; 915 paddr = phdr_ptr->p_offset;
1125 start = rounddown(paddr, PAGE 916 start = rounddown(paddr, PAGE_SIZE);
1126 end = roundup(paddr + phdr_pt 917 end = roundup(paddr + phdr_ptr->p_memsz, PAGE_SIZE);
1127 size = end - start; 918 size = end - start;
1128 919
1129 /* Add this contiguous chunk 920 /* Add this contiguous chunk of memory to vmcore list.*/
1130 new = get_new_element(); 921 new = get_new_element();
1131 if (!new) 922 if (!new)
1132 return -ENOMEM; 923 return -ENOMEM;
1133 new->paddr = start; 924 new->paddr = start;
1134 new->size = size; 925 new->size = size;
1135 list_add_tail(&new->list, vc_ 926 list_add_tail(&new->list, vc_list);
1136 927
1137 /* Update the program header 928 /* Update the program header offset. */
1138 phdr_ptr->p_offset = vmcore_o 929 phdr_ptr->p_offset = vmcore_off + (paddr - start);
1139 vmcore_off = vmcore_off + siz 930 vmcore_off = vmcore_off + size;
1140 } 931 }
1141 return 0; 932 return 0;
1142 } 933 }
1143 934
1144 static int __init process_ptload_program_head 935 static int __init process_ptload_program_headers_elf32(char *elfptr,
1145 936 size_t elfsz,
1146 937 size_t elfnotes_sz,
1147 938 struct list_head *vc_list)
1148 { 939 {
1149 int i; 940 int i;
1150 Elf32_Ehdr *ehdr_ptr; 941 Elf32_Ehdr *ehdr_ptr;
1151 Elf32_Phdr *phdr_ptr; 942 Elf32_Phdr *phdr_ptr;
1152 loff_t vmcore_off; 943 loff_t vmcore_off;
1153 struct vmcore *new; 944 struct vmcore *new;
1154 945
1155 ehdr_ptr = (Elf32_Ehdr *)elfptr; 946 ehdr_ptr = (Elf32_Ehdr *)elfptr;
1156 phdr_ptr = (Elf32_Phdr*)(elfptr + siz 947 phdr_ptr = (Elf32_Phdr*)(elfptr + sizeof(Elf32_Ehdr)); /* PT_NOTE hdr */
1157 948
1158 /* Skip ELF header, program headers a !! 949 /* Skip Elf header, program headers and Elf note segment. */
1159 vmcore_off = elfsz + elfnotes_sz; 950 vmcore_off = elfsz + elfnotes_sz;
1160 951
1161 for (i = 0; i < ehdr_ptr->e_phnum; i+ 952 for (i = 0; i < ehdr_ptr->e_phnum; i++, phdr_ptr++) {
1162 u64 paddr, start, end, size; 953 u64 paddr, start, end, size;
1163 954
1164 if (phdr_ptr->p_type != PT_LO 955 if (phdr_ptr->p_type != PT_LOAD)
1165 continue; 956 continue;
1166 957
1167 paddr = phdr_ptr->p_offset; 958 paddr = phdr_ptr->p_offset;
1168 start = rounddown(paddr, PAGE 959 start = rounddown(paddr, PAGE_SIZE);
1169 end = roundup(paddr + phdr_pt 960 end = roundup(paddr + phdr_ptr->p_memsz, PAGE_SIZE);
1170 size = end - start; 961 size = end - start;
1171 962
1172 /* Add this contiguous chunk 963 /* Add this contiguous chunk of memory to vmcore list.*/
1173 new = get_new_element(); 964 new = get_new_element();
1174 if (!new) 965 if (!new)
1175 return -ENOMEM; 966 return -ENOMEM;
1176 new->paddr = start; 967 new->paddr = start;
1177 new->size = size; 968 new->size = size;
1178 list_add_tail(&new->list, vc_ 969 list_add_tail(&new->list, vc_list);
1179 970
1180 /* Update the program header 971 /* Update the program header offset */
1181 phdr_ptr->p_offset = vmcore_o 972 phdr_ptr->p_offset = vmcore_off + (paddr - start);
1182 vmcore_off = vmcore_off + siz 973 vmcore_off = vmcore_off + size;
1183 } 974 }
1184 return 0; 975 return 0;
1185 } 976 }
1186 977
1187 /* Sets offset fields of vmcore elements. */ 978 /* Sets offset fields of vmcore elements. */
1188 static void set_vmcore_list_offsets(size_t el !! 979 static void __init set_vmcore_list_offsets(size_t elfsz, size_t elfnotes_sz,
1189 struct li !! 980 struct list_head *vc_list)
1190 { 981 {
1191 loff_t vmcore_off; 982 loff_t vmcore_off;
1192 struct vmcore *m; 983 struct vmcore *m;
1193 984
1194 /* Skip ELF header, program headers a !! 985 /* Skip Elf header, program headers and Elf note segment. */
1195 vmcore_off = elfsz + elfnotes_sz; 986 vmcore_off = elfsz + elfnotes_sz;
1196 987
1197 list_for_each_entry(m, vc_list, list) 988 list_for_each_entry(m, vc_list, list) {
1198 m->offset = vmcore_off; 989 m->offset = vmcore_off;
1199 vmcore_off += m->size; 990 vmcore_off += m->size;
1200 } 991 }
1201 } 992 }
1202 993
1203 static void free_elfcorebuf(void) 994 static void free_elfcorebuf(void)
1204 { 995 {
1205 free_pages((unsigned long)elfcorebuf, 996 free_pages((unsigned long)elfcorebuf, get_order(elfcorebuf_sz_orig));
1206 elfcorebuf = NULL; 997 elfcorebuf = NULL;
1207 vfree(elfnotes_buf); 998 vfree(elfnotes_buf);
1208 elfnotes_buf = NULL; 999 elfnotes_buf = NULL;
1209 } 1000 }
1210 1001
1211 static int __init parse_crash_elf64_headers(v 1002 static int __init parse_crash_elf64_headers(void)
1212 { 1003 {
1213 int rc=0; 1004 int rc=0;
1214 Elf64_Ehdr ehdr; 1005 Elf64_Ehdr ehdr;
1215 u64 addr; 1006 u64 addr;
1216 1007
1217 addr = elfcorehdr_addr; 1008 addr = elfcorehdr_addr;
1218 1009
1219 /* Read ELF header */ !! 1010 /* Read Elf header */
1220 rc = elfcorehdr_read((char *)&ehdr, s 1011 rc = elfcorehdr_read((char *)&ehdr, sizeof(Elf64_Ehdr), &addr);
1221 if (rc < 0) 1012 if (rc < 0)
1222 return rc; 1013 return rc;
1223 1014
1224 /* Do some basic Verification. */ 1015 /* Do some basic Verification. */
1225 if (memcmp(ehdr.e_ident, ELFMAG, SELF 1016 if (memcmp(ehdr.e_ident, ELFMAG, SELFMAG) != 0 ||
1226 (ehdr.e_type != ET_CORE) || 1017 (ehdr.e_type != ET_CORE) ||
1227 !vmcore_elf64_check_arch(&ehd 1018 !vmcore_elf64_check_arch(&ehdr) ||
1228 ehdr.e_ident[EI_CLASS] != ELF 1019 ehdr.e_ident[EI_CLASS] != ELFCLASS64 ||
1229 ehdr.e_ident[EI_VERSION] != E 1020 ehdr.e_ident[EI_VERSION] != EV_CURRENT ||
1230 ehdr.e_version != EV_CURRENT 1021 ehdr.e_version != EV_CURRENT ||
1231 ehdr.e_ehsize != sizeof(Elf64 1022 ehdr.e_ehsize != sizeof(Elf64_Ehdr) ||
1232 ehdr.e_phentsize != sizeof(El 1023 ehdr.e_phentsize != sizeof(Elf64_Phdr) ||
1233 ehdr.e_phnum == 0) { 1024 ehdr.e_phnum == 0) {
1234 pr_warn("Warning: Core image 1025 pr_warn("Warning: Core image elf header is not sane\n");
1235 return -EINVAL; 1026 return -EINVAL;
1236 } 1027 }
1237 1028
1238 /* Read in all elf headers. */ 1029 /* Read in all elf headers. */
1239 elfcorebuf_sz_orig = sizeof(Elf64_Ehd 1030 elfcorebuf_sz_orig = sizeof(Elf64_Ehdr) +
1240 ehdr.e_phnum 1031 ehdr.e_phnum * sizeof(Elf64_Phdr);
1241 elfcorebuf_sz = elfcorebuf_sz_orig; 1032 elfcorebuf_sz = elfcorebuf_sz_orig;
1242 elfcorebuf = (void *)__get_free_pages 1033 elfcorebuf = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1243 1034 get_order(elfcorebuf_sz_orig));
1244 if (!elfcorebuf) 1035 if (!elfcorebuf)
1245 return -ENOMEM; 1036 return -ENOMEM;
1246 addr = elfcorehdr_addr; 1037 addr = elfcorehdr_addr;
1247 rc = elfcorehdr_read(elfcorebuf, elfc 1038 rc = elfcorehdr_read(elfcorebuf, elfcorebuf_sz_orig, &addr);
1248 if (rc < 0) 1039 if (rc < 0)
1249 goto fail; 1040 goto fail;
1250 1041
1251 /* Merge all PT_NOTE headers into one 1042 /* Merge all PT_NOTE headers into one. */
1252 rc = merge_note_headers_elf64(elfcore 1043 rc = merge_note_headers_elf64(elfcorebuf, &elfcorebuf_sz,
1253 &elfnot 1044 &elfnotes_buf, &elfnotes_sz);
1254 if (rc) 1045 if (rc)
1255 goto fail; 1046 goto fail;
1256 rc = process_ptload_program_headers_e 1047 rc = process_ptload_program_headers_elf64(elfcorebuf, elfcorebuf_sz,
1257 1048 elfnotes_sz, &vmcore_list);
1258 if (rc) 1049 if (rc)
1259 goto fail; 1050 goto fail;
1260 set_vmcore_list_offsets(elfcorebuf_sz 1051 set_vmcore_list_offsets(elfcorebuf_sz, elfnotes_sz, &vmcore_list);
1261 return 0; 1052 return 0;
1262 fail: 1053 fail:
1263 free_elfcorebuf(); 1054 free_elfcorebuf();
1264 return rc; 1055 return rc;
1265 } 1056 }
1266 1057
1267 static int __init parse_crash_elf32_headers(v 1058 static int __init parse_crash_elf32_headers(void)
1268 { 1059 {
1269 int rc=0; 1060 int rc=0;
1270 Elf32_Ehdr ehdr; 1061 Elf32_Ehdr ehdr;
1271 u64 addr; 1062 u64 addr;
1272 1063
1273 addr = elfcorehdr_addr; 1064 addr = elfcorehdr_addr;
1274 1065
1275 /* Read ELF header */ !! 1066 /* Read Elf header */
1276 rc = elfcorehdr_read((char *)&ehdr, s 1067 rc = elfcorehdr_read((char *)&ehdr, sizeof(Elf32_Ehdr), &addr);
1277 if (rc < 0) 1068 if (rc < 0)
1278 return rc; 1069 return rc;
1279 1070
1280 /* Do some basic Verification. */ 1071 /* Do some basic Verification. */
1281 if (memcmp(ehdr.e_ident, ELFMAG, SELF 1072 if (memcmp(ehdr.e_ident, ELFMAG, SELFMAG) != 0 ||
1282 (ehdr.e_type != ET_CORE) || 1073 (ehdr.e_type != ET_CORE) ||
1283 !vmcore_elf32_check_arch(&ehd 1074 !vmcore_elf32_check_arch(&ehdr) ||
1284 ehdr.e_ident[EI_CLASS] != ELF 1075 ehdr.e_ident[EI_CLASS] != ELFCLASS32||
1285 ehdr.e_ident[EI_VERSION] != E 1076 ehdr.e_ident[EI_VERSION] != EV_CURRENT ||
1286 ehdr.e_version != EV_CURRENT 1077 ehdr.e_version != EV_CURRENT ||
1287 ehdr.e_ehsize != sizeof(Elf32 1078 ehdr.e_ehsize != sizeof(Elf32_Ehdr) ||
1288 ehdr.e_phentsize != sizeof(El 1079 ehdr.e_phentsize != sizeof(Elf32_Phdr) ||
1289 ehdr.e_phnum == 0) { 1080 ehdr.e_phnum == 0) {
1290 pr_warn("Warning: Core image 1081 pr_warn("Warning: Core image elf header is not sane\n");
1291 return -EINVAL; 1082 return -EINVAL;
1292 } 1083 }
1293 1084
1294 /* Read in all elf headers. */ 1085 /* Read in all elf headers. */
1295 elfcorebuf_sz_orig = sizeof(Elf32_Ehd 1086 elfcorebuf_sz_orig = sizeof(Elf32_Ehdr) + ehdr.e_phnum * sizeof(Elf32_Phdr);
1296 elfcorebuf_sz = elfcorebuf_sz_orig; 1087 elfcorebuf_sz = elfcorebuf_sz_orig;
1297 elfcorebuf = (void *)__get_free_pages 1088 elfcorebuf = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1298 1089 get_order(elfcorebuf_sz_orig));
1299 if (!elfcorebuf) 1090 if (!elfcorebuf)
1300 return -ENOMEM; 1091 return -ENOMEM;
1301 addr = elfcorehdr_addr; 1092 addr = elfcorehdr_addr;
1302 rc = elfcorehdr_read(elfcorebuf, elfc 1093 rc = elfcorehdr_read(elfcorebuf, elfcorebuf_sz_orig, &addr);
1303 if (rc < 0) 1094 if (rc < 0)
1304 goto fail; 1095 goto fail;
1305 1096
1306 /* Merge all PT_NOTE headers into one 1097 /* Merge all PT_NOTE headers into one. */
1307 rc = merge_note_headers_elf32(elfcore 1098 rc = merge_note_headers_elf32(elfcorebuf, &elfcorebuf_sz,
1308 &elfnot 1099 &elfnotes_buf, &elfnotes_sz);
1309 if (rc) 1100 if (rc)
1310 goto fail; 1101 goto fail;
1311 rc = process_ptload_program_headers_e 1102 rc = process_ptload_program_headers_elf32(elfcorebuf, elfcorebuf_sz,
1312 1103 elfnotes_sz, &vmcore_list);
1313 if (rc) 1104 if (rc)
1314 goto fail; 1105 goto fail;
1315 set_vmcore_list_offsets(elfcorebuf_sz 1106 set_vmcore_list_offsets(elfcorebuf_sz, elfnotes_sz, &vmcore_list);
1316 return 0; 1107 return 0;
1317 fail: 1108 fail:
1318 free_elfcorebuf(); 1109 free_elfcorebuf();
1319 return rc; 1110 return rc;
1320 } 1111 }
1321 1112
1322 static int __init parse_crash_elf_headers(voi 1113 static int __init parse_crash_elf_headers(void)
1323 { 1114 {
1324 unsigned char e_ident[EI_NIDENT]; 1115 unsigned char e_ident[EI_NIDENT];
1325 u64 addr; 1116 u64 addr;
1326 int rc=0; 1117 int rc=0;
1327 1118
1328 addr = elfcorehdr_addr; 1119 addr = elfcorehdr_addr;
1329 rc = elfcorehdr_read(e_ident, EI_NIDE 1120 rc = elfcorehdr_read(e_ident, EI_NIDENT, &addr);
1330 if (rc < 0) 1121 if (rc < 0)
1331 return rc; 1122 return rc;
1332 if (memcmp(e_ident, ELFMAG, SELFMAG) 1123 if (memcmp(e_ident, ELFMAG, SELFMAG) != 0) {
1333 pr_warn("Warning: Core image 1124 pr_warn("Warning: Core image elf header not found\n");
1334 return -EINVAL; 1125 return -EINVAL;
1335 } 1126 }
1336 1127
1337 if (e_ident[EI_CLASS] == ELFCLASS64) 1128 if (e_ident[EI_CLASS] == ELFCLASS64) {
1338 rc = parse_crash_elf64_header 1129 rc = parse_crash_elf64_headers();
1339 if (rc) 1130 if (rc)
1340 return rc; 1131 return rc;
1341 } else if (e_ident[EI_CLASS] == ELFCL 1132 } else if (e_ident[EI_CLASS] == ELFCLASS32) {
1342 rc = parse_crash_elf32_header 1133 rc = parse_crash_elf32_headers();
1343 if (rc) 1134 if (rc)
1344 return rc; 1135 return rc;
1345 } else { 1136 } else {
1346 pr_warn("Warning: Core image 1137 pr_warn("Warning: Core image elf header is not sane\n");
1347 return -EINVAL; 1138 return -EINVAL;
1348 } 1139 }
1349 1140
1350 /* Determine vmcore size. */ 1141 /* Determine vmcore size. */
1351 vmcore_size = get_vmcore_size(elfcore 1142 vmcore_size = get_vmcore_size(elfcorebuf_sz, elfnotes_sz,
1352 &vmcore 1143 &vmcore_list);
1353 1144
1354 return 0; 1145 return 0;
1355 } 1146 }
1356 1147
1357 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP <<
1358 /** <<
1359 * vmcoredd_write_header - Write vmcore devic <<
1360 * beginning of the dump's buffer. <<
1361 * @buf: Output buffer where the note is writ <<
1362 * @data: Dump info <<
1363 * @size: Size of the dump <<
1364 * <<
1365 * Fills beginning of the dump's buffer with <<
1366 */ <<
1367 static void vmcoredd_write_header(void *buf, <<
1368 u32 size) <<
1369 { <<
1370 struct vmcoredd_header *vdd_hdr = (st <<
1371 <<
1372 vdd_hdr->n_namesz = sizeof(vdd_hdr->n <<
1373 vdd_hdr->n_descsz = size + sizeof(vdd <<
1374 vdd_hdr->n_type = NT_VMCOREDD; <<
1375 <<
1376 strscpy_pad(vdd_hdr->name, VMCOREDD_N <<
1377 strscpy_pad(vdd_hdr->dump_name, data- <<
1378 } <<
1379 <<
1380 /** <<
1381 * vmcoredd_update_program_headers - Update a <<
1382 * @elfptr: Pointer to elf header <<
1383 * @elfnotesz: Size of elf notes aligned to p <<
1384 * @vmcoreddsz: Size of device dumps to be ad <<
1385 * <<
1386 * Determine type of ELF header (Elf64 or Elf <<
1387 * Also update the offsets of all the program <<
1388 */ <<
1389 static void vmcoredd_update_program_headers(c <<
1390 s <<
1391 { <<
1392 unsigned char *e_ident = (unsigned ch <<
1393 u64 start, end, size; <<
1394 loff_t vmcore_off; <<
1395 u32 i; <<
1396 <<
1397 vmcore_off = elfcorebuf_sz + elfnotes <<
1398 <<
1399 if (e_ident[EI_CLASS] == ELFCLASS64) <<
1400 Elf64_Ehdr *ehdr = (Elf64_Ehd <<
1401 Elf64_Phdr *phdr = (Elf64_Phd <<
1402 <<
1403 /* Update all program headers <<
1404 for (i = 0; i < ehdr->e_phnum <<
1405 if (phdr->p_type == P <<
1406 /* Update not <<
1407 phdr->p_memsz <<
1408 phdr->p_files <<
1409 continue; <<
1410 } <<
1411 <<
1412 start = rounddown(phd <<
1413 end = roundup(phdr->p <<
1414 PAGE_SI <<
1415 size = end - start; <<
1416 phdr->p_offset = vmco <<
1417 vmcore_off += size; <<
1418 } <<
1419 } else { <<
1420 Elf32_Ehdr *ehdr = (Elf32_Ehd <<
1421 Elf32_Phdr *phdr = (Elf32_Phd <<
1422 <<
1423 /* Update all program headers <<
1424 for (i = 0; i < ehdr->e_phnum <<
1425 if (phdr->p_type == P <<
1426 /* Update not <<
1427 phdr->p_memsz <<
1428 phdr->p_files <<
1429 continue; <<
1430 } <<
1431 <<
1432 start = rounddown(phd <<
1433 end = roundup(phdr->p <<
1434 PAGE_SI <<
1435 size = end - start; <<
1436 phdr->p_offset = vmco <<
1437 vmcore_off += size; <<
1438 } <<
1439 } <<
1440 } <<
1441 <<
1442 /** <<
1443 * vmcoredd_update_size - Update the total si <<
1444 * ELF header <<
1445 * @dump_size: Size of the current device dum <<
1446 * <<
1447 * Update the total size of all the device du <<
1448 * headers. Calculate the new offsets for the <<
1449 * total vmcore size. <<
1450 */ <<
1451 static void vmcoredd_update_size(size_t dump_ <<
1452 { <<
1453 vmcoredd_orig_sz += dump_size; <<
1454 elfnotes_sz = roundup(elfnotes_orig_s <<
1455 vmcoredd_update_program_headers(elfco <<
1456 vmcor <<
1457 <<
1458 /* Update vmcore list offsets */ <<
1459 set_vmcore_list_offsets(elfcorebuf_sz <<
1460 <<
1461 vmcore_size = get_vmcore_size(elfcore <<
1462 &vmcore <<
1463 proc_vmcore->size = vmcore_size; <<
1464 } <<
1465 <<
1466 /** <<
1467 * vmcore_add_device_dump - Add a buffer cont <<
1468 * @data: dump info. <<
1469 * <<
1470 * Allocate a buffer and invoke the calling d <<
1471 * Write ELF note at the beginning of the buf <<
1472 * dump and add the dump to global list. <<
1473 */ <<
1474 int vmcore_add_device_dump(struct vmcoredd_da <<
1475 { <<
1476 struct vmcoredd_node *dump; <<
1477 void *buf = NULL; <<
1478 size_t data_size; <<
1479 int ret; <<
1480 <<
1481 if (vmcoredd_disabled) { <<
1482 pr_err_once("Device dump is d <<
1483 return -EINVAL; <<
1484 } <<
1485 <<
1486 if (!data || !strlen(data->dump_name) <<
1487 !data->vmcoredd_callback || !data <<
1488 return -EINVAL; <<
1489 <<
1490 dump = vzalloc(sizeof(*dump)); <<
1491 if (!dump) { <<
1492 ret = -ENOMEM; <<
1493 goto out_err; <<
1494 } <<
1495 <<
1496 /* Keep size of the buffer page align <<
1497 data_size = roundup(sizeof(struct vmc <<
1498 PAGE_SIZE); <<
1499 <<
1500 /* Allocate buffer for driver's to wr <<
1501 buf = vmcore_alloc_buf(data_size); <<
1502 if (!buf) { <<
1503 ret = -ENOMEM; <<
1504 goto out_err; <<
1505 } <<
1506 <<
1507 vmcoredd_write_header(buf, data, data <<
1508 sizeof(struct v <<
1509 <<
1510 /* Invoke the driver's dump collectio <<
1511 ret = data->vmcoredd_callback(data, b <<
1512 sizeof( <<
1513 if (ret) <<
1514 goto out_err; <<
1515 <<
1516 dump->buf = buf; <<
1517 dump->size = data_size; <<
1518 <<
1519 /* Add the dump to driver sysfs list <<
1520 mutex_lock(&vmcoredd_mutex); <<
1521 list_add_tail(&dump->list, &vmcoredd_ <<
1522 mutex_unlock(&vmcoredd_mutex); <<
1523 <<
1524 vmcoredd_update_size(data_size); <<
1525 return 0; <<
1526 <<
1527 out_err: <<
1528 vfree(buf); <<
1529 vfree(dump); <<
1530 <<
1531 return ret; <<
1532 } <<
1533 EXPORT_SYMBOL(vmcore_add_device_dump); <<
1534 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
1535 <<
1536 /* Free all dumps in vmcore device dump list <<
1537 static void vmcore_free_device_dumps(void) <<
1538 { <<
1539 #ifdef CONFIG_PROC_VMCORE_DEVICE_DUMP <<
1540 mutex_lock(&vmcoredd_mutex); <<
1541 while (!list_empty(&vmcoredd_list)) { <<
1542 struct vmcoredd_node *dump; <<
1543 <<
1544 dump = list_first_entry(&vmco <<
1545 list) <<
1546 list_del(&dump->list); <<
1547 vfree(dump->buf); <<
1548 vfree(dump); <<
1549 } <<
1550 mutex_unlock(&vmcoredd_mutex); <<
1551 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */ <<
1552 } <<
1553 <<
1554 /* Init function for vmcore module. */ 1148 /* Init function for vmcore module. */
1555 static int __init vmcore_init(void) 1149 static int __init vmcore_init(void)
1556 { 1150 {
1557 int rc = 0; 1151 int rc = 0;
1558 1152
1559 /* Allow architectures to allocate EL 1153 /* Allow architectures to allocate ELF header in 2nd kernel */
1560 rc = elfcorehdr_alloc(&elfcorehdr_add 1154 rc = elfcorehdr_alloc(&elfcorehdr_addr, &elfcorehdr_size);
1561 if (rc) 1155 if (rc)
1562 return rc; 1156 return rc;
1563 /* 1157 /*
1564 * If elfcorehdr= has been passed in 1158 * If elfcorehdr= has been passed in cmdline or created in 2nd kernel,
1565 * then capture the dump. 1159 * then capture the dump.
1566 */ 1160 */
1567 if (!(is_vmcore_usable())) 1161 if (!(is_vmcore_usable()))
1568 return rc; 1162 return rc;
1569 rc = parse_crash_elf_headers(); 1163 rc = parse_crash_elf_headers();
1570 if (rc) { 1164 if (rc) {
1571 elfcorehdr_free(elfcorehdr_ad <<
1572 pr_warn("Kdump: vmcore not in 1165 pr_warn("Kdump: vmcore not initialized\n");
1573 return rc; 1166 return rc;
1574 } 1167 }
1575 elfcorehdr_free(elfcorehdr_addr); 1168 elfcorehdr_free(elfcorehdr_addr);
1576 elfcorehdr_addr = ELFCORE_ADDR_ERR; 1169 elfcorehdr_addr = ELFCORE_ADDR_ERR;
1577 1170
1578 proc_vmcore = proc_create("vmcore", S !! 1171 proc_vmcore = proc_create("vmcore", S_IRUSR, NULL, &proc_vmcore_operations);
1579 if (proc_vmcore) 1172 if (proc_vmcore)
1580 proc_vmcore->size = vmcore_si 1173 proc_vmcore->size = vmcore_size;
1581 return 0; 1174 return 0;
1582 } 1175 }
1583 fs_initcall(vmcore_init); 1176 fs_initcall(vmcore_init);
1584 1177
1585 /* Cleanup function for vmcore module. */ 1178 /* Cleanup function for vmcore module. */
1586 void vmcore_cleanup(void) 1179 void vmcore_cleanup(void)
1587 { 1180 {
>> 1181 struct list_head *pos, *next;
>> 1182
1588 if (proc_vmcore) { 1183 if (proc_vmcore) {
1589 proc_remove(proc_vmcore); 1184 proc_remove(proc_vmcore);
1590 proc_vmcore = NULL; 1185 proc_vmcore = NULL;
1591 } 1186 }
1592 1187
1593 /* clear the vmcore list. */ 1188 /* clear the vmcore list. */
1594 while (!list_empty(&vmcore_list)) { !! 1189 list_for_each_safe(pos, next, &vmcore_list) {
1595 struct vmcore *m; 1190 struct vmcore *m;
1596 1191
1597 m = list_first_entry(&vmcore_ !! 1192 m = list_entry(pos, struct vmcore, list);
1598 list_del(&m->list); 1193 list_del(&m->list);
1599 kfree(m); 1194 kfree(m);
1600 } 1195 }
1601 free_elfcorebuf(); 1196 free_elfcorebuf();
1602 <<
1603 /* clear vmcore device dump list */ <<
1604 vmcore_free_device_dumps(); <<
1605 } 1197 }
1606 1198
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.