aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/kexec.c
diff options
context:
space:
mode:
authorDave Young <dyoung@redhat.com>2015-09-09 18:38:51 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2015-09-10 16:29:01 -0400
commita43cac0d9dc2073ff2245a171429ddbe1accece7 (patch)
tree2d4c4e98d70b12a451d5b790cc6a369c7eefe85b /kernel/kexec.c
parenta202fbbf56e819de83876827c4bf5da2bfbac5ec (diff)
kexec: split kexec_file syscall code to kexec_file.c
Split kexec_file syscall related code to another file kernel/kexec_file.c so that the #ifdef CONFIG_KEXEC_FILE in kexec.c can be dropped. Sharing variables and functions are moved to kernel/kexec_internal.h per suggestion from Vivek and Petr. [akpm@linux-foundation.org: fix bisectability] [akpm@linux-foundation.org: declare the various arch_kexec functions] [akpm@linux-foundation.org: fix build] Signed-off-by: Dave Young <dyoung@redhat.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Petr Tesarik <ptesarik@suse.cz> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Josh Boyer <jwboyer@fedoraproject.org> Cc: David Howells <dhowells@redhat.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'kernel/kexec.c')
-rw-r--r--kernel/kexec.c1056
1 files changed, 11 insertions, 1045 deletions
diff --git a/kernel/kexec.c b/kernel/kexec.c
index a785c1015e25..2d73ecfa5505 100644
--- a/kernel/kexec.c
+++ b/kernel/kexec.c
@@ -28,10 +28,10 @@
28#include <linux/suspend.h> 28#include <linux/suspend.h>
29#include <linux/device.h> 29#include <linux/device.h>
30#include <linux/freezer.h> 30#include <linux/freezer.h>
31#include <linux/vmalloc.h>
31#include <linux/pm.h> 32#include <linux/pm.h>
32#include <linux/cpu.h> 33#include <linux/cpu.h>
33#include <linux/console.h> 34#include <linux/console.h>
34#include <linux/vmalloc.h>
35#include <linux/swap.h> 35#include <linux/swap.h>
36#include <linux/syscore_ops.h> 36#include <linux/syscore_ops.h>
37#include <linux/compiler.h> 37#include <linux/compiler.h>
@@ -44,6 +44,9 @@
44 44
45#include <crypto/hash.h> 45#include <crypto/hash.h>
46#include <crypto/sha.h> 46#include <crypto/sha.h>
47#include "kexec_internal.h"
48
49DEFINE_MUTEX(kexec_mutex);
47 50
48/* Per cpu memory for storing cpu states in case of system crash. */ 51/* Per cpu memory for storing cpu states in case of system crash. */
49note_buf_t __percpu *crash_notes; 52note_buf_t __percpu *crash_notes;
@@ -57,16 +60,6 @@ size_t vmcoreinfo_max_size = sizeof(vmcoreinfo_data);
57/* Flag to indicate we are going to kexec a new kernel */ 60/* Flag to indicate we are going to kexec a new kernel */
58bool kexec_in_progress = false; 61bool kexec_in_progress = false;
59 62
60/*
61 * Declare these symbols weak so that if architecture provides a purgatory,
62 * these will be overridden.
63 */
64char __weak kexec_purgatory[0];
65size_t __weak kexec_purgatory_size = 0;
66
67#ifdef CONFIG_KEXEC_FILE
68static int kexec_calculate_store_digests(struct kimage *image);
69#endif
70 63
71/* Location of the reserved area for the crash kernel */ 64/* Location of the reserved area for the crash kernel */
72struct resource crashk_res = { 65struct resource crashk_res = {
@@ -146,8 +139,6 @@ int kexec_should_crash(struct task_struct *p)
146 */ 139 */
147#define KIMAGE_NO_DEST (-1UL) 140#define KIMAGE_NO_DEST (-1UL)
148 141
149static int kimage_is_destination_range(struct kimage *image,
150 unsigned long start, unsigned long end);
151static struct page *kimage_alloc_page(struct kimage *image, 142static struct page *kimage_alloc_page(struct kimage *image,
152 gfp_t gfp_mask, 143 gfp_t gfp_mask,
153 unsigned long dest); 144 unsigned long dest);
@@ -169,7 +160,7 @@ static int copy_user_segment_list(struct kimage *image,
169 return ret; 160 return ret;
170} 161}
171 162
172static int sanity_check_segment_list(struct kimage *image) 163int sanity_check_segment_list(struct kimage *image)
173{ 164{
174 int result, i; 165 int result, i;
175 unsigned long nr_segments = image->nr_segments; 166 unsigned long nr_segments = image->nr_segments;
@@ -259,7 +250,7 @@ static int sanity_check_segment_list(struct kimage *image)
259 return 0; 250 return 0;
260} 251}
261 252
262static struct kimage *do_kimage_alloc_init(void) 253struct kimage *do_kimage_alloc_init(void)
263{ 254{
264 struct kimage *image; 255 struct kimage *image;
265 256
@@ -286,8 +277,6 @@ static struct kimage *do_kimage_alloc_init(void)
286 return image; 277 return image;
287} 278}
288 279
289static void kimage_free_page_list(struct list_head *list);
290
291static int kimage_alloc_init(struct kimage **rimage, unsigned long entry, 280static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
292 unsigned long nr_segments, 281 unsigned long nr_segments,
293 struct kexec_segment __user *segments, 282 struct kexec_segment __user *segments,
@@ -354,283 +343,7 @@ out_free_image:
354 return ret; 343 return ret;
355} 344}
356 345
357#ifdef CONFIG_KEXEC_FILE 346int kimage_is_destination_range(struct kimage *image,
358static int copy_file_from_fd(int fd, void **buf, unsigned long *buf_len)
359{
360 struct fd f = fdget(fd);
361 int ret;
362 struct kstat stat;
363 loff_t pos;
364 ssize_t bytes = 0;
365
366 if (!f.file)
367 return -EBADF;
368
369 ret = vfs_getattr(&f.file->f_path, &stat);
370 if (ret)
371 goto out;
372
373 if (stat.size > INT_MAX) {
374 ret = -EFBIG;
375 goto out;
376 }
377
378 /* Don't hand 0 to vmalloc, it whines. */
379 if (stat.size == 0) {
380 ret = -EINVAL;
381 goto out;
382 }
383
384 *buf = vmalloc(stat.size);
385 if (!*buf) {
386 ret = -ENOMEM;
387 goto out;
388 }
389
390 pos = 0;
391 while (pos < stat.size) {
392 bytes = kernel_read(f.file, pos, (char *)(*buf) + pos,
393 stat.size - pos);
394 if (bytes < 0) {
395 vfree(*buf);
396 ret = bytes;
397 goto out;
398 }
399
400 if (bytes == 0)
401 break;
402 pos += bytes;
403 }
404
405 if (pos != stat.size) {
406 ret = -EBADF;
407 vfree(*buf);
408 goto out;
409 }
410
411 *buf_len = pos;
412out:
413 fdput(f);
414 return ret;
415}
416
417/* Architectures can provide this probe function */
418int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,
419 unsigned long buf_len)
420{
421 return -ENOEXEC;
422}
423
424void * __weak arch_kexec_kernel_image_load(struct kimage *image)
425{
426 return ERR_PTR(-ENOEXEC);
427}
428
429void __weak arch_kimage_file_post_load_cleanup(struct kimage *image)
430{
431}
432
433int __weak arch_kexec_kernel_verify_sig(struct kimage *image, void *buf,
434 unsigned long buf_len)
435{
436 return -EKEYREJECTED;
437}
438
439/* Apply relocations of type RELA */
440int __weak
441arch_kexec_apply_relocations_add(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
442 unsigned int relsec)
443{
444 pr_err("RELA relocation unsupported.\n");
445 return -ENOEXEC;
446}
447
448/* Apply relocations of type REL */
449int __weak
450arch_kexec_apply_relocations(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
451 unsigned int relsec)
452{
453 pr_err("REL relocation unsupported.\n");
454 return -ENOEXEC;
455}
456
457/*
458 * Free up memory used by kernel, initrd, and command line. This is temporary
459 * memory allocation which is not needed any more after these buffers have
460 * been loaded into separate segments and have been copied elsewhere.
461 */
462static void kimage_file_post_load_cleanup(struct kimage *image)
463{
464 struct purgatory_info *pi = &image->purgatory_info;
465
466 vfree(image->kernel_buf);
467 image->kernel_buf = NULL;
468
469 vfree(image->initrd_buf);
470 image->initrd_buf = NULL;
471
472 kfree(image->cmdline_buf);
473 image->cmdline_buf = NULL;
474
475 vfree(pi->purgatory_buf);
476 pi->purgatory_buf = NULL;
477
478 vfree(pi->sechdrs);
479 pi->sechdrs = NULL;
480
481 /* See if architecture has anything to cleanup post load */
482 arch_kimage_file_post_load_cleanup(image);
483
484 /*
485 * Above call should have called into bootloader to free up
486 * any data stored in kimage->image_loader_data. It should
487 * be ok now to free it up.
488 */
489 kfree(image->image_loader_data);
490 image->image_loader_data = NULL;
491}
492
493/*
494 * In file mode list of segments is prepared by kernel. Copy relevant
495 * data from user space, do error checking, prepare segment list
496 */
497static int
498kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,
499 const char __user *cmdline_ptr,
500 unsigned long cmdline_len, unsigned flags)
501{
502 int ret = 0;
503 void *ldata;
504
505 ret = copy_file_from_fd(kernel_fd, &image->kernel_buf,
506 &image->kernel_buf_len);
507 if (ret)
508 return ret;
509
510 /* Call arch image probe handlers */
511 ret = arch_kexec_kernel_image_probe(image, image->kernel_buf,
512 image->kernel_buf_len);
513
514 if (ret)
515 goto out;
516
517#ifdef CONFIG_KEXEC_VERIFY_SIG
518 ret = arch_kexec_kernel_verify_sig(image, image->kernel_buf,
519 image->kernel_buf_len);
520 if (ret) {
521 pr_debug("kernel signature verification failed.\n");
522 goto out;
523 }
524 pr_debug("kernel signature verification successful.\n");
525#endif
526 /* It is possible that there no initramfs is being loaded */
527 if (!(flags & KEXEC_FILE_NO_INITRAMFS)) {
528 ret = copy_file_from_fd(initrd_fd, &image->initrd_buf,
529 &image->initrd_buf_len);
530 if (ret)
531 goto out;
532 }
533
534 if (cmdline_len) {
535 image->cmdline_buf = kzalloc(cmdline_len, GFP_KERNEL);
536 if (!image->cmdline_buf) {
537 ret = -ENOMEM;
538 goto out;
539 }
540
541 ret = copy_from_user(image->cmdline_buf, cmdline_ptr,
542 cmdline_len);
543 if (ret) {
544 ret = -EFAULT;
545 goto out;
546 }
547
548 image->cmdline_buf_len = cmdline_len;
549
550 /* command line should be a string with last byte null */
551 if (image->cmdline_buf[cmdline_len - 1] != '\0') {
552 ret = -EINVAL;
553 goto out;
554 }
555 }
556
557 /* Call arch image load handlers */
558 ldata = arch_kexec_kernel_image_load(image);
559
560 if (IS_ERR(ldata)) {
561 ret = PTR_ERR(ldata);
562 goto out;
563 }
564
565 image->image_loader_data = ldata;
566out:
567 /* In case of error, free up all allocated memory in this function */
568 if (ret)
569 kimage_file_post_load_cleanup(image);
570 return ret;
571}
572
573static int
574kimage_file_alloc_init(struct kimage **rimage, int kernel_fd,
575 int initrd_fd, const char __user *cmdline_ptr,
576 unsigned long cmdline_len, unsigned long flags)
577{
578 int ret;
579 struct kimage *image;
580 bool kexec_on_panic = flags & KEXEC_FILE_ON_CRASH;
581
582 image = do_kimage_alloc_init();
583 if (!image)
584 return -ENOMEM;
585
586 image->file_mode = 1;
587
588 if (kexec_on_panic) {
589 /* Enable special crash kernel control page alloc policy. */
590 image->control_page = crashk_res.start;
591 image->type = KEXEC_TYPE_CRASH;
592 }
593
594 ret = kimage_file_prepare_segments(image, kernel_fd, initrd_fd,
595 cmdline_ptr, cmdline_len, flags);
596 if (ret)
597 goto out_free_image;
598
599 ret = sanity_check_segment_list(image);
600 if (ret)
601 goto out_free_post_load_bufs;
602
603 ret = -ENOMEM;
604 image->control_code_page = kimage_alloc_control_pages(image,
605 get_order(KEXEC_CONTROL_PAGE_SIZE));
606 if (!image->control_code_page) {
607 pr_err("Could not allocate control_code_buffer\n");
608 goto out_free_post_load_bufs;
609 }
610
611 if (!kexec_on_panic) {
612 image->swap_page = kimage_alloc_control_pages(image, 0);
613 if (!image->swap_page) {
614 pr_err("Could not allocate swap buffer\n");
615 goto out_free_control_pages;
616 }
617 }
618
619 *rimage = image;
620 return 0;
621out_free_control_pages:
622 kimage_free_page_list(&image->control_pages);
623out_free_post_load_bufs:
624 kimage_file_post_load_cleanup(image);
625out_free_image:
626 kfree(image);
627 return ret;
628}
629#else /* CONFIG_KEXEC_FILE */
630static inline void kimage_file_post_load_cleanup(struct kimage *image) { }
631#endif /* CONFIG_KEXEC_FILE */
632
633static int kimage_is_destination_range(struct kimage *image,
634 unsigned long start, 347 unsigned long start,
635 unsigned long end) 348 unsigned long end)
636{ 349{
@@ -676,7 +389,7 @@ static void kimage_free_pages(struct page *page)
676 __free_pages(page, order); 389 __free_pages(page, order);
677} 390}
678 391
679static void kimage_free_page_list(struct list_head *list) 392void kimage_free_page_list(struct list_head *list)
680{ 393{
681 struct list_head *pos, *next; 394 struct list_head *pos, *next;
682 395
@@ -892,7 +605,7 @@ static void kimage_free_extra_pages(struct kimage *image)
892 kimage_free_page_list(&image->unusable_pages); 605 kimage_free_page_list(&image->unusable_pages);
893 606
894} 607}
895static void kimage_terminate(struct kimage *image) 608void kimage_terminate(struct kimage *image)
896{ 609{
897 if (*image->entry != 0) 610 if (*image->entry != 0)
898 image->entry++; 611 image->entry++;
@@ -913,7 +626,7 @@ static void kimage_free_entry(kimage_entry_t entry)
913 kimage_free_pages(page); 626 kimage_free_pages(page);
914} 627}
915 628
916static void kimage_free(struct kimage *image) 629void kimage_free(struct kimage *image)
917{ 630{
918 kimage_entry_t *ptr, entry; 631 kimage_entry_t *ptr, entry;
919 kimage_entry_t ind = 0; 632 kimage_entry_t ind = 0;
@@ -1204,7 +917,7 @@ out:
1204 return result; 917 return result;
1205} 918}
1206 919
1207static int kimage_load_segment(struct kimage *image, 920int kimage_load_segment(struct kimage *image,
1208 struct kexec_segment *segment) 921 struct kexec_segment *segment)
1209{ 922{
1210 int result = -ENOMEM; 923 int result = -ENOMEM;
@@ -1245,8 +958,6 @@ struct kimage *kexec_image;
1245struct kimage *kexec_crash_image; 958struct kimage *kexec_crash_image;
1246int kexec_load_disabled; 959int kexec_load_disabled;
1247 960
1248static DEFINE_MUTEX(kexec_mutex);
1249
1250SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments, 961SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
1251 struct kexec_segment __user *, segments, unsigned long, flags) 962 struct kexec_segment __user *, segments, unsigned long, flags)
1252{ 963{
@@ -1391,85 +1102,6 @@ COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
1391} 1102}
1392#endif 1103#endif
1393 1104
1394#ifdef CONFIG_KEXEC_FILE
1395SYSCALL_DEFINE5(kexec_file_load, int, kernel_fd, int, initrd_fd,
1396 unsigned long, cmdline_len, const char __user *, cmdline_ptr,
1397 unsigned long, flags)
1398{
1399 int ret = 0, i;
1400 struct kimage **dest_image, *image;
1401
1402 /* We only trust the superuser with rebooting the system. */
1403 if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
1404 return -EPERM;
1405
1406 /* Make sure we have a legal set of flags */
1407 if (flags != (flags & KEXEC_FILE_FLAGS))
1408 return -EINVAL;
1409
1410 image = NULL;
1411
1412 if (!mutex_trylock(&kexec_mutex))
1413 return -EBUSY;
1414
1415 dest_image = &kexec_image;
1416 if (flags & KEXEC_FILE_ON_CRASH)
1417 dest_image = &kexec_crash_image;
1418
1419 if (flags & KEXEC_FILE_UNLOAD)
1420 goto exchange;
1421
1422 /*
1423 * In case of crash, new kernel gets loaded in reserved region. It is
1424 * same memory where old crash kernel might be loaded. Free any
1425 * current crash dump kernel before we corrupt it.
1426 */
1427 if (flags & KEXEC_FILE_ON_CRASH)
1428 kimage_free(xchg(&kexec_crash_image, NULL));
1429
1430 ret = kimage_file_alloc_init(&image, kernel_fd, initrd_fd, cmdline_ptr,
1431 cmdline_len, flags);
1432 if (ret)
1433 goto out;
1434
1435 ret = machine_kexec_prepare(image);
1436 if (ret)
1437 goto out;
1438
1439 ret = kexec_calculate_store_digests(image);
1440 if (ret)
1441 goto out;
1442
1443 for (i = 0; i < image->nr_segments; i++) {
1444 struct kexec_segment *ksegment;
1445
1446 ksegment = &image->segment[i];
1447 pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
1448 i, ksegment->buf, ksegment->bufsz, ksegment->mem,
1449 ksegment->memsz);
1450
1451 ret = kimage_load_segment(image, &image->segment[i]);
1452 if (ret)
1453 goto out;
1454 }
1455
1456 kimage_terminate(image);
1457
1458 /*
1459 * Free up any temporary buffers allocated which are not needed
1460 * after image has been loaded
1461 */
1462 kimage_file_post_load_cleanup(image);
1463exchange:
1464 image = xchg(dest_image, image);
1465out:
1466 mutex_unlock(&kexec_mutex);
1467 kimage_free(image);
1468 return ret;
1469}
1470
1471#endif /* CONFIG_KEXEC_FILE */
1472
1473void crash_kexec(struct pt_regs *regs) 1105void crash_kexec(struct pt_regs *regs)
1474{ 1106{
1475 /* Take the kexec_mutex here to prevent sys_kexec_load 1107 /* Take the kexec_mutex here to prevent sys_kexec_load
@@ -2024,672 +1656,6 @@ static int __init crash_save_vmcoreinfo_init(void)
2024 1656
2025subsys_initcall(crash_save_vmcoreinfo_init); 1657subsys_initcall(crash_save_vmcoreinfo_init);
2026 1658
2027#ifdef CONFIG_KEXEC_FILE
2028static int locate_mem_hole_top_down(unsigned long start, unsigned long end,
2029 struct kexec_buf *kbuf)
2030{
2031 struct kimage *image = kbuf->image;
2032 unsigned long temp_start, temp_end;
2033
2034 temp_end = min(end, kbuf->buf_max);
2035 temp_start = temp_end - kbuf->memsz;
2036
2037 do {
2038 /* align down start */
2039 temp_start = temp_start & (~(kbuf->buf_align - 1));
2040
2041 if (temp_start < start || temp_start < kbuf->buf_min)
2042 return 0;
2043
2044 temp_end = temp_start + kbuf->memsz - 1;
2045
2046 /*
2047 * Make sure this does not conflict with any of existing
2048 * segments
2049 */
2050 if (kimage_is_destination_range(image, temp_start, temp_end)) {
2051 temp_start = temp_start - PAGE_SIZE;
2052 continue;
2053 }
2054
2055 /* We found a suitable memory range */
2056 break;
2057 } while (1);
2058
2059 /* If we are here, we found a suitable memory range */
2060 kbuf->mem = temp_start;
2061
2062 /* Success, stop navigating through remaining System RAM ranges */
2063 return 1;
2064}
2065
2066static int locate_mem_hole_bottom_up(unsigned long start, unsigned long end,
2067 struct kexec_buf *kbuf)
2068{
2069 struct kimage *image = kbuf->image;
2070 unsigned long temp_start, temp_end;
2071
2072 temp_start = max(start, kbuf->buf_min);
2073
2074 do {
2075 temp_start = ALIGN(temp_start, kbuf->buf_align);
2076 temp_end = temp_start + kbuf->memsz - 1;
2077
2078 if (temp_end > end || temp_end > kbuf->buf_max)
2079 return 0;
2080 /*
2081 * Make sure this does not conflict with any of existing
2082 * segments
2083 */
2084 if (kimage_is_destination_range(image, temp_start, temp_end)) {
2085 temp_start = temp_start + PAGE_SIZE;
2086 continue;
2087 }
2088
2089 /* We found a suitable memory range */
2090 break;
2091 } while (1);
2092
2093 /* If we are here, we found a suitable memory range */
2094 kbuf->mem = temp_start;
2095
2096 /* Success, stop navigating through remaining System RAM ranges */
2097 return 1;
2098}
2099
2100static int locate_mem_hole_callback(u64 start, u64 end, void *arg)
2101{
2102 struct kexec_buf *kbuf = (struct kexec_buf *)arg;
2103 unsigned long sz = end - start + 1;
2104
2105 /* Returning 0 will take to next memory range */
2106 if (sz < kbuf->memsz)
2107 return 0;
2108
2109 if (end < kbuf->buf_min || start > kbuf->buf_max)
2110 return 0;
2111
2112 /*
2113 * Allocate memory top down with-in ram range. Otherwise bottom up
2114 * allocation.
2115 */
2116 if (kbuf->top_down)
2117 return locate_mem_hole_top_down(start, end, kbuf);
2118 return locate_mem_hole_bottom_up(start, end, kbuf);
2119}
2120
2121/*
2122 * Helper function for placing a buffer in a kexec segment. This assumes
2123 * that kexec_mutex is held.
2124 */
2125int kexec_add_buffer(struct kimage *image, char *buffer, unsigned long bufsz,
2126 unsigned long memsz, unsigned long buf_align,
2127 unsigned long buf_min, unsigned long buf_max,
2128 bool top_down, unsigned long *load_addr)
2129{
2130
2131 struct kexec_segment *ksegment;
2132 struct kexec_buf buf, *kbuf;
2133 int ret;
2134
2135 /* Currently adding segment this way is allowed only in file mode */
2136 if (!image->file_mode)
2137 return -EINVAL;
2138
2139 if (image->nr_segments >= KEXEC_SEGMENT_MAX)
2140 return -EINVAL;
2141
2142 /*
2143 * Make sure we are not trying to add buffer after allocating
2144 * control pages. All segments need to be placed first before
2145 * any control pages are allocated. As control page allocation
2146 * logic goes through list of segments to make sure there are
2147 * no destination overlaps.
2148 */
2149 if (!list_empty(&image->control_pages)) {
2150 WARN_ON(1);
2151 return -EINVAL;
2152 }
2153
2154 memset(&buf, 0, sizeof(struct kexec_buf));
2155 kbuf = &buf;
2156 kbuf->image = image;
2157 kbuf->buffer = buffer;
2158 kbuf->bufsz = bufsz;
2159
2160 kbuf->memsz = ALIGN(memsz, PAGE_SIZE);
2161 kbuf->buf_align = max(buf_align, PAGE_SIZE);
2162 kbuf->buf_min = buf_min;
2163 kbuf->buf_max = buf_max;
2164 kbuf->top_down = top_down;
2165
2166 /* Walk the RAM ranges and allocate a suitable range for the buffer */
2167 if (image->type == KEXEC_TYPE_CRASH)
2168 ret = walk_iomem_res("Crash kernel",
2169 IORESOURCE_MEM | IORESOURCE_BUSY,
2170 crashk_res.start, crashk_res.end, kbuf,
2171 locate_mem_hole_callback);
2172 else
2173 ret = walk_system_ram_res(0, -1, kbuf,
2174 locate_mem_hole_callback);
2175 if (ret != 1) {
2176 /* A suitable memory range could not be found for buffer */
2177 return -EADDRNOTAVAIL;
2178 }
2179
2180 /* Found a suitable memory range */
2181 ksegment = &image->segment[image->nr_segments];
2182 ksegment->kbuf = kbuf->buffer;
2183 ksegment->bufsz = kbuf->bufsz;
2184 ksegment->mem = kbuf->mem;
2185 ksegment->memsz = kbuf->memsz;
2186 image->nr_segments++;
2187 *load_addr = ksegment->mem;
2188 return 0;
2189}
2190
2191/* Calculate and store the digest of segments */
2192static int kexec_calculate_store_digests(struct kimage *image)
2193{
2194 struct crypto_shash *tfm;
2195 struct shash_desc *desc;
2196 int ret = 0, i, j, zero_buf_sz, sha_region_sz;
2197 size_t desc_size, nullsz;
2198 char *digest;
2199 void *zero_buf;
2200 struct kexec_sha_region *sha_regions;
2201 struct purgatory_info *pi = &image->purgatory_info;
2202
2203 zero_buf = __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT);
2204 zero_buf_sz = PAGE_SIZE;
2205
2206 tfm = crypto_alloc_shash("sha256", 0, 0);
2207 if (IS_ERR(tfm)) {
2208 ret = PTR_ERR(tfm);
2209 goto out;
2210 }
2211
2212 desc_size = crypto_shash_descsize(tfm) + sizeof(*desc);
2213 desc = kzalloc(desc_size, GFP_KERNEL);
2214 if (!desc) {
2215 ret = -ENOMEM;
2216 goto out_free_tfm;
2217 }
2218
2219 sha_region_sz = KEXEC_SEGMENT_MAX * sizeof(struct kexec_sha_region);
2220 sha_regions = vzalloc(sha_region_sz);
2221 if (!sha_regions)
2222 goto out_free_desc;
2223
2224 desc->tfm = tfm;
2225 desc->flags = 0;
2226
2227 ret = crypto_shash_init(desc);
2228 if (ret < 0)
2229 goto out_free_sha_regions;
2230
2231 digest = kzalloc(SHA256_DIGEST_SIZE, GFP_KERNEL);
2232 if (!digest) {
2233 ret = -ENOMEM;
2234 goto out_free_sha_regions;
2235 }
2236
2237 for (j = i = 0; i < image->nr_segments; i++) {
2238 struct kexec_segment *ksegment;
2239
2240 ksegment = &image->segment[i];
2241 /*
2242 * Skip purgatory as it will be modified once we put digest
2243 * info in purgatory.
2244 */
2245 if (ksegment->kbuf == pi->purgatory_buf)
2246 continue;
2247
2248 ret = crypto_shash_update(desc, ksegment->kbuf,
2249 ksegment->bufsz);
2250 if (ret)
2251 break;
2252
2253 /*
2254 * Assume rest of the buffer is filled with zero and
2255 * update digest accordingly.
2256 */
2257 nullsz = ksegment->memsz - ksegment->bufsz;
2258 while (nullsz) {
2259 unsigned long bytes = nullsz;
2260
2261 if (bytes > zero_buf_sz)
2262 bytes = zero_buf_sz;
2263 ret = crypto_shash_update(desc, zero_buf, bytes);
2264 if (ret)
2265 break;
2266 nullsz -= bytes;
2267 }
2268
2269 if (ret)
2270 break;
2271
2272 sha_regions[j].start = ksegment->mem;
2273 sha_regions[j].len = ksegment->memsz;
2274 j++;
2275 }
2276
2277 if (!ret) {
2278 ret = crypto_shash_final(desc, digest);
2279 if (ret)
2280 goto out_free_digest;
2281 ret = kexec_purgatory_get_set_symbol(image, "sha_regions",
2282 sha_regions, sha_region_sz, 0);
2283 if (ret)
2284 goto out_free_digest;
2285
2286 ret = kexec_purgatory_get_set_symbol(image, "sha256_digest",
2287 digest, SHA256_DIGEST_SIZE, 0);
2288 if (ret)
2289 goto out_free_digest;
2290 }
2291
2292out_free_digest:
2293 kfree(digest);
2294out_free_sha_regions:
2295 vfree(sha_regions);
2296out_free_desc:
2297 kfree(desc);
2298out_free_tfm:
2299 kfree(tfm);
2300out:
2301 return ret;
2302}
2303
2304/* Actually load purgatory. Lot of code taken from kexec-tools */
2305static int __kexec_load_purgatory(struct kimage *image, unsigned long min,
2306 unsigned long max, int top_down)
2307{
2308 struct purgatory_info *pi = &image->purgatory_info;
2309 unsigned long align, buf_align, bss_align, buf_sz, bss_sz, bss_pad;
2310 unsigned long memsz, entry, load_addr, curr_load_addr, bss_addr, offset;
2311 unsigned char *buf_addr, *src;
2312 int i, ret = 0, entry_sidx = -1;
2313 const Elf_Shdr *sechdrs_c;
2314 Elf_Shdr *sechdrs = NULL;
2315 void *purgatory_buf = NULL;
2316
2317 /*
2318 * sechdrs_c points to section headers in purgatory and are read
2319 * only. No modifications allowed.
2320 */
2321 sechdrs_c = (void *)pi->ehdr + pi->ehdr->e_shoff;
2322
2323 /*
2324 * We can not modify sechdrs_c[] and its fields. It is read only.
2325 * Copy it over to a local copy where one can store some temporary
2326 * data and free it at the end. We need to modify ->sh_addr and
2327 * ->sh_offset fields to keep track of permanent and temporary
2328 * locations of sections.
2329 */
2330 sechdrs = vzalloc(pi->ehdr->e_shnum * sizeof(Elf_Shdr));
2331 if (!sechdrs)
2332 return -ENOMEM;
2333
2334 memcpy(sechdrs, sechdrs_c, pi->ehdr->e_shnum * sizeof(Elf_Shdr));
2335
2336 /*
2337 * We seem to have multiple copies of sections. First copy is which
2338 * is embedded in kernel in read only section. Some of these sections
2339 * will be copied to a temporary buffer and relocated. And these
2340 * sections will finally be copied to their final destination at
2341 * segment load time.
2342 *
2343 * Use ->sh_offset to reflect section address in memory. It will
2344 * point to original read only copy if section is not allocatable.
2345 * Otherwise it will point to temporary copy which will be relocated.
2346 *
2347 * Use ->sh_addr to contain final address of the section where it
2348 * will go during execution time.
2349 */
2350 for (i = 0; i < pi->ehdr->e_shnum; i++) {
2351 if (sechdrs[i].sh_type == SHT_NOBITS)
2352 continue;
2353
2354 sechdrs[i].sh_offset = (unsigned long)pi->ehdr +
2355 sechdrs[i].sh_offset;
2356 }
2357
2358 /*
2359 * Identify entry point section and make entry relative to section
2360 * start.
2361 */
2362 entry = pi->ehdr->e_entry;
2363 for (i = 0; i < pi->ehdr->e_shnum; i++) {
2364 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
2365 continue;
2366
2367 if (!(sechdrs[i].sh_flags & SHF_EXECINSTR))
2368 continue;
2369
2370 /* Make entry section relative */
2371 if (sechdrs[i].sh_addr <= pi->ehdr->e_entry &&
2372 ((sechdrs[i].sh_addr + sechdrs[i].sh_size) >
2373 pi->ehdr->e_entry)) {
2374 entry_sidx = i;
2375 entry -= sechdrs[i].sh_addr;
2376 break;
2377 }
2378 }
2379
2380 /* Determine how much memory is needed to load relocatable object. */
2381 buf_align = 1;
2382 bss_align = 1;
2383 buf_sz = 0;
2384 bss_sz = 0;
2385
2386 for (i = 0; i < pi->ehdr->e_shnum; i++) {
2387 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
2388 continue;
2389
2390 align = sechdrs[i].sh_addralign;
2391 if (sechdrs[i].sh_type != SHT_NOBITS) {
2392 if (buf_align < align)
2393 buf_align = align;
2394 buf_sz = ALIGN(buf_sz, align);
2395 buf_sz += sechdrs[i].sh_size;
2396 } else {
2397 /* bss section */
2398 if (bss_align < align)
2399 bss_align = align;
2400 bss_sz = ALIGN(bss_sz, align);
2401 bss_sz += sechdrs[i].sh_size;
2402 }
2403 }
2404
2405 /* Determine the bss padding required to align bss properly */
2406 bss_pad = 0;
2407 if (buf_sz & (bss_align - 1))
2408 bss_pad = bss_align - (buf_sz & (bss_align - 1));
2409
2410 memsz = buf_sz + bss_pad + bss_sz;
2411
2412 /* Allocate buffer for purgatory */
2413 purgatory_buf = vzalloc(buf_sz);
2414 if (!purgatory_buf) {
2415 ret = -ENOMEM;
2416 goto out;
2417 }
2418
2419 if (buf_align < bss_align)
2420 buf_align = bss_align;
2421
2422 /* Add buffer to segment list */
2423 ret = kexec_add_buffer(image, purgatory_buf, buf_sz, memsz,
2424 buf_align, min, max, top_down,
2425 &pi->purgatory_load_addr);
2426 if (ret)
2427 goto out;
2428
2429 /* Load SHF_ALLOC sections */
2430 buf_addr = purgatory_buf;
2431 load_addr = curr_load_addr = pi->purgatory_load_addr;
2432 bss_addr = load_addr + buf_sz + bss_pad;
2433
2434 for (i = 0; i < pi->ehdr->e_shnum; i++) {
2435 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
2436 continue;
2437
2438 align = sechdrs[i].sh_addralign;
2439 if (sechdrs[i].sh_type != SHT_NOBITS) {
2440 curr_load_addr = ALIGN(curr_load_addr, align);
2441 offset = curr_load_addr - load_addr;
2442 /* We already modifed ->sh_offset to keep src addr */
2443 src = (char *) sechdrs[i].sh_offset;
2444 memcpy(buf_addr + offset, src, sechdrs[i].sh_size);
2445
2446 /* Store load address and source address of section */
2447 sechdrs[i].sh_addr = curr_load_addr;
2448
2449 /*
2450 * This section got copied to temporary buffer. Update
2451 * ->sh_offset accordingly.
2452 */
2453 sechdrs[i].sh_offset = (unsigned long)(buf_addr + offset);
2454
2455 /* Advance to the next address */
2456 curr_load_addr += sechdrs[i].sh_size;
2457 } else {
2458 bss_addr = ALIGN(bss_addr, align);
2459 sechdrs[i].sh_addr = bss_addr;
2460 bss_addr += sechdrs[i].sh_size;
2461 }
2462 }
2463
2464 /* Update entry point based on load address of text section */
2465 if (entry_sidx >= 0)
2466 entry += sechdrs[entry_sidx].sh_addr;
2467
2468 /* Make kernel jump to purgatory after shutdown */
2469 image->start = entry;
2470
2471 /* Used later to get/set symbol values */
2472 pi->sechdrs = sechdrs;
2473
2474 /*
2475 * Used later to identify which section is purgatory and skip it
2476 * from checksumming.
2477 */
2478 pi->purgatory_buf = purgatory_buf;
2479 return ret;
2480out:
2481 vfree(sechdrs);
2482 vfree(purgatory_buf);
2483 return ret;
2484}
2485
2486static int kexec_apply_relocations(struct kimage *image)
2487{
2488 int i, ret;
2489 struct purgatory_info *pi = &image->purgatory_info;
2490 Elf_Shdr *sechdrs = pi->sechdrs;
2491
2492 /* Apply relocations */
2493 for (i = 0; i < pi->ehdr->e_shnum; i++) {
2494 Elf_Shdr *section, *symtab;
2495
2496 if (sechdrs[i].sh_type != SHT_RELA &&
2497 sechdrs[i].sh_type != SHT_REL)
2498 continue;
2499
2500 /*
2501 * For section of type SHT_RELA/SHT_REL,
2502 * ->sh_link contains section header index of associated
2503 * symbol table. And ->sh_info contains section header
2504 * index of section to which relocations apply.
2505 */
2506 if (sechdrs[i].sh_info >= pi->ehdr->e_shnum ||
2507 sechdrs[i].sh_link >= pi->ehdr->e_shnum)
2508 return -ENOEXEC;
2509
2510 section = &sechdrs[sechdrs[i].sh_info];
2511 symtab = &sechdrs[sechdrs[i].sh_link];
2512
2513 if (!(section->sh_flags & SHF_ALLOC))
2514 continue;
2515
2516 /*
2517 * symtab->sh_link contain section header index of associated
2518 * string table.
2519 */
2520 if (symtab->sh_link >= pi->ehdr->e_shnum)
2521 /* Invalid section number? */
2522 continue;
2523
2524 /*
2525 * Respective architecture needs to provide support for applying
2526 * relocations of type SHT_RELA/SHT_REL.
2527 */
2528 if (sechdrs[i].sh_type == SHT_RELA)
2529 ret = arch_kexec_apply_relocations_add(pi->ehdr,
2530 sechdrs, i);
2531 else if (sechdrs[i].sh_type == SHT_REL)
2532 ret = arch_kexec_apply_relocations(pi->ehdr,
2533 sechdrs, i);
2534 if (ret)
2535 return ret;
2536 }
2537
2538 return 0;
2539}
2540
2541/* Load relocatable purgatory object and relocate it appropriately */
2542int kexec_load_purgatory(struct kimage *image, unsigned long min,
2543 unsigned long max, int top_down,
2544 unsigned long *load_addr)
2545{
2546 struct purgatory_info *pi = &image->purgatory_info;
2547 int ret;
2548
2549 if (kexec_purgatory_size <= 0)
2550 return -EINVAL;
2551
2552 if (kexec_purgatory_size < sizeof(Elf_Ehdr))
2553 return -ENOEXEC;
2554
2555 pi->ehdr = (Elf_Ehdr *)kexec_purgatory;
2556
2557 if (memcmp(pi->ehdr->e_ident, ELFMAG, SELFMAG) != 0
2558 || pi->ehdr->e_type != ET_REL
2559 || !elf_check_arch(pi->ehdr)
2560 || pi->ehdr->e_shentsize != sizeof(Elf_Shdr))
2561 return -ENOEXEC;
2562
2563 if (pi->ehdr->e_shoff >= kexec_purgatory_size
2564 || (pi->ehdr->e_shnum * sizeof(Elf_Shdr) >
2565 kexec_purgatory_size - pi->ehdr->e_shoff))
2566 return -ENOEXEC;
2567
2568 ret = __kexec_load_purgatory(image, min, max, top_down);
2569 if (ret)
2570 return ret;
2571
2572 ret = kexec_apply_relocations(image);
2573 if (ret)
2574 goto out;
2575
2576 *load_addr = pi->purgatory_load_addr;
2577 return 0;
2578out:
2579 vfree(pi->sechdrs);
2580 vfree(pi->purgatory_buf);
2581 return ret;
2582}
2583
2584static Elf_Sym *kexec_purgatory_find_symbol(struct purgatory_info *pi,
2585 const char *name)
2586{
2587 Elf_Sym *syms;
2588 Elf_Shdr *sechdrs;
2589 Elf_Ehdr *ehdr;
2590 int i, k;
2591 const char *strtab;
2592
2593 if (!pi->sechdrs || !pi->ehdr)
2594 return NULL;
2595
2596 sechdrs = pi->sechdrs;
2597 ehdr = pi->ehdr;
2598
2599 for (i = 0; i < ehdr->e_shnum; i++) {
2600 if (sechdrs[i].sh_type != SHT_SYMTAB)
2601 continue;
2602
2603 if (sechdrs[i].sh_link >= ehdr->e_shnum)
2604 /* Invalid strtab section number */
2605 continue;
2606 strtab = (char *)sechdrs[sechdrs[i].sh_link].sh_offset;
2607 syms = (Elf_Sym *)sechdrs[i].sh_offset;
2608
2609 /* Go through symbols for a match */
2610 for (k = 0; k < sechdrs[i].sh_size/sizeof(Elf_Sym); k++) {
2611 if (ELF_ST_BIND(syms[k].st_info) != STB_GLOBAL)
2612 continue;
2613
2614 if (strcmp(strtab + syms[k].st_name, name) != 0)
2615 continue;
2616
2617 if (syms[k].st_shndx == SHN_UNDEF ||
2618 syms[k].st_shndx >= ehdr->e_shnum) {
2619 pr_debug("Symbol: %s has bad section index %d.\n",
2620 name, syms[k].st_shndx);
2621 return NULL;
2622 }
2623
2624 /* Found the symbol we are looking for */
2625 return &syms[k];
2626 }
2627 }
2628
2629 return NULL;
2630}
2631
2632void *kexec_purgatory_get_symbol_addr(struct kimage *image, const char *name)
2633{
2634 struct purgatory_info *pi = &image->purgatory_info;
2635 Elf_Sym *sym;
2636 Elf_Shdr *sechdr;
2637
2638 sym = kexec_purgatory_find_symbol(pi, name);
2639 if (!sym)
2640 return ERR_PTR(-EINVAL);
2641
2642 sechdr = &pi->sechdrs[sym->st_shndx];
2643
2644 /*
2645 * Returns the address where symbol will finally be loaded after
2646 * kexec_load_segment()
2647 */
2648 return (void *)(sechdr->sh_addr + sym->st_value);
2649}
2650
2651/*
2652 * Get or set value of a symbol. If "get_value" is true, symbol value is
2653 * returned in buf otherwise symbol value is set based on value in buf.
2654 */
2655int kexec_purgatory_get_set_symbol(struct kimage *image, const char *name,
2656 void *buf, unsigned int size, bool get_value)
2657{
2658 Elf_Sym *sym;
2659 Elf_Shdr *sechdrs;
2660 struct purgatory_info *pi = &image->purgatory_info;
2661 char *sym_buf;
2662
2663 sym = kexec_purgatory_find_symbol(pi, name);
2664 if (!sym)
2665 return -EINVAL;
2666
2667 if (sym->st_size != size) {
2668 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
2669 name, (unsigned long)sym->st_size, size);
2670 return -EINVAL;
2671 }
2672
2673 sechdrs = pi->sechdrs;
2674
2675 if (sechdrs[sym->st_shndx].sh_type == SHT_NOBITS) {
2676 pr_err("symbol %s is in a bss section. Cannot %s\n", name,
2677 get_value ? "get" : "set");
2678 return -EINVAL;
2679 }
2680
2681 sym_buf = (unsigned char *)sechdrs[sym->st_shndx].sh_offset +
2682 sym->st_value;
2683
2684 if (get_value)
2685 memcpy((void *)buf, sym_buf, size);
2686 else
2687 memcpy((void *)sym_buf, buf, size);
2688
2689 return 0;
2690}
2691#endif /* CONFIG_KEXEC_FILE */
2692
2693/* 1659/*
2694 * Move into place and start executing a preloaded standalone 1660 * Move into place and start executing a preloaded standalone
2695 * executable. If nothing was preloaded return an error. 1661 * executable. If nothing was preloaded return an error.
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-26 11:59:21 -0500