1 // SPDX-License-Identifier: GPL-2.0-only
3 * kexec: kexec_file_load system call
5 * Copyright (C) 2014 Red Hat Inc.
7 * Vivek Goyal <vgoyal@redhat.com>
10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 #include <linux/capability.h>
14 #include <linux/file.h>
15 #include <linux/slab.h>
16 #include <linux/kexec.h>
17 #include <linux/memblock.h>
18 #include <linux/mutex.h>
19 #include <linux/list.h>
21 #include <linux/ima.h>
22 #include <crypto/hash.h>
23 #include <crypto/sha.h>
24 #include <linux/elf.h>
25 #include <linux/elfcore.h>
26 #include <linux/kernel.h>
27 #include <linux/syscalls.h>
28 #include <linux/vmalloc.h>
29 #include "kexec_internal.h"
31 static int kexec_calculate_store_digests(struct kimage
*image
);
34 * Currently this is the only default function that is exported as some
35 * architectures need it to do additional handlings.
36 * In the future, other default functions may be exported too if required.
38 int kexec_image_probe_default(struct kimage
*image
, void *buf
,
39 unsigned long buf_len
)
41 const struct kexec_file_ops
* const *fops
;
44 for (fops
= &kexec_file_loaders
[0]; *fops
&& (*fops
)->probe
; ++fops
) {
45 ret
= (*fops
)->probe(buf
, buf_len
);
55 /* Architectures can provide this probe function */
56 int __weak
arch_kexec_kernel_image_probe(struct kimage
*image
, void *buf
,
57 unsigned long buf_len
)
59 return kexec_image_probe_default(image
, buf
, buf_len
);
62 static void *kexec_image_load_default(struct kimage
*image
)
64 if (!image
->fops
|| !image
->fops
->load
)
65 return ERR_PTR(-ENOEXEC
);
67 return image
->fops
->load(image
, image
->kernel_buf
,
68 image
->kernel_buf_len
, image
->initrd_buf
,
69 image
->initrd_buf_len
, image
->cmdline_buf
,
70 image
->cmdline_buf_len
);
73 void * __weak
arch_kexec_kernel_image_load(struct kimage
*image
)
75 return kexec_image_load_default(image
);
78 int kexec_image_post_load_cleanup_default(struct kimage
*image
)
80 if (!image
->fops
|| !image
->fops
->cleanup
)
83 return image
->fops
->cleanup(image
->image_loader_data
);
86 int __weak
arch_kimage_file_post_load_cleanup(struct kimage
*image
)
88 return kexec_image_post_load_cleanup_default(image
);
91 #ifdef CONFIG_KEXEC_SIG
92 static int kexec_image_verify_sig_default(struct kimage
*image
, void *buf
,
93 unsigned long buf_len
)
95 if (!image
->fops
|| !image
->fops
->verify_sig
) {
96 pr_debug("kernel loader does not support signature verification.\n");
100 return image
->fops
->verify_sig(buf
, buf_len
);
103 int __weak
arch_kexec_kernel_verify_sig(struct kimage
*image
, void *buf
,
104 unsigned long buf_len
)
106 return kexec_image_verify_sig_default(image
, buf
, buf_len
);
111 * arch_kexec_apply_relocations_add - apply relocations of type RELA
112 * @pi: Purgatory to be relocated.
113 * @section: Section relocations applying to.
114 * @relsec: Section containing RELAs.
115 * @symtab: Corresponding symtab.
117 * Return: 0 on success, negative errno on error.
120 arch_kexec_apply_relocations_add(struct purgatory_info
*pi
, Elf_Shdr
*section
,
121 const Elf_Shdr
*relsec
, const Elf_Shdr
*symtab
)
123 pr_err("RELA relocation unsupported.\n");
128 * arch_kexec_apply_relocations - apply relocations of type REL
129 * @pi: Purgatory to be relocated.
130 * @section: Section relocations applying to.
131 * @relsec: Section containing RELs.
132 * @symtab: Corresponding symtab.
134 * Return: 0 on success, negative errno on error.
137 arch_kexec_apply_relocations(struct purgatory_info
*pi
, Elf_Shdr
*section
,
138 const Elf_Shdr
*relsec
, const Elf_Shdr
*symtab
)
140 pr_err("REL relocation unsupported.\n");
145 * Free up memory used by kernel, initrd, and command line. This is temporary
146 * memory allocation which is not needed any more after these buffers have
147 * been loaded into separate segments and have been copied elsewhere.
149 void kimage_file_post_load_cleanup(struct kimage
*image
)
151 struct purgatory_info
*pi
= &image
->purgatory_info
;
153 vfree(image
->kernel_buf
);
154 image
->kernel_buf
= NULL
;
156 vfree(image
->initrd_buf
);
157 image
->initrd_buf
= NULL
;
159 kfree(image
->cmdline_buf
);
160 image
->cmdline_buf
= NULL
;
162 vfree(pi
->purgatory_buf
);
163 pi
->purgatory_buf
= NULL
;
168 /* See if architecture has anything to cleanup post load */
169 arch_kimage_file_post_load_cleanup(image
);
172 * Above call should have called into bootloader to free up
173 * any data stored in kimage->image_loader_data. It should
174 * be ok now to free it up.
176 kfree(image
->image_loader_data
);
177 image
->image_loader_data
= NULL
;
180 #ifdef CONFIG_KEXEC_SIG
182 kimage_validate_signature(struct kimage
*image
)
186 ret
= arch_kexec_kernel_verify_sig(image
, image
->kernel_buf
,
187 image
->kernel_buf_len
);
190 if (IS_ENABLED(CONFIG_KEXEC_SIG_FORCE
)) {
191 pr_notice("Enforced kernel signature verification failed (%d).\n", ret
);
196 * If IMA is guaranteed to appraise a signature on the kexec
197 * image, permit it even if the kernel is otherwise locked
200 if (!ima_appraise_signature(READING_KEXEC_IMAGE
) &&
201 security_locked_down(LOCKDOWN_KEXEC
))
204 pr_debug("kernel signature verification failed (%d).\n", ret
);
212 * In file mode list of segments is prepared by kernel. Copy relevant
213 * data from user space, do error checking, prepare segment list
216 kimage_file_prepare_segments(struct kimage
*image
, int kernel_fd
, int initrd_fd
,
217 const char __user
*cmdline_ptr
,
218 unsigned long cmdline_len
, unsigned flags
)
224 ret
= kernel_read_file_from_fd(kernel_fd
, &image
->kernel_buf
,
225 &size
, INT_MAX
, READING_KEXEC_IMAGE
);
228 image
->kernel_buf_len
= size
;
230 /* Call arch image probe handlers */
231 ret
= arch_kexec_kernel_image_probe(image
, image
->kernel_buf
,
232 image
->kernel_buf_len
);
236 #ifdef CONFIG_KEXEC_SIG
237 ret
= kimage_validate_signature(image
);
242 /* It is possible that there no initramfs is being loaded */
243 if (!(flags
& KEXEC_FILE_NO_INITRAMFS
)) {
244 ret
= kernel_read_file_from_fd(initrd_fd
, &image
->initrd_buf
,
246 READING_KEXEC_INITRAMFS
);
249 image
->initrd_buf_len
= size
;
253 image
->cmdline_buf
= memdup_user(cmdline_ptr
, cmdline_len
);
254 if (IS_ERR(image
->cmdline_buf
)) {
255 ret
= PTR_ERR(image
->cmdline_buf
);
256 image
->cmdline_buf
= NULL
;
260 image
->cmdline_buf_len
= cmdline_len
;
262 /* command line should be a string with last byte null */
263 if (image
->cmdline_buf
[cmdline_len
- 1] != '\0') {
268 ima_kexec_cmdline(kernel_fd
, image
->cmdline_buf
,
269 image
->cmdline_buf_len
- 1);
272 /* IMA needs to pass the measurement list to the next kernel. */
273 ima_add_kexec_buffer(image
);
275 /* Call arch image load handlers */
276 ldata
= arch_kexec_kernel_image_load(image
);
279 ret
= PTR_ERR(ldata
);
283 image
->image_loader_data
= ldata
;
285 /* In case of error, free up all allocated memory in this function */
287 kimage_file_post_load_cleanup(image
);
292 kimage_file_alloc_init(struct kimage
**rimage
, int kernel_fd
,
293 int initrd_fd
, const char __user
*cmdline_ptr
,
294 unsigned long cmdline_len
, unsigned long flags
)
297 struct kimage
*image
;
298 bool kexec_on_panic
= flags
& KEXEC_FILE_ON_CRASH
;
300 image
= do_kimage_alloc_init();
304 image
->file_mode
= 1;
306 if (kexec_on_panic
) {
307 /* Enable special crash kernel control page alloc policy. */
308 image
->control_page
= crashk_res
.start
;
309 image
->type
= KEXEC_TYPE_CRASH
;
312 ret
= kimage_file_prepare_segments(image
, kernel_fd
, initrd_fd
,
313 cmdline_ptr
, cmdline_len
, flags
);
317 ret
= sanity_check_segment_list(image
);
319 goto out_free_post_load_bufs
;
322 image
->control_code_page
= kimage_alloc_control_pages(image
,
323 get_order(KEXEC_CONTROL_PAGE_SIZE
));
324 if (!image
->control_code_page
) {
325 pr_err("Could not allocate control_code_buffer\n");
326 goto out_free_post_load_bufs
;
329 if (!kexec_on_panic
) {
330 image
->swap_page
= kimage_alloc_control_pages(image
, 0);
331 if (!image
->swap_page
) {
332 pr_err("Could not allocate swap buffer\n");
333 goto out_free_control_pages
;
339 out_free_control_pages
:
340 kimage_free_page_list(&image
->control_pages
);
341 out_free_post_load_bufs
:
342 kimage_file_post_load_cleanup(image
);
348 SYSCALL_DEFINE5(kexec_file_load
, int, kernel_fd
, int, initrd_fd
,
349 unsigned long, cmdline_len
, const char __user
*, cmdline_ptr
,
350 unsigned long, flags
)
353 struct kimage
**dest_image
, *image
;
355 /* We only trust the superuser with rebooting the system. */
356 if (!capable(CAP_SYS_BOOT
) || kexec_load_disabled
)
359 /* Make sure we have a legal set of flags */
360 if (flags
!= (flags
& KEXEC_FILE_FLAGS
))
365 if (!mutex_trylock(&kexec_mutex
))
368 dest_image
= &kexec_image
;
369 if (flags
& KEXEC_FILE_ON_CRASH
) {
370 dest_image
= &kexec_crash_image
;
371 if (kexec_crash_image
)
372 arch_kexec_unprotect_crashkres();
375 if (flags
& KEXEC_FILE_UNLOAD
)
379 * In case of crash, new kernel gets loaded in reserved region. It is
380 * same memory where old crash kernel might be loaded. Free any
381 * current crash dump kernel before we corrupt it.
383 if (flags
& KEXEC_FILE_ON_CRASH
)
384 kimage_free(xchg(&kexec_crash_image
, NULL
));
386 ret
= kimage_file_alloc_init(&image
, kernel_fd
, initrd_fd
, cmdline_ptr
,
391 ret
= machine_kexec_prepare(image
);
396 * Some architecture(like S390) may touch the crash memory before
397 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
399 ret
= kimage_crash_copy_vmcoreinfo(image
);
403 ret
= kexec_calculate_store_digests(image
);
407 for (i
= 0; i
< image
->nr_segments
; i
++) {
408 struct kexec_segment
*ksegment
;
410 ksegment
= &image
->segment
[i
];
411 pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
412 i
, ksegment
->buf
, ksegment
->bufsz
, ksegment
->mem
,
415 ret
= kimage_load_segment(image
, &image
->segment
[i
]);
420 kimage_terminate(image
);
422 ret
= machine_kexec_post_load(image
);
427 * Free up any temporary buffers allocated which are not needed
428 * after image has been loaded
430 kimage_file_post_load_cleanup(image
);
432 image
= xchg(dest_image
, image
);
434 if ((flags
& KEXEC_FILE_ON_CRASH
) && kexec_crash_image
)
435 arch_kexec_protect_crashkres();
437 mutex_unlock(&kexec_mutex
);
442 static int locate_mem_hole_top_down(unsigned long start
, unsigned long end
,
443 struct kexec_buf
*kbuf
)
445 struct kimage
*image
= kbuf
->image
;
446 unsigned long temp_start
, temp_end
;
448 temp_end
= min(end
, kbuf
->buf_max
);
449 temp_start
= temp_end
- kbuf
->memsz
;
452 /* align down start */
453 temp_start
= temp_start
& (~(kbuf
->buf_align
- 1));
455 if (temp_start
< start
|| temp_start
< kbuf
->buf_min
)
458 temp_end
= temp_start
+ kbuf
->memsz
- 1;
461 * Make sure this does not conflict with any of existing
464 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
465 temp_start
= temp_start
- PAGE_SIZE
;
469 /* We found a suitable memory range */
473 /* If we are here, we found a suitable memory range */
474 kbuf
->mem
= temp_start
;
476 /* Success, stop navigating through remaining System RAM ranges */
480 static int locate_mem_hole_bottom_up(unsigned long start
, unsigned long end
,
481 struct kexec_buf
*kbuf
)
483 struct kimage
*image
= kbuf
->image
;
484 unsigned long temp_start
, temp_end
;
486 temp_start
= max(start
, kbuf
->buf_min
);
489 temp_start
= ALIGN(temp_start
, kbuf
->buf_align
);
490 temp_end
= temp_start
+ kbuf
->memsz
- 1;
492 if (temp_end
> end
|| temp_end
> kbuf
->buf_max
)
495 * Make sure this does not conflict with any of existing
498 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
499 temp_start
= temp_start
+ PAGE_SIZE
;
503 /* We found a suitable memory range */
507 /* If we are here, we found a suitable memory range */
508 kbuf
->mem
= temp_start
;
510 /* Success, stop navigating through remaining System RAM ranges */
514 static int locate_mem_hole_callback(struct resource
*res
, void *arg
)
516 struct kexec_buf
*kbuf
= (struct kexec_buf
*)arg
;
517 u64 start
= res
->start
, end
= res
->end
;
518 unsigned long sz
= end
- start
+ 1;
520 /* Returning 0 will take to next memory range */
522 /* Don't use memory that will be detected and handled by a driver. */
523 if (res
->flags
& IORESOURCE_MEM_DRIVER_MANAGED
)
526 if (sz
< kbuf
->memsz
)
529 if (end
< kbuf
->buf_min
|| start
> kbuf
->buf_max
)
533 * Allocate memory top down with-in ram range. Otherwise bottom up
537 return locate_mem_hole_top_down(start
, end
, kbuf
);
538 return locate_mem_hole_bottom_up(start
, end
, kbuf
);
541 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
542 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
543 int (*func
)(struct resource
*, void *))
547 phys_addr_t mstart
, mend
;
548 struct resource res
= { };
550 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
551 return func(&crashk_res
, kbuf
);
553 if (kbuf
->top_down
) {
554 for_each_free_mem_range_reverse(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
555 &mstart
, &mend
, NULL
) {
557 * In memblock, end points to the first byte after the
558 * range while in kexec, end points to the last byte
563 ret
= func(&res
, kbuf
);
568 for_each_free_mem_range(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
569 &mstart
, &mend
, NULL
) {
571 * In memblock, end points to the first byte after the
572 * range while in kexec, end points to the last byte
577 ret
= func(&res
, kbuf
);
586 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
587 int (*func
)(struct resource
*, void *))
594 * kexec_walk_resources - call func(data) on free memory regions
595 * @kbuf: Context info for the search. Also passed to @func.
596 * @func: Function to call for each memory region.
598 * Return: The memory walk will stop when func returns a non-zero value
599 * and that value will be returned. If all free regions are visited without
600 * func returning non-zero, then zero will be returned.
602 static int kexec_walk_resources(struct kexec_buf
*kbuf
,
603 int (*func
)(struct resource
*, void *))
605 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
606 return walk_iomem_res_desc(crashk_res
.desc
,
607 IORESOURCE_SYSTEM_RAM
| IORESOURCE_BUSY
,
608 crashk_res
.start
, crashk_res
.end
,
611 return walk_system_ram_res(0, ULONG_MAX
, kbuf
, func
);
615 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
616 * @kbuf: Parameters for the memory search.
618 * On success, kbuf->mem will have the start address of the memory region found.
620 * Return: 0 on success, negative errno on error.
622 int kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
626 /* Arch knows where to place */
627 if (kbuf
->mem
!= KEXEC_BUF_MEM_UNKNOWN
)
630 if (!IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK
))
631 ret
= kexec_walk_resources(kbuf
, locate_mem_hole_callback
);
633 ret
= kexec_walk_memblock(kbuf
, locate_mem_hole_callback
);
635 return ret
== 1 ? 0 : -EADDRNOTAVAIL
;
639 * arch_kexec_locate_mem_hole - Find free memory to place the segments.
640 * @kbuf: Parameters for the memory search.
642 * On success, kbuf->mem will have the start address of the memory region found.
644 * Return: 0 on success, negative errno on error.
646 int __weak
arch_kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
648 return kexec_locate_mem_hole(kbuf
);
652 * kexec_add_buffer - place a buffer in a kexec segment
653 * @kbuf: Buffer contents and memory parameters.
655 * This function assumes that kexec_mutex is held.
656 * On successful return, @kbuf->mem will have the physical address of
657 * the buffer in memory.
659 * Return: 0 on success, negative errno on error.
661 int kexec_add_buffer(struct kexec_buf
*kbuf
)
663 struct kexec_segment
*ksegment
;
666 /* Currently adding segment this way is allowed only in file mode */
667 if (!kbuf
->image
->file_mode
)
670 if (kbuf
->image
->nr_segments
>= KEXEC_SEGMENT_MAX
)
674 * Make sure we are not trying to add buffer after allocating
675 * control pages. All segments need to be placed first before
676 * any control pages are allocated. As control page allocation
677 * logic goes through list of segments to make sure there are
678 * no destination overlaps.
680 if (!list_empty(&kbuf
->image
->control_pages
)) {
685 /* Ensure minimum alignment needed for segments. */
686 kbuf
->memsz
= ALIGN(kbuf
->memsz
, PAGE_SIZE
);
687 kbuf
->buf_align
= max(kbuf
->buf_align
, PAGE_SIZE
);
689 /* Walk the RAM ranges and allocate a suitable range for the buffer */
690 ret
= arch_kexec_locate_mem_hole(kbuf
);
694 /* Found a suitable memory range */
695 ksegment
= &kbuf
->image
->segment
[kbuf
->image
->nr_segments
];
696 ksegment
->kbuf
= kbuf
->buffer
;
697 ksegment
->bufsz
= kbuf
->bufsz
;
698 ksegment
->mem
= kbuf
->mem
;
699 ksegment
->memsz
= kbuf
->memsz
;
700 kbuf
->image
->nr_segments
++;
704 /* Calculate and store the digest of segments */
705 static int kexec_calculate_store_digests(struct kimage
*image
)
707 struct crypto_shash
*tfm
;
708 struct shash_desc
*desc
;
709 int ret
= 0, i
, j
, zero_buf_sz
, sha_region_sz
;
710 size_t desc_size
, nullsz
;
713 struct kexec_sha_region
*sha_regions
;
714 struct purgatory_info
*pi
= &image
->purgatory_info
;
716 if (!IS_ENABLED(CONFIG_ARCH_HAS_KEXEC_PURGATORY
))
719 zero_buf
= __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT
);
720 zero_buf_sz
= PAGE_SIZE
;
722 tfm
= crypto_alloc_shash("sha256", 0, 0);
728 desc_size
= crypto_shash_descsize(tfm
) + sizeof(*desc
);
729 desc
= kzalloc(desc_size
, GFP_KERNEL
);
735 sha_region_sz
= KEXEC_SEGMENT_MAX
* sizeof(struct kexec_sha_region
);
736 sha_regions
= vzalloc(sha_region_sz
);
742 ret
= crypto_shash_init(desc
);
744 goto out_free_sha_regions
;
746 digest
= kzalloc(SHA256_DIGEST_SIZE
, GFP_KERNEL
);
749 goto out_free_sha_regions
;
752 for (j
= i
= 0; i
< image
->nr_segments
; i
++) {
753 struct kexec_segment
*ksegment
;
755 ksegment
= &image
->segment
[i
];
757 * Skip purgatory as it will be modified once we put digest
760 if (ksegment
->kbuf
== pi
->purgatory_buf
)
763 ret
= crypto_shash_update(desc
, ksegment
->kbuf
,
769 * Assume rest of the buffer is filled with zero and
770 * update digest accordingly.
772 nullsz
= ksegment
->memsz
- ksegment
->bufsz
;
774 unsigned long bytes
= nullsz
;
776 if (bytes
> zero_buf_sz
)
778 ret
= crypto_shash_update(desc
, zero_buf
, bytes
);
787 sha_regions
[j
].start
= ksegment
->mem
;
788 sha_regions
[j
].len
= ksegment
->memsz
;
793 ret
= crypto_shash_final(desc
, digest
);
795 goto out_free_digest
;
796 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha_regions",
797 sha_regions
, sha_region_sz
, 0);
799 goto out_free_digest
;
801 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha256_digest",
802 digest
, SHA256_DIGEST_SIZE
, 0);
804 goto out_free_digest
;
809 out_free_sha_regions
:
819 #ifdef CONFIG_ARCH_HAS_KEXEC_PURGATORY
821 * kexec_purgatory_setup_kbuf - prepare buffer to load purgatory.
822 * @pi: Purgatory to be loaded.
823 * @kbuf: Buffer to setup.
825 * Allocates the memory needed for the buffer. Caller is responsible to free
826 * the memory after use.
828 * Return: 0 on success, negative errno on error.
830 static int kexec_purgatory_setup_kbuf(struct purgatory_info
*pi
,
831 struct kexec_buf
*kbuf
)
833 const Elf_Shdr
*sechdrs
;
834 unsigned long bss_align
;
835 unsigned long bss_sz
;
839 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
840 kbuf
->buf_align
= bss_align
= 1;
841 kbuf
->bufsz
= bss_sz
= 0;
843 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
844 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
847 align
= sechdrs
[i
].sh_addralign
;
848 if (sechdrs
[i
].sh_type
!= SHT_NOBITS
) {
849 if (kbuf
->buf_align
< align
)
850 kbuf
->buf_align
= align
;
851 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, align
);
852 kbuf
->bufsz
+= sechdrs
[i
].sh_size
;
854 if (bss_align
< align
)
856 bss_sz
= ALIGN(bss_sz
, align
);
857 bss_sz
+= sechdrs
[i
].sh_size
;
860 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, bss_align
);
861 kbuf
->memsz
= kbuf
->bufsz
+ bss_sz
;
862 if (kbuf
->buf_align
< bss_align
)
863 kbuf
->buf_align
= bss_align
;
865 kbuf
->buffer
= vzalloc(kbuf
->bufsz
);
868 pi
->purgatory_buf
= kbuf
->buffer
;
870 ret
= kexec_add_buffer(kbuf
);
876 vfree(pi
->purgatory_buf
);
877 pi
->purgatory_buf
= NULL
;
882 * kexec_purgatory_setup_sechdrs - prepares the pi->sechdrs buffer.
883 * @pi: Purgatory to be loaded.
884 * @kbuf: Buffer prepared to store purgatory.
886 * Allocates the memory needed for the buffer. Caller is responsible to free
887 * the memory after use.
889 * Return: 0 on success, negative errno on error.
891 static int kexec_purgatory_setup_sechdrs(struct purgatory_info
*pi
,
892 struct kexec_buf
*kbuf
)
894 unsigned long bss_addr
;
895 unsigned long offset
;
900 * The section headers in kexec_purgatory are read-only. In order to
901 * have them modifiable make a temporary copy.
903 sechdrs
= vzalloc(array_size(sizeof(Elf_Shdr
), pi
->ehdr
->e_shnum
));
906 memcpy(sechdrs
, (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
,
907 pi
->ehdr
->e_shnum
* sizeof(Elf_Shdr
));
908 pi
->sechdrs
= sechdrs
;
911 bss_addr
= kbuf
->mem
+ kbuf
->bufsz
;
912 kbuf
->image
->start
= pi
->ehdr
->e_entry
;
914 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
918 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
921 align
= sechdrs
[i
].sh_addralign
;
922 if (sechdrs
[i
].sh_type
== SHT_NOBITS
) {
923 bss_addr
= ALIGN(bss_addr
, align
);
924 sechdrs
[i
].sh_addr
= bss_addr
;
925 bss_addr
+= sechdrs
[i
].sh_size
;
929 offset
= ALIGN(offset
, align
);
930 if (sechdrs
[i
].sh_flags
& SHF_EXECINSTR
&&
931 pi
->ehdr
->e_entry
>= sechdrs
[i
].sh_addr
&&
932 pi
->ehdr
->e_entry
< (sechdrs
[i
].sh_addr
933 + sechdrs
[i
].sh_size
)) {
934 kbuf
->image
->start
-= sechdrs
[i
].sh_addr
;
935 kbuf
->image
->start
+= kbuf
->mem
+ offset
;
938 src
= (void *)pi
->ehdr
+ sechdrs
[i
].sh_offset
;
939 dst
= pi
->purgatory_buf
+ offset
;
940 memcpy(dst
, src
, sechdrs
[i
].sh_size
);
942 sechdrs
[i
].sh_addr
= kbuf
->mem
+ offset
;
943 sechdrs
[i
].sh_offset
= offset
;
944 offset
+= sechdrs
[i
].sh_size
;
950 static int kexec_apply_relocations(struct kimage
*image
)
953 struct purgatory_info
*pi
= &image
->purgatory_info
;
954 const Elf_Shdr
*sechdrs
;
956 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
958 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
959 const Elf_Shdr
*relsec
;
960 const Elf_Shdr
*symtab
;
963 relsec
= sechdrs
+ i
;
965 if (relsec
->sh_type
!= SHT_RELA
&&
966 relsec
->sh_type
!= SHT_REL
)
970 * For section of type SHT_RELA/SHT_REL,
971 * ->sh_link contains section header index of associated
972 * symbol table. And ->sh_info contains section header
973 * index of section to which relocations apply.
975 if (relsec
->sh_info
>= pi
->ehdr
->e_shnum
||
976 relsec
->sh_link
>= pi
->ehdr
->e_shnum
)
979 section
= pi
->sechdrs
+ relsec
->sh_info
;
980 symtab
= sechdrs
+ relsec
->sh_link
;
982 if (!(section
->sh_flags
& SHF_ALLOC
))
986 * symtab->sh_link contain section header index of associated
989 if (symtab
->sh_link
>= pi
->ehdr
->e_shnum
)
990 /* Invalid section number? */
994 * Respective architecture needs to provide support for applying
995 * relocations of type SHT_RELA/SHT_REL.
997 if (relsec
->sh_type
== SHT_RELA
)
998 ret
= arch_kexec_apply_relocations_add(pi
, section
,
1000 else if (relsec
->sh_type
== SHT_REL
)
1001 ret
= arch_kexec_apply_relocations(pi
, section
,
1011 * kexec_load_purgatory - Load and relocate the purgatory object.
1012 * @image: Image to add the purgatory to.
1013 * @kbuf: Memory parameters to use.
1015 * Allocates the memory needed for image->purgatory_info.sechdrs and
1016 * image->purgatory_info.purgatory_buf/kbuf->buffer. Caller is responsible
1017 * to free the memory after use.
1019 * Return: 0 on success, negative errno on error.
1021 int kexec_load_purgatory(struct kimage
*image
, struct kexec_buf
*kbuf
)
1023 struct purgatory_info
*pi
= &image
->purgatory_info
;
1026 if (kexec_purgatory_size
<= 0)
1029 pi
->ehdr
= (const Elf_Ehdr
*)kexec_purgatory
;
1031 ret
= kexec_purgatory_setup_kbuf(pi
, kbuf
);
1035 ret
= kexec_purgatory_setup_sechdrs(pi
, kbuf
);
1039 ret
= kexec_apply_relocations(image
);
1048 vfree(pi
->purgatory_buf
);
1049 pi
->purgatory_buf
= NULL
;
1054 * kexec_purgatory_find_symbol - find a symbol in the purgatory
1055 * @pi: Purgatory to search in.
1056 * @name: Name of the symbol.
1058 * Return: pointer to symbol in read-only symtab on success, NULL on error.
1060 static const Elf_Sym
*kexec_purgatory_find_symbol(struct purgatory_info
*pi
,
1063 const Elf_Shdr
*sechdrs
;
1064 const Elf_Ehdr
*ehdr
;
1065 const Elf_Sym
*syms
;
1073 sechdrs
= (void *)ehdr
+ ehdr
->e_shoff
;
1075 for (i
= 0; i
< ehdr
->e_shnum
; i
++) {
1076 if (sechdrs
[i
].sh_type
!= SHT_SYMTAB
)
1079 if (sechdrs
[i
].sh_link
>= ehdr
->e_shnum
)
1080 /* Invalid strtab section number */
1082 strtab
= (void *)ehdr
+ sechdrs
[sechdrs
[i
].sh_link
].sh_offset
;
1083 syms
= (void *)ehdr
+ sechdrs
[i
].sh_offset
;
1085 /* Go through symbols for a match */
1086 for (k
= 0; k
< sechdrs
[i
].sh_size
/sizeof(Elf_Sym
); k
++) {
1087 if (ELF_ST_BIND(syms
[k
].st_info
) != STB_GLOBAL
)
1090 if (strcmp(strtab
+ syms
[k
].st_name
, name
) != 0)
1093 if (syms
[k
].st_shndx
== SHN_UNDEF
||
1094 syms
[k
].st_shndx
>= ehdr
->e_shnum
) {
1095 pr_debug("Symbol: %s has bad section index %d.\n",
1096 name
, syms
[k
].st_shndx
);
1100 /* Found the symbol we are looking for */
1108 void *kexec_purgatory_get_symbol_addr(struct kimage
*image
, const char *name
)
1110 struct purgatory_info
*pi
= &image
->purgatory_info
;
1114 sym
= kexec_purgatory_find_symbol(pi
, name
);
1116 return ERR_PTR(-EINVAL
);
1118 sechdr
= &pi
->sechdrs
[sym
->st_shndx
];
1121 * Returns the address where symbol will finally be loaded after
1122 * kexec_load_segment()
1124 return (void *)(sechdr
->sh_addr
+ sym
->st_value
);
1128 * Get or set value of a symbol. If "get_value" is true, symbol value is
1129 * returned in buf otherwise symbol value is set based on value in buf.
1131 int kexec_purgatory_get_set_symbol(struct kimage
*image
, const char *name
,
1132 void *buf
, unsigned int size
, bool get_value
)
1134 struct purgatory_info
*pi
= &image
->purgatory_info
;
1139 sym
= kexec_purgatory_find_symbol(pi
, name
);
1143 if (sym
->st_size
!= size
) {
1144 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1145 name
, (unsigned long)sym
->st_size
, size
);
1149 sec
= pi
->sechdrs
+ sym
->st_shndx
;
1151 if (sec
->sh_type
== SHT_NOBITS
) {
1152 pr_err("symbol %s is in a bss section. Cannot %s\n", name
,
1153 get_value
? "get" : "set");
1157 sym_buf
= (char *)pi
->purgatory_buf
+ sec
->sh_offset
+ sym
->st_value
;
1160 memcpy((void *)buf
, sym_buf
, size
);
1162 memcpy((void *)sym_buf
, buf
, size
);
1166 #endif /* CONFIG_ARCH_HAS_KEXEC_PURGATORY */
1168 int crash_exclude_mem_range(struct crash_mem
*mem
,
1169 unsigned long long mstart
, unsigned long long mend
)
1172 unsigned long long start
, end
, p_start
, p_end
;
1173 struct crash_mem_range temp_range
= {0, 0};
1175 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1176 start
= mem
->ranges
[i
].start
;
1177 end
= mem
->ranges
[i
].end
;
1181 if (mstart
> end
|| mend
< start
)
1184 /* Truncate any area outside of range */
1190 /* Found completely overlapping range */
1191 if (p_start
== start
&& p_end
== end
) {
1192 mem
->ranges
[i
].start
= 0;
1193 mem
->ranges
[i
].end
= 0;
1194 if (i
< mem
->nr_ranges
- 1) {
1195 /* Shift rest of the ranges to left */
1196 for (j
= i
; j
< mem
->nr_ranges
- 1; j
++) {
1197 mem
->ranges
[j
].start
=
1198 mem
->ranges
[j
+1].start
;
1199 mem
->ranges
[j
].end
=
1200 mem
->ranges
[j
+1].end
;
1204 * Continue to check if there are another overlapping ranges
1205 * from the current position because of shifting the above
1216 if (p_start
> start
&& p_end
< end
) {
1217 /* Split original range */
1218 mem
->ranges
[i
].end
= p_start
- 1;
1219 temp_range
.start
= p_end
+ 1;
1220 temp_range
.end
= end
;
1221 } else if (p_start
!= start
)
1222 mem
->ranges
[i
].end
= p_start
- 1;
1224 mem
->ranges
[i
].start
= p_end
+ 1;
1228 /* If a split happened, add the split to array */
1229 if (!temp_range
.end
)
1232 /* Split happened */
1233 if (i
== mem
->max_nr_ranges
- 1)
1236 /* Location where new range should go */
1238 if (j
< mem
->nr_ranges
) {
1239 /* Move over all ranges one slot towards the end */
1240 for (i
= mem
->nr_ranges
- 1; i
>= j
; i
--)
1241 mem
->ranges
[i
+ 1] = mem
->ranges
[i
];
1244 mem
->ranges
[j
].start
= temp_range
.start
;
1245 mem
->ranges
[j
].end
= temp_range
.end
;
1250 int crash_prepare_elf64_headers(struct crash_mem
*mem
, int kernel_map
,
1251 void **addr
, unsigned long *sz
)
1255 unsigned long nr_cpus
= num_possible_cpus(), nr_phdr
, elf_sz
;
1257 unsigned int cpu
, i
;
1258 unsigned long long notes_addr
;
1259 unsigned long mstart
, mend
;
1261 /* extra phdr for vmcoreinfo ELF note */
1262 nr_phdr
= nr_cpus
+ 1;
1263 nr_phdr
+= mem
->nr_ranges
;
1266 * kexec-tools creates an extra PT_LOAD phdr for kernel text mapping
1267 * area (for example, ffffffff80000000 - ffffffffa0000000 on x86_64).
1268 * I think this is required by tools like gdb. So same physical
1269 * memory will be mapped in two ELF headers. One will contain kernel
1270 * text virtual addresses and other will have __va(physical) addresses.
1274 elf_sz
= sizeof(Elf64_Ehdr
) + nr_phdr
* sizeof(Elf64_Phdr
);
1275 elf_sz
= ALIGN(elf_sz
, ELF_CORE_HEADER_ALIGN
);
1277 buf
= vzalloc(elf_sz
);
1281 ehdr
= (Elf64_Ehdr
*)buf
;
1282 phdr
= (Elf64_Phdr
*)(ehdr
+ 1);
1283 memcpy(ehdr
->e_ident
, ELFMAG
, SELFMAG
);
1284 ehdr
->e_ident
[EI_CLASS
] = ELFCLASS64
;
1285 ehdr
->e_ident
[EI_DATA
] = ELFDATA2LSB
;
1286 ehdr
->e_ident
[EI_VERSION
] = EV_CURRENT
;
1287 ehdr
->e_ident
[EI_OSABI
] = ELF_OSABI
;
1288 memset(ehdr
->e_ident
+ EI_PAD
, 0, EI_NIDENT
- EI_PAD
);
1289 ehdr
->e_type
= ET_CORE
;
1290 ehdr
->e_machine
= ELF_ARCH
;
1291 ehdr
->e_version
= EV_CURRENT
;
1292 ehdr
->e_phoff
= sizeof(Elf64_Ehdr
);
1293 ehdr
->e_ehsize
= sizeof(Elf64_Ehdr
);
1294 ehdr
->e_phentsize
= sizeof(Elf64_Phdr
);
1296 /* Prepare one phdr of type PT_NOTE for each present CPU */
1297 for_each_present_cpu(cpu
) {
1298 phdr
->p_type
= PT_NOTE
;
1299 notes_addr
= per_cpu_ptr_to_phys(per_cpu_ptr(crash_notes
, cpu
));
1300 phdr
->p_offset
= phdr
->p_paddr
= notes_addr
;
1301 phdr
->p_filesz
= phdr
->p_memsz
= sizeof(note_buf_t
);
1306 /* Prepare one PT_NOTE header for vmcoreinfo */
1307 phdr
->p_type
= PT_NOTE
;
1308 phdr
->p_offset
= phdr
->p_paddr
= paddr_vmcoreinfo_note();
1309 phdr
->p_filesz
= phdr
->p_memsz
= VMCOREINFO_NOTE_SIZE
;
1313 /* Prepare PT_LOAD type program header for kernel text region */
1315 phdr
->p_type
= PT_LOAD
;
1316 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1317 phdr
->p_vaddr
= (unsigned long) _text
;
1318 phdr
->p_filesz
= phdr
->p_memsz
= _end
- _text
;
1319 phdr
->p_offset
= phdr
->p_paddr
= __pa_symbol(_text
);
1324 /* Go through all the ranges in mem->ranges[] and prepare phdr */
1325 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1326 mstart
= mem
->ranges
[i
].start
;
1327 mend
= mem
->ranges
[i
].end
;
1329 phdr
->p_type
= PT_LOAD
;
1330 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1331 phdr
->p_offset
= mstart
;
1333 phdr
->p_paddr
= mstart
;
1334 phdr
->p_vaddr
= (unsigned long) __va(mstart
);
1335 phdr
->p_filesz
= phdr
->p_memsz
= mend
- mstart
+ 1;
1338 pr_debug("Crash PT_LOAD ELF header. phdr=%p vaddr=0x%llx, paddr=0x%llx, sz=0x%llx e_phnum=%d p_offset=0x%llx\n",
1339 phdr
, phdr
->p_vaddr
, phdr
->p_paddr
, phdr
->p_filesz
,
1340 ehdr
->e_phnum
, phdr
->p_offset
);