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_VERIFY_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
;
181 * In file mode list of segments is prepared by kernel. Copy relevant
182 * data from user space, do error checking, prepare segment list
185 kimage_file_prepare_segments(struct kimage
*image
, int kernel_fd
, int initrd_fd
,
186 const char __user
*cmdline_ptr
,
187 unsigned long cmdline_len
, unsigned flags
)
193 ret
= kernel_read_file_from_fd(kernel_fd
, &image
->kernel_buf
,
194 &size
, INT_MAX
, READING_KEXEC_IMAGE
);
197 image
->kernel_buf_len
= size
;
199 /* Call arch image probe handlers */
200 ret
= arch_kexec_kernel_image_probe(image
, image
->kernel_buf
,
201 image
->kernel_buf_len
);
205 #ifdef CONFIG_KEXEC_VERIFY_SIG
206 ret
= arch_kexec_kernel_verify_sig(image
, image
->kernel_buf
,
207 image
->kernel_buf_len
);
209 pr_debug("kernel signature verification failed.\n");
212 pr_debug("kernel signature verification successful.\n");
214 /* It is possible that there no initramfs is being loaded */
215 if (!(flags
& KEXEC_FILE_NO_INITRAMFS
)) {
216 ret
= kernel_read_file_from_fd(initrd_fd
, &image
->initrd_buf
,
218 READING_KEXEC_INITRAMFS
);
221 image
->initrd_buf_len
= size
;
225 image
->cmdline_buf
= memdup_user(cmdline_ptr
, cmdline_len
);
226 if (IS_ERR(image
->cmdline_buf
)) {
227 ret
= PTR_ERR(image
->cmdline_buf
);
228 image
->cmdline_buf
= NULL
;
232 image
->cmdline_buf_len
= cmdline_len
;
234 /* command line should be a string with last byte null */
235 if (image
->cmdline_buf
[cmdline_len
- 1] != '\0') {
240 ima_kexec_cmdline(image
->cmdline_buf
,
241 image
->cmdline_buf_len
- 1);
244 /* IMA needs to pass the measurement list to the next kernel. */
245 ima_add_kexec_buffer(image
);
247 /* Call arch image load handlers */
248 ldata
= arch_kexec_kernel_image_load(image
);
251 ret
= PTR_ERR(ldata
);
255 image
->image_loader_data
= ldata
;
257 /* In case of error, free up all allocated memory in this function */
259 kimage_file_post_load_cleanup(image
);
264 kimage_file_alloc_init(struct kimage
**rimage
, int kernel_fd
,
265 int initrd_fd
, const char __user
*cmdline_ptr
,
266 unsigned long cmdline_len
, unsigned long flags
)
269 struct kimage
*image
;
270 bool kexec_on_panic
= flags
& KEXEC_FILE_ON_CRASH
;
272 image
= do_kimage_alloc_init();
276 image
->file_mode
= 1;
278 if (kexec_on_panic
) {
279 /* Enable special crash kernel control page alloc policy. */
280 image
->control_page
= crashk_res
.start
;
281 image
->type
= KEXEC_TYPE_CRASH
;
284 ret
= kimage_file_prepare_segments(image
, kernel_fd
, initrd_fd
,
285 cmdline_ptr
, cmdline_len
, flags
);
289 ret
= sanity_check_segment_list(image
);
291 goto out_free_post_load_bufs
;
294 image
->control_code_page
= kimage_alloc_control_pages(image
,
295 get_order(KEXEC_CONTROL_PAGE_SIZE
));
296 if (!image
->control_code_page
) {
297 pr_err("Could not allocate control_code_buffer\n");
298 goto out_free_post_load_bufs
;
301 if (!kexec_on_panic
) {
302 image
->swap_page
= kimage_alloc_control_pages(image
, 0);
303 if (!image
->swap_page
) {
304 pr_err("Could not allocate swap buffer\n");
305 goto out_free_control_pages
;
311 out_free_control_pages
:
312 kimage_free_page_list(&image
->control_pages
);
313 out_free_post_load_bufs
:
314 kimage_file_post_load_cleanup(image
);
320 SYSCALL_DEFINE5(kexec_file_load
, int, kernel_fd
, int, initrd_fd
,
321 unsigned long, cmdline_len
, const char __user
*, cmdline_ptr
,
322 unsigned long, flags
)
325 struct kimage
**dest_image
, *image
;
327 /* We only trust the superuser with rebooting the system. */
328 if (!capable(CAP_SYS_BOOT
) || kexec_load_disabled
)
331 /* Make sure we have a legal set of flags */
332 if (flags
!= (flags
& KEXEC_FILE_FLAGS
))
337 if (!mutex_trylock(&kexec_mutex
))
340 dest_image
= &kexec_image
;
341 if (flags
& KEXEC_FILE_ON_CRASH
) {
342 dest_image
= &kexec_crash_image
;
343 if (kexec_crash_image
)
344 arch_kexec_unprotect_crashkres();
347 if (flags
& KEXEC_FILE_UNLOAD
)
351 * In case of crash, new kernel gets loaded in reserved region. It is
352 * same memory where old crash kernel might be loaded. Free any
353 * current crash dump kernel before we corrupt it.
355 if (flags
& KEXEC_FILE_ON_CRASH
)
356 kimage_free(xchg(&kexec_crash_image
, NULL
));
358 ret
= kimage_file_alloc_init(&image
, kernel_fd
, initrd_fd
, cmdline_ptr
,
363 ret
= machine_kexec_prepare(image
);
368 * Some architecture(like S390) may touch the crash memory before
369 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
371 ret
= kimage_crash_copy_vmcoreinfo(image
);
375 ret
= kexec_calculate_store_digests(image
);
379 for (i
= 0; i
< image
->nr_segments
; i
++) {
380 struct kexec_segment
*ksegment
;
382 ksegment
= &image
->segment
[i
];
383 pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
384 i
, ksegment
->buf
, ksegment
->bufsz
, ksegment
->mem
,
387 ret
= kimage_load_segment(image
, &image
->segment
[i
]);
392 kimage_terminate(image
);
395 * Free up any temporary buffers allocated which are not needed
396 * after image has been loaded
398 kimage_file_post_load_cleanup(image
);
400 image
= xchg(dest_image
, image
);
402 if ((flags
& KEXEC_FILE_ON_CRASH
) && kexec_crash_image
)
403 arch_kexec_protect_crashkres();
405 mutex_unlock(&kexec_mutex
);
410 static int locate_mem_hole_top_down(unsigned long start
, unsigned long end
,
411 struct kexec_buf
*kbuf
)
413 struct kimage
*image
= kbuf
->image
;
414 unsigned long temp_start
, temp_end
;
416 temp_end
= min(end
, kbuf
->buf_max
);
417 temp_start
= temp_end
- kbuf
->memsz
;
420 /* align down start */
421 temp_start
= temp_start
& (~(kbuf
->buf_align
- 1));
423 if (temp_start
< start
|| temp_start
< kbuf
->buf_min
)
426 temp_end
= temp_start
+ kbuf
->memsz
- 1;
429 * Make sure this does not conflict with any of existing
432 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
433 temp_start
= temp_start
- PAGE_SIZE
;
437 /* We found a suitable memory range */
441 /* If we are here, we found a suitable memory range */
442 kbuf
->mem
= temp_start
;
444 /* Success, stop navigating through remaining System RAM ranges */
448 static int locate_mem_hole_bottom_up(unsigned long start
, unsigned long end
,
449 struct kexec_buf
*kbuf
)
451 struct kimage
*image
= kbuf
->image
;
452 unsigned long temp_start
, temp_end
;
454 temp_start
= max(start
, kbuf
->buf_min
);
457 temp_start
= ALIGN(temp_start
, kbuf
->buf_align
);
458 temp_end
= temp_start
+ kbuf
->memsz
- 1;
460 if (temp_end
> end
|| temp_end
> kbuf
->buf_max
)
463 * Make sure this does not conflict with any of existing
466 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
467 temp_start
= temp_start
+ PAGE_SIZE
;
471 /* We found a suitable memory range */
475 /* If we are here, we found a suitable memory range */
476 kbuf
->mem
= temp_start
;
478 /* Success, stop navigating through remaining System RAM ranges */
482 static int locate_mem_hole_callback(struct resource
*res
, void *arg
)
484 struct kexec_buf
*kbuf
= (struct kexec_buf
*)arg
;
485 u64 start
= res
->start
, end
= res
->end
;
486 unsigned long sz
= end
- start
+ 1;
488 /* Returning 0 will take to next memory range */
489 if (sz
< kbuf
->memsz
)
492 if (end
< kbuf
->buf_min
|| start
> kbuf
->buf_max
)
496 * Allocate memory top down with-in ram range. Otherwise bottom up
500 return locate_mem_hole_top_down(start
, end
, kbuf
);
501 return locate_mem_hole_bottom_up(start
, end
, kbuf
);
504 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
505 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
506 int (*func
)(struct resource
*, void *))
510 phys_addr_t mstart
, mend
;
511 struct resource res
= { };
513 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
514 return func(&crashk_res
, kbuf
);
516 if (kbuf
->top_down
) {
517 for_each_free_mem_range_reverse(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
518 &mstart
, &mend
, NULL
) {
520 * In memblock, end points to the first byte after the
521 * range while in kexec, end points to the last byte
526 ret
= func(&res
, kbuf
);
531 for_each_free_mem_range(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
532 &mstart
, &mend
, NULL
) {
534 * In memblock, end points to the first byte after the
535 * range while in kexec, end points to the last byte
540 ret
= func(&res
, kbuf
);
549 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
550 int (*func
)(struct resource
*, void *))
557 * kexec_walk_resources - call func(data) on free memory regions
558 * @kbuf: Context info for the search. Also passed to @func.
559 * @func: Function to call for each memory region.
561 * Return: The memory walk will stop when func returns a non-zero value
562 * and that value will be returned. If all free regions are visited without
563 * func returning non-zero, then zero will be returned.
565 static int kexec_walk_resources(struct kexec_buf
*kbuf
,
566 int (*func
)(struct resource
*, void *))
568 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
569 return walk_iomem_res_desc(crashk_res
.desc
,
570 IORESOURCE_SYSTEM_RAM
| IORESOURCE_BUSY
,
571 crashk_res
.start
, crashk_res
.end
,
574 return walk_system_ram_res(0, ULONG_MAX
, kbuf
, func
);
578 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
579 * @kbuf: Parameters for the memory search.
581 * On success, kbuf->mem will have the start address of the memory region found.
583 * Return: 0 on success, negative errno on error.
585 int kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
589 /* Arch knows where to place */
590 if (kbuf
->mem
!= KEXEC_BUF_MEM_UNKNOWN
)
593 if (!IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK
))
594 ret
= kexec_walk_resources(kbuf
, locate_mem_hole_callback
);
596 ret
= kexec_walk_memblock(kbuf
, locate_mem_hole_callback
);
598 return ret
== 1 ? 0 : -EADDRNOTAVAIL
;
602 * kexec_add_buffer - place a buffer in a kexec segment
603 * @kbuf: Buffer contents and memory parameters.
605 * This function assumes that kexec_mutex is held.
606 * On successful return, @kbuf->mem will have the physical address of
607 * the buffer in memory.
609 * Return: 0 on success, negative errno on error.
611 int kexec_add_buffer(struct kexec_buf
*kbuf
)
614 struct kexec_segment
*ksegment
;
617 /* Currently adding segment this way is allowed only in file mode */
618 if (!kbuf
->image
->file_mode
)
621 if (kbuf
->image
->nr_segments
>= KEXEC_SEGMENT_MAX
)
625 * Make sure we are not trying to add buffer after allocating
626 * control pages. All segments need to be placed first before
627 * any control pages are allocated. As control page allocation
628 * logic goes through list of segments to make sure there are
629 * no destination overlaps.
631 if (!list_empty(&kbuf
->image
->control_pages
)) {
636 /* Ensure minimum alignment needed for segments. */
637 kbuf
->memsz
= ALIGN(kbuf
->memsz
, PAGE_SIZE
);
638 kbuf
->buf_align
= max(kbuf
->buf_align
, PAGE_SIZE
);
640 /* Walk the RAM ranges and allocate a suitable range for the buffer */
641 ret
= kexec_locate_mem_hole(kbuf
);
645 /* Found a suitable memory range */
646 ksegment
= &kbuf
->image
->segment
[kbuf
->image
->nr_segments
];
647 ksegment
->kbuf
= kbuf
->buffer
;
648 ksegment
->bufsz
= kbuf
->bufsz
;
649 ksegment
->mem
= kbuf
->mem
;
650 ksegment
->memsz
= kbuf
->memsz
;
651 kbuf
->image
->nr_segments
++;
655 /* Calculate and store the digest of segments */
656 static int kexec_calculate_store_digests(struct kimage
*image
)
658 struct crypto_shash
*tfm
;
659 struct shash_desc
*desc
;
660 int ret
= 0, i
, j
, zero_buf_sz
, sha_region_sz
;
661 size_t desc_size
, nullsz
;
664 struct kexec_sha_region
*sha_regions
;
665 struct purgatory_info
*pi
= &image
->purgatory_info
;
667 if (!IS_ENABLED(CONFIG_ARCH_HAS_KEXEC_PURGATORY
))
670 zero_buf
= __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT
);
671 zero_buf_sz
= PAGE_SIZE
;
673 tfm
= crypto_alloc_shash("sha256", 0, 0);
679 desc_size
= crypto_shash_descsize(tfm
) + sizeof(*desc
);
680 desc
= kzalloc(desc_size
, GFP_KERNEL
);
686 sha_region_sz
= KEXEC_SEGMENT_MAX
* sizeof(struct kexec_sha_region
);
687 sha_regions
= vzalloc(sha_region_sz
);
693 ret
= crypto_shash_init(desc
);
695 goto out_free_sha_regions
;
697 digest
= kzalloc(SHA256_DIGEST_SIZE
, GFP_KERNEL
);
700 goto out_free_sha_regions
;
703 for (j
= i
= 0; i
< image
->nr_segments
; i
++) {
704 struct kexec_segment
*ksegment
;
706 ksegment
= &image
->segment
[i
];
708 * Skip purgatory as it will be modified once we put digest
711 if (ksegment
->kbuf
== pi
->purgatory_buf
)
714 ret
= crypto_shash_update(desc
, ksegment
->kbuf
,
720 * Assume rest of the buffer is filled with zero and
721 * update digest accordingly.
723 nullsz
= ksegment
->memsz
- ksegment
->bufsz
;
725 unsigned long bytes
= nullsz
;
727 if (bytes
> zero_buf_sz
)
729 ret
= crypto_shash_update(desc
, zero_buf
, bytes
);
738 sha_regions
[j
].start
= ksegment
->mem
;
739 sha_regions
[j
].len
= ksegment
->memsz
;
744 ret
= crypto_shash_final(desc
, digest
);
746 goto out_free_digest
;
747 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha_regions",
748 sha_regions
, sha_region_sz
, 0);
750 goto out_free_digest
;
752 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha256_digest",
753 digest
, SHA256_DIGEST_SIZE
, 0);
755 goto out_free_digest
;
760 out_free_sha_regions
:
770 #ifdef CONFIG_ARCH_HAS_KEXEC_PURGATORY
772 * kexec_purgatory_setup_kbuf - prepare buffer to load purgatory.
773 * @pi: Purgatory to be loaded.
774 * @kbuf: Buffer to setup.
776 * Allocates the memory needed for the buffer. Caller is responsible to free
777 * the memory after use.
779 * Return: 0 on success, negative errno on error.
781 static int kexec_purgatory_setup_kbuf(struct purgatory_info
*pi
,
782 struct kexec_buf
*kbuf
)
784 const Elf_Shdr
*sechdrs
;
785 unsigned long bss_align
;
786 unsigned long bss_sz
;
790 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
791 kbuf
->buf_align
= bss_align
= 1;
792 kbuf
->bufsz
= bss_sz
= 0;
794 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
795 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
798 align
= sechdrs
[i
].sh_addralign
;
799 if (sechdrs
[i
].sh_type
!= SHT_NOBITS
) {
800 if (kbuf
->buf_align
< align
)
801 kbuf
->buf_align
= align
;
802 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, align
);
803 kbuf
->bufsz
+= sechdrs
[i
].sh_size
;
805 if (bss_align
< align
)
807 bss_sz
= ALIGN(bss_sz
, align
);
808 bss_sz
+= sechdrs
[i
].sh_size
;
811 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, bss_align
);
812 kbuf
->memsz
= kbuf
->bufsz
+ bss_sz
;
813 if (kbuf
->buf_align
< bss_align
)
814 kbuf
->buf_align
= bss_align
;
816 kbuf
->buffer
= vzalloc(kbuf
->bufsz
);
819 pi
->purgatory_buf
= kbuf
->buffer
;
821 ret
= kexec_add_buffer(kbuf
);
827 vfree(pi
->purgatory_buf
);
828 pi
->purgatory_buf
= NULL
;
833 * kexec_purgatory_setup_sechdrs - prepares the pi->sechdrs buffer.
834 * @pi: Purgatory to be loaded.
835 * @kbuf: Buffer prepared to store purgatory.
837 * Allocates the memory needed for the buffer. Caller is responsible to free
838 * the memory after use.
840 * Return: 0 on success, negative errno on error.
842 static int kexec_purgatory_setup_sechdrs(struct purgatory_info
*pi
,
843 struct kexec_buf
*kbuf
)
845 unsigned long bss_addr
;
846 unsigned long offset
;
851 * The section headers in kexec_purgatory are read-only. In order to
852 * have them modifiable make a temporary copy.
854 sechdrs
= vzalloc(array_size(sizeof(Elf_Shdr
), pi
->ehdr
->e_shnum
));
857 memcpy(sechdrs
, (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
,
858 pi
->ehdr
->e_shnum
* sizeof(Elf_Shdr
));
859 pi
->sechdrs
= sechdrs
;
862 bss_addr
= kbuf
->mem
+ kbuf
->bufsz
;
863 kbuf
->image
->start
= pi
->ehdr
->e_entry
;
865 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
869 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
872 align
= sechdrs
[i
].sh_addralign
;
873 if (sechdrs
[i
].sh_type
== SHT_NOBITS
) {
874 bss_addr
= ALIGN(bss_addr
, align
);
875 sechdrs
[i
].sh_addr
= bss_addr
;
876 bss_addr
+= sechdrs
[i
].sh_size
;
880 offset
= ALIGN(offset
, align
);
881 if (sechdrs
[i
].sh_flags
& SHF_EXECINSTR
&&
882 pi
->ehdr
->e_entry
>= sechdrs
[i
].sh_addr
&&
883 pi
->ehdr
->e_entry
< (sechdrs
[i
].sh_addr
884 + sechdrs
[i
].sh_size
)) {
885 kbuf
->image
->start
-= sechdrs
[i
].sh_addr
;
886 kbuf
->image
->start
+= kbuf
->mem
+ offset
;
889 src
= (void *)pi
->ehdr
+ sechdrs
[i
].sh_offset
;
890 dst
= pi
->purgatory_buf
+ offset
;
891 memcpy(dst
, src
, sechdrs
[i
].sh_size
);
893 sechdrs
[i
].sh_addr
= kbuf
->mem
+ offset
;
894 sechdrs
[i
].sh_offset
= offset
;
895 offset
+= sechdrs
[i
].sh_size
;
901 static int kexec_apply_relocations(struct kimage
*image
)
904 struct purgatory_info
*pi
= &image
->purgatory_info
;
905 const Elf_Shdr
*sechdrs
;
907 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
909 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
910 const Elf_Shdr
*relsec
;
911 const Elf_Shdr
*symtab
;
914 relsec
= sechdrs
+ i
;
916 if (relsec
->sh_type
!= SHT_RELA
&&
917 relsec
->sh_type
!= SHT_REL
)
921 * For section of type SHT_RELA/SHT_REL,
922 * ->sh_link contains section header index of associated
923 * symbol table. And ->sh_info contains section header
924 * index of section to which relocations apply.
926 if (relsec
->sh_info
>= pi
->ehdr
->e_shnum
||
927 relsec
->sh_link
>= pi
->ehdr
->e_shnum
)
930 section
= pi
->sechdrs
+ relsec
->sh_info
;
931 symtab
= sechdrs
+ relsec
->sh_link
;
933 if (!(section
->sh_flags
& SHF_ALLOC
))
937 * symtab->sh_link contain section header index of associated
940 if (symtab
->sh_link
>= pi
->ehdr
->e_shnum
)
941 /* Invalid section number? */
945 * Respective architecture needs to provide support for applying
946 * relocations of type SHT_RELA/SHT_REL.
948 if (relsec
->sh_type
== SHT_RELA
)
949 ret
= arch_kexec_apply_relocations_add(pi
, section
,
951 else if (relsec
->sh_type
== SHT_REL
)
952 ret
= arch_kexec_apply_relocations(pi
, section
,
962 * kexec_load_purgatory - Load and relocate the purgatory object.
963 * @image: Image to add the purgatory to.
964 * @kbuf: Memory parameters to use.
966 * Allocates the memory needed for image->purgatory_info.sechdrs and
967 * image->purgatory_info.purgatory_buf/kbuf->buffer. Caller is responsible
968 * to free the memory after use.
970 * Return: 0 on success, negative errno on error.
972 int kexec_load_purgatory(struct kimage
*image
, struct kexec_buf
*kbuf
)
974 struct purgatory_info
*pi
= &image
->purgatory_info
;
977 if (kexec_purgatory_size
<= 0)
980 pi
->ehdr
= (const Elf_Ehdr
*)kexec_purgatory
;
982 ret
= kexec_purgatory_setup_kbuf(pi
, kbuf
);
986 ret
= kexec_purgatory_setup_sechdrs(pi
, kbuf
);
990 ret
= kexec_apply_relocations(image
);
999 vfree(pi
->purgatory_buf
);
1000 pi
->purgatory_buf
= NULL
;
1005 * kexec_purgatory_find_symbol - find a symbol in the purgatory
1006 * @pi: Purgatory to search in.
1007 * @name: Name of the symbol.
1009 * Return: pointer to symbol in read-only symtab on success, NULL on error.
1011 static const Elf_Sym
*kexec_purgatory_find_symbol(struct purgatory_info
*pi
,
1014 const Elf_Shdr
*sechdrs
;
1015 const Elf_Ehdr
*ehdr
;
1016 const Elf_Sym
*syms
;
1024 sechdrs
= (void *)ehdr
+ ehdr
->e_shoff
;
1026 for (i
= 0; i
< ehdr
->e_shnum
; i
++) {
1027 if (sechdrs
[i
].sh_type
!= SHT_SYMTAB
)
1030 if (sechdrs
[i
].sh_link
>= ehdr
->e_shnum
)
1031 /* Invalid strtab section number */
1033 strtab
= (void *)ehdr
+ sechdrs
[sechdrs
[i
].sh_link
].sh_offset
;
1034 syms
= (void *)ehdr
+ sechdrs
[i
].sh_offset
;
1036 /* Go through symbols for a match */
1037 for (k
= 0; k
< sechdrs
[i
].sh_size
/sizeof(Elf_Sym
); k
++) {
1038 if (ELF_ST_BIND(syms
[k
].st_info
) != STB_GLOBAL
)
1041 if (strcmp(strtab
+ syms
[k
].st_name
, name
) != 0)
1044 if (syms
[k
].st_shndx
== SHN_UNDEF
||
1045 syms
[k
].st_shndx
>= ehdr
->e_shnum
) {
1046 pr_debug("Symbol: %s has bad section index %d.\n",
1047 name
, syms
[k
].st_shndx
);
1051 /* Found the symbol we are looking for */
1059 void *kexec_purgatory_get_symbol_addr(struct kimage
*image
, const char *name
)
1061 struct purgatory_info
*pi
= &image
->purgatory_info
;
1065 sym
= kexec_purgatory_find_symbol(pi
, name
);
1067 return ERR_PTR(-EINVAL
);
1069 sechdr
= &pi
->sechdrs
[sym
->st_shndx
];
1072 * Returns the address where symbol will finally be loaded after
1073 * kexec_load_segment()
1075 return (void *)(sechdr
->sh_addr
+ sym
->st_value
);
1079 * Get or set value of a symbol. If "get_value" is true, symbol value is
1080 * returned in buf otherwise symbol value is set based on value in buf.
1082 int kexec_purgatory_get_set_symbol(struct kimage
*image
, const char *name
,
1083 void *buf
, unsigned int size
, bool get_value
)
1085 struct purgatory_info
*pi
= &image
->purgatory_info
;
1090 sym
= kexec_purgatory_find_symbol(pi
, name
);
1094 if (sym
->st_size
!= size
) {
1095 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1096 name
, (unsigned long)sym
->st_size
, size
);
1100 sec
= pi
->sechdrs
+ sym
->st_shndx
;
1102 if (sec
->sh_type
== SHT_NOBITS
) {
1103 pr_err("symbol %s is in a bss section. Cannot %s\n", name
,
1104 get_value
? "get" : "set");
1108 sym_buf
= (char *)pi
->purgatory_buf
+ sec
->sh_offset
+ sym
->st_value
;
1111 memcpy((void *)buf
, sym_buf
, size
);
1113 memcpy((void *)sym_buf
, buf
, size
);
1117 #endif /* CONFIG_ARCH_HAS_KEXEC_PURGATORY */
1119 int crash_exclude_mem_range(struct crash_mem
*mem
,
1120 unsigned long long mstart
, unsigned long long mend
)
1123 unsigned long long start
, end
;
1124 struct crash_mem_range temp_range
= {0, 0};
1126 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1127 start
= mem
->ranges
[i
].start
;
1128 end
= mem
->ranges
[i
].end
;
1130 if (mstart
> end
|| mend
< start
)
1133 /* Truncate any area outside of range */
1139 /* Found completely overlapping range */
1140 if (mstart
== start
&& mend
== end
) {
1141 mem
->ranges
[i
].start
= 0;
1142 mem
->ranges
[i
].end
= 0;
1143 if (i
< mem
->nr_ranges
- 1) {
1144 /* Shift rest of the ranges to left */
1145 for (j
= i
; j
< mem
->nr_ranges
- 1; j
++) {
1146 mem
->ranges
[j
].start
=
1147 mem
->ranges
[j
+1].start
;
1148 mem
->ranges
[j
].end
=
1149 mem
->ranges
[j
+1].end
;
1156 if (mstart
> start
&& mend
< end
) {
1157 /* Split original range */
1158 mem
->ranges
[i
].end
= mstart
- 1;
1159 temp_range
.start
= mend
+ 1;
1160 temp_range
.end
= end
;
1161 } else if (mstart
!= start
)
1162 mem
->ranges
[i
].end
= mstart
- 1;
1164 mem
->ranges
[i
].start
= mend
+ 1;
1168 /* If a split happened, add the split to array */
1169 if (!temp_range
.end
)
1172 /* Split happened */
1173 if (i
== mem
->max_nr_ranges
- 1)
1176 /* Location where new range should go */
1178 if (j
< mem
->nr_ranges
) {
1179 /* Move over all ranges one slot towards the end */
1180 for (i
= mem
->nr_ranges
- 1; i
>= j
; i
--)
1181 mem
->ranges
[i
+ 1] = mem
->ranges
[i
];
1184 mem
->ranges
[j
].start
= temp_range
.start
;
1185 mem
->ranges
[j
].end
= temp_range
.end
;
1190 int crash_prepare_elf64_headers(struct crash_mem
*mem
, int kernel_map
,
1191 void **addr
, unsigned long *sz
)
1195 unsigned long nr_cpus
= num_possible_cpus(), nr_phdr
, elf_sz
;
1197 unsigned int cpu
, i
;
1198 unsigned long long notes_addr
;
1199 unsigned long mstart
, mend
;
1201 /* extra phdr for vmcoreinfo elf note */
1202 nr_phdr
= nr_cpus
+ 1;
1203 nr_phdr
+= mem
->nr_ranges
;
1206 * kexec-tools creates an extra PT_LOAD phdr for kernel text mapping
1207 * area (for example, ffffffff80000000 - ffffffffa0000000 on x86_64).
1208 * I think this is required by tools like gdb. So same physical
1209 * memory will be mapped in two elf headers. One will contain kernel
1210 * text virtual addresses and other will have __va(physical) addresses.
1214 elf_sz
= sizeof(Elf64_Ehdr
) + nr_phdr
* sizeof(Elf64_Phdr
);
1215 elf_sz
= ALIGN(elf_sz
, ELF_CORE_HEADER_ALIGN
);
1217 buf
= vzalloc(elf_sz
);
1221 ehdr
= (Elf64_Ehdr
*)buf
;
1222 phdr
= (Elf64_Phdr
*)(ehdr
+ 1);
1223 memcpy(ehdr
->e_ident
, ELFMAG
, SELFMAG
);
1224 ehdr
->e_ident
[EI_CLASS
] = ELFCLASS64
;
1225 ehdr
->e_ident
[EI_DATA
] = ELFDATA2LSB
;
1226 ehdr
->e_ident
[EI_VERSION
] = EV_CURRENT
;
1227 ehdr
->e_ident
[EI_OSABI
] = ELF_OSABI
;
1228 memset(ehdr
->e_ident
+ EI_PAD
, 0, EI_NIDENT
- EI_PAD
);
1229 ehdr
->e_type
= ET_CORE
;
1230 ehdr
->e_machine
= ELF_ARCH
;
1231 ehdr
->e_version
= EV_CURRENT
;
1232 ehdr
->e_phoff
= sizeof(Elf64_Ehdr
);
1233 ehdr
->e_ehsize
= sizeof(Elf64_Ehdr
);
1234 ehdr
->e_phentsize
= sizeof(Elf64_Phdr
);
1236 /* Prepare one phdr of type PT_NOTE for each present cpu */
1237 for_each_present_cpu(cpu
) {
1238 phdr
->p_type
= PT_NOTE
;
1239 notes_addr
= per_cpu_ptr_to_phys(per_cpu_ptr(crash_notes
, cpu
));
1240 phdr
->p_offset
= phdr
->p_paddr
= notes_addr
;
1241 phdr
->p_filesz
= phdr
->p_memsz
= sizeof(note_buf_t
);
1246 /* Prepare one PT_NOTE header for vmcoreinfo */
1247 phdr
->p_type
= PT_NOTE
;
1248 phdr
->p_offset
= phdr
->p_paddr
= paddr_vmcoreinfo_note();
1249 phdr
->p_filesz
= phdr
->p_memsz
= VMCOREINFO_NOTE_SIZE
;
1253 /* Prepare PT_LOAD type program header for kernel text region */
1255 phdr
->p_type
= PT_LOAD
;
1256 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1257 phdr
->p_vaddr
= (Elf64_Addr
)_text
;
1258 phdr
->p_filesz
= phdr
->p_memsz
= _end
- _text
;
1259 phdr
->p_offset
= phdr
->p_paddr
= __pa_symbol(_text
);
1264 /* Go through all the ranges in mem->ranges[] and prepare phdr */
1265 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1266 mstart
= mem
->ranges
[i
].start
;
1267 mend
= mem
->ranges
[i
].end
;
1269 phdr
->p_type
= PT_LOAD
;
1270 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1271 phdr
->p_offset
= mstart
;
1273 phdr
->p_paddr
= mstart
;
1274 phdr
->p_vaddr
= (unsigned long long) __va(mstart
);
1275 phdr
->p_filesz
= phdr
->p_memsz
= mend
- mstart
+ 1;
1279 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",
1280 phdr
, phdr
->p_vaddr
, phdr
->p_paddr
, phdr
->p_filesz
,
1281 ehdr
->e_phnum
, phdr
->p_offset
);