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/sha2.h>
24 #include <linux/elf.h>
25 #include <linux/elfcore.h>
26 #include <linux/kernel.h>
27 #include <linux/kernel_read_file.h>
28 #include <linux/syscalls.h>
29 #include <linux/vmalloc.h>
30 #include "kexec_internal.h"
32 #ifdef CONFIG_KEXEC_SIG
33 static bool sig_enforce
= IS_ENABLED(CONFIG_KEXEC_SIG_FORCE
);
35 void set_kexec_sig_enforced(void)
41 static int kexec_calculate_store_digests(struct kimage
*image
);
43 /* Maximum size in bytes for kernel/initrd files. */
44 #define KEXEC_FILE_SIZE_MAX min_t(s64, 4LL << 30, SSIZE_MAX)
47 * Currently this is the only default function that is exported as some
48 * architectures need it to do additional handlings.
49 * In the future, other default functions may be exported too if required.
51 int kexec_image_probe_default(struct kimage
*image
, void *buf
,
52 unsigned long buf_len
)
54 const struct kexec_file_ops
* const *fops
;
57 for (fops
= &kexec_file_loaders
[0]; *fops
&& (*fops
)->probe
; ++fops
) {
58 ret
= (*fops
)->probe(buf
, buf_len
);
68 static void *kexec_image_load_default(struct kimage
*image
)
70 if (!image
->fops
|| !image
->fops
->load
)
71 return ERR_PTR(-ENOEXEC
);
73 return image
->fops
->load(image
, image
->kernel_buf
,
74 image
->kernel_buf_len
, image
->initrd_buf
,
75 image
->initrd_buf_len
, image
->cmdline_buf
,
76 image
->cmdline_buf_len
);
79 int kexec_image_post_load_cleanup_default(struct kimage
*image
)
81 if (!image
->fops
|| !image
->fops
->cleanup
)
84 return image
->fops
->cleanup(image
->image_loader_data
);
88 * Free up memory used by kernel, initrd, and command line. This is temporary
89 * memory allocation which is not needed any more after these buffers have
90 * been loaded into separate segments and have been copied elsewhere.
92 void kimage_file_post_load_cleanup(struct kimage
*image
)
94 struct purgatory_info
*pi
= &image
->purgatory_info
;
96 vfree(image
->kernel_buf
);
97 image
->kernel_buf
= NULL
;
99 vfree(image
->initrd_buf
);
100 image
->initrd_buf
= NULL
;
102 kfree(image
->cmdline_buf
);
103 image
->cmdline_buf
= NULL
;
105 vfree(pi
->purgatory_buf
);
106 pi
->purgatory_buf
= NULL
;
111 #ifdef CONFIG_IMA_KEXEC
112 vfree(image
->ima_buffer
);
113 image
->ima_buffer
= NULL
;
114 #endif /* CONFIG_IMA_KEXEC */
116 /* See if architecture has anything to cleanup post load */
117 arch_kimage_file_post_load_cleanup(image
);
120 * Above call should have called into bootloader to free up
121 * any data stored in kimage->image_loader_data. It should
122 * be ok now to free it up.
124 kfree(image
->image_loader_data
);
125 image
->image_loader_data
= NULL
;
128 #ifdef CONFIG_KEXEC_SIG
129 #ifdef CONFIG_SIGNED_PE_FILE_VERIFICATION
130 int kexec_kernel_verify_pe_sig(const char *kernel
, unsigned long kernel_len
)
134 ret
= verify_pefile_signature(kernel
, kernel_len
,
135 VERIFY_USE_SECONDARY_KEYRING
,
136 VERIFYING_KEXEC_PE_SIGNATURE
);
137 if (ret
== -ENOKEY
&& IS_ENABLED(CONFIG_INTEGRITY_PLATFORM_KEYRING
)) {
138 ret
= verify_pefile_signature(kernel
, kernel_len
,
139 VERIFY_USE_PLATFORM_KEYRING
,
140 VERIFYING_KEXEC_PE_SIGNATURE
);
146 static int kexec_image_verify_sig(struct kimage
*image
, void *buf
,
147 unsigned long buf_len
)
149 if (!image
->fops
|| !image
->fops
->verify_sig
) {
150 pr_debug("kernel loader does not support signature verification.\n");
151 return -EKEYREJECTED
;
154 return image
->fops
->verify_sig(buf
, buf_len
);
158 kimage_validate_signature(struct kimage
*image
)
162 ret
= kexec_image_verify_sig(image
, image
->kernel_buf
,
163 image
->kernel_buf_len
);
167 pr_notice("Enforced kernel signature verification failed (%d).\n", ret
);
172 * If IMA is guaranteed to appraise a signature on the kexec
173 * image, permit it even if the kernel is otherwise locked
176 if (!ima_appraise_signature(READING_KEXEC_IMAGE
) &&
177 security_locked_down(LOCKDOWN_KEXEC
))
180 pr_debug("kernel signature verification failed (%d).\n", ret
);
188 * In file mode list of segments is prepared by kernel. Copy relevant
189 * data from user space, do error checking, prepare segment list
192 kimage_file_prepare_segments(struct kimage
*image
, int kernel_fd
, int initrd_fd
,
193 const char __user
*cmdline_ptr
,
194 unsigned long cmdline_len
, unsigned flags
)
199 ret
= kernel_read_file_from_fd(kernel_fd
, 0, &image
->kernel_buf
,
200 KEXEC_FILE_SIZE_MAX
, NULL
,
201 READING_KEXEC_IMAGE
);
204 image
->kernel_buf_len
= ret
;
206 /* Call arch image probe handlers */
207 ret
= arch_kexec_kernel_image_probe(image
, image
->kernel_buf
,
208 image
->kernel_buf_len
);
212 #ifdef CONFIG_KEXEC_SIG
213 ret
= kimage_validate_signature(image
);
218 /* It is possible that there no initramfs is being loaded */
219 if (!(flags
& KEXEC_FILE_NO_INITRAMFS
)) {
220 ret
= kernel_read_file_from_fd(initrd_fd
, 0, &image
->initrd_buf
,
221 KEXEC_FILE_SIZE_MAX
, NULL
,
222 READING_KEXEC_INITRAMFS
);
225 image
->initrd_buf_len
= ret
;
230 image
->cmdline_buf
= memdup_user(cmdline_ptr
, cmdline_len
);
231 if (IS_ERR(image
->cmdline_buf
)) {
232 ret
= PTR_ERR(image
->cmdline_buf
);
233 image
->cmdline_buf
= NULL
;
237 image
->cmdline_buf_len
= cmdline_len
;
239 /* command line should be a string with last byte null */
240 if (image
->cmdline_buf
[cmdline_len
- 1] != '\0') {
245 ima_kexec_cmdline(kernel_fd
, image
->cmdline_buf
,
246 image
->cmdline_buf_len
- 1);
249 /* IMA needs to pass the measurement list to the next kernel. */
250 ima_add_kexec_buffer(image
);
252 /* Call image load handler */
253 ldata
= kexec_image_load_default(image
);
256 ret
= PTR_ERR(ldata
);
260 image
->image_loader_data
= ldata
;
262 /* In case of error, free up all allocated memory in this function */
264 kimage_file_post_load_cleanup(image
);
269 kimage_file_alloc_init(struct kimage
**rimage
, int kernel_fd
,
270 int initrd_fd
, const char __user
*cmdline_ptr
,
271 unsigned long cmdline_len
, unsigned long flags
)
274 struct kimage
*image
;
275 bool kexec_on_panic
= flags
& KEXEC_FILE_ON_CRASH
;
277 image
= do_kimage_alloc_init();
281 image
->file_mode
= 1;
283 if (kexec_on_panic
) {
284 /* Enable special crash kernel control page alloc policy. */
285 image
->control_page
= crashk_res
.start
;
286 image
->type
= KEXEC_TYPE_CRASH
;
289 ret
= kimage_file_prepare_segments(image
, kernel_fd
, initrd_fd
,
290 cmdline_ptr
, cmdline_len
, flags
);
294 ret
= sanity_check_segment_list(image
);
296 goto out_free_post_load_bufs
;
299 image
->control_code_page
= kimage_alloc_control_pages(image
,
300 get_order(KEXEC_CONTROL_PAGE_SIZE
));
301 if (!image
->control_code_page
) {
302 pr_err("Could not allocate control_code_buffer\n");
303 goto out_free_post_load_bufs
;
306 if (!kexec_on_panic
) {
307 image
->swap_page
= kimage_alloc_control_pages(image
, 0);
308 if (!image
->swap_page
) {
309 pr_err("Could not allocate swap buffer\n");
310 goto out_free_control_pages
;
316 out_free_control_pages
:
317 kimage_free_page_list(&image
->control_pages
);
318 out_free_post_load_bufs
:
319 kimage_file_post_load_cleanup(image
);
325 SYSCALL_DEFINE5(kexec_file_load
, int, kernel_fd
, int, initrd_fd
,
326 unsigned long, cmdline_len
, const char __user
*, cmdline_ptr
,
327 unsigned long, flags
)
329 int image_type
= (flags
& KEXEC_FILE_ON_CRASH
) ?
330 KEXEC_TYPE_CRASH
: KEXEC_TYPE_DEFAULT
;
331 struct kimage
**dest_image
, *image
;
334 /* We only trust the superuser with rebooting the system. */
335 if (!kexec_load_permitted(image_type
))
338 /* Make sure we have a legal set of flags */
339 if (flags
!= (flags
& KEXEC_FILE_FLAGS
))
344 if (!kexec_trylock())
347 if (image_type
== KEXEC_TYPE_CRASH
) {
348 dest_image
= &kexec_crash_image
;
349 if (kexec_crash_image
)
350 arch_kexec_unprotect_crashkres();
352 dest_image
= &kexec_image
;
355 if (flags
& KEXEC_FILE_UNLOAD
)
359 * In case of crash, new kernel gets loaded in reserved region. It is
360 * same memory where old crash kernel might be loaded. Free any
361 * current crash dump kernel before we corrupt it.
363 if (flags
& KEXEC_FILE_ON_CRASH
)
364 kimage_free(xchg(&kexec_crash_image
, NULL
));
366 ret
= kimage_file_alloc_init(&image
, kernel_fd
, initrd_fd
, cmdline_ptr
,
371 ret
= machine_kexec_prepare(image
);
376 * Some architecture(like S390) may touch the crash memory before
377 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
379 ret
= kimage_crash_copy_vmcoreinfo(image
);
383 ret
= kexec_calculate_store_digests(image
);
387 for (i
= 0; i
< image
->nr_segments
; i
++) {
388 struct kexec_segment
*ksegment
;
390 ksegment
= &image
->segment
[i
];
391 pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
392 i
, ksegment
->buf
, ksegment
->bufsz
, ksegment
->mem
,
395 ret
= kimage_load_segment(image
, &image
->segment
[i
]);
400 kimage_terminate(image
);
402 ret
= machine_kexec_post_load(image
);
407 * Free up any temporary buffers allocated which are not needed
408 * after image has been loaded
410 kimage_file_post_load_cleanup(image
);
412 image
= xchg(dest_image
, image
);
414 if ((flags
& KEXEC_FILE_ON_CRASH
) && kexec_crash_image
)
415 arch_kexec_protect_crashkres();
422 static int locate_mem_hole_top_down(unsigned long start
, unsigned long end
,
423 struct kexec_buf
*kbuf
)
425 struct kimage
*image
= kbuf
->image
;
426 unsigned long temp_start
, temp_end
;
428 temp_end
= min(end
, kbuf
->buf_max
);
429 temp_start
= temp_end
- kbuf
->memsz
;
432 /* align down start */
433 temp_start
= temp_start
& (~(kbuf
->buf_align
- 1));
435 if (temp_start
< start
|| temp_start
< kbuf
->buf_min
)
438 temp_end
= temp_start
+ kbuf
->memsz
- 1;
441 * Make sure this does not conflict with any of existing
444 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
445 temp_start
= temp_start
- PAGE_SIZE
;
449 /* We found a suitable memory range */
453 /* If we are here, we found a suitable memory range */
454 kbuf
->mem
= temp_start
;
456 /* Success, stop navigating through remaining System RAM ranges */
460 static int locate_mem_hole_bottom_up(unsigned long start
, unsigned long end
,
461 struct kexec_buf
*kbuf
)
463 struct kimage
*image
= kbuf
->image
;
464 unsigned long temp_start
, temp_end
;
466 temp_start
= max(start
, kbuf
->buf_min
);
469 temp_start
= ALIGN(temp_start
, kbuf
->buf_align
);
470 temp_end
= temp_start
+ kbuf
->memsz
- 1;
472 if (temp_end
> end
|| temp_end
> kbuf
->buf_max
)
475 * Make sure this does not conflict with any of existing
478 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
479 temp_start
= temp_start
+ PAGE_SIZE
;
483 /* We found a suitable memory range */
487 /* If we are here, we found a suitable memory range */
488 kbuf
->mem
= temp_start
;
490 /* Success, stop navigating through remaining System RAM ranges */
494 static int locate_mem_hole_callback(struct resource
*res
, void *arg
)
496 struct kexec_buf
*kbuf
= (struct kexec_buf
*)arg
;
497 u64 start
= res
->start
, end
= res
->end
;
498 unsigned long sz
= end
- start
+ 1;
500 /* Returning 0 will take to next memory range */
502 /* Don't use memory that will be detected and handled by a driver. */
503 if (res
->flags
& IORESOURCE_SYSRAM_DRIVER_MANAGED
)
506 if (sz
< kbuf
->memsz
)
509 if (end
< kbuf
->buf_min
|| start
> kbuf
->buf_max
)
513 * Allocate memory top down with-in ram range. Otherwise bottom up
517 return locate_mem_hole_top_down(start
, end
, kbuf
);
518 return locate_mem_hole_bottom_up(start
, end
, kbuf
);
521 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
522 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
523 int (*func
)(struct resource
*, void *))
527 phys_addr_t mstart
, mend
;
528 struct resource res
= { };
530 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
531 return func(&crashk_res
, kbuf
);
534 * Using MEMBLOCK_NONE will properly skip MEMBLOCK_DRIVER_MANAGED. See
535 * IORESOURCE_SYSRAM_DRIVER_MANAGED handling in
536 * locate_mem_hole_callback().
538 if (kbuf
->top_down
) {
539 for_each_free_mem_range_reverse(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
540 &mstart
, &mend
, NULL
) {
542 * In memblock, end points to the first byte after the
543 * range while in kexec, end points to the last byte
548 ret
= func(&res
, kbuf
);
553 for_each_free_mem_range(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
554 &mstart
, &mend
, NULL
) {
556 * In memblock, end points to the first byte after the
557 * range while in kexec, end points to the last byte
562 ret
= func(&res
, kbuf
);
571 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
572 int (*func
)(struct resource
*, void *))
579 * kexec_walk_resources - call func(data) on free memory regions
580 * @kbuf: Context info for the search. Also passed to @func.
581 * @func: Function to call for each memory region.
583 * Return: The memory walk will stop when func returns a non-zero value
584 * and that value will be returned. If all free regions are visited without
585 * func returning non-zero, then zero will be returned.
587 static int kexec_walk_resources(struct kexec_buf
*kbuf
,
588 int (*func
)(struct resource
*, void *))
590 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
591 return walk_iomem_res_desc(crashk_res
.desc
,
592 IORESOURCE_SYSTEM_RAM
| IORESOURCE_BUSY
,
593 crashk_res
.start
, crashk_res
.end
,
596 return walk_system_ram_res(0, ULONG_MAX
, kbuf
, func
);
600 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
601 * @kbuf: Parameters for the memory search.
603 * On success, kbuf->mem will have the start address of the memory region found.
605 * Return: 0 on success, negative errno on error.
607 int kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
611 /* Arch knows where to place */
612 if (kbuf
->mem
!= KEXEC_BUF_MEM_UNKNOWN
)
615 if (!IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK
))
616 ret
= kexec_walk_resources(kbuf
, locate_mem_hole_callback
);
618 ret
= kexec_walk_memblock(kbuf
, locate_mem_hole_callback
);
620 return ret
== 1 ? 0 : -EADDRNOTAVAIL
;
624 * kexec_add_buffer - place a buffer in a kexec segment
625 * @kbuf: Buffer contents and memory parameters.
627 * This function assumes that kexec_lock is held.
628 * On successful return, @kbuf->mem will have the physical address of
629 * the buffer in memory.
631 * Return: 0 on success, negative errno on error.
633 int kexec_add_buffer(struct kexec_buf
*kbuf
)
635 struct kexec_segment
*ksegment
;
638 /* Currently adding segment this way is allowed only in file mode */
639 if (!kbuf
->image
->file_mode
)
642 if (kbuf
->image
->nr_segments
>= KEXEC_SEGMENT_MAX
)
646 * Make sure we are not trying to add buffer after allocating
647 * control pages. All segments need to be placed first before
648 * any control pages are allocated. As control page allocation
649 * logic goes through list of segments to make sure there are
650 * no destination overlaps.
652 if (!list_empty(&kbuf
->image
->control_pages
)) {
657 /* Ensure minimum alignment needed for segments. */
658 kbuf
->memsz
= ALIGN(kbuf
->memsz
, PAGE_SIZE
);
659 kbuf
->buf_align
= max(kbuf
->buf_align
, PAGE_SIZE
);
661 /* Walk the RAM ranges and allocate a suitable range for the buffer */
662 ret
= arch_kexec_locate_mem_hole(kbuf
);
666 /* Found a suitable memory range */
667 ksegment
= &kbuf
->image
->segment
[kbuf
->image
->nr_segments
];
668 ksegment
->kbuf
= kbuf
->buffer
;
669 ksegment
->bufsz
= kbuf
->bufsz
;
670 ksegment
->mem
= kbuf
->mem
;
671 ksegment
->memsz
= kbuf
->memsz
;
672 kbuf
->image
->nr_segments
++;
676 /* Calculate and store the digest of segments */
677 static int kexec_calculate_store_digests(struct kimage
*image
)
679 struct crypto_shash
*tfm
;
680 struct shash_desc
*desc
;
681 int ret
= 0, i
, j
, zero_buf_sz
, sha_region_sz
;
682 size_t desc_size
, nullsz
;
685 struct kexec_sha_region
*sha_regions
;
686 struct purgatory_info
*pi
= &image
->purgatory_info
;
688 if (!IS_ENABLED(CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY
))
691 zero_buf
= __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT
);
692 zero_buf_sz
= PAGE_SIZE
;
694 tfm
= crypto_alloc_shash("sha256", 0, 0);
700 desc_size
= crypto_shash_descsize(tfm
) + sizeof(*desc
);
701 desc
= kzalloc(desc_size
, GFP_KERNEL
);
707 sha_region_sz
= KEXEC_SEGMENT_MAX
* sizeof(struct kexec_sha_region
);
708 sha_regions
= vzalloc(sha_region_sz
);
716 ret
= crypto_shash_init(desc
);
718 goto out_free_sha_regions
;
720 digest
= kzalloc(SHA256_DIGEST_SIZE
, GFP_KERNEL
);
723 goto out_free_sha_regions
;
726 for (j
= i
= 0; i
< image
->nr_segments
; i
++) {
727 struct kexec_segment
*ksegment
;
729 #ifdef CONFIG_CRASH_HOTPLUG
730 /* Exclude elfcorehdr segment to allow future changes via hotplug */
731 if (j
== image
->elfcorehdr_index
)
735 ksegment
= &image
->segment
[i
];
737 * Skip purgatory as it will be modified once we put digest
740 if (ksegment
->kbuf
== pi
->purgatory_buf
)
743 ret
= crypto_shash_update(desc
, ksegment
->kbuf
,
749 * Assume rest of the buffer is filled with zero and
750 * update digest accordingly.
752 nullsz
= ksegment
->memsz
- ksegment
->bufsz
;
754 unsigned long bytes
= nullsz
;
756 if (bytes
> zero_buf_sz
)
758 ret
= crypto_shash_update(desc
, zero_buf
, bytes
);
767 sha_regions
[j
].start
= ksegment
->mem
;
768 sha_regions
[j
].len
= ksegment
->memsz
;
773 ret
= crypto_shash_final(desc
, digest
);
775 goto out_free_digest
;
776 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha_regions",
777 sha_regions
, sha_region_sz
, 0);
779 goto out_free_digest
;
781 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha256_digest",
782 digest
, SHA256_DIGEST_SIZE
, 0);
784 goto out_free_digest
;
789 out_free_sha_regions
:
799 #ifdef CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY
801 * kexec_purgatory_setup_kbuf - prepare buffer to load purgatory.
802 * @pi: Purgatory to be loaded.
803 * @kbuf: Buffer to setup.
805 * Allocates the memory needed for the buffer. Caller is responsible to free
806 * the memory after use.
808 * Return: 0 on success, negative errno on error.
810 static int kexec_purgatory_setup_kbuf(struct purgatory_info
*pi
,
811 struct kexec_buf
*kbuf
)
813 const Elf_Shdr
*sechdrs
;
814 unsigned long bss_align
;
815 unsigned long bss_sz
;
819 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
820 kbuf
->buf_align
= bss_align
= 1;
821 kbuf
->bufsz
= bss_sz
= 0;
823 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
824 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
827 align
= sechdrs
[i
].sh_addralign
;
828 if (sechdrs
[i
].sh_type
!= SHT_NOBITS
) {
829 if (kbuf
->buf_align
< align
)
830 kbuf
->buf_align
= align
;
831 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, align
);
832 kbuf
->bufsz
+= sechdrs
[i
].sh_size
;
834 if (bss_align
< align
)
836 bss_sz
= ALIGN(bss_sz
, align
);
837 bss_sz
+= sechdrs
[i
].sh_size
;
840 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, bss_align
);
841 kbuf
->memsz
= kbuf
->bufsz
+ bss_sz
;
842 if (kbuf
->buf_align
< bss_align
)
843 kbuf
->buf_align
= bss_align
;
845 kbuf
->buffer
= vzalloc(kbuf
->bufsz
);
848 pi
->purgatory_buf
= kbuf
->buffer
;
850 ret
= kexec_add_buffer(kbuf
);
856 vfree(pi
->purgatory_buf
);
857 pi
->purgatory_buf
= NULL
;
862 * kexec_purgatory_setup_sechdrs - prepares the pi->sechdrs buffer.
863 * @pi: Purgatory to be loaded.
864 * @kbuf: Buffer prepared to store purgatory.
866 * Allocates the memory needed for the buffer. Caller is responsible to free
867 * the memory after use.
869 * Return: 0 on success, negative errno on error.
871 static int kexec_purgatory_setup_sechdrs(struct purgatory_info
*pi
,
872 struct kexec_buf
*kbuf
)
874 unsigned long bss_addr
;
875 unsigned long offset
;
881 * The section headers in kexec_purgatory are read-only. In order to
882 * have them modifiable make a temporary copy.
884 sechdrs_size
= array_size(sizeof(Elf_Shdr
), pi
->ehdr
->e_shnum
);
885 sechdrs
= vzalloc(sechdrs_size
);
888 memcpy(sechdrs
, (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
, sechdrs_size
);
889 pi
->sechdrs
= sechdrs
;
892 bss_addr
= kbuf
->mem
+ kbuf
->bufsz
;
893 kbuf
->image
->start
= pi
->ehdr
->e_entry
;
895 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
899 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
902 align
= sechdrs
[i
].sh_addralign
;
903 if (sechdrs
[i
].sh_type
== SHT_NOBITS
) {
904 bss_addr
= ALIGN(bss_addr
, align
);
905 sechdrs
[i
].sh_addr
= bss_addr
;
906 bss_addr
+= sechdrs
[i
].sh_size
;
910 offset
= ALIGN(offset
, align
);
913 * Check if the segment contains the entry point, if so,
914 * calculate the value of image->start based on it.
915 * If the compiler has produced more than one .text section
916 * (Eg: .text.hot), they are generally after the main .text
917 * section, and they shall not be used to calculate
918 * image->start. So do not re-calculate image->start if it
919 * is not set to the initial value, and warn the user so they
920 * have a chance to fix their purgatory's linker script.
922 if (sechdrs
[i
].sh_flags
& SHF_EXECINSTR
&&
923 pi
->ehdr
->e_entry
>= sechdrs
[i
].sh_addr
&&
924 pi
->ehdr
->e_entry
< (sechdrs
[i
].sh_addr
925 + sechdrs
[i
].sh_size
) &&
926 !WARN_ON(kbuf
->image
->start
!= pi
->ehdr
->e_entry
)) {
927 kbuf
->image
->start
-= sechdrs
[i
].sh_addr
;
928 kbuf
->image
->start
+= kbuf
->mem
+ offset
;
931 src
= (void *)pi
->ehdr
+ sechdrs
[i
].sh_offset
;
932 dst
= pi
->purgatory_buf
+ offset
;
933 memcpy(dst
, src
, sechdrs
[i
].sh_size
);
935 sechdrs
[i
].sh_addr
= kbuf
->mem
+ offset
;
936 sechdrs
[i
].sh_offset
= offset
;
937 offset
+= sechdrs
[i
].sh_size
;
943 static int kexec_apply_relocations(struct kimage
*image
)
946 struct purgatory_info
*pi
= &image
->purgatory_info
;
947 const Elf_Shdr
*sechdrs
;
949 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
951 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
952 const Elf_Shdr
*relsec
;
953 const Elf_Shdr
*symtab
;
956 relsec
= sechdrs
+ i
;
958 if (relsec
->sh_type
!= SHT_RELA
&&
959 relsec
->sh_type
!= SHT_REL
)
963 * For section of type SHT_RELA/SHT_REL,
964 * ->sh_link contains section header index of associated
965 * symbol table. And ->sh_info contains section header
966 * index of section to which relocations apply.
968 if (relsec
->sh_info
>= pi
->ehdr
->e_shnum
||
969 relsec
->sh_link
>= pi
->ehdr
->e_shnum
)
972 section
= pi
->sechdrs
+ relsec
->sh_info
;
973 symtab
= sechdrs
+ relsec
->sh_link
;
975 if (!(section
->sh_flags
& SHF_ALLOC
))
979 * symtab->sh_link contain section header index of associated
982 if (symtab
->sh_link
>= pi
->ehdr
->e_shnum
)
983 /* Invalid section number? */
987 * Respective architecture needs to provide support for applying
988 * relocations of type SHT_RELA/SHT_REL.
990 if (relsec
->sh_type
== SHT_RELA
)
991 ret
= arch_kexec_apply_relocations_add(pi
, section
,
993 else if (relsec
->sh_type
== SHT_REL
)
994 ret
= arch_kexec_apply_relocations(pi
, section
,
1004 * kexec_load_purgatory - Load and relocate the purgatory object.
1005 * @image: Image to add the purgatory to.
1006 * @kbuf: Memory parameters to use.
1008 * Allocates the memory needed for image->purgatory_info.sechdrs and
1009 * image->purgatory_info.purgatory_buf/kbuf->buffer. Caller is responsible
1010 * to free the memory after use.
1012 * Return: 0 on success, negative errno on error.
1014 int kexec_load_purgatory(struct kimage
*image
, struct kexec_buf
*kbuf
)
1016 struct purgatory_info
*pi
= &image
->purgatory_info
;
1019 if (kexec_purgatory_size
<= 0)
1022 pi
->ehdr
= (const Elf_Ehdr
*)kexec_purgatory
;
1024 ret
= kexec_purgatory_setup_kbuf(pi
, kbuf
);
1028 ret
= kexec_purgatory_setup_sechdrs(pi
, kbuf
);
1032 ret
= kexec_apply_relocations(image
);
1041 vfree(pi
->purgatory_buf
);
1042 pi
->purgatory_buf
= NULL
;
1047 * kexec_purgatory_find_symbol - find a symbol in the purgatory
1048 * @pi: Purgatory to search in.
1049 * @name: Name of the symbol.
1051 * Return: pointer to symbol in read-only symtab on success, NULL on error.
1053 static const Elf_Sym
*kexec_purgatory_find_symbol(struct purgatory_info
*pi
,
1056 const Elf_Shdr
*sechdrs
;
1057 const Elf_Ehdr
*ehdr
;
1058 const Elf_Sym
*syms
;
1066 sechdrs
= (void *)ehdr
+ ehdr
->e_shoff
;
1068 for (i
= 0; i
< ehdr
->e_shnum
; i
++) {
1069 if (sechdrs
[i
].sh_type
!= SHT_SYMTAB
)
1072 if (sechdrs
[i
].sh_link
>= ehdr
->e_shnum
)
1073 /* Invalid strtab section number */
1075 strtab
= (void *)ehdr
+ sechdrs
[sechdrs
[i
].sh_link
].sh_offset
;
1076 syms
= (void *)ehdr
+ sechdrs
[i
].sh_offset
;
1078 /* Go through symbols for a match */
1079 for (k
= 0; k
< sechdrs
[i
].sh_size
/sizeof(Elf_Sym
); k
++) {
1080 if (ELF_ST_BIND(syms
[k
].st_info
) != STB_GLOBAL
)
1083 if (strcmp(strtab
+ syms
[k
].st_name
, name
) != 0)
1086 if (syms
[k
].st_shndx
== SHN_UNDEF
||
1087 syms
[k
].st_shndx
>= ehdr
->e_shnum
) {
1088 pr_debug("Symbol: %s has bad section index %d.\n",
1089 name
, syms
[k
].st_shndx
);
1093 /* Found the symbol we are looking for */
1101 void *kexec_purgatory_get_symbol_addr(struct kimage
*image
, const char *name
)
1103 struct purgatory_info
*pi
= &image
->purgatory_info
;
1107 sym
= kexec_purgatory_find_symbol(pi
, name
);
1109 return ERR_PTR(-EINVAL
);
1111 sechdr
= &pi
->sechdrs
[sym
->st_shndx
];
1114 * Returns the address where symbol will finally be loaded after
1115 * kexec_load_segment()
1117 return (void *)(sechdr
->sh_addr
+ sym
->st_value
);
1121 * Get or set value of a symbol. If "get_value" is true, symbol value is
1122 * returned in buf otherwise symbol value is set based on value in buf.
1124 int kexec_purgatory_get_set_symbol(struct kimage
*image
, const char *name
,
1125 void *buf
, unsigned int size
, bool get_value
)
1127 struct purgatory_info
*pi
= &image
->purgatory_info
;
1132 sym
= kexec_purgatory_find_symbol(pi
, name
);
1136 if (sym
->st_size
!= size
) {
1137 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1138 name
, (unsigned long)sym
->st_size
, size
);
1142 sec
= pi
->sechdrs
+ sym
->st_shndx
;
1144 if (sec
->sh_type
== SHT_NOBITS
) {
1145 pr_err("symbol %s is in a bss section. Cannot %s\n", name
,
1146 get_value
? "get" : "set");
1150 sym_buf
= (char *)pi
->purgatory_buf
+ sec
->sh_offset
+ sym
->st_value
;
1153 memcpy((void *)buf
, sym_buf
, size
);
1155 memcpy((void *)sym_buf
, buf
, size
);
1159 #endif /* CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY */