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
;
127 kexec_file_dbg_print
= false;
130 #ifdef CONFIG_KEXEC_SIG
131 #ifdef CONFIG_SIGNED_PE_FILE_VERIFICATION
132 int kexec_kernel_verify_pe_sig(const char *kernel
, unsigned long kernel_len
)
136 ret
= verify_pefile_signature(kernel
, kernel_len
,
137 VERIFY_USE_SECONDARY_KEYRING
,
138 VERIFYING_KEXEC_PE_SIGNATURE
);
139 if (ret
== -ENOKEY
&& IS_ENABLED(CONFIG_INTEGRITY_PLATFORM_KEYRING
)) {
140 ret
= verify_pefile_signature(kernel
, kernel_len
,
141 VERIFY_USE_PLATFORM_KEYRING
,
142 VERIFYING_KEXEC_PE_SIGNATURE
);
148 static int kexec_image_verify_sig(struct kimage
*image
, void *buf
,
149 unsigned long buf_len
)
151 if (!image
->fops
|| !image
->fops
->verify_sig
) {
152 pr_debug("kernel loader does not support signature verification.\n");
153 return -EKEYREJECTED
;
156 return image
->fops
->verify_sig(buf
, buf_len
);
160 kimage_validate_signature(struct kimage
*image
)
164 ret
= kexec_image_verify_sig(image
, image
->kernel_buf
,
165 image
->kernel_buf_len
);
169 pr_notice("Enforced kernel signature verification failed (%d).\n", ret
);
174 * If IMA is guaranteed to appraise a signature on the kexec
175 * image, permit it even if the kernel is otherwise locked
178 if (!ima_appraise_signature(READING_KEXEC_IMAGE
) &&
179 security_locked_down(LOCKDOWN_KEXEC
))
182 pr_debug("kernel signature verification failed (%d).\n", ret
);
190 * In file mode list of segments is prepared by kernel. Copy relevant
191 * data from user space, do error checking, prepare segment list
194 kimage_file_prepare_segments(struct kimage
*image
, int kernel_fd
, int initrd_fd
,
195 const char __user
*cmdline_ptr
,
196 unsigned long cmdline_len
, unsigned flags
)
201 ret
= kernel_read_file_from_fd(kernel_fd
, 0, &image
->kernel_buf
,
202 KEXEC_FILE_SIZE_MAX
, NULL
,
203 READING_KEXEC_IMAGE
);
206 image
->kernel_buf_len
= ret
;
207 kexec_dprintk("kernel: %p kernel_size: %#lx\n",
208 image
->kernel_buf
, image
->kernel_buf_len
);
210 /* Call arch image probe handlers */
211 ret
= arch_kexec_kernel_image_probe(image
, image
->kernel_buf
,
212 image
->kernel_buf_len
);
216 #ifdef CONFIG_KEXEC_SIG
217 ret
= kimage_validate_signature(image
);
222 /* It is possible that there no initramfs is being loaded */
223 if (!(flags
& KEXEC_FILE_NO_INITRAMFS
)) {
224 ret
= kernel_read_file_from_fd(initrd_fd
, 0, &image
->initrd_buf
,
225 KEXEC_FILE_SIZE_MAX
, NULL
,
226 READING_KEXEC_INITRAMFS
);
229 image
->initrd_buf_len
= ret
;
234 image
->cmdline_buf
= memdup_user(cmdline_ptr
, cmdline_len
);
235 if (IS_ERR(image
->cmdline_buf
)) {
236 ret
= PTR_ERR(image
->cmdline_buf
);
237 image
->cmdline_buf
= NULL
;
241 image
->cmdline_buf_len
= cmdline_len
;
243 /* command line should be a string with last byte null */
244 if (image
->cmdline_buf
[cmdline_len
- 1] != '\0') {
249 ima_kexec_cmdline(kernel_fd
, image
->cmdline_buf
,
250 image
->cmdline_buf_len
- 1);
253 /* IMA needs to pass the measurement list to the next kernel. */
254 ima_add_kexec_buffer(image
);
256 /* Call image load handler */
257 ldata
= kexec_image_load_default(image
);
260 ret
= PTR_ERR(ldata
);
264 image
->image_loader_data
= ldata
;
266 /* In case of error, free up all allocated memory in this function */
268 kimage_file_post_load_cleanup(image
);
273 kimage_file_alloc_init(struct kimage
**rimage
, int kernel_fd
,
274 int initrd_fd
, const char __user
*cmdline_ptr
,
275 unsigned long cmdline_len
, unsigned long flags
)
278 struct kimage
*image
;
279 bool kexec_on_panic
= flags
& KEXEC_FILE_ON_CRASH
;
281 image
= do_kimage_alloc_init();
285 kexec_file_dbg_print
= !!(flags
& KEXEC_FILE_DEBUG
);
286 image
->file_mode
= 1;
288 #ifdef CONFIG_CRASH_DUMP
289 if (kexec_on_panic
) {
290 /* Enable special crash kernel control page alloc policy. */
291 image
->control_page
= crashk_res
.start
;
292 image
->type
= KEXEC_TYPE_CRASH
;
296 ret
= kimage_file_prepare_segments(image
, kernel_fd
, initrd_fd
,
297 cmdline_ptr
, cmdline_len
, flags
);
301 ret
= sanity_check_segment_list(image
);
303 goto out_free_post_load_bufs
;
306 image
->control_code_page
= kimage_alloc_control_pages(image
,
307 get_order(KEXEC_CONTROL_PAGE_SIZE
));
308 if (!image
->control_code_page
) {
309 pr_err("Could not allocate control_code_buffer\n");
310 goto out_free_post_load_bufs
;
313 if (!kexec_on_panic
) {
314 image
->swap_page
= kimage_alloc_control_pages(image
, 0);
315 if (!image
->swap_page
) {
316 pr_err("Could not allocate swap buffer\n");
317 goto out_free_control_pages
;
323 out_free_control_pages
:
324 kimage_free_page_list(&image
->control_pages
);
325 out_free_post_load_bufs
:
326 kimage_file_post_load_cleanup(image
);
332 SYSCALL_DEFINE5(kexec_file_load
, int, kernel_fd
, int, initrd_fd
,
333 unsigned long, cmdline_len
, const char __user
*, cmdline_ptr
,
334 unsigned long, flags
)
336 int image_type
= (flags
& KEXEC_FILE_ON_CRASH
) ?
337 KEXEC_TYPE_CRASH
: KEXEC_TYPE_DEFAULT
;
338 struct kimage
**dest_image
, *image
;
341 /* We only trust the superuser with rebooting the system. */
342 if (!kexec_load_permitted(image_type
))
345 /* Make sure we have a legal set of flags */
346 if (flags
!= (flags
& KEXEC_FILE_FLAGS
))
351 if (!kexec_trylock())
354 #ifdef CONFIG_CRASH_DUMP
355 if (image_type
== KEXEC_TYPE_CRASH
) {
356 dest_image
= &kexec_crash_image
;
357 if (kexec_crash_image
)
358 arch_kexec_unprotect_crashkres();
361 dest_image
= &kexec_image
;
363 if (flags
& KEXEC_FILE_UNLOAD
)
367 * In case of crash, new kernel gets loaded in reserved region. It is
368 * same memory where old crash kernel might be loaded. Free any
369 * current crash dump kernel before we corrupt it.
371 if (flags
& KEXEC_FILE_ON_CRASH
)
372 kimage_free(xchg(&kexec_crash_image
, NULL
));
374 ret
= kimage_file_alloc_init(&image
, kernel_fd
, initrd_fd
, cmdline_ptr
,
379 #ifdef CONFIG_CRASH_HOTPLUG
380 if ((flags
& KEXEC_FILE_ON_CRASH
) && arch_crash_hotplug_support(image
, flags
))
381 image
->hotplug_support
= 1;
384 ret
= machine_kexec_prepare(image
);
389 * Some architecture(like S390) may touch the crash memory before
390 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
392 ret
= kimage_crash_copy_vmcoreinfo(image
);
396 ret
= kexec_calculate_store_digests(image
);
400 kexec_dprintk("nr_segments = %lu\n", image
->nr_segments
);
401 for (i
= 0; i
< image
->nr_segments
; i
++) {
402 struct kexec_segment
*ksegment
;
404 ksegment
= &image
->segment
[i
];
405 kexec_dprintk("segment[%d]: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
406 i
, ksegment
->buf
, ksegment
->bufsz
, ksegment
->mem
,
409 ret
= kimage_load_segment(image
, &image
->segment
[i
]);
414 kimage_terminate(image
);
416 ret
= machine_kexec_post_load(image
);
420 kexec_dprintk("kexec_file_load: type:%u, start:0x%lx head:0x%lx flags:0x%lx\n",
421 image
->type
, image
->start
, image
->head
, flags
);
423 * Free up any temporary buffers allocated which are not needed
424 * after image has been loaded
426 kimage_file_post_load_cleanup(image
);
428 image
= xchg(dest_image
, image
);
430 #ifdef CONFIG_CRASH_DUMP
431 if ((flags
& KEXEC_FILE_ON_CRASH
) && kexec_crash_image
)
432 arch_kexec_protect_crashkres();
440 static int locate_mem_hole_top_down(unsigned long start
, unsigned long end
,
441 struct kexec_buf
*kbuf
)
443 struct kimage
*image
= kbuf
->image
;
444 unsigned long temp_start
, temp_end
;
446 temp_end
= min(end
, kbuf
->buf_max
);
447 temp_start
= temp_end
- kbuf
->memsz
+ 1;
450 /* align down start */
451 temp_start
= ALIGN_DOWN(temp_start
, kbuf
->buf_align
);
453 if (temp_start
< start
|| temp_start
< kbuf
->buf_min
)
456 temp_end
= temp_start
+ kbuf
->memsz
- 1;
459 * Make sure this does not conflict with any of existing
462 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
463 temp_start
= temp_start
- PAGE_SIZE
;
467 /* We found a suitable memory range */
471 /* If we are here, we found a suitable memory range */
472 kbuf
->mem
= temp_start
;
474 /* Success, stop navigating through remaining System RAM ranges */
478 static int locate_mem_hole_bottom_up(unsigned long start
, unsigned long end
,
479 struct kexec_buf
*kbuf
)
481 struct kimage
*image
= kbuf
->image
;
482 unsigned long temp_start
, temp_end
;
484 temp_start
= max(start
, kbuf
->buf_min
);
487 temp_start
= ALIGN(temp_start
, kbuf
->buf_align
);
488 temp_end
= temp_start
+ kbuf
->memsz
- 1;
490 if (temp_end
> end
|| temp_end
> kbuf
->buf_max
)
493 * Make sure this does not conflict with any of existing
496 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
497 temp_start
= temp_start
+ PAGE_SIZE
;
501 /* We found a suitable memory range */
505 /* If we are here, we found a suitable memory range */
506 kbuf
->mem
= temp_start
;
508 /* Success, stop navigating through remaining System RAM ranges */
512 static int locate_mem_hole_callback(struct resource
*res
, void *arg
)
514 struct kexec_buf
*kbuf
= (struct kexec_buf
*)arg
;
515 u64 start
= res
->start
, end
= res
->end
;
516 unsigned long sz
= end
- start
+ 1;
518 /* Returning 0 will take to next memory range */
520 /* Don't use memory that will be detected and handled by a driver. */
521 if (res
->flags
& IORESOURCE_SYSRAM_DRIVER_MANAGED
)
524 if (sz
< kbuf
->memsz
)
527 if (end
< kbuf
->buf_min
|| start
> kbuf
->buf_max
)
531 * Allocate memory top down with-in ram range. Otherwise bottom up
535 return locate_mem_hole_top_down(start
, end
, kbuf
);
536 return locate_mem_hole_bottom_up(start
, end
, kbuf
);
539 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
540 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
541 int (*func
)(struct resource
*, void *))
545 phys_addr_t mstart
, mend
;
546 struct resource res
= { };
548 #ifdef CONFIG_CRASH_DUMP
549 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
550 return func(&crashk_res
, kbuf
);
554 * Using MEMBLOCK_NONE will properly skip MEMBLOCK_DRIVER_MANAGED. See
555 * IORESOURCE_SYSRAM_DRIVER_MANAGED handling in
556 * locate_mem_hole_callback().
558 if (kbuf
->top_down
) {
559 for_each_free_mem_range_reverse(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
560 &mstart
, &mend
, NULL
) {
562 * In memblock, end points to the first byte after the
563 * range while in kexec, end points to the last byte
568 ret
= func(&res
, kbuf
);
573 for_each_free_mem_range(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
574 &mstart
, &mend
, NULL
) {
576 * In memblock, end points to the first byte after the
577 * range while in kexec, end points to the last byte
582 ret
= func(&res
, kbuf
);
591 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
592 int (*func
)(struct resource
*, void *))
599 * kexec_walk_resources - call func(data) on free memory regions
600 * @kbuf: Context info for the search. Also passed to @func.
601 * @func: Function to call for each memory region.
603 * Return: The memory walk will stop when func returns a non-zero value
604 * and that value will be returned. If all free regions are visited without
605 * func returning non-zero, then zero will be returned.
607 static int kexec_walk_resources(struct kexec_buf
*kbuf
,
608 int (*func
)(struct resource
*, void *))
610 #ifdef CONFIG_CRASH_DUMP
611 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
612 return walk_iomem_res_desc(crashk_res
.desc
,
613 IORESOURCE_SYSTEM_RAM
| IORESOURCE_BUSY
,
614 crashk_res
.start
, crashk_res
.end
,
618 return walk_system_ram_res_rev(0, ULONG_MAX
, kbuf
, func
);
620 return walk_system_ram_res(0, ULONG_MAX
, kbuf
, func
);
624 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
625 * @kbuf: Parameters for the memory search.
627 * On success, kbuf->mem will have the start address of the memory region found.
629 * Return: 0 on success, negative errno on error.
631 int kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
635 /* Arch knows where to place */
636 if (kbuf
->mem
!= KEXEC_BUF_MEM_UNKNOWN
)
639 if (!IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK
))
640 ret
= kexec_walk_resources(kbuf
, locate_mem_hole_callback
);
642 ret
= kexec_walk_memblock(kbuf
, locate_mem_hole_callback
);
644 return ret
== 1 ? 0 : -EADDRNOTAVAIL
;
648 * kexec_add_buffer - place a buffer in a kexec segment
649 * @kbuf: Buffer contents and memory parameters.
651 * This function assumes that kexec_lock is held.
652 * On successful return, @kbuf->mem will have the physical address of
653 * the buffer in memory.
655 * Return: 0 on success, negative errno on error.
657 int kexec_add_buffer(struct kexec_buf
*kbuf
)
659 struct kexec_segment
*ksegment
;
662 /* Currently adding segment this way is allowed only in file mode */
663 if (!kbuf
->image
->file_mode
)
666 if (kbuf
->image
->nr_segments
>= KEXEC_SEGMENT_MAX
)
670 * Make sure we are not trying to add buffer after allocating
671 * control pages. All segments need to be placed first before
672 * any control pages are allocated. As control page allocation
673 * logic goes through list of segments to make sure there are
674 * no destination overlaps.
676 if (!list_empty(&kbuf
->image
->control_pages
)) {
681 /* Ensure minimum alignment needed for segments. */
682 kbuf
->memsz
= ALIGN(kbuf
->memsz
, PAGE_SIZE
);
683 kbuf
->buf_align
= max(kbuf
->buf_align
, PAGE_SIZE
);
685 /* Walk the RAM ranges and allocate a suitable range for the buffer */
686 ret
= arch_kexec_locate_mem_hole(kbuf
);
690 /* Found a suitable memory range */
691 ksegment
= &kbuf
->image
->segment
[kbuf
->image
->nr_segments
];
692 ksegment
->kbuf
= kbuf
->buffer
;
693 ksegment
->bufsz
= kbuf
->bufsz
;
694 ksegment
->mem
= kbuf
->mem
;
695 ksegment
->memsz
= kbuf
->memsz
;
696 kbuf
->image
->nr_segments
++;
700 /* Calculate and store the digest of segments */
701 static int kexec_calculate_store_digests(struct kimage
*image
)
703 struct crypto_shash
*tfm
;
704 struct shash_desc
*desc
;
705 int ret
= 0, i
, j
, zero_buf_sz
, sha_region_sz
;
706 size_t desc_size
, nullsz
;
709 struct kexec_sha_region
*sha_regions
;
710 struct purgatory_info
*pi
= &image
->purgatory_info
;
712 if (!IS_ENABLED(CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY
))
715 zero_buf
= __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT
);
716 zero_buf_sz
= PAGE_SIZE
;
718 tfm
= crypto_alloc_shash("sha256", 0, 0);
724 desc_size
= crypto_shash_descsize(tfm
) + sizeof(*desc
);
725 desc
= kzalloc(desc_size
, GFP_KERNEL
);
731 sha_region_sz
= KEXEC_SEGMENT_MAX
* sizeof(struct kexec_sha_region
);
732 sha_regions
= vzalloc(sha_region_sz
);
740 ret
= crypto_shash_init(desc
);
742 goto out_free_sha_regions
;
744 digest
= kzalloc(SHA256_DIGEST_SIZE
, GFP_KERNEL
);
747 goto out_free_sha_regions
;
750 for (j
= i
= 0; i
< image
->nr_segments
; i
++) {
751 struct kexec_segment
*ksegment
;
753 #ifdef CONFIG_CRASH_HOTPLUG
754 /* Exclude elfcorehdr segment to allow future changes via hotplug */
755 if (i
== image
->elfcorehdr_index
)
759 ksegment
= &image
->segment
[i
];
761 * Skip purgatory as it will be modified once we put digest
764 if (ksegment
->kbuf
== pi
->purgatory_buf
)
767 ret
= crypto_shash_update(desc
, ksegment
->kbuf
,
773 * Assume rest of the buffer is filled with zero and
774 * update digest accordingly.
776 nullsz
= ksegment
->memsz
- ksegment
->bufsz
;
778 unsigned long bytes
= nullsz
;
780 if (bytes
> zero_buf_sz
)
782 ret
= crypto_shash_update(desc
, zero_buf
, bytes
);
791 sha_regions
[j
].start
= ksegment
->mem
;
792 sha_regions
[j
].len
= ksegment
->memsz
;
797 ret
= crypto_shash_final(desc
, digest
);
799 goto out_free_digest
;
800 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha_regions",
801 sha_regions
, sha_region_sz
, 0);
803 goto out_free_digest
;
805 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha256_digest",
806 digest
, SHA256_DIGEST_SIZE
, 0);
808 goto out_free_digest
;
813 out_free_sha_regions
:
823 #ifdef CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY
825 * kexec_purgatory_setup_kbuf - prepare buffer to load purgatory.
826 * @pi: Purgatory to be loaded.
827 * @kbuf: Buffer to setup.
829 * Allocates the memory needed for the buffer. Caller is responsible to free
830 * the memory after use.
832 * Return: 0 on success, negative errno on error.
834 static int kexec_purgatory_setup_kbuf(struct purgatory_info
*pi
,
835 struct kexec_buf
*kbuf
)
837 const Elf_Shdr
*sechdrs
;
838 unsigned long bss_align
;
839 unsigned long bss_sz
;
843 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
844 kbuf
->buf_align
= bss_align
= 1;
845 kbuf
->bufsz
= bss_sz
= 0;
847 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
848 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
851 align
= sechdrs
[i
].sh_addralign
;
852 if (sechdrs
[i
].sh_type
!= SHT_NOBITS
) {
853 if (kbuf
->buf_align
< align
)
854 kbuf
->buf_align
= align
;
855 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, align
);
856 kbuf
->bufsz
+= sechdrs
[i
].sh_size
;
858 if (bss_align
< align
)
860 bss_sz
= ALIGN(bss_sz
, align
);
861 bss_sz
+= sechdrs
[i
].sh_size
;
864 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, bss_align
);
865 kbuf
->memsz
= kbuf
->bufsz
+ bss_sz
;
866 if (kbuf
->buf_align
< bss_align
)
867 kbuf
->buf_align
= bss_align
;
869 kbuf
->buffer
= vzalloc(kbuf
->bufsz
);
872 pi
->purgatory_buf
= kbuf
->buffer
;
874 ret
= kexec_add_buffer(kbuf
);
880 vfree(pi
->purgatory_buf
);
881 pi
->purgatory_buf
= NULL
;
886 * kexec_purgatory_setup_sechdrs - prepares the pi->sechdrs buffer.
887 * @pi: Purgatory to be loaded.
888 * @kbuf: Buffer prepared to store purgatory.
890 * Allocates the memory needed for the buffer. Caller is responsible to free
891 * the memory after use.
893 * Return: 0 on success, negative errno on error.
895 static int kexec_purgatory_setup_sechdrs(struct purgatory_info
*pi
,
896 struct kexec_buf
*kbuf
)
898 unsigned long bss_addr
;
899 unsigned long offset
;
905 * The section headers in kexec_purgatory are read-only. In order to
906 * have them modifiable make a temporary copy.
908 sechdrs_size
= array_size(sizeof(Elf_Shdr
), pi
->ehdr
->e_shnum
);
909 sechdrs
= vzalloc(sechdrs_size
);
912 memcpy(sechdrs
, (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
, sechdrs_size
);
913 pi
->sechdrs
= sechdrs
;
916 bss_addr
= kbuf
->mem
+ kbuf
->bufsz
;
917 kbuf
->image
->start
= pi
->ehdr
->e_entry
;
919 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
923 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
926 align
= sechdrs
[i
].sh_addralign
;
927 if (sechdrs
[i
].sh_type
== SHT_NOBITS
) {
928 bss_addr
= ALIGN(bss_addr
, align
);
929 sechdrs
[i
].sh_addr
= bss_addr
;
930 bss_addr
+= sechdrs
[i
].sh_size
;
934 offset
= ALIGN(offset
, align
);
937 * Check if the segment contains the entry point, if so,
938 * calculate the value of image->start based on it.
939 * If the compiler has produced more than one .text section
940 * (Eg: .text.hot), they are generally after the main .text
941 * section, and they shall not be used to calculate
942 * image->start. So do not re-calculate image->start if it
943 * is not set to the initial value, and warn the user so they
944 * have a chance to fix their purgatory's linker script.
946 if (sechdrs
[i
].sh_flags
& SHF_EXECINSTR
&&
947 pi
->ehdr
->e_entry
>= sechdrs
[i
].sh_addr
&&
948 pi
->ehdr
->e_entry
< (sechdrs
[i
].sh_addr
949 + sechdrs
[i
].sh_size
) &&
950 !WARN_ON(kbuf
->image
->start
!= pi
->ehdr
->e_entry
)) {
951 kbuf
->image
->start
-= sechdrs
[i
].sh_addr
;
952 kbuf
->image
->start
+= kbuf
->mem
+ offset
;
955 src
= (void *)pi
->ehdr
+ sechdrs
[i
].sh_offset
;
956 dst
= pi
->purgatory_buf
+ offset
;
957 memcpy(dst
, src
, sechdrs
[i
].sh_size
);
959 sechdrs
[i
].sh_addr
= kbuf
->mem
+ offset
;
960 sechdrs
[i
].sh_offset
= offset
;
961 offset
+= sechdrs
[i
].sh_size
;
967 static int kexec_apply_relocations(struct kimage
*image
)
970 struct purgatory_info
*pi
= &image
->purgatory_info
;
971 const Elf_Shdr
*sechdrs
;
973 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
975 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
976 const Elf_Shdr
*relsec
;
977 const Elf_Shdr
*symtab
;
980 relsec
= sechdrs
+ i
;
982 if (relsec
->sh_type
!= SHT_RELA
&&
983 relsec
->sh_type
!= SHT_REL
)
987 * For section of type SHT_RELA/SHT_REL,
988 * ->sh_link contains section header index of associated
989 * symbol table. And ->sh_info contains section header
990 * index of section to which relocations apply.
992 if (relsec
->sh_info
>= pi
->ehdr
->e_shnum
||
993 relsec
->sh_link
>= pi
->ehdr
->e_shnum
)
996 section
= pi
->sechdrs
+ relsec
->sh_info
;
997 symtab
= sechdrs
+ relsec
->sh_link
;
999 if (!(section
->sh_flags
& SHF_ALLOC
))
1003 * symtab->sh_link contain section header index of associated
1006 if (symtab
->sh_link
>= pi
->ehdr
->e_shnum
)
1007 /* Invalid section number? */
1011 * Respective architecture needs to provide support for applying
1012 * relocations of type SHT_RELA/SHT_REL.
1014 if (relsec
->sh_type
== SHT_RELA
)
1015 ret
= arch_kexec_apply_relocations_add(pi
, section
,
1017 else if (relsec
->sh_type
== SHT_REL
)
1018 ret
= arch_kexec_apply_relocations(pi
, section
,
1028 * kexec_load_purgatory - Load and relocate the purgatory object.
1029 * @image: Image to add the purgatory to.
1030 * @kbuf: Memory parameters to use.
1032 * Allocates the memory needed for image->purgatory_info.sechdrs and
1033 * image->purgatory_info.purgatory_buf/kbuf->buffer. Caller is responsible
1034 * to free the memory after use.
1036 * Return: 0 on success, negative errno on error.
1038 int kexec_load_purgatory(struct kimage
*image
, struct kexec_buf
*kbuf
)
1040 struct purgatory_info
*pi
= &image
->purgatory_info
;
1043 if (kexec_purgatory_size
<= 0)
1046 pi
->ehdr
= (const Elf_Ehdr
*)kexec_purgatory
;
1048 ret
= kexec_purgatory_setup_kbuf(pi
, kbuf
);
1052 ret
= kexec_purgatory_setup_sechdrs(pi
, kbuf
);
1056 ret
= kexec_apply_relocations(image
);
1065 vfree(pi
->purgatory_buf
);
1066 pi
->purgatory_buf
= NULL
;
1071 * kexec_purgatory_find_symbol - find a symbol in the purgatory
1072 * @pi: Purgatory to search in.
1073 * @name: Name of the symbol.
1075 * Return: pointer to symbol in read-only symtab on success, NULL on error.
1077 static const Elf_Sym
*kexec_purgatory_find_symbol(struct purgatory_info
*pi
,
1080 const Elf_Shdr
*sechdrs
;
1081 const Elf_Ehdr
*ehdr
;
1082 const Elf_Sym
*syms
;
1090 sechdrs
= (void *)ehdr
+ ehdr
->e_shoff
;
1092 for (i
= 0; i
< ehdr
->e_shnum
; i
++) {
1093 if (sechdrs
[i
].sh_type
!= SHT_SYMTAB
)
1096 if (sechdrs
[i
].sh_link
>= ehdr
->e_shnum
)
1097 /* Invalid strtab section number */
1099 strtab
= (void *)ehdr
+ sechdrs
[sechdrs
[i
].sh_link
].sh_offset
;
1100 syms
= (void *)ehdr
+ sechdrs
[i
].sh_offset
;
1102 /* Go through symbols for a match */
1103 for (k
= 0; k
< sechdrs
[i
].sh_size
/sizeof(Elf_Sym
); k
++) {
1104 if (ELF_ST_BIND(syms
[k
].st_info
) != STB_GLOBAL
)
1107 if (strcmp(strtab
+ syms
[k
].st_name
, name
) != 0)
1110 if (syms
[k
].st_shndx
== SHN_UNDEF
||
1111 syms
[k
].st_shndx
>= ehdr
->e_shnum
) {
1112 pr_debug("Symbol: %s has bad section index %d.\n",
1113 name
, syms
[k
].st_shndx
);
1117 /* Found the symbol we are looking for */
1125 void *kexec_purgatory_get_symbol_addr(struct kimage
*image
, const char *name
)
1127 struct purgatory_info
*pi
= &image
->purgatory_info
;
1131 sym
= kexec_purgatory_find_symbol(pi
, name
);
1133 return ERR_PTR(-EINVAL
);
1135 sechdr
= &pi
->sechdrs
[sym
->st_shndx
];
1138 * Returns the address where symbol will finally be loaded after
1139 * kexec_load_segment()
1141 return (void *)(sechdr
->sh_addr
+ sym
->st_value
);
1145 * Get or set value of a symbol. If "get_value" is true, symbol value is
1146 * returned in buf otherwise symbol value is set based on value in buf.
1148 int kexec_purgatory_get_set_symbol(struct kimage
*image
, const char *name
,
1149 void *buf
, unsigned int size
, bool get_value
)
1151 struct purgatory_info
*pi
= &image
->purgatory_info
;
1156 sym
= kexec_purgatory_find_symbol(pi
, name
);
1160 if (sym
->st_size
!= size
) {
1161 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1162 name
, (unsigned long)sym
->st_size
, size
);
1166 sec
= pi
->sechdrs
+ sym
->st_shndx
;
1168 if (sec
->sh_type
== SHT_NOBITS
) {
1169 pr_err("symbol %s is in a bss section. Cannot %s\n", name
,
1170 get_value
? "get" : "set");
1174 sym_buf
= (char *)pi
->purgatory_buf
+ sec
->sh_offset
+ sym
->st_value
;
1177 memcpy((void *)buf
, sym_buf
, size
);
1179 memcpy((void *)sym_buf
, buf
, size
);
1183 #endif /* CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY */