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 static int kexec_calculate_store_digests(struct kimage
*image
);
35 * Currently this is the only default function that is exported as some
36 * architectures need it to do additional handlings.
37 * In the future, other default functions may be exported too if required.
39 int kexec_image_probe_default(struct kimage
*image
, void *buf
,
40 unsigned long buf_len
)
42 const struct kexec_file_ops
* const *fops
;
45 for (fops
= &kexec_file_loaders
[0]; *fops
&& (*fops
)->probe
; ++fops
) {
46 ret
= (*fops
)->probe(buf
, buf_len
);
56 /* Architectures can provide this probe function */
57 int __weak
arch_kexec_kernel_image_probe(struct kimage
*image
, void *buf
,
58 unsigned long buf_len
)
60 return kexec_image_probe_default(image
, buf
, buf_len
);
63 static void *kexec_image_load_default(struct kimage
*image
)
65 if (!image
->fops
|| !image
->fops
->load
)
66 return ERR_PTR(-ENOEXEC
);
68 return image
->fops
->load(image
, image
->kernel_buf
,
69 image
->kernel_buf_len
, image
->initrd_buf
,
70 image
->initrd_buf_len
, image
->cmdline_buf
,
71 image
->cmdline_buf_len
);
74 void * __weak
arch_kexec_kernel_image_load(struct kimage
*image
)
76 return kexec_image_load_default(image
);
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
);
87 int __weak
arch_kimage_file_post_load_cleanup(struct kimage
*image
)
89 return kexec_image_post_load_cleanup_default(image
);
92 #ifdef CONFIG_KEXEC_SIG
93 static int kexec_image_verify_sig_default(struct kimage
*image
, void *buf
,
94 unsigned long buf_len
)
96 if (!image
->fops
|| !image
->fops
->verify_sig
) {
97 pr_debug("kernel loader does not support signature verification.\n");
101 return image
->fops
->verify_sig(buf
, buf_len
);
104 int __weak
arch_kexec_kernel_verify_sig(struct kimage
*image
, void *buf
,
105 unsigned long buf_len
)
107 return kexec_image_verify_sig_default(image
, buf
, buf_len
);
112 * arch_kexec_apply_relocations_add - apply relocations of type RELA
113 * @pi: Purgatory to be relocated.
114 * @section: Section relocations applying to.
115 * @relsec: Section containing RELAs.
116 * @symtab: Corresponding symtab.
118 * Return: 0 on success, negative errno on error.
121 arch_kexec_apply_relocations_add(struct purgatory_info
*pi
, Elf_Shdr
*section
,
122 const Elf_Shdr
*relsec
, const Elf_Shdr
*symtab
)
124 pr_err("RELA relocation unsupported.\n");
129 * arch_kexec_apply_relocations - apply relocations of type REL
130 * @pi: Purgatory to be relocated.
131 * @section: Section relocations applying to.
132 * @relsec: Section containing RELs.
133 * @symtab: Corresponding symtab.
135 * Return: 0 on success, negative errno on error.
138 arch_kexec_apply_relocations(struct purgatory_info
*pi
, Elf_Shdr
*section
,
139 const Elf_Shdr
*relsec
, const Elf_Shdr
*symtab
)
141 pr_err("REL relocation unsupported.\n");
146 * Free up memory used by kernel, initrd, and command line. This is temporary
147 * memory allocation which is not needed any more after these buffers have
148 * been loaded into separate segments and have been copied elsewhere.
150 void kimage_file_post_load_cleanup(struct kimage
*image
)
152 struct purgatory_info
*pi
= &image
->purgatory_info
;
154 vfree(image
->kernel_buf
);
155 image
->kernel_buf
= NULL
;
157 vfree(image
->initrd_buf
);
158 image
->initrd_buf
= NULL
;
160 kfree(image
->cmdline_buf
);
161 image
->cmdline_buf
= NULL
;
163 vfree(pi
->purgatory_buf
);
164 pi
->purgatory_buf
= NULL
;
169 #ifdef CONFIG_IMA_KEXEC
170 vfree(image
->ima_buffer
);
171 image
->ima_buffer
= NULL
;
172 #endif /* CONFIG_IMA_KEXEC */
174 /* See if architecture has anything to cleanup post load */
175 arch_kimage_file_post_load_cleanup(image
);
178 * Above call should have called into bootloader to free up
179 * any data stored in kimage->image_loader_data. It should
180 * be ok now to free it up.
182 kfree(image
->image_loader_data
);
183 image
->image_loader_data
= NULL
;
186 #ifdef CONFIG_KEXEC_SIG
188 kimage_validate_signature(struct kimage
*image
)
192 ret
= arch_kexec_kernel_verify_sig(image
, image
->kernel_buf
,
193 image
->kernel_buf_len
);
196 if (IS_ENABLED(CONFIG_KEXEC_SIG_FORCE
)) {
197 pr_notice("Enforced kernel signature verification failed (%d).\n", ret
);
202 * If IMA is guaranteed to appraise a signature on the kexec
203 * image, permit it even if the kernel is otherwise locked
206 if (!ima_appraise_signature(READING_KEXEC_IMAGE
) &&
207 security_locked_down(LOCKDOWN_KEXEC
))
210 pr_debug("kernel signature verification failed (%d).\n", ret
);
218 * In file mode list of segments is prepared by kernel. Copy relevant
219 * data from user space, do error checking, prepare segment list
222 kimage_file_prepare_segments(struct kimage
*image
, int kernel_fd
, int initrd_fd
,
223 const char __user
*cmdline_ptr
,
224 unsigned long cmdline_len
, unsigned flags
)
229 ret
= kernel_read_file_from_fd(kernel_fd
, 0, &image
->kernel_buf
,
230 INT_MAX
, NULL
, READING_KEXEC_IMAGE
);
233 image
->kernel_buf_len
= ret
;
235 /* Call arch image probe handlers */
236 ret
= arch_kexec_kernel_image_probe(image
, image
->kernel_buf
,
237 image
->kernel_buf_len
);
241 #ifdef CONFIG_KEXEC_SIG
242 ret
= kimage_validate_signature(image
);
247 /* It is possible that there no initramfs is being loaded */
248 if (!(flags
& KEXEC_FILE_NO_INITRAMFS
)) {
249 ret
= kernel_read_file_from_fd(initrd_fd
, 0, &image
->initrd_buf
,
251 READING_KEXEC_INITRAMFS
);
254 image
->initrd_buf_len
= ret
;
259 image
->cmdline_buf
= memdup_user(cmdline_ptr
, cmdline_len
);
260 if (IS_ERR(image
->cmdline_buf
)) {
261 ret
= PTR_ERR(image
->cmdline_buf
);
262 image
->cmdline_buf
= NULL
;
266 image
->cmdline_buf_len
= cmdline_len
;
268 /* command line should be a string with last byte null */
269 if (image
->cmdline_buf
[cmdline_len
- 1] != '\0') {
274 ima_kexec_cmdline(kernel_fd
, image
->cmdline_buf
,
275 image
->cmdline_buf_len
- 1);
278 /* IMA needs to pass the measurement list to the next kernel. */
279 ima_add_kexec_buffer(image
);
281 /* Call arch image load handlers */
282 ldata
= arch_kexec_kernel_image_load(image
);
285 ret
= PTR_ERR(ldata
);
289 image
->image_loader_data
= ldata
;
291 /* In case of error, free up all allocated memory in this function */
293 kimage_file_post_load_cleanup(image
);
298 kimage_file_alloc_init(struct kimage
**rimage
, int kernel_fd
,
299 int initrd_fd
, const char __user
*cmdline_ptr
,
300 unsigned long cmdline_len
, unsigned long flags
)
303 struct kimage
*image
;
304 bool kexec_on_panic
= flags
& KEXEC_FILE_ON_CRASH
;
306 image
= do_kimage_alloc_init();
310 image
->file_mode
= 1;
312 if (kexec_on_panic
) {
313 /* Enable special crash kernel control page alloc policy. */
314 image
->control_page
= crashk_res
.start
;
315 image
->type
= KEXEC_TYPE_CRASH
;
318 ret
= kimage_file_prepare_segments(image
, kernel_fd
, initrd_fd
,
319 cmdline_ptr
, cmdline_len
, flags
);
323 ret
= sanity_check_segment_list(image
);
325 goto out_free_post_load_bufs
;
328 image
->control_code_page
= kimage_alloc_control_pages(image
,
329 get_order(KEXEC_CONTROL_PAGE_SIZE
));
330 if (!image
->control_code_page
) {
331 pr_err("Could not allocate control_code_buffer\n");
332 goto out_free_post_load_bufs
;
335 if (!kexec_on_panic
) {
336 image
->swap_page
= kimage_alloc_control_pages(image
, 0);
337 if (!image
->swap_page
) {
338 pr_err("Could not allocate swap buffer\n");
339 goto out_free_control_pages
;
345 out_free_control_pages
:
346 kimage_free_page_list(&image
->control_pages
);
347 out_free_post_load_bufs
:
348 kimage_file_post_load_cleanup(image
);
354 SYSCALL_DEFINE5(kexec_file_load
, int, kernel_fd
, int, initrd_fd
,
355 unsigned long, cmdline_len
, const char __user
*, cmdline_ptr
,
356 unsigned long, flags
)
359 struct kimage
**dest_image
, *image
;
361 /* We only trust the superuser with rebooting the system. */
362 if (!capable(CAP_SYS_BOOT
) || kexec_load_disabled
)
365 /* Make sure we have a legal set of flags */
366 if (flags
!= (flags
& KEXEC_FILE_FLAGS
))
371 if (!mutex_trylock(&kexec_mutex
))
374 dest_image
= &kexec_image
;
375 if (flags
& KEXEC_FILE_ON_CRASH
) {
376 dest_image
= &kexec_crash_image
;
377 if (kexec_crash_image
)
378 arch_kexec_unprotect_crashkres();
381 if (flags
& KEXEC_FILE_UNLOAD
)
385 * In case of crash, new kernel gets loaded in reserved region. It is
386 * same memory where old crash kernel might be loaded. Free any
387 * current crash dump kernel before we corrupt it.
389 if (flags
& KEXEC_FILE_ON_CRASH
)
390 kimage_free(xchg(&kexec_crash_image
, NULL
));
392 ret
= kimage_file_alloc_init(&image
, kernel_fd
, initrd_fd
, cmdline_ptr
,
397 ret
= machine_kexec_prepare(image
);
402 * Some architecture(like S390) may touch the crash memory before
403 * machine_kexec_prepare(), we must copy vmcoreinfo data after it.
405 ret
= kimage_crash_copy_vmcoreinfo(image
);
409 ret
= kexec_calculate_store_digests(image
);
413 for (i
= 0; i
< image
->nr_segments
; i
++) {
414 struct kexec_segment
*ksegment
;
416 ksegment
= &image
->segment
[i
];
417 pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
418 i
, ksegment
->buf
, ksegment
->bufsz
, ksegment
->mem
,
421 ret
= kimage_load_segment(image
, &image
->segment
[i
]);
426 kimage_terminate(image
);
428 ret
= machine_kexec_post_load(image
);
433 * Free up any temporary buffers allocated which are not needed
434 * after image has been loaded
436 kimage_file_post_load_cleanup(image
);
438 image
= xchg(dest_image
, image
);
440 if ((flags
& KEXEC_FILE_ON_CRASH
) && kexec_crash_image
)
441 arch_kexec_protect_crashkres();
443 mutex_unlock(&kexec_mutex
);
448 static int locate_mem_hole_top_down(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_end
= min(end
, kbuf
->buf_max
);
455 temp_start
= temp_end
- kbuf
->memsz
;
458 /* align down start */
459 temp_start
= temp_start
& (~(kbuf
->buf_align
- 1));
461 if (temp_start
< start
|| temp_start
< kbuf
->buf_min
)
464 temp_end
= temp_start
+ kbuf
->memsz
- 1;
467 * Make sure this does not conflict with any of existing
470 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
471 temp_start
= temp_start
- PAGE_SIZE
;
475 /* We found a suitable memory range */
479 /* If we are here, we found a suitable memory range */
480 kbuf
->mem
= temp_start
;
482 /* Success, stop navigating through remaining System RAM ranges */
486 static int locate_mem_hole_bottom_up(unsigned long start
, unsigned long end
,
487 struct kexec_buf
*kbuf
)
489 struct kimage
*image
= kbuf
->image
;
490 unsigned long temp_start
, temp_end
;
492 temp_start
= max(start
, kbuf
->buf_min
);
495 temp_start
= ALIGN(temp_start
, kbuf
->buf_align
);
496 temp_end
= temp_start
+ kbuf
->memsz
- 1;
498 if (temp_end
> end
|| temp_end
> kbuf
->buf_max
)
501 * Make sure this does not conflict with any of existing
504 if (kimage_is_destination_range(image
, temp_start
, temp_end
)) {
505 temp_start
= temp_start
+ PAGE_SIZE
;
509 /* We found a suitable memory range */
513 /* If we are here, we found a suitable memory range */
514 kbuf
->mem
= temp_start
;
516 /* Success, stop navigating through remaining System RAM ranges */
520 static int locate_mem_hole_callback(struct resource
*res
, void *arg
)
522 struct kexec_buf
*kbuf
= (struct kexec_buf
*)arg
;
523 u64 start
= res
->start
, end
= res
->end
;
524 unsigned long sz
= end
- start
+ 1;
526 /* Returning 0 will take to next memory range */
528 /* Don't use memory that will be detected and handled by a driver. */
529 if (res
->flags
& IORESOURCE_SYSRAM_DRIVER_MANAGED
)
532 if (sz
< kbuf
->memsz
)
535 if (end
< kbuf
->buf_min
|| start
> kbuf
->buf_max
)
539 * Allocate memory top down with-in ram range. Otherwise bottom up
543 return locate_mem_hole_top_down(start
, end
, kbuf
);
544 return locate_mem_hole_bottom_up(start
, end
, kbuf
);
547 #ifdef CONFIG_ARCH_KEEP_MEMBLOCK
548 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
549 int (*func
)(struct resource
*, void *))
553 phys_addr_t mstart
, mend
;
554 struct resource res
= { };
556 if (kbuf
->image
->type
== KEXEC_TYPE_CRASH
)
557 return func(&crashk_res
, kbuf
);
559 if (kbuf
->top_down
) {
560 for_each_free_mem_range_reverse(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
561 &mstart
, &mend
, NULL
) {
563 * In memblock, end points to the first byte after the
564 * range while in kexec, end points to the last byte
569 ret
= func(&res
, kbuf
);
574 for_each_free_mem_range(i
, NUMA_NO_NODE
, MEMBLOCK_NONE
,
575 &mstart
, &mend
, NULL
) {
577 * In memblock, end points to the first byte after the
578 * range while in kexec, end points to the last byte
583 ret
= func(&res
, kbuf
);
592 static int kexec_walk_memblock(struct kexec_buf
*kbuf
,
593 int (*func
)(struct resource
*, void *))
600 * kexec_walk_resources - call func(data) on free memory regions
601 * @kbuf: Context info for the search. Also passed to @func.
602 * @func: Function to call for each memory region.
604 * Return: The memory walk will stop when func returns a non-zero value
605 * and that value will be returned. If all free regions are visited without
606 * func returning non-zero, then zero will be returned.
608 static int kexec_walk_resources(struct kexec_buf
*kbuf
,
609 int (*func
)(struct resource
*, void *))
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
,
617 return walk_system_ram_res(0, ULONG_MAX
, kbuf
, func
);
621 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
622 * @kbuf: Parameters for the memory search.
624 * On success, kbuf->mem will have the start address of the memory region found.
626 * Return: 0 on success, negative errno on error.
628 int kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
632 /* Arch knows where to place */
633 if (kbuf
->mem
!= KEXEC_BUF_MEM_UNKNOWN
)
636 if (!IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK
))
637 ret
= kexec_walk_resources(kbuf
, locate_mem_hole_callback
);
639 ret
= kexec_walk_memblock(kbuf
, locate_mem_hole_callback
);
641 return ret
== 1 ? 0 : -EADDRNOTAVAIL
;
645 * arch_kexec_locate_mem_hole - Find free memory to place the segments.
646 * @kbuf: Parameters for the memory search.
648 * On success, kbuf->mem will have the start address of the memory region found.
650 * Return: 0 on success, negative errno on error.
652 int __weak
arch_kexec_locate_mem_hole(struct kexec_buf
*kbuf
)
654 return kexec_locate_mem_hole(kbuf
);
658 * kexec_add_buffer - place a buffer in a kexec segment
659 * @kbuf: Buffer contents and memory parameters.
661 * This function assumes that kexec_mutex is held.
662 * On successful return, @kbuf->mem will have the physical address of
663 * the buffer in memory.
665 * Return: 0 on success, negative errno on error.
667 int kexec_add_buffer(struct kexec_buf
*kbuf
)
669 struct kexec_segment
*ksegment
;
672 /* Currently adding segment this way is allowed only in file mode */
673 if (!kbuf
->image
->file_mode
)
676 if (kbuf
->image
->nr_segments
>= KEXEC_SEGMENT_MAX
)
680 * Make sure we are not trying to add buffer after allocating
681 * control pages. All segments need to be placed first before
682 * any control pages are allocated. As control page allocation
683 * logic goes through list of segments to make sure there are
684 * no destination overlaps.
686 if (!list_empty(&kbuf
->image
->control_pages
)) {
691 /* Ensure minimum alignment needed for segments. */
692 kbuf
->memsz
= ALIGN(kbuf
->memsz
, PAGE_SIZE
);
693 kbuf
->buf_align
= max(kbuf
->buf_align
, PAGE_SIZE
);
695 /* Walk the RAM ranges and allocate a suitable range for the buffer */
696 ret
= arch_kexec_locate_mem_hole(kbuf
);
700 /* Found a suitable memory range */
701 ksegment
= &kbuf
->image
->segment
[kbuf
->image
->nr_segments
];
702 ksegment
->kbuf
= kbuf
->buffer
;
703 ksegment
->bufsz
= kbuf
->bufsz
;
704 ksegment
->mem
= kbuf
->mem
;
705 ksegment
->memsz
= kbuf
->memsz
;
706 kbuf
->image
->nr_segments
++;
710 /* Calculate and store the digest of segments */
711 static int kexec_calculate_store_digests(struct kimage
*image
)
713 struct crypto_shash
*tfm
;
714 struct shash_desc
*desc
;
715 int ret
= 0, i
, j
, zero_buf_sz
, sha_region_sz
;
716 size_t desc_size
, nullsz
;
719 struct kexec_sha_region
*sha_regions
;
720 struct purgatory_info
*pi
= &image
->purgatory_info
;
722 if (!IS_ENABLED(CONFIG_ARCH_HAS_KEXEC_PURGATORY
))
725 zero_buf
= __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT
);
726 zero_buf_sz
= PAGE_SIZE
;
728 tfm
= crypto_alloc_shash("sha256", 0, 0);
734 desc_size
= crypto_shash_descsize(tfm
) + sizeof(*desc
);
735 desc
= kzalloc(desc_size
, GFP_KERNEL
);
741 sha_region_sz
= KEXEC_SEGMENT_MAX
* sizeof(struct kexec_sha_region
);
742 sha_regions
= vzalloc(sha_region_sz
);
750 ret
= crypto_shash_init(desc
);
752 goto out_free_sha_regions
;
754 digest
= kzalloc(SHA256_DIGEST_SIZE
, GFP_KERNEL
);
757 goto out_free_sha_regions
;
760 for (j
= i
= 0; i
< image
->nr_segments
; i
++) {
761 struct kexec_segment
*ksegment
;
763 ksegment
= &image
->segment
[i
];
765 * Skip purgatory as it will be modified once we put digest
768 if (ksegment
->kbuf
== pi
->purgatory_buf
)
771 ret
= crypto_shash_update(desc
, ksegment
->kbuf
,
777 * Assume rest of the buffer is filled with zero and
778 * update digest accordingly.
780 nullsz
= ksegment
->memsz
- ksegment
->bufsz
;
782 unsigned long bytes
= nullsz
;
784 if (bytes
> zero_buf_sz
)
786 ret
= crypto_shash_update(desc
, zero_buf
, bytes
);
795 sha_regions
[j
].start
= ksegment
->mem
;
796 sha_regions
[j
].len
= ksegment
->memsz
;
801 ret
= crypto_shash_final(desc
, digest
);
803 goto out_free_digest
;
804 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha_regions",
805 sha_regions
, sha_region_sz
, 0);
807 goto out_free_digest
;
809 ret
= kexec_purgatory_get_set_symbol(image
, "purgatory_sha256_digest",
810 digest
, SHA256_DIGEST_SIZE
, 0);
812 goto out_free_digest
;
817 out_free_sha_regions
:
827 #ifdef CONFIG_ARCH_HAS_KEXEC_PURGATORY
829 * kexec_purgatory_setup_kbuf - prepare buffer to load purgatory.
830 * @pi: Purgatory to be loaded.
831 * @kbuf: Buffer to setup.
833 * Allocates the memory needed for the buffer. Caller is responsible to free
834 * the memory after use.
836 * Return: 0 on success, negative errno on error.
838 static int kexec_purgatory_setup_kbuf(struct purgatory_info
*pi
,
839 struct kexec_buf
*kbuf
)
841 const Elf_Shdr
*sechdrs
;
842 unsigned long bss_align
;
843 unsigned long bss_sz
;
847 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
848 kbuf
->buf_align
= bss_align
= 1;
849 kbuf
->bufsz
= bss_sz
= 0;
851 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
852 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
855 align
= sechdrs
[i
].sh_addralign
;
856 if (sechdrs
[i
].sh_type
!= SHT_NOBITS
) {
857 if (kbuf
->buf_align
< align
)
858 kbuf
->buf_align
= align
;
859 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, align
);
860 kbuf
->bufsz
+= sechdrs
[i
].sh_size
;
862 if (bss_align
< align
)
864 bss_sz
= ALIGN(bss_sz
, align
);
865 bss_sz
+= sechdrs
[i
].sh_size
;
868 kbuf
->bufsz
= ALIGN(kbuf
->bufsz
, bss_align
);
869 kbuf
->memsz
= kbuf
->bufsz
+ bss_sz
;
870 if (kbuf
->buf_align
< bss_align
)
871 kbuf
->buf_align
= bss_align
;
873 kbuf
->buffer
= vzalloc(kbuf
->bufsz
);
876 pi
->purgatory_buf
= kbuf
->buffer
;
878 ret
= kexec_add_buffer(kbuf
);
884 vfree(pi
->purgatory_buf
);
885 pi
->purgatory_buf
= NULL
;
890 * kexec_purgatory_setup_sechdrs - prepares the pi->sechdrs buffer.
891 * @pi: Purgatory to be loaded.
892 * @kbuf: Buffer prepared to store purgatory.
894 * Allocates the memory needed for the buffer. Caller is responsible to free
895 * the memory after use.
897 * Return: 0 on success, negative errno on error.
899 static int kexec_purgatory_setup_sechdrs(struct purgatory_info
*pi
,
900 struct kexec_buf
*kbuf
)
902 unsigned long bss_addr
;
903 unsigned long offset
;
908 * The section headers in kexec_purgatory are read-only. In order to
909 * have them modifiable make a temporary copy.
911 sechdrs
= vzalloc(array_size(sizeof(Elf_Shdr
), pi
->ehdr
->e_shnum
));
914 memcpy(sechdrs
, (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
,
915 pi
->ehdr
->e_shnum
* sizeof(Elf_Shdr
));
916 pi
->sechdrs
= sechdrs
;
919 bss_addr
= kbuf
->mem
+ kbuf
->bufsz
;
920 kbuf
->image
->start
= pi
->ehdr
->e_entry
;
922 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
926 if (!(sechdrs
[i
].sh_flags
& SHF_ALLOC
))
929 align
= sechdrs
[i
].sh_addralign
;
930 if (sechdrs
[i
].sh_type
== SHT_NOBITS
) {
931 bss_addr
= ALIGN(bss_addr
, align
);
932 sechdrs
[i
].sh_addr
= bss_addr
;
933 bss_addr
+= sechdrs
[i
].sh_size
;
937 offset
= ALIGN(offset
, align
);
938 if (sechdrs
[i
].sh_flags
& SHF_EXECINSTR
&&
939 pi
->ehdr
->e_entry
>= sechdrs
[i
].sh_addr
&&
940 pi
->ehdr
->e_entry
< (sechdrs
[i
].sh_addr
941 + sechdrs
[i
].sh_size
)) {
942 kbuf
->image
->start
-= sechdrs
[i
].sh_addr
;
943 kbuf
->image
->start
+= kbuf
->mem
+ offset
;
946 src
= (void *)pi
->ehdr
+ sechdrs
[i
].sh_offset
;
947 dst
= pi
->purgatory_buf
+ offset
;
948 memcpy(dst
, src
, sechdrs
[i
].sh_size
);
950 sechdrs
[i
].sh_addr
= kbuf
->mem
+ offset
;
951 sechdrs
[i
].sh_offset
= offset
;
952 offset
+= sechdrs
[i
].sh_size
;
958 static int kexec_apply_relocations(struct kimage
*image
)
961 struct purgatory_info
*pi
= &image
->purgatory_info
;
962 const Elf_Shdr
*sechdrs
;
964 sechdrs
= (void *)pi
->ehdr
+ pi
->ehdr
->e_shoff
;
966 for (i
= 0; i
< pi
->ehdr
->e_shnum
; i
++) {
967 const Elf_Shdr
*relsec
;
968 const Elf_Shdr
*symtab
;
971 relsec
= sechdrs
+ i
;
973 if (relsec
->sh_type
!= SHT_RELA
&&
974 relsec
->sh_type
!= SHT_REL
)
978 * For section of type SHT_RELA/SHT_REL,
979 * ->sh_link contains section header index of associated
980 * symbol table. And ->sh_info contains section header
981 * index of section to which relocations apply.
983 if (relsec
->sh_info
>= pi
->ehdr
->e_shnum
||
984 relsec
->sh_link
>= pi
->ehdr
->e_shnum
)
987 section
= pi
->sechdrs
+ relsec
->sh_info
;
988 symtab
= sechdrs
+ relsec
->sh_link
;
990 if (!(section
->sh_flags
& SHF_ALLOC
))
994 * symtab->sh_link contain section header index of associated
997 if (symtab
->sh_link
>= pi
->ehdr
->e_shnum
)
998 /* Invalid section number? */
1002 * Respective architecture needs to provide support for applying
1003 * relocations of type SHT_RELA/SHT_REL.
1005 if (relsec
->sh_type
== SHT_RELA
)
1006 ret
= arch_kexec_apply_relocations_add(pi
, section
,
1008 else if (relsec
->sh_type
== SHT_REL
)
1009 ret
= arch_kexec_apply_relocations(pi
, section
,
1019 * kexec_load_purgatory - Load and relocate the purgatory object.
1020 * @image: Image to add the purgatory to.
1021 * @kbuf: Memory parameters to use.
1023 * Allocates the memory needed for image->purgatory_info.sechdrs and
1024 * image->purgatory_info.purgatory_buf/kbuf->buffer. Caller is responsible
1025 * to free the memory after use.
1027 * Return: 0 on success, negative errno on error.
1029 int kexec_load_purgatory(struct kimage
*image
, struct kexec_buf
*kbuf
)
1031 struct purgatory_info
*pi
= &image
->purgatory_info
;
1034 if (kexec_purgatory_size
<= 0)
1037 pi
->ehdr
= (const Elf_Ehdr
*)kexec_purgatory
;
1039 ret
= kexec_purgatory_setup_kbuf(pi
, kbuf
);
1043 ret
= kexec_purgatory_setup_sechdrs(pi
, kbuf
);
1047 ret
= kexec_apply_relocations(image
);
1056 vfree(pi
->purgatory_buf
);
1057 pi
->purgatory_buf
= NULL
;
1062 * kexec_purgatory_find_symbol - find a symbol in the purgatory
1063 * @pi: Purgatory to search in.
1064 * @name: Name of the symbol.
1066 * Return: pointer to symbol in read-only symtab on success, NULL on error.
1068 static const Elf_Sym
*kexec_purgatory_find_symbol(struct purgatory_info
*pi
,
1071 const Elf_Shdr
*sechdrs
;
1072 const Elf_Ehdr
*ehdr
;
1073 const Elf_Sym
*syms
;
1081 sechdrs
= (void *)ehdr
+ ehdr
->e_shoff
;
1083 for (i
= 0; i
< ehdr
->e_shnum
; i
++) {
1084 if (sechdrs
[i
].sh_type
!= SHT_SYMTAB
)
1087 if (sechdrs
[i
].sh_link
>= ehdr
->e_shnum
)
1088 /* Invalid strtab section number */
1090 strtab
= (void *)ehdr
+ sechdrs
[sechdrs
[i
].sh_link
].sh_offset
;
1091 syms
= (void *)ehdr
+ sechdrs
[i
].sh_offset
;
1093 /* Go through symbols for a match */
1094 for (k
= 0; k
< sechdrs
[i
].sh_size
/sizeof(Elf_Sym
); k
++) {
1095 if (ELF_ST_BIND(syms
[k
].st_info
) != STB_GLOBAL
)
1098 if (strcmp(strtab
+ syms
[k
].st_name
, name
) != 0)
1101 if (syms
[k
].st_shndx
== SHN_UNDEF
||
1102 syms
[k
].st_shndx
>= ehdr
->e_shnum
) {
1103 pr_debug("Symbol: %s has bad section index %d.\n",
1104 name
, syms
[k
].st_shndx
);
1108 /* Found the symbol we are looking for */
1116 void *kexec_purgatory_get_symbol_addr(struct kimage
*image
, const char *name
)
1118 struct purgatory_info
*pi
= &image
->purgatory_info
;
1122 sym
= kexec_purgatory_find_symbol(pi
, name
);
1124 return ERR_PTR(-EINVAL
);
1126 sechdr
= &pi
->sechdrs
[sym
->st_shndx
];
1129 * Returns the address where symbol will finally be loaded after
1130 * kexec_load_segment()
1132 return (void *)(sechdr
->sh_addr
+ sym
->st_value
);
1136 * Get or set value of a symbol. If "get_value" is true, symbol value is
1137 * returned in buf otherwise symbol value is set based on value in buf.
1139 int kexec_purgatory_get_set_symbol(struct kimage
*image
, const char *name
,
1140 void *buf
, unsigned int size
, bool get_value
)
1142 struct purgatory_info
*pi
= &image
->purgatory_info
;
1147 sym
= kexec_purgatory_find_symbol(pi
, name
);
1151 if (sym
->st_size
!= size
) {
1152 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1153 name
, (unsigned long)sym
->st_size
, size
);
1157 sec
= pi
->sechdrs
+ sym
->st_shndx
;
1159 if (sec
->sh_type
== SHT_NOBITS
) {
1160 pr_err("symbol %s is in a bss section. Cannot %s\n", name
,
1161 get_value
? "get" : "set");
1165 sym_buf
= (char *)pi
->purgatory_buf
+ sec
->sh_offset
+ sym
->st_value
;
1168 memcpy((void *)buf
, sym_buf
, size
);
1170 memcpy((void *)sym_buf
, buf
, size
);
1174 #endif /* CONFIG_ARCH_HAS_KEXEC_PURGATORY */
1176 int crash_exclude_mem_range(struct crash_mem
*mem
,
1177 unsigned long long mstart
, unsigned long long mend
)
1180 unsigned long long start
, end
, p_start
, p_end
;
1181 struct crash_mem_range temp_range
= {0, 0};
1183 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1184 start
= mem
->ranges
[i
].start
;
1185 end
= mem
->ranges
[i
].end
;
1189 if (mstart
> end
|| mend
< start
)
1192 /* Truncate any area outside of range */
1198 /* Found completely overlapping range */
1199 if (p_start
== start
&& p_end
== end
) {
1200 mem
->ranges
[i
].start
= 0;
1201 mem
->ranges
[i
].end
= 0;
1202 if (i
< mem
->nr_ranges
- 1) {
1203 /* Shift rest of the ranges to left */
1204 for (j
= i
; j
< mem
->nr_ranges
- 1; j
++) {
1205 mem
->ranges
[j
].start
=
1206 mem
->ranges
[j
+1].start
;
1207 mem
->ranges
[j
].end
=
1208 mem
->ranges
[j
+1].end
;
1212 * Continue to check if there are another overlapping ranges
1213 * from the current position because of shifting the above
1224 if (p_start
> start
&& p_end
< end
) {
1225 /* Split original range */
1226 mem
->ranges
[i
].end
= p_start
- 1;
1227 temp_range
.start
= p_end
+ 1;
1228 temp_range
.end
= end
;
1229 } else if (p_start
!= start
)
1230 mem
->ranges
[i
].end
= p_start
- 1;
1232 mem
->ranges
[i
].start
= p_end
+ 1;
1236 /* If a split happened, add the split to array */
1237 if (!temp_range
.end
)
1240 /* Split happened */
1241 if (i
== mem
->max_nr_ranges
- 1)
1244 /* Location where new range should go */
1246 if (j
< mem
->nr_ranges
) {
1247 /* Move over all ranges one slot towards the end */
1248 for (i
= mem
->nr_ranges
- 1; i
>= j
; i
--)
1249 mem
->ranges
[i
+ 1] = mem
->ranges
[i
];
1252 mem
->ranges
[j
].start
= temp_range
.start
;
1253 mem
->ranges
[j
].end
= temp_range
.end
;
1258 int crash_prepare_elf64_headers(struct crash_mem
*mem
, int kernel_map
,
1259 void **addr
, unsigned long *sz
)
1263 unsigned long nr_cpus
= num_possible_cpus(), nr_phdr
, elf_sz
;
1265 unsigned int cpu
, i
;
1266 unsigned long long notes_addr
;
1267 unsigned long mstart
, mend
;
1269 /* extra phdr for vmcoreinfo ELF note */
1270 nr_phdr
= nr_cpus
+ 1;
1271 nr_phdr
+= mem
->nr_ranges
;
1274 * kexec-tools creates an extra PT_LOAD phdr for kernel text mapping
1275 * area (for example, ffffffff80000000 - ffffffffa0000000 on x86_64).
1276 * I think this is required by tools like gdb. So same physical
1277 * memory will be mapped in two ELF headers. One will contain kernel
1278 * text virtual addresses and other will have __va(physical) addresses.
1282 elf_sz
= sizeof(Elf64_Ehdr
) + nr_phdr
* sizeof(Elf64_Phdr
);
1283 elf_sz
= ALIGN(elf_sz
, ELF_CORE_HEADER_ALIGN
);
1285 buf
= vzalloc(elf_sz
);
1289 ehdr
= (Elf64_Ehdr
*)buf
;
1290 phdr
= (Elf64_Phdr
*)(ehdr
+ 1);
1291 memcpy(ehdr
->e_ident
, ELFMAG
, SELFMAG
);
1292 ehdr
->e_ident
[EI_CLASS
] = ELFCLASS64
;
1293 ehdr
->e_ident
[EI_DATA
] = ELFDATA2LSB
;
1294 ehdr
->e_ident
[EI_VERSION
] = EV_CURRENT
;
1295 ehdr
->e_ident
[EI_OSABI
] = ELF_OSABI
;
1296 memset(ehdr
->e_ident
+ EI_PAD
, 0, EI_NIDENT
- EI_PAD
);
1297 ehdr
->e_type
= ET_CORE
;
1298 ehdr
->e_machine
= ELF_ARCH
;
1299 ehdr
->e_version
= EV_CURRENT
;
1300 ehdr
->e_phoff
= sizeof(Elf64_Ehdr
);
1301 ehdr
->e_ehsize
= sizeof(Elf64_Ehdr
);
1302 ehdr
->e_phentsize
= sizeof(Elf64_Phdr
);
1304 /* Prepare one phdr of type PT_NOTE for each present CPU */
1305 for_each_present_cpu(cpu
) {
1306 phdr
->p_type
= PT_NOTE
;
1307 notes_addr
= per_cpu_ptr_to_phys(per_cpu_ptr(crash_notes
, cpu
));
1308 phdr
->p_offset
= phdr
->p_paddr
= notes_addr
;
1309 phdr
->p_filesz
= phdr
->p_memsz
= sizeof(note_buf_t
);
1314 /* Prepare one PT_NOTE header for vmcoreinfo */
1315 phdr
->p_type
= PT_NOTE
;
1316 phdr
->p_offset
= phdr
->p_paddr
= paddr_vmcoreinfo_note();
1317 phdr
->p_filesz
= phdr
->p_memsz
= VMCOREINFO_NOTE_SIZE
;
1321 /* Prepare PT_LOAD type program header for kernel text region */
1323 phdr
->p_type
= PT_LOAD
;
1324 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1325 phdr
->p_vaddr
= (unsigned long) _text
;
1326 phdr
->p_filesz
= phdr
->p_memsz
= _end
- _text
;
1327 phdr
->p_offset
= phdr
->p_paddr
= __pa_symbol(_text
);
1332 /* Go through all the ranges in mem->ranges[] and prepare phdr */
1333 for (i
= 0; i
< mem
->nr_ranges
; i
++) {
1334 mstart
= mem
->ranges
[i
].start
;
1335 mend
= mem
->ranges
[i
].end
;
1337 phdr
->p_type
= PT_LOAD
;
1338 phdr
->p_flags
= PF_R
|PF_W
|PF_X
;
1339 phdr
->p_offset
= mstart
;
1341 phdr
->p_paddr
= mstart
;
1342 phdr
->p_vaddr
= (unsigned long) __va(mstart
);
1343 phdr
->p_filesz
= phdr
->p_memsz
= mend
- mstart
+ 1;
1346 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",
1347 phdr
, phdr
->p_vaddr
, phdr
->p_paddr
, phdr
->p_filesz
,
1348 ehdr
->e_phnum
, phdr
->p_offset
);