2 * kexec: kexec_file_load system call
4 * Copyright (C) 2014 Red Hat Inc.
6 * Vivek Goyal <vgoyal@redhat.com>
8 * This source code is licensed under the GNU General Public License,
9 * Version 2. See the file COPYING for more details.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <linux/capability.h>
16 #include <linux/file.h>
17 #include <linux/slab.h>
18 #include <linux/kexec.h>
19 #include <linux/mutex.h>
20 #include <linux/list.h>
22 #include <linux/ima.h>
23 #include <crypto/hash.h>
24 #include <crypto/sha.h>
25 #include <linux/elf.h>
26 #include <linux/elfcore.h>
27 #include <linux/kernel.h>
28 #include <linux/kexec.h>
29 #include <linux/slab.h>
30 #include <linux/syscalls.h>
31 #include <linux/vmalloc.h>
32 #include "kexec_internal.h"
34 static int kexec_calculate_store_digests(struct kimage *image);
37 * Currently this is the only default function that is exported as some
38 * architectures need it to do additional handlings.
39 * In the future, other default functions may be exported too if required.
41 int kexec_image_probe_default(struct kimage *image, void *buf,
42 unsigned long buf_len)
44 const struct kexec_file_ops * const *fops;
47 for (fops = &kexec_file_loaders[0]; *fops && (*fops)->probe; ++fops) {
48 ret = (*fops)->probe(buf, buf_len);
58 /* Architectures can provide this probe function */
59 int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,
60 unsigned long buf_len)
62 return kexec_image_probe_default(image, buf, buf_len);
65 static void *kexec_image_load_default(struct kimage *image)
67 if (!image->fops || !image->fops->load)
68 return ERR_PTR(-ENOEXEC);
70 return image->fops->load(image, image->kernel_buf,
71 image->kernel_buf_len, image->initrd_buf,
72 image->initrd_buf_len, image->cmdline_buf,
73 image->cmdline_buf_len);
76 void * __weak arch_kexec_kernel_image_load(struct kimage *image)
78 return kexec_image_load_default(image);
81 static int kexec_image_post_load_cleanup_default(struct kimage *image)
83 if (!image->fops || !image->fops->cleanup)
86 return image->fops->cleanup(image->image_loader_data);
89 int __weak arch_kimage_file_post_load_cleanup(struct kimage *image)
91 return kexec_image_post_load_cleanup_default(image);
94 #ifdef CONFIG_KEXEC_VERIFY_SIG
95 static int kexec_image_verify_sig_default(struct kimage *image, void *buf,
96 unsigned long buf_len)
98 if (!image->fops || !image->fops->verify_sig) {
99 pr_debug("kernel loader does not support signature verification.\n");
100 return -EKEYREJECTED;
103 return image->fops->verify_sig(buf, buf_len);
106 int __weak arch_kexec_kernel_verify_sig(struct kimage *image, void *buf,
107 unsigned long buf_len)
109 return kexec_image_verify_sig_default(image, buf, buf_len);
114 * arch_kexec_apply_relocations_add - apply relocations of type RELA
115 * @pi: Purgatory to be relocated.
116 * @section: Section relocations applying to.
117 * @relsec: Section containing RELAs.
118 * @symtab: Corresponding symtab.
120 * Return: 0 on success, negative errno on error.
123 arch_kexec_apply_relocations_add(struct purgatory_info *pi, Elf_Shdr *section,
124 const Elf_Shdr *relsec, const Elf_Shdr *symtab)
126 pr_err("RELA relocation unsupported.\n");
131 * arch_kexec_apply_relocations - apply relocations of type REL
132 * @pi: Purgatory to be relocated.
133 * @section: Section relocations applying to.
134 * @relsec: Section containing RELs.
135 * @symtab: Corresponding symtab.
137 * Return: 0 on success, negative errno on error.
140 arch_kexec_apply_relocations(struct purgatory_info *pi, Elf_Shdr *section,
141 const Elf_Shdr *relsec, const Elf_Shdr *symtab)
143 pr_err("REL relocation unsupported.\n");
148 * Free up memory used by kernel, initrd, and command line. This is temporary
149 * memory allocation which is not needed any more after these buffers have
150 * been loaded into separate segments and have been copied elsewhere.
152 void kimage_file_post_load_cleanup(struct kimage *image)
154 struct purgatory_info *pi = &image->purgatory_info;
156 vfree(image->kernel_buf);
157 image->kernel_buf = NULL;
159 vfree(image->initrd_buf);
160 image->initrd_buf = NULL;
162 kfree(image->cmdline_buf);
163 image->cmdline_buf = NULL;
165 vfree(pi->purgatory_buf);
166 pi->purgatory_buf = NULL;
171 /* See if architecture has anything to cleanup post load */
172 arch_kimage_file_post_load_cleanup(image);
175 * Above call should have called into bootloader to free up
176 * any data stored in kimage->image_loader_data. It should
177 * be ok now to free it up.
179 kfree(image->image_loader_data);
180 image->image_loader_data = NULL;
184 * In file mode list of segments is prepared by kernel. Copy relevant
185 * data from user space, do error checking, prepare segment list
188 kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,
189 const char __user *cmdline_ptr,
190 unsigned long cmdline_len, unsigned flags)
196 ret = kernel_read_file_from_fd(kernel_fd, &image->kernel_buf,
197 &size, INT_MAX, READING_KEXEC_IMAGE);
200 image->kernel_buf_len = size;
202 /* IMA needs to pass the measurement list to the next kernel. */
203 ima_add_kexec_buffer(image);
205 /* Call arch image probe handlers */
206 ret = arch_kexec_kernel_image_probe(image, image->kernel_buf,
207 image->kernel_buf_len);
211 #ifdef CONFIG_KEXEC_VERIFY_SIG
212 ret = arch_kexec_kernel_verify_sig(image, image->kernel_buf,
213 image->kernel_buf_len);
215 pr_debug("kernel signature verification failed.\n");
218 pr_debug("kernel signature verification successful.\n");
220 /* It is possible that there no initramfs is being loaded */
221 if (!(flags & KEXEC_FILE_NO_INITRAMFS)) {
222 ret = kernel_read_file_from_fd(initrd_fd, &image->initrd_buf,
224 READING_KEXEC_INITRAMFS);
227 image->initrd_buf_len = size;
231 image->cmdline_buf = memdup_user(cmdline_ptr, cmdline_len);
232 if (IS_ERR(image->cmdline_buf)) {
233 ret = PTR_ERR(image->cmdline_buf);
234 image->cmdline_buf = NULL;
238 image->cmdline_buf_len = cmdline_len;
240 /* command line should be a string with last byte null */
241 if (image->cmdline_buf[cmdline_len - 1] != '\0') {
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);
505 * arch_kexec_walk_mem - call func(data) on free memory regions
506 * @kbuf: Context info for the search. Also passed to @func.
507 * @func: Function to call for each memory region.
509 * Return: The memory walk will stop when func returns a non-zero value
510 * and that value will be returned. If all free regions are visited without
511 * func returning non-zero, then zero will be returned.
513 int __weak arch_kexec_walk_mem(struct kexec_buf *kbuf,
514 int (*func)(struct resource *, void *))
516 if (kbuf->image->type == KEXEC_TYPE_CRASH)
517 return walk_iomem_res_desc(crashk_res.desc,
518 IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY,
519 crashk_res.start, crashk_res.end,
522 return walk_system_ram_res(0, ULONG_MAX, kbuf, func);
526 * kexec_locate_mem_hole - find free memory for the purgatory or the next kernel
527 * @kbuf: Parameters for the memory search.
529 * On success, kbuf->mem will have the start address of the memory region found.
531 * Return: 0 on success, negative errno on error.
533 int kexec_locate_mem_hole(struct kexec_buf *kbuf)
537 ret = arch_kexec_walk_mem(kbuf, locate_mem_hole_callback);
539 return ret == 1 ? 0 : -EADDRNOTAVAIL;
543 * kexec_add_buffer - place a buffer in a kexec segment
544 * @kbuf: Buffer contents and memory parameters.
546 * This function assumes that kexec_mutex is held.
547 * On successful return, @kbuf->mem will have the physical address of
548 * the buffer in memory.
550 * Return: 0 on success, negative errno on error.
552 int kexec_add_buffer(struct kexec_buf *kbuf)
555 struct kexec_segment *ksegment;
558 /* Currently adding segment this way is allowed only in file mode */
559 if (!kbuf->image->file_mode)
562 if (kbuf->image->nr_segments >= KEXEC_SEGMENT_MAX)
566 * Make sure we are not trying to add buffer after allocating
567 * control pages. All segments need to be placed first before
568 * any control pages are allocated. As control page allocation
569 * logic goes through list of segments to make sure there are
570 * no destination overlaps.
572 if (!list_empty(&kbuf->image->control_pages)) {
577 /* Ensure minimum alignment needed for segments. */
578 kbuf->memsz = ALIGN(kbuf->memsz, PAGE_SIZE);
579 kbuf->buf_align = max(kbuf->buf_align, PAGE_SIZE);
581 /* Walk the RAM ranges and allocate a suitable range for the buffer */
582 ret = kexec_locate_mem_hole(kbuf);
586 /* Found a suitable memory range */
587 ksegment = &kbuf->image->segment[kbuf->image->nr_segments];
588 ksegment->kbuf = kbuf->buffer;
589 ksegment->bufsz = kbuf->bufsz;
590 ksegment->mem = kbuf->mem;
591 ksegment->memsz = kbuf->memsz;
592 kbuf->image->nr_segments++;
596 /* Calculate and store the digest of segments */
597 static int kexec_calculate_store_digests(struct kimage *image)
599 struct crypto_shash *tfm;
600 struct shash_desc *desc;
601 int ret = 0, i, j, zero_buf_sz, sha_region_sz;
602 size_t desc_size, nullsz;
605 struct kexec_sha_region *sha_regions;
606 struct purgatory_info *pi = &image->purgatory_info;
608 if (!IS_ENABLED(CONFIG_ARCH_HAS_KEXEC_PURGATORY))
611 zero_buf = __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT);
612 zero_buf_sz = PAGE_SIZE;
614 tfm = crypto_alloc_shash("sha256", 0, 0);
620 desc_size = crypto_shash_descsize(tfm) + sizeof(*desc);
621 desc = kzalloc(desc_size, GFP_KERNEL);
627 sha_region_sz = KEXEC_SEGMENT_MAX * sizeof(struct kexec_sha_region);
628 sha_regions = vzalloc(sha_region_sz);
635 ret = crypto_shash_init(desc);
637 goto out_free_sha_regions;
639 digest = kzalloc(SHA256_DIGEST_SIZE, GFP_KERNEL);
642 goto out_free_sha_regions;
645 for (j = i = 0; i < image->nr_segments; i++) {
646 struct kexec_segment *ksegment;
648 ksegment = &image->segment[i];
650 * Skip purgatory as it will be modified once we put digest
653 if (ksegment->kbuf == pi->purgatory_buf)
656 ret = crypto_shash_update(desc, ksegment->kbuf,
662 * Assume rest of the buffer is filled with zero and
663 * update digest accordingly.
665 nullsz = ksegment->memsz - ksegment->bufsz;
667 unsigned long bytes = nullsz;
669 if (bytes > zero_buf_sz)
671 ret = crypto_shash_update(desc, zero_buf, bytes);
680 sha_regions[j].start = ksegment->mem;
681 sha_regions[j].len = ksegment->memsz;
686 ret = crypto_shash_final(desc, digest);
688 goto out_free_digest;
689 ret = kexec_purgatory_get_set_symbol(image, "purgatory_sha_regions",
690 sha_regions, sha_region_sz, 0);
692 goto out_free_digest;
694 ret = kexec_purgatory_get_set_symbol(image, "purgatory_sha256_digest",
695 digest, SHA256_DIGEST_SIZE, 0);
697 goto out_free_digest;
702 out_free_sha_regions:
712 #ifdef CONFIG_ARCH_HAS_KEXEC_PURGATORY
713 /* Actually load purgatory. Lot of code taken from kexec-tools */
714 static int __kexec_load_purgatory(struct kimage *image, unsigned long min,
715 unsigned long max, int top_down)
717 struct purgatory_info *pi = &image->purgatory_info;
718 unsigned long align, bss_align, bss_sz, bss_pad;
719 unsigned long entry, load_addr, curr_load_addr, bss_addr, offset;
720 unsigned char *buf_addr, *src;
721 int i, ret = 0, entry_sidx = -1;
722 const Elf_Shdr *sechdrs_c;
723 Elf_Shdr *sechdrs = NULL;
724 struct kexec_buf kbuf = { .image = image, .bufsz = 0, .buf_align = 1,
725 .buf_min = min, .buf_max = max,
726 .top_down = top_down };
729 * sechdrs_c points to section headers in purgatory and are read
730 * only. No modifications allowed.
732 sechdrs_c = (void *)pi->ehdr + pi->ehdr->e_shoff;
735 * We can not modify sechdrs_c[] and its fields. It is read only.
736 * Copy it over to a local copy where one can store some temporary
737 * data and free it at the end. We need to modify ->sh_addr and
738 * ->sh_offset fields to keep track of permanent and temporary
739 * locations of sections.
741 sechdrs = vzalloc(pi->ehdr->e_shnum * sizeof(Elf_Shdr));
745 memcpy(sechdrs, sechdrs_c, pi->ehdr->e_shnum * sizeof(Elf_Shdr));
748 * We seem to have multiple copies of sections. First copy is which
749 * is embedded in kernel in read only section. Some of these sections
750 * will be copied to a temporary buffer and relocated. And these
751 * sections will finally be copied to their final destination at
754 * Use ->sh_offset to reflect section address in memory. It will
755 * point to original read only copy if section is not allocatable.
756 * Otherwise it will point to temporary copy which will be relocated.
758 * Use ->sh_addr to contain final address of the section where it
759 * will go during execution time.
761 for (i = 0; i < pi->ehdr->e_shnum; i++) {
762 if (sechdrs[i].sh_type == SHT_NOBITS)
765 sechdrs[i].sh_offset = (unsigned long)pi->ehdr +
766 sechdrs[i].sh_offset;
770 * Identify entry point section and make entry relative to section
773 entry = pi->ehdr->e_entry;
774 for (i = 0; i < pi->ehdr->e_shnum; i++) {
775 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
778 if (!(sechdrs[i].sh_flags & SHF_EXECINSTR))
781 /* Make entry section relative */
782 if (sechdrs[i].sh_addr <= pi->ehdr->e_entry &&
783 ((sechdrs[i].sh_addr + sechdrs[i].sh_size) >
784 pi->ehdr->e_entry)) {
786 entry -= sechdrs[i].sh_addr;
791 /* Determine how much memory is needed to load relocatable object. */
795 for (i = 0; i < pi->ehdr->e_shnum; i++) {
796 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
799 align = sechdrs[i].sh_addralign;
800 if (sechdrs[i].sh_type != SHT_NOBITS) {
801 if (kbuf.buf_align < align)
802 kbuf.buf_align = align;
803 kbuf.bufsz = ALIGN(kbuf.bufsz, align);
804 kbuf.bufsz += sechdrs[i].sh_size;
807 if (bss_align < align)
809 bss_sz = ALIGN(bss_sz, align);
810 bss_sz += sechdrs[i].sh_size;
814 /* Determine the bss padding required to align bss properly */
816 if (kbuf.bufsz & (bss_align - 1))
817 bss_pad = bss_align - (kbuf.bufsz & (bss_align - 1));
819 kbuf.memsz = kbuf.bufsz + bss_pad + bss_sz;
821 /* Allocate buffer for purgatory */
822 kbuf.buffer = vzalloc(kbuf.bufsz);
828 if (kbuf.buf_align < bss_align)
829 kbuf.buf_align = bss_align;
831 /* Add buffer to segment list */
832 ret = kexec_add_buffer(&kbuf);
835 pi->purgatory_load_addr = kbuf.mem;
837 /* Load SHF_ALLOC sections */
838 buf_addr = kbuf.buffer;
839 load_addr = curr_load_addr = pi->purgatory_load_addr;
840 bss_addr = load_addr + kbuf.bufsz + bss_pad;
842 for (i = 0; i < pi->ehdr->e_shnum; i++) {
843 if (!(sechdrs[i].sh_flags & SHF_ALLOC))
846 align = sechdrs[i].sh_addralign;
847 if (sechdrs[i].sh_type != SHT_NOBITS) {
848 curr_load_addr = ALIGN(curr_load_addr, align);
849 offset = curr_load_addr - load_addr;
850 /* We already modifed ->sh_offset to keep src addr */
851 src = (char *) sechdrs[i].sh_offset;
852 memcpy(buf_addr + offset, src, sechdrs[i].sh_size);
854 /* Store load address and source address of section */
855 sechdrs[i].sh_addr = curr_load_addr;
858 * This section got copied to temporary buffer. Update
859 * ->sh_offset accordingly.
861 sechdrs[i].sh_offset = (unsigned long)(buf_addr + offset);
863 /* Advance to the next address */
864 curr_load_addr += sechdrs[i].sh_size;
866 bss_addr = ALIGN(bss_addr, align);
867 sechdrs[i].sh_addr = bss_addr;
868 bss_addr += sechdrs[i].sh_size;
872 /* Update entry point based on load address of text section */
874 entry += sechdrs[entry_sidx].sh_addr;
876 /* Make kernel jump to purgatory after shutdown */
877 image->start = entry;
879 /* Used later to get/set symbol values */
880 pi->sechdrs = sechdrs;
883 * Used later to identify which section is purgatory and skip it
886 pi->purgatory_buf = kbuf.buffer;
894 static int kexec_apply_relocations(struct kimage *image)
897 struct purgatory_info *pi = &image->purgatory_info;
898 const Elf_Shdr *sechdrs;
900 sechdrs = (void *)pi->ehdr + pi->ehdr->e_shoff;
902 for (i = 0; i < pi->ehdr->e_shnum; i++) {
903 const Elf_Shdr *relsec;
904 const Elf_Shdr *symtab;
907 relsec = sechdrs + i;
909 if (relsec->sh_type != SHT_RELA &&
910 relsec->sh_type != SHT_REL)
914 * For section of type SHT_RELA/SHT_REL,
915 * ->sh_link contains section header index of associated
916 * symbol table. And ->sh_info contains section header
917 * index of section to which relocations apply.
919 if (relsec->sh_info >= pi->ehdr->e_shnum ||
920 relsec->sh_link >= pi->ehdr->e_shnum)
923 section = pi->sechdrs + relsec->sh_info;
924 symtab = sechdrs + relsec->sh_link;
926 if (!(section->sh_flags & SHF_ALLOC))
930 * symtab->sh_link contain section header index of associated
933 if (symtab->sh_link >= pi->ehdr->e_shnum)
934 /* Invalid section number? */
938 * Respective architecture needs to provide support for applying
939 * relocations of type SHT_RELA/SHT_REL.
941 if (relsec->sh_type == SHT_RELA)
942 ret = arch_kexec_apply_relocations_add(pi, section,
944 else if (relsec->sh_type == SHT_REL)
945 ret = arch_kexec_apply_relocations(pi, section,
954 /* Load relocatable purgatory object and relocate it appropriately */
955 int kexec_load_purgatory(struct kimage *image, unsigned long min,
956 unsigned long max, int top_down,
957 unsigned long *load_addr)
959 struct purgatory_info *pi = &image->purgatory_info;
962 if (kexec_purgatory_size <= 0)
965 pi->ehdr = (const Elf_Ehdr *)kexec_purgatory;
967 ret = __kexec_load_purgatory(image, min, max, top_down);
971 ret = kexec_apply_relocations(image);
975 *load_addr = pi->purgatory_load_addr;
981 vfree(pi->purgatory_buf);
982 pi->purgatory_buf = NULL;
987 * kexec_purgatory_find_symbol - find a symbol in the purgatory
988 * @pi: Purgatory to search in.
989 * @name: Name of the symbol.
991 * Return: pointer to symbol in read-only symtab on success, NULL on error.
993 static const Elf_Sym *kexec_purgatory_find_symbol(struct purgatory_info *pi,
996 const Elf_Shdr *sechdrs;
997 const Elf_Ehdr *ehdr;
1006 sechdrs = (void *)ehdr + ehdr->e_shoff;
1008 for (i = 0; i < ehdr->e_shnum; i++) {
1009 if (sechdrs[i].sh_type != SHT_SYMTAB)
1012 if (sechdrs[i].sh_link >= ehdr->e_shnum)
1013 /* Invalid strtab section number */
1015 strtab = (void *)ehdr + sechdrs[sechdrs[i].sh_link].sh_offset;
1016 syms = (void *)ehdr + sechdrs[i].sh_offset;
1018 /* Go through symbols for a match */
1019 for (k = 0; k < sechdrs[i].sh_size/sizeof(Elf_Sym); k++) {
1020 if (ELF_ST_BIND(syms[k].st_info) != STB_GLOBAL)
1023 if (strcmp(strtab + syms[k].st_name, name) != 0)
1026 if (syms[k].st_shndx == SHN_UNDEF ||
1027 syms[k].st_shndx >= ehdr->e_shnum) {
1028 pr_debug("Symbol: %s has bad section index %d.\n",
1029 name, syms[k].st_shndx);
1033 /* Found the symbol we are looking for */
1041 void *kexec_purgatory_get_symbol_addr(struct kimage *image, const char *name)
1043 struct purgatory_info *pi = &image->purgatory_info;
1047 sym = kexec_purgatory_find_symbol(pi, name);
1049 return ERR_PTR(-EINVAL);
1051 sechdr = &pi->sechdrs[sym->st_shndx];
1054 * Returns the address where symbol will finally be loaded after
1055 * kexec_load_segment()
1057 return (void *)(sechdr->sh_addr + sym->st_value);
1061 * Get or set value of a symbol. If "get_value" is true, symbol value is
1062 * returned in buf otherwise symbol value is set based on value in buf.
1064 int kexec_purgatory_get_set_symbol(struct kimage *image, const char *name,
1065 void *buf, unsigned int size, bool get_value)
1067 struct purgatory_info *pi = &image->purgatory_info;
1072 sym = kexec_purgatory_find_symbol(pi, name);
1076 if (sym->st_size != size) {
1077 pr_err("symbol %s size mismatch: expected %lu actual %u\n",
1078 name, (unsigned long)sym->st_size, size);
1082 sec = pi->sechdrs + sym->st_shndx;
1084 if (sec->sh_type == SHT_NOBITS) {
1085 pr_err("symbol %s is in a bss section. Cannot %s\n", name,
1086 get_value ? "get" : "set");
1090 sym_buf = (char *)sec->sh_offset + sym->st_value;
1093 memcpy((void *)buf, sym_buf, size);
1095 memcpy((void *)sym_buf, buf, size);
1099 #endif /* CONFIG_ARCH_HAS_KEXEC_PURGATORY */
1101 int crash_exclude_mem_range(struct crash_mem *mem,
1102 unsigned long long mstart, unsigned long long mend)
1105 unsigned long long start, end;
1106 struct crash_mem_range temp_range = {0, 0};
1108 for (i = 0; i < mem->nr_ranges; i++) {
1109 start = mem->ranges[i].start;
1110 end = mem->ranges[i].end;
1112 if (mstart > end || mend < start)
1115 /* Truncate any area outside of range */
1121 /* Found completely overlapping range */
1122 if (mstart == start && mend == end) {
1123 mem->ranges[i].start = 0;
1124 mem->ranges[i].end = 0;
1125 if (i < mem->nr_ranges - 1) {
1126 /* Shift rest of the ranges to left */
1127 for (j = i; j < mem->nr_ranges - 1; j++) {
1128 mem->ranges[j].start =
1129 mem->ranges[j+1].start;
1130 mem->ranges[j].end =
1131 mem->ranges[j+1].end;
1138 if (mstart > start && mend < end) {
1139 /* Split original range */
1140 mem->ranges[i].end = mstart - 1;
1141 temp_range.start = mend + 1;
1142 temp_range.end = end;
1143 } else if (mstart != start)
1144 mem->ranges[i].end = mstart - 1;
1146 mem->ranges[i].start = mend + 1;
1150 /* If a split happened, add the split to array */
1151 if (!temp_range.end)
1154 /* Split happened */
1155 if (i == mem->max_nr_ranges - 1)
1158 /* Location where new range should go */
1160 if (j < mem->nr_ranges) {
1161 /* Move over all ranges one slot towards the end */
1162 for (i = mem->nr_ranges - 1; i >= j; i--)
1163 mem->ranges[i + 1] = mem->ranges[i];
1166 mem->ranges[j].start = temp_range.start;
1167 mem->ranges[j].end = temp_range.end;
1172 int crash_prepare_elf64_headers(struct crash_mem *mem, int kernel_map,
1173 void **addr, unsigned long *sz)
1177 unsigned long nr_cpus = num_possible_cpus(), nr_phdr, elf_sz;
1179 unsigned int cpu, i;
1180 unsigned long long notes_addr;
1181 unsigned long mstart, mend;
1183 /* extra phdr for vmcoreinfo elf note */
1184 nr_phdr = nr_cpus + 1;
1185 nr_phdr += mem->nr_ranges;
1188 * kexec-tools creates an extra PT_LOAD phdr for kernel text mapping
1189 * area (for example, ffffffff80000000 - ffffffffa0000000 on x86_64).
1190 * I think this is required by tools like gdb. So same physical
1191 * memory will be mapped in two elf headers. One will contain kernel
1192 * text virtual addresses and other will have __va(physical) addresses.
1196 elf_sz = sizeof(Elf64_Ehdr) + nr_phdr * sizeof(Elf64_Phdr);
1197 elf_sz = ALIGN(elf_sz, ELF_CORE_HEADER_ALIGN);
1199 buf = vzalloc(elf_sz);
1203 ehdr = (Elf64_Ehdr *)buf;
1204 phdr = (Elf64_Phdr *)(ehdr + 1);
1205 memcpy(ehdr->e_ident, ELFMAG, SELFMAG);
1206 ehdr->e_ident[EI_CLASS] = ELFCLASS64;
1207 ehdr->e_ident[EI_DATA] = ELFDATA2LSB;
1208 ehdr->e_ident[EI_VERSION] = EV_CURRENT;
1209 ehdr->e_ident[EI_OSABI] = ELF_OSABI;
1210 memset(ehdr->e_ident + EI_PAD, 0, EI_NIDENT - EI_PAD);
1211 ehdr->e_type = ET_CORE;
1212 ehdr->e_machine = ELF_ARCH;
1213 ehdr->e_version = EV_CURRENT;
1214 ehdr->e_phoff = sizeof(Elf64_Ehdr);
1215 ehdr->e_ehsize = sizeof(Elf64_Ehdr);
1216 ehdr->e_phentsize = sizeof(Elf64_Phdr);
1218 /* Prepare one phdr of type PT_NOTE for each present cpu */
1219 for_each_present_cpu(cpu) {
1220 phdr->p_type = PT_NOTE;
1221 notes_addr = per_cpu_ptr_to_phys(per_cpu_ptr(crash_notes, cpu));
1222 phdr->p_offset = phdr->p_paddr = notes_addr;
1223 phdr->p_filesz = phdr->p_memsz = sizeof(note_buf_t);
1228 /* Prepare one PT_NOTE header for vmcoreinfo */
1229 phdr->p_type = PT_NOTE;
1230 phdr->p_offset = phdr->p_paddr = paddr_vmcoreinfo_note();
1231 phdr->p_filesz = phdr->p_memsz = VMCOREINFO_NOTE_SIZE;
1235 /* Prepare PT_LOAD type program header for kernel text region */
1237 phdr->p_type = PT_LOAD;
1238 phdr->p_flags = PF_R|PF_W|PF_X;
1239 phdr->p_vaddr = (Elf64_Addr)_text;
1240 phdr->p_filesz = phdr->p_memsz = _end - _text;
1241 phdr->p_offset = phdr->p_paddr = __pa_symbol(_text);
1246 /* Go through all the ranges in mem->ranges[] and prepare phdr */
1247 for (i = 0; i < mem->nr_ranges; i++) {
1248 mstart = mem->ranges[i].start;
1249 mend = mem->ranges[i].end;
1251 phdr->p_type = PT_LOAD;
1252 phdr->p_flags = PF_R|PF_W|PF_X;
1253 phdr->p_offset = mstart;
1255 phdr->p_paddr = mstart;
1256 phdr->p_vaddr = (unsigned long long) __va(mstart);
1257 phdr->p_filesz = phdr->p_memsz = mend - mstart + 1;
1261 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",
1262 phdr, phdr->p_vaddr, phdr->p_paddr, phdr->p_filesz,
1263 ehdr->e_phnum, phdr->p_offset);