kexec: split kexec_file syscall code to kexec_file.c
authorDave Young <dyoung@redhat.com>
Wed, 9 Sep 2015 22:38:51 +0000 (15:38 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 10 Sep 2015 20:29:01 +0000 (13:29 -0700)
Split kexec_file syscall related code to another file kernel/kexec_file.c
so that the #ifdef CONFIG_KEXEC_FILE in kexec.c can be dropped.

Sharing variables and functions are moved to kernel/kexec_internal.h per
suggestion from Vivek and Petr.

[akpm@linux-foundation.org: fix bisectability]
[akpm@linux-foundation.org: declare the various arch_kexec functions]
[akpm@linux-foundation.org: fix build]
Signed-off-by: Dave Young <dyoung@redhat.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: Petr Tesarik <ptesarik@suse.cz>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/kexec.h
kernel/Makefile
kernel/kexec.c
kernel/kexec_file.c [new file with mode: 0644]
kernel/kexec_internal.h [new file with mode: 0644]

index b63218f68c4b5a2c2862b082f62a1fe6caa5d128..ab150ade0d1879d02e3a29a19f5139eca78d592a 100644 (file)
@@ -318,6 +318,17 @@ int crash_shrink_memory(unsigned long new_size);
 size_t crash_get_memory_size(void);
 void crash_free_reserved_phys_range(unsigned long begin, unsigned long end);
 
+int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,
+                                        unsigned long buf_len);
+void * __weak arch_kexec_kernel_image_load(struct kimage *image);
+int __weak arch_kimage_file_post_load_cleanup(struct kimage *image);
+int __weak arch_kexec_kernel_verify_sig(struct kimage *image, void *buf,
+                                       unsigned long buf_len);
+int __weak arch_kexec_apply_relocations_add(const Elf_Ehdr *ehdr,
+                                       Elf_Shdr *sechdrs, unsigned int relsec);
+int __weak arch_kexec_apply_relocations(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
+                                       unsigned int relsec);
+
 #else /* !CONFIG_KEXEC */
 struct pt_regs;
 struct task_struct;
index e0d7587e7684e7b4408b69a189870eca3f9bce13..1b4890af5a659c616bae04757a21e6aeaecebe60 100644 (file)
@@ -50,6 +50,7 @@ obj-$(CONFIG_MODULE_SIG) += module_signing.o
 obj-$(CONFIG_KALLSYMS) += kallsyms.o
 obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o
 obj-$(CONFIG_KEXEC) += kexec.o
+obj-$(CONFIG_KEXEC_FILE) += kexec_file.o
 obj-$(CONFIG_BACKTRACE_SELF_TEST) += backtracetest.o
 obj-$(CONFIG_COMPAT) += compat.o
 obj-$(CONFIG_CGROUPS) += cgroup.o
index a785c1015e25bf1ecacd3a6d92956e3e630e7f37..2d73ecfa550581ae86d7851d29dfe0d8dd44b313 100644 (file)
 #include <linux/suspend.h>
 #include <linux/device.h>
 #include <linux/freezer.h>
+#include <linux/vmalloc.h>
 #include <linux/pm.h>
 #include <linux/cpu.h>
 #include <linux/console.h>
-#include <linux/vmalloc.h>
 #include <linux/swap.h>
 #include <linux/syscore_ops.h>
 #include <linux/compiler.h>
@@ -44,6 +44,9 @@
 
 #include <crypto/hash.h>
 #include <crypto/sha.h>
+#include "kexec_internal.h"
+
+DEFINE_MUTEX(kexec_mutex);
 
 /* Per cpu memory for storing cpu states in case of system crash. */
 note_buf_t __percpu *crash_notes;
@@ -57,16 +60,6 @@ size_t vmcoreinfo_max_size = sizeof(vmcoreinfo_data);
 /* Flag to indicate we are going to kexec a new kernel */
 bool kexec_in_progress = false;
 
-/*
- * Declare these symbols weak so that if architecture provides a purgatory,
- * these will be overridden.
- */
-char __weak kexec_purgatory[0];
-size_t __weak kexec_purgatory_size = 0;
-
-#ifdef CONFIG_KEXEC_FILE
-static int kexec_calculate_store_digests(struct kimage *image);
-#endif
 
 /* Location of the reserved area for the crash kernel */
 struct resource crashk_res = {
@@ -146,8 +139,6 @@ int kexec_should_crash(struct task_struct *p)
  */
 #define KIMAGE_NO_DEST (-1UL)
 
-static int kimage_is_destination_range(struct kimage *image,
-                                      unsigned long start, unsigned long end);
 static struct page *kimage_alloc_page(struct kimage *image,
                                       gfp_t gfp_mask,
                                       unsigned long dest);
@@ -169,7 +160,7 @@ static int copy_user_segment_list(struct kimage *image,
        return ret;
 }
 
-static int sanity_check_segment_list(struct kimage *image)
+int sanity_check_segment_list(struct kimage *image)
 {
        int result, i;
        unsigned long nr_segments = image->nr_segments;
@@ -259,7 +250,7 @@ static int sanity_check_segment_list(struct kimage *image)
        return 0;
 }
 
-static struct kimage *do_kimage_alloc_init(void)
+struct kimage *do_kimage_alloc_init(void)
 {
        struct kimage *image;
 
@@ -286,8 +277,6 @@ static struct kimage *do_kimage_alloc_init(void)
        return image;
 }
 
-static void kimage_free_page_list(struct list_head *list);
-
 static int kimage_alloc_init(struct kimage **rimage, unsigned long entry,
                             unsigned long nr_segments,
                             struct kexec_segment __user *segments,
@@ -354,283 +343,7 @@ out_free_image:
        return ret;
 }
 
-#ifdef CONFIG_KEXEC_FILE
-static int copy_file_from_fd(int fd, void **buf, unsigned long *buf_len)
-{
-       struct fd f = fdget(fd);
-       int ret;
-       struct kstat stat;
-       loff_t pos;
-       ssize_t bytes = 0;
-
-       if (!f.file)
-               return -EBADF;
-
-       ret = vfs_getattr(&f.file->f_path, &stat);
-       if (ret)
-               goto out;
-
-       if (stat.size > INT_MAX) {
-               ret = -EFBIG;
-               goto out;
-       }
-
-       /* Don't hand 0 to vmalloc, it whines. */
-       if (stat.size == 0) {
-               ret = -EINVAL;
-               goto out;
-       }
-
-       *buf = vmalloc(stat.size);
-       if (!*buf) {
-               ret = -ENOMEM;
-               goto out;
-       }
-
-       pos = 0;
-       while (pos < stat.size) {
-               bytes = kernel_read(f.file, pos, (char *)(*buf) + pos,
-                                   stat.size - pos);
-               if (bytes < 0) {
-                       vfree(*buf);
-                       ret = bytes;
-                       goto out;
-               }
-
-               if (bytes == 0)
-                       break;
-               pos += bytes;
-       }
-
-       if (pos != stat.size) {
-               ret = -EBADF;
-               vfree(*buf);
-               goto out;
-       }
-
-       *buf_len = pos;
-out:
-       fdput(f);
-       return ret;
-}
-
-/* Architectures can provide this probe function */
-int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,
-                                        unsigned long buf_len)
-{
-       return -ENOEXEC;
-}
-
-void * __weak arch_kexec_kernel_image_load(struct kimage *image)
-{
-       return ERR_PTR(-ENOEXEC);
-}
-
-void __weak arch_kimage_file_post_load_cleanup(struct kimage *image)
-{
-}
-
-int __weak arch_kexec_kernel_verify_sig(struct kimage *image, void *buf,
-                                       unsigned long buf_len)
-{
-       return -EKEYREJECTED;
-}
-
-/* Apply relocations of type RELA */
-int __weak
-arch_kexec_apply_relocations_add(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
-                                unsigned int relsec)
-{
-       pr_err("RELA relocation unsupported.\n");
-       return -ENOEXEC;
-}
-
-/* Apply relocations of type REL */
-int __weak
-arch_kexec_apply_relocations(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
-                            unsigned int relsec)
-{
-       pr_err("REL relocation unsupported.\n");
-       return -ENOEXEC;
-}
-
-/*
- * Free up memory used by kernel, initrd, and command line. This is temporary
- * memory allocation which is not needed any more after these buffers have
- * been loaded into separate segments and have been copied elsewhere.
- */
-static void kimage_file_post_load_cleanup(struct kimage *image)
-{
-       struct purgatory_info *pi = &image->purgatory_info;
-
-       vfree(image->kernel_buf);
-       image->kernel_buf = NULL;
-
-       vfree(image->initrd_buf);
-       image->initrd_buf = NULL;
-
-       kfree(image->cmdline_buf);
-       image->cmdline_buf = NULL;
-
-       vfree(pi->purgatory_buf);
-       pi->purgatory_buf = NULL;
-
-       vfree(pi->sechdrs);
-       pi->sechdrs = NULL;
-
-       /* See if architecture has anything to cleanup post load */
-       arch_kimage_file_post_load_cleanup(image);
-
-       /*
-        * Above call should have called into bootloader to free up
-        * any data stored in kimage->image_loader_data. It should
-        * be ok now to free it up.
-        */
-       kfree(image->image_loader_data);
-       image->image_loader_data = NULL;
-}
-
-/*
- * In file mode list of segments is prepared by kernel. Copy relevant
- * data from user space, do error checking, prepare segment list
- */
-static int
-kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,
-                            const char __user *cmdline_ptr,
-                            unsigned long cmdline_len, unsigned flags)
-{
-       int ret = 0;
-       void *ldata;
-
-       ret = copy_file_from_fd(kernel_fd, &image->kernel_buf,
-                               &image->kernel_buf_len);
-       if (ret)
-               return ret;
-
-       /* Call arch image probe handlers */
-       ret = arch_kexec_kernel_image_probe(image, image->kernel_buf,
-                                           image->kernel_buf_len);
-
-       if (ret)
-               goto out;
-
-#ifdef CONFIG_KEXEC_VERIFY_SIG
-       ret = arch_kexec_kernel_verify_sig(image, image->kernel_buf,
-                                          image->kernel_buf_len);
-       if (ret) {
-               pr_debug("kernel signature verification failed.\n");
-               goto out;
-       }
-       pr_debug("kernel signature verification successful.\n");
-#endif
-       /* It is possible that there no initramfs is being loaded */
-       if (!(flags & KEXEC_FILE_NO_INITRAMFS)) {
-               ret = copy_file_from_fd(initrd_fd, &image->initrd_buf,
-                                       &image->initrd_buf_len);
-               if (ret)
-                       goto out;
-       }
-
-       if (cmdline_len) {
-               image->cmdline_buf = kzalloc(cmdline_len, GFP_KERNEL);
-               if (!image->cmdline_buf) {
-                       ret = -ENOMEM;
-                       goto out;
-               }
-
-               ret = copy_from_user(image->cmdline_buf, cmdline_ptr,
-                                    cmdline_len);
-               if (ret) {
-                       ret = -EFAULT;
-                       goto out;
-               }
-
-               image->cmdline_buf_len = cmdline_len;
-
-               /* command line should be a string with last byte null */
-               if (image->cmdline_buf[cmdline_len - 1] != '\0') {
-                       ret = -EINVAL;
-                       goto out;
-               }
-       }
-
-       /* Call arch image load handlers */
-       ldata = arch_kexec_kernel_image_load(image);
-
-       if (IS_ERR(ldata)) {
-               ret = PTR_ERR(ldata);
-               goto out;
-       }
-
-       image->image_loader_data = ldata;
-out:
-       /* In case of error, free up all allocated memory in this function */
-       if (ret)
-               kimage_file_post_load_cleanup(image);
-       return ret;
-}
-
-static int
-kimage_file_alloc_init(struct kimage **rimage, int kernel_fd,
-                      int initrd_fd, const char __user *cmdline_ptr,
-                      unsigned long cmdline_len, unsigned long flags)
-{
-       int ret;
-       struct kimage *image;
-       bool kexec_on_panic = flags & KEXEC_FILE_ON_CRASH;
-
-       image = do_kimage_alloc_init();
-       if (!image)
-               return -ENOMEM;
-
-       image->file_mode = 1;
-
-       if (kexec_on_panic) {
-               /* Enable special crash kernel control page alloc policy. */
-               image->control_page = crashk_res.start;
-               image->type = KEXEC_TYPE_CRASH;
-       }
-
-       ret = kimage_file_prepare_segments(image, kernel_fd, initrd_fd,
-                                          cmdline_ptr, cmdline_len, flags);
-       if (ret)
-               goto out_free_image;
-
-       ret = sanity_check_segment_list(image);
-       if (ret)
-               goto out_free_post_load_bufs;
-
-       ret = -ENOMEM;
-       image->control_code_page = kimage_alloc_control_pages(image,
-                                          get_order(KEXEC_CONTROL_PAGE_SIZE));
-       if (!image->control_code_page) {
-               pr_err("Could not allocate control_code_buffer\n");
-               goto out_free_post_load_bufs;
-       }
-
-       if (!kexec_on_panic) {
-               image->swap_page = kimage_alloc_control_pages(image, 0);
-               if (!image->swap_page) {
-                       pr_err("Could not allocate swap buffer\n");
-                       goto out_free_control_pages;
-               }
-       }
-
-       *rimage = image;
-       return 0;
-out_free_control_pages:
-       kimage_free_page_list(&image->control_pages);
-out_free_post_load_bufs:
-       kimage_file_post_load_cleanup(image);
-out_free_image:
-       kfree(image);
-       return ret;
-}
-#else /* CONFIG_KEXEC_FILE */
-static inline void kimage_file_post_load_cleanup(struct kimage *image) { }
-#endif /* CONFIG_KEXEC_FILE */
-
-static int kimage_is_destination_range(struct kimage *image,
+int kimage_is_destination_range(struct kimage *image,
                                        unsigned long start,
                                        unsigned long end)
 {
@@ -676,7 +389,7 @@ static void kimage_free_pages(struct page *page)
        __free_pages(page, order);
 }
 
-static void kimage_free_page_list(struct list_head *list)
+void kimage_free_page_list(struct list_head *list)
 {
        struct list_head *pos, *next;
 
@@ -892,7 +605,7 @@ static void kimage_free_extra_pages(struct kimage *image)
        kimage_free_page_list(&image->unusable_pages);
 
 }
-static void kimage_terminate(struct kimage *image)
+void kimage_terminate(struct kimage *image)
 {
        if (*image->entry != 0)
                image->entry++;
@@ -913,7 +626,7 @@ static void kimage_free_entry(kimage_entry_t entry)
        kimage_free_pages(page);
 }
 
-static void kimage_free(struct kimage *image)
+void kimage_free(struct kimage *image)
 {
        kimage_entry_t *ptr, entry;
        kimage_entry_t ind = 0;
@@ -1204,7 +917,7 @@ out:
        return result;
 }
 
-static int kimage_load_segment(struct kimage *image,
+int kimage_load_segment(struct kimage *image,
                                struct kexec_segment *segment)
 {
        int result = -ENOMEM;
@@ -1245,8 +958,6 @@ struct kimage *kexec_image;
 struct kimage *kexec_crash_image;
 int kexec_load_disabled;
 
-static DEFINE_MUTEX(kexec_mutex);
-
 SYSCALL_DEFINE4(kexec_load, unsigned long, entry, unsigned long, nr_segments,
                struct kexec_segment __user *, segments, unsigned long, flags)
 {
@@ -1391,85 +1102,6 @@ COMPAT_SYSCALL_DEFINE4(kexec_load, compat_ulong_t, entry,
 }
 #endif
 
-#ifdef CONFIG_KEXEC_FILE
-SYSCALL_DEFINE5(kexec_file_load, int, kernel_fd, int, initrd_fd,
-               unsigned long, cmdline_len, const char __user *, cmdline_ptr,
-               unsigned long, flags)
-{
-       int ret = 0, i;
-       struct kimage **dest_image, *image;
-
-       /* We only trust the superuser with rebooting the system. */
-       if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
-               return -EPERM;
-
-       /* Make sure we have a legal set of flags */
-       if (flags != (flags & KEXEC_FILE_FLAGS))
-               return -EINVAL;
-
-       image = NULL;
-
-       if (!mutex_trylock(&kexec_mutex))
-               return -EBUSY;
-
-       dest_image = &kexec_image;
-       if (flags & KEXEC_FILE_ON_CRASH)
-               dest_image = &kexec_crash_image;
-
-       if (flags & KEXEC_FILE_UNLOAD)
-               goto exchange;
-
-       /*
-        * In case of crash, new kernel gets loaded in reserved region. It is
-        * same memory where old crash kernel might be loaded. Free any
-        * current crash dump kernel before we corrupt it.
-        */
-       if (flags & KEXEC_FILE_ON_CRASH)
-               kimage_free(xchg(&kexec_crash_image, NULL));
-
-       ret = kimage_file_alloc_init(&image, kernel_fd, initrd_fd, cmdline_ptr,
-                                    cmdline_len, flags);
-       if (ret)
-               goto out;
-
-       ret = machine_kexec_prepare(image);
-       if (ret)
-               goto out;
-
-       ret = kexec_calculate_store_digests(image);
-       if (ret)
-               goto out;
-
-       for (i = 0; i < image->nr_segments; i++) {
-               struct kexec_segment *ksegment;
-
-               ksegment = &image->segment[i];
-               pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
-                        i, ksegment->buf, ksegment->bufsz, ksegment->mem,
-                        ksegment->memsz);
-
-               ret = kimage_load_segment(image, &image->segment[i]);
-               if (ret)
-                       goto out;
-       }
-
-       kimage_terminate(image);
-
-       /*
-        * Free up any temporary buffers allocated which are not needed
-        * after image has been loaded
-        */
-       kimage_file_post_load_cleanup(image);
-exchange:
-       image = xchg(dest_image, image);
-out:
-       mutex_unlock(&kexec_mutex);
-       kimage_free(image);
-       return ret;
-}
-
-#endif /* CONFIG_KEXEC_FILE */
-
 void crash_kexec(struct pt_regs *regs)
 {
        /* Take the kexec_mutex here to prevent sys_kexec_load
@@ -2024,672 +1656,6 @@ static int __init crash_save_vmcoreinfo_init(void)
 
 subsys_initcall(crash_save_vmcoreinfo_init);
 
-#ifdef CONFIG_KEXEC_FILE
-static int locate_mem_hole_top_down(unsigned long start, unsigned long end,
-                                   struct kexec_buf *kbuf)
-{
-       struct kimage *image = kbuf->image;
-       unsigned long temp_start, temp_end;
-
-       temp_end = min(end, kbuf->buf_max);
-       temp_start = temp_end - kbuf->memsz;
-
-       do {
-               /* align down start */
-               temp_start = temp_start & (~(kbuf->buf_align - 1));
-
-               if (temp_start < start || temp_start < kbuf->buf_min)
-                       return 0;
-
-               temp_end = temp_start + kbuf->memsz - 1;
-
-               /*
-                * Make sure this does not conflict with any of existing
-                * segments
-                */
-               if (kimage_is_destination_range(image, temp_start, temp_end)) {
-                       temp_start = temp_start - PAGE_SIZE;
-                       continue;
-               }
-
-               /* We found a suitable memory range */
-               break;
-       } while (1);
-
-       /* If we are here, we found a suitable memory range */
-       kbuf->mem = temp_start;
-
-       /* Success, stop navigating through remaining System RAM ranges */
-       return 1;
-}
-
-static int locate_mem_hole_bottom_up(unsigned long start, unsigned long end,
-                                    struct kexec_buf *kbuf)
-{
-       struct kimage *image = kbuf->image;
-       unsigned long temp_start, temp_end;
-
-       temp_start = max(start, kbuf->buf_min);
-
-       do {
-               temp_start = ALIGN(temp_start, kbuf->buf_align);
-               temp_end = temp_start + kbuf->memsz - 1;
-
-               if (temp_end > end || temp_end > kbuf->buf_max)
-                       return 0;
-               /*
-                * Make sure this does not conflict with any of existing
-                * segments
-                */
-               if (kimage_is_destination_range(image, temp_start, temp_end)) {
-                       temp_start = temp_start + PAGE_SIZE;
-                       continue;
-               }
-
-               /* We found a suitable memory range */
-               break;
-       } while (1);
-
-       /* If we are here, we found a suitable memory range */
-       kbuf->mem = temp_start;
-
-       /* Success, stop navigating through remaining System RAM ranges */
-       return 1;
-}
-
-static int locate_mem_hole_callback(u64 start, u64 end, void *arg)
-{
-       struct kexec_buf *kbuf = (struct kexec_buf *)arg;
-       unsigned long sz = end - start + 1;
-
-       /* Returning 0 will take to next memory range */
-       if (sz < kbuf->memsz)
-               return 0;
-
-       if (end < kbuf->buf_min || start > kbuf->buf_max)
-               return 0;
-
-       /*
-        * Allocate memory top down with-in ram range. Otherwise bottom up
-        * allocation.
-        */
-       if (kbuf->top_down)
-               return locate_mem_hole_top_down(start, end, kbuf);
-       return locate_mem_hole_bottom_up(start, end, kbuf);
-}
-
-/*
- * Helper function for placing a buffer in a kexec segment. This assumes
- * that kexec_mutex is held.
- */
-int kexec_add_buffer(struct kimage *image, char *buffer, unsigned long bufsz,
-                    unsigned long memsz, unsigned long buf_align,
-                    unsigned long buf_min, unsigned long buf_max,
-                    bool top_down, unsigned long *load_addr)
-{
-
-       struct kexec_segment *ksegment;
-       struct kexec_buf buf, *kbuf;
-       int ret;
-
-       /* Currently adding segment this way is allowed only in file mode */
-       if (!image->file_mode)
-               return -EINVAL;
-
-       if (image->nr_segments >= KEXEC_SEGMENT_MAX)
-               return -EINVAL;
-
-       /*
-        * Make sure we are not trying to add buffer after allocating
-        * control pages. All segments need to be placed first before
-        * any control pages are allocated. As control page allocation
-        * logic goes through list of segments to make sure there are
-        * no destination overlaps.
-        */
-       if (!list_empty(&image->control_pages)) {
-               WARN_ON(1);
-               return -EINVAL;
-       }
-
-       memset(&buf, 0, sizeof(struct kexec_buf));
-       kbuf = &buf;
-       kbuf->image = image;
-       kbuf->buffer = buffer;
-       kbuf->bufsz = bufsz;
-
-       kbuf->memsz = ALIGN(memsz, PAGE_SIZE);
-       kbuf->buf_align = max(buf_align, PAGE_SIZE);
-       kbuf->buf_min = buf_min;
-       kbuf->buf_max = buf_max;
-       kbuf->top_down = top_down;
-
-       /* Walk the RAM ranges and allocate a suitable range for the buffer */
-       if (image->type == KEXEC_TYPE_CRASH)
-               ret = walk_iomem_res("Crash kernel",
-                                    IORESOURCE_MEM | IORESOURCE_BUSY,
-                                    crashk_res.start, crashk_res.end, kbuf,
-                                    locate_mem_hole_callback);
-       else
-               ret = walk_system_ram_res(0, -1, kbuf,
-                                         locate_mem_hole_callback);
-       if (ret != 1) {
-               /* A suitable memory range could not be found for buffer */
-               return -EADDRNOTAVAIL;
-       }
-
-       /* Found a suitable memory range */
-       ksegment = &image->segment[image->nr_segments];
-       ksegment->kbuf = kbuf->buffer;
-       ksegment->bufsz = kbuf->bufsz;
-       ksegment->mem = kbuf->mem;
-       ksegment->memsz = kbuf->memsz;
-       image->nr_segments++;
-       *load_addr = ksegment->mem;
-       return 0;
-}
-
-/* Calculate and store the digest of segments */
-static int kexec_calculate_store_digests(struct kimage *image)
-{
-       struct crypto_shash *tfm;
-       struct shash_desc *desc;
-       int ret = 0, i, j, zero_buf_sz, sha_region_sz;
-       size_t desc_size, nullsz;
-       char *digest;
-       void *zero_buf;
-       struct kexec_sha_region *sha_regions;
-       struct purgatory_info *pi = &image->purgatory_info;
-
-       zero_buf = __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT);
-       zero_buf_sz = PAGE_SIZE;
-
-       tfm = crypto_alloc_shash("sha256", 0, 0);
-       if (IS_ERR(tfm)) {
-               ret = PTR_ERR(tfm);
-               goto out;
-       }
-
-       desc_size = crypto_shash_descsize(tfm) + sizeof(*desc);
-       desc = kzalloc(desc_size, GFP_KERNEL);
-       if (!desc) {
-               ret = -ENOMEM;
-               goto out_free_tfm;
-       }
-
-       sha_region_sz = KEXEC_SEGMENT_MAX * sizeof(struct kexec_sha_region);
-       sha_regions = vzalloc(sha_region_sz);
-       if (!sha_regions)
-               goto out_free_desc;
-
-       desc->tfm   = tfm;
-       desc->flags = 0;
-
-       ret = crypto_shash_init(desc);
-       if (ret < 0)
-               goto out_free_sha_regions;
-
-       digest = kzalloc(SHA256_DIGEST_SIZE, GFP_KERNEL);
-       if (!digest) {
-               ret = -ENOMEM;
-               goto out_free_sha_regions;
-       }
-
-       for (j = i = 0; i < image->nr_segments; i++) {
-               struct kexec_segment *ksegment;
-
-               ksegment = &image->segment[i];
-               /*
-                * Skip purgatory as it will be modified once we put digest
-                * info in purgatory.
-                */
-               if (ksegment->kbuf == pi->purgatory_buf)
-                       continue;
-
-               ret = crypto_shash_update(desc, ksegment->kbuf,
-                                         ksegment->bufsz);
-               if (ret)
-                       break;
-
-               /*
-                * Assume rest of the buffer is filled with zero and
-                * update digest accordingly.
-                */
-               nullsz = ksegment->memsz - ksegment->bufsz;
-               while (nullsz) {
-                       unsigned long bytes = nullsz;
-
-                       if (bytes > zero_buf_sz)
-                               bytes = zero_buf_sz;
-                       ret = crypto_shash_update(desc, zero_buf, bytes);
-                       if (ret)
-                               break;
-                       nullsz -= bytes;
-               }
-
-               if (ret)
-                       break;
-
-               sha_regions[j].start = ksegment->mem;
-               sha_regions[j].len = ksegment->memsz;
-               j++;
-       }
-
-       if (!ret) {
-               ret = crypto_shash_final(desc, digest);
-               if (ret)
-                       goto out_free_digest;
-               ret = kexec_purgatory_get_set_symbol(image, "sha_regions",
-                                               sha_regions, sha_region_sz, 0);
-               if (ret)
-                       goto out_free_digest;
-
-               ret = kexec_purgatory_get_set_symbol(image, "sha256_digest",
-                                               digest, SHA256_DIGEST_SIZE, 0);
-               if (ret)
-                       goto out_free_digest;
-       }
-
-out_free_digest:
-       kfree(digest);
-out_free_sha_regions:
-       vfree(sha_regions);
-out_free_desc:
-       kfree(desc);
-out_free_tfm:
-       kfree(tfm);
-out:
-       return ret;
-}
-
-/* Actually load purgatory. Lot of code taken from kexec-tools */
-static int __kexec_load_purgatory(struct kimage *image, unsigned long min,
-                                 unsigned long max, int top_down)
-{
-       struct purgatory_info *pi = &image->purgatory_info;
-       unsigned long align, buf_align, bss_align, buf_sz, bss_sz, bss_pad;
-       unsigned long memsz, entry, load_addr, curr_load_addr, bss_addr, offset;
-       unsigned char *buf_addr, *src;
-       int i, ret = 0, entry_sidx = -1;
-       const Elf_Shdr *sechdrs_c;
-       Elf_Shdr *sechdrs = NULL;
-       void *purgatory_buf = NULL;
-
-       /*
-        * sechdrs_c points to section headers in purgatory and are read
-        * only. No modifications allowed.
-        */
-       sechdrs_c = (void *)pi->ehdr + pi->ehdr->e_shoff;
-
-       /*
-        * We can not modify sechdrs_c[] and its fields. It is read only.
-        * Copy it over to a local copy where one can store some temporary
-        * data and free it at the end. We need to modify ->sh_addr and
-        * ->sh_offset fields to keep track of permanent and temporary
-        * locations of sections.
-        */
-       sechdrs = vzalloc(pi->ehdr->e_shnum * sizeof(Elf_Shdr));
-       if (!sechdrs)
-               return -ENOMEM;
-
-       memcpy(sechdrs, sechdrs_c, pi->ehdr->e_shnum * sizeof(Elf_Shdr));
-
-       /*
-        * We seem to have multiple copies of sections. First copy is which
-        * is embedded in kernel in read only section. Some of these sections
-        * will be copied to a temporary buffer and relocated. And these
-        * sections will finally be copied to their final destination at
-        * segment load time.
-        *
-        * Use ->sh_offset to reflect section address in memory. It will
-        * point to original read only copy if section is not allocatable.
-        * Otherwise it will point to temporary copy which will be relocated.
-        *
-        * Use ->sh_addr to contain final address of the section where it
-        * will go during execution time.
-        */
-       for (i = 0; i < pi->ehdr->e_shnum; i++) {
-               if (sechdrs[i].sh_type == SHT_NOBITS)
-                       continue;
-
-               sechdrs[i].sh_offset = (unsigned long)pi->ehdr +
-                                               sechdrs[i].sh_offset;
-       }
-
-       /*
-        * Identify entry point section and make entry relative to section
-        * start.
-        */
-       entry = pi->ehdr->e_entry;
-       for (i = 0; i < pi->ehdr->e_shnum; i++) {
-               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
-                       continue;
-
-               if (!(sechdrs[i].sh_flags & SHF_EXECINSTR))
-                       continue;
-
-               /* Make entry section relative */
-               if (sechdrs[i].sh_addr <= pi->ehdr->e_entry &&
-                   ((sechdrs[i].sh_addr + sechdrs[i].sh_size) >
-                    pi->ehdr->e_entry)) {
-                       entry_sidx = i;
-                       entry -= sechdrs[i].sh_addr;
-                       break;
-               }
-       }
-
-       /* Determine how much memory is needed to load relocatable object. */
-       buf_align = 1;
-       bss_align = 1;
-       buf_sz = 0;
-       bss_sz = 0;
-
-       for (i = 0; i < pi->ehdr->e_shnum; i++) {
-               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
-                       continue;
-
-               align = sechdrs[i].sh_addralign;
-               if (sechdrs[i].sh_type != SHT_NOBITS) {
-                       if (buf_align < align)
-                               buf_align = align;
-                       buf_sz = ALIGN(buf_sz, align);
-                       buf_sz += sechdrs[i].sh_size;
-               } else {
-                       /* bss section */
-                       if (bss_align < align)
-                               bss_align = align;
-                       bss_sz = ALIGN(bss_sz, align);
-                       bss_sz += sechdrs[i].sh_size;
-               }
-       }
-
-       /* Determine the bss padding required to align bss properly */
-       bss_pad = 0;
-       if (buf_sz & (bss_align - 1))
-               bss_pad = bss_align - (buf_sz & (bss_align - 1));
-
-       memsz = buf_sz + bss_pad + bss_sz;
-
-       /* Allocate buffer for purgatory */
-       purgatory_buf = vzalloc(buf_sz);
-       if (!purgatory_buf) {
-               ret = -ENOMEM;
-               goto out;
-       }
-
-       if (buf_align < bss_align)
-               buf_align = bss_align;
-
-       /* Add buffer to segment list */
-       ret = kexec_add_buffer(image, purgatory_buf, buf_sz, memsz,
-                               buf_align, min, max, top_down,
-                               &pi->purgatory_load_addr);
-       if (ret)
-               goto out;
-
-       /* Load SHF_ALLOC sections */
-       buf_addr = purgatory_buf;
-       load_addr = curr_load_addr = pi->purgatory_load_addr;
-       bss_addr = load_addr + buf_sz + bss_pad;
-
-       for (i = 0; i < pi->ehdr->e_shnum; i++) {
-               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
-                       continue;
-
-               align = sechdrs[i].sh_addralign;
-               if (sechdrs[i].sh_type != SHT_NOBITS) {
-                       curr_load_addr = ALIGN(curr_load_addr, align);
-                       offset = curr_load_addr - load_addr;
-                       /* We already modifed ->sh_offset to keep src addr */
-                       src = (char *) sechdrs[i].sh_offset;
-                       memcpy(buf_addr + offset, src, sechdrs[i].sh_size);
-
-                       /* Store load address and source address of section */
-                       sechdrs[i].sh_addr = curr_load_addr;
-
-                       /*
-                        * This section got copied to temporary buffer. Update
-                        * ->sh_offset accordingly.
-                        */
-                       sechdrs[i].sh_offset = (unsigned long)(buf_addr + offset);
-
-                       /* Advance to the next address */
-                       curr_load_addr += sechdrs[i].sh_size;
-               } else {
-                       bss_addr = ALIGN(bss_addr, align);
-                       sechdrs[i].sh_addr = bss_addr;
-                       bss_addr += sechdrs[i].sh_size;
-               }
-       }
-
-       /* Update entry point based on load address of text section */
-       if (entry_sidx >= 0)
-               entry += sechdrs[entry_sidx].sh_addr;
-
-       /* Make kernel jump to purgatory after shutdown */
-       image->start = entry;
-
-       /* Used later to get/set symbol values */
-       pi->sechdrs = sechdrs;
-
-       /*
-        * Used later to identify which section is purgatory and skip it
-        * from checksumming.
-        */
-       pi->purgatory_buf = purgatory_buf;
-       return ret;
-out:
-       vfree(sechdrs);
-       vfree(purgatory_buf);
-       return ret;
-}
-
-static int kexec_apply_relocations(struct kimage *image)
-{
-       int i, ret;
-       struct purgatory_info *pi = &image->purgatory_info;
-       Elf_Shdr *sechdrs = pi->sechdrs;
-
-       /* Apply relocations */
-       for (i = 0; i < pi->ehdr->e_shnum; i++) {
-               Elf_Shdr *section, *symtab;
-
-               if (sechdrs[i].sh_type != SHT_RELA &&
-                   sechdrs[i].sh_type != SHT_REL)
-                       continue;
-
-               /*
-                * For section of type SHT_RELA/SHT_REL,
-                * ->sh_link contains section header index of associated
-                * symbol table. And ->sh_info contains section header
-                * index of section to which relocations apply.
-                */
-               if (sechdrs[i].sh_info >= pi->ehdr->e_shnum ||
-                   sechdrs[i].sh_link >= pi->ehdr->e_shnum)
-                       return -ENOEXEC;
-
-               section = &sechdrs[sechdrs[i].sh_info];
-               symtab = &sechdrs[sechdrs[i].sh_link];
-
-               if (!(section->sh_flags & SHF_ALLOC))
-                       continue;
-
-               /*
-                * symtab->sh_link contain section header index of associated
-                * string table.
-                */
-               if (symtab->sh_link >= pi->ehdr->e_shnum)
-                       /* Invalid section number? */
-                       continue;
-
-               /*
-                * Respective architecture needs to provide support for applying
-                * relocations of type SHT_RELA/SHT_REL.
-                */
-               if (sechdrs[i].sh_type == SHT_RELA)
-                       ret = arch_kexec_apply_relocations_add(pi->ehdr,
-                                                              sechdrs, i);
-               else if (sechdrs[i].sh_type == SHT_REL)
-                       ret = arch_kexec_apply_relocations(pi->ehdr,
-                                                          sechdrs, i);
-               if (ret)
-                       return ret;
-       }
-
-       return 0;
-}
-
-/* Load relocatable purgatory object and relocate it appropriately */
-int kexec_load_purgatory(struct kimage *image, unsigned long min,
-                        unsigned long max, int top_down,
-                        unsigned long *load_addr)
-{
-       struct purgatory_info *pi = &image->purgatory_info;
-       int ret;
-
-       if (kexec_purgatory_size <= 0)
-               return -EINVAL;
-
-       if (kexec_purgatory_size < sizeof(Elf_Ehdr))
-               return -ENOEXEC;
-
-       pi->ehdr = (Elf_Ehdr *)kexec_purgatory;
-
-       if (memcmp(pi->ehdr->e_ident, ELFMAG, SELFMAG) != 0
-           || pi->ehdr->e_type != ET_REL
-           || !elf_check_arch(pi->ehdr)
-           || pi->ehdr->e_shentsize != sizeof(Elf_Shdr))
-               return -ENOEXEC;
-
-       if (pi->ehdr->e_shoff >= kexec_purgatory_size
-           || (pi->ehdr->e_shnum * sizeof(Elf_Shdr) >
-           kexec_purgatory_size - pi->ehdr->e_shoff))
-               return -ENOEXEC;
-
-       ret = __kexec_load_purgatory(image, min, max, top_down);
-       if (ret)
-               return ret;
-
-       ret = kexec_apply_relocations(image);
-       if (ret)
-               goto out;
-
-       *load_addr = pi->purgatory_load_addr;
-       return 0;
-out:
-       vfree(pi->sechdrs);
-       vfree(pi->purgatory_buf);
-       return ret;
-}
-
-static Elf_Sym *kexec_purgatory_find_symbol(struct purgatory_info *pi,
-                                           const char *name)
-{
-       Elf_Sym *syms;
-       Elf_Shdr *sechdrs;
-       Elf_Ehdr *ehdr;
-       int i, k;
-       const char *strtab;
-
-       if (!pi->sechdrs || !pi->ehdr)
-               return NULL;
-
-       sechdrs = pi->sechdrs;
-       ehdr = pi->ehdr;
-
-       for (i = 0; i < ehdr->e_shnum; i++) {
-               if (sechdrs[i].sh_type != SHT_SYMTAB)
-                       continue;
-
-               if (sechdrs[i].sh_link >= ehdr->e_shnum)
-                       /* Invalid strtab section number */
-                       continue;
-               strtab = (char *)sechdrs[sechdrs[i].sh_link].sh_offset;
-               syms = (Elf_Sym *)sechdrs[i].sh_offset;
-
-               /* Go through symbols for a match */
-               for (k = 0; k < sechdrs[i].sh_size/sizeof(Elf_Sym); k++) {
-                       if (ELF_ST_BIND(syms[k].st_info) != STB_GLOBAL)
-                               continue;
-
-                       if (strcmp(strtab + syms[k].st_name, name) != 0)
-                               continue;
-
-                       if (syms[k].st_shndx == SHN_UNDEF ||
-                           syms[k].st_shndx >= ehdr->e_shnum) {
-                               pr_debug("Symbol: %s has bad section index %d.\n",
-                                               name, syms[k].st_shndx);
-                               return NULL;
-                       }
-
-                       /* Found the symbol we are looking for */
-                       return &syms[k];
-               }
-       }
-
-       return NULL;
-}
-
-void *kexec_purgatory_get_symbol_addr(struct kimage *image, const char *name)
-{
-       struct purgatory_info *pi = &image->purgatory_info;
-       Elf_Sym *sym;
-       Elf_Shdr *sechdr;
-
-       sym = kexec_purgatory_find_symbol(pi, name);
-       if (!sym)
-               return ERR_PTR(-EINVAL);
-
-       sechdr = &pi->sechdrs[sym->st_shndx];
-
-       /*
-        * Returns the address where symbol will finally be loaded after
-        * kexec_load_segment()
-        */
-       return (void *)(sechdr->sh_addr + sym->st_value);
-}
-
-/*
- * Get or set value of a symbol. If "get_value" is true, symbol value is
- * returned in buf otherwise symbol value is set based on value in buf.
- */
-int kexec_purgatory_get_set_symbol(struct kimage *image, const char *name,
-                                  void *buf, unsigned int size, bool get_value)
-{
-       Elf_Sym *sym;
-       Elf_Shdr *sechdrs;
-       struct purgatory_info *pi = &image->purgatory_info;
-       char *sym_buf;
-
-       sym = kexec_purgatory_find_symbol(pi, name);
-       if (!sym)
-               return -EINVAL;
-
-       if (sym->st_size != size) {
-               pr_err("symbol %s size mismatch: expected %lu actual %u\n",
-                      name, (unsigned long)sym->st_size, size);
-               return -EINVAL;
-       }
-
-       sechdrs = pi->sechdrs;
-
-       if (sechdrs[sym->st_shndx].sh_type == SHT_NOBITS) {
-               pr_err("symbol %s is in a bss section. Cannot %s\n", name,
-                      get_value ? "get" : "set");
-               return -EINVAL;
-       }
-
-       sym_buf = (unsigned char *)sechdrs[sym->st_shndx].sh_offset +
-                                       sym->st_value;
-
-       if (get_value)
-               memcpy((void *)buf, sym_buf, size);
-       else
-               memcpy((void *)sym_buf, buf, size);
-
-       return 0;
-}
-#endif /* CONFIG_KEXEC_FILE */
-
 /*
  * Move into place and start executing a preloaded standalone
  * executable.  If nothing was preloaded return an error.
diff --git a/kernel/kexec_file.c b/kernel/kexec_file.c
new file mode 100644 (file)
index 0000000..6a9a3f2
--- /dev/null
@@ -0,0 +1,1045 @@
+/*
+ * kexec: kexec_file_load system call
+ *
+ * Copyright (C) 2014 Red Hat Inc.
+ * Authors:
+ *      Vivek Goyal <vgoyal@redhat.com>
+ *
+ * This source code is licensed under the GNU General Public License,
+ * Version 2.  See the file COPYING for more details.
+ */
+
+#include <linux/capability.h>
+#include <linux/mm.h>
+#include <linux/file.h>
+#include <linux/slab.h>
+#include <linux/kexec.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <crypto/hash.h>
+#include <crypto/sha.h>
+#include <linux/syscalls.h>
+#include <linux/vmalloc.h>
+#include "kexec_internal.h"
+
+/*
+ * Declare these symbols weak so that if architecture provides a purgatory,
+ * these will be overridden.
+ */
+char __weak kexec_purgatory[0];
+size_t __weak kexec_purgatory_size = 0;
+
+static int kexec_calculate_store_digests(struct kimage *image);
+
+static int copy_file_from_fd(int fd, void **buf, unsigned long *buf_len)
+{
+       struct fd f = fdget(fd);
+       int ret;
+       struct kstat stat;
+       loff_t pos;
+       ssize_t bytes = 0;
+
+       if (!f.file)
+               return -EBADF;
+
+       ret = vfs_getattr(&f.file->f_path, &stat);
+       if (ret)
+               goto out;
+
+       if (stat.size > INT_MAX) {
+               ret = -EFBIG;
+               goto out;
+       }
+
+       /* Don't hand 0 to vmalloc, it whines. */
+       if (stat.size == 0) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       *buf = vmalloc(stat.size);
+       if (!*buf) {
+               ret = -ENOMEM;
+               goto out;
+       }
+
+       pos = 0;
+       while (pos < stat.size) {
+               bytes = kernel_read(f.file, pos, (char *)(*buf) + pos,
+                                   stat.size - pos);
+               if (bytes < 0) {
+                       vfree(*buf);
+                       ret = bytes;
+                       goto out;
+               }
+
+               if (bytes == 0)
+                       break;
+               pos += bytes;
+       }
+
+       if (pos != stat.size) {
+               ret = -EBADF;
+               vfree(*buf);
+               goto out;
+       }
+
+       *buf_len = pos;
+out:
+       fdput(f);
+       return ret;
+}
+
+/* Architectures can provide this probe function */
+int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,
+                                        unsigned long buf_len)
+{
+       return -ENOEXEC;
+}
+
+void * __weak arch_kexec_kernel_image_load(struct kimage *image)
+{
+       return ERR_PTR(-ENOEXEC);
+}
+
+int __weak arch_kimage_file_post_load_cleanup(struct kimage *image)
+{
+       return -EINVAL;
+}
+
+int __weak arch_kexec_kernel_verify_sig(struct kimage *image, void *buf,
+                                       unsigned long buf_len)
+{
+       return -EKEYREJECTED;
+}
+
+/* Apply relocations of type RELA */
+int __weak
+arch_kexec_apply_relocations_add(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
+                                unsigned int relsec)
+{
+       pr_err("RELA relocation unsupported.\n");
+       return -ENOEXEC;
+}
+
+/* Apply relocations of type REL */
+int __weak
+arch_kexec_apply_relocations(const Elf_Ehdr *ehdr, Elf_Shdr *sechdrs,
+                            unsigned int relsec)
+{
+       pr_err("REL relocation unsupported.\n");
+       return -ENOEXEC;
+}
+
+/*
+ * Free up memory used by kernel, initrd, and command line. This is temporary
+ * memory allocation which is not needed any more after these buffers have
+ * been loaded into separate segments and have been copied elsewhere.
+ */
+void kimage_file_post_load_cleanup(struct kimage *image)
+{
+       struct purgatory_info *pi = &image->purgatory_info;
+
+       vfree(image->kernel_buf);
+       image->kernel_buf = NULL;
+
+       vfree(image->initrd_buf);
+       image->initrd_buf = NULL;
+
+       kfree(image->cmdline_buf);
+       image->cmdline_buf = NULL;
+
+       vfree(pi->purgatory_buf);
+       pi->purgatory_buf = NULL;
+
+       vfree(pi->sechdrs);
+       pi->sechdrs = NULL;
+
+       /* See if architecture has anything to cleanup post load */
+       arch_kimage_file_post_load_cleanup(image);
+
+       /*
+        * Above call should have called into bootloader to free up
+        * any data stored in kimage->image_loader_data. It should
+        * be ok now to free it up.
+        */
+       kfree(image->image_loader_data);
+       image->image_loader_data = NULL;
+}
+
+/*
+ * In file mode list of segments is prepared by kernel. Copy relevant
+ * data from user space, do error checking, prepare segment list
+ */
+static int
+kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,
+                            const char __user *cmdline_ptr,
+                            unsigned long cmdline_len, unsigned flags)
+{
+       int ret = 0;
+       void *ldata;
+
+       ret = copy_file_from_fd(kernel_fd, &image->kernel_buf,
+                               &image->kernel_buf_len);
+       if (ret)
+               return ret;
+
+       /* Call arch image probe handlers */
+       ret = arch_kexec_kernel_image_probe(image, image->kernel_buf,
+                                           image->kernel_buf_len);
+
+       if (ret)
+               goto out;
+
+#ifdef CONFIG_KEXEC_VERIFY_SIG
+       ret = arch_kexec_kernel_verify_sig(image, image->kernel_buf,
+                                          image->kernel_buf_len);
+       if (ret) {
+               pr_debug("kernel signature verification failed.\n");
+               goto out;
+       }
+       pr_debug("kernel signature verification successful.\n");
+#endif
+       /* It is possible that there no initramfs is being loaded */
+       if (!(flags & KEXEC_FILE_NO_INITRAMFS)) {
+               ret = copy_file_from_fd(initrd_fd, &image->initrd_buf,
+                                       &image->initrd_buf_len);
+               if (ret)
+                       goto out;
+       }
+
+       if (cmdline_len) {
+               image->cmdline_buf = kzalloc(cmdline_len, GFP_KERNEL);
+               if (!image->cmdline_buf) {
+                       ret = -ENOMEM;
+                       goto out;
+               }
+
+               ret = copy_from_user(image->cmdline_buf, cmdline_ptr,
+                                    cmdline_len);
+               if (ret) {
+                       ret = -EFAULT;
+                       goto out;
+               }
+
+               image->cmdline_buf_len = cmdline_len;
+
+               /* command line should be a string with last byte null */
+               if (image->cmdline_buf[cmdline_len - 1] != '\0') {
+                       ret = -EINVAL;
+                       goto out;
+               }
+       }
+
+       /* Call arch image load handlers */
+       ldata = arch_kexec_kernel_image_load(image);
+
+       if (IS_ERR(ldata)) {
+               ret = PTR_ERR(ldata);
+               goto out;
+       }
+
+       image->image_loader_data = ldata;
+out:
+       /* In case of error, free up all allocated memory in this function */
+       if (ret)
+               kimage_file_post_load_cleanup(image);
+       return ret;
+}
+
+static int
+kimage_file_alloc_init(struct kimage **rimage, int kernel_fd,
+                      int initrd_fd, const char __user *cmdline_ptr,
+                      unsigned long cmdline_len, unsigned long flags)
+{
+       int ret;
+       struct kimage *image;
+       bool kexec_on_panic = flags & KEXEC_FILE_ON_CRASH;
+
+       image = do_kimage_alloc_init();
+       if (!image)
+               return -ENOMEM;
+
+       image->file_mode = 1;
+
+       if (kexec_on_panic) {
+               /* Enable special crash kernel control page alloc policy. */
+               image->control_page = crashk_res.start;
+               image->type = KEXEC_TYPE_CRASH;
+       }
+
+       ret = kimage_file_prepare_segments(image, kernel_fd, initrd_fd,
+                                          cmdline_ptr, cmdline_len, flags);
+       if (ret)
+               goto out_free_image;
+
+       ret = sanity_check_segment_list(image);
+       if (ret)
+               goto out_free_post_load_bufs;
+
+       ret = -ENOMEM;
+       image->control_code_page = kimage_alloc_control_pages(image,
+                                          get_order(KEXEC_CONTROL_PAGE_SIZE));
+       if (!image->control_code_page) {
+               pr_err("Could not allocate control_code_buffer\n");
+               goto out_free_post_load_bufs;
+       }
+
+       if (!kexec_on_panic) {
+               image->swap_page = kimage_alloc_control_pages(image, 0);
+               if (!image->swap_page) {
+                       pr_err("Could not allocate swap buffer\n");
+                       goto out_free_control_pages;
+               }
+       }
+
+       *rimage = image;
+       return 0;
+out_free_control_pages:
+       kimage_free_page_list(&image->control_pages);
+out_free_post_load_bufs:
+       kimage_file_post_load_cleanup(image);
+out_free_image:
+       kfree(image);
+       return ret;
+}
+
+SYSCALL_DEFINE5(kexec_file_load, int, kernel_fd, int, initrd_fd,
+               unsigned long, cmdline_len, const char __user *, cmdline_ptr,
+               unsigned long, flags)
+{
+       int ret = 0, i;
+       struct kimage **dest_image, *image;
+
+       /* We only trust the superuser with rebooting the system. */
+       if (!capable(CAP_SYS_BOOT) || kexec_load_disabled)
+               return -EPERM;
+
+       /* Make sure we have a legal set of flags */
+       if (flags != (flags & KEXEC_FILE_FLAGS))
+               return -EINVAL;
+
+       image = NULL;
+
+       if (!mutex_trylock(&kexec_mutex))
+               return -EBUSY;
+
+       dest_image = &kexec_image;
+       if (flags & KEXEC_FILE_ON_CRASH)
+               dest_image = &kexec_crash_image;
+
+       if (flags & KEXEC_FILE_UNLOAD)
+               goto exchange;
+
+       /*
+        * In case of crash, new kernel gets loaded in reserved region. It is
+        * same memory where old crash kernel might be loaded. Free any
+        * current crash dump kernel before we corrupt it.
+        */
+       if (flags & KEXEC_FILE_ON_CRASH)
+               kimage_free(xchg(&kexec_crash_image, NULL));
+
+       ret = kimage_file_alloc_init(&image, kernel_fd, initrd_fd, cmdline_ptr,
+                                    cmdline_len, flags);
+       if (ret)
+               goto out;
+
+       ret = machine_kexec_prepare(image);
+       if (ret)
+               goto out;
+
+       ret = kexec_calculate_store_digests(image);
+       if (ret)
+               goto out;
+
+       for (i = 0; i < image->nr_segments; i++) {
+               struct kexec_segment *ksegment;
+
+               ksegment = &image->segment[i];
+               pr_debug("Loading segment %d: buf=0x%p bufsz=0x%zx mem=0x%lx memsz=0x%zx\n",
+                        i, ksegment->buf, ksegment->bufsz, ksegment->mem,
+                        ksegment->memsz);
+
+               ret = kimage_load_segment(image, &image->segment[i]);
+               if (ret)
+                       goto out;
+       }
+
+       kimage_terminate(image);
+
+       /*
+        * Free up any temporary buffers allocated which are not needed
+        * after image has been loaded
+        */
+       kimage_file_post_load_cleanup(image);
+exchange:
+       image = xchg(dest_image, image);
+out:
+       mutex_unlock(&kexec_mutex);
+       kimage_free(image);
+       return ret;
+}
+
+static int locate_mem_hole_top_down(unsigned long start, unsigned long end,
+                                   struct kexec_buf *kbuf)
+{
+       struct kimage *image = kbuf->image;
+       unsigned long temp_start, temp_end;
+
+       temp_end = min(end, kbuf->buf_max);
+       temp_start = temp_end - kbuf->memsz;
+
+       do {
+               /* align down start */
+               temp_start = temp_start & (~(kbuf->buf_align - 1));
+
+               if (temp_start < start || temp_start < kbuf->buf_min)
+                       return 0;
+
+               temp_end = temp_start + kbuf->memsz - 1;
+
+               /*
+                * Make sure this does not conflict with any of existing
+                * segments
+                */
+               if (kimage_is_destination_range(image, temp_start, temp_end)) {
+                       temp_start = temp_start - PAGE_SIZE;
+                       continue;
+               }
+
+               /* We found a suitable memory range */
+               break;
+       } while (1);
+
+       /* If we are here, we found a suitable memory range */
+       kbuf->mem = temp_start;
+
+       /* Success, stop navigating through remaining System RAM ranges */
+       return 1;
+}
+
+static int locate_mem_hole_bottom_up(unsigned long start, unsigned long end,
+                                    struct kexec_buf *kbuf)
+{
+       struct kimage *image = kbuf->image;
+       unsigned long temp_start, temp_end;
+
+       temp_start = max(start, kbuf->buf_min);
+
+       do {
+               temp_start = ALIGN(temp_start, kbuf->buf_align);
+               temp_end = temp_start + kbuf->memsz - 1;
+
+               if (temp_end > end || temp_end > kbuf->buf_max)
+                       return 0;
+               /*
+                * Make sure this does not conflict with any of existing
+                * segments
+                */
+               if (kimage_is_destination_range(image, temp_start, temp_end)) {
+                       temp_start = temp_start + PAGE_SIZE;
+                       continue;
+               }
+
+               /* We found a suitable memory range */
+               break;
+       } while (1);
+
+       /* If we are here, we found a suitable memory range */
+       kbuf->mem = temp_start;
+
+       /* Success, stop navigating through remaining System RAM ranges */
+       return 1;
+}
+
+static int locate_mem_hole_callback(u64 start, u64 end, void *arg)
+{
+       struct kexec_buf *kbuf = (struct kexec_buf *)arg;
+       unsigned long sz = end - start + 1;
+
+       /* Returning 0 will take to next memory range */
+       if (sz < kbuf->memsz)
+               return 0;
+
+       if (end < kbuf->buf_min || start > kbuf->buf_max)
+               return 0;
+
+       /*
+        * Allocate memory top down with-in ram range. Otherwise bottom up
+        * allocation.
+        */
+       if (kbuf->top_down)
+               return locate_mem_hole_top_down(start, end, kbuf);
+       return locate_mem_hole_bottom_up(start, end, kbuf);
+}
+
+/*
+ * Helper function for placing a buffer in a kexec segment. This assumes
+ * that kexec_mutex is held.
+ */
+int kexec_add_buffer(struct kimage *image, char *buffer, unsigned long bufsz,
+                    unsigned long memsz, unsigned long buf_align,
+                    unsigned long buf_min, unsigned long buf_max,
+                    bool top_down, unsigned long *load_addr)
+{
+
+       struct kexec_segment *ksegment;
+       struct kexec_buf buf, *kbuf;
+       int ret;
+
+       /* Currently adding segment this way is allowed only in file mode */
+       if (!image->file_mode)
+               return -EINVAL;
+
+       if (image->nr_segments >= KEXEC_SEGMENT_MAX)
+               return -EINVAL;
+
+       /*
+        * Make sure we are not trying to add buffer after allocating
+        * control pages. All segments need to be placed first before
+        * any control pages are allocated. As control page allocation
+        * logic goes through list of segments to make sure there are
+        * no destination overlaps.
+        */
+       if (!list_empty(&image->control_pages)) {
+               WARN_ON(1);
+               return -EINVAL;
+       }
+
+       memset(&buf, 0, sizeof(struct kexec_buf));
+       kbuf = &buf;
+       kbuf->image = image;
+       kbuf->buffer = buffer;
+       kbuf->bufsz = bufsz;
+
+       kbuf->memsz = ALIGN(memsz, PAGE_SIZE);
+       kbuf->buf_align = max(buf_align, PAGE_SIZE);
+       kbuf->buf_min = buf_min;
+       kbuf->buf_max = buf_max;
+       kbuf->top_down = top_down;
+
+       /* Walk the RAM ranges and allocate a suitable range for the buffer */
+       if (image->type == KEXEC_TYPE_CRASH)
+               ret = walk_iomem_res("Crash kernel",
+                                    IORESOURCE_MEM | IORESOURCE_BUSY,
+                                    crashk_res.start, crashk_res.end, kbuf,
+                                    locate_mem_hole_callback);
+       else
+               ret = walk_system_ram_res(0, -1, kbuf,
+                                         locate_mem_hole_callback);
+       if (ret != 1) {
+               /* A suitable memory range could not be found for buffer */
+               return -EADDRNOTAVAIL;
+       }
+
+       /* Found a suitable memory range */
+       ksegment = &image->segment[image->nr_segments];
+       ksegment->kbuf = kbuf->buffer;
+       ksegment->bufsz = kbuf->bufsz;
+       ksegment->mem = kbuf->mem;
+       ksegment->memsz = kbuf->memsz;
+       image->nr_segments++;
+       *load_addr = ksegment->mem;
+       return 0;
+}
+
+/* Calculate and store the digest of segments */
+static int kexec_calculate_store_digests(struct kimage *image)
+{
+       struct crypto_shash *tfm;
+       struct shash_desc *desc;
+       int ret = 0, i, j, zero_buf_sz, sha_region_sz;
+       size_t desc_size, nullsz;
+       char *digest;
+       void *zero_buf;
+       struct kexec_sha_region *sha_regions;
+       struct purgatory_info *pi = &image->purgatory_info;
+
+       zero_buf = __va(page_to_pfn(ZERO_PAGE(0)) << PAGE_SHIFT);
+       zero_buf_sz = PAGE_SIZE;
+
+       tfm = crypto_alloc_shash("sha256", 0, 0);
+       if (IS_ERR(tfm)) {
+               ret = PTR_ERR(tfm);
+               goto out;
+       }
+
+       desc_size = crypto_shash_descsize(tfm) + sizeof(*desc);
+       desc = kzalloc(desc_size, GFP_KERNEL);
+       if (!desc) {
+               ret = -ENOMEM;
+               goto out_free_tfm;
+       }
+
+       sha_region_sz = KEXEC_SEGMENT_MAX * sizeof(struct kexec_sha_region);
+       sha_regions = vzalloc(sha_region_sz);
+       if (!sha_regions)
+               goto out_free_desc;
+
+       desc->tfm   = tfm;
+       desc->flags = 0;
+
+       ret = crypto_shash_init(desc);
+       if (ret < 0)
+               goto out_free_sha_regions;
+
+       digest = kzalloc(SHA256_DIGEST_SIZE, GFP_KERNEL);
+       if (!digest) {
+               ret = -ENOMEM;
+               goto out_free_sha_regions;
+       }
+
+       for (j = i = 0; i < image->nr_segments; i++) {
+               struct kexec_segment *ksegment;
+
+               ksegment = &image->segment[i];
+               /*
+                * Skip purgatory as it will be modified once we put digest
+                * info in purgatory.
+                */
+               if (ksegment->kbuf == pi->purgatory_buf)
+                       continue;
+
+               ret = crypto_shash_update(desc, ksegment->kbuf,
+                                         ksegment->bufsz);
+               if (ret)
+                       break;
+
+               /*
+                * Assume rest of the buffer is filled with zero and
+                * update digest accordingly.
+                */
+               nullsz = ksegment->memsz - ksegment->bufsz;
+               while (nullsz) {
+                       unsigned long bytes = nullsz;
+
+                       if (bytes > zero_buf_sz)
+                               bytes = zero_buf_sz;
+                       ret = crypto_shash_update(desc, zero_buf, bytes);
+                       if (ret)
+                               break;
+                       nullsz -= bytes;
+               }
+
+               if (ret)
+                       break;
+
+               sha_regions[j].start = ksegment->mem;
+               sha_regions[j].len = ksegment->memsz;
+               j++;
+       }
+
+       if (!ret) {
+               ret = crypto_shash_final(desc, digest);
+               if (ret)
+                       goto out_free_digest;
+               ret = kexec_purgatory_get_set_symbol(image, "sha_regions",
+                                               sha_regions, sha_region_sz, 0);
+               if (ret)
+                       goto out_free_digest;
+
+               ret = kexec_purgatory_get_set_symbol(image, "sha256_digest",
+                                               digest, SHA256_DIGEST_SIZE, 0);
+               if (ret)
+                       goto out_free_digest;
+       }
+
+out_free_digest:
+       kfree(digest);
+out_free_sha_regions:
+       vfree(sha_regions);
+out_free_desc:
+       kfree(desc);
+out_free_tfm:
+       kfree(tfm);
+out:
+       return ret;
+}
+
+/* Actually load purgatory. Lot of code taken from kexec-tools */
+static int __kexec_load_purgatory(struct kimage *image, unsigned long min,
+                                 unsigned long max, int top_down)
+{
+       struct purgatory_info *pi = &image->purgatory_info;
+       unsigned long align, buf_align, bss_align, buf_sz, bss_sz, bss_pad;
+       unsigned long memsz, entry, load_addr, curr_load_addr, bss_addr, offset;
+       unsigned char *buf_addr, *src;
+       int i, ret = 0, entry_sidx = -1;
+       const Elf_Shdr *sechdrs_c;
+       Elf_Shdr *sechdrs = NULL;
+       void *purgatory_buf = NULL;
+
+       /*
+        * sechdrs_c points to section headers in purgatory and are read
+        * only. No modifications allowed.
+        */
+       sechdrs_c = (void *)pi->ehdr + pi->ehdr->e_shoff;
+
+       /*
+        * We can not modify sechdrs_c[] and its fields. It is read only.
+        * Copy it over to a local copy where one can store some temporary
+        * data and free it at the end. We need to modify ->sh_addr and
+        * ->sh_offset fields to keep track of permanent and temporary
+        * locations of sections.
+        */
+       sechdrs = vzalloc(pi->ehdr->e_shnum * sizeof(Elf_Shdr));
+       if (!sechdrs)
+               return -ENOMEM;
+
+       memcpy(sechdrs, sechdrs_c, pi->ehdr->e_shnum * sizeof(Elf_Shdr));
+
+       /*
+        * We seem to have multiple copies of sections. First copy is which
+        * is embedded in kernel in read only section. Some of these sections
+        * will be copied to a temporary buffer and relocated. And these
+        * sections will finally be copied to their final destination at
+        * segment load time.
+        *
+        * Use ->sh_offset to reflect section address in memory. It will
+        * point to original read only copy if section is not allocatable.
+        * Otherwise it will point to temporary copy which will be relocated.
+        *
+        * Use ->sh_addr to contain final address of the section where it
+        * will go during execution time.
+        */
+       for (i = 0; i < pi->ehdr->e_shnum; i++) {
+               if (sechdrs[i].sh_type == SHT_NOBITS)
+                       continue;
+
+               sechdrs[i].sh_offset = (unsigned long)pi->ehdr +
+                                               sechdrs[i].sh_offset;
+       }
+
+       /*
+        * Identify entry point section and make entry relative to section
+        * start.
+        */
+       entry = pi->ehdr->e_entry;
+       for (i = 0; i < pi->ehdr->e_shnum; i++) {
+               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
+                       continue;
+
+               if (!(sechdrs[i].sh_flags & SHF_EXECINSTR))
+                       continue;
+
+               /* Make entry section relative */
+               if (sechdrs[i].sh_addr <= pi->ehdr->e_entry &&
+                   ((sechdrs[i].sh_addr + sechdrs[i].sh_size) >
+                    pi->ehdr->e_entry)) {
+                       entry_sidx = i;
+                       entry -= sechdrs[i].sh_addr;
+                       break;
+               }
+       }
+
+       /* Determine how much memory is needed to load relocatable object. */
+       buf_align = 1;
+       bss_align = 1;
+       buf_sz = 0;
+       bss_sz = 0;
+
+       for (i = 0; i < pi->ehdr->e_shnum; i++) {
+               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
+                       continue;
+
+               align = sechdrs[i].sh_addralign;
+               if (sechdrs[i].sh_type != SHT_NOBITS) {
+                       if (buf_align < align)
+                               buf_align = align;
+                       buf_sz = ALIGN(buf_sz, align);
+                       buf_sz += sechdrs[i].sh_size;
+               } else {
+                       /* bss section */
+                       if (bss_align < align)
+                               bss_align = align;
+                       bss_sz = ALIGN(bss_sz, align);
+                       bss_sz += sechdrs[i].sh_size;
+               }
+       }
+
+       /* Determine the bss padding required to align bss properly */
+       bss_pad = 0;
+       if (buf_sz & (bss_align - 1))
+               bss_pad = bss_align - (buf_sz & (bss_align - 1));
+
+       memsz = buf_sz + bss_pad + bss_sz;
+
+       /* Allocate buffer for purgatory */
+       purgatory_buf = vzalloc(buf_sz);
+       if (!purgatory_buf) {
+               ret = -ENOMEM;
+               goto out;
+       }
+
+       if (buf_align < bss_align)
+               buf_align = bss_align;
+
+       /* Add buffer to segment list */
+       ret = kexec_add_buffer(image, purgatory_buf, buf_sz, memsz,
+                               buf_align, min, max, top_down,
+                               &pi->purgatory_load_addr);
+       if (ret)
+               goto out;
+
+       /* Load SHF_ALLOC sections */
+       buf_addr = purgatory_buf;
+       load_addr = curr_load_addr = pi->purgatory_load_addr;
+       bss_addr = load_addr + buf_sz + bss_pad;
+
+       for (i = 0; i < pi->ehdr->e_shnum; i++) {
+               if (!(sechdrs[i].sh_flags & SHF_ALLOC))
+                       continue;
+
+               align = sechdrs[i].sh_addralign;
+               if (sechdrs[i].sh_type != SHT_NOBITS) {
+                       curr_load_addr = ALIGN(curr_load_addr, align);
+                       offset = curr_load_addr - load_addr;
+                       /* We already modifed ->sh_offset to keep src addr */
+                       src = (char *) sechdrs[i].sh_offset;
+                       memcpy(buf_addr + offset, src, sechdrs[i].sh_size);
+
+                       /* Store load address and source address of section */
+                       sechdrs[i].sh_addr = curr_load_addr;
+
+                       /*
+                        * This section got copied to temporary buffer. Update
+                        * ->sh_offset accordingly.
+                        */
+                       sechdrs[i].sh_offset = (unsigned long)(buf_addr + offset);
+
+                       /* Advance to the next address */
+                       curr_load_addr += sechdrs[i].sh_size;
+               } else {
+                       bss_addr = ALIGN(bss_addr, align);
+                       sechdrs[i].sh_addr = bss_addr;
+                       bss_addr += sechdrs[i].sh_size;
+               }
+       }
+
+       /* Update entry point based on load address of text section */
+       if (entry_sidx >= 0)
+               entry += sechdrs[entry_sidx].sh_addr;
+
+       /* Make kernel jump to purgatory after shutdown */
+       image->start = entry;
+
+       /* Used later to get/set symbol values */
+       pi->sechdrs = sechdrs;
+
+       /*
+        * Used later to identify which section is purgatory and skip it
+        * from checksumming.
+        */
+       pi->purgatory_buf = purgatory_buf;
+       return ret;
+out:
+       vfree(sechdrs);
+       vfree(purgatory_buf);
+       return ret;
+}
+
+static int kexec_apply_relocations(struct kimage *image)
+{
+       int i, ret;
+       struct purgatory_info *pi = &image->purgatory_info;
+       Elf_Shdr *sechdrs = pi->sechdrs;
+
+       /* Apply relocations */
+       for (i = 0; i < pi->ehdr->e_shnum; i++) {
+               Elf_Shdr *section, *symtab;
+
+               if (sechdrs[i].sh_type != SHT_RELA &&
+                   sechdrs[i].sh_type != SHT_REL)
+                       continue;
+
+               /*
+                * For section of type SHT_RELA/SHT_REL,
+                * ->sh_link contains section header index of associated
+                * symbol table. And ->sh_info contains section header
+                * index of section to which relocations apply.
+                */
+               if (sechdrs[i].sh_info >= pi->ehdr->e_shnum ||
+                   sechdrs[i].sh_link >= pi->ehdr->e_shnum)
+                       return -ENOEXEC;
+
+               section = &sechdrs[sechdrs[i].sh_info];
+               symtab = &sechdrs[sechdrs[i].sh_link];
+
+               if (!(section->sh_flags & SHF_ALLOC))
+                       continue;
+
+               /*
+                * symtab->sh_link contain section header index of associated
+                * string table.
+                */
+               if (symtab->sh_link >= pi->ehdr->e_shnum)
+                       /* Invalid section number? */
+                       continue;
+
+               /*
+                * Respective architecture needs to provide support for applying
+                * relocations of type SHT_RELA/SHT_REL.
+                */
+               if (sechdrs[i].sh_type == SHT_RELA)
+                       ret = arch_kexec_apply_relocations_add(pi->ehdr,
+                                                              sechdrs, i);
+               else if (sechdrs[i].sh_type == SHT_REL)
+                       ret = arch_kexec_apply_relocations(pi->ehdr,
+                                                          sechdrs, i);
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
+}
+
+/* Load relocatable purgatory object and relocate it appropriately */
+int kexec_load_purgatory(struct kimage *image, unsigned long min,
+                        unsigned long max, int top_down,
+                        unsigned long *load_addr)
+{
+       struct purgatory_info *pi = &image->purgatory_info;
+       int ret;
+
+       if (kexec_purgatory_size <= 0)
+               return -EINVAL;
+
+       if (kexec_purgatory_size < sizeof(Elf_Ehdr))
+               return -ENOEXEC;
+
+       pi->ehdr = (Elf_Ehdr *)kexec_purgatory;
+
+       if (memcmp(pi->ehdr->e_ident, ELFMAG, SELFMAG) != 0
+           || pi->ehdr->e_type != ET_REL
+           || !elf_check_arch(pi->ehdr)
+           || pi->ehdr->e_shentsize != sizeof(Elf_Shdr))
+               return -ENOEXEC;
+
+       if (pi->ehdr->e_shoff >= kexec_purgatory_size
+           || (pi->ehdr->e_shnum * sizeof(Elf_Shdr) >
+           kexec_purgatory_size - pi->ehdr->e_shoff))
+               return -ENOEXEC;
+
+       ret = __kexec_load_purgatory(image, min, max, top_down);
+       if (ret)
+               return ret;
+
+       ret = kexec_apply_relocations(image);
+       if (ret)
+               goto out;
+
+       *load_addr = pi->purgatory_load_addr;
+       return 0;
+out:
+       vfree(pi->sechdrs);
+       vfree(pi->purgatory_buf);
+       return ret;
+}
+
+static Elf_Sym *kexec_purgatory_find_symbol(struct purgatory_info *pi,
+                                           const char *name)
+{
+       Elf_Sym *syms;
+       Elf_Shdr *sechdrs;
+       Elf_Ehdr *ehdr;
+       int i, k;
+       const char *strtab;
+
+       if (!pi->sechdrs || !pi->ehdr)
+               return NULL;
+
+       sechdrs = pi->sechdrs;
+       ehdr = pi->ehdr;
+
+       for (i = 0; i < ehdr->e_shnum; i++) {
+               if (sechdrs[i].sh_type != SHT_SYMTAB)
+                       continue;
+
+               if (sechdrs[i].sh_link >= ehdr->e_shnum)
+                       /* Invalid strtab section number */
+                       continue;
+               strtab = (char *)sechdrs[sechdrs[i].sh_link].sh_offset;
+               syms = (Elf_Sym *)sechdrs[i].sh_offset;
+
+               /* Go through symbols for a match */
+               for (k = 0; k < sechdrs[i].sh_size/sizeof(Elf_Sym); k++) {
+                       if (ELF_ST_BIND(syms[k].st_info) != STB_GLOBAL)
+                               continue;
+
+                       if (strcmp(strtab + syms[k].st_name, name) != 0)
+                               continue;
+
+                       if (syms[k].st_shndx == SHN_UNDEF ||
+                           syms[k].st_shndx >= ehdr->e_shnum) {
+                               pr_debug("Symbol: %s has bad section index %d.\n",
+                                               name, syms[k].st_shndx);
+                               return NULL;
+                       }
+
+                       /* Found the symbol we are looking for */
+                       return &syms[k];
+               }
+       }
+
+       return NULL;
+}
+
+void *kexec_purgatory_get_symbol_addr(struct kimage *image, const char *name)
+{
+       struct purgatory_info *pi = &image->purgatory_info;
+       Elf_Sym *sym;
+       Elf_Shdr *sechdr;
+
+       sym = kexec_purgatory_find_symbol(pi, name);
+       if (!sym)
+               return ERR_PTR(-EINVAL);
+
+       sechdr = &pi->sechdrs[sym->st_shndx];
+
+       /*
+        * Returns the address where symbol will finally be loaded after
+        * kexec_load_segment()
+        */
+       return (void *)(sechdr->sh_addr + sym->st_value);
+}
+
+/*
+ * Get or set value of a symbol. If "get_value" is true, symbol value is
+ * returned in buf otherwise symbol value is set based on value in buf.
+ */
+int kexec_purgatory_get_set_symbol(struct kimage *image, const char *name,
+                                  void *buf, unsigned int size, bool get_value)
+{
+       Elf_Sym *sym;
+       Elf_Shdr *sechdrs;
+       struct purgatory_info *pi = &image->purgatory_info;
+       char *sym_buf;
+
+       sym = kexec_purgatory_find_symbol(pi, name);
+       if (!sym)
+               return -EINVAL;
+
+       if (sym->st_size != size) {
+               pr_err("symbol %s size mismatch: expected %lu actual %u\n",
+                      name, (unsigned long)sym->st_size, size);
+               return -EINVAL;
+       }
+
+       sechdrs = pi->sechdrs;
+
+       if (sechdrs[sym->st_shndx].sh_type == SHT_NOBITS) {
+               pr_err("symbol %s is in a bss section. Cannot %s\n", name,
+                      get_value ? "get" : "set");
+               return -EINVAL;
+       }
+
+       sym_buf = (unsigned char *)sechdrs[sym->st_shndx].sh_offset +
+                                       sym->st_value;
+
+       if (get_value)
+               memcpy((void *)buf, sym_buf, size);
+       else
+               memcpy((void *)sym_buf, buf, size);
+
+       return 0;
+}
diff --git a/kernel/kexec_internal.h b/kernel/kexec_internal.h
new file mode 100644 (file)
index 0000000..e4392a6
--- /dev/null
@@ -0,0 +1,22 @@
+#ifndef LINUX_KEXEC_INTERNAL_H
+#define LINUX_KEXEC_INTERNAL_H
+
+#include <linux/kexec.h>
+
+struct kimage *do_kimage_alloc_init(void);
+int sanity_check_segment_list(struct kimage *image);
+void kimage_free_page_list(struct list_head *list);
+void kimage_free(struct kimage *image);
+int kimage_load_segment(struct kimage *image, struct kexec_segment *segment);
+void kimage_terminate(struct kimage *image);
+int kimage_is_destination_range(struct kimage *image,
+                               unsigned long start, unsigned long end);
+
+extern struct mutex kexec_mutex;
+
+#ifdef CONFIG_KEXEC_FILE
+void kimage_file_post_load_cleanup(struct kimage *image);
+#else /* CONFIG_KEXEC_FILE */
+static inline void kimage_file_post_load_cleanup(struct kimage *image) { }
+#endif /* CONFIG_KEXEC_FILE */
+#endif /* LINUX_KEXEC_INTERNAL_H */