perf: Add Linux perf's symbol-elf.c
authorBarret Rhoden <brho@cs.berkeley.edu>
Wed, 25 May 2016 20:56:07 +0000 (16:56 -0400)
committerBarret Rhoden <brho@cs.berkeley.edu>
Thu, 16 Jun 2016 15:48:38 +0000 (11:48 -0400)
Unmodified other than a CR header, from commit c13c81006314 ("Merge tag
'rtc-v4.2-1' of
git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux")

Signed-off-by: Barret Rhoden <brho@cs.berkeley.edu>
tools/profile/perf/symbol-elf.c [new file with mode: 0644]

diff --git a/tools/profile/perf/symbol-elf.c b/tools/profile/perf/symbol-elf.c
new file mode 100644 (file)
index 0000000..5d8c614
--- /dev/null
@@ -0,0 +1,1795 @@
+/* Copyright (c) 2011-2015 Linux Perf Authors
+ *
+ * This source code is licensed under the GNU General Public License Version 2.
+ * See the file LICENSE-gpl-2.0.txt for more details. */
+
+#include <fcntl.h>
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <unistd.h>
+#include <inttypes.h>
+
+#include "symbol.h"
+#include "machine.h"
+#include "vdso.h"
+#include <symbol/kallsyms.h>
+#include "debug.h"
+
+#ifndef EM_AARCH64
+#define EM_AARCH64     183  /* ARM 64 bit */
+#endif
+
+
+#ifdef HAVE_CPLUS_DEMANGLE_SUPPORT
+extern char *cplus_demangle(const char *, int);
+
+static inline char *bfd_demangle(void __maybe_unused *v, const char *c, int i)
+{
+       return cplus_demangle(c, i);
+}
+#else
+#ifdef NO_DEMANGLE
+static inline char *bfd_demangle(void __maybe_unused *v,
+                                const char __maybe_unused *c,
+                                int __maybe_unused i)
+{
+       return NULL;
+}
+#else
+#define PACKAGE 'perf'
+#include <bfd.h>
+#endif
+#endif
+
+#ifndef HAVE_ELF_GETPHDRNUM_SUPPORT
+static int elf_getphdrnum(Elf *elf, size_t *dst)
+{
+       GElf_Ehdr gehdr;
+       GElf_Ehdr *ehdr;
+
+       ehdr = gelf_getehdr(elf, &gehdr);
+       if (!ehdr)
+               return -1;
+
+       *dst = ehdr->e_phnum;
+
+       return 0;
+}
+#endif
+
+#ifndef NT_GNU_BUILD_ID
+#define NT_GNU_BUILD_ID 3
+#endif
+
+/**
+ * elf_symtab__for_each_symbol - iterate thru all the symbols
+ *
+ * @syms: struct elf_symtab instance to iterate
+ * @idx: uint32_t idx
+ * @sym: GElf_Sym iterator
+ */
+#define elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) \
+       for (idx = 0, gelf_getsym(syms, idx, &sym);\
+            idx < nr_syms; \
+            idx++, gelf_getsym(syms, idx, &sym))
+
+static inline uint8_t elf_sym__type(const GElf_Sym *sym)
+{
+       return GELF_ST_TYPE(sym->st_info);
+}
+
+#ifndef STT_GNU_IFUNC
+#define STT_GNU_IFUNC 10
+#endif
+
+static inline int elf_sym__is_function(const GElf_Sym *sym)
+{
+       return (elf_sym__type(sym) == STT_FUNC ||
+               elf_sym__type(sym) == STT_GNU_IFUNC) &&
+              sym->st_name != 0 &&
+              sym->st_shndx != SHN_UNDEF;
+}
+
+static inline bool elf_sym__is_object(const GElf_Sym *sym)
+{
+       return elf_sym__type(sym) == STT_OBJECT &&
+               sym->st_name != 0 &&
+               sym->st_shndx != SHN_UNDEF;
+}
+
+static inline int elf_sym__is_label(const GElf_Sym *sym)
+{
+       return elf_sym__type(sym) == STT_NOTYPE &&
+               sym->st_name != 0 &&
+               sym->st_shndx != SHN_UNDEF &&
+               sym->st_shndx != SHN_ABS;
+}
+
+static bool elf_sym__is_a(GElf_Sym *sym, enum map_type type)
+{
+       switch (type) {
+       case MAP__FUNCTION:
+               return elf_sym__is_function(sym);
+       case MAP__VARIABLE:
+               return elf_sym__is_object(sym);
+       default:
+               return false;
+       }
+}
+
+static inline const char *elf_sym__name(const GElf_Sym *sym,
+                                       const Elf_Data *symstrs)
+{
+       return symstrs->d_buf + sym->st_name;
+}
+
+static inline const char *elf_sec__name(const GElf_Shdr *shdr,
+                                       const Elf_Data *secstrs)
+{
+       return secstrs->d_buf + shdr->sh_name;
+}
+
+static inline int elf_sec__is_text(const GElf_Shdr *shdr,
+                                       const Elf_Data *secstrs)
+{
+       return strstr(elf_sec__name(shdr, secstrs), "text") != NULL;
+}
+
+static inline bool elf_sec__is_data(const GElf_Shdr *shdr,
+                                   const Elf_Data *secstrs)
+{
+       return strstr(elf_sec__name(shdr, secstrs), "data") != NULL;
+}
+
+static bool elf_sec__is_a(GElf_Shdr *shdr, Elf_Data *secstrs,
+                         enum map_type type)
+{
+       switch (type) {
+       case MAP__FUNCTION:
+               return elf_sec__is_text(shdr, secstrs);
+       case MAP__VARIABLE:
+               return elf_sec__is_data(shdr, secstrs);
+       default:
+               return false;
+       }
+}
+
+static size_t elf_addr_to_index(Elf *elf, GElf_Addr addr)
+{
+       Elf_Scn *sec = NULL;
+       GElf_Shdr shdr;
+       size_t cnt = 1;
+
+       while ((sec = elf_nextscn(elf, sec)) != NULL) {
+               gelf_getshdr(sec, &shdr);
+
+               if ((addr >= shdr.sh_addr) &&
+                   (addr < (shdr.sh_addr + shdr.sh_size)))
+                       return cnt;
+
+               ++cnt;
+       }
+
+       return -1;
+}
+
+Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep,
+                            GElf_Shdr *shp, const char *name, size_t *idx)
+{
+       Elf_Scn *sec = NULL;
+       size_t cnt = 1;
+
+       /* Elf is corrupted/truncated, avoid calling elf_strptr. */
+       if (!elf_rawdata(elf_getscn(elf, ep->e_shstrndx), NULL))
+               return NULL;
+
+       while ((sec = elf_nextscn(elf, sec)) != NULL) {
+               char *str;
+
+               gelf_getshdr(sec, shp);
+               str = elf_strptr(elf, ep->e_shstrndx, shp->sh_name);
+               if (str && !strcmp(name, str)) {
+                       if (idx)
+                               *idx = cnt;
+                       return sec;
+               }
+               ++cnt;
+       }
+
+       return NULL;
+}
+
+#define elf_section__for_each_rel(reldata, pos, pos_mem, idx, nr_entries) \
+       for (idx = 0, pos = gelf_getrel(reldata, 0, &pos_mem); \
+            idx < nr_entries; \
+            ++idx, pos = gelf_getrel(reldata, idx, &pos_mem))
+
+#define elf_section__for_each_rela(reldata, pos, pos_mem, idx, nr_entries) \
+       for (idx = 0, pos = gelf_getrela(reldata, 0, &pos_mem); \
+            idx < nr_entries; \
+            ++idx, pos = gelf_getrela(reldata, idx, &pos_mem))
+
+/*
+ * We need to check if we have a .dynsym, so that we can handle the
+ * .plt, synthesizing its symbols, that aren't on the symtabs (be it
+ * .dynsym or .symtab).
+ * And always look at the original dso, not at debuginfo packages, that
+ * have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS).
+ */
+int dso__synthesize_plt_symbols(struct dso *dso, struct symsrc *ss, struct map *map,
+                               symbol_filter_t filter)
+{
+       uint32_t nr_rel_entries, idx;
+       GElf_Sym sym;
+       u64 plt_offset;
+       GElf_Shdr shdr_plt;
+       struct symbol *f;
+       GElf_Shdr shdr_rel_plt, shdr_dynsym;
+       Elf_Data *reldata, *syms, *symstrs;
+       Elf_Scn *scn_plt_rel, *scn_symstrs, *scn_dynsym;
+       size_t dynsym_idx;
+       GElf_Ehdr ehdr;
+       char sympltname[1024];
+       Elf *elf;
+       int nr = 0, symidx, err = 0;
+
+       if (!ss->dynsym)
+               return 0;
+
+       elf = ss->elf;
+       ehdr = ss->ehdr;
+
+       scn_dynsym = ss->dynsym;
+       shdr_dynsym = ss->dynshdr;
+       dynsym_idx = ss->dynsym_idx;
+
+       if (scn_dynsym == NULL)
+               goto out_elf_end;
+
+       scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
+                                         ".rela.plt", NULL);
+       if (scn_plt_rel == NULL) {
+               scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
+                                                 ".rel.plt", NULL);
+               if (scn_plt_rel == NULL)
+                       goto out_elf_end;
+       }
+
+       err = -1;
+
+       if (shdr_rel_plt.sh_link != dynsym_idx)
+               goto out_elf_end;
+
+       if (elf_section_by_name(elf, &ehdr, &shdr_plt, ".plt", NULL) == NULL)
+               goto out_elf_end;
+
+       /*
+        * Fetch the relocation section to find the idxes to the GOT
+        * and the symbols in the .dynsym they refer to.
+        */
+       reldata = elf_getdata(scn_plt_rel, NULL);
+       if (reldata == NULL)
+               goto out_elf_end;
+
+       syms = elf_getdata(scn_dynsym, NULL);
+       if (syms == NULL)
+               goto out_elf_end;
+
+       scn_symstrs = elf_getscn(elf, shdr_dynsym.sh_link);
+       if (scn_symstrs == NULL)
+               goto out_elf_end;
+
+       symstrs = elf_getdata(scn_symstrs, NULL);
+       if (symstrs == NULL)
+               goto out_elf_end;
+
+       if (symstrs->d_size == 0)
+               goto out_elf_end;
+
+       nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
+       plt_offset = shdr_plt.sh_offset;
+
+       if (shdr_rel_plt.sh_type == SHT_RELA) {
+               GElf_Rela pos_mem, *pos;
+
+               elf_section__for_each_rela(reldata, pos, pos_mem, idx,
+                                          nr_rel_entries) {
+                       symidx = GELF_R_SYM(pos->r_info);
+                       plt_offset += shdr_plt.sh_entsize;
+                       gelf_getsym(syms, symidx, &sym);
+                       snprintf(sympltname, sizeof(sympltname),
+                                "%s@plt", elf_sym__name(&sym, symstrs));
+
+                       f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+                                       STB_GLOBAL, sympltname);
+                       if (!f)
+                               goto out_elf_end;
+
+                       if (filter && filter(map, f))
+                               symbol__delete(f);
+                       else {
+                               symbols__insert(&dso->symbols[map->type], f);
+                               ++nr;
+                       }
+               }
+       } else if (shdr_rel_plt.sh_type == SHT_REL) {
+               GElf_Rel pos_mem, *pos;
+               elf_section__for_each_rel(reldata, pos, pos_mem, idx,
+                                         nr_rel_entries) {
+                       symidx = GELF_R_SYM(pos->r_info);
+                       plt_offset += shdr_plt.sh_entsize;
+                       gelf_getsym(syms, symidx, &sym);
+                       snprintf(sympltname, sizeof(sympltname),
+                                "%s@plt", elf_sym__name(&sym, symstrs));
+
+                       f = symbol__new(plt_offset, shdr_plt.sh_entsize,
+                                       STB_GLOBAL, sympltname);
+                       if (!f)
+                               goto out_elf_end;
+
+                       if (filter && filter(map, f))
+                               symbol__delete(f);
+                       else {
+                               symbols__insert(&dso->symbols[map->type], f);
+                               ++nr;
+                       }
+               }
+       }
+
+       err = 0;
+out_elf_end:
+       if (err == 0)
+               return nr;
+       pr_debug("%s: problems reading %s PLT info.\n",
+                __func__, dso->long_name);
+       return 0;
+}
+
+/*
+ * Align offset to 4 bytes as needed for note name and descriptor data.
+ */
+#define NOTE_ALIGN(n) (((n) + 3) & -4U)
+
+static int elf_read_build_id(Elf *elf, void *bf, size_t size)
+{
+       int err = -1;
+       GElf_Ehdr ehdr;
+       GElf_Shdr shdr;
+       Elf_Data *data;
+       Elf_Scn *sec;
+       Elf_Kind ek;
+       void *ptr;
+
+       if (size < BUILD_ID_SIZE)
+               goto out;
+
+       ek = elf_kind(elf);
+       if (ek != ELF_K_ELF)
+               goto out;
+
+       if (gelf_getehdr(elf, &ehdr) == NULL) {
+               pr_err("%s: cannot get elf header.\n", __func__);
+               goto out;
+       }
+
+       /*
+        * Check following sections for notes:
+        *   '.note.gnu.build-id'
+        *   '.notes'
+        *   '.note' (VDSO specific)
+        */
+       do {
+               sec = elf_section_by_name(elf, &ehdr, &shdr,
+                                         ".note.gnu.build-id", NULL);
+               if (sec)
+                       break;
+
+               sec = elf_section_by_name(elf, &ehdr, &shdr,
+                                         ".notes", NULL);
+               if (sec)
+                       break;
+
+               sec = elf_section_by_name(elf, &ehdr, &shdr,
+                                         ".note", NULL);
+               if (sec)
+                       break;
+
+               return err;
+
+       } while (0);
+
+       data = elf_getdata(sec, NULL);
+       if (data == NULL)
+               goto out;
+
+       ptr = data->d_buf;
+       while (ptr < (data->d_buf + data->d_size)) {
+               GElf_Nhdr *nhdr = ptr;
+               size_t namesz = NOTE_ALIGN(nhdr->n_namesz),
+                      descsz = NOTE_ALIGN(nhdr->n_descsz);
+               const char *name;
+
+               ptr += sizeof(*nhdr);
+               name = ptr;
+               ptr += namesz;
+               if (nhdr->n_type == NT_GNU_BUILD_ID &&
+                   nhdr->n_namesz == sizeof("GNU")) {
+                       if (memcmp(name, "GNU", sizeof("GNU")) == 0) {
+                               size_t sz = min(size, descsz);
+                               memcpy(bf, ptr, sz);
+                               memset(bf + sz, 0, size - sz);
+                               err = descsz;
+                               break;
+                       }
+               }
+               ptr += descsz;
+       }
+
+out:
+       return err;
+}
+
+int filename__read_build_id(const char *filename, void *bf, size_t size)
+{
+       int fd, err = -1;
+       Elf *elf;
+
+       if (size < BUILD_ID_SIZE)
+               goto out;
+
+       fd = open(filename, O_RDONLY);
+       if (fd < 0)
+               goto out;
+
+       elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
+       if (elf == NULL) {
+               pr_debug2("%s: cannot read %s ELF file.\n", __func__, filename);
+               goto out_close;
+       }
+
+       err = elf_read_build_id(elf, bf, size);
+
+       elf_end(elf);
+out_close:
+       close(fd);
+out:
+       return err;
+}
+
+int sysfs__read_build_id(const char *filename, void *build_id, size_t size)
+{
+       int fd, err = -1;
+
+       if (size < BUILD_ID_SIZE)
+               goto out;
+
+       fd = open(filename, O_RDONLY);
+       if (fd < 0)
+               goto out;
+
+       while (1) {
+               char bf[BUFSIZ];
+               GElf_Nhdr nhdr;
+               size_t namesz, descsz;
+
+               if (read(fd, &nhdr, sizeof(nhdr)) != sizeof(nhdr))
+                       break;
+
+               namesz = NOTE_ALIGN(nhdr.n_namesz);
+               descsz = NOTE_ALIGN(nhdr.n_descsz);
+               if (nhdr.n_type == NT_GNU_BUILD_ID &&
+                   nhdr.n_namesz == sizeof("GNU")) {
+                       if (read(fd, bf, namesz) != (ssize_t)namesz)
+                               break;
+                       if (memcmp(bf, "GNU", sizeof("GNU")) == 0) {
+                               size_t sz = min(descsz, size);
+                               if (read(fd, build_id, sz) == (ssize_t)sz) {
+                                       memset(build_id + sz, 0, size - sz);
+                                       err = 0;
+                                       break;
+                               }
+                       } else if (read(fd, bf, descsz) != (ssize_t)descsz)
+                               break;
+               } else {
+                       int n = namesz + descsz;
+                       if (read(fd, bf, n) != n)
+                               break;
+               }
+       }
+       close(fd);
+out:
+       return err;
+}
+
+int filename__read_debuglink(const char *filename, char *debuglink,
+                            size_t size)
+{
+       int fd, err = -1;
+       Elf *elf;
+       GElf_Ehdr ehdr;
+       GElf_Shdr shdr;
+       Elf_Data *data;
+       Elf_Scn *sec;
+       Elf_Kind ek;
+
+       fd = open(filename, O_RDONLY);
+       if (fd < 0)
+               goto out;
+
+       elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
+       if (elf == NULL) {
+               pr_debug2("%s: cannot read %s ELF file.\n", __func__, filename);
+               goto out_close;
+       }
+
+       ek = elf_kind(elf);
+       if (ek != ELF_K_ELF)
+               goto out_elf_end;
+
+       if (gelf_getehdr(elf, &ehdr) == NULL) {
+               pr_err("%s: cannot get elf header.\n", __func__);
+               goto out_elf_end;
+       }
+
+       sec = elf_section_by_name(elf, &ehdr, &shdr,
+                                 ".gnu_debuglink", NULL);
+       if (sec == NULL)
+               goto out_elf_end;
+
+       data = elf_getdata(sec, NULL);
+       if (data == NULL)
+               goto out_elf_end;
+
+       /* the start of this section is a zero-terminated string */
+       strncpy(debuglink, data->d_buf, size);
+
+       err = 0;
+
+out_elf_end:
+       elf_end(elf);
+out_close:
+       close(fd);
+out:
+       return err;
+}
+
+static int dso__swap_init(struct dso *dso, unsigned char eidata)
+{
+       static unsigned int const endian = 1;
+
+       dso->needs_swap = DSO_SWAP__NO;
+
+       switch (eidata) {
+       case ELFDATA2LSB:
+               /* We are big endian, DSO is little endian. */
+               if (*(unsigned char const *)&endian != 1)
+                       dso->needs_swap = DSO_SWAP__YES;
+               break;
+
+       case ELFDATA2MSB:
+               /* We are little endian, DSO is big endian. */
+               if (*(unsigned char const *)&endian != 0)
+                       dso->needs_swap = DSO_SWAP__YES;
+               break;
+
+       default:
+               pr_err("unrecognized DSO data encoding %d\n", eidata);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int decompress_kmodule(struct dso *dso, const char *name,
+                             enum dso_binary_type type)
+{
+       int fd = -1;
+       char tmpbuf[] = "/tmp/perf-kmod-XXXXXX";
+       struct kmod_path m;
+
+       if (type != DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE_COMP &&
+           type != DSO_BINARY_TYPE__GUEST_KMODULE_COMP &&
+           type != DSO_BINARY_TYPE__BUILD_ID_CACHE)
+               return -1;
+
+       if (type == DSO_BINARY_TYPE__BUILD_ID_CACHE)
+               name = dso->long_name;
+
+       if (kmod_path__parse_ext(&m, name) || !m.comp)
+               return -1;
+
+       fd = mkstemp(tmpbuf);
+       if (fd < 0) {
+               dso->load_errno = errno;
+               goto out;
+       }
+
+       if (!decompress_to_file(m.ext, name, fd)) {
+               dso->load_errno = DSO_LOAD_ERRNO__DECOMPRESSION_FAILURE;
+               close(fd);
+               fd = -1;
+       }
+
+       unlink(tmpbuf);
+
+out:
+       free(m.ext);
+       return fd;
+}
+
+bool symsrc__possibly_runtime(struct symsrc *ss)
+{
+       return ss->dynsym || ss->opdsec;
+}
+
+bool symsrc__has_symtab(struct symsrc *ss)
+{
+       return ss->symtab != NULL;
+}
+
+void symsrc__destroy(struct symsrc *ss)
+{
+       zfree(&ss->name);
+       elf_end(ss->elf);
+       close(ss->fd);
+}
+
+bool __weak elf__needs_adjust_symbols(GElf_Ehdr ehdr)
+{
+       return ehdr.e_type == ET_EXEC || ehdr.e_type == ET_REL;
+}
+
+int symsrc__init(struct symsrc *ss, struct dso *dso, const char *name,
+                enum dso_binary_type type)
+{
+       int err = -1;
+       GElf_Ehdr ehdr;
+       Elf *elf;
+       int fd;
+
+       if (dso__needs_decompress(dso)) {
+               fd = decompress_kmodule(dso, name, type);
+               if (fd < 0)
+                       return -1;
+       } else {
+               fd = open(name, O_RDONLY);
+               if (fd < 0) {
+                       dso->load_errno = errno;
+                       return -1;
+               }
+       }
+
+       elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
+       if (elf == NULL) {
+               pr_debug("%s: cannot read %s ELF file.\n", __func__, name);
+               dso->load_errno = DSO_LOAD_ERRNO__INVALID_ELF;
+               goto out_close;
+       }
+
+       if (gelf_getehdr(elf, &ehdr) == NULL) {
+               dso->load_errno = DSO_LOAD_ERRNO__INVALID_ELF;
+               pr_debug("%s: cannot get elf header.\n", __func__);
+               goto out_elf_end;
+       }
+
+       if (dso__swap_init(dso, ehdr.e_ident[EI_DATA])) {
+               dso->load_errno = DSO_LOAD_ERRNO__INTERNAL_ERROR;
+               goto out_elf_end;
+       }
+
+       /* Always reject images with a mismatched build-id: */
+       if (dso->has_build_id) {
+               u8 build_id[BUILD_ID_SIZE];
+
+               if (elf_read_build_id(elf, build_id, BUILD_ID_SIZE) < 0) {
+                       dso->load_errno = DSO_LOAD_ERRNO__CANNOT_READ_BUILDID;
+                       goto out_elf_end;
+               }
+
+               if (!dso__build_id_equal(dso, build_id)) {
+                       pr_debug("%s: build id mismatch for %s.\n", __func__, name);
+                       dso->load_errno = DSO_LOAD_ERRNO__MISMATCHING_BUILDID;
+                       goto out_elf_end;
+               }
+       }
+
+       ss->is_64_bit = (gelf_getclass(elf) == ELFCLASS64);
+
+       ss->symtab = elf_section_by_name(elf, &ehdr, &ss->symshdr, ".symtab",
+                       NULL);
+       if (ss->symshdr.sh_type != SHT_SYMTAB)
+               ss->symtab = NULL;
+
+       ss->dynsym_idx = 0;
+       ss->dynsym = elf_section_by_name(elf, &ehdr, &ss->dynshdr, ".dynsym",
+                       &ss->dynsym_idx);
+       if (ss->dynshdr.sh_type != SHT_DYNSYM)
+               ss->dynsym = NULL;
+
+       ss->opdidx = 0;
+       ss->opdsec = elf_section_by_name(elf, &ehdr, &ss->opdshdr, ".opd",
+                       &ss->opdidx);
+       if (ss->opdshdr.sh_type != SHT_PROGBITS)
+               ss->opdsec = NULL;
+
+       if (dso->kernel == DSO_TYPE_USER) {
+               GElf_Shdr shdr;
+               ss->adjust_symbols = (ehdr.e_type == ET_EXEC ||
+                               ehdr.e_type == ET_REL ||
+                               dso__is_vdso(dso) ||
+                               elf_section_by_name(elf, &ehdr, &shdr,
+                                                    ".gnu.prelink_undo",
+                                                    NULL) != NULL);
+       } else {
+               ss->adjust_symbols = elf__needs_adjust_symbols(ehdr);
+       }
+
+       ss->name   = strdup(name);
+       if (!ss->name) {
+               dso->load_errno = errno;
+               goto out_elf_end;
+       }
+
+       ss->elf    = elf;
+       ss->fd     = fd;
+       ss->ehdr   = ehdr;
+       ss->type   = type;
+
+       return 0;
+
+out_elf_end:
+       elf_end(elf);
+out_close:
+       close(fd);
+       return err;
+}
+
+/**
+ * ref_reloc_sym_not_found - has kernel relocation symbol been found.
+ * @kmap: kernel maps and relocation reference symbol
+ *
+ * This function returns %true if we are dealing with the kernel maps and the
+ * relocation reference symbol has not yet been found.  Otherwise %false is
+ * returned.
+ */
+static bool ref_reloc_sym_not_found(struct kmap *kmap)
+{
+       return kmap && kmap->ref_reloc_sym && kmap->ref_reloc_sym->name &&
+              !kmap->ref_reloc_sym->unrelocated_addr;
+}
+
+/**
+ * ref_reloc - kernel relocation offset.
+ * @kmap: kernel maps and relocation reference symbol
+ *
+ * This function returns the offset of kernel addresses as determined by using
+ * the relocation reference symbol i.e. if the kernel has not been relocated
+ * then the return value is zero.
+ */
+static u64 ref_reloc(struct kmap *kmap)
+{
+       if (kmap && kmap->ref_reloc_sym &&
+           kmap->ref_reloc_sym->unrelocated_addr)
+               return kmap->ref_reloc_sym->addr -
+                      kmap->ref_reloc_sym->unrelocated_addr;
+       return 0;
+}
+
+static bool want_demangle(bool is_kernel_sym)
+{
+       return is_kernel_sym ? symbol_conf.demangle_kernel : symbol_conf.demangle;
+}
+
+void __weak arch__elf_sym_adjust(GElf_Sym *sym __maybe_unused) { }
+
+int dso__load_sym(struct dso *dso, struct map *map,
+                 struct symsrc *syms_ss, struct symsrc *runtime_ss,
+                 symbol_filter_t filter, int kmodule)
+{
+       struct kmap *kmap = dso->kernel ? map__kmap(map) : NULL;
+       struct map_groups *kmaps = kmap ? map__kmaps(map) : NULL;
+       struct map *curr_map = map;
+       struct dso *curr_dso = dso;
+       Elf_Data *symstrs, *secstrs;
+       uint32_t nr_syms;
+       int err = -1;
+       uint32_t idx;
+       GElf_Ehdr ehdr;
+       GElf_Shdr shdr;
+       Elf_Data *syms, *opddata = NULL;
+       GElf_Sym sym;
+       Elf_Scn *sec, *sec_strndx;
+       Elf *elf;
+       int nr = 0;
+       bool remap_kernel = false, adjust_kernel_syms = false;
+
+       if (kmap && !kmaps)
+               return -1;
+
+       dso->symtab_type = syms_ss->type;
+       dso->is_64_bit = syms_ss->is_64_bit;
+       dso->rel = syms_ss->ehdr.e_type == ET_REL;
+
+       /*
+        * Modules may already have symbols from kallsyms, but those symbols
+        * have the wrong values for the dso maps, so remove them.
+        */
+       if (kmodule && syms_ss->symtab)
+               symbols__delete(&dso->symbols[map->type]);
+
+       if (!syms_ss->symtab) {
+               /*
+                * If the vmlinux is stripped, fail so we will fall back
+                * to using kallsyms. The vmlinux runtime symbols aren't
+                * of much use.
+                */
+               if (dso->kernel)
+                       goto out_elf_end;
+
+               syms_ss->symtab  = syms_ss->dynsym;
+               syms_ss->symshdr = syms_ss->dynshdr;
+       }
+
+       elf = syms_ss->elf;
+       ehdr = syms_ss->ehdr;
+       sec = syms_ss->symtab;
+       shdr = syms_ss->symshdr;
+
+       if (runtime_ss->opdsec)
+               opddata = elf_rawdata(runtime_ss->opdsec, NULL);
+
+       syms = elf_getdata(sec, NULL);
+       if (syms == NULL)
+               goto out_elf_end;
+
+       sec = elf_getscn(elf, shdr.sh_link);
+       if (sec == NULL)
+               goto out_elf_end;
+
+       symstrs = elf_getdata(sec, NULL);
+       if (symstrs == NULL)
+               goto out_elf_end;
+
+       sec_strndx = elf_getscn(runtime_ss->elf, runtime_ss->ehdr.e_shstrndx);
+       if (sec_strndx == NULL)
+               goto out_elf_end;
+
+       secstrs = elf_getdata(sec_strndx, NULL);
+       if (secstrs == NULL)
+               goto out_elf_end;
+
+       nr_syms = shdr.sh_size / shdr.sh_entsize;
+
+       memset(&sym, 0, sizeof(sym));
+
+       /*
+        * The kernel relocation symbol is needed in advance in order to adjust
+        * kernel maps correctly.
+        */
+       if (ref_reloc_sym_not_found(kmap)) {
+               elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) {
+                       const char *elf_name = elf_sym__name(&sym, symstrs);
+
+                       if (strcmp(elf_name, kmap->ref_reloc_sym->name))
+                               continue;
+                       kmap->ref_reloc_sym->unrelocated_addr = sym.st_value;
+                       map->reloc = kmap->ref_reloc_sym->addr -
+                                    kmap->ref_reloc_sym->unrelocated_addr;
+                       break;
+               }
+       }
+
+       /*
+        * Handle any relocation of vdso necessary because older kernels
+        * attempted to prelink vdso to its virtual address.
+        */
+       if (dso__is_vdso(dso)) {
+               GElf_Shdr tshdr;
+
+               if (elf_section_by_name(elf, &ehdr, &tshdr, ".text", NULL))
+                       map->reloc = map->start - tshdr.sh_addr + tshdr.sh_offset;
+       }
+
+       dso->adjust_symbols = runtime_ss->adjust_symbols || ref_reloc(kmap);
+       /*
+        * Initial kernel and module mappings do not map to the dso.  For
+        * function mappings, flag the fixups.
+        */
+       if (map->type == MAP__FUNCTION && (dso->kernel || kmodule)) {
+               remap_kernel = true;
+               adjust_kernel_syms = dso->adjust_symbols;
+       }
+       elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) {
+               struct symbol *f;
+               const char *elf_name = elf_sym__name(&sym, symstrs);
+               char *demangled = NULL;
+               int is_label = elf_sym__is_label(&sym);
+               const char *section_name;
+               bool used_opd = false;
+
+               if (!is_label && !elf_sym__is_a(&sym, map->type))
+                       continue;
+
+               /* Reject ARM ELF "mapping symbols": these aren't unique and
+                * don't identify functions, so will confuse the profile
+                * output: */
+               if (ehdr.e_machine == EM_ARM || ehdr.e_machine == EM_AARCH64) {
+                       if (elf_name[0] == '$' && strchr("adtx", elf_name[1])
+                           && (elf_name[2] == '\0' || elf_name[2] == '.'))
+                               continue;
+               }
+
+               if (runtime_ss->opdsec && sym.st_shndx == runtime_ss->opdidx) {
+                       u32 offset = sym.st_value - syms_ss->opdshdr.sh_addr;
+                       u64 *opd = opddata->d_buf + offset;
+                       sym.st_value = DSO__SWAP(dso, u64, *opd);
+                       sym.st_shndx = elf_addr_to_index(runtime_ss->elf,
+                                       sym.st_value);
+                       used_opd = true;
+               }
+               /*
+                * When loading symbols in a data mapping, ABS symbols (which
+                * has a value of SHN_ABS in its st_shndx) failed at
+                * elf_getscn().  And it marks the loading as a failure so
+                * already loaded symbols cannot be fixed up.
+                *
+                * I'm not sure what should be done. Just ignore them for now.
+                * - Namhyung Kim
+                */
+               if (sym.st_shndx == SHN_ABS)
+                       continue;
+
+               sec = elf_getscn(runtime_ss->elf, sym.st_shndx);
+               if (!sec)
+                       goto out_elf_end;
+
+               gelf_getshdr(sec, &shdr);
+
+               if (is_label && !elf_sec__is_a(&shdr, secstrs, map->type))
+                       continue;
+
+               section_name = elf_sec__name(&shdr, secstrs);
+
+               /* On ARM, symbols for thumb functions have 1 added to
+                * the symbol address as a flag - remove it */
+               if ((ehdr.e_machine == EM_ARM) &&
+                   (map->type == MAP__FUNCTION) &&
+                   (sym.st_value & 1))
+                       --sym.st_value;
+
+               arch__elf_sym_adjust(&sym);
+
+               if (dso->kernel || kmodule) {
+                       char dso_name[PATH_MAX];
+
+                       /* Adjust symbol to map to file offset */
+                       if (adjust_kernel_syms)
+                               sym.st_value -= shdr.sh_addr - shdr.sh_offset;
+
+                       if (strcmp(section_name,
+                                  (curr_dso->short_name +
+                                   dso->short_name_len)) == 0)
+                               goto new_symbol;
+
+                       if (strcmp(section_name, ".text") == 0) {
+                               /*
+                                * The initial kernel mapping is based on
+                                * kallsyms and identity maps.  Overwrite it to
+                                * map to the kernel dso.
+                                */
+                               if (remap_kernel && dso->kernel) {
+                                       remap_kernel = false;
+                                       map->start = shdr.sh_addr +
+                                                    ref_reloc(kmap);
+                                       map->end = map->start + shdr.sh_size;
+                                       map->pgoff = shdr.sh_offset;
+                                       map->map_ip = map__map_ip;
+                                       map->unmap_ip = map__unmap_ip;
+                                       /* Ensure maps are correctly ordered */
+                                       if (kmaps) {
+                                               map__get(map);
+                                               map_groups__remove(kmaps, map);
+                                               map_groups__insert(kmaps, map);
+                                               map__put(map);
+                                       }
+                               }
+
+                               /*
+                                * The initial module mapping is based on
+                                * /proc/modules mapped to offset zero.
+                                * Overwrite it to map to the module dso.
+                                */
+                               if (remap_kernel && kmodule) {
+                                       remap_kernel = false;
+                                       map->pgoff = shdr.sh_offset;
+                               }
+
+                               curr_map = map;
+                               curr_dso = dso;
+                               goto new_symbol;
+                       }
+
+                       if (!kmap)
+                               goto new_symbol;
+
+                       snprintf(dso_name, sizeof(dso_name),
+                                "%s%s", dso->short_name, section_name);
+
+                       curr_map = map_groups__find_by_name(kmaps, map->type, dso_name);
+                       if (curr_map == NULL) {
+                               u64 start = sym.st_value;
+
+                               if (kmodule)
+                                       start += map->start + shdr.sh_offset;
+
+                               curr_dso = dso__new(dso_name);
+                               if (curr_dso == NULL)
+                                       goto out_elf_end;
+                               curr_dso->kernel = dso->kernel;
+                               curr_dso->long_name = dso->long_name;
+                               curr_dso->long_name_len = dso->long_name_len;
+                               curr_map = map__new2(start, curr_dso,
+                                                    map->type);
+                               dso__put(curr_dso);
+                               if (curr_map == NULL) {
+                                       goto out_elf_end;
+                               }
+                               if (adjust_kernel_syms) {
+                                       curr_map->start = shdr.sh_addr +
+                                                         ref_reloc(kmap);
+                                       curr_map->end = curr_map->start +
+                                                       shdr.sh_size;
+                                       curr_map->pgoff = shdr.sh_offset;
+                               } else {
+                                       curr_map->map_ip = identity__map_ip;
+                                       curr_map->unmap_ip = identity__map_ip;
+                               }
+                               curr_dso->symtab_type = dso->symtab_type;
+                               map_groups__insert(kmaps, curr_map);
+                               /*
+                                * Add it before we drop the referece to curr_map,
+                                * i.e. while we still are sure to have a reference
+                                * to this DSO via curr_map->dso.
+                                */
+                               dsos__add(&map->groups->machine->dsos, curr_dso);
+                               /* kmaps already got it */
+                               map__put(curr_map);
+                               dso__set_loaded(curr_dso, map->type);
+                       } else
+                               curr_dso = curr_map->dso;
+
+                       goto new_symbol;
+               }
+
+               if ((used_opd && runtime_ss->adjust_symbols)
+                               || (!used_opd && syms_ss->adjust_symbols)) {
+                       pr_debug4("%s: adjusting symbol: st_value: %#" PRIx64 " "
+                                 "sh_addr: %#" PRIx64 " sh_offset: %#" PRIx64 "\n", __func__,
+                                 (u64)sym.st_value, (u64)shdr.sh_addr,
+                                 (u64)shdr.sh_offset);
+                       sym.st_value -= shdr.sh_addr - shdr.sh_offset;
+               }
+new_symbol:
+               /*
+                * We need to figure out if the object was created from C++ sources
+                * DWARF DW_compile_unit has this, but we don't always have access
+                * to it...
+                */
+               if (want_demangle(dso->kernel || kmodule)) {
+                       int demangle_flags = DMGL_NO_OPTS;
+                       if (verbose)
+                               demangle_flags = DMGL_PARAMS | DMGL_ANSI;
+
+                       demangled = bfd_demangle(NULL, elf_name, demangle_flags);
+                       if (demangled != NULL)
+                               elf_name = demangled;
+               }
+               f = symbol__new(sym.st_value, sym.st_size,
+                               GELF_ST_BIND(sym.st_info), elf_name);
+               free(demangled);
+               if (!f)
+                       goto out_elf_end;
+
+               if (filter && filter(curr_map, f))
+                       symbol__delete(f);
+               else {
+                       symbols__insert(&curr_dso->symbols[curr_map->type], f);
+                       nr++;
+               }
+       }
+
+       /*
+        * For misannotated, zeroed, ASM function sizes.
+        */
+       if (nr > 0) {
+               if (!symbol_conf.allow_aliases)
+                       symbols__fixup_duplicate(&dso->symbols[map->type]);
+               symbols__fixup_end(&dso->symbols[map->type]);
+               if (kmap) {
+                       /*
+                        * We need to fixup this here too because we create new
+                        * maps here, for things like vsyscall sections.
+                        */
+                       __map_groups__fixup_end(kmaps, map->type);
+               }
+       }
+       err = nr;
+out_elf_end:
+       return err;
+}
+
+static int elf_read_maps(Elf *elf, bool exe, mapfn_t mapfn, void *data)
+{
+       GElf_Phdr phdr;
+       size_t i, phdrnum;
+       int err;
+       u64 sz;
+
+       if (elf_getphdrnum(elf, &phdrnum))
+               return -1;
+
+       for (i = 0; i < phdrnum; i++) {
+               if (gelf_getphdr(elf, i, &phdr) == NULL)
+                       return -1;
+               if (phdr.p_type != PT_LOAD)
+                       continue;
+               if (exe) {
+                       if (!(phdr.p_flags & PF_X))
+                               continue;
+               } else {
+                       if (!(phdr.p_flags & PF_R))
+                               continue;
+               }
+               sz = min(phdr.p_memsz, phdr.p_filesz);
+               if (!sz)
+                       continue;
+               err = mapfn(phdr.p_vaddr, sz, phdr.p_offset, data);
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+int file__read_maps(int fd, bool exe, mapfn_t mapfn, void *data,
+                   bool *is_64_bit)
+{
+       int err;
+       Elf *elf;
+
+       elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
+       if (elf == NULL)
+               return -1;
+
+       if (is_64_bit)
+               *is_64_bit = (gelf_getclass(elf) == ELFCLASS64);
+
+       err = elf_read_maps(elf, exe, mapfn, data);
+
+       elf_end(elf);
+       return err;
+}
+
+enum dso_type dso__type_fd(int fd)
+{
+       enum dso_type dso_type = DSO__TYPE_UNKNOWN;
+       GElf_Ehdr ehdr;
+       Elf_Kind ek;
+       Elf *elf;
+
+       elf = elf_begin(fd, PERF_ELF_C_READ_MMAP, NULL);
+       if (elf == NULL)
+               goto out;
+
+       ek = elf_kind(elf);
+       if (ek != ELF_K_ELF)
+               goto out_end;
+
+       if (gelf_getclass(elf) == ELFCLASS64) {
+               dso_type = DSO__TYPE_64BIT;
+               goto out_end;
+       }
+
+       if (gelf_getehdr(elf, &ehdr) == NULL)
+               goto out_end;
+
+       if (ehdr.e_machine == EM_X86_64)
+               dso_type = DSO__TYPE_X32BIT;
+       else
+               dso_type = DSO__TYPE_32BIT;
+out_end:
+       elf_end(elf);
+out:
+       return dso_type;
+}
+
+static int copy_bytes(int from, off_t from_offs, int to, off_t to_offs, u64 len)
+{
+       ssize_t r;
+       size_t n;
+       int err = -1;
+       char *buf = malloc(page_size);
+
+       if (buf == NULL)
+               return -1;
+
+       if (lseek(to, to_offs, SEEK_SET) != to_offs)
+               goto out;
+
+       if (lseek(from, from_offs, SEEK_SET) != from_offs)
+               goto out;
+
+       while (len) {
+               n = page_size;
+               if (len < n)
+                       n = len;
+               /* Use read because mmap won't work on proc files */
+               r = read(from, buf, n);
+               if (r < 0)
+                       goto out;
+               if (!r)
+                       break;
+               n = r;
+               r = write(to, buf, n);
+               if (r < 0)
+                       goto out;
+               if ((size_t)r != n)
+                       goto out;
+               len -= n;
+       }
+
+       err = 0;
+out:
+       free(buf);
+       return err;
+}
+
+struct kcore {
+       int fd;
+       int elfclass;
+       Elf *elf;
+       GElf_Ehdr ehdr;
+};
+
+static int kcore__open(struct kcore *kcore, const char *filename)
+{
+       GElf_Ehdr *ehdr;
+
+       kcore->fd = open(filename, O_RDONLY);
+       if (kcore->fd == -1)
+               return -1;
+
+       kcore->elf = elf_begin(kcore->fd, ELF_C_READ, NULL);
+       if (!kcore->elf)
+               goto out_close;
+
+       kcore->elfclass = gelf_getclass(kcore->elf);
+       if (kcore->elfclass == ELFCLASSNONE)
+               goto out_end;
+
+       ehdr = gelf_getehdr(kcore->elf, &kcore->ehdr);
+       if (!ehdr)
+               goto out_end;
+
+       return 0;
+
+out_end:
+       elf_end(kcore->elf);
+out_close:
+       close(kcore->fd);
+       return -1;
+}
+
+static int kcore__init(struct kcore *kcore, char *filename, int elfclass,
+                      bool temp)
+{
+       kcore->elfclass = elfclass;
+
+       if (temp)
+               kcore->fd = mkstemp(filename);
+       else
+               kcore->fd = open(filename, O_WRONLY | O_CREAT | O_EXCL, 0400);
+       if (kcore->fd == -1)
+               return -1;
+
+       kcore->elf = elf_begin(kcore->fd, ELF_C_WRITE, NULL);
+       if (!kcore->elf)
+               goto out_close;
+
+       if (!gelf_newehdr(kcore->elf, elfclass))
+               goto out_end;
+
+       memset(&kcore->ehdr, 0, sizeof(GElf_Ehdr));
+
+       return 0;
+
+out_end:
+       elf_end(kcore->elf);
+out_close:
+       close(kcore->fd);
+       unlink(filename);
+       return -1;
+}
+
+static void kcore__close(struct kcore *kcore)
+{
+       elf_end(kcore->elf);
+       close(kcore->fd);
+}
+
+static int kcore__copy_hdr(struct kcore *from, struct kcore *to, size_t count)
+{
+       GElf_Ehdr *ehdr = &to->ehdr;
+       GElf_Ehdr *kehdr = &from->ehdr;
+
+       memcpy(ehdr->e_ident, kehdr->e_ident, EI_NIDENT);
+       ehdr->e_type      = kehdr->e_type;
+       ehdr->e_machine   = kehdr->e_machine;
+       ehdr->e_version   = kehdr->e_version;
+       ehdr->e_entry     = 0;
+       ehdr->e_shoff     = 0;
+       ehdr->e_flags     = kehdr->e_flags;
+       ehdr->e_phnum     = count;
+       ehdr->e_shentsize = 0;
+       ehdr->e_shnum     = 0;
+       ehdr->e_shstrndx  = 0;
+
+       if (from->elfclass == ELFCLASS32) {
+               ehdr->e_phoff     = sizeof(Elf32_Ehdr);
+               ehdr->e_ehsize    = sizeof(Elf32_Ehdr);
+               ehdr->e_phentsize = sizeof(Elf32_Phdr);
+       } else {
+               ehdr->e_phoff     = sizeof(Elf64_Ehdr);
+               ehdr->e_ehsize    = sizeof(Elf64_Ehdr);
+               ehdr->e_phentsize = sizeof(Elf64_Phdr);
+       }
+
+       if (!gelf_update_ehdr(to->elf, ehdr))
+               return -1;
+
+       if (!gelf_newphdr(to->elf, count))
+               return -1;
+
+       return 0;
+}
+
+static int kcore__add_phdr(struct kcore *kcore, int idx, off_t offset,
+                          u64 addr, u64 len)
+{
+       GElf_Phdr phdr = {
+               .p_type         = PT_LOAD,
+               .p_flags        = PF_R | PF_W | PF_X,
+               .p_offset       = offset,
+               .p_vaddr        = addr,
+               .p_paddr        = 0,
+               .p_filesz       = len,
+               .p_memsz        = len,
+               .p_align        = page_size,
+       };
+
+       if (!gelf_update_phdr(kcore->elf, idx, &phdr))
+               return -1;
+
+       return 0;
+}
+
+static off_t kcore__write(struct kcore *kcore)
+{
+       return elf_update(kcore->elf, ELF_C_WRITE);
+}
+
+struct phdr_data {
+       off_t offset;
+       u64 addr;
+       u64 len;
+};
+
+struct kcore_copy_info {
+       u64 stext;
+       u64 etext;
+       u64 first_symbol;
+       u64 last_symbol;
+       u64 first_module;
+       u64 last_module_symbol;
+       struct phdr_data kernel_map;
+       struct phdr_data modules_map;
+};
+
+static int kcore_copy__process_kallsyms(void *arg, const char *name, char type,
+                                       u64 start)
+{
+       struct kcore_copy_info *kci = arg;
+
+       if (!symbol_type__is_a(type, MAP__FUNCTION))
+               return 0;
+
+       if (strchr(name, '[')) {
+               if (start > kci->last_module_symbol)
+                       kci->last_module_symbol = start;
+               return 0;
+       }
+
+       if (!kci->first_symbol || start < kci->first_symbol)
+               kci->first_symbol = start;
+
+       if (!kci->last_symbol || start > kci->last_symbol)
+               kci->last_symbol = start;
+
+       if (!strcmp(name, "_stext")) {
+               kci->stext = start;
+               return 0;
+       }
+
+       if (!strcmp(name, "_etext")) {
+               kci->etext = start;
+               return 0;
+       }
+
+       return 0;
+}
+
+static int kcore_copy__parse_kallsyms(struct kcore_copy_info *kci,
+                                     const char *dir)
+{
+       char kallsyms_filename[PATH_MAX];
+
+       scnprintf(kallsyms_filename, PATH_MAX, "%s/kallsyms", dir);
+
+       if (symbol__restricted_filename(kallsyms_filename, "/proc/kallsyms"))
+               return -1;
+
+       if (kallsyms__parse(kallsyms_filename, kci,
+                           kcore_copy__process_kallsyms) < 0)
+               return -1;
+
+       return 0;
+}
+
+static int kcore_copy__process_modules(void *arg,
+                                      const char *name __maybe_unused,
+                                      u64 start)
+{
+       struct kcore_copy_info *kci = arg;
+
+       if (!kci->first_module || start < kci->first_module)
+               kci->first_module = start;
+
+       return 0;
+}
+
+static int kcore_copy__parse_modules(struct kcore_copy_info *kci,
+                                    const char *dir)
+{
+       char modules_filename[PATH_MAX];
+
+       scnprintf(modules_filename, PATH_MAX, "%s/modules", dir);
+
+       if (symbol__restricted_filename(modules_filename, "/proc/modules"))
+               return -1;
+
+       if (modules__parse(modules_filename, kci,
+                          kcore_copy__process_modules) < 0)
+               return -1;
+
+       return 0;
+}
+
+static void kcore_copy__map(struct phdr_data *p, u64 start, u64 end, u64 pgoff,
+                           u64 s, u64 e)
+{
+       if (p->addr || s < start || s >= end)
+               return;
+
+       p->addr = s;
+       p->offset = (s - start) + pgoff;
+       p->len = e < end ? e - s : end - s;
+}
+
+static int kcore_copy__read_map(u64 start, u64 len, u64 pgoff, void *data)
+{
+       struct kcore_copy_info *kci = data;
+       u64 end = start + len;
+
+       kcore_copy__map(&kci->kernel_map, start, end, pgoff, kci->stext,
+                       kci->etext);
+
+       kcore_copy__map(&kci->modules_map, start, end, pgoff, kci->first_module,
+                       kci->last_module_symbol);
+
+       return 0;
+}
+
+static int kcore_copy__read_maps(struct kcore_copy_info *kci, Elf *elf)
+{
+       if (elf_read_maps(elf, true, kcore_copy__read_map, kci) < 0)
+               return -1;
+
+       return 0;
+}
+
+static int kcore_copy__calc_maps(struct kcore_copy_info *kci, const char *dir,
+                                Elf *elf)
+{
+       if (kcore_copy__parse_kallsyms(kci, dir))
+               return -1;
+
+       if (kcore_copy__parse_modules(kci, dir))
+               return -1;
+
+       if (kci->stext)
+               kci->stext = round_down(kci->stext, page_size);
+       else
+               kci->stext = round_down(kci->first_symbol, page_size);
+
+       if (kci->etext) {
+               kci->etext = round_up(kci->etext, page_size);
+       } else if (kci->last_symbol) {
+               kci->etext = round_up(kci->last_symbol, page_size);
+               kci->etext += page_size;
+       }
+
+       kci->first_module = round_down(kci->first_module, page_size);
+
+       if (kci->last_module_symbol) {
+               kci->last_module_symbol = round_up(kci->last_module_symbol,
+                                                  page_size);
+               kci->last_module_symbol += page_size;
+       }
+
+       if (!kci->stext || !kci->etext)
+               return -1;
+
+       if (kci->first_module && !kci->last_module_symbol)
+               return -1;
+
+       return kcore_copy__read_maps(kci, elf);
+}
+
+static int kcore_copy__copy_file(const char *from_dir, const char *to_dir,
+                                const char *name)
+{
+       char from_filename[PATH_MAX];
+       char to_filename[PATH_MAX];
+
+       scnprintf(from_filename, PATH_MAX, "%s/%s", from_dir, name);
+       scnprintf(to_filename, PATH_MAX, "%s/%s", to_dir, name);
+
+       return copyfile_mode(from_filename, to_filename, 0400);
+}
+
+static int kcore_copy__unlink(const char *dir, const char *name)
+{
+       char filename[PATH_MAX];
+
+       scnprintf(filename, PATH_MAX, "%s/%s", dir, name);
+
+       return unlink(filename);
+}
+
+static int kcore_copy__compare_fds(int from, int to)
+{
+       char *buf_from;
+       char *buf_to;
+       ssize_t ret;
+       size_t len;
+       int err = -1;
+
+       buf_from = malloc(page_size);
+       buf_to = malloc(page_size);
+       if (!buf_from || !buf_to)
+               goto out;
+
+       while (1) {
+               /* Use read because mmap won't work on proc files */
+               ret = read(from, buf_from, page_size);
+               if (ret < 0)
+                       goto out;
+
+               if (!ret)
+                       break;
+
+               len = ret;
+
+               if (readn(to, buf_to, len) != (int)len)
+                       goto out;
+
+               if (memcmp(buf_from, buf_to, len))
+                       goto out;
+       }
+
+       err = 0;
+out:
+       free(buf_to);
+       free(buf_from);
+       return err;
+}
+
+static int kcore_copy__compare_files(const char *from_filename,
+                                    const char *to_filename)
+{
+       int from, to, err = -1;
+
+       from = open(from_filename, O_RDONLY);
+       if (from < 0)
+               return -1;
+
+       to = open(to_filename, O_RDONLY);
+       if (to < 0)
+               goto out_close_from;
+
+       err = kcore_copy__compare_fds(from, to);
+
+       close(to);
+out_close_from:
+       close(from);
+       return err;
+}
+
+static int kcore_copy__compare_file(const char *from_dir, const char *to_dir,
+                                   const char *name)
+{
+       char from_filename[PATH_MAX];
+       char to_filename[PATH_MAX];
+
+       scnprintf(from_filename, PATH_MAX, "%s/%s", from_dir, name);
+       scnprintf(to_filename, PATH_MAX, "%s/%s", to_dir, name);
+
+       return kcore_copy__compare_files(from_filename, to_filename);
+}
+
+/**
+ * kcore_copy - copy kallsyms, modules and kcore from one directory to another.
+ * @from_dir: from directory
+ * @to_dir: to directory
+ *
+ * This function copies kallsyms, modules and kcore files from one directory to
+ * another.  kallsyms and modules are copied entirely.  Only code segments are
+ * copied from kcore.  It is assumed that two segments suffice: one for the
+ * kernel proper and one for all the modules.  The code segments are determined
+ * from kallsyms and modules files.  The kernel map starts at _stext or the
+ * lowest function symbol, and ends at _etext or the highest function symbol.
+ * The module map starts at the lowest module address and ends at the highest
+ * module symbol.  Start addresses are rounded down to the nearest page.  End
+ * addresses are rounded up to the nearest page.  An extra page is added to the
+ * highest kernel symbol and highest module symbol to, hopefully, encompass that
+ * symbol too.  Because it contains only code sections, the resulting kcore is
+ * unusual.  One significant peculiarity is that the mapping (start -> pgoff)
+ * is not the same for the kernel map and the modules map.  That happens because
+ * the data is copied adjacently whereas the original kcore has gaps.  Finally,
+ * kallsyms and modules files are compared with their copies to check that
+ * modules have not been loaded or unloaded while the copies were taking place.
+ *
+ * Return: %0 on success, %-1 on failure.
+ */
+int kcore_copy(const char *from_dir, const char *to_dir)
+{
+       struct kcore kcore;
+       struct kcore extract;
+       size_t count = 2;
+       int idx = 0, err = -1;
+       off_t offset = page_size, sz, modules_offset = 0;
+       struct kcore_copy_info kci = { .stext = 0, };
+       char kcore_filename[PATH_MAX];
+       char extract_filename[PATH_MAX];
+
+       if (kcore_copy__copy_file(from_dir, to_dir, "kallsyms"))
+               return -1;
+
+       if (kcore_copy__copy_file(from_dir, to_dir, "modules"))
+               goto out_unlink_kallsyms;
+
+       scnprintf(kcore_filename, PATH_MAX, "%s/kcore", from_dir);
+       scnprintf(extract_filename, PATH_MAX, "%s/kcore", to_dir);
+
+       if (kcore__open(&kcore, kcore_filename))
+               goto out_unlink_modules;
+
+       if (kcore_copy__calc_maps(&kci, from_dir, kcore.elf))
+               goto out_kcore_close;
+
+       if (kcore__init(&extract, extract_filename, kcore.elfclass, false))
+               goto out_kcore_close;
+
+       if (!kci.modules_map.addr)
+               count -= 1;
+
+       if (kcore__copy_hdr(&kcore, &extract, count))
+               goto out_extract_close;
+
+       if (kcore__add_phdr(&extract, idx++, offset, kci.kernel_map.addr,
+                           kci.kernel_map.len))
+               goto out_extract_close;
+
+       if (kci.modules_map.addr) {
+               modules_offset = offset + kci.kernel_map.len;
+               if (kcore__add_phdr(&extract, idx, modules_offset,
+                                   kci.modules_map.addr, kci.modules_map.len))
+                       goto out_extract_close;
+       }
+
+       sz = kcore__write(&extract);
+       if (sz < 0 || sz > offset)
+               goto out_extract_close;
+
+       if (copy_bytes(kcore.fd, kci.kernel_map.offset, extract.fd, offset,
+                      kci.kernel_map.len))
+               goto out_extract_close;
+
+       if (modules_offset && copy_bytes(kcore.fd, kci.modules_map.offset,
+                                        extract.fd, modules_offset,
+                                        kci.modules_map.len))
+               goto out_extract_close;
+
+       if (kcore_copy__compare_file(from_dir, to_dir, "modules"))
+               goto out_extract_close;
+
+       if (kcore_copy__compare_file(from_dir, to_dir, "kallsyms"))
+               goto out_extract_close;
+
+       err = 0;
+
+out_extract_close:
+       kcore__close(&extract);
+       if (err)
+               unlink(extract_filename);
+out_kcore_close:
+       kcore__close(&kcore);
+out_unlink_modules:
+       if (err)
+               kcore_copy__unlink(to_dir, "modules");
+out_unlink_kallsyms:
+       if (err)
+               kcore_copy__unlink(to_dir, "kallsyms");
+
+       return err;
+}
+
+int kcore_extract__create(struct kcore_extract *kce)
+{
+       struct kcore kcore;
+       struct kcore extract;
+       size_t count = 1;
+       int idx = 0, err = -1;
+       off_t offset = page_size, sz;
+
+       if (kcore__open(&kcore, kce->kcore_filename))
+               return -1;
+
+       strcpy(kce->extract_filename, PERF_KCORE_EXTRACT);
+       if (kcore__init(&extract, kce->extract_filename, kcore.elfclass, true))
+               goto out_kcore_close;
+
+       if (kcore__copy_hdr(&kcore, &extract, count))
+               goto out_extract_close;
+
+       if (kcore__add_phdr(&extract, idx, offset, kce->addr, kce->len))
+               goto out_extract_close;
+
+       sz = kcore__write(&extract);
+       if (sz < 0 || sz > offset)
+               goto out_extract_close;
+
+       if (copy_bytes(kcore.fd, kce->offs, extract.fd, offset, kce->len))
+               goto out_extract_close;
+
+       err = 0;
+
+out_extract_close:
+       kcore__close(&extract);
+       if (err)
+               unlink(kce->extract_filename);
+out_kcore_close:
+       kcore__close(&kcore);
+
+       return err;
+}
+
+void kcore_extract__delete(struct kcore_extract *kce)
+{
+       unlink(kce->extract_filename);
+}
+
+void symbol__elf_init(void)
+{
+       elf_version(EV_CURRENT);
+}