diff options
Diffstat (limited to 'toolchain/sstrip/src')
-rw-r--r-- | toolchain/sstrip/src/sstrip.c | 492 |
1 files changed, 492 insertions, 0 deletions
diff --git a/toolchain/sstrip/src/sstrip.c b/toolchain/sstrip/src/sstrip.c new file mode 100644 index 0000000000..60c12c972c --- /dev/null +++ b/toolchain/sstrip/src/sstrip.c @@ -0,0 +1,492 @@ +/* http://www.muppetlabs.com/~breadbox/software/elfkickers.html */ + +/* sstrip: Copyright (C) 1999-2001 by Brian Raiter, under the GNU + * General Public License. No warranty. See COPYING for details. + * + * Aug 23, 2004 Hacked by Manuel Novoa III <mjn3@codepoet.org> to + * handle targets of different endianness and/or elf class, making + * it more useful in a cross-devel environment. + */ + +/* ============== original README =================== + * + * sstrip is a small utility that removes the contents at the end of an + * ELF file that are not part of the program's memory image. + * + * Most ELF executables are built with both a program header table and a + * section header table. However, only the former is required in order + * for the OS to load, link and execute a program. sstrip attempts to + * extract the ELF header, the program header table, and its contents, + * leaving everything else in the bit bucket. It can only remove parts of + * the file that occur at the end, after the parts to be saved. However, + * this almost always includes the section header table, and occasionally + * a few random sections that are not used when running a program. + * + * It should be noted that the GNU bfd library is (understandably) + * dependent on the section header table as an index to the file's + * contents. Thus, an executable file that has no section header table + * cannot be used with gdb, objdump, or any other program based upon the + * bfd library, at all. In fact, the program will not even recognize the + * file as a valid executable. (This limitation is noted in the source + * code comments for bfd, and is marked "FIXME", so this may change at + * some future date. However, I would imagine that it is a pretty + * low-priority item, as executables without a section header table are + * rare in the extreme.) This probably also explains why strip doesn't + * offer the option to do this. + * + * Shared library files may also have their section header table removed. + * Such a library will still function; however, it will no longer be + * possible for a compiler to link a new program against it. + * + * As an added bonus, sstrip also tries to removes trailing zero bytes + * from the end of the file. (This normally cannot be done with an + * executable that has a section header table.) + * + * sstrip is a very simplistic program. It depends upon the common + * practice of putting the parts of the file that contribute to the + * memory image at the front, and the remaining material at the end. This + * permits it to discard the latter material without affecting file + * offsets and memory addresses in what remains. Of course, the ELF + * standard permits files to be organized in almost any order, so if a + * pathological linker decided to put its section headers at the top, + * sstrip would be useless on such executables. + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <errno.h> +#include <unistd.h> +#include <fcntl.h> +#include <elf.h> + +#ifdef __FreeBSD__ +/** + * This seems to work on FreeBSD 5.3, should + * work on all newer versions as well. I have + * no idea if it will work on versions < 5.3 + * + * Joe Estock (guru) <jestock at nutextonline.com> + */ +#include <sys/endian.h> +#define bswap_64 __bswap64 +#define bswap_32 __bswap32 +#define bswap_16 __bswap16 +#elif defined(__APPLE__) +#include <machine/endian.h> +#include <machine/byte_order.h> +#define __BYTE_ORDER BYTE_ORDER +#define __BIG_ENDIAN BIG_ENDIAN +#define bswap_16(x) NXSwapShort(x) +#define bswap_32(x) NXSwapInt(x) +#define bswap_64(x) NXSwapLongLong(x) +#else +#include <endian.h> +#include <byteswap.h> +#endif + + +#ifndef TRUE +#define TRUE 1 +#define FALSE 0 +#endif + +/* The name of the program. + */ +static char const *progname; + +/* The name of the current file. + */ +static char const *filename; + + +/* A simple error-handling function. FALSE is always returned for the + * convenience of the caller. + */ +static int err(char const *errmsg) +{ + fprintf(stderr, "%s: %s: %s\n", progname, filename, errmsg); + return FALSE; +} + +/* A flag to signal the need for endian reversal. + */ +static int do_reverse_endian; + +/* Get a value from the elf header, compensating for endianness. + */ +#define EGET(X) \ + (__extension__ ({ \ + uint64_t __res; \ + if (!do_reverse_endian) { \ + __res = (X); \ + } else if (sizeof(X) == 1) { \ + __res = (X); \ + } else if (sizeof(X) == 2) { \ + __res = bswap_16((X)); \ + } else if (sizeof(X) == 4) { \ + __res = bswap_32((X)); \ + } else if (sizeof(X) == 8) { \ + __res = bswap_64((X)); \ + } else { \ + fprintf(stderr, "%s: %s: EGET failed for size %d\n", \ + progname, filename, sizeof(X)); \ + exit(EXIT_FAILURE); \ + } \ + __res; \ + })) + +/* Set a value 'Y' in the elf header to 'X', compensating for endianness. + */ +#define ESET(Y,X) \ + do if (!do_reverse_endian) { \ + Y = (X); \ + } else if (sizeof(Y) == 1) { \ + Y = (X); \ + } else if (sizeof(Y) == 2) { \ + Y = bswap_16((uint16_t)(X)); \ + } else if (sizeof(Y) == 4) { \ + Y = bswap_32((uint32_t)(X)); \ + } else if (sizeof(Y) == 8) { \ + Y = bswap_64((uint64_t)(X)); \ + } else { \ + fprintf(stderr, "%s: %s: ESET failed for size %d\n", \ + progname, filename, sizeof(Y)); \ + exit(EXIT_FAILURE); \ + } while (0) + + +/* A macro for I/O errors: The given error message is used only when + * errno is not set. + */ +#define ferr(msg) (err(errno ? strerror(errno) : (msg))) + + + +#define HEADER_FUNCTIONS(CLASS) \ + \ +/* readelfheader() reads the ELF header into our global variable, and \ + * checks to make sure that this is in fact a file that we should be \ + * munging. \ + */ \ +static int readelfheader ## CLASS (int fd, Elf ## CLASS ## _Ehdr *ehdr) \ +{ \ + if (read(fd, ((char *)ehdr)+EI_NIDENT, sizeof(*ehdr) - EI_NIDENT) \ + != sizeof(*ehdr) - EI_NIDENT) \ + return ferr("missing or incomplete ELF header."); \ + \ + /* Verify the sizes of the ELF header and the program segment \ + * header table entries. \ + */ \ + if (EGET(ehdr->e_ehsize) != sizeof(Elf ## CLASS ## _Ehdr)) \ + return err("unrecognized ELF header size."); \ + if (EGET(ehdr->e_phentsize) != sizeof(Elf ## CLASS ## _Phdr)) \ + return err("unrecognized program segment header size."); \ + \ + /* Finally, check the file type. \ + */ \ + if (EGET(ehdr->e_type) != ET_EXEC && EGET(ehdr->e_type) != ET_DYN) \ + return err("not an executable or shared-object library."); \ + \ + return TRUE; \ +} \ + \ +/* readphdrtable() loads the program segment header table into memory. \ + */ \ +static int readphdrtable ## CLASS (int fd, Elf ## CLASS ## _Ehdr const *ehdr, \ + Elf ## CLASS ## _Phdr **phdrs) \ +{ \ + size_t size; \ + \ + if (!EGET(ehdr->e_phoff) || !EGET(ehdr->e_phnum) \ +) return err("ELF file has no program header table."); \ + \ + size = EGET(ehdr->e_phnum) * sizeof **phdrs; \ + if (!(*phdrs = malloc(size))) \ + return err("Out of memory!"); \ + \ + errno = 0; \ + if (read(fd, *phdrs, size) != (ssize_t)size) \ + return ferr("missing or incomplete program segment header table."); \ + \ + return TRUE; \ +} \ + \ +/* getmemorysize() determines the offset of the last byte of the file \ + * that is referenced by an entry in the program segment header table. \ + * (Anything in the file after that point is not used when the program \ + * is executing, and thus can be safely discarded.) \ + */ \ +static int getmemorysize ## CLASS (Elf ## CLASS ## _Ehdr const *ehdr, \ + Elf ## CLASS ## _Phdr const *phdrs, \ + unsigned long *newsize) \ +{ \ + Elf ## CLASS ## _Phdr const *phdr; \ + unsigned long size, n; \ + int i; \ + \ + /* Start by setting the size to include the ELF header and the \ + * complete program segment header table. \ + */ \ + size = EGET(ehdr->e_phoff) + EGET(ehdr->e_phnum) * sizeof *phdrs; \ + if (size < sizeof *ehdr) \ + size = sizeof *ehdr; \ + \ + /* Then keep extending the size to include whatever data the \ + * program segment header table references. \ + */ \ + for (i = 0, phdr = phdrs ; i < EGET(ehdr->e_phnum) ; ++i, ++phdr) { \ + if (EGET(phdr->p_type) != PT_NULL) { \ + n = EGET(phdr->p_offset) + EGET(phdr->p_filesz); \ + if (n > size) \ + size = n; \ + } \ + } \ + \ + *newsize = size; \ + return TRUE; \ +} \ + \ +/* modifyheaders() removes references to the section header table if \ + * it was stripped, and reduces program header table entries that \ + * included truncated bytes at the end of the file. \ + */ \ +static int modifyheaders ## CLASS (Elf ## CLASS ## _Ehdr *ehdr, \ + Elf ## CLASS ## _Phdr *phdrs, \ + unsigned long newsize) \ +{ \ + Elf ## CLASS ## _Phdr *phdr; \ + int i; \ + \ + /* If the section header table is gone, then remove all references \ + * to it in the ELF header. \ + */ \ + if (EGET(ehdr->e_shoff) >= newsize) { \ + ESET(ehdr->e_shoff,0); \ + ESET(ehdr->e_shnum,0); \ + ESET(ehdr->e_shentsize,0); \ + ESET(ehdr->e_shstrndx,0); \ + } \ + \ + /* The program adjusts the file size of any segment that was \ + * truncated. The case of a segment being completely stripped out \ + * is handled separately. \ + */ \ + for (i = 0, phdr = phdrs ; i < EGET(ehdr->e_phnum) ; ++i, ++phdr) { \ + if (EGET(phdr->p_offset) >= newsize) { \ + ESET(phdr->p_offset,newsize); \ + ESET(phdr->p_filesz,0); \ + } else if (EGET(phdr->p_offset) + EGET(phdr->p_filesz) > newsize) { \ + newsize -= EGET(phdr->p_offset); \ + ESET(phdr->p_filesz, newsize); \ + } \ + } \ + \ + return TRUE; \ +} \ + \ +/* commitchanges() writes the new headers back to the original file \ + * and sets the file to its new size. \ + */ \ +static int commitchanges ## CLASS (int fd, Elf ## CLASS ## _Ehdr const *ehdr, \ + Elf ## CLASS ## _Phdr *phdrs, \ + unsigned long newsize) \ +{ \ + size_t n; \ + \ + /* Save the changes to the ELF header, if any. \ + */ \ + if (lseek(fd, 0, SEEK_SET)) \ + return ferr("could not rewind file"); \ + errno = 0; \ + if (write(fd, ehdr, sizeof *ehdr) != sizeof *ehdr) \ + return err("could not modify file"); \ + \ + /* Save the changes to the program segment header table, if any. \ + */ \ + if (lseek(fd, EGET(ehdr->e_phoff), SEEK_SET) == (off_t)-1) { \ + err("could not seek in file."); \ + goto warning; \ + } \ + n = EGET(ehdr->e_phnum) * sizeof *phdrs; \ + if (write(fd, phdrs, n) != (ssize_t)n) { \ + err("could not write to file"); \ + goto warning; \ + } \ + \ + /* Eleventh-hour sanity check: don't truncate before the end of \ + * the program segment header table. \ + */ \ + if (newsize < EGET(ehdr->e_phoff) + n) \ + newsize = EGET(ehdr->e_phoff) + n; \ + \ + /* Chop off the end of the file. \ + */ \ + if (ftruncate(fd, newsize)) { \ + err("could not resize file"); \ + goto warning; \ + } \ + \ + return TRUE; \ + \ + warning: \ + return err("ELF file may have been corrupted!"); \ +} + + +/* First elements of Elf32_Ehdr and Elf64_Ehdr are common. + */ +static int readelfheaderident(int fd, Elf32_Ehdr *ehdr) +{ + errno = 0; + if (read(fd, ehdr, EI_NIDENT) != EI_NIDENT) + return ferr("missing or incomplete ELF header."); + + /* Check the ELF signature. + */ + if (!(ehdr->e_ident[EI_MAG0] == ELFMAG0 && + ehdr->e_ident[EI_MAG1] == ELFMAG1 && + ehdr->e_ident[EI_MAG2] == ELFMAG2 && + ehdr->e_ident[EI_MAG3] == ELFMAG3)) + { + err("missing ELF signature."); + return -1; + } + + /* Compare the file's class and endianness with the program's. + */ +#if __BYTE_ORDER == __LITTLE_ENDIAN + if (ehdr->e_ident[EI_DATA] == ELFDATA2LSB) { + do_reverse_endian = 0; + } else if (ehdr->e_ident[EI_DATA] == ELFDATA2MSB) { +/* fprintf(stderr, "ELF file has different endianness.\n"); */ + do_reverse_endian = 1; + } +#elif __BYTE_ORDER == __BIG_ENDIAN + if (ehdr->e_ident[EI_DATA] == ELFDATA2LSB) { +/* fprintf(stderr, "ELF file has different endianness.\n"); */ + do_reverse_endian = 1; + } else if (ehdr->e_ident[EI_DATA] == ELFDATA2MSB) { + do_reverse_endian = 0; + } +#else +#error unkown endianness +#endif + else { + err("Unsupported endianness"); + return -1; + } + + /* Check the target architecture. + */ +/* if (EGET(ehdr->e_machine) != ELF_ARCH) { */ +/* /\* return err("ELF file created for different architecture."); *\/ */ +/* fprintf(stderr, "ELF file created for different architecture.\n"); */ +/* } */ + return ehdr->e_ident[EI_CLASS]; +} + + +HEADER_FUNCTIONS(32) + +HEADER_FUNCTIONS(64) + +/* truncatezeros() examines the bytes at the end of the file's + * size-to-be, and reduces the size to exclude any trailing zero + * bytes. + */ +static int truncatezeros(int fd, unsigned long *newsize) +{ + unsigned char contents[1024]; + unsigned long size, n; + + size = *newsize; + do { + n = sizeof contents; + if (n > size) + n = size; + if (lseek(fd, size - n, SEEK_SET) == (off_t)-1) + return ferr("cannot seek in file."); + if (read(fd, contents, n) != (ssize_t)n) + return ferr("cannot read file contents"); + while (n && !contents[--n]) + --size; + } while (size && !n); + + /* Sanity check. + */ + if (!size) + return err("ELF file is completely blank!"); + + *newsize = size; + return TRUE; +} + +/* main() loops over the cmdline arguments, leaving all the real work + * to the other functions. + */ +int main(int argc, char *argv[]) +{ + int fd; + union { + Elf32_Ehdr ehdr32; + Elf64_Ehdr ehdr64; + } e; + union { + Elf32_Phdr *phdrs32; + Elf64_Phdr *phdrs64; + } p; + unsigned long newsize; + char **arg; + int failures = 0; + + if (argc < 2 || argv[1][0] == '-') { + printf("Usage: sstrip FILE...\n" + "sstrip discards all nonessential bytes from an executable.\n\n" + "Version 2.0-X Copyright (C) 2000,2001 Brian Raiter.\n" + "Cross-devel hacks Copyright (C) 2004 Manuel Novoa III.\n" + "This program is free software, licensed under the GNU\n" + "General Public License. There is absolutely no warranty.\n"); + return EXIT_SUCCESS; + } + + progname = argv[0]; + + for (arg = argv + 1 ; *arg != NULL ; ++arg) { + filename = *arg; + + fd = open(*arg, O_RDWR); + if (fd < 0) { + ferr("can't open"); + ++failures; + continue; + } + + switch (readelfheaderident(fd, &e.ehdr32)) { + case ELFCLASS32: + if (!(readelfheader32(fd, &e.ehdr32) && + readphdrtable32(fd, &e.ehdr32, &p.phdrs32) && + getmemorysize32(&e.ehdr32, p.phdrs32, &newsize) && + truncatezeros(fd, &newsize) && + modifyheaders32(&e.ehdr32, p.phdrs32, newsize) && + commitchanges32(fd, &e.ehdr32, p.phdrs32, newsize))) + ++failures; + break; + case ELFCLASS64: + if (!(readelfheader64(fd, &e.ehdr64) && + readphdrtable64(fd, &e.ehdr64, &p.phdrs64) && + getmemorysize64(&e.ehdr64, p.phdrs64, &newsize) && + truncatezeros(fd, &newsize) && + modifyheaders64(&e.ehdr64, p.phdrs64, newsize) && + commitchanges64(fd, &e.ehdr64, p.phdrs64, newsize))) + ++failures; + break; + default: + ++failures; + break; + } + close(fd); + } + + return failures ? EXIT_FAILURE : EXIT_SUCCESS; +} |