busybox/modutils/modutils-24.c

3951 lines
95 KiB
C
Raw Normal View History

/* vi: set sw=4 ts=4: */
/*
* Mini insmod implementation for busybox
*
* This version of insmod supports ARM, CRIS, H8/300, x86, ia64, x86_64,
* m68k, MIPS, PowerPC, S390, SH3/4/5, Sparc, v850e, and x86_64.
*
* Copyright (C) 1999-2004 by Erik Andersen <andersen@codepoet.org>
* and Ron Alder <alder@lineo.com>
*
* Rodney Radford <rradford@mindspring.com> 17-Aug-2004.
* Added x86_64 support.
*
* Miles Bader <miles@gnu.org> added NEC V850E support.
*
* Modified by Bryan Rittmeyer <bryan@ixiacom.com> to support SH4
* and (theoretically) SH3. I have only tested SH4 in little endian mode.
*
* Modified by Alcove, Julien Gaulmin <julien.gaulmin@alcove.fr> and
* Nicolas Ferre <nicolas.ferre@alcove.fr> to support ARM7TDMI. Only
* very minor changes required to also work with StrongArm and presumably
* all ARM based systems.
*
* Yoshinori Sato <ysato@users.sourceforge.jp> 19-May-2004.
* added Renesas H8/300 support.
*
* Paul Mundt <lethal@linux-sh.org> 08-Aug-2003.
* Integrated support for sh64 (SH-5), from preliminary modutils
* patches from Benedict Gaster <benedict.gaster@superh.com>.
* Currently limited to support for 32bit ABI.
*
* Magnus Damm <damm@opensource.se> 22-May-2002.
* The plt and got code are now using the same structs.
* Added generic linked list code to fully support PowerPC.
* Replaced the mess in arch_apply_relocation() with architecture blocks.
* The arch_create_got() function got cleaned up with architecture blocks.
* These blocks should be easy maintain and sync with obj_xxx.c in modutils.
*
* Magnus Damm <damm@opensource.se> added PowerPC support 20-Feb-2001.
* PowerPC specific code stolen from modutils-2.3.16,
* written by Paul Mackerras, Copyright 1996, 1997 Linux International.
* I've only tested the code on mpc8xx platforms in big-endian mode.
* Did some cleanup and added USE_xxx_ENTRIES...
*
* Quinn Jensen <jensenq@lineo.com> added MIPS support 23-Feb-2001.
* based on modutils-2.4.2
* MIPS specific support for Elf loading and relocation.
* Copyright 1996, 1997 Linux International.
* Contributed by Ralf Baechle <ralf@gnu.ai.mit.edu>
*
* Based almost entirely on the Linux modutils-2.3.11 implementation.
* Copyright 1996, 1997 Linux International.
* New implementation contributed by Richard Henderson <rth@tamu.edu>
* Based on original work by Bjorn Ekwall <bj0rn@blox.se>
* Restructured (and partly rewritten) by:
* Bj<EFBFBD>rn Ekwall <bj0rn@blox.se> February 1999
*
* Licensed under GPLv2 or later, see file LICENSE in this source tree.
*/
//kbuild:lib-$(CONFIG_FEATURE_2_4_MODULES) += modutils-24.o
#include "libbb.h"
#include "modutils.h"
#include <sys/utsname.h>
#if ENABLE_FEATURE_INSMOD_LOADINKMEM
#define LOADBITS 0
#else
#define LOADBITS 1
#endif
/* Alpha */
#if defined(__alpha__)
#define MATCH_MACHINE(x) (x == EM_ALPHA)
#define SHT_RELM SHT_RELA
#define Elf64_RelM Elf64_Rela
#define ELFCLASSM ELFCLASS64
#endif
/* ARM support */
#if defined(__arm__)
#define MATCH_MACHINE(x) (x == EM_ARM)
#define SHT_RELM SHT_REL
#define Elf32_RelM Elf32_Rel
#define ELFCLASSM ELFCLASS32
#define USE_PLT_ENTRIES
#define PLT_ENTRY_SIZE 8
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 8
#define USE_SINGLE
#endif
/* NDS32 support */
#if defined(__nds32__) || defined(__NDS32__)
#define CONFIG_USE_GOT_ENTRIES
#define CONFIG_GOT_ENTRY_SIZE 4
#define CONFIG_USE_SINGLE
#if defined(__NDS32_EB__)
#define MATCH_MACHINE(x) (x == EM_NDS32)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
#if defined(__NDS32_EL__)
#define MATCH_MACHINE(x) (x == EM_NDS32)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
#endif
/* blackfin */
#if defined(BFIN)
#define MATCH_MACHINE(x) (x == EM_BLACKFIN)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
/* CRIS */
#if defined(__cris__)
#define MATCH_MACHINE(x) (x == EM_CRIS)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#ifndef EM_CRIS
#define EM_CRIS 76
#define R_CRIS_NONE 0
#define R_CRIS_32 3
#endif
#endif
/* H8/300 */
#if defined(__H8300H__) || defined(__H8300S__)
#define MATCH_MACHINE(x) (x == EM_H8_300)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_SINGLE
#define SYMBOL_PREFIX "_"
#endif
/* PA-RISC / HP-PA */
#if defined(__hppa__)
#define MATCH_MACHINE(x) (x == EM_PARISC)
#define SHT_RELM SHT_RELA
#if defined(__LP64__)
#define Elf64_RelM Elf64_Rela
#define ELFCLASSM ELFCLASS64
#else
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
#endif
/* x86 */
#if defined(__i386__)
#ifndef EM_486
#define MATCH_MACHINE(x) (x == EM_386)
#else
#define MATCH_MACHINE(x) (x == EM_386 || x == EM_486)
#endif
#define SHT_RELM SHT_REL
#define Elf32_RelM Elf32_Rel
#define ELFCLASSM ELFCLASS32
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 4
#define USE_SINGLE
#endif
/* IA64, aka Itanium */
#if defined(__ia64__)
#define MATCH_MACHINE(x) (x == EM_IA_64)
#define SHT_RELM SHT_RELA
#define Elf64_RelM Elf64_Rela
#define ELFCLASSM ELFCLASS64
#endif
/* m68k */
#if defined(__mc68000__)
#define MATCH_MACHINE(x) (x == EM_68K)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 4
#define USE_SINGLE
#endif
/* Microblaze */
#if defined(__microblaze__)
#define USE_SINGLE
#include <linux/elf-em.h>
#define MATCH_MACHINE(x) (x == EM_XILINX_MICROBLAZE)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
/* MIPS */
#if defined(__mips__)
#define MATCH_MACHINE(x) (x == EM_MIPS || x == EM_MIPS_RS3_LE)
#define SHT_RELM SHT_REL
#define Elf32_RelM Elf32_Rel
#define ELFCLASSM ELFCLASS32
/* Account for ELF spec changes. */
#ifndef EM_MIPS_RS3_LE
#ifdef EM_MIPS_RS4_BE
#define EM_MIPS_RS3_LE EM_MIPS_RS4_BE
#else
#define EM_MIPS_RS3_LE 10
#endif
#endif /* !EM_MIPS_RS3_LE */
#define ARCHDATAM "__dbe_table"
#endif
/* Nios II */
#if defined(__nios2__)
#define MATCH_MACHINE(x) (x == EM_ALTERA_NIOS2)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
/* PowerPC */
#if defined(__powerpc64__)
#define MATCH_MACHINE(x) (x == EM_PPC64)
#define SHT_RELM SHT_RELA
#define Elf64_RelM Elf64_Rela
#define ELFCLASSM ELFCLASS64
#elif defined(__powerpc__)
#define MATCH_MACHINE(x) (x == EM_PPC)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_PLT_ENTRIES
#define PLT_ENTRY_SIZE 16
#define USE_PLT_LIST
#define LIST_ARCHTYPE ElfW(Addr)
#define USE_LIST
#define ARCHDATAM "__ftr_fixup"
#endif
/* S390 */
#if defined(__s390__)
#define MATCH_MACHINE(x) (x == EM_S390)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_PLT_ENTRIES
#define PLT_ENTRY_SIZE 8
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 8
#define USE_SINGLE
#endif
/* SuperH */
#if defined(__sh__)
#define MATCH_MACHINE(x) (x == EM_SH)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 4
#define USE_SINGLE
/* the SH changes have only been tested in =little endian= mode */
/* I'm not sure about big endian, so let's warn: */
#if defined(__sh__) && BB_BIG_ENDIAN
# error insmod.c may require changes for use on big endian SH
#endif
/* it may or may not work on the SH1/SH2... Error on those also */
#if ((!(defined(__SH3__) || defined(__SH4__) || defined(__SH5__)))) && (defined(__sh__))
#error insmod.c may require changes for SH1 or SH2 use
#endif
#endif
/* Sparc */
#if defined(__sparc__)
#define MATCH_MACHINE(x) (x == EM_SPARC)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#endif
/* v850e */
#if defined(__v850e__)
#define MATCH_MACHINE(x) ((x) == EM_V850 || (x) == EM_CYGNUS_V850)
#define SHT_RELM SHT_RELA
#define Elf32_RelM Elf32_Rela
#define ELFCLASSM ELFCLASS32
#define USE_PLT_ENTRIES
#define PLT_ENTRY_SIZE 8
#define USE_SINGLE
#ifndef EM_CYGNUS_V850 /* grumble */
#define EM_CYGNUS_V850 0x9080
#endif
#define SYMBOL_PREFIX "_"
#endif
/* X86_64 */
#if defined(__x86_64__)
#define MATCH_MACHINE(x) (x == EM_X86_64)
#define SHT_RELM SHT_RELA
#define USE_GOT_ENTRIES
#define GOT_ENTRY_SIZE 8
#define USE_SINGLE
#define Elf64_RelM Elf64_Rela
#define ELFCLASSM ELFCLASS64
#endif
#ifndef SHT_RELM
#error Sorry, but insmod.c does not yet support this architecture...
#endif
//----------------------------------------------------------------------------
//--------modutils module.h, lines 45-242
//----------------------------------------------------------------------------
/* Definitions for the Linux module syscall interface.
Copyright 1996, 1997 Linux International.
Contributed by Richard Henderson <rth@tamu.edu>
This file is part of the Linux modutils.
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software Foundation,
Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
#ifndef MODUTILS_MODULE_H
/*======================================================================*/
/* For sizeof() which are related to the module platform and not to the
environment isnmod is running in, use sizeof_xx instead of sizeof(xx). */
#define tgt_sizeof_char sizeof(char)
#define tgt_sizeof_short sizeof(short)
#define tgt_sizeof_int sizeof(int)
#define tgt_sizeof_long sizeof(long)
#define tgt_sizeof_char_p sizeof(char *)
#define tgt_sizeof_void_p sizeof(void *)
#define tgt_long long
#if defined(__sparc__) && !defined(__sparc_v9__) && defined(ARCH_sparc64)
#undef tgt_sizeof_long
#undef tgt_sizeof_char_p
#undef tgt_sizeof_void_p
#undef tgt_long
enum {
tgt_sizeof_long = 8,
tgt_sizeof_char_p = 8,
tgt_sizeof_void_p = 8
};
#define tgt_long long long
#endif
/*======================================================================*/
/* The structures used in Linux 2.1. */
/* Note: new_module_symbol does not use tgt_long intentionally */
struct new_module_symbol {
unsigned long value;
unsigned long name;
};
struct new_module_persist;
struct new_module_ref {
unsigned tgt_long dep; /* kernel addresses */
unsigned tgt_long ref;
unsigned tgt_long next_ref;
};
struct new_module {
unsigned tgt_long size_of_struct; /* == sizeof(module) */
unsigned tgt_long next;
unsigned tgt_long name;
unsigned tgt_long size;
tgt_long usecount;
unsigned tgt_long flags; /* AUTOCLEAN et al */
unsigned nsyms;
unsigned ndeps;
unsigned tgt_long syms;
unsigned tgt_long deps;
unsigned tgt_long refs;
unsigned tgt_long init;
unsigned tgt_long cleanup;
unsigned tgt_long ex_table_start;
unsigned tgt_long ex_table_end;
#ifdef __alpha__
unsigned tgt_long gp;
#endif
/* Everything after here is extension. */
unsigned tgt_long persist_start;
unsigned tgt_long persist_end;
unsigned tgt_long can_unload;
unsigned tgt_long runsize;
const char *kallsyms_start; /* All symbols for kernel debugging */
const char *kallsyms_end;
const char *archdata_start; /* arch specific data for module */
const char *archdata_end;
const char *kernel_data; /* Reserved for kernel internal use */
};
#ifdef ARCHDATAM
#define ARCHDATA_SEC_NAME ARCHDATAM
#else
#define ARCHDATA_SEC_NAME "__archdata"
#endif
#define KALLSYMS_SEC_NAME "__kallsyms"
struct new_module_info {
unsigned long addr;
unsigned long size;
unsigned long flags;
long usecount;
};
/* Bits of module.flags. */
enum {
NEW_MOD_RUNNING = 1,
NEW_MOD_DELETED = 2,
NEW_MOD_AUTOCLEAN = 4,
NEW_MOD_VISITED = 8,
NEW_MOD_USED_ONCE = 16
};
int init_module(const char *name, const struct new_module *);
int query_module(const char *name, int which, void *buf,
size_t bufsize, size_t *ret);
/* Values for query_module's which. */
enum {
QM_MODULES = 1,
QM_DEPS = 2,
QM_REFS = 3,
QM_SYMBOLS = 4,
QM_INFO = 5
};
/*======================================================================*/
/* The system calls unchanged between 2.0 and 2.1. */
unsigned long create_module(const char *, size_t);
int delete_module(const char *module, unsigned int flags);
#endif /* module.h */
//----------------------------------------------------------------------------
//--------end of modutils module.h
//----------------------------------------------------------------------------
//----------------------------------------------------------------------------
//--------modutils obj.h, lines 253-462
//----------------------------------------------------------------------------
/* Elf object file loading and relocation routines.
Copyright 1996, 1997 Linux International.
Contributed by Richard Henderson <rth@tamu.edu>
This file is part of the Linux modutils.
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software Foundation,
Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
#ifndef MODUTILS_OBJ_H
/* The relocatable object is manipulated using elfin types. */
#include <elf.h>
#include <endian.h>
#ifndef ElfW
# if ELFCLASSM == ELFCLASS32
# define ElfW(x) Elf32_ ## x
# define ELFW(x) ELF32_ ## x
# else
# define ElfW(x) Elf64_ ## x
# define ELFW(x) ELF64_ ## x
# endif
#endif
/* For some reason this is missing from some ancient C libraries.... */
#ifndef ELF32_ST_INFO
# define ELF32_ST_INFO(bind, type) (((bind) << 4) + ((type) & 0xf))
#endif
#ifndef ELF64_ST_INFO
# define ELF64_ST_INFO(bind, type) (((bind) << 4) + ((type) & 0xf))
#endif
#define ELF_ST_BIND(info) ELFW(ST_BIND)(info)
#define ELF_ST_TYPE(info) ELFW(ST_TYPE)(info)
#define ELF_ST_INFO(bind, type) ELFW(ST_INFO)(bind, type)
#define ELF_R_TYPE(val) ELFW(R_TYPE)(val)
#define ELF_R_SYM(val) ELFW(R_SYM)(val)
struct obj_string_patch;
struct obj_symbol_patch;
struct obj_section {
ElfW(Shdr) header;
const char *name;
char *contents;
struct obj_section *load_next;
int idx;
};
struct obj_symbol {
struct obj_symbol *next; /* hash table link */
const char *name;
unsigned long value;
unsigned long size;
int secidx; /* the defining section index/module */
int info;
int ksymidx; /* for export to the kernel symtab */
int referenced; /* actually used in the link */
};
/* Hardcode the hash table size. We shouldn't be needing so many
symbols that we begin to degrade performance, and we get a big win
by giving the compiler a constant divisor. */
#define HASH_BUCKETS 521
struct obj_file {
ElfW(Ehdr) header;
ElfW(Addr) baseaddr;
struct obj_section **sections;
struct obj_section *load_order;
struct obj_section **load_order_search_start;
struct obj_string_patch *string_patches;
struct obj_symbol_patch *symbol_patches;
int (*symbol_cmp)(const char *, const char *); /* cant be FAST_FUNC */
unsigned long (*symbol_hash)(const char *) FAST_FUNC;
unsigned long local_symtab_size;
struct obj_symbol **local_symtab;
struct obj_symbol *symtab[HASH_BUCKETS];
};
enum obj_reloc {
obj_reloc_ok,
obj_reloc_overflow,
obj_reloc_dangerous,
obj_reloc_unhandled
};
struct obj_string_patch {
struct obj_string_patch *next;
int reloc_secidx;
ElfW(Addr) reloc_offset;
ElfW(Addr) string_offset;
};
struct obj_symbol_patch {
struct obj_symbol_patch *next;
int reloc_secidx;
ElfW(Addr) reloc_offset;
struct obj_symbol *sym;
};
/* Generic object manipulation routines. */
static unsigned long FAST_FUNC obj_elf_hash(const char *);
static unsigned long obj_elf_hash_n(const char *, unsigned long len);
static struct obj_symbol *obj_find_symbol(struct obj_file *f,
const char *name);
static ElfW(Addr) obj_symbol_final_value(struct obj_file *f,
struct obj_symbol *sym);
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
static void obj_set_symbol_compare(struct obj_file *f,
int (*cmp)(const char *, const char *),
unsigned long (*hash)(const char *) FAST_FUNC);
#endif
static struct obj_section *obj_find_section(struct obj_file *f,
const char *name);
static void obj_insert_section_load_order(struct obj_file *f,
struct obj_section *sec);
static struct obj_section *obj_create_alloced_section(struct obj_file *f,
const char *name,
unsigned long align,
unsigned long size);
static struct obj_section *obj_create_alloced_section_first(struct obj_file *f,
const char *name,
unsigned long align,
unsigned long size);
static void *obj_extend_section(struct obj_section *sec, unsigned long more);
static void obj_string_patch(struct obj_file *f, int secidx, ElfW(Addr) offset,
const char *string);
static void obj_symbol_patch(struct obj_file *f, int secidx, ElfW(Addr) offset,
struct obj_symbol *sym);
static void obj_check_undefineds(struct obj_file *f);
static void obj_allocate_commons(struct obj_file *f);
static unsigned long obj_load_size(struct obj_file *f);
static int obj_relocate(struct obj_file *f, ElfW(Addr) base);
#if !LOADBITS
#define obj_load(image, image_size, loadprogbits) \
obj_load(image, image_size)
#endif
static struct obj_file *obj_load(char *image, size_t image_size, int loadprogbits);
static int obj_create_image(struct obj_file *f, char *image);
/* Architecture specific manipulation routines. */
static struct obj_file *arch_new_file(void);
static struct obj_section *arch_new_section(void);
static struct obj_symbol *arch_new_symbol(void);
static enum obj_reloc arch_apply_relocation(struct obj_file *f,
struct obj_section *targsec,
/*struct obj_section *symsec,*/
struct obj_symbol *sym,
ElfW(RelM) *rel, ElfW(Addr) value);
static void arch_create_got(struct obj_file *f);
#if ENABLE_FEATURE_CHECK_TAINTED_MODULE
static int obj_gpl_license(struct obj_file *f, const char **license);
#endif
#endif /* obj.h */
//----------------------------------------------------------------------------
//--------end of modutils obj.h
//----------------------------------------------------------------------------
/* SPFX is always a string, so it can be concatenated to string constants. */
#ifdef SYMBOL_PREFIX
#define SPFX SYMBOL_PREFIX
#else
#define SPFX ""
#endif
enum { STRVERSIONLEN = 64 };
/*======================================================================*/
#define flag_force_load (option_mask32 & INSMOD_OPT_FORCE)
#define flag_autoclean (option_mask32 & INSMOD_OPT_KERNELD)
#define flag_verbose (option_mask32 & INSMOD_OPT_VERBOSE)
#define flag_quiet (option_mask32 & INSMOD_OPT_SILENT)
#define flag_noexport (option_mask32 & INSMOD_OPT_NO_EXPORT)
#define flag_print_load_map (option_mask32 & INSMOD_OPT_PRINT_MAP)
/*======================================================================*/
#if defined(USE_LIST)
struct arch_list_entry {
struct arch_list_entry *next;
LIST_ARCHTYPE addend;
int offset;
int inited : 1;
};
#endif
#if defined(USE_SINGLE)
struct arch_single_entry {
int offset;
int inited : 1;
int allocated : 1;
};
#endif
#if defined(__mips__)
struct mips_hi16 {
struct mips_hi16 *next;
ElfW(Addr) *addr;
ElfW(Addr) value;
};
#endif
struct arch_file {
struct obj_file root;
#if defined(USE_PLT_ENTRIES)
struct obj_section *plt;
#endif
#if defined(USE_GOT_ENTRIES)
struct obj_section *got;
#endif
#if defined(__mips__)
struct mips_hi16 *mips_hi16_list;
#endif
};
struct arch_symbol {
struct obj_symbol root;
#if defined(USE_PLT_ENTRIES)
#if defined(USE_PLT_LIST)
struct arch_list_entry *pltent;
#else
struct arch_single_entry pltent;
#endif
#endif
#if defined(USE_GOT_ENTRIES)
struct arch_single_entry gotent;
#endif
};
struct external_module {
const char *name;
ElfW(Addr) addr;
int used;
size_t nsyms;
struct new_module_symbol *syms;
};
static struct new_module_symbol *ksyms;
static size_t nksyms;
static struct external_module *ext_modules;
static int n_ext_modules;
static int n_ext_modules_used;
/*======================================================================*/
static struct obj_file *arch_new_file(void)
{
struct arch_file *f;
f = xzalloc(sizeof(*f));
return &f->root; /* it's a first member */
}
static struct obj_section *arch_new_section(void)
{
return xzalloc(sizeof(struct obj_section));
}
static struct obj_symbol *arch_new_symbol(void)
{
struct arch_symbol *sym;
sym = xzalloc(sizeof(*sym));
return &sym->root;
}
static enum obj_reloc
arch_apply_relocation(struct obj_file *f,
struct obj_section *targsec,
/*struct obj_section *symsec,*/
struct obj_symbol *sym,
ElfW(RelM) *rel, ElfW(Addr) v)
{
#if defined(__arm__) || defined(__i386__) || defined(__mc68000__) \
|| defined(__sh__) || defined(__s390__) || defined(__x86_64__) \
|| defined(__powerpc__) || defined(__mips__)
struct arch_file *ifile = (struct arch_file *) f;
#endif
enum obj_reloc ret = obj_reloc_ok;
ElfW(Addr) *loc = (ElfW(Addr) *) (targsec->contents + rel->r_offset);
#if defined(__arm__) || defined(__H8300H__) || defined(__H8300S__) \
|| defined(__i386__) || defined(__mc68000__) || defined(__microblaze__) \
|| defined(__mips__) || defined(__nios2__) || defined(__powerpc__) \
|| defined(__s390__) || defined(__sh__) || defined(__x86_64__)
ElfW(Addr) dot = targsec->header.sh_addr + rel->r_offset;
#endif
#if defined(USE_GOT_ENTRIES) || defined(USE_PLT_ENTRIES)
struct arch_symbol *isym = (struct arch_symbol *) sym;
#endif
#if defined(__arm__) || defined(__i386__) || defined(__mc68000__) \
|| defined(__sh__) || defined(__s390__)
#if defined(USE_GOT_ENTRIES)
ElfW(Addr) got = ifile->got ? ifile->got->header.sh_addr : 0;
#endif
#endif
#if defined(USE_PLT_ENTRIES)
ElfW(Addr) plt = ifile->plt ? ifile->plt->header.sh_addr : 0;
unsigned long *ip;
# if defined(USE_PLT_LIST)
struct arch_list_entry *pe;
# else
struct arch_single_entry *pe;
# endif
#endif
switch (ELF_R_TYPE(rel->r_info)) {
#if defined(__arm__)
case R_ARM_NONE:
break;
case R_ARM_ABS32:
*loc += v;
break;
case R_ARM_GOT32:
goto bb_use_got;
case R_ARM_GOTPC:
/* relative reloc, always to _GLOBAL_OFFSET_TABLE_
* (which is .got) similar to branch,
* but is full 32 bits relative */
*loc += got - dot;
break;
case R_ARM_PC24:
case R_ARM_PLT32:
goto bb_use_plt;
case R_ARM_GOTOFF: /* address relative to the got */
*loc += v - got;
break;
#elif defined(__cris__)
case R_CRIS_NONE:
break;
case R_CRIS_32:
/* CRIS keeps the relocation value in the r_addend field and
* should not use whats in *loc at all
*/
*loc = v;
break;
#elif defined(__H8300H__) || defined(__H8300S__)
case R_H8_DIR24R8:
loc = (ElfW(Addr) *)((ElfW(Addr))loc - 1);
*loc = (*loc & 0xff000000) | ((*loc & 0xffffff) + v);
break;
case R_H8_DIR24A8:
*loc += v;
break;
case R_H8_DIR32:
case R_H8_DIR32A16:
*loc += v;
break;
case R_H8_PCREL16:
v -= dot + 2;
if ((ElfW(Sword))v > 0x7fff
|| (ElfW(Sword))v < -(ElfW(Sword))0x8000
) {
ret = obj_reloc_overflow;
} else {
*(unsigned short *)loc = v;
}
break;
case R_H8_PCREL8:
v -= dot + 1;
if ((ElfW(Sword))v > 0x7f
|| (ElfW(Sword))v < -(ElfW(Sword))0x80
) {
ret = obj_reloc_overflow;
} else {
*(unsigned char *)loc = v;
}
break;
#elif defined(__i386__)
case R_386_NONE:
break;
case R_386_32:
*loc += v;
break;
case R_386_PLT32:
case R_386_PC32:
case R_386_GOTOFF:
*loc += v - dot;
break;
case R_386_GLOB_DAT:
case R_386_JMP_SLOT:
*loc = v;
break;
case R_386_RELATIVE:
*loc += f->baseaddr;
break;
case R_386_GOTPC:
*loc += got - dot;
break;
case R_386_GOT32:
goto bb_use_got;
break;
#elif defined(__microblaze__)
case R_MICROBLAZE_NONE:
case R_MICROBLAZE_64_NONE:
case R_MICROBLAZE_32_SYM_OP_SYM:
case R_MICROBLAZE_32_PCREL:
break;
case R_MICROBLAZE_64_PCREL: {
/* dot is the address of the current instruction.
* v is the target symbol address.
* So we need to extract the offset in the code,
* adding v, then subtrating the current address
* of this instruction.
* Ex: "IMM 0xFFFE bralid 0x0000" = "bralid 0xFFFE0000"
*/
/* Get split offset stored in code */
unsigned int temp = (loc[0] & 0xFFFF) << 16 |
(loc[1] & 0xFFFF);
/* Adjust relative offset. -4 adjustment required
* because dot points to the IMM insn, but branch
* is computed relative to the branch instruction itself.
*/
temp += v - dot - 4;
/* Store back into code */
loc[0] = (loc[0] & 0xFFFF0000) | temp >> 16;
loc[1] = (loc[1] & 0xFFFF0000) | (temp & 0xFFFF);
break;
}
case R_MICROBLAZE_32:
*loc += v;
break;
case R_MICROBLAZE_64: {
/* Get split pointer stored in code */
unsigned int temp1 = (loc[0] & 0xFFFF) << 16 |
(loc[1] & 0xFFFF);
/* Add reloc offset */
temp1+=v;
/* Store back into code */
loc[0] = (loc[0] & 0xFFFF0000) | temp1 >> 16;
loc[1] = (loc[1] & 0xFFFF0000) | (temp1 & 0xFFFF);
break;
}
case R_MICROBLAZE_32_PCREL_LO:
case R_MICROBLAZE_32_LO:
case R_MICROBLAZE_SRO32:
case R_MICROBLAZE_SRW32:
ret = obj_reloc_unhandled;
break;
#elif defined(__mc68000__)
case R_68K_NONE:
break;
case R_68K_32:
*loc += v;
break;
case R_68K_8:
if (v > 0xff) {
ret = obj_reloc_overflow;
}
*(char *)loc = v;
break;
case R_68K_16:
if (v > 0xffff) {
ret = obj_reloc_overflow;
}
*(short *)loc = v;
break;
case R_68K_PC8:
v -= dot;
if ((ElfW(Sword))v > 0x7f
|| (ElfW(Sword))v < -(ElfW(Sword))0x80
) {
ret = obj_reloc_overflow;
}
*(char *)loc = v;
break;
case R_68K_PC16:
v -= dot;
if ((ElfW(Sword))v > 0x7fff
|| (ElfW(Sword))v < -(ElfW(Sword))0x8000
) {
ret = obj_reloc_overflow;
}
*(short *)loc = v;
break;
case R_68K_PC32:
*(int *)loc = v - dot;
break;
case R_68K_GLOB_DAT:
case R_68K_JMP_SLOT:
*loc = v;
break;
case R_68K_RELATIVE:
*(int *)loc += f->baseaddr;
break;
case R_68K_GOT32:
goto bb_use_got;
# ifdef R_68K_GOTOFF
case R_68K_GOTOFF:
*loc += v - got;
break;
# endif
#elif defined(__mips__)
case R_MIPS_NONE:
break;
case R_MIPS_32:
*loc += v;
break;
case R_MIPS_26:
if (v % 4)
ret = obj_reloc_dangerous;
if ((v & 0xf0000000) != ((dot + 4) & 0xf0000000))
ret = obj_reloc_overflow;
*loc =
(*loc & ~0x03ffffff) | ((*loc + (v >> 2)) &
0x03ffffff);
break;
case R_MIPS_HI16:
{
struct mips_hi16 *n;
/* We cannot relocate this one now because we don't know the value
of the carry we need to add. Save the information, and let LO16
do the actual relocation. */
n = xmalloc(sizeof *n);
n->addr = loc;
n->value = v;
n->next = ifile->mips_hi16_list;
ifile->mips_hi16_list = n;
break;
}
case R_MIPS_LO16:
{
unsigned long insnlo = *loc;
ElfW(Addr) val, vallo;
/* Sign extend the addend we extract from the lo insn. */
vallo = ((insnlo & 0xffff) ^ 0x8000) - 0x8000;
if (ifile->mips_hi16_list != NULL) {
struct mips_hi16 *l;
l = ifile->mips_hi16_list;
while (l != NULL) {
struct mips_hi16 *next;
unsigned long insn;
/* Do the HI16 relocation. Note that we actually don't
need to know anything about the LO16 itself, except where
to find the low 16 bits of the addend needed by the LO16. */
insn = *l->addr;
val =
((insn & 0xffff) << 16) +
vallo;
val += v;
/* Account for the sign extension that will happen in the
low bits. */
val =
((val >> 16) +
((val & 0x8000) !=
0)) & 0xffff;
insn = (insn & ~0xffff) | val;
*l->addr = insn;
next = l->next;
free(l);
l = next;
}
ifile->mips_hi16_list = NULL;
}
/* Ok, we're done with the HI16 relocs. Now deal with the LO16. */
val = v + vallo;
insnlo = (insnlo & ~0xffff) | (val & 0xffff);
*loc = insnlo;
break;
}
#elif defined(__nios2__)
case R_NIOS2_NONE:
break;
case R_NIOS2_BFD_RELOC_32:
*loc += v;
break;
case R_NIOS2_BFD_RELOC_16:
if (v > 0xffff) {
ret = obj_reloc_overflow;
}
*(short *)loc = v;
break;
case R_NIOS2_BFD_RELOC_8:
if (v > 0xff) {
ret = obj_reloc_overflow;
}
*(char *)loc = v;
break;
case R_NIOS2_S16:
{
Elf32_Addr word;
if ((Elf32_Sword)v > 0x7fff
|| (Elf32_Sword)v < -(Elf32_Sword)0x8000
) {
ret = obj_reloc_overflow;
}
word = *loc;
*loc = ((((word >> 22) << 16) | (v & 0xffff)) << 6) |
(word & 0x3f);
}
break;
case R_NIOS2_U16:
{
Elf32_Addr word;
if (v > 0xffff) {
ret = obj_reloc_overflow;
}
word = *loc;
*loc = ((((word >> 22) << 16) | (v & 0xffff)) << 6) |
(word & 0x3f);
}
break;
case R_NIOS2_PCREL16:
{
Elf32_Addr word;
v -= dot + 4;
if ((Elf32_Sword)v > 0x7fff
|| (Elf32_Sword)v < -(Elf32_Sword)0x8000
) {
ret = obj_reloc_overflow;
}
word = *loc;
*loc = ((((word >> 22) << 16) | (v & 0xffff)) << 6) | (word & 0x3f);
}
break;
case R_NIOS2_GPREL:
{
Elf32_Addr word, gp;
/* get _gp */
gp = obj_symbol_final_value(f, obj_find_symbol(f, SPFX "_gp"));
v -= gp;
if ((Elf32_Sword)v > 0x7fff
|| (Elf32_Sword)v < -(Elf32_Sword)0x8000
) {
ret = obj_reloc_overflow;
}
word = *loc;
*loc = ((((word >> 22) << 16) | (v & 0xffff)) << 6) | (word & 0x3f);
}
break;
case R_NIOS2_CALL26:
if (v & 3)
ret = obj_reloc_dangerous;
if ((v >> 28) != (dot >> 28))
ret = obj_reloc_overflow;
*loc = (*loc & 0x3f) | ((v >> 2) << 6);
break;
case R_NIOS2_IMM5:
{
Elf32_Addr word;
if (v > 0x1f) {
ret = obj_reloc_overflow;
}
word = *loc & ~0x7c0;
*loc = word | ((v & 0x1f) << 6);
}
break;
case R_NIOS2_IMM6:
{
Elf32_Addr word;
if (v > 0x3f) {
ret = obj_reloc_overflow;
}
word = *loc & ~0xfc0;
*loc = word | ((v & 0x3f) << 6);
}
break;
case R_NIOS2_IMM8:
{
Elf32_Addr word;
if (v > 0xff) {
ret = obj_reloc_overflow;
}
word = *loc & ~0x3fc0;
*loc = word | ((v & 0xff) << 6);
}
break;
case R_NIOS2_HI16:
{
Elf32_Addr word;
word = *loc;
*loc = ((((word >> 22) << 16) | ((v >>16) & 0xffff)) << 6) |
(word & 0x3f);
}
break;
case R_NIOS2_LO16:
{
Elf32_Addr word;
word = *loc;
*loc = ((((word >> 22) << 16) | (v & 0xffff)) << 6) |
(word & 0x3f);
}
break;
case R_NIOS2_HIADJ16:
{
Elf32_Addr word1, word2;
word1 = *loc;
word2 = ((v >> 16) + ((v >> 15) & 1)) & 0xffff;
*loc = ((((word1 >> 22) << 16) | word2) << 6) |
(word1 & 0x3f);
}
break;
#elif defined(__powerpc64__)
/* PPC64 needs a 2.6 kernel, 2.4 module relocation irrelevant */
#elif defined(__powerpc__)
case R_PPC_ADDR16_HA:
*(unsigned short *)loc = (v + 0x8000) >> 16;
break;
case R_PPC_ADDR16_HI:
*(unsigned short *)loc = v >> 16;
break;
case R_PPC_ADDR16_LO:
*(unsigned short *)loc = v;
break;
case R_PPC_REL24:
goto bb_use_plt;
case R_PPC_REL32:
*loc = v - dot;
break;
case R_PPC_ADDR32:
*loc = v;
break;
#elif defined(__s390__)
case R_390_32:
*(unsigned int *) loc += v;
break;
case R_390_16:
*(unsigned short *) loc += v;
break;
case R_390_8:
*(unsigned char *) loc += v;
break;
case R_390_PC32:
*(unsigned int *) loc += v - dot;
break;
case R_390_PC16DBL:
*(unsigned short *) loc += (v - dot) >> 1;
break;
case R_390_PC16:
*(unsigned short *) loc += v - dot;
break;
case R_390_PLT32:
case R_390_PLT16DBL:
/* find the plt entry and initialize it. */
pe = (struct arch_single_entry *) &isym->pltent;
if (pe->inited == 0) {
ip = (unsigned long *)(ifile->plt->contents + pe->offset);
ip[0] = 0x0d105810; /* basr 1,0; lg 1,10(1); br 1 */
ip[1] = 0x100607f1;
if (ELF_R_TYPE(rel->r_info) == R_390_PLT16DBL)
ip[2] = v - 2;
else
ip[2] = v;
pe->inited = 1;
}
/* Insert relative distance to target. */
v = plt + pe->offset - dot;
if (ELF_R_TYPE(rel->r_info) == R_390_PLT32)
*(unsigned int *) loc = (unsigned int) v;
else if (ELF_R_TYPE(rel->r_info) == R_390_PLT16DBL)
*(unsigned short *) loc = (unsigned short) ((v + 2) >> 1);
break;
case R_390_GLOB_DAT:
case R_390_JMP_SLOT:
*loc = v;
break;
case R_390_RELATIVE:
*loc += f->baseaddr;
break;
case R_390_GOTPC:
*(unsigned long *) loc += got - dot;
break;
case R_390_GOT12:
case R_390_GOT16:
case R_390_GOT32:
if (!isym->gotent.inited)
{
isym->gotent.inited = 1;
*(ElfW(Addr) *)(ifile->got->contents + isym->gotent.offset) = v;
}
if (ELF_R_TYPE(rel->r_info) == R_390_GOT12)
*(unsigned short *) loc |= (*(unsigned short *) loc + isym->gotent.offset) & 0xfff;
else if (ELF_R_TYPE(rel->r_info) == R_390_GOT16)
*(unsigned short *) loc += isym->gotent.offset;
else if (ELF_R_TYPE(rel->r_info) == R_390_GOT32)
*(unsigned int *) loc += isym->gotent.offset;
break;
# ifndef R_390_GOTOFF32
# define R_390_GOTOFF32 R_390_GOTOFF
# endif
case R_390_GOTOFF32:
*loc += v - got;
break;
#elif defined(__sh__)
case R_SH_NONE:
break;
case R_SH_DIR32:
*loc += v;
break;
case R_SH_REL32:
*loc += v - dot;
break;
case R_SH_PLT32:
*loc = v - dot;
break;
case R_SH_GLOB_DAT:
case R_SH_JMP_SLOT:
*loc = v;
break;
case R_SH_RELATIVE:
*loc = f->baseaddr + rel->r_addend;
break;
case R_SH_GOTPC:
*loc = got - dot + rel->r_addend;
break;
case R_SH_GOT32:
goto bb_use_got;
case R_SH_GOTOFF:
*loc = v - got;
break;
# if defined(__SH5__)
case R_SH_IMM_MEDLOW16:
case R_SH_IMM_LOW16:
{
ElfW(Addr) word;
if (ELF_R_TYPE(rel->r_info) == R_SH_IMM_MEDLOW16)
v >>= 16;
/*
* movi and shori have the format:
*
* | op | imm | reg | reserved |
* 31..26 25..10 9.. 4 3 .. 0
*
* so we simply mask and or in imm.
*/
word = *loc & ~0x3fffc00;
word |= (v & 0xffff) << 10;
*loc = word;
break;
}
case R_SH_IMM_MEDLOW16_PCREL:
case R_SH_IMM_LOW16_PCREL:
{
ElfW(Addr) word;
word = *loc & ~0x3fffc00;
v -= dot;
if (ELF_R_TYPE(rel->r_info) == R_SH_IMM_MEDLOW16_PCREL)
v >>= 16;
word |= (v & 0xffff) << 10;
*loc = word;
break;
}
# endif /* __SH5__ */
#elif defined(__v850e__)
case R_V850_NONE:
break;
case R_V850_32:
/* We write two shorts instead of a long because even
32-bit insns only need half-word alignment, but
32-bit data needs to be long-word aligned. */
v += ((unsigned short *)loc)[0];
v += ((unsigned short *)loc)[1] << 16;
((unsigned short *)loc)[0] = v & 0xffff;
((unsigned short *)loc)[1] = (v >> 16) & 0xffff;
break;
case R_V850_22_PCREL:
goto bb_use_plt;
#elif defined(__x86_64__)
case R_X86_64_NONE:
break;
case R_X86_64_64:
*loc += v;
break;
case R_X86_64_32:
*(unsigned int *) loc += v;
if (v > 0xffffffff)
{
ret = obj_reloc_overflow; /* Kernel module compiled without -mcmodel=kernel. */
/* error("Possibly is module compiled without -mcmodel=kernel!"); */
}
break;
case R_X86_64_32S:
*(signed int *) loc += v;
break;
case R_X86_64_16:
*(unsigned short *) loc += v;
break;
case R_X86_64_8:
*(unsigned char *) loc += v;
break;
case R_X86_64_PC32:
*(unsigned int *) loc += v - dot;
break;
case R_X86_64_PC16:
*(unsigned short *) loc += v - dot;
break;
case R_X86_64_PC8:
*(unsigned char *) loc += v - dot;
break;
case R_X86_64_GLOB_DAT:
case R_X86_64_JUMP_SLOT:
*loc = v;
break;
case R_X86_64_RELATIVE:
*loc += f->baseaddr;
break;
case R_X86_64_GOT32:
case R_X86_64_GOTPCREL:
goto bb_use_got;
# if 0
if (!isym->gotent.reloc_done)
{
isym->gotent.reloc_done = 1;
*(Elf64_Addr *)(ifile->got->contents + isym->gotent.offset) = v;
}
/* XXX are these really correct? */
if (ELF64_R_TYPE(rel->r_info) == R_X86_64_GOTPCREL)
*(unsigned int *) loc += v + isym->gotent.offset;
else
*loc += isym->gotent.offset;
break;
# endif
#else
# warning "no idea how to handle relocations on your arch"
#endif
default:
printf("Warning: unhandled reloc %d\n", (int)ELF_R_TYPE(rel->r_info));
ret = obj_reloc_unhandled;
break;
#if defined(USE_PLT_ENTRIES)
bb_use_plt:
/* find the plt entry and initialize it if necessary */
#if defined(USE_PLT_LIST)
for (pe = isym->pltent; pe != NULL && pe->addend != rel->r_addend;)
pe = pe->next;
#else
pe = &isym->pltent;
#endif
if (! pe->inited) {
ip = (unsigned long *) (ifile->plt->contents + pe->offset);
/* generate some machine code */
#if defined(__arm__)
ip[0] = 0xe51ff004; /* ldr pc,[pc,#-4] */
ip[1] = v; /* sym@ */
#endif
#if defined(__powerpc__)
ip[0] = 0x3d600000 + ((v + 0x8000) >> 16); /* lis r11,sym@ha */
ip[1] = 0x396b0000 + (v & 0xffff); /* addi r11,r11,sym@l */
ip[2] = 0x7d6903a6; /* mtctr r11 */
ip[3] = 0x4e800420; /* bctr */
#endif
#if defined(__v850e__)
/* We have to trash a register, so we assume that any control
transfer more than 21-bits away must be a function call
(so we can use a call-clobbered register). */
ip[0] = 0x0621 + ((v & 0xffff) << 16); /* mov sym, r1 ... */
ip[1] = ((v >> 16) & 0xffff) + 0x610000; /* ...; jmp r1 */
#endif
pe->inited = 1;
}
/* relative distance to target */
v -= dot;
/* if the target is too far away.... */
#if defined(__arm__) || defined(__powerpc__)
if ((int)v < -0x02000000 || (int)v >= 0x02000000)
#elif defined(__v850e__)
if ((ElfW(Sword))v > 0x1fffff || (ElfW(Sword))v < (ElfW(Sword))-0x200000)
#endif
/* go via the plt */
v = plt + pe->offset - dot;
#if defined(__v850e__)
if (v & 1)
#else
if (v & 3)
#endif
ret = obj_reloc_dangerous;
/* merge the offset into the instruction. */
#if defined(__arm__)
/* Convert to words. */
v >>= 2;
*loc = (*loc & ~0x00ffffff) | ((v + *loc) & 0x00ffffff);
#endif
#if defined(__powerpc__)
*loc = (*loc & ~0x03fffffc) | (v & 0x03fffffc);
#endif
#if defined(__v850e__)
/* We write two shorts instead of a long because even 32-bit insns
only need half-word alignment, but the 32-bit data write needs
to be long-word aligned. */
((unsigned short *)loc)[0] =
(*(unsigned short *)loc & 0xffc0) /* opcode + reg */
| ((v >> 16) & 0x3f); /* offs high part */
((unsigned short *)loc)[1] =
(v & 0xffff); /* offs low part */
#endif
break;
#endif /* USE_PLT_ENTRIES */
#if defined(USE_GOT_ENTRIES)
bb_use_got:
/* needs an entry in the .got: set it, once */
if (!isym->gotent.inited) {
isym->gotent.inited = 1;
*(ElfW(Addr) *) (ifile->got->contents + isym->gotent.offset) = v;
}
/* make the reloc with_respect_to_.got */
#if defined(__sh__)
*loc += isym->gotent.offset + rel->r_addend;
#elif defined(__i386__) || defined(__arm__) || defined(__mc68000__)
*loc += isym->gotent.offset;
#endif
break;
#endif /* USE_GOT_ENTRIES */
}
return ret;
}
#if defined(USE_LIST)
static int arch_list_add(ElfW(RelM) *rel, struct arch_list_entry **list,
int offset, int size)
{
struct arch_list_entry *pe;
for (pe = *list; pe != NULL; pe = pe->next) {
if (pe->addend == rel->r_addend) {
break;
}
}
if (pe == NULL) {
pe = xzalloc(sizeof(struct arch_list_entry));
pe->next = *list;
pe->addend = rel->r_addend;
pe->offset = offset;
/*pe->inited = 0;*/
*list = pe;
return size;
}
return 0;
}
#endif
#if defined(USE_SINGLE)
static int arch_single_init(/*ElfW(RelM) *rel,*/ struct arch_single_entry *single,
int offset, int size)
{
if (single->allocated == 0) {
single->allocated = 1;
single->offset = offset;
single->inited = 0;
return size;
}
return 0;
}
#endif
#if defined(USE_GOT_ENTRIES) || defined(USE_PLT_ENTRIES)
static struct obj_section *arch_xsect_init(struct obj_file *f, const char *name,
int offset, int size)
{
struct obj_section *myrelsec = obj_find_section(f, name);
if (offset == 0) {
offset += size;
}
if (myrelsec) {
obj_extend_section(myrelsec, offset);
} else {
myrelsec = obj_create_alloced_section(f, name,
size, offset);
}
return myrelsec;
}
#endif
static void arch_create_got(struct obj_file *f)
{
#if defined(USE_GOT_ENTRIES) || defined(USE_PLT_ENTRIES)
struct arch_file *ifile = (struct arch_file *) f;
int i;
#if defined(USE_GOT_ENTRIES)
int got_offset = 0, got_needed = 0, got_allocate;
#endif
#if defined(USE_PLT_ENTRIES)
int plt_offset = 0, plt_needed = 0, plt_allocate;
#endif
struct obj_section *relsec, *symsec, *strsec;
ElfW(RelM) *rel, *relend;
ElfW(Sym) *symtab, *extsym;
const char *strtab, *name;
struct arch_symbol *intsym;
for (i = 0; i < f->header.e_shnum; ++i) {
relsec = f->sections[i];
if (relsec->header.sh_type != SHT_RELM)
continue;
symsec = f->sections[relsec->header.sh_link];
strsec = f->sections[symsec->header.sh_link];
rel = (ElfW(RelM) *) relsec->contents;
relend = rel + (relsec->header.sh_size / sizeof(ElfW(RelM)));
symtab = (ElfW(Sym) *) symsec->contents;
strtab = (const char *) strsec->contents;
for (; rel < relend; ++rel) {
extsym = &symtab[ELF_R_SYM(rel->r_info)];
#if defined(USE_GOT_ENTRIES)
got_allocate = 0;
#endif
#if defined(USE_PLT_ENTRIES)
plt_allocate = 0;
#endif
switch (ELF_R_TYPE(rel->r_info)) {
#if defined(__arm__)
case R_ARM_PC24:
case R_ARM_PLT32:
plt_allocate = 1;
break;
case R_ARM_GOTOFF:
case R_ARM_GOTPC:
got_needed = 1;
continue;
case R_ARM_GOT32:
got_allocate = 1;
break;
#elif defined(__i386__)
case R_386_GOTPC:
case R_386_GOTOFF:
got_needed = 1;
continue;
case R_386_GOT32:
got_allocate = 1;
break;
#elif defined(__powerpc__)
case R_PPC_REL24:
plt_allocate = 1;
break;
#elif defined(__mc68000__)
case R_68K_GOT32:
got_allocate = 1;
break;
#ifdef R_68K_GOTOFF
case R_68K_GOTOFF:
got_needed = 1;
continue;
#endif
#elif defined(__sh__)
case R_SH_GOT32:
got_allocate = 1;
break;
case R_SH_GOTPC:
case R_SH_GOTOFF:
got_needed = 1;
continue;
#elif defined(__v850e__)
case R_V850_22_PCREL:
plt_needed = 1;
break;
#endif
default:
continue;
}
if (extsym->st_name != 0) {
name = strtab + extsym->st_name;
} else {
name = f->sections[extsym->st_shndx]->name;
}
intsym = (struct arch_symbol *) obj_find_symbol(f, name);
#if defined(USE_GOT_ENTRIES)
if (got_allocate) {
got_offset += arch_single_init(
/*rel,*/ &intsym->gotent,
got_offset, GOT_ENTRY_SIZE);
got_needed = 1;
}
#endif
#if defined(USE_PLT_ENTRIES)
if (plt_allocate) {
#if defined(USE_PLT_LIST)
plt_offset += arch_list_add(
rel, &intsym->pltent,
plt_offset, PLT_ENTRY_SIZE);
#else
plt_offset += arch_single_init(
/*rel,*/ &intsym->pltent,
plt_offset, PLT_ENTRY_SIZE);
#endif
plt_needed = 1;
}
#endif
}
}
#if defined(USE_GOT_ENTRIES)
if (got_needed) {
ifile->got = arch_xsect_init(f, ".got", got_offset,
GOT_ENTRY_SIZE);
}
#endif
#if defined(USE_PLT_ENTRIES)
if (plt_needed) {
ifile->plt = arch_xsect_init(f, ".plt", plt_offset,
PLT_ENTRY_SIZE);
}
#endif
#endif /* defined(USE_GOT_ENTRIES) || defined(USE_PLT_ENTRIES) */
}
/*======================================================================*/
/* Standard ELF hash function. */
static unsigned long obj_elf_hash_n(const char *name, unsigned long n)
{
unsigned long h = 0;
unsigned long g;
unsigned char ch;
while (n > 0) {
ch = *name++;
h = (h << 4) + ch;
g = (h & 0xf0000000);
if (g != 0) {
h ^= g >> 24;
h &= ~g;
}
n--;
}
return h;
}
static unsigned long FAST_FUNC obj_elf_hash(const char *name)
{
return obj_elf_hash_n(name, strlen(name));
}
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
/* String comparison for non-co-versioned kernel and module. */
static int ncv_strcmp(const char *a, const char *b)
{
size_t alen = strlen(a), blen = strlen(b);
if (blen == alen + 10 && b[alen] == '_' && b[alen + 1] == 'R')
return strncmp(a, b, alen);
else if (alen == blen + 10 && a[blen] == '_' && a[blen + 1] == 'R')
return strncmp(a, b, blen);
else
return strcmp(a, b);
}
/* String hashing for non-co-versioned kernel and module. Here
we are simply forced to drop the crc from the hash. */
static unsigned long FAST_FUNC ncv_symbol_hash(const char *str)
{
size_t len = strlen(str);
if (len > 10 && str[len - 10] == '_' && str[len - 9] == 'R')
len -= 10;
return obj_elf_hash_n(str, len);
}
static void
obj_set_symbol_compare(struct obj_file *f,
int (*cmp) (const char *, const char *),
unsigned long (*hash) (const char *) FAST_FUNC)
{
if (cmp)
f->symbol_cmp = cmp;
if (hash) {
struct obj_symbol *tmptab[HASH_BUCKETS], *sym, *next;
int i;
f->symbol_hash = hash;
memcpy(tmptab, f->symtab, sizeof(tmptab));
memset(f->symtab, 0, sizeof(f->symtab));
for (i = 0; i < HASH_BUCKETS; ++i) {
for (sym = tmptab[i]; sym; sym = next) {
unsigned long h = hash(sym->name) % HASH_BUCKETS;
next = sym->next;
sym->next = f->symtab[h];
f->symtab[h] = sym;
}
}
}
}
#endif /* FEATURE_INSMOD_VERSION_CHECKING */
static struct obj_symbol *
obj_add_symbol(struct obj_file *f, const char *name,
unsigned long symidx, int info,
int secidx, ElfW(Addr) value,
unsigned long size)
{
struct obj_symbol *sym;
unsigned long hash = f->symbol_hash(name) % HASH_BUCKETS;
int n_type = ELF_ST_TYPE(info);
int n_binding = ELF_ST_BIND(info);
for (sym = f->symtab[hash]; sym; sym = sym->next) {
if (f->symbol_cmp(sym->name, name) == 0) {
int o_secidx = sym->secidx;
int o_info = sym->info;
int o_type = ELF_ST_TYPE(o_info);
int o_binding = ELF_ST_BIND(o_info);
/* A redefinition! Is it legal? */
if (secidx == SHN_UNDEF)
return sym;
else if (o_secidx == SHN_UNDEF)
goto found;
else if (n_binding == STB_GLOBAL && o_binding == STB_LOCAL) {
/* Cope with local and global symbols of the same name
in the same object file, as might have been created
by ld -r. The only reason locals are now seen at this
level at all is so that we can do semi-sensible things
with parameters. */
struct obj_symbol *nsym, **p;
nsym = arch_new_symbol();
nsym->next = sym->next;
nsym->ksymidx = -1;
/* Excise the old (local) symbol from the hash chain. */
for (p = &f->symtab[hash]; *p != sym; p = &(*p)->next)
continue;
*p = sym = nsym;
goto found;
} else if (n_binding == STB_LOCAL) {
/* Another symbol of the same name has already been defined.
Just add this to the local table. */
sym = arch_new_symbol();
sym->next = NULL;
sym->ksymidx = -1;
f->local_symtab[symidx] = sym;
goto found;
} else if (n_binding == STB_WEAK)
return sym;
else if (o_binding == STB_WEAK)
goto found;
/* Don't unify COMMON symbols with object types the programmer
doesn't expect. */
else if (secidx == SHN_COMMON
&& (o_type == STT_NOTYPE || o_type == STT_OBJECT))
return sym;
else if (o_secidx == SHN_COMMON
&& (n_type == STT_NOTYPE || n_type == STT_OBJECT))
goto found;
else {
/* Don't report an error if the symbol is coming from
the kernel or some external module. */
if (secidx <= SHN_HIRESERVE)
bb_error_msg("%s multiply defined", name);
return sym;
}
}
}
/* Completely new symbol. */
sym = arch_new_symbol();
sym->next = f->symtab[hash];
f->symtab[hash] = sym;
sym->ksymidx = -1;
if (ELF_ST_BIND(info) == STB_LOCAL && symidx != (unsigned long)(-1)) {
if (symidx >= f->local_symtab_size)
bb_error_msg("local symbol %s with index %ld exceeds local_symtab_size %ld",
name, (long) symidx, (long) f->local_symtab_size);
else
f->local_symtab[symidx] = sym;
}
found:
sym->name = name;
sym->value = value;
sym->size = size;
sym->secidx = secidx;
sym->info = info;
return sym;
}
static struct obj_symbol *
obj_find_symbol(struct obj_file *f, const char *name)
{
struct obj_symbol *sym;
unsigned long hash = f->symbol_hash(name) % HASH_BUCKETS;
for (sym = f->symtab[hash]; sym; sym = sym->next)
if (f->symbol_cmp(sym->name, name) == 0)
return sym;
return NULL;
}
static ElfW(Addr) obj_symbol_final_value(struct obj_file * f, struct obj_symbol * sym)
{
if (sym) {
if (sym->secidx >= SHN_LORESERVE)
return sym->value;
return sym->value + f->sections[sym->secidx]->header.sh_addr;
}
/* As a special case, a NULL sym has value zero. */
return 0;
}
static struct obj_section *obj_find_section(struct obj_file *f, const char *name)
{
int i, n = f->header.e_shnum;
for (i = 0; i < n; ++i)
if (strcmp(f->sections[i]->name, name) == 0)
return f->sections[i];
return NULL;
}
static int obj_load_order_prio(struct obj_section *a)
{
unsigned long af, ac;
af = a->header.sh_flags;
ac = 0;
if (a->name[0] != '.' || strlen(a->name) != 10
|| strcmp(a->name + 5, ".init") != 0
) {
ac |= 32;
}
if (af & SHF_ALLOC)
ac |= 16;
if (!(af & SHF_WRITE))
ac |= 8;
if (af & SHF_EXECINSTR)
ac |= 4;
if (a->header.sh_type != SHT_NOBITS)
ac |= 2;
return ac;
}
static void
obj_insert_section_load_order(struct obj_file *f, struct obj_section *sec)
{
struct obj_section **p;
int prio = obj_load_order_prio(sec);
for (p = f->load_order_search_start; *p; p = &(*p)->load_next)
if (obj_load_order_prio(*p) < prio)
break;
sec->load_next = *p;
*p = sec;
}
static struct obj_section *helper_create_alloced_section(struct obj_file *f,
const char *name,
unsigned long align,
unsigned long size)
{
int newidx = f->header.e_shnum++;
struct obj_section *sec;
f->sections = xrealloc_vector(f->sections, 2, newidx);
f->sections[newidx] = sec = arch_new_section();
sec->header.sh_type = SHT_PROGBITS;
sec->header.sh_flags = SHF_WRITE | SHF_ALLOC;
sec->header.sh_size = size;
sec->header.sh_addralign = align;
sec->name = name;
sec->idx = newidx;
if (size)
sec->contents = xzalloc(size);
return sec;
}
static struct obj_section *obj_create_alloced_section(struct obj_file *f,
const char *name,
unsigned long align,
unsigned long size)
{
struct obj_section *sec;
sec = helper_create_alloced_section(f, name, align, size);
obj_insert_section_load_order(f, sec);
return sec;
}
static struct obj_section *obj_create_alloced_section_first(struct obj_file *f,
const char *name,
unsigned long align,
unsigned long size)
{
struct obj_section *sec;
sec = helper_create_alloced_section(f, name, align, size);
sec->load_next = f->load_order;
f->load_order = sec;
if (f->load_order_search_start == &f->load_order)
f->load_order_search_start = &sec->load_next;
return sec;
}
static void *obj_extend_section(struct obj_section *sec, unsigned long more)
{
unsigned long oldsize = sec->header.sh_size;
if (more) {
sec->header.sh_size += more;
sec->contents = xrealloc(sec->contents, sec->header.sh_size);
}
return sec->contents + oldsize;
}
/* Conditionally add the symbols from the given symbol set to the
new module. */
static int add_symbols_from(struct obj_file *f,
int idx,
struct new_module_symbol *syms,
size_t nsyms)
{
struct new_module_symbol *s;
size_t i;
int used = 0;
#ifdef SYMBOL_PREFIX
char *name_buf = NULL;
size_t name_alloced_size = 0;
#endif
#if ENABLE_FEATURE_CHECK_TAINTED_MODULE
int gpl;
gpl = obj_gpl_license(f, NULL) == 0;
#endif
for (i = 0, s = syms; i < nsyms; ++i, ++s) {
/* Only add symbols that are already marked external.
If we override locals we may cause problems for
argument initialization. We will also create a false
dependency on the module. */
struct obj_symbol *sym;
char *name;
/* GPL licensed modules can use symbols exported with
* EXPORT_SYMBOL_GPL, so ignore any GPLONLY_ prefix on the
* exported names. Non-GPL modules never see any GPLONLY_
* symbols so they cannot fudge it by adding the prefix on
* their references.
*/
libbb: introduce and use is_prefixed_with() function old new delta is_prefixed_with - 18 +18 complete_username 78 77 -1 man_main 737 735 -2 fsck_device 429 427 -2 unpack_ar_archive 80 76 -4 strip_unsafe_prefix 105 101 -4 singlemount 1054 1050 -4 rtc_adjtime_is_utc 90 86 -4 resolve_mount_spec 88 84 -4 parse_one_line 1029 1025 -4 parse_conf 1460 1456 -4 may_wakeup 83 79 -4 loadkmap_main 219 215 -4 get_irqs_from_stat 103 99 -4 get_header_cpio 913 909 -4 findfs_main 79 75 -4 fbsplash_main 1230 1226 -4 load_crontab 776 771 -5 expand_vars_to_list 1151 1146 -5 date_main 881 876 -5 skip_dev_pfx 30 24 -6 make_device 2199 2193 -6 complete_cmd_dir_file 773 767 -6 run_applet_and_exit 715 708 -7 uudecode_main 321 313 -8 pwdx_main 197 189 -8 execute 568 560 -8 i2cdetect_main 1186 1176 -10 procps_scan 1242 1230 -12 procps_read_smaps 1017 1005 -12 process_module 746 734 -12 patch_main 1903 1891 -12 nfsmount 3572 3560 -12 stack_machine 126 112 -14 process_timer_stats 449 435 -14 match_fstype 111 97 -14 do_ipaddr 1344 1330 -14 open_list_and_close 359 343 -16 get_header_tar 1795 1779 -16 prepend_new_eth_table 340 323 -17 fsck_main 1811 1794 -17 find_iface_state 56 38 -18 dnsd_main 1321 1303 -18 base_device 179 158 -21 find_keyword 104 82 -22 handle_incoming_and_exit 2785 2762 -23 parse_and_put_prompt 774 746 -28 modinfo 347 317 -30 find_action 204 171 -33 update_passwd 1470 1436 -34 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 0/49 up/down: 18/-540) Total: -522 bytes Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2015-03-12 22:18:34 +05:30
if (is_prefixed_with((char *)s->name, "GPLONLY_")) {
#if ENABLE_FEATURE_CHECK_TAINTED_MODULE
if (gpl)
s->name += 8;
else
#endif
continue;
}
name = (char *)s->name;
#ifdef SYMBOL_PREFIX
/* Prepend SYMBOL_PREFIX to the symbol's name (the
kernel exports 'C names', but module object files
reference 'linker names'). */
size_t extra = sizeof SYMBOL_PREFIX;
size_t name_size = strlen(name) + extra;
if (name_size > name_alloced_size) {
name_alloced_size = name_size * 2;
name_buf = alloca(name_alloced_size);
}
strcpy(name_buf, SYMBOL_PREFIX);
strcpy(name_buf + extra - 1, name);
name = name_buf;
#endif
sym = obj_find_symbol(f, name);
if (sym && !(ELF_ST_BIND(sym->info) == STB_LOCAL)) {
#ifdef SYMBOL_PREFIX
/* Put NAME_BUF into more permanent storage. */
name = xmalloc(name_size);
strcpy(name, name_buf);
#endif
sym = obj_add_symbol(f, name, -1,
ELF_ST_INFO(STB_GLOBAL,
STT_NOTYPE),
idx, s->value, 0);
/* Did our symbol just get installed? If so, mark the
module as "used". */
if (sym->secidx == idx)
used = 1;
}
}
return used;
}
static void add_kernel_symbols(struct obj_file *f)
{
struct external_module *m;
int i, nused = 0;
/* Add module symbols first. */
for (i = 0, m = ext_modules; i < n_ext_modules; ++i, ++m) {
if (m->nsyms
&& add_symbols_from(f, SHN_HIRESERVE + 2 + i, m->syms, m->nsyms)
) {
m->used = 1;
++nused;
}
}
n_ext_modules_used = nused;
/* And finally the symbols from the kernel proper. */
if (nksyms)
add_symbols_from(f, SHN_HIRESERVE + 1, ksyms, nksyms);
}
static char *get_modinfo_value(struct obj_file *f, const char *key)
{
struct obj_section *sec;
char *p, *v, *n, *ep;
size_t klen = strlen(key);
sec = obj_find_section(f, ".modinfo");
if (sec == NULL)
return NULL;
p = sec->contents;
ep = p + sec->header.sh_size;
while (p < ep) {
v = strchr(p, '=');
n = strchr(p, '\0');
if (v) {
if (p + klen == v && strncmp(p, key, klen) == 0)
return v + 1;
} else {
if (p + klen == n && strcmp(p, key) == 0)
return n;
}
p = n + 1;
}
return NULL;
}
/*======================================================================*/
/* Functions relating to module loading after 2.1.18. */
/* From Linux-2.6 sources */
/* You can use " around spaces, but can't escape ". */
/* Hyphens and underscores equivalent in parameter names. */
static char *next_arg(char *args, char **param, char **val)
{
unsigned int i, equals = 0;
int in_quote = 0, quoted = 0;
char *next;
if (*args == '"') {
args++;
in_quote = 1;
quoted = 1;
}
for (i = 0; args[i]; i++) {
if (args[i] == ' ' && !in_quote)
break;
if (equals == 0) {
if (args[i] == '=')
equals = i;
}
if (args[i] == '"')
in_quote = !in_quote;
}
*param = args;
if (!equals)
*val = NULL;
else {
args[equals] = '\0';
*val = args + equals + 1;
/* Don't include quotes in value. */
if (**val == '"') {
(*val)++;
if (args[i-1] == '"')
args[i-1] = '\0';
}
if (quoted && args[i-1] == '"')
args[i-1] = '\0';
}
if (args[i]) {
args[i] = '\0';
next = args + i + 1;
} else
next = args + i;
/* Chew up trailing spaces. */
return skip_whitespace(next);
}
static void
new_process_module_arguments(struct obj_file *f, const char *options)
{
char *xoptions, *pos;
char *param, *val;
xoptions = pos = xstrdup(skip_whitespace(options));
while (*pos) {
unsigned long charssize = 0;
char *tmp, *contents, *loc, *pinfo, *p;
struct obj_symbol *sym;
int min, max, n, len;
pos = next_arg(pos, &param, &val);
tmp = xasprintf("parm_%s", param);
pinfo = get_modinfo_value(f, tmp);
free(tmp);
if (pinfo == NULL)
bb_error_msg_and_die("invalid parameter %s", param);
#ifdef SYMBOL_PREFIX
tmp = xasprintf(SYMBOL_PREFIX "%s", param);
sym = obj_find_symbol(f, tmp);
free(tmp);
#else
sym = obj_find_symbol(f, param);
#endif
/* Also check that the parameter was not resolved from the kernel. */
if (sym == NULL || sym->secidx > SHN_HIRESERVE)
bb_error_msg_and_die("symbol for parameter %s not found", param);
/* Number of parameters */
min = max = 1;
if (isdigit(*pinfo)) {
min = max = strtoul(pinfo, &pinfo, 10);
if (*pinfo == '-')
max = strtoul(pinfo + 1, &pinfo, 10);
}
contents = f->sections[sym->secidx]->contents;
loc = contents + sym->value;
if (*pinfo == 'c') {
if (!isdigit(pinfo[1])) {
bb_error_msg_and_die("parameter type 'c' for %s must be followed by"
" the maximum size", param);
}
charssize = strtoul(pinfo + 1, NULL, 10);
}
if (val == NULL) {
if (*pinfo != 'b')
bb_error_msg_and_die("argument expected for parameter %s", param);
val = (char *) "1";
}
/* Parse parameter values */
n = 0;
p = val;
while (*p) {
char sv_ch;
char *endp;
if (++n > max)
bb_error_msg_and_die("too many values for %s (max %d)", param, max);
switch (*pinfo) {
case 's':
len = strcspn(p, ",");
sv_ch = p[len];
p[len] = '\0';
obj_string_patch(f, sym->secidx,
loc - contents, p);
loc += tgt_sizeof_char_p;
p += len;
*p = sv_ch;
break;
case 'c':
len = strcspn(p, ",");
sv_ch = p[len];
p[len] = '\0';
if (len >= charssize)
bb_error_msg_and_die("string too long for %s (max %ld)", param,
charssize - 1);
strcpy((char *) loc, p);
loc += charssize;
p += len;
*p = sv_ch;
break;
case 'b':
*loc++ = strtoul(p, &endp, 0);
p = endp; /* gcc likes temp var for &endp */
break;
case 'h':
*(short *) loc = strtoul(p, &endp, 0);
loc += tgt_sizeof_short;
p = endp;
break;
case 'i':
*(int *) loc = strtoul(p, &endp, 0);
loc += tgt_sizeof_int;
p = endp;
break;
case 'l':
*(long *) loc = strtoul(p, &endp, 0);
loc += tgt_sizeof_long;
p = endp;
break;
default:
bb_error_msg_and_die("unknown parameter type '%c' for %s",
*pinfo, param);
}
p = skip_whitespace(p);
if (*p != ',')
break;
p = skip_whitespace(p + 1);
}
if (n < min)
bb_error_msg_and_die("parameter %s requires at least %d arguments", param, min);
if (*p != '\0')
bb_error_msg_and_die("invalid argument syntax for %s", param);
}
free(xoptions);
}
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
static int new_is_module_checksummed(struct obj_file *f)
{
const char *p = get_modinfo_value(f, "using_checksums");
if (p)
return xatoi(p);
return 0;
}
/* Get the module's kernel version in the canonical integer form. */
static int
new_get_module_version(struct obj_file *f, char str[STRVERSIONLEN])
{
char *p, *q;
int a, b, c;
p = get_modinfo_value(f, "kernel_version");
if (p == NULL)
return -1;
safe_strncpy(str, p, STRVERSIONLEN);
a = strtoul(p, &p, 10);
if (*p != '.')
return -1;
b = strtoul(p + 1, &p, 10);
if (*p != '.')
return -1;
c = strtoul(p + 1, &q, 10);
if (p + 1 == q)
return -1;
return a << 16 | b << 8 | c;
}
#endif /* FEATURE_INSMOD_VERSION_CHECKING */
/* Fetch the loaded modules, and all currently exported symbols. */
static void new_get_kernel_symbols(void)
{
char *module_names, *mn;
struct external_module *modules, *m;
struct new_module_symbol *syms, *s;
size_t ret, bufsize, nmod, nsyms, i, j;
/* Collect the loaded modules. */
bufsize = 256;
module_names = xmalloc(bufsize);
retry_modules_load:
if (query_module(NULL, QM_MODULES, module_names, bufsize, &ret)) {
if (errno == ENOSPC && bufsize < ret) {
bufsize = ret;
module_names = xrealloc(module_names, bufsize);
goto retry_modules_load;
}
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_perror_msg_and_die("QM_MODULES");
}
n_ext_modules = nmod = ret;
/* Collect the modules' symbols. */
if (nmod) {
ext_modules = modules = xzalloc(nmod * sizeof(*modules));
for (i = 0, mn = module_names, m = modules;
i < nmod; ++i, ++m, mn += strlen(mn) + 1) {
struct new_module_info info;
if (query_module(mn, QM_INFO, &info, sizeof(info), &ret)) {
if (errno == ENOENT) {
/* The module was removed out from underneath us. */
continue;
}
bb_perror_msg_and_die("query_module: QM_INFO: %s", mn);
}
bufsize = 1024;
syms = xmalloc(bufsize);
retry_mod_sym_load:
if (query_module(mn, QM_SYMBOLS, syms, bufsize, &ret)) {
switch (errno) {
case ENOSPC:
bufsize = ret;
syms = xrealloc(syms, bufsize);
goto retry_mod_sym_load;
case ENOENT:
/* The module was removed out from underneath us. */
continue;
default:
bb_perror_msg_and_die("query_module: QM_SYMBOLS: %s", mn);
}
}
nsyms = ret;
m->name = mn;
m->addr = info.addr;
m->nsyms = nsyms;
m->syms = syms;
for (j = 0, s = syms; j < nsyms; ++j, ++s) {
s->name += (unsigned long) syms;
}
}
}
/* Collect the kernel's symbols. */
bufsize = 16 * 1024;
syms = xmalloc(bufsize);
retry_kern_sym_load:
if (query_module(NULL, QM_SYMBOLS, syms, bufsize, &ret)) {
if (errno == ENOSPC && bufsize < ret) {
bufsize = ret;
syms = xrealloc(syms, bufsize);
goto retry_kern_sym_load;
}
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_perror_msg_and_die("kernel: QM_SYMBOLS");
}
nksyms = nsyms = ret;
ksyms = syms;
for (j = 0, s = syms; j < nsyms; ++j, ++s) {
s->name += (unsigned long) syms;
}
}
/* Return the kernel symbol checksum version, or zero if not used. */
static int new_is_kernel_checksummed(void)
{
struct new_module_symbol *s;
size_t i;
/* Using_Versions is not the first symbol, but it should be in there. */
for (i = 0, s = ksyms; i < nksyms; ++i, ++s)
if (strcmp((char *) s->name, "Using_Versions") == 0)
return s->value;
return 0;
}
static void new_create_this_module(struct obj_file *f, const char *m_name)
{
struct obj_section *sec;
sec = obj_create_alloced_section_first(f, ".this", tgt_sizeof_long,
sizeof(struct new_module));
/* done by obj_create_alloced_section_first: */
/*memset(sec->contents, 0, sizeof(struct new_module));*/
obj_add_symbol(f, SPFX "__this_module", -1,
ELF_ST_INFO(STB_LOCAL, STT_OBJECT), sec->idx, 0,
sizeof(struct new_module));
obj_string_patch(f, sec->idx, offsetof(struct new_module, name),
m_name);
}
#if ENABLE_FEATURE_INSMOD_KSYMOOPS_SYMBOLS
/* add an entry to the __ksymtab section, creating it if necessary */
static void new_add_ksymtab(struct obj_file *f, struct obj_symbol *sym)
{
struct obj_section *sec;
ElfW(Addr) ofs;
/* ensure __ksymtab is allocated, EXPORT_NOSYMBOLS creates a non-alloc section.
* If __ksymtab is defined but not marked alloc, x out the first character
* (no obj_delete routine) and create a new __ksymtab with the correct
* characteristics.
*/
sec = obj_find_section(f, "__ksymtab");
if (sec && !(sec->header.sh_flags & SHF_ALLOC)) {
*((char *)(sec->name)) = 'x'; /* override const */
sec = NULL;
}
if (!sec)
sec = obj_create_alloced_section(f, "__ksymtab",
tgt_sizeof_void_p, 0);
if (!sec)
return;
sec->header.sh_flags |= SHF_ALLOC;
/* Empty section might be byte-aligned */
sec->header.sh_addralign = tgt_sizeof_void_p;
ofs = sec->header.sh_size;
obj_symbol_patch(f, sec->idx, ofs, sym);
obj_string_patch(f, sec->idx, ofs + tgt_sizeof_void_p, sym->name);
obj_extend_section(sec, 2 * tgt_sizeof_char_p);
}
#endif /* FEATURE_INSMOD_KSYMOOPS_SYMBOLS */
static int new_create_module_ksymtab(struct obj_file *f)
{
struct obj_section *sec;
int i;
/* We must always add the module references. */
if (n_ext_modules_used) {
struct new_module_ref *dep;
struct obj_symbol *tm;
sec = obj_create_alloced_section(f, ".kmodtab", tgt_sizeof_void_p,
(sizeof(struct new_module_ref)
* n_ext_modules_used));
if (!sec)
return 0;
tm = obj_find_symbol(f, SPFX "__this_module");
dep = (struct new_module_ref *) sec->contents;
for (i = 0; i < n_ext_modules; ++i)
if (ext_modules[i].used) {
dep->dep = ext_modules[i].addr;
obj_symbol_patch(f, sec->idx,
(char *) &dep->ref - sec->contents, tm);
dep->next_ref = 0;
++dep;
}
}
if (!flag_noexport && !obj_find_section(f, "__ksymtab")) {
size_t nsyms;
int *loaded;
sec = obj_create_alloced_section(f, "__ksymtab", tgt_sizeof_void_p, 0);
/* We don't want to export symbols residing in sections that
aren't loaded. There are a number of these created so that
we make sure certain module options don't appear twice. */
i = f->header.e_shnum;
loaded = alloca(sizeof(int) * i);
while (--i >= 0)
loaded[i] = (f->sections[i]->header.sh_flags & SHF_ALLOC) != 0;
for (nsyms = i = 0; i < HASH_BUCKETS; ++i) {
struct obj_symbol *sym;
for (sym = f->symtab[i]; sym; sym = sym->next) {
if (ELF_ST_BIND(sym->info) != STB_LOCAL
&& sym->secidx <= SHN_HIRESERVE
&& (sym->secidx >= SHN_LORESERVE || loaded[sym->secidx])
) {
ElfW(Addr) ofs = nsyms * 2 * tgt_sizeof_void_p;
obj_symbol_patch(f, sec->idx, ofs, sym);
obj_string_patch(f, sec->idx, ofs + tgt_sizeof_void_p,
sym->name);
nsyms++;
}
}
}
obj_extend_section(sec, nsyms * 2 * tgt_sizeof_char_p);
}
return 1;
}
static int
new_init_module(const char *m_name, struct obj_file *f, unsigned long m_size)
{
struct new_module *module;
struct obj_section *sec;
void *image;
int ret;
tgt_long m_addr;
sec = obj_find_section(f, ".this");
if (!sec || !sec->contents) {
bb_perror_msg_and_die("corrupt module %s?", m_name);
}
module = (struct new_module *) sec->contents;
m_addr = sec->header.sh_addr;
module->size_of_struct = sizeof(*module);
module->size = m_size;
module->flags = flag_autoclean ? NEW_MOD_AUTOCLEAN : 0;
sec = obj_find_section(f, "__ksymtab");
if (sec && sec->header.sh_size) {
module->syms = sec->header.sh_addr;
module->nsyms = sec->header.sh_size / (2 * tgt_sizeof_char_p);
}
if (n_ext_modules_used) {
sec = obj_find_section(f, ".kmodtab");
module->deps = sec->header.sh_addr;
module->ndeps = n_ext_modules_used;
}
module->init = obj_symbol_final_value(f, obj_find_symbol(f, SPFX "init_module"));
module->cleanup = obj_symbol_final_value(f, obj_find_symbol(f, SPFX "cleanup_module"));
sec = obj_find_section(f, "__ex_table");
if (sec) {
module->ex_table_start = sec->header.sh_addr;
module->ex_table_end = sec->header.sh_addr + sec->header.sh_size;
}
sec = obj_find_section(f, ".text.init");
if (sec) {
module->runsize = sec->header.sh_addr - m_addr;
}
sec = obj_find_section(f, ".data.init");
if (sec) {
if (!module->runsize
|| module->runsize > sec->header.sh_addr - m_addr
) {
module->runsize = sec->header.sh_addr - m_addr;
}
}
sec = obj_find_section(f, ARCHDATA_SEC_NAME);
if (sec && sec->header.sh_size) {
module->archdata_start = (void*)sec->header.sh_addr;
module->archdata_end = module->archdata_start + sec->header.sh_size;
}
sec = obj_find_section(f, KALLSYMS_SEC_NAME);
if (sec && sec->header.sh_size) {
module->kallsyms_start = (void*)sec->header.sh_addr;
module->kallsyms_end = module->kallsyms_start + sec->header.sh_size;
}
/* Whew! All of the initialization is complete. Collect the final
module image and give it to the kernel. */
image = xmalloc(m_size);
obj_create_image(f, image);
ret = init_module(m_name, (struct new_module *) image);
if (ret)
bb_perror_msg("init_module: %s", m_name);
free(image);
return ret == 0;
}
/*======================================================================*/
static void
obj_string_patch(struct obj_file *f, int secidx, ElfW(Addr) offset,
const char *string)
{
struct obj_string_patch *p;
struct obj_section *strsec;
size_t len = strlen(string) + 1;
char *loc;
p = xzalloc(sizeof(*p));
p->next = f->string_patches;
p->reloc_secidx = secidx;
p->reloc_offset = offset;
f->string_patches = p;
strsec = obj_find_section(f, ".kstrtab");
if (strsec == NULL) {
strsec = obj_create_alloced_section(f, ".kstrtab", 1, len);
/*p->string_offset = 0;*/
loc = strsec->contents;
} else {
p->string_offset = strsec->header.sh_size;
loc = obj_extend_section(strsec, len);
}
memcpy(loc, string, len);
}
static void
obj_symbol_patch(struct obj_file *f, int secidx, ElfW(Addr) offset,
struct obj_symbol *sym)
{
struct obj_symbol_patch *p;
p = xmalloc(sizeof(*p));
p->next = f->symbol_patches;
p->reloc_secidx = secidx;
p->reloc_offset = offset;
p->sym = sym;
f->symbol_patches = p;
}
static void obj_check_undefineds(struct obj_file *f)
{
unsigned i;
for (i = 0; i < HASH_BUCKETS; ++i) {
struct obj_symbol *sym;
for (sym = f->symtab[i]; sym; sym = sym->next) {
if (sym->secidx == SHN_UNDEF) {
if (ELF_ST_BIND(sym->info) == STB_WEAK) {
sym->secidx = SHN_ABS;
sym->value = 0;
} else {
if (!flag_quiet)
bb_error_msg_and_die("unresolved symbol %s", sym->name);
}
}
}
}
}
static void obj_allocate_commons(struct obj_file *f)
{
struct common_entry {
struct common_entry *next;
struct obj_symbol *sym;
} *common_head = NULL;
unsigned long i;
for (i = 0; i < HASH_BUCKETS; ++i) {
struct obj_symbol *sym;
for (sym = f->symtab[i]; sym; sym = sym->next) {
if (sym->secidx == SHN_COMMON) {
/* Collect all COMMON symbols and sort them by size so as to
minimize space wasted by alignment requirements. */
struct common_entry **p, *n;
for (p = &common_head; *p; p = &(*p)->next)
if (sym->size <= (*p)->sym->size)
break;
n = alloca(sizeof(*n));
n->next = *p;
n->sym = sym;
*p = n;
}
}
}
for (i = 1; i < f->local_symtab_size; ++i) {
struct obj_symbol *sym = f->local_symtab[i];
if (sym && sym->secidx == SHN_COMMON) {
struct common_entry **p, *n;
for (p = &common_head; *p; p = &(*p)->next) {
if (sym == (*p)->sym)
break;
if (sym->size < (*p)->sym->size) {
n = alloca(sizeof(*n));
n->next = *p;
n->sym = sym;
*p = n;
break;
}
}
}
}
if (common_head) {
/* Find the bss section. */
for (i = 0; i < f->header.e_shnum; ++i)
if (f->sections[i]->header.sh_type == SHT_NOBITS)
break;
/* If for some reason there hadn't been one, create one. */
if (i == f->header.e_shnum) {
struct obj_section *sec;
f->header.e_shnum++;
f->sections = xrealloc_vector(f->sections, 2, i);
f->sections[i] = sec = arch_new_section();
sec->header.sh_type = SHT_PROGBITS;
sec->header.sh_flags = SHF_WRITE | SHF_ALLOC;
sec->name = ".bss";
sec->idx = i;
}
/* Allocate the COMMONS. */
{
ElfW(Addr) bss_size = f->sections[i]->header.sh_size;
ElfW(Addr) max_align = f->sections[i]->header.sh_addralign;
struct common_entry *c;
for (c = common_head; c; c = c->next) {
ElfW(Addr) align = c->sym->value;
if (align > max_align)
max_align = align;
if (bss_size & (align - 1))
bss_size = (bss_size | (align - 1)) + 1;
c->sym->secidx = i;
c->sym->value = bss_size;
bss_size += c->sym->size;
}
f->sections[i]->header.sh_size = bss_size;
f->sections[i]->header.sh_addralign = max_align;
}
}
/* For the sake of patch relocation and parameter initialization,
allocate zeroed data for NOBITS sections now. Note that after
this we cannot assume NOBITS are really empty. */
for (i = 0; i < f->header.e_shnum; ++i) {
struct obj_section *s = f->sections[i];
if (s->header.sh_type == SHT_NOBITS) {
s->contents = NULL;
if (s->header.sh_size != 0)
s->contents = xzalloc(s->header.sh_size);
s->header.sh_type = SHT_PROGBITS;
}
}
}
static unsigned long obj_load_size(struct obj_file *f)
{
unsigned long dot = 0;
struct obj_section *sec;
/* Finalize the positions of the sections relative to one another. */
for (sec = f->load_order; sec; sec = sec->load_next) {
ElfW(Addr) align;
align = sec->header.sh_addralign;
if (align && (dot & (align - 1)))
dot = (dot | (align - 1)) + 1;
sec->header.sh_addr = dot;
dot += sec->header.sh_size;
}
return dot;
}
static int obj_relocate(struct obj_file *f, ElfW(Addr) base)
{
int i, n = f->header.e_shnum;
int ret = 1;
/* Finalize the addresses of the sections. */
f->baseaddr = base;
for (i = 0; i < n; ++i)
f->sections[i]->header.sh_addr += base;
/* And iterate over all of the relocations. */
for (i = 0; i < n; ++i) {
struct obj_section *relsec, *symsec, *targsec, *strsec;
ElfW(RelM) * rel, *relend;
ElfW(Sym) * symtab;
const char *strtab;
relsec = f->sections[i];
if (relsec->header.sh_type != SHT_RELM)
continue;
symsec = f->sections[relsec->header.sh_link];
targsec = f->sections[relsec->header.sh_info];
strsec = f->sections[symsec->header.sh_link];
rel = (ElfW(RelM) *) relsec->contents;
relend = rel + (relsec->header.sh_size / sizeof(ElfW(RelM)));
symtab = (ElfW(Sym) *) symsec->contents;
strtab = (const char *) strsec->contents;
for (; rel < relend; ++rel) {
ElfW(Addr) value = 0;
struct obj_symbol *intsym = NULL;
unsigned long symndx;
ElfW(Sym) *extsym = NULL;
const char *errmsg;
/* Attempt to find a value to use for this relocation. */
symndx = ELF_R_SYM(rel->r_info);
if (symndx) {
/* Note we've already checked for undefined symbols. */
extsym = &symtab[symndx];
if (ELF_ST_BIND(extsym->st_info) == STB_LOCAL) {
/* Local symbols we look up in the local table to be sure
we get the one that is really intended. */
intsym = f->local_symtab[symndx];
} else {
/* Others we look up in the hash table. */
const char *name;
if (extsym->st_name)
name = strtab + extsym->st_name;
else
name = f->sections[extsym->st_shndx]->name;
intsym = obj_find_symbol(f, name);
}
value = obj_symbol_final_value(f, intsym);
intsym->referenced = 1;
}
#if SHT_RELM == SHT_RELA
#if defined(__alpha__) && defined(AXP_BROKEN_GAS)
/* Work around a nasty GAS bug, that is fixed as of 2.7.0.9. */
if (!extsym || !extsym->st_name
|| ELF_ST_BIND(extsym->st_info) != STB_LOCAL)
#endif
value += rel->r_addend;
#endif
/* Do it! */
switch (arch_apply_relocation
(f, targsec, /*symsec,*/ intsym, rel, value)
) {
case obj_reloc_ok:
break;
case obj_reloc_overflow:
errmsg = "Relocation overflow";
goto bad_reloc;
case obj_reloc_dangerous:
errmsg = "Dangerous relocation";
goto bad_reloc;
case obj_reloc_unhandled:
errmsg = "Unhandled relocation";
bad_reloc:
if (extsym) {
bb_error_msg("%s of type %ld for %s", errmsg,
(long) ELF_R_TYPE(rel->r_info),
strtab + extsym->st_name);
} else {
bb_error_msg("%s of type %ld", errmsg,
(long) ELF_R_TYPE(rel->r_info));
}
ret = 0;
break;
}
}
}
/* Finally, take care of the patches. */
if (f->string_patches) {
struct obj_string_patch *p;
struct obj_section *strsec;
ElfW(Addr) strsec_base;
strsec = obj_find_section(f, ".kstrtab");
strsec_base = strsec->header.sh_addr;
for (p = f->string_patches; p; p = p->next) {
struct obj_section *targsec = f->sections[p->reloc_secidx];
*(ElfW(Addr) *) (targsec->contents + p->reloc_offset)
= strsec_base + p->string_offset;
}
}
if (f->symbol_patches) {
struct obj_symbol_patch *p;
for (p = f->symbol_patches; p; p = p->next) {
struct obj_section *targsec = f->sections[p->reloc_secidx];
*(ElfW(Addr) *) (targsec->contents + p->reloc_offset)
= obj_symbol_final_value(f, p->sym);
}
}
return ret;
}
static int obj_create_image(struct obj_file *f, char *image)
{
struct obj_section *sec;
ElfW(Addr) base = f->baseaddr;
for (sec = f->load_order; sec; sec = sec->load_next) {
char *secimg;
if (sec->contents == 0 || sec->header.sh_size == 0)
continue;
secimg = image + (sec->header.sh_addr - base);
/* Note that we allocated data for NOBITS sections earlier. */
memcpy(secimg, sec->contents, sec->header.sh_size);
}
return 1;
}
/*======================================================================*/
static struct obj_file *obj_load(char *image, size_t image_size, int loadprogbits)
{
typedef uint32_t aliased_uint32_t FIX_ALIASING;
#if BB_LITTLE_ENDIAN
# define ELFMAG_U32 ((uint32_t)(ELFMAG0 + 0x100 * (ELFMAG1 + (0x100 * (ELFMAG2 + 0x100 * ELFMAG3)))))
#else
# define ELFMAG_U32 ((uint32_t)((((ELFMAG0 * 0x100) + ELFMAG1) * 0x100 + ELFMAG2) * 0x100 + ELFMAG3))
#endif
struct obj_file *f;
ElfW(Shdr) * section_headers;
size_t shnum, i;
char *shstrtab;
/* Read the file header. */
f = arch_new_file();
f->symbol_cmp = strcmp;
f->symbol_hash = obj_elf_hash;
f->load_order_search_start = &f->load_order;
if (image_size < sizeof(f->header))
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("error while loading ELF header");
memcpy(&f->header, image, sizeof(f->header));
if (*(aliased_uint32_t*)(&f->header.e_ident) != ELFMAG_U32) {
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("not an ELF file");
}
if (f->header.e_ident[EI_CLASS] != ELFCLASSM
|| f->header.e_ident[EI_DATA] != (BB_BIG_ENDIAN ? ELFDATA2MSB : ELFDATA2LSB)
|| f->header.e_ident[EI_VERSION] != EV_CURRENT
|| !MATCH_MACHINE(f->header.e_machine)
) {
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("ELF file not for this architecture");
}
if (f->header.e_type != ET_REL) {
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("ELF file not a relocatable object");
}
/* Read the section headers. */
if (f->header.e_shentsize != sizeof(ElfW(Shdr))) {
bb_error_msg_and_die("section header size mismatch: %lu != %lu",
(unsigned long) f->header.e_shentsize,
(unsigned long) sizeof(ElfW(Shdr)));
}
shnum = f->header.e_shnum;
/* Growth of ->sections vector will be done by
* xrealloc_vector(..., 2, ...), therefore we must allocate
* at least 2^2 = 4 extra elements here. */
f->sections = xzalloc(sizeof(f->sections[0]) * (shnum + 4));
section_headers = alloca(sizeof(ElfW(Shdr)) * shnum);
if (image_size < f->header.e_shoff + sizeof(ElfW(Shdr)) * shnum)
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("error while loading section headers");
memcpy(section_headers, image + f->header.e_shoff, sizeof(ElfW(Shdr)) * shnum);
/* Read the section data. */
for (i = 0; i < shnum; ++i) {
struct obj_section *sec;
f->sections[i] = sec = arch_new_section();
sec->header = section_headers[i];
sec->idx = i;
if (sec->header.sh_size) {
switch (sec->header.sh_type) {
case SHT_NULL:
case SHT_NOTE:
case SHT_NOBITS:
/* ignore */
break;
case SHT_PROGBITS:
#if LOADBITS
if (!loadprogbits) {
sec->contents = NULL;
break;
}
#endif
case SHT_SYMTAB:
case SHT_STRTAB:
case SHT_RELM:
#if defined(__mips__)
case SHT_MIPS_DWARF:
#endif
sec->contents = NULL;
if (sec->header.sh_size > 0) {
sec->contents = xmalloc(sec->header.sh_size);
if (image_size < (sec->header.sh_offset + sec->header.sh_size))
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("error while loading section data");
memcpy(sec->contents, image + sec->header.sh_offset, sec->header.sh_size);
}
break;
#if SHT_RELM == SHT_REL
case SHT_RELA:
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("RELA relocations not supported on this architecture");
#else
case SHT_REL:
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("REL relocations not supported on this architecture");
#endif
default:
if (sec->header.sh_type >= SHT_LOPROC) {
/* Assume processor specific section types are debug
info and can safely be ignored. If this is ever not
the case (Hello MIPS?), don't put ifdefs here but
create an arch_load_proc_section(). */
break;
}
bb_error_msg_and_die("can't handle sections of type %ld",
(long) sec->header.sh_type);
}
}
}
/* Do what sort of interpretation as needed by each section. */
shstrtab = f->sections[f->header.e_shstrndx]->contents;
for (i = 0; i < shnum; ++i) {
struct obj_section *sec = f->sections[i];
sec->name = shstrtab + sec->header.sh_name;
}
for (i = 0; i < shnum; ++i) {
struct obj_section *sec = f->sections[i];
/* .modinfo should be contents only but gcc has no attribute for that.
* The kernel may have marked .modinfo as ALLOC, ignore this bit.
*/
if (strcmp(sec->name, ".modinfo") == 0)
sec->header.sh_flags &= ~SHF_ALLOC;
if (sec->header.sh_flags & SHF_ALLOC)
obj_insert_section_load_order(f, sec);
switch (sec->header.sh_type) {
case SHT_SYMTAB:
{
unsigned long nsym, j;
char *strtab;
ElfW(Sym) * sym;
if (sec->header.sh_entsize != sizeof(ElfW(Sym))) {
bb_error_msg_and_die("symbol size mismatch: %lu != %lu",
(unsigned long) sec->header.sh_entsize,
(unsigned long) sizeof(ElfW(Sym)));
}
nsym = sec->header.sh_size / sizeof(ElfW(Sym));
strtab = f->sections[sec->header.sh_link]->contents;
sym = (ElfW(Sym) *) sec->contents;
/* Allocate space for a table of local symbols. */
j = f->local_symtab_size = sec->header.sh_info;
f->local_symtab = xzalloc(j * sizeof(struct obj_symbol *));
/* Insert all symbols into the hash table. */
for (j = 1, ++sym; j < nsym; ++j, ++sym) {
ElfW(Addr) val = sym->st_value;
const char *name;
if (sym->st_name)
name = strtab + sym->st_name;
else if (sym->st_shndx < shnum)
name = f->sections[sym->st_shndx]->name;
else
continue;
#if defined(__SH5__)
/*
* For sh64 it is possible that the target of a branch
* requires a mode switch (32 to 16 and back again).
*
* This is implied by the lsb being set in the target
* address for SHmedia mode and clear for SHcompact.
*/
val |= sym->st_other & 4;
#endif
obj_add_symbol(f, name, j, sym->st_info, sym->st_shndx,
val, sym->st_size);
}
}
break;
case SHT_RELM:
if (sec->header.sh_entsize != sizeof(ElfW(RelM))) {
bb_error_msg_and_die("relocation entry size mismatch: %lu != %lu",
(unsigned long) sec->header.sh_entsize,
(unsigned long) sizeof(ElfW(RelM)));
}
break;
/* XXX Relocation code from modutils-2.3.19 is not here.
* Why? That's about 20 lines of code from obj/obj_load.c,
* which gets done in a second pass through the sections.
* This BusyBox insmod does similar work in obj_relocate(). */
}
}
return f;
}
#if ENABLE_FEATURE_INSMOD_LOADINKMEM
/*
* load the unloaded sections directly into the memory allocated by
* kernel for the module
*/
static int obj_load_progbits(char *image, size_t image_size, struct obj_file *f, char *imagebase)
{
ElfW(Addr) base = f->baseaddr;
struct obj_section* sec;
for (sec = f->load_order; sec; sec = sec->load_next) {
/* section already loaded? */
if (sec->contents != NULL)
continue;
if (sec->header.sh_size == 0)
continue;
sec->contents = imagebase + (sec->header.sh_addr - base);
if (image_size < (sec->header.sh_offset + sec->header.sh_size)) {
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg("error reading ELF section data");
return 0; /* need to delete half-loaded module! */
}
memcpy(sec->contents, image + sec->header.sh_offset, sec->header.sh_size);
}
return 1;
}
#endif
static void hide_special_symbols(struct obj_file *f)
{
static const char *const specials[] ALIGN_PTR = {
SPFX "cleanup_module",
SPFX "init_module",
SPFX "kernel_version",
NULL
};
struct obj_symbol *sym;
const char *const *p;
for (p = specials; *p; ++p) {
sym = obj_find_symbol(f, *p);
if (sym != NULL)
sym->info = ELF_ST_INFO(STB_LOCAL, ELF_ST_TYPE(sym->info));
}
}
#if ENABLE_FEATURE_CHECK_TAINTED_MODULE
static int obj_gpl_license(struct obj_file *f, const char **license)
{
struct obj_section *sec;
/* This list must match *exactly* the list of allowable licenses in
* linux/include/linux/module.h. Checking for leading "GPL" will not
* work, somebody will use "GPL sucks, this is proprietary".
*/
static const char *const gpl_licenses[] ALIGN_PTR = {
"GPL",
"GPL v2",
"GPL and additional rights",
"Dual BSD/GPL",
"Dual MPL/GPL"
};
sec = obj_find_section(f, ".modinfo");
if (sec) {
const char *value, *ptr, *endptr;
ptr = sec->contents;
endptr = ptr + sec->header.sh_size;
while (ptr < endptr) {
value = strchr(ptr, '=');
if (value && strncmp(ptr, "license", value-ptr) == 0) {
unsigned i;
if (license)
*license = value+1;
for (i = 0; i < ARRAY_SIZE(gpl_licenses); ++i) {
if (strcmp(value+1, gpl_licenses[i]) == 0)
return 0;
}
return 2;
}
ptr = strchr(ptr, '\0');
if (ptr)
ptr++;
else
ptr = endptr;
}
}
return 1;
}
#define TAINT_FILENAME "/proc/sys/kernel/tainted"
#define TAINT_PROPRIETORY_MODULE (1 << 0)
#define TAINT_FORCED_MODULE (1 << 1)
#define TAINT_UNSAFE_SMP (1 << 2)
#define TAINT_URL "http://www.tux.org/lkml/#export-tainted"
static void set_tainted(int fd, const char *m_name,
int kernel_has_tainted, int taint,
const char *text1, const char *text2)
{
static smallint printed_info;
char buf[80];
int oldval;
if (fd < 0 && !kernel_has_tainted)
return; /* New modutils on old kernel */
printf("Warning: loading %s will taint the kernel: %s%s\n",
m_name, text1, text2);
if (!printed_info) {
printf(" See %s for information about tainted modules\n", TAINT_URL);
printed_info = 1;
}
if (fd >= 0) {
read(fd, buf, sizeof(buf)-1);
buf[sizeof(buf)-1] = '\0';
oldval = strtoul(buf, NULL, 10);
sprintf(buf, "%d\n", oldval | taint);
xwrite_str(fd, buf);
}
}
/* Check if loading this module will taint the kernel. */
static void check_tainted_module(struct obj_file *f, const char *m_name)
{
int fd, kernel_has_tainted;
const char *ptr;
kernel_has_tainted = 1;
fd = open(TAINT_FILENAME, O_RDWR);
if (fd < 0) {
if (errno == ENOENT)
kernel_has_tainted = 0;
else if (errno == EACCES)
kernel_has_tainted = 1;
else {
bb_simple_perror_msg(TAINT_FILENAME);
kernel_has_tainted = 0;
}
}
switch (obj_gpl_license(f, &ptr)) {
case 0:
break;
case 1:
set_tainted(fd, m_name, kernel_has_tainted, TAINT_PROPRIETORY_MODULE, "no license", "");
break;
default: /* case 2: */
/* The module has a non-GPL license so we pretend that the
* kernel always has a taint flag to get a warning even on
* kernels without the proc flag.
*/
set_tainted(fd, m_name, 1, TAINT_PROPRIETORY_MODULE, "non-GPL license - ", ptr);
break;
}
if (flag_force_load)
set_tainted(fd, m_name, 1, TAINT_FORCED_MODULE, "forced load", "");
if (fd >= 0)
close(fd);
}
#else /* !FEATURE_CHECK_TAINTED_MODULE */
#define check_tainted_module(x, y) do { } while (0);
#endif
#if ENABLE_FEATURE_INSMOD_KSYMOOPS_SYMBOLS
/* add module source, timestamp, kernel version and a symbol for the
* start of some sections. this info is used by ksymoops to do better
* debugging.
*/
#if !ENABLE_FEATURE_INSMOD_VERSION_CHECKING
#define get_module_version(f, str) get_module_version(str)
#endif
static int
get_module_version(struct obj_file *f, char str[STRVERSIONLEN])
{
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
return new_get_module_version(f, str);
#else
strncpy(str, "???", sizeof(str));
return -1;
#endif
}
/* add module source, timestamp, kernel version and a symbol for the
* start of some sections. this info is used by ksymoops to do better
* debugging.
*/
static void
add_ksymoops_symbols(struct obj_file *f, const char *filename,
const char *m_name)
{
static const char symprefix[] ALIGN1 = "__insmod_";
static const char section_names[][8] = {
".text",
".rodata",
".data",
".bss",
".sbss"
};
struct obj_section *sec;
struct obj_symbol *sym;
char *name, *absolute_filename;
char str[STRVERSIONLEN];
unsigned i;
int lm_name, lfilename, use_ksymtab, version;
struct stat statbuf;
/* WARNING: was using realpath, but replaced by readlink to stop using
* lots of stack. But here it seems to be able to cause problems? */
absolute_filename = xmalloc_readlink(filename);
if (!absolute_filename)
absolute_filename = xstrdup(filename);
lm_name = strlen(m_name);
lfilename = strlen(absolute_filename);
/* add to ksymtab if it already exists or there is no ksymtab and other symbols
* are not to be exported. otherwise leave ksymtab alone for now, the
* "export all symbols" compatibility code will export these symbols later.
*/
use_ksymtab = obj_find_section(f, "__ksymtab") || flag_noexport;
sec = obj_find_section(f, ".this");
if (sec) {
/* tag the module header with the object name, last modified
* timestamp and module version. worst case for module version
* is 0xffffff, decimal 16777215. putting all three fields in
* one symbol is less readable but saves kernel space.
*/
if (stat(absolute_filename, &statbuf) != 0)
statbuf.st_mtime = 0;
version = get_module_version(f, str); /* -1 if not found */
name = xasprintf("%s%s_O%s_M%0*lX_V%d",
symprefix, m_name, absolute_filename,
(int)(2 * sizeof(statbuf.st_mtime)),
(long)statbuf.st_mtime,
version);
sym = obj_add_symbol(f, name, -1,
ELF_ST_INFO(STB_GLOBAL, STT_NOTYPE),
sec->idx, sec->header.sh_addr, 0);
if (use_ksymtab)
new_add_ksymtab(f, sym);
}
free(absolute_filename);
#ifdef _NOT_SUPPORTED_
/* record where the persistent data is going, same address as previous symbol */
if (f->persist) {
name = xasprintf("%s%s_P%s",
symprefix, m_name, f->persist);
sym = obj_add_symbol(f, name, -1, ELF_ST_INFO(STB_GLOBAL, STT_NOTYPE),
sec->idx, sec->header.sh_addr, 0);
if (use_ksymtab)
new_add_ksymtab(f, sym);
}
#endif
/* tag the desired sections if size is non-zero */
for (i = 0; i < ARRAY_SIZE(section_names); ++i) {
sec = obj_find_section(f, section_names[i]);
if (sec && sec->header.sh_size) {
name = xasprintf("%s%s_S%s_L%ld",
symprefix, m_name, sec->name,
(long)sec->header.sh_size);
sym = obj_add_symbol(f, name, -1, ELF_ST_INFO(STB_GLOBAL, STT_NOTYPE),
sec->idx, sec->header.sh_addr, 0);
if (use_ksymtab)
new_add_ksymtab(f, sym);
}
}
}
#endif /* FEATURE_INSMOD_KSYMOOPS_SYMBOLS */
#if ENABLE_FEATURE_INSMOD_LOAD_MAP
static void print_load_map(struct obj_file *f)
{
struct obj_section *sec;
#if ENABLE_FEATURE_INSMOD_LOAD_MAP_FULL
struct obj_symbol **all, **p;
int i, nsyms;
char *loaded; /* array of booleans */
struct obj_symbol *sym;
#endif
/* Report on the section layout. */
printf("Sections: Size %-*s Align\n",
(int) (2 * sizeof(void *)), "Address");
for (sec = f->load_order; sec; sec = sec->load_next) {
int a;
unsigned long tmp;
for (a = -1, tmp = sec->header.sh_addralign; tmp; ++a)
tmp >>= 1;
if (a == -1)
a = 0;
printf("%-15s %08lx %0*lx 2**%d\n",
sec->name,
(long)sec->header.sh_size,
(int) (2 * sizeof(void *)),
(long)sec->header.sh_addr,
a);
}
#if ENABLE_FEATURE_INSMOD_LOAD_MAP_FULL
/* Quick reference which section indices are loaded. */
i = f->header.e_shnum;
loaded = alloca(i * sizeof(loaded[0]));
while (--i >= 0)
loaded[i] = ((f->sections[i]->header.sh_flags & SHF_ALLOC) != 0);
/* Collect the symbols we'll be listing. */
for (nsyms = i = 0; i < HASH_BUCKETS; ++i)
for (sym = f->symtab[i]; sym; sym = sym->next)
if (sym->secidx <= SHN_HIRESERVE
&& (sym->secidx >= SHN_LORESERVE || loaded[sym->secidx])
) {
++nsyms;
}
all = alloca(nsyms * sizeof(all[0]));
for (i = 0, p = all; i < HASH_BUCKETS; ++i)
for (sym = f->symtab[i]; sym; sym = sym->next)
if (sym->secidx <= SHN_HIRESERVE
&& (sym->secidx >= SHN_LORESERVE || loaded[sym->secidx])
) {
*p++ = sym;
}
/* And list them. */
printf("\nSymbols:\n");
for (p = all; p < all + nsyms; ++p) {
char type = '?';
unsigned long value;
sym = *p;
if (sym->secidx == SHN_ABS) {
type = 'A';
value = sym->value;
} else if (sym->secidx == SHN_UNDEF) {
type = 'U';
value = 0;
} else {
sec = f->sections[sym->secidx];
if (sec->header.sh_type == SHT_NOBITS)
type = 'B';
else if (sec->header.sh_flags & SHF_ALLOC) {
if (sec->header.sh_flags & SHF_EXECINSTR)
type = 'T';
else if (sec->header.sh_flags & SHF_WRITE)
type = 'D';
else
type = 'R';
}
value = sym->value + sec->header.sh_addr;
}
if (ELF_ST_BIND(sym->info) == STB_LOCAL)
type |= 0x20; /* tolower. safe for '?' too */
printf("%0*lx %c %s\n", (int) (2 * sizeof(void *)), value,
type, sym->name);
}
#endif
}
#else /* !FEATURE_INSMOD_LOAD_MAP */
static void print_load_map(struct obj_file *f UNUSED_PARAM)
{
}
#endif
int FAST_FUNC bb_init_module_24(const char *m_filename, const char *options)
{
int k_crcs;
unsigned long m_size;
ElfW(Addr) m_addr;
struct obj_file *f;
int exit_status = EXIT_FAILURE;
char *m_name;
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
int m_has_modinfo;
#endif
char *image;
size_t image_size;
bool mmaped;
image_size = INT_MAX - 4095;
mmaped = 0;
image = try_to_mmap_module(m_filename, &image_size);
if (image) {
mmaped = 1;
} else {
/* Load module into memory and unzip if compressed */
image = xmalloc_open_zipped_read_close(m_filename, &image_size);
if (!image)
return EXIT_FAILURE;
}
m_name = xstrdup(bb_basename(m_filename));
/* "module.o[.gz]" -> "module" */
*strchrnul(m_name, '.') = '\0';
f = obj_load(image, image_size, LOADBITS);
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
/* Version correspondence? */
m_has_modinfo = (get_modinfo_value(f, "kernel_version") != NULL);
if (!flag_quiet) {
char m_strversion[STRVERSIONLEN];
struct utsname uts;
if (m_has_modinfo) {
int m_version = new_get_module_version(f, m_strversion);
if (m_version == -1) {
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("can't find the kernel version "
"the module was compiled for");
}
}
uname(&uts);
if (strncmp(uts.release, m_strversion, STRVERSIONLEN) != 0) {
bb_error_msg("%skernel-module version mismatch\n"
"\t%s was compiled for kernel version %s\n"
"\twhile this kernel is version %s",
flag_force_load ? "warning: " : "",
m_name, m_strversion, uts.release);
if (!flag_force_load)
goto out;
}
}
#endif
if (query_module(NULL, 0, NULL, 0, NULL))
libbb: reduce the overhead of single parameter bb_error_msg() calls Back in 2007, commit 0c97c9d43707 ("'simple' error message functions by Loic Grenie") introduced bb_simple_perror_msg() to allow for a lower overhead call to bb_perror_msg() when only a string was being printed with no parameters. This saves space for some CPU architectures because it avoids the overhead of a call to a variadic function. However there has never been a simple version of bb_error_msg(), and since 2007 many new calls to bb_perror_msg() have been added that only take a single parameter and so could have been using bb_simple_perror_message(). This changeset introduces 'simple' versions of bb_info_msg(), bb_error_msg(), bb_error_msg_and_die(), bb_herror_msg() and bb_herror_msg_and_die(), and replaces all calls that only take a single parameter, or use something like ("%s", arg), with calls to the corresponding 'simple' version. Since it is likely that single parameter calls to the variadic functions may be accidentally reintroduced in the future a new debugging config option WARN_SIMPLE_MSG has been introduced. This uses some macro magic which will cause any such calls to generate a warning, but this is turned off by default to avoid use of the unpleasant macros in normal circumstances. This is a large changeset due to the number of calls that have been replaced. The only files that contain changes other than simple substitution of function calls are libbb.h, libbb/herror_msg.c, libbb/verror_msg.c and libbb/xfuncs_printf.c. In miscutils/devfsd.c, networking/udhcp/common.h and util-linux/mdev.c additonal macros have been added for logging so that single parameter and multiple parameter logging variants exist. The amount of space saved varies considerably by architecture, and was found to be as follows (for 'defconfig' using GCC 7.4): Arm: -92 bytes MIPS: -52 bytes PPC: -1836 bytes x86_64: -938 bytes Note that for the MIPS architecture only an exception had to be made disabling the 'simple' calls for 'udhcp' (in networking/udhcp/common.h) because it made these files larger on MIPS. Signed-off-by: James Byrne <james.byrne@origamienergy.com> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2019-07-02 15:05:03 +05:30
bb_simple_error_msg_and_die("old (unsupported) kernel");
new_get_kernel_symbols();
k_crcs = new_is_kernel_checksummed();
#if ENABLE_FEATURE_INSMOD_VERSION_CHECKING
{
int m_crcs = 0;
if (m_has_modinfo)
m_crcs = new_is_module_checksummed(f);
if (m_crcs != k_crcs)
obj_set_symbol_compare(f, ncv_strcmp, ncv_symbol_hash);
}
#endif
/* Let the module know about the kernel symbols. */
add_kernel_symbols(f);
/* Allocate common symbols, symbol tables, and string tables. */
new_create_this_module(f, m_name);
obj_check_undefineds(f);
obj_allocate_commons(f);
check_tainted_module(f, m_name);
/* Done with the module name, on to the optional var=value arguments */
new_process_module_arguments(f, options);
arch_create_got(f);
hide_special_symbols(f);
#if ENABLE_FEATURE_INSMOD_KSYMOOPS_SYMBOLS
add_ksymoops_symbols(f, m_filename, m_name);
#endif
new_create_module_ksymtab(f);
/* Find current size of the module */
m_size = obj_load_size(f);
m_addr = create_module(m_name, m_size);
if (m_addr == (ElfW(Addr))(-1)) switch (errno) {
case EEXIST:
bb_error_msg_and_die("a module named %s already exists", m_name);
case ENOMEM:
bb_error_msg_and_die("can't allocate kernel memory for module; needed %lu bytes",
m_size);
default:
bb_perror_msg_and_die("create_module: %s", m_name);
}
#if !LOADBITS
/*
* the PROGBITS section was not loaded by the obj_load
* now we can load them directly into the kernel memory
*/
if (!obj_load_progbits(image, image_size, f, (char*)m_addr)) {
delete_module(m_name, 0);
goto out;
}
#endif
if (!obj_relocate(f, m_addr)) {
delete_module(m_name, 0);
goto out;
}
if (!new_init_module(m_name, f, m_size)) {
delete_module(m_name, 0);
goto out;
}
if (flag_print_load_map)
print_load_map(f);
exit_status = EXIT_SUCCESS;
out:
if (mmaped)
munmap(image, image_size);
else
free(image);
free(m_name);
return exit_status;
}