modpost.c

linux 内核源代码

 **/
static int strrcmp(const char *s, const char *sub)
{
        int slen, sublen;

	if (!s || !sub)
		return 1;

	slen = strlen(s);
        sublen = strlen(sub);

	if ((slen == 0) || (sublen == 0))
		return 1;

        if (sublen > slen)
                return 1;

        return memcmp(s + slen - sublen, sub, sublen);
}

/*
 * Functions used only during module init is marked __init and is stored in
 * a .init.text section. Likewise data is marked __initdata and stored in
 * a .init.data section.
 * If this section is one of these sections return 1
 * See include/linux/init.h for the details
 */
static int init_section(const char *name)
{
	if (strcmp(name, ".init") == 0)
		return 1;
	if (strncmp(name, ".init.", strlen(".init.")) == 0)
		return 1;
	return 0;
}

/*
 * Functions used only during module exit is marked __exit and is stored in
 * a .exit.text section. Likewise data is marked __exitdata and stored in
 * a .exit.data section.
 * If this section is one of these sections return 1
 * See include/linux/init.h for the details
 **/
static int exit_section(const char *name)
{
	if (strcmp(name, ".exit.text") == 0)
		return 1;
	if (strcmp(name, ".exit.data") == 0)
		return 1;
	return 0;

}

/*
 * Data sections are named like this:
 * .data | .data.rel | .data.rel.*
 * Return 1 if the specified section is a data section
 */
static int data_section(const char *name)
{
	if ((strcmp(name, ".data") == 0) ||
	    (strcmp(name, ".data.rel") == 0) ||
	    (strncmp(name, ".data.rel.", strlen(".data.rel.")) == 0))
		return 1;
	else
		return 0;
}

/**
 * Whitelist to allow certain references to pass with no warning.
 *
 * Pattern 0:
 *   Do not warn if funtion/data are marked with __init_refok/__initdata_refok.
 *   The pattern is identified by:
 *   fromsec = .text.init.refok* | .data.init.refok*
 *
 * Pattern 1:
 *   If a module parameter is declared __initdata and permissions=0
 *   then this is legal despite the warning generated.
 *   We cannot see value of permissions here, so just ignore
 *   this pattern.
 *   The pattern is identified by:
 *   tosec   = .init.data
 *   fromsec = .data*
 *   atsym   =__param*
 *
 * Pattern 2:
 *   Many drivers utilise a *driver container with references to
 *   add, remove, probe functions etc.
 *   These functions may often be marked __init and we do not want to
 *   warn here.
 *   the pattern is identified by:
 *   tosec   = init or exit section
 *   fromsec = data section
 *   atsym = *driver, *_template, *_sht, *_ops, *_probe, *probe_one, *_console, *_timer
 *
 * Pattern 3:
 *   Whitelist all refereces from .text.head to .init.data
 *   Whitelist all refereces from .text.head to .init.text
 *
 * Pattern 4:
 *   Some symbols belong to init section but still it is ok to reference
 *   these from non-init sections as these symbols don't have any memory
 *   allocated for them and symbol address and value are same. So even
 *   if init section is freed, its ok to reference those symbols.
 *   For ex. symbols marking the init section boundaries.
 *   This pattern is identified by
 *   refsymname = __init_begin, _sinittext, _einittext
 *
 * Pattern 5:
 *   Xtensa uses literal sections for constants that are accessed PC-relative.
 *   Literal sections may safely reference their text sections.
 *   (Note that the name for the literal section omits any trailing '.text')
 *   tosec = <section>[.text]
 *   fromsec = <section>.literal
 **/
static int secref_whitelist(const char *modname, const char *tosec,
			    const char *fromsec, const char *atsym,
			    const char *refsymname)
{
	int len;
	const char **s;
	const char *pat2sym[] = {
		"driver",
		"_template", /* scsi uses *_template a lot */
		"_timer",    /* arm uses ops structures named _timer a lot */
		"_sht",      /* scsi also used *_sht to some extent */
		"_ops",
		"_probe",
		"_probe_one",
		"_console",
		NULL
	};

	const char *pat3refsym[] = {
		"__init_begin",
		"_sinittext",
		"_einittext",
		NULL
	};

	/* Check for pattern 0 */
	if ((strncmp(fromsec, ".text.init.refok", strlen(".text.init.refok")) == 0) ||
	    (strncmp(fromsec, ".exit.text.refok", strlen(".exit.text.refok")) == 0) ||
	    (strncmp(fromsec, ".data.init.refok", strlen(".data.init.refok")) == 0))
		return 1;

	/* Check for pattern 1 */
	if ((strcmp(tosec, ".init.data") == 0) &&
	    (strncmp(fromsec, ".data", strlen(".data")) == 0) &&
	    (strncmp(atsym, "__param", strlen("__param")) == 0))
		return 1;

	/* Check for pattern 2 */
	if ((init_section(tosec) || exit_section(tosec)) && data_section(fromsec))
		for (s = pat2sym; *s; s++)
			if (strrcmp(atsym, *s) == 0)
				return 1;

	/* Check for pattern 3 */
	if ((strcmp(fromsec, ".text.head") == 0) &&
		((strcmp(tosec, ".init.data") == 0) ||
		(strcmp(tosec, ".init.text") == 0)))
	return 1;

	/* Check for pattern 4 */
	for (s = pat3refsym; *s; s++)
		if (strcmp(refsymname, *s) == 0)
			return 1;

	/* Check for pattern 5 */
	if (strrcmp(tosec, ".text") == 0)
		len = strlen(tosec) - strlen(".text");
	else
		len = strlen(tosec);
	if ((strncmp(tosec, fromsec, len) == 0) && (strlen(fromsec) > len) &&
	    (strcmp(fromsec + len, ".literal") == 0))
		return 1;

	return 0;
}

/**
 * Find symbol based on relocation record info.
 * In some cases the symbol supplied is a valid symbol so
 * return refsym. If st_name != 0 we assume this is a valid symbol.
 * In other cases the symbol needs to be looked up in the symbol table
 * based on section and address.
 *  **/
static Elf_Sym *find_elf_symbol(struct elf_info *elf, Elf_Addr addr,
				Elf_Sym *relsym)
{
	Elf_Sym *sym;

	if (relsym->st_name != 0)
		return relsym;
	for (sym = elf->symtab_start; sym < elf->symtab_stop; sym++) {
		if (sym->st_shndx != relsym->st_shndx)
			continue;
		if (ELF_ST_TYPE(sym->st_info) == STT_SECTION)
			continue;
		if (sym->st_value == addr)
			return sym;
	}
	return NULL;
}

static inline int is_arm_mapping_symbol(const char *str)
{
	return str[0] == '$' && strchr("atd", str[1])
	       && (str[2] == '\0' || str[2] == '.');
}

/*
 * If there's no name there, ignore it; likewise, ignore it if it's
 * one of the magic symbols emitted used by current ARM tools.
 *
 * Otherwise if find_symbols_between() returns those symbols, they'll
 * fail the whitelist tests and cause lots of false alarms ... fixable
 * only by merging __exit and __init sections into __text, bloating
 * the kernel (which is especially evil on embedded platforms).
 */
static inline int is_valid_name(struct elf_info *elf, Elf_Sym *sym)
{
	const char *name = elf->strtab + sym->st_name;

	if (!name || !strlen(name))
		return 0;
	return !is_arm_mapping_symbol(name);
}

/*
 * Find symbols before or equal addr and after addr - in the section sec.
 * If we find two symbols with equal offset prefer one with a valid name.
 * The ELF format may have a better way to detect what type of symbol
 * it is, but this works for now.
 **/
static void find_symbols_between(struct elf_info *elf, Elf_Addr addr,
				 const char *sec,
			         Elf_Sym **before, Elf_Sym **after)
{
	Elf_Sym *sym;
	Elf_Ehdr *hdr = elf->hdr;
	Elf_Addr beforediff = ~0;
	Elf_Addr afterdiff = ~0;
	const char *secstrings = (void *)hdr +
				 elf->sechdrs[hdr->e_shstrndx].sh_offset;

	*before = NULL;
	*after = NULL;

	for (sym = elf->symtab_start; sym < elf->symtab_stop; sym++) {
		const char *symsec;

		if (sym->st_shndx >= SHN_LORESERVE)
			continue;
		symsec = secstrings + elf->sechdrs[sym->st_shndx].sh_name;
		if (strcmp(symsec, sec) != 0)
			continue;
		if (!is_valid_name(elf, sym))
			continue;
		if (sym->st_value <= addr) {
			if ((addr - sym->st_value) < beforediff) {
				beforediff = addr - sym->st_value;
				*before = sym;
			}
			else if ((addr - sym->st_value) == beforediff) {
				*before = sym;
			}
		}
		else
		{
			if ((sym->st_value - addr) < afterdiff) {
				afterdiff = sym->st_value - addr;
				*after = sym;
			}
			else if ((sym->st_value - addr) == afterdiff) {
				*after = sym;
			}
		}
	}
}

/**
 * Print a warning about a section mismatch.
 * Try to find symbols near it so user can find it.
 * Check whitelist before warning - it may be a false positive.
 **/
static void warn_sec_mismatch(const char *modname, const char *fromsec,
			      struct elf_info *elf, Elf_Sym *sym, Elf_Rela r)
{
	const char *refsymname = "";
	Elf_Sym *before, *after;
	Elf_Sym *refsym;
	Elf_Ehdr *hdr = elf->hdr;
	Elf_Shdr *sechdrs = elf->sechdrs;
	const char *secstrings = (void *)hdr +
				 sechdrs[hdr->e_shstrndx].sh_offset;
	const char *secname = secstrings + sechdrs[sym->st_shndx].sh_name;

	find_symbols_between(elf, r.r_offset, fromsec, &before, &after);

	refsym = find_elf_symbol(elf, r.r_addend, sym);
	if (refsym && strlen(elf->strtab + refsym->st_name))
		refsymname = elf->strtab + refsym->st_name;

	/* check whitelist - we may ignore it */
	if (secref_whitelist(modname, secname, fromsec,
			     before ? elf->strtab + before->st_name : "",
	                     refsymname))
		return;

	if (before && after) {
		warn("%s(%s+0x%llx): Section mismatch: reference to %s:%s "
		     "(between '%s' and '%s')\n",
		     modname, fromsec, (unsigned long long)r.r_offset,
		     secname, refsymname,
		     elf->strtab + before->st_name,
		     elf->strtab + after->st_name);
	} else if (before) {
		warn("%s(%s+0x%llx): Section mismatch: reference to %s:%s "
		     "(after '%s')\n",
		     modname, fromsec, (unsigned long long)r.r_offset,
		     secname, refsymname,
		     elf->strtab + before->st_name);
	} else if (after) {
		warn("%s(%s+0x%llx): Section mismatch: reference to %s:%s "
		     "before '%s' (at offset -0x%llx)\n",
		     modname, fromsec, (unsigned long long)r.r_offset,
		     secname, refsymname,
		     elf->strtab + after->st_name);
	} else {
		warn("%s(%s+0x%llx): Section mismatch: reference to %s:%s\n",
		     modname, fromsec, (unsigned long long)r.r_offset,
		     secname, refsymname);
	}
}

static unsigned int *reloc_location(struct elf_info *elf,
					   int rsection, Elf_Rela *r)
{
	Elf_Shdr *sechdrs = elf->sechdrs;
	int section = sechdrs[rsection].sh_info;

	return (void *)elf->hdr + sechdrs[section].sh_offset +
		(r->r_offset - sechdrs[section].sh_addr);
}

static int addend_386_rel(struct elf_info *elf, int rsection, Elf_Rela *r)
{
	unsigned int r_typ = ELF_R_TYPE(r->r_info);
	unsigned int *location = reloc_location(elf, rsection, r);

	switch (r_typ) {
	case R_386_32:
		r->r_addend = TO_NATIVE(*location);
		break;
	case R_386_PC32:
		r->r_addend = TO_NATIVE(*location) + 4;
		/* For CONFIG_RELOCATABLE=y */
		if (elf->hdr->e_type == ET_EXEC)
			r->r_addend += r->r_offset;
		break;
	}
	return 0;
}

static int addend_arm_rel(struct elf_info *elf, int rsection, Elf_Rela *r)
{
	unsigned int r_typ = ELF_R_TYPE(r->r_info);

	switch (r_typ) {
	case R_ARM_ABS32:
		/* From ARM ABI: (S + A) | T */
		r->r_addend = (int)(long)(elf->symtab_start + ELF_R_SYM(r->r_info));
		break;
	case R_ARM_PC24:
		/* From ARM ABI: ((S + A) | T) - P */
		r->r_addend = (int)(long)(elf->hdr + elf->sechdrs[rsection].sh_offset +
		                          (r->r_offset - elf->sechdrs[rsection].sh_addr));
		break;
	default:
		return 1;
	}
	return 0;
}

static int addend_mips_rel(struct elf_info *elf, int rsection, Elf_Rela *r)
{
	unsigned int r_typ = ELF_R_TYPE(r->r_info);
	unsigned int *location = reloc_location(elf, rsection, r);
	unsigned int inst;

	if (r_typ == R_MIPS_HI16)
		return 1;	/* skip this */
	inst = TO_NATIVE(*location);
	switch (r_typ) {
	case R_MIPS_LO16:
		r->r_addend = inst & 0xffff;
		break;
	case R_MIPS_26:
		r->r_addend = (inst & 0x03ffffff) << 2;
		break;
	case R_MIPS_32:
		r->r_addend = inst;
		break;
	}
	return 0;
}

/**
 * A module includes a number of sections that are discarded
 * either when loaded or when used as built-in.
 * For loaded modules all functions marked __init and all data
 * marked __initdata will be discarded when the module has been intialized.
 * Likewise for modules used built-in the sections marked __exit
 * are discarded because __exit marked function are supposed to be called
 * only when a moduel is unloaded which never happes for built-in modules.
 * The check_sec_ref() function traverses all relocation records
 * to find all references to a section that reference a section that will
 * be discarded and warns about it.
 **/
static void check_sec_ref(struct module *mod, const char *modname,
			  struct elf_info *elf,
			  int section(const char*),
			  int section_ref_ok(const char *))
{
	int i;
	Elf_Sym  *sym;
	Elf_Ehdr *hdr = elf->hdr;
	Elf_Shdr *sechdrs = elf->sechdrs;
	const char *secstrings = (void *)hdr +
				 sechdrs[hdr->e_shstrndx].sh_offset;

	/* Walk through all sections */
	for (i = 0; i < hdr->e_shnum; i++) {
		const char *name = secstrings + sechdrs[i].sh_name;
		const char *secname;
		Elf_Rela r;
		unsigned int r_sym;
		/* We want to process only relocation sections and not .init */
		if (sechdrs[i].sh_type == SHT_RELA) {
			Elf_Rela *rela;
			Elf_Rela *start = (void *)hdr + sechdrs[i].sh_offset;
			Elf_Rela *stop  = (void*)start + sechdrs[i].sh_size;
			name += strlen(".rela");
			if (section_ref_ok(name))
				continue;

			for (rela = start; rela < stop; rela++) {
				r.r_offset = TO_NATIVE(rela->r_offset);
#if KERNEL_ELFCLASS == ELFCLASS64
				if (hdr->e_machine == EM_MIPS) {
					unsigned int r_typ;
					r_sym = ELF64_MIPS_R_SYM(rela->r_info);
					r_sym = TO_NATIVE(r_sym);
					r_typ = ELF64_MIPS_R_TYPE(rela->r_info);
					r.r_info = ELF64_R_INFO(r_sym, r_typ);
				} else {
					r.r_info = TO_NATIVE(rela->r_info);
					r_sym = ELF_R_SYM(r.r_info);
				}
#else
				r.r_info = TO_NATIVE(rela->r_info);
				r_sym = ELF_R_SYM(r.r_info);
#endif
				r.r_addend = TO_NATIVE(rela->r_addend);
				sym = elf->symtab_start + r_sym;
				/* Skip special sections */
				if (sym->st_shndx >= SHN_LORESERVE)
					continue;

				secname = secstrings +
					sechdrs[sym->st_shndx].sh_name;
				if (section(secname))
					warn_sec_mismatch(modname, name,
							  elf, sym, r);
			}
		} else if (sechdrs[i].sh_type == SHT_REL) {
			Elf_Rel *rel;
			Elf_Rel *start = (void *)hdr + sechdrs[i].sh_offset;
			Elf_Rel *stop  = (void*)start + sechdrs[i].sh_size;
			name += strlen(".rel");
			if (section_ref_ok(name))
				continue;

			for (rel = start; rel < stop; rel++) {
				r.r_offset = TO_NATIVE(rel->r_offset);
#if KERNEL_ELFCLASS == ELFCLASS64
				if (hdr->e_machine == EM_MIPS) {
					unsigned int r_typ;
					r_sym = ELF64_MIPS_R_SYM(rel->r_info);
					r_sym = TO_NATIVE(r_sym);
					r_typ = ELF64_MIPS_R_TYPE(rel->r_info);
					r.r_info = ELF64_R_INFO(r_sym, r_typ);
				} else {
					r.r_info = TO_NATIVE(rel->r_info);
					r_sym = ELF_R_SYM(r.r_info);
				}
#else
				r.r_info = TO_NATIVE(rel->r_info);
				r_sym = ELF_R_SYM(r.r_info);
#endif
				r.r_addend = 0;
				switch (hdr->e_machine) {
				case EM_386:
					if (addend_386_rel(elf, i, &r))
						continue;
					break;
				case EM_ARM:
					if(addend_arm_rel(elf, i, &r))
						continue;
					break;
				case EM_MIPS:
					if (addend_mips_rel(elf, i, &r))
						continue;
					break;
				}
				sym = elf->symtab_start + r_sym;
				/* Skip special sections */
				if (sym->st_shndx >= SHN_LORESERVE)
					continue;

				secname = secstrings +
					sechdrs[sym->st_shndx].sh_name;
				if (section(secname))
					warn_sec_mismatch(modname, name,
							  elf, sym, r);
			}
		}
	}
}

/*
 * Identify sections from which references to either a
 * .init or a .exit section is OK.
 *
 * [OPD] Keith Ownes <[e-mail]> commented:
 * For our future {in}sanity, add a comment that this is the ppc .opd
 * section, not the ia64 .opd section.
 * ia64 .opd should not point to discarded sections.
 * [.rodata] like for .init.text we ignore .rodata references -same reason
 */
static int initexit_section_ref_ok(const char *name)
{
	const char **s;
	/* Absolute section names */
	const char *namelist1[] = {
		"__bug_table",		/* used by powerpc for BUG() */
		"__ex_table",
		".altinstructions",
		".cranges",		/* used by sh64 */
		".fixup",
		".machvec",		/* ia64 + powerpc uses these */
		".machine.desc",
		".opd",			/* See comment [OPD] */
		"__dbe_table",
		".parainstructions",
		".pdr",
		".plt",			/* seen on ARCH=um build on x86_64. Harmless */
		".smp_locks",
		".stab",
		".m68k_fixup",
		".xt.prop",		/* xtensa informational section */
		".xt.lit",		/* xtensa informational section */
		NULL
	};
	/* Start of section names */
	const char *namelist2[] = {
		".debug",
		".eh_frame",
		".note",		/* ignore ELF notes - may contain anything */
		".got",			/* powerpc - global offset table */
		".toc",			/* powerpc - table of contents */
		NULL
	};
	/* part of section name */