#include <sys/wait.h>
#endif
-#if defined(linux_HOST_OS ) || defined(freebsd_HOST_OS) || \
- defined(dragonfly_HOST_OS) || defined(netbsd_HOST_OS ) || \
- defined(openbsd_HOST_OS ) || \
- ( defined(darwin_HOST_OS ) && !defined(powerpc_HOST_ARCH) ) || \
- defined(kfreebsdgnu_HOST_OS)
-/* Don't use mmap on powerpc-apple-darwin as mmap doesn't support
+#if !defined(powerpc_HOST_ARCH) && \
+ ( defined(linux_HOST_OS ) || defined(freebsd_HOST_OS) || \
+ defined(dragonfly_HOST_OS) || defined(netbsd_HOST_OS ) || \
+ defined(openbsd_HOST_OS ) || defined(darwin_HOST_OS ) || \
+ defined(kfreebsdgnu_HOST_OS) )
+/* Don't use mmap on powerpc_HOST_ARCH as mmap doesn't support
* reallocating but we need to allocate jump islands just after each
* object images. Otherwise relative branches to jump islands can fail
* due to 24-bits displacement overflow.
#elif defined(darwin_HOST_OS)
# define OBJFORMAT_MACHO
# include <regex.h>
+# include <mach/machine.h>
+# include <mach-o/fat.h>
# include <mach-o/loader.h>
# include <mach-o/nlist.h>
# include <mach-o/reloc.h>
SymI_HasProto(stg_asyncReadzh) \
SymI_HasProto(stg_asyncWritezh) \
SymI_HasProto(stg_asyncDoProczh) \
+ SymI_HasProto(getWin32ProgArgv) \
+ SymI_HasProto(setWin32ProgArgv) \
SymI_HasProto(memset) \
SymI_HasProto(inet_ntoa) \
SymI_HasProto(inet_addr) \
SymI_HasProto(stg_newTVarzh) \
SymI_HasProto(stg_noDuplicatezh) \
SymI_HasProto(stg_atomicModifyMutVarzh) \
+ SymI_HasProto(stg_casMutVarzh) \
SymI_HasProto(stg_newPinnedByteArrayzh) \
SymI_HasProto(stg_newAlignedPinnedByteArrayzh) \
SymI_HasProto(newSpark) \
# endif /* RTLD_DEFAULT */
compileResult = regcomp(&re_invalid,
- "(([^ \t()])+\\.so([^ \t:()])*):([ \t])*invalid ELF header",
+ "(([^ \t()])+\\.so([^ \t:()])*):([ \t])*(invalid ELF header|file too short)",
REG_EXTENDED);
ASSERT( compileResult == 0 );
compileResult = regcomp(&re_realso,
- "GROUP *\\( *(([^ )])+)",
+ "(GROUP|INPUT) *\\( *(([^ )])+)",
REG_EXTENDED);
ASSERT( compileResult == 0 );
# endif
if (regexec(&re_realso, line, (size_t) NMATCH, match, 0) == 0) {
// success -- try to dlopen the first named file
IF_DEBUG(linker, debugBelch("match%s\n",""));
- line[match[1].rm_eo] = '\0';
- errmsg = internal_dlopen(line+match[1].rm_so);
+ line[match[2].rm_eo] = '\0';
+ errmsg = internal_dlopen(line+match[2].rm_so);
break;
}
// if control reaches here, no GROUP ( ... ) directive was found
char *fileName;
size_t fileNameSize;
int isObject, isGnuIndex;
- char tmp[12];
+ char tmp[20];
char *gnuFileIndex;
int gnuFileIndexSize;
-#if !defined(USE_MMAP) && defined(darwin_HOST_OS)
+#if defined(darwin_HOST_OS)
+ int i;
+ uint32_t nfat_arch, nfat_offset, cputype, cpusubtype;
+#if defined(i386_HOST_ARCH)
+ const uint32_t mycputype = CPU_TYPE_X86;
+ const uint32_t mycpusubtype = CPU_SUBTYPE_X86_ALL;
+#elif defined(x86_64_HOST_ARCH)
+ const uint32_t mycputype = CPU_TYPE_X86_64;
+ const uint32_t mycpusubtype = CPU_SUBTYPE_X86_64_ALL;
+#elif defined(powerpc_HOST_ARCH)
+ const uint32_t mycputype = CPU_TYPE_POWERPC;
+ const uint32_t mycpusubtype = CPU_SUBTYPE_POWERPC_ALL;
+#elif defined(powerpc64_HOST_ARCH)
+ const uint32_t mycputype = CPU_TYPE_POWERPC64;
+ const uint32_t mycpusubtype = CPU_SUBTYPE_POWERPC_ALL;
+#else
+#error Unknown Darwin architecture
+#endif
+#if !defined(USE_MMAP)
int misalignment;
#endif
+#endif
IF_DEBUG(linker, debugBelch("loadArchive: start\n"));
IF_DEBUG(linker, debugBelch("loadArchive: Loading archive `%s'\n", path));
if (!f)
barf("loadObj: can't read `%s'", path);
+ /* Check if this is an archive by looking for the magic "!<arch>\n"
+ * string. Usually, if this fails, we barf and quit. On Darwin however,
+ * we may have a fat archive, which contains archives for more than
+ * one architecture. Fat archives start with the magic number 0xcafebabe,
+ * always stored big endian. If we find a fat_header, we scan through
+ * the fat_arch structs, searching through for one for our host
+ * architecture. If a matching struct is found, we read the offset
+ * of our archive data (nfat_offset) and seek forward nfat_offset bytes
+ * from the start of the file.
+ *
+ * A subtlety is that all of the members of the fat_header and fat_arch
+ * structs are stored big endian, so we need to call byte order
+ * conversion functions.
+ *
+ * If we find the appropriate architecture in a fat archive, we gobble
+ * its magic "!<arch>\n" string and continue processing just as if
+ * we had a single architecture archive.
+ */
+
n = fread ( tmp, 1, 8, f );
+ if (n != 8)
+ barf("loadArchive: Failed reading header from `%s'", path);
if (strncmp(tmp, "!<arch>\n", 8) != 0) {
+
+#if defined(darwin_HOST_OS)
+ /* Not a standard archive, look for a fat archive magic number: */
+ if (ntohl(*(uint32_t *)tmp) == FAT_MAGIC) {
+ nfat_arch = ntohl(*(uint32_t *)(tmp + 4));
+ IF_DEBUG(linker, debugBelch("loadArchive: found a fat archive containing %d architectures\n", nfat_arch));
+ nfat_offset = 0;
+
+ for (i = 0; i < (int)nfat_arch; i++) {
+ /* search for the right arch */
+ n = fread( tmp, 1, 20, f );
+ if (n != 8)
+ barf("loadArchive: Failed reading arch from `%s'", path);
+ cputype = ntohl(*(uint32_t *)tmp);
+ cpusubtype = ntohl(*(uint32_t *)(tmp + 4));
+
+ if (cputype == mycputype && cpusubtype == mycpusubtype) {
+ IF_DEBUG(linker, debugBelch("loadArchive: found my archive in a fat archive\n"));
+ nfat_offset = ntohl(*(uint32_t *)(tmp + 8));
+ break;
+ }
+ }
+
+ if (nfat_offset == 0) {
+ barf ("loadArchive: searched %d architectures, but no host arch found", (int)nfat_arch);
+ }
+ else {
+ n = fseek( f, nfat_offset, SEEK_SET );
+ if (n != 0)
+ barf("loadArchive: Failed to seek to arch in `%s'", path);
+ n = fread ( tmp, 1, 8, f );
+ if (n != 8)
+ barf("loadArchive: Failed reading header from `%s'", path);
+ if (strncmp(tmp, "!<arch>\n", 8) != 0) {
+ barf("loadArchive: couldn't find archive in `%s' at offset %d", path, nfat_offset);
+ }
+ }
+ }
+ else {
+ barf("loadArchive: Neither an archive, nor a fat archive: `%s'", path);
+ }
+
+#else
barf("loadArchive: Not an archive: `%s'", path);
+#endif
}
IF_DEBUG(linker, debugBelch("loadArchive: loading archive contents\n"));
barf("loadArchive: Failed reading file name from `%s'", path);
}
}
+
#if defined(darwin_HOST_OS)
- else {
- if (strncmp(fileName, "!<arch>\n", 8) == 0) {
- IF_DEBUG(linker, debugBelch("loadArchive: found the start of another archive, breaking\n"));
- break;
- }
+ if (strncmp(fileName, "!<arch>\n", 8) == 0) {
+ IF_DEBUG(linker, debugBelch("loadArchive: found the start of another archive, breaking\n"));
+ break;
}
#endif
IF_DEBUG(linker, debugBelch("loadArchive: size of this archive member is %d\n", memberSize));
n = fread ( tmp, 1, 2, f );
+ if (n != 2)
+ barf("loadArchive: Failed reading magic from `%s'", path);
if (strncmp(tmp, "\x60\x0A", 2) != 0)
barf("loadArchive: Failed reading magic from `%s' at %ld. Got %c%c",
path, ftell(f), tmp[0], tmp[1]);
// stgFree(oc->image);
// #endif
stgFree(oc->fileName);
+ stgFree(oc->archiveMemberName);
stgFree(oc->symbols);
stgFree(oc->sections);
stgFree(oc);
ocFlushInstructionCache( ObjectCode *oc )
{
/* The main object code */
- ocFlushInstructionCacheFrom(oc->image + oc->misalignment, oc->fileSize);
+ ocFlushInstructionCacheFrom(oc->image
+#ifdef darwin_HOST_OS
+ + oc->misalignment
+#endif
+ , oc->fileSize);
/* Jump Islands */
ocFlushInstructionCacheFrom(oc->symbol_extras, sizeof(SymbolExtra) * oc->n_symbol_extras);
* Generic ELF functions
*/
-static char *
-findElfSection ( void* objImage, Elf_Word sh_type )
-{
- char* ehdrC = (char*)objImage;
- Elf_Ehdr* ehdr = (Elf_Ehdr*)ehdrC;
- Elf_Shdr* shdr = (Elf_Shdr*)(ehdrC + ehdr->e_shoff);
- char* sh_strtab = ehdrC + shdr[ehdr->e_shstrndx].sh_offset;
- char* ptr = NULL;
- int i;
-
- for (i = 0; i < ehdr->e_shnum; i++) {
- if (shdr[i].sh_type == sh_type
- /* Ignore the section header's string table. */
- && i != ehdr->e_shstrndx
- /* Ignore string tables named .stabstr, as they contain
- debugging info. */
- && 0 != memcmp(".stabstr", sh_strtab + shdr[i].sh_name, 8)
- ) {
- ptr = ehdrC + shdr[i].sh_offset;
- break;
- }
- }
- return ptr;
-}
-
static int
ocVerifyImage_ELF ( ObjectCode* oc )
{
Elf_Sym* stab;
int i, j, nent, nstrtab, nsymtabs;
char* sh_strtab;
- char* strtab;
char* ehdrC = (char*)(oc->image);
Elf_Ehdr* ehdr = (Elf_Ehdr*)ehdrC;
ehdrC + shdr[i].sh_offset,
ehdrC + shdr[i].sh_offset + shdr[i].sh_size - 1));
- if (shdr[i].sh_type == SHT_REL) {
- IF_DEBUG(linker,debugBelch("Rel " ));
- } else if (shdr[i].sh_type == SHT_RELA) {
- IF_DEBUG(linker,debugBelch("RelA " ));
- } else {
- IF_DEBUG(linker,debugBelch(" "));
+#define SECTION_INDEX_VALID(ndx) (ndx > SHN_UNDEF && ndx < ehdr->e_shnum)
+
+ switch (shdr[i].sh_type) {
+
+ case SHT_REL:
+ case SHT_RELA:
+ IF_DEBUG(linker,debugBelch( shdr[i].sh_type == SHT_REL ? "Rel " : "RelA "));
+
+ if (!SECTION_INDEX_VALID(shdr[i].sh_link)) {
+ if (shdr[i].sh_link == SHN_UNDEF)
+ errorBelch("\n%s: relocation section #%d has no symbol table\n"
+ "This object file has probably been fully striped. "
+ "Such files cannot be linked.\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName, i);
+ else
+ errorBelch("\n%s: relocation section #%d has an invalid link field (%d)\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName,
+ i, shdr[i].sh_link);
+ return 0;
+ }
+ if (shdr[shdr[i].sh_link].sh_type != SHT_SYMTAB) {
+ errorBelch("\n%s: relocation section #%d does not link to a symbol table\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName, i);
+ return 0;
+ }
+ if (!SECTION_INDEX_VALID(shdr[i].sh_info)) {
+ errorBelch("\n%s: relocation section #%d has an invalid info field (%d)\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName,
+ i, shdr[i].sh_info);
+ return 0;
+ }
+
+ break;
+ case SHT_SYMTAB:
+ IF_DEBUG(linker,debugBelch("Sym "));
+
+ if (!SECTION_INDEX_VALID(shdr[i].sh_link)) {
+ errorBelch("\n%s: symbol table section #%d has an invalid link field (%d)\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName,
+ i, shdr[i].sh_link);
+ return 0;
+ }
+ if (shdr[shdr[i].sh_link].sh_type != SHT_STRTAB) {
+ errorBelch("\n%s: symbol table section #%d does not link to a string table\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName, i);
+
+ return 0;
+ }
+ break;
+ case SHT_STRTAB: IF_DEBUG(linker,debugBelch("Str ")); break;
+ default: IF_DEBUG(linker,debugBelch(" ")); break;
}
if (sh_strtab) {
IF_DEBUG(linker,debugBelch("sname=%s\n", sh_strtab + shdr[i].sh_name ));
}
}
- IF_DEBUG(linker,debugBelch( "\nString tables" ));
- strtab = NULL;
+ IF_DEBUG(linker,debugBelch( "\nString tables\n" ));
nstrtab = 0;
for (i = 0; i < ehdr->e_shnum; i++) {
if (shdr[i].sh_type == SHT_STRTAB
debugging info. */
&& 0 != memcmp(".stabstr", sh_strtab + shdr[i].sh_name, 8)
) {
- IF_DEBUG(linker,debugBelch(" section %d is a normal string table", i ));
- strtab = ehdrC + shdr[i].sh_offset;
+ IF_DEBUG(linker,debugBelch(" section %d is a normal string table\n", i ));
nstrtab++;
}
}
- if (nstrtab != 1) {
- errorBelch("%s: no string tables, or too many", oc->fileName);
- return 0;
+ if (nstrtab == 0) {
+ IF_DEBUG(linker,debugBelch(" no normal string tables (potentially, but not necessarily a problem)\n"));
}
nsymtabs = 0;
- IF_DEBUG(linker,debugBelch( "\nSymbol tables" ));
+ IF_DEBUG(linker,debugBelch( "Symbol tables\n" ));
for (i = 0; i < ehdr->e_shnum; i++) {
if (shdr[i].sh_type != SHT_SYMTAB) continue;
IF_DEBUG(linker,debugBelch( "section %d is a symbol table\n", i ));
}
IF_DEBUG(linker,debugBelch(" " ));
- IF_DEBUG(linker,debugBelch("name=%s\n", strtab + stab[j].st_name ));
+ IF_DEBUG(linker,debugBelch("name=%s\n",
+ ehdrC + shdr[shdr[i].sh_link].sh_offset
+ + stab[j].st_name ));
}
}
if (nsymtabs == 0) {
- errorBelch("%s: didn't find any symbol tables", oc->fileName);
- return 0;
+ // Not having a symbol table is not in principle a problem.
+ // When an object file has no symbols then the 'strip' program
+ // typically will remove the symbol table entirely.
+ IF_DEBUG(linker,debugBelch(" no symbol tables (potentially, but not necessarily a problem)\n"));
}
return 1;
char* ehdrC = (char*)(oc->image);
Elf_Ehdr* ehdr = (Elf_Ehdr*)ehdrC;
- char* strtab = findElfSection ( ehdrC, SHT_STRTAB );
+ char* strtab;
Elf_Shdr* shdr = (Elf_Shdr*) (ehdrC + ehdr->e_shoff);
ASSERT(symhash != NULL);
- if (!strtab) {
- errorBelch("%s: no strtab", oc->fileName);
- return 0;
- }
-
k = 0;
for (i = 0; i < ehdr->e_shnum; i++) {
/* Figure out what kind of section it is. Logic derived from
/* copy stuff into this module's object symbol table */
stab = (Elf_Sym*) (ehdrC + shdr[i].sh_offset);
+ strtab = ehdrC + shdr[shdr[i].sh_link].sh_offset;
nent = shdr[i].sh_size / sizeof(Elf_Sym);
oc->n_symbols = nent;
oc->symbols = stgMallocBytes(oc->n_symbols * sizeof(char*),
"ocGetNames_ELF(oc->symbols)");
+ //TODO: we ignore local symbols anyway right? So we can use the
+ // shdr[i].sh_info to get the index of the first non-local symbol
+ // ie we should use j = shdr[i].sh_info
for (j = 0; j < nent; j++) {
char isLocal = FALSE; /* avoids uninit-var warning */
relocations appear to be of this form. */
static int
do_Elf_Rel_relocations ( ObjectCode* oc, char* ehdrC,
- Elf_Shdr* shdr, int shnum,
- Elf_Sym* stab, char* strtab )
+ Elf_Shdr* shdr, int shnum )
{
int j;
char *symbol;
Elf_Word* targ;
Elf_Rel* rtab = (Elf_Rel*) (ehdrC + shdr[shnum].sh_offset);
+ Elf_Sym* stab;
+ char* strtab;
int nent = shdr[shnum].sh_size / sizeof(Elf_Rel);
int target_shndx = shdr[shnum].sh_info;
int symtab_shndx = shdr[shnum].sh_link;
+ int strtab_shndx = shdr[symtab_shndx].sh_link;
stab = (Elf_Sym*) (ehdrC + shdr[ symtab_shndx ].sh_offset);
+ strtab= (char*) (ehdrC + shdr[ strtab_shndx ].sh_offset);
targ = (Elf_Word*)(ehdrC + shdr[ target_shndx ].sh_offset);
- IF_DEBUG(linker,debugBelch( "relocations for section %d using symtab %d\n",
- target_shndx, symtab_shndx ));
+ IF_DEBUG(linker,debugBelch( "relocations for section %d using symtab %d and strtab %d\n",
+ target_shndx, symtab_shndx, strtab_shndx ));
/* Skip sections that we're not interested in. */
{
sparc-solaris relocations appear to be of this form. */
static int
do_Elf_Rela_relocations ( ObjectCode* oc, char* ehdrC,
- Elf_Shdr* shdr, int shnum,
- Elf_Sym* stab, char* strtab )
+ Elf_Shdr* shdr, int shnum )
{
int j;
char *symbol = NULL;
Elf_Addr targ;
Elf_Rela* rtab = (Elf_Rela*) (ehdrC + shdr[shnum].sh_offset);
+ Elf_Sym* stab;
+ char* strtab;
int nent = shdr[shnum].sh_size / sizeof(Elf_Rela);
int target_shndx = shdr[shnum].sh_info;
int symtab_shndx = shdr[shnum].sh_link;
+ int strtab_shndx = shdr[symtab_shndx].sh_link;
stab = (Elf_Sym*) (ehdrC + shdr[ symtab_shndx ].sh_offset);
+ strtab= (char*) (ehdrC + shdr[ strtab_shndx ].sh_offset);
targ = (Elf_Addr) (ehdrC + shdr[ target_shndx ].sh_offset);
IF_DEBUG(linker,debugBelch( "relocations for section %d using symtab %d\n",
target_shndx, symtab_shndx ));
static int
ocResolve_ELF ( ObjectCode* oc )
{
- char *strtab;
int shnum, ok;
- Elf_Sym* stab = NULL;
char* ehdrC = (char*)(oc->image);
Elf_Ehdr* ehdr = (Elf_Ehdr*) ehdrC;
Elf_Shdr* shdr = (Elf_Shdr*) (ehdrC + ehdr->e_shoff);
- /* first find "the" symbol table */
- stab = (Elf_Sym*) findElfSection ( ehdrC, SHT_SYMTAB );
-
- /* also go find the string table */
- strtab = findElfSection ( ehdrC, SHT_STRTAB );
-
- if (stab == NULL || strtab == NULL) {
- errorBelch("%s: can't find string or symbol table", oc->fileName);
- return 0;
- }
-
/* Process the relocation sections. */
for (shnum = 0; shnum < ehdr->e_shnum; shnum++) {
if (shdr[shnum].sh_type == SHT_REL) {
- ok = do_Elf_Rel_relocations ( oc, ehdrC, shdr,
- shnum, stab, strtab );
+ ok = do_Elf_Rel_relocations ( oc, ehdrC, shdr, shnum );
if (!ok) return ok;
}
else
if (shdr[shnum].sh_type == SHT_RELA) {
- ok = do_Elf_Rela_relocations ( oc, ehdrC, shdr,
- shnum, stab, strtab );
+ ok = do_Elf_Rela_relocations ( oc, ehdrC, shdr, shnum );
if (!ok) return ok;
}
}
if( i == ehdr->e_shnum )
{
- errorBelch( "This ELF file contains no symtab" );
- return 0;
+ // Not having a symbol table is not in principle a problem.
+ // When an object file has no symbols then the 'strip' program
+ // typically will remove the symbol table entirely.
+ IF_DEBUG(linker, debugBelch( "The ELF file %s contains no symtab\n",
+ oc->archiveMemberName ? oc->archiveMemberName : oc->fileName ));
+ return 1;
}
if( shdr[i].sh_entsize != sizeof( Elf_Sym ) )