/** Initialize by parsing a file. */
SgAsmElfEHFrameSection *
SgAsmElfEHFrameSection::parse()
{
SgAsmElfSection::parse();
SgAsmElfFileHeader *fhdr = get_elf_header();
ROSE_ASSERT(fhdr!=NULL);
rose_addr_t record_offset=0;
std::map<rose_addr_t, SgAsmElfEHFrameEntryCI*> cies;
while (record_offset<get_size()) {
rose_addr_t at = record_offset;
unsigned char u8_disk;
uint32_t u32_disk;
uint64_t u64_disk;
/* Length or extended length */
rose_addr_t length_field_size = 4; /*number of bytes not counted in length*/
read_content_local(at, &u32_disk, 4); at += 4;
rose_addr_t record_size = disk_to_host(fhdr->get_sex(), u32_disk);
if (record_size==0xffffffff) {
read_content_local(at, &u64_disk, 8); at += 8;
record_size = disk_to_host(fhdr->get_sex(), u64_disk);
length_field_size += 8; /*FIXME: it's not entirely clear whether ExtendedLength includes this field*/
}
if (0==record_size)
break;
/* Backward offset to CIE record, or zero if this is a CIE record. */
read_content_local(at, &u32_disk, 4); at += 4;
rose_addr_t cie_back_offset = disk_to_host(fhdr->get_sex(), u32_disk);
if (0==cie_back_offset) {
/* This is a CIE record */
SgAsmElfEHFrameEntryCI *cie = new SgAsmElfEHFrameEntryCI(this);
cies[record_offset] = cie;
/* Version */
uint8_t cie_version;
read_content_local(at++, &cie_version, 1);
cie->set_version(cie_version);
/* Augmentation String */
std::string astr = read_content_local_str(at);
at += astr.size() + 1;
cie->set_augmentation_string(astr);
/* Alignment factors */
cie->set_code_alignment_factor(read_content_local_uleb128(&at));
cie->set_data_alignment_factor(read_content_local_sleb128(&at));
/* Augmentation data length. This is apparently the length of the data described by the Augmentation String plus
* the Initial Instructions plus any padding. [RPM 2009-01-15] */
cie->set_augmentation_data_length(read_content_local_uleb128(&at));
/* Augmentation data. The format of the augmentation data in the CIE record is determined by reading the
* characters of the augmentation string. */
if (!astr.empty() && astr[0]=='z') {
for (size_t i=1; i<astr.size(); i++) {
if ('L'==astr[i]) {
read_content_local(at++, &u8_disk, 1);
cie->set_lsda_encoding(u8_disk);
} else if ('P'==astr[i]) {
/* The first byte is an encoding method which describes the following bytes, which are the address of
* a Personality Routine Handler. There appears to be very little documentation about these fields. */
read_content_local(at++, &u8_disk, 1);
cie->set_prh_encoding(u8_disk);
switch (cie->get_prh_encoding()) {
case 0x05: /* See Ubuntu 32bit /usr/bin/aptitude */
case 0x06: /* See second CIE record for Gentoo-Amd64 /usr/bin/addftinfo */
case 0x07: /* See first CIE record for Gentoo-Amd64 /usr/bin/addftinfo */
read_content_local(at++, &u8_disk, 1); /* not sure what this is; arg for __gxx_personality_v0? */
cie->set_prh_arg(u8_disk);
read_content_local(at, &u32_disk, 4); at+=4; /* address of <__gxx_personality_v0@plt> */
cie->set_prh_addr(ByteOrder::le_to_host(u32_disk));
break;
case 0x09: /* *.o file generated by gcc-4.0.x */
/* FIXME: Cannot find any info about this entry. Fix SgAsmElfEHFrameSection::parse() if we
* ever figure this out. [RPM 2009-09-29] */
/*fallthrough*/
default: {
if (++nwarnings<=WARNING_LIMIT) {
fprintf(stderr, "%s:%u: warning: ELF CIE 0x%08"PRIx64" has unknown PRH encoding 0x%02x\n",
__FILE__, __LINE__, get_offset()+record_offset, cie->get_prh_encoding());
if (WARNING_LIMIT==nwarnings)
fprintf(stderr, " (additional frame warnings will be suppressed)\n");
}
break;
}
}
} else if ('R'==astr[i]) {
read_content_local(at++, &u8_disk, 1);
cie->set_addr_encoding(u8_disk);
} else if ('S'==astr[i]) {
/* See http://lkml.indiana.edu/hypermail/linux/kernel/0602.3/1144.html and GCC PR #26208*/
cie->set_sig_frame(true);
} else {
/* Some stuff we don't handle yet. Warn about it and don't read anything. */
if (++nwarnings<=WARNING_LIMIT) {
fprintf(stderr, "%s:%u: warning: ELF CIE 0x%08"PRIx64" has invalid augmentation string \"%s\"\n",
__FILE__, __LINE__, get_offset()+record_offset, escapeString(astr).c_str());
if (WARNING_LIMIT==nwarnings)
fprintf(stderr, " (additional frame warnings will be suppressed)\n");
}
}
}
}
/* Initial instructions. These are apparently included in the augmentation_data_length. The final instructions can
* be zero padding (no-op instructions) to bring the record up to a multiple of the word size. */
rose_addr_t init_insn_size = (length_field_size + record_size) - (at - record_offset);
cie->get_instructions() = read_content_local_ucl(at, init_insn_size);
ROSE_ASSERT(cie->get_instructions().size()==init_insn_size);
} else {
/* This is a FDE record */
rose_addr_t cie_offset = record_offset + length_field_size - cie_back_offset;
assert(cies.find(cie_offset)!=cies.end());
SgAsmElfEHFrameEntryCI *cie = cies[cie_offset];
SgAsmElfEHFrameEntryFD *fde = new SgAsmElfEHFrameEntryFD(cie);
/* PC Begin (begin_rva) and size */
switch (cie->get_addr_encoding()) {
case -1: /* No address encoding specified */
case 0x01:
case 0x03:
case 0x1b: /* Address doesn't look valid (e.g., 0xfffd74e8) but still four bytes [RPM 2008-01-16]*/
{
read_content_local(at, &u32_disk, 4); at+=4;
fde->set_begin_rva(ByteOrder::le_to_host(u32_disk));
read_content_local(at, &u32_disk, 4); at+=4;
fde->set_size(ByteOrder::le_to_host(u32_disk));
break;
}
default:
fprintf(stderr, "%s:%u: ELF CIE 0x%08"PRIx64", FDE 0x%08"PRIx64": unknown address encoding: 0x%02x\n",
__FILE__, __LINE__, get_offset()+cie_offset, get_offset()+record_offset, cie->get_addr_encoding());
abort();
}
/* Augmentation Data */
std::string astring = cie->get_augmentation_string();
if (astring.size()>0 && astring[0]=='z') {
rose_addr_t aug_length = read_content_local_uleb128(&at);
fde->get_augmentation_data() = read_content_local_ucl(at, aug_length);
at += aug_length;
ROSE_ASSERT(fde->get_augmentation_data().size()==aug_length);
}
/* Call frame instructions */
rose_addr_t cf_insn_size = (length_field_size + record_size) - (at - record_offset);
fde->get_instructions() = read_content_local_ucl(at, cf_insn_size);
ROSE_ASSERT(fde->get_instructions().size()==cf_insn_size);
}
record_offset += length_field_size + record_size;
}
return this;
}
/** Initialize by parsing a file. */
SgAsmElfEHFrameSection *
SgAsmElfEHFrameSection::parse()
{
SgAsmElfSection::parse();
SgAsmElfFileHeader *fhdr = get_elf_header();
ROSE_ASSERT(fhdr!=NULL);
rose_addr_t record_offset=0;
std::map<rose_addr_t, SgAsmElfEHFrameEntryCI*> cies;
// This section consists of a Common Information Entry (CIE) followed by one or more Frame Description Entry (FDE) and this
// pattern is repeated for the entire section or until a termination record is reached. There is typically one CIE per
// object file and one FDE per function. The CIE and FDE together describe how to unwind the caller during an exception if
// the current instruction pointer is in the range covered by the FDE.
Sawyer::Optional<rose_addr_t> prevCieOffset;
while (record_offset<get_size()) {
rose_addr_t at = record_offset;
unsigned char u8_disk;
uint32_t u32_disk;
uint64_t u64_disk;
/* Length or extended length */
rose_addr_t length_field_size = 4; /*number of bytes not counted in length*/
read_content_local(at, &u32_disk, 4); at += 4;
rose_addr_t record_size = disk_to_host(fhdr->get_sex(), u32_disk);
if (record_size==0xffffffff) {
read_content_local(at, &u64_disk, 8); at += 8;
record_size = disk_to_host(fhdr->get_sex(), u64_disk);
length_field_size += 8; // length field size is both "length" and "extended length"
}
if (0==record_size) {
record_offset += length_field_size;
break;
}
/* Backward offset to CIE record, or zero if this is a CIE record. */
read_content_local(at, &u32_disk, 4); at += 4;
rose_addr_t cie_back_offset = disk_to_host(fhdr->get_sex(), u32_disk);
if (0==cie_back_offset) {
/* This is a CIE record */
SgAsmElfEHFrameEntryCI *cie = new SgAsmElfEHFrameEntryCI(this);
cies[record_offset] = cie;
prevCieOffset = record_offset;
/* Version. This parser was written based on version 1 documentation. */
uint8_t cie_version;
read_content_local(at++, &cie_version, 1);
cie->set_version(cie_version);
/* Augmentation String */
std::string astr = read_content_local_str(at);
at += astr.size() + 1;
cie->set_augmentation_string(astr);
/* EH data: 4 or 8 bytes depending on word size, only present "if the augmentation string contains 'eh'". The word
* "contains" in the documentation apparently means the string is _equal_ to "eh". */
if (astr == "eh") {
uint64_t eh_data;
if (4==fhdr->get_exec_format()->get_word_size()) {
read_content_local(at, &u32_disk, 4); at += 4;
eh_data = disk_to_host(fhdr->get_sex(), u32_disk);
} else {
read_content_local(at, &u64_disk, 8); at += 8;
eh_data = disk_to_host(fhdr->get_sex(), u64_disk);
}
cie->set_eh_data(eh_data);
}
/* Alignment factors:
* + code_alignment_factor: An unsigned LEB128 encoded value that is factored out of all advance location
* instructions that are associated with this CIE or its FDEs. This value shall be multiplied by the delta
* argument of an adavance location instruction to obtain the new location value.
* + data_alignment_factor: A signed LEB128 encoded value that is factored out of all offset instructions that are
* associated with this CIE or its FDEs. This value shall be multiplied by the register offset argument of an
* offset instruction to obtain the new offset value. */
cie->set_code_alignment_factor(read_content_local_uleb128(&at));
cie->set_data_alignment_factor(read_content_local_sleb128(&at));
/* Augmentation data length. This is apparently the length of the data described by the Augmentation String plus
* the Initial Instructions plus any padding [RPM 2009-01-15]. This field is present only if the augmentation
* string starts with the character "z" [Robb P. Matzke 2014-12-15]. */
if (!astr.empty() && astr[0]=='z')
cie->set_augmentation_data_length(read_content_local_uleb128(&at));
/* Augmentation data. The format of the augmentation data in the CIE record is determined by reading the
* characters of the augmentation string. */
if (!astr.empty() && astr[0]=='z') {
for (size_t i=1; i<astr.size(); i++) {
if ('L'==astr[i]) {
read_content_local(at++, &u8_disk, 1);
cie->set_lsda_encoding(u8_disk);
} else if ('P'==astr[i]) {
/* The first byte is an encoding method which describes the following bytes, which are the address of
* a Personality Routine Handler. There appears to be very little documentation about these fields. */
read_content_local(at++, &u8_disk, 1);
cie->set_prh_encoding(u8_disk);
switch (cie->get_prh_encoding()) {
case 0x05: /* See Ubuntu 32bit /usr/bin/aptitude */
case 0x06: /* See second CIE record for Gentoo-Amd64 /usr/bin/addftinfo */
case 0x07: /* See first CIE record for Gentoo-Amd64 /usr/bin/addftinfo */
read_content_local(at++, &u8_disk, 1); /* not sure what this is; arg for __gxx_personality_v0? */
cie->set_prh_arg(u8_disk);
read_content_local(at, &u32_disk, 4); at+=4; /* address of <__gxx_personality_v0@plt> */
cie->set_prh_addr(ByteOrder::le_to_host(u32_disk));
break;
case 0x09: /* *.o file generated by gcc-4.0.x */
/* FIXME: Cannot find any info about this entry. Fix SgAsmElfEHFrameSection::parse() if we
* ever figure this out. [RPM 2009-09-29] */
/*fallthrough*/
case 0x10: /* See 32-bit version of skype library */
/* FIXME[Robb P. Matzke 2014-12-15]: cannot find any info for this entry. */
/*fallthrough*/
default: {
if (++nwarnings<=WARNING_LIMIT) {
fprintf(stderr, "%s:%u: warning: ELF CIE 0x%08"PRIx64" has unknown PRH encoding 0x%02x\n",
__FILE__, __LINE__, get_offset()+record_offset, cie->get_prh_encoding());
if (WARNING_LIMIT==nwarnings)
fprintf(stderr, " (additional frame warnings will be suppressed)\n");
}
break;
}
}
} else if ('R'==astr[i]) {
read_content_local(at++, &u8_disk, 1);
cie->set_addr_encoding(u8_disk);
} else if ('S'==astr[i]) {
/* See http://lkml.indiana.edu/hypermail/linux/kernel/0602.3/1144.html and GCC PR #26208*/
cie->set_sig_frame(true);
} else {
/* Some stuff we don't handle yet. Warn about it and don't read anything. */
if (++nwarnings<=WARNING_LIMIT) {
fprintf(stderr, "%s:%u: warning: ELF CIE 0x%08"PRIx64" has invalid augmentation string \"%s\"\n",
__FILE__, __LINE__, get_offset()+record_offset, escapeString(astr).c_str());
if (WARNING_LIMIT==nwarnings)
fprintf(stderr, " (additional frame warnings will be suppressed)\n");
}
}
}
}
/* Initial instructions. These are apparently included in the augmentation_data_length. The final instructions can
* be zero padding (no-op instructions) to bring the record up to a multiple of the word size. */
rose_addr_t init_insn_size = (length_field_size + record_size) - (at - record_offset);
cie->get_instructions() = read_content_local_ucl(at, init_insn_size);
ROSE_ASSERT(cie->get_instructions().size()==init_insn_size);
} else {
/* This is a FDE record */
bool fde_parse_error = false;
rose_addr_t cie_offset = record_offset + length_field_size - cie_back_offset;
if (cies.find(cie_offset) == cies.end()) {
mlog[ERROR] <<"bad CIE offset " <<cie_offset
<<" in section \"" <<StringUtility::cEscape(get_name()->get_string()) <<"\"\n";
mlog[ERROR] <<" referenced by FDE at offset " <<record_offset
<<" having CIE back offset " <<cie_back_offset <<"\n";
if (prevCieOffset) {
mlog[ERROR] <<" previous CIE was at offset " <<*prevCieOffset <<"\n";
} else {
mlog[ERROR] <<" there was no previous CIE in this section\n";
}
mlog[ERROR] <<" bailing out early\n";
break;
}
assert(cies.find(cie_offset)!=cies.end());
SgAsmElfEHFrameEntryCI *cie = cies[cie_offset];
SgAsmElfEHFrameEntryFD *fde = new SgAsmElfEHFrameEntryFD(cie);
/* PC Begin (begin_rva) and size */
switch (cie->get_addr_encoding()) {
case -1: /* No address encoding specified */
case 0x01:
case 0x03:
case 0x1b: /* Address doesn't look valid (e.g., 0xfffd74e8) but still four bytes [RPM 2008-01-16]*/
{
read_content_local(at, &u32_disk, 4); at+=4;
fde->set_begin_rva(ByteOrder::le_to_host(u32_disk));
read_content_local(at, &u32_disk, 4); at+=4;
fde->set_size(ByteOrder::le_to_host(u32_disk));
break;
}
default:
fprintf(stderr, "%s:%u: warning: ELF CIE 0x%08"PRIx64", FDE 0x%08"PRIx64": unknown address encoding: 0x%02x\n",
__FILE__, __LINE__, get_offset()+cie_offset, get_offset()+record_offset, cie->get_addr_encoding());
fde_parse_error = true;
break;
}
// Location of following fields depends on having correctly unparsed the previous stuff which is RLE.
if (!fde_parse_error) {
/* Augmentation Data */
std::string astring = cie->get_augmentation_string();
if (astring.size()>0 && astring[0]=='z') {
rose_addr_t aug_length = read_content_local_uleb128(&at);
fde->get_augmentation_data() = read_content_local_ucl(at, aug_length);
at += aug_length;
ROSE_ASSERT(fde->get_augmentation_data().size()==aug_length);
}
/* Call frame instructions */
rose_addr_t cf_insn_size = (length_field_size + record_size) - (at - record_offset);
fde->get_instructions() = read_content_local_ucl(at, cf_insn_size);
ROSE_ASSERT(fde->get_instructions().size()==cf_insn_size);
}
}
record_offset += length_field_size + record_size;
}
if (record_offset < get_size()) {
mlog[WARN] <<"ELF error handling section \"" <<StringUtility::cEscape(get_name()->get_string()) <<"\""
<<" contains " <<StringUtility::plural(get_size()-record_offset, "more bytes")
<<" that were not parsed\n";
}
return this;
}
