llvm-project/lldb/source/Symbol/DWARFCallFrameInfo.cpp

1345 lines
51 KiB
C++

//===-- DWARFCallFrameInfo.cpp ----------------------------------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
// C Includes
// C++ Includes
#include <list>
// Other libraries and framework includes
// Project includes
#include "lldb/Symbol/DWARFCallFrameInfo.h"
#include "lldb/Core/ArchSpec.h"
#include "lldb/Core/Module.h"
#include "lldb/Symbol/ObjectFile.h"
#include "lldb/Target/RegisterContext.h"
#include "lldb/Core/Section.h"
#include "lldb/Target/Thread.h"
using namespace lldb;
using namespace lldb_private;
static void
DumpRegisterName (Stream *s, Thread *thread, const ArchSpec *arch, uint32_t reg_kind, uint32_t reg_num)
{
const char *reg_name = NULL;
RegisterContext *reg_ctx = NULL;
if (thread)
{
reg_ctx = thread->GetRegisterContext();
if (reg_ctx)
reg_name = reg_ctx->GetRegisterName (reg_ctx->ConvertRegisterKindToRegisterNumber (reg_kind, reg_num));
}
if (reg_name == NULL && arch != NULL)
{
switch (reg_kind)
{
case eRegisterKindDWARF: reg_name = arch->GetRegisterName(reg_num, eRegisterKindDWARF); break;
case eRegisterKindGCC: reg_name = arch->GetRegisterName(reg_num, eRegisterKindGCC); break;
default:
break;
}
}
if (reg_name)
s->PutCString(reg_name);
else
{
const char *reg_kind_name = NULL;
switch (reg_kind)
{
case eRegisterKindDWARF: reg_kind_name = "dwarf-reg"; break;
case eRegisterKindGCC: reg_kind_name = "compiler-reg"; break;
case eRegisterKindGeneric: reg_kind_name = "generic-reg"; break;
default:
break;
}
if (reg_kind_name)
s->Printf("%s(%u)", reg_kind_name, reg_num);
else
s->Printf("reg(%d.%u)", reg_kind, reg_num);
}
}
#pragma mark DWARFCallFrameInfo::RegisterLocation
DWARFCallFrameInfo::RegisterLocation::RegisterLocation() :
m_type(isSame)
{
}
bool
DWARFCallFrameInfo::RegisterLocation::operator == (const DWARFCallFrameInfo::RegisterLocation& rhs) const
{
if (m_type != rhs.m_type)
return false;
switch (m_type)
{
case unspecified:
case isUndefined:
case isSame:
return true;
case atCFAPlusOffset:
return m_location.offset == rhs.m_location.offset;
case isCFAPlusOffset:
return m_location.offset == rhs.m_location.offset;
case inOtherRegister:
return m_location.reg_num == rhs.m_location.reg_num;
default:
break;
}
return false;
}
void
DWARFCallFrameInfo::RegisterLocation::SetUnspecified()
{
m_type = unspecified;
}
void
DWARFCallFrameInfo::RegisterLocation::SetUndefined()
{
m_type = isUndefined;
}
void
DWARFCallFrameInfo::RegisterLocation::SetSame()
{
m_type = isSame;
}
void
DWARFCallFrameInfo::RegisterLocation::SetAtCFAPlusOffset(int64_t offset)
{
m_type = atCFAPlusOffset;
m_location.offset = offset;
}
void
DWARFCallFrameInfo::RegisterLocation::SetIsCFAPlusOffset(int64_t offset)
{
m_type = isCFAPlusOffset;
m_location.offset = offset;
}
void
DWARFCallFrameInfo::RegisterLocation::SetInRegister (uint32_t reg_num)
{
m_type = inOtherRegister;
m_location.reg_num = reg_num;
}
void
DWARFCallFrameInfo::RegisterLocation::SetAtDWARFExpression(const uint8_t *opcodes, uint32_t len)
{
m_type = atDWARFExpression;
m_location.expr.opcodes = opcodes;
m_location.expr.length = len;
}
void
DWARFCallFrameInfo::RegisterLocation::SetIsDWARFExpression(const uint8_t *opcodes, uint32_t len)
{
m_type = isDWARFExpression;
m_location.expr.opcodes = opcodes;
m_location.expr.length = len;
}
void
DWARFCallFrameInfo::RegisterLocation::Dump(Stream *s, const DWARFCallFrameInfo &cfi, Thread *thread, const Row *row, uint32_t reg_num) const
{
const ArchSpec *arch = cfi.GetArchitecture();
const uint32_t reg_kind = cfi.GetRegisterKind();
DumpRegisterName (s, thread, arch, reg_kind, reg_num);
s->PutChar('=');
switch (m_type)
{
case unspecified:
s->PutChar('?');
break;
case isUndefined:
s->PutCString("undefined");
break;
case isSame:
s->PutCString("same");
break;
case atCFAPlusOffset:
s->PutChar('[');
// Fall through to isCFAPlusOffset...
case isCFAPlusOffset:
{
DumpRegisterName (s, thread, arch, reg_kind, row->GetCFARegister());
int32_t offset = row->GetCFAOffset() + m_location.offset;
if (offset != 0)
s->Printf("%-+d", offset);
if (m_type == atCFAPlusOffset)
s->PutChar(']');
}
break;
case inOtherRegister:
DumpRegisterName (s, thread, arch, reg_kind, m_location.reg_num);
break;
case atDWARFExpression:
s->PutCString("[EXPR] ");
break;
case isDWARFExpression:
s->PutCString("EXPR ");
break;
}
}
#pragma mark DWARFCallFrameInfo::Row
DWARFCallFrameInfo::Row::Row() :
m_offset(0),
m_cfa_reg_num(0),
m_cfa_offset(0),
m_register_locations()
{
}
DWARFCallFrameInfo::Row::~Row()
{
}
void
DWARFCallFrameInfo::Row::Clear()
{
m_register_locations.clear();
}
bool
DWARFCallFrameInfo::Row::GetRegisterInfo (uint32_t reg_num, DWARFCallFrameInfo::RegisterLocation& register_location) const
{
collection::const_iterator pos = m_register_locations.find(reg_num);
if (pos != m_register_locations.end())
{
register_location = pos->second;
return true;
}
return false;
}
void
DWARFCallFrameInfo::Row::SetRegisterInfo (uint32_t reg_num, const RegisterLocation& register_location)
{
m_register_locations[reg_num] = register_location;
}
void
DWARFCallFrameInfo::Row::Dump(Stream* s, const DWARFCallFrameInfo &cfi, Thread *thread, lldb::addr_t base_addr) const
{
const ArchSpec *arch = cfi.GetArchitecture();
const uint32_t reg_kind = cfi.GetRegisterKind();
collection::const_iterator pos, end = m_register_locations.end();
s->Indent();
s->Printf("0x%16.16llx: CFA=", m_offset + base_addr);
DumpRegisterName(s, thread, arch, reg_kind, m_cfa_reg_num);
if (m_cfa_offset != 0)
s->Printf("%-+lld", m_cfa_offset);
for (pos = m_register_locations.begin(); pos != end; ++pos)
{
s->PutChar(' ');
pos->second.Dump(s, cfi, thread, this, pos->first);
}
s->EOL();
}
#pragma mark DWARFCallFrameInfo::FDE
DWARFCallFrameInfo::FDE::FDE (dw_offset_t offset, const AddressRange &range) :
m_fde_offset (offset),
m_range (range),
m_row_list ()
{
}
DWARFCallFrameInfo::FDE::~FDE()
{
}
void
DWARFCallFrameInfo::FDE::AppendRow (const Row &row)
{
if (m_row_list.empty() || m_row_list.back().GetOffset() != row.GetOffset())
m_row_list.push_back(row);
else
m_row_list.back() = row;
}
void
DWARFCallFrameInfo::FDE::Dump (Stream *s, const DWARFCallFrameInfo &cfi, Thread* thread) const
{
s->Indent();
s->Printf("FDE{0x%8.8x} ", m_fde_offset);
m_range.Dump(s, NULL, Address::DumpStyleFileAddress);
lldb::addr_t fde_base_addr = m_range.GetBaseAddress().GetFileAddress();
s->EOL();
s->IndentMore();
collection::const_iterator pos, end = m_row_list.end();
for (pos = m_row_list.begin(); pos != end; ++pos)
{
pos->Dump(s, cfi, thread, fde_base_addr);
}
s->IndentLess();
}
const AddressRange &
DWARFCallFrameInfo::FDE::GetAddressRange() const
{
return m_range;
}
bool
DWARFCallFrameInfo::FDE::IsValidRowIndex (uint32_t idx) const
{
return idx < m_row_list.size();
}
const DWARFCallFrameInfo::Row&
DWARFCallFrameInfo::FDE::GetRowAtIndex (uint32_t idx)
{
// You must call IsValidRowIndex(idx) first before calling this!!!
return m_row_list[idx];
}
#pragma mark DWARFCallFrameInfo::FDEInfo
DWARFCallFrameInfo::FDEInfo::FDEInfo () :
fde_offset (0),
fde_sp()
{
}
DWARFCallFrameInfo::FDEInfo::FDEInfo (off_t offset) :
fde_offset(offset),
fde_sp()
{
}
#pragma mark DWARFCallFrameInfo::CIE
DWARFCallFrameInfo::CIE::CIE(dw_offset_t offset) :
cie_offset (offset),
version (0),
augmentation(),
code_align (0),
data_align (0),
return_addr_reg_num (0),
inst_offset (0),
inst_length (0),
ptr_encoding (DW_GNU_EH_PE_absptr)
{
}
DWARFCallFrameInfo::CIE::~CIE()
{
}
void
DWARFCallFrameInfo::CIE::Dump(Stream *s, Thread* thread, const ArchSpec *arch, uint32_t reg_kind) const
{
s->Indent();
s->Printf("CIE{0x%8.8x} version=%u, code_align=%u, data_align=%d, return_addr_reg=", cie_offset, version, code_align, data_align);
DumpRegisterName(s, thread, arch, reg_kind, return_addr_reg_num);
s->Printf(", instr_offset=0x%8.8x, instr_length=%u, ptr_encoding=0x%02x\n",
inst_offset,
inst_length,
ptr_encoding);
}
#pragma mark DWARFCallFrameInfo::CIE
DWARFCallFrameInfo::DWARFCallFrameInfo(ObjectFile *objfile, Section *section, uint32_t reg_kind) :
m_objfile (objfile),
m_section (section),
m_reg_kind (reg_kind), // The flavor of registers that the CFI data uses (One of the defines that starts with "LLDB_REGKIND_")
m_cfi_data (),
m_cie_map (),
m_fde_map ()
{
if (objfile && section)
{
section->ReadSectionDataFromObjectFile (objfile, m_cfi_data);
}
}
DWARFCallFrameInfo::~DWARFCallFrameInfo()
{
}
bool
DWARFCallFrameInfo::IsEHFrame() const
{
return (m_reg_kind == eRegisterKindGCC);
}
const ArchSpec *
DWARFCallFrameInfo::GetArchitecture() const
{
if (m_objfile && m_objfile->GetModule())
return &m_objfile->GetModule()->GetArchitecture();
return NULL;
}
uint32_t
DWARFCallFrameInfo::GetRegisterKind () const
{
return m_reg_kind;
}
void
DWARFCallFrameInfo::SetRegisterKind (uint32_t reg_kind)
{
m_reg_kind = reg_kind;
}
const DWARFCallFrameInfo::CIE*
DWARFCallFrameInfo::GetCIE(dw_offset_t cie_offset)
{
Index ();
cie_map_t::iterator pos = m_cie_map.find(cie_offset);
if (pos != m_cie_map.end())
{
// Parse and cache the CIE
if (pos->second.get() == NULL)
pos->second = ParseCIE (cie_offset);
return pos->second.get();
}
return NULL;
}
DWARFCallFrameInfo::CIE::shared_ptr
DWARFCallFrameInfo::ParseCIE (const dw_offset_t cie_offset)
{
CIE::shared_ptr cie_sp(new CIE(cie_offset));
const bool for_eh_frame = IsEHFrame();
dw_offset_t offset = cie_offset;
const uint32_t length = m_cfi_data.GetU32(&offset);
const dw_offset_t cie_id = m_cfi_data.GetU32(&offset);
const dw_offset_t end_offset = cie_offset + length + 4;
if (length > 0 && (!for_eh_frame && cie_id == 0xfffffffful) || (for_eh_frame && cie_id == 0ul))
{
size_t i;
// cie.offset = cie_offset;
// cie.length = length;
// cie.cieID = cieID;
cie_sp->ptr_encoding = DW_GNU_EH_PE_absptr;
cie_sp->version = m_cfi_data.GetU8(&offset);
for (i=0; i<CFI_AUG_MAX_SIZE; ++i)
{
cie_sp->augmentation[i] = m_cfi_data.GetU8(&offset);
if (cie_sp->augmentation[i] == '\0')
{
// Zero out remaining bytes in augmentation string
for (size_t j = i+1; j<CFI_AUG_MAX_SIZE; ++j)
cie_sp->augmentation[j] = '\0';
break;
}
}
if (i == CFI_AUG_MAX_SIZE && cie_sp->augmentation[CFI_AUG_MAX_SIZE-1] != '\0')
{
fprintf(stderr, "CIE parse error: CIE augmentation string was too large for the fixed sized buffer of %d bytes.\n", CFI_AUG_MAX_SIZE);
return cie_sp;
}
cie_sp->code_align = (uint32_t)m_cfi_data.GetULEB128(&offset);
cie_sp->data_align = (int32_t)m_cfi_data.GetSLEB128(&offset);
cie_sp->return_addr_reg_num = m_cfi_data.GetU8(&offset);
if (cie_sp->augmentation[0])
{
// Get the length of the eh_frame augmentation data
// which starts with a ULEB128 length in bytes
const size_t aug_data_len = (size_t)m_cfi_data.GetULEB128(&offset);
const size_t aug_data_end = offset + aug_data_len;
const size_t aug_str_len = strlen(cie_sp->augmentation);
// A 'z' may be present as the first character of the string.
// If present, the Augmentation Data field shall be present.
// The contents of the Augmentation Data shall be intepreted
// according to other characters in the Augmentation String.
if (cie_sp->augmentation[0] == 'z')
{
// Extract the Augmentation Data
size_t aug_str_idx = 0;
for (aug_str_idx = 1; aug_str_idx < aug_str_len; aug_str_idx++)
{
char aug = cie_sp->augmentation[aug_str_idx];
switch (aug)
{
case 'L':
// Indicates the presence of one argument in the
// Augmentation Data of the CIE, and a corresponding
// argument in the Augmentation Data of the FDE. The
// argument in the Augmentation Data of the CIE is
// 1-byte and represents the pointer encoding used
// for the argument in the Augmentation Data of the
// FDE, which is the address of a language-specific
// data area (LSDA). The size of the LSDA pointer is
// specified by the pointer encoding used.
m_cfi_data.GetU8(&offset);
break;
case 'P':
// Indicates the presence of two arguments in the
// Augmentation Data of the cie_sp-> The first argument
// is 1-byte and represents the pointer encoding
// used for the second argument, which is the
// address of a personality routine handler. The
// size of the personality routine pointer is
// specified by the pointer encoding used.
{
uint8_t arg_ptr_encoding = m_cfi_data.GetU8(&offset);
m_cfi_data.GetGNUEHPointer(&offset, arg_ptr_encoding, LLDB_INVALID_ADDRESS, LLDB_INVALID_ADDRESS, LLDB_INVALID_ADDRESS);
}
break;
case 'R':
// A 'R' may be present at any position after the
// first character of the string. The Augmentation
// Data shall include a 1 byte argument that
// represents the pointer encoding for the address
// pointers used in the FDE.
cie_sp->ptr_encoding = m_cfi_data.GetU8(&offset);
break;
}
}
}
else if (strcmp(cie_sp->augmentation, "eh") == 0)
{
// If the Augmentation string has the value "eh", then
// the EH Data field shall be present
}
// Set the offset to be the end of the augmentation data just in case
// we didn't understand any of the data.
offset = (uint32_t)aug_data_end;
}
if (end_offset > offset)
{
cie_sp->inst_offset = offset;
cie_sp->inst_length = end_offset - offset;
}
}
return cie_sp;
}
DWARFCallFrameInfo::FDE::shared_ptr
DWARFCallFrameInfo::ParseFDE(const dw_offset_t fde_offset)
{
const bool for_eh_frame = IsEHFrame();
FDE::shared_ptr fde_sp;
dw_offset_t offset = fde_offset;
const uint32_t length = m_cfi_data.GetU32(&offset);
dw_offset_t cie_offset = m_cfi_data.GetU32(&offset);
const dw_offset_t end_offset = fde_offset + length + 4;
// Translate the CIE_id from the eh_frame format, which
// is relative to the FDE offset, into a __eh_frame section
// offset
if (for_eh_frame)
cie_offset = offset - (cie_offset + 4);
const CIE* cie = GetCIE(cie_offset);
if (cie)
{
const lldb::addr_t pc_rel_addr = m_section->GetFileAddress();
const lldb::addr_t text_addr = LLDB_INVALID_ADDRESS;
const lldb::addr_t data_addr = LLDB_INVALID_ADDRESS;
lldb::addr_t range_base = m_cfi_data.GetGNUEHPointer(&offset, cie->ptr_encoding, pc_rel_addr, text_addr, data_addr);
lldb::addr_t range_len = m_cfi_data.GetGNUEHPointer(&offset, cie->ptr_encoding & DW_GNU_EH_PE_MASK_ENCODING, pc_rel_addr, text_addr, data_addr);
if (cie->augmentation[0] == 'z')
{
uint32_t aug_data_len = (uint32_t)m_cfi_data.GetULEB128(&offset);
offset += aug_data_len;
}
AddressRange fde_range (range_base, range_len, m_objfile->GetSectionList ());
fde_sp.reset(new FDE(fde_offset, fde_range));
if (offset < end_offset)
{
dw_offset_t fde_instr_offset = offset;
uint32_t fde_instr_length = end_offset - offset;
if (cie->inst_length > 0)
ParseInstructions(cie, fde_sp.get(), cie->inst_offset, cie->inst_length);
ParseInstructions(cie, fde_sp.get(), fde_instr_offset, fde_instr_length);
}
}
return fde_sp;
}
const DWARFCallFrameInfo::FDE *
DWARFCallFrameInfo::FindFDE(const Address &addr)
{
Index ();
VMRange find_range(addr.GetFileAddress(), 0);
fde_map_t::iterator pos = m_fde_map.lower_bound (find_range);
fde_map_t::iterator end = m_fde_map.end();
if (pos != end)
{
if (pos->first.Contains(find_range.GetBaseAddress()))
{
// Parse and cache the FDE if we already haven't
if (pos->second.fde_sp.get() == NULL)
pos->second.fde_sp = ParseFDE(pos->second.fde_offset);
return pos->second.fde_sp.get();
}
}
return NULL;
}
void
DWARFCallFrameInfo::Index ()
{
if (m_flags.IsClear(eFlagParsedIndex))
{
m_flags.Set (eFlagParsedIndex);
const bool for_eh_frame = IsEHFrame();
CIE::shared_ptr empty_cie_sp;
dw_offset_t offset = 0;
// Parse all of the CIEs first since we will need them to be able to
// properly parse the FDE addresses due to them possibly having
// GNU pointer encodings in their augmentations...
while (m_cfi_data.ValidOffsetForDataOfSize(offset, 8))
{
const dw_offset_t curr_offset = offset;
const uint32_t length = m_cfi_data.GetU32(&offset);
const dw_offset_t next_offset = offset + length;
const dw_offset_t cie_id = m_cfi_data.GetU32(&offset);
bool is_cie = for_eh_frame ? cie_id == 0 : cie_id == UINT32_MAX;
if (is_cie)
m_cie_map[curr_offset]= ParseCIE(curr_offset);
offset = next_offset;
}
// Now go back through and index all FDEs
offset = 0;
const lldb::addr_t pc_rel_addr = m_section->GetFileAddress();
const lldb::addr_t text_addr = LLDB_INVALID_ADDRESS;
const lldb::addr_t data_addr = LLDB_INVALID_ADDRESS;
while (m_cfi_data.ValidOffsetForDataOfSize(offset, 8))
{
const dw_offset_t curr_offset = offset;
const uint32_t length = m_cfi_data.GetU32(&offset);
const dw_offset_t next_offset = offset + length;
const dw_offset_t cie_id = m_cfi_data.GetU32(&offset);
bool is_fde = for_eh_frame ? cie_id != 0 : cie_id != UINT32_MAX;
if (is_fde)
{
dw_offset_t cie_offset;
if (for_eh_frame)
cie_offset = offset - (cie_id + 4);
else
cie_offset = cie_id;
const CIE* cie = GetCIE(cie_offset);
assert(cie);
lldb::addr_t addr = m_cfi_data.GetGNUEHPointer(&offset, cie->ptr_encoding, pc_rel_addr, text_addr, data_addr);
lldb::addr_t length = m_cfi_data.GetGNUEHPointer(&offset, cie->ptr_encoding & DW_GNU_EH_PE_MASK_ENCODING, pc_rel_addr, text_addr, data_addr);
m_fde_map[VMRange(addr, addr + length)] = FDEInfo(curr_offset);
}
offset = next_offset;
}
}
}
//----------------------------------------------------------------------
// Parse instructions for a FDE. The initial instruction for the CIE
// are parsed first, then the instructions for the FDE are parsed
//----------------------------------------------------------------------
void
DWARFCallFrameInfo::ParseInstructions(const CIE *cie, FDE *fde, dw_offset_t instr_offset, uint32_t instr_length)
{
if (cie != NULL && fde == NULL)
return;
uint32_t reg_num = 0;
int32_t op_offset = 0;
uint32_t tmp_uval32;
uint32_t code_align = cie->code_align;
int32_t data_align = cie->data_align;
typedef std::list<Row> RowStack;
RowStack row_stack;
Row row;
if (fde->IsValidRowIndex(0))
row = fde->GetRowAtIndex(0);
dw_offset_t offset = instr_offset;
const dw_offset_t end_offset = instr_offset + instr_length;
RegisterLocation reg_location;
while (m_cfi_data.ValidOffset(offset) && offset < end_offset)
{
uint8_t inst = m_cfi_data.GetU8(&offset);
uint8_t primary_opcode = inst & 0xC0;
uint8_t extended_opcode = inst & 0x3F;
if (primary_opcode)
{
switch (primary_opcode)
{
case DW_CFA_advance_loc : // (Row Creation Instruction)
{ // 0x40 - high 2 bits are 0x1, lower 6 bits are delta
// takes a single argument that represents a constant delta. The
// required action is to create a new table row with a location
// value that is computed by taking the current entry's location
// value and adding (delta * code_align). All other
// values in the new row are initially identical to the current row.
fde->AppendRow(row);
row.SlideOffset(extended_opcode * code_align);
}
break;
case DW_CFA_offset :
{ // 0x80 - high 2 bits are 0x2, lower 6 bits are register
// takes two arguments: an unsigned LEB128 constant representing a
// factored offset and a register number. The required action is to
// change the rule for the register indicated by the register number
// to be an offset(N) rule with a value of
// (N = factored offset * data_align).
reg_num = extended_opcode;
op_offset = (int32_t)m_cfi_data.GetULEB128(&offset) * data_align;
reg_location.SetAtCFAPlusOffset(op_offset);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_restore :
{ // 0xC0 - high 2 bits are 0x3, lower 6 bits are register
// takes a single argument that represents a register number. The
// required action is to change the rule for the indicated register
// to the rule assigned it by the initial_instructions in the CIE.
reg_num = extended_opcode;
// We only keep enough register locations around to
// unwind what is in our thread, and these are organized
// by the register index in that state, so we need to convert our
// GCC register number from the EH frame info, to a registe index
if (fde->IsValidRowIndex(0) && fde->GetRowAtIndex(0).GetRegisterInfo(reg_num, reg_location))
row.SetRegisterInfo (reg_num, reg_location);
}
break;
}
}
else
{
switch (extended_opcode)
{
case DW_CFA_nop : // 0x0
break;
case DW_CFA_set_loc : // 0x1 (Row Creation Instruction)
{
// DW_CFA_set_loc takes a single argument that represents an address.
// The required action is to create a new table row using the
// specified address as the location. All other values in the new row
// are initially identical to the current row. The new location value
// should always be greater than the current one.
fde->AppendRow(row);
row.SetOffset(m_cfi_data.GetPointer(&offset) - fde->GetAddressRange().GetBaseAddress().GetFileAddress());
}
break;
case DW_CFA_advance_loc1 : // 0x2 (Row Creation Instruction)
{
// takes a single uword argument that represents a constant delta.
// This instruction is identical to DW_CFA_advance_loc except for the
// encoding and size of the delta argument.
fde->AppendRow(row);
row.SlideOffset (m_cfi_data.GetU8(&offset) * code_align);
}
break;
case DW_CFA_advance_loc2 : // 0x3 (Row Creation Instruction)
{
// takes a single uword argument that represents a constant delta.
// This instruction is identical to DW_CFA_advance_loc except for the
// encoding and size of the delta argument.
fde->AppendRow(row);
row.SlideOffset (m_cfi_data.GetU16(&offset) * code_align);
}
break;
case DW_CFA_advance_loc4 : // 0x4 (Row Creation Instruction)
{
// takes a single uword argument that represents a constant delta.
// This instruction is identical to DW_CFA_advance_loc except for the
// encoding and size of the delta argument.
fde->AppendRow(row);
row.SlideOffset (m_cfi_data.GetU32(&offset) * code_align);
}
break;
case DW_CFA_offset_extended : // 0x5
{
// takes two unsigned LEB128 arguments representing a register number
// and a factored offset. This instruction is identical to DW_CFA_offset
// except for the encoding and size of the register argument.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
op_offset = (int32_t)m_cfi_data.GetULEB128(&offset) * data_align;
reg_location.SetAtCFAPlusOffset(op_offset);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_restore_extended : // 0x6
{
// takes a single unsigned LEB128 argument that represents a register
// number. This instruction is identical to DW_CFA_restore except for
// the encoding and size of the register argument.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
if (fde->IsValidRowIndex(0) && fde->GetRowAtIndex(0).GetRegisterInfo(reg_num, reg_location))
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_undefined : // 0x7
{
// takes a single unsigned LEB128 argument that represents a register
// number. The required action is to set the rule for the specified
// register to undefined.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
reg_location.SetUndefined();
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_same_value : // 0x8
{
// takes a single unsigned LEB128 argument that represents a register
// number. The required action is to set the rule for the specified
// register to same value.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
reg_location.SetSame();
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_register : // 0x9
{
// takes two unsigned LEB128 arguments representing register numbers.
// The required action is to set the rule for the first register to be
// the second register.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
uint32_t other_reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
reg_location.SetInRegister(other_reg_num);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_remember_state : // 0xA
// These instructions define a stack of information. Encountering the
// DW_CFA_remember_state instruction means to save the rules for every
// register on the current row on the stack. Encountering the
// DW_CFA_restore_state instruction means to pop the set of rules off
// the stack and place them in the current row. (This operation is
// useful for compilers that move epilogue code into the body of a
// function.)
row_stack.push_back(row);
break;
case DW_CFA_restore_state : // 0xB
// These instructions define a stack of information. Encountering the
// DW_CFA_remember_state instruction means to save the rules for every
// register on the current row on the stack. Encountering the
// DW_CFA_restore_state instruction means to pop the set of rules off
// the stack and place them in the current row. (This operation is
// useful for compilers that move epilogue code into the body of a
// function.)
{
row = row_stack.back();
row_stack.pop_back();
}
break;
case DW_CFA_def_cfa : // 0xC (CFA Definition Instruction)
{
// Takes two unsigned LEB128 operands representing a register
// number and a (non-factored) offset. The required action
// is to define the current CFA rule to use the provided
// register and offset.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
op_offset = (int32_t)m_cfi_data.GetULEB128(&offset);
row.SetCFARegister (reg_num);
row.SetCFAOffset (op_offset);
}
break;
case DW_CFA_def_cfa_register : // 0xD (CFA Definition Instruction)
{
// takes a single unsigned LEB128 argument representing a register
// number. The required action is to define the current CFA rule to
// use the provided register (but to keep the old offset).
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
row.SetCFARegister (reg_num);
}
break;
case DW_CFA_def_cfa_offset : // 0xE (CFA Definition Instruction)
{
// Takes a single unsigned LEB128 operand representing a
// (non-factored) offset. The required action is to define
// the current CFA rule to use the provided offset (but
// to keep the old register).
op_offset = (int32_t)m_cfi_data.GetULEB128(&offset);
row.SetCFAOffset (op_offset);
}
break;
case DW_CFA_def_cfa_expression : // 0xF (CFA Definition Instruction)
{
size_t block_len = (size_t)m_cfi_data.GetULEB128(&offset);
offset += (uint32_t)block_len;
}
break;
case DW_CFA_expression : // 0x10
{
// Takes two operands: an unsigned LEB128 value representing
// a register number, and a DW_FORM_block value representing a DWARF
// expression. The required action is to change the rule for the
// register indicated by the register number to be an expression(E)
// rule where E is the DWARF expression. That is, the DWARF
// expression computes the address. The value of the CFA is
// pushed on the DWARF evaluation stack prior to execution of
// the DWARF expression.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
uint32_t block_len = (uint32_t)m_cfi_data.GetULEB128(&offset);
const uint8_t *block_data = (uint8_t *)m_cfi_data.GetData(&offset, block_len);
reg_location.SetAtDWARFExpression(block_data, block_len);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_offset_extended_sf : // 0x11
{
// takes two operands: an unsigned LEB128 value representing a
// register number and a signed LEB128 factored offset. This
// instruction is identical to DW_CFA_offset_extended except
//that the second operand is signed and factored.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
op_offset = (int32_t)m_cfi_data.GetSLEB128(&offset) * data_align;
reg_location.SetAtCFAPlusOffset(op_offset);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_def_cfa_sf : // 0x12 (CFA Definition Instruction)
{
// Takes two operands: an unsigned LEB128 value representing
// a register number and a signed LEB128 factored offset.
// This instruction is identical to DW_CFA_def_cfa except
// that the second operand is signed and factored.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
op_offset = (int32_t)m_cfi_data.GetSLEB128(&offset) * data_align;
row.SetCFARegister (reg_num);
row.SetCFAOffset (op_offset);
}
break;
case DW_CFA_def_cfa_offset_sf : // 0x13 (CFA Definition Instruction)
{
// takes a signed LEB128 operand representing a factored
// offset. This instruction is identical to DW_CFA_def_cfa_offset
// except that the operand is signed and factored.
op_offset = (int32_t)m_cfi_data.GetSLEB128(&offset) * data_align;
row.SetCFAOffset (op_offset);
}
break;
case DW_CFA_val_expression : // 0x16
{
// takes two operands: an unsigned LEB128 value representing a register
// number, and a DW_FORM_block value representing a DWARF expression.
// The required action is to change the rule for the register indicated
// by the register number to be a val_expression(E) rule where E is the
// DWARF expression. That is, the DWARF expression computes the value of
// the given register. The value of the CFA is pushed on the DWARF
// evaluation stack prior to execution of the DWARF expression.
reg_num = (uint32_t)m_cfi_data.GetULEB128(&offset);
uint32_t block_len = (uint32_t)m_cfi_data.GetULEB128(&offset);
const uint8_t* block_data = (uint8_t*)m_cfi_data.GetData(&offset, block_len);
//#if defined(__i386__) || defined(__x86_64__)
// // The EH frame info for EIP and RIP contains code that looks for traps to
// // be a specific type and increments the PC.
// // For i386:
// // DW_CFA_val_expression where:
// // eip = DW_OP_breg6(+28), DW_OP_deref, DW_OP_dup, DW_OP_plus_uconst(0x34),
// // DW_OP_deref, DW_OP_swap, DW_OP_plus_uconst(0), DW_OP_deref,
// // DW_OP_dup, DW_OP_lit3, DW_OP_ne, DW_OP_swap, DW_OP_lit4, DW_OP_ne,
// // DW_OP_and, DW_OP_plus
// // This basically does a:
// // eip = ucontenxt.mcontext32->gpr.eip;
// // if (ucontenxt.mcontext32->exc.trapno != 3 && ucontenxt.mcontext32->exc.trapno != 4)
// // eip++;
// //
// // For x86_64:
// // DW_CFA_val_expression where:
// // rip = DW_OP_breg3(+48), DW_OP_deref, DW_OP_dup, DW_OP_plus_uconst(0x90), DW_OP_deref,
// // DW_OP_swap, DW_OP_plus_uconst(0), DW_OP_deref_size(4), DW_OP_dup, DW_OP_lit3,
// // DW_OP_ne, DW_OP_swap, DW_OP_lit4, DW_OP_ne, DW_OP_and, DW_OP_plus
// // This basically does a:
// // rip = ucontenxt.mcontext64->gpr.rip;
// // if (ucontenxt.mcontext64->exc.trapno != 3 && ucontenxt.mcontext64->exc.trapno != 4)
// // rip++;
// // The trap comparisons and increments are not needed as it hoses up the unwound PC which
// // is expected to point at least past the instruction that causes the fault/trap. So we
// // take it out by trimming the expression right at the first "DW_OP_swap" opcodes
// if (block_data != NULL && thread->GetPCRegNum(Thread::GCC) == reg_num)
// {
// if (thread->Is64Bit())
// {
// if (block_len > 9 && block_data[8] == DW_OP_swap && block_data[9] == DW_OP_plus_uconst)
// block_len = 8;
// }
// else
// {
// if (block_len > 8 && block_data[7] == DW_OP_swap && block_data[8] == DW_OP_plus_uconst)
// block_len = 7;
// }
// }
//#endif
reg_location.SetIsDWARFExpression(block_data, block_len);
row.SetRegisterInfo (reg_num, reg_location);
}
break;
case DW_CFA_val_offset : // 0x14
case DW_CFA_val_offset_sf : // 0x15
default:
tmp_uval32 = extended_opcode;
break;
}
}
}
fde->AppendRow(row);
}
void
DWARFCallFrameInfo::ParseAll()
{
Index();
fde_map_t::iterator pos, end = m_fde_map.end();
for (pos = m_fde_map.begin(); pos != end; ++ pos)
{
if (pos->second.fde_sp.get() == NULL)
pos->second.fde_sp = ParseFDE(pos->second.fde_offset);
}
}
//bool
//DWARFCallFrameInfo::UnwindRegisterAtIndex
//(
// const uint32_t reg_idx,
// const Thread* currState,
// const DWARFCallFrameInfo::Row* row,
// mapped_memory_t * memCache,
// Thread* unwindState
//)
//{
// bool get_reg_success = false;
//
// const RegLocation* regLocation = row->regs.GetRegisterInfo(reg_idx);
//
// // On some systems, we may not get unwind info for the program counter,
// // but the return address register can be used to get that information.
// if (reg_idx == currState->GetPCRegNum(Thread::Index))
// {
// const RegLocation* returnAddrRegLocation = row->regs.GetRegisterInfo(currState->GetRARegNum(Thread::Index));
// if (regLocation == NULL)
// {
// // We have nothing to the program counter, so lets see if this
// // thread state has a return address (link register) that can
// // help us track down the previous PC
// regLocation = returnAddrRegLocation;
// }
// else if (regLocation->type == RegLocation::unspecified)
// {
// // We did have a location that didn't specify a value for unwinding
// // the PC, so if there is a info for the return return address
// // register (link register) lets use that
// if (returnAddrRegLocation)
// regLocation = returnAddrRegLocation;
// }
// }
//
// if (regLocation)
// {
// mach_vm_address_t unwoundRegValue = INVALID_VMADDR;
// switch (regLocation->type)
// {
// case RegLocation::undefined:
// // Register is not available, mark it as invalid
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// return true;
//
// case RegLocation::unspecified:
// // Nothing to do if it is the same
// return true;
//
// case RegLocation::same:
// // Nothing to do if it is the same
// return true;
//
// case RegLocation::atFPPlusOffset:
// case RegLocation::isFPPlusOffset:
// {
// uint64_t unwindAddress = currState->GetRegisterValue(row->cfa_register, Thread::GCC, INVALID_VMADDR, &get_reg_success);
//
// if (get_reg_success)
// {
// unwindAddress += row->cfa_offset + regLocation->location.offset;
//
// if (regLocation->type == RegLocation::isFPPlusOffset)
// {
// unwindState->SetRegisterValue(reg_idx, Thread::Index, unwindAddress);
// return true;
// }
// else
// {
// kern_return_t err = mapped_memory_read_pointer(memCache, unwindAddress, &unwoundRegValue);
// if (err != KERN_SUCCESS)
// {
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// return false;
// }
// unwindState->SetRegisterValue(reg_idx, Thread::Index, unwoundRegValue);
// return true;
// }
// }
// else
// {
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// }
// return false;
// }
// break;
//
// case RegLocation::atDWARFExpression:
// case RegLocation::isDWARFExpression:
// {
// bool swap = false;
// DWARFExpressionBaton baton = { currState, memCache, swap };
// uint64_t expr_result = 0;
// CSBinaryDataRef opcodes(regLocation->location.expr.opcodes, regLocation->location.expr.length, swap);
// opcodes.SetPointerSize(currState->Is64Bit() ? 8 : 4);
// const char * expr_err = CSDWARFExpression::Evaluate(DWARFExpressionReadMemoryDCScriptInterpreter::Type,
// DWARFExpressionReadRegisterDCScriptInterpreter::Type,
// &baton,
// opcodes,
// 0,
// regLocation->location.expr.length,
// NULL,
// expr_result);
// if (expr_err == NULL)
// {
// // SUCCESS!
// if (regLocation->type == RegLocation::isDWARFExpression)
// {
// unwindState->SetRegisterValue(reg_idx, Thread::Index, expr_result);
// return true;
// }
// else
// {
// kern_return_t err = mapped_memory_read_pointer(memCache, expr_result, &unwoundRegValue);
// if (err != KERN_SUCCESS)
// {
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// return false;
// }
// unwindState->SetRegisterValue(reg_idx, Thread::Index, unwoundRegValue);
// return true;
// }
// }
// else
// {
// // FAIL
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// }
// return false;
// }
// break;
//
//
// case RegLocation::inRegister:
// // The value is in another register.
// unwoundRegValue = currState->GetRegisterValue(regLocation->location.reg, Thread::GCC, 0, &get_reg_success);
// if (get_reg_success)
// {
// unwindState->SetRegisterValue(reg_idx, Thread::Index, unwoundRegValue);
// return true;
// }
// return false;
//
// default:
// break;
// }
// }
//
// if (reg_idx == currState->GetSPRegNum(Thread::Index))
// {
// uint64_t cfa = currState->GetRegisterValue(row->cfa_register, Thread::GCC, 0, &get_reg_success);
// if (get_reg_success)
// {
// return unwindState->SetSP(cfa + row->cfa_offset);
// }
// else
// {
// unwindState->SetRegisterIsValid(reg_idx, Thread::Index, false);
// return false;
// }
// }
//
// return false;
//}
void
DWARFCallFrameInfo::Dump(Stream *s, Thread *thread) const
{
s->Indent();
s->Printf("DWARFCallFrameInfo for ");
*s << m_objfile->GetFileSpec();
if (m_flags.IsSet(eFlagParsedIndex))
{
s->Printf(" (CIE[%zu], FDE[%zu])\n", m_cie_map.size(), m_fde_map.size());
s->IndentMore();
cie_map_t::const_iterator cie_pos, cie_end = m_cie_map.end();
const ArchSpec *arch = &m_objfile->GetModule()->GetArchitecture();
for (cie_pos = m_cie_map.begin(); cie_pos != cie_end; ++ cie_pos)
{
if (cie_pos->second.get() == NULL)
{
s->Indent();
s->Printf("CIE{0x%8.8x} - unparsed\n", cie_pos->first);
}
else
{
cie_pos->second->Dump(s, thread, arch, m_reg_kind);
}
}
fde_map_t::const_iterator fde_pos, fde_end = m_fde_map.end();
for (fde_pos = m_fde_map.begin(); fde_pos != fde_end; ++ fde_pos)
{
if (fde_pos->second.fde_sp.get() == NULL)
{
s->Indent();
s->Printf("FDE{0x%8.8x} - unparsed\n", fde_pos->second.fde_offset);
}
else
{
fde_pos->second.fde_sp->Dump(s, *this, thread);
}
}
s->IndentLess();
}
else
{
s->PutCString(" (not indexed yet)\n");
}
}
//uint32_t
//DWARFCallFrameInfo::UnwindThreadState(const Thread* currState, mapped_memory_t *memCache, bool is_first_frame, Thread* unwindState)
//{
// if (currState == NULL || unwindState == NULL)
// return 0;
//
// *unwindState = *currState;
// uint32_t numRegisterUnwound = 0;
// uint64_t currPC = currState->GetPC(INVALID_VMADDR);
//
// if (currPC != INVALID_VMADDR)
// {
// // If this is not the first frame, we care about the previous instruction
// // since it will be at the instruction following the instruction that
// // made the function call.
// uint64_t unwindPC = currPC;
// if (unwindPC > 0 && !is_first_frame)
// --unwindPC;
//
//#if defined(__i386__) || defined(__x86_64__)
// // Only on i386 do we have __IMPORT segments that contain trampolines
// if (!currState->Is64Bit() && ImportRangesContainsAddress(unwindPC))
// {
// uint64_t curr_sp = currState->GetSP(INVALID_VMADDR);
// mach_vm_address_t pc = INVALID_VMADDR;
// unwindState->SetSP(curr_sp + 4);
// kern_return_t err = mapped_memory_read_pointer(memCache, curr_sp, &pc);
// if (err == KERN_SUCCESS)
// {
// unwindState->SetPC(pc);
// return 2;
// }
// }
//#endif
// FDE *fde = FindFDE(unwindPC);
// if (fde)
// {
// FindRowUserData rowUserData (currState, unwindPC);
// ParseInstructions (currState, fde, FindRowForAddress, &rowUserData);
//
// const uint32_t numRegs = currState->NumRegisters();
// for (uint32_t regNum = 0; regNum < numRegs; regNum++)
// {
// if (UnwindRegisterAtIndex(regNum, currState, &rowUserData.state, memCache, unwindState))
// numRegisterUnwound++;
// }
// }
// }
// return numRegisterUnwound;
//}