citra-emu
/
citra
Archived
1
0
Fork 0

GDB Modernization:

- Can be used in either DynCom or Dynarmic mode
- Added support for threads
- Proper support for FPU registers
- Fix for NibbleToHex conversion that used to produce false error codes
- Fix for clang-format failing under Windows
This commit is contained in:
Jarek Syrylak 2018-08-16 10:40:52 +01:00
parent 5c5aad09ce
commit bd658a8801
9 changed files with 281 additions and 62 deletions

View File

@ -321,7 +321,7 @@ if (CLANG_FORMAT)
set(CCOMMENT "Running clang format against all the .h and .cpp files in src/")
if (WIN32)
add_custom_target(clang-format
COMMAND powershell.exe -Command "${CLANG_FORMAT} -i @(Get-ChildItem -Recurse ${SRCS}/* -Include \'*.h\', \'*.cpp\')"
COMMAND powershell.exe -Command "Get-ChildItem ${SRCS}/* -Include *.cpp,*.h -Recurse | Foreach {${CLANG_FORMAT} -i $_.fullname}"
COMMENT ${CCOMMENT})
elseif(MINGW)
add_custom_target(clang-format

View File

@ -22,13 +22,6 @@
<string>GDB</string>
</property>
<layout class="QVBoxLayout" name="verticalLayout_3">
<item>
<widget class="QLabel" name="label_1">
<property name="text">
<string>The GDB Stub only works correctly when the CPU JIT is off.</string>
</property>
</widget>
</item>
<item>
<layout class="QHBoxLayout" name="horizontalLayout_1">
<item>

View File

@ -87,6 +87,8 @@ static void InterpreterFallback(u32 pc, Dynarmic::Jit* jit, void* user_arg) {
jit->SetCpsr(state->Cpsr);
jit->ExtRegs() = state->ExtReg;
jit->SetFpscr(state->VFP[VFP_FPSCR]);
state->ServeBreak();
}
static bool IsReadOnlyMemory(u32 vaddr) {
@ -233,6 +235,7 @@ void ARM_Dynarmic::ClearInstructionCache() {
for (const auto& j : jits) {
j.second->ClearCache();
}
interpreter_state->instruction_cache.clear();
}
void ARM_Dynarmic::InvalidateCacheRange(u32 start_address, size_t length) {

View File

@ -147,6 +147,7 @@ void ARM_DynCom::ExecuteInstructions(u64 num_instructions) {
state->NumInstrsToExecute = num_instructions;
unsigned ticks_executed = InterpreterMainLoop(state.get());
CoreTiming::AddTicks(ticks_executed);
state.get()->ServeBreak();
}
std::unique_ptr<ARM_Interface::ThreadContext> ARM_DynCom::NewContext() const {

View File

@ -956,7 +956,7 @@ unsigned InterpreterMainLoop(ARMul_State* cpu) {
if (GDBStub::IsServerEnabled()) { \
if (GDBStub::IsMemoryBreak() || (breakpoint_data.type != GDBStub::BreakpointType::None && \
PC == breakpoint_data.address)) { \
GDBStub::Break(); \
cpu->RecordBreak(breakpoint_data); \
goto END; \
} \
}

View File

@ -21,6 +21,8 @@
#include <unordered_map>
#include "common/common_types.h"
#include "core/arm/skyeye_common/arm_regformat.h"
#include "core/core.h"
#include "core/gdbstub/gdbstub.h"
// Signal levels
enum { LOW = 0, HIGH = 1, LOWHIGH = 1, HIGHLOW = 2 };
@ -189,6 +191,26 @@ public:
return TFlag ? 2 : 4;
}
void RecordBreak(GDBStub::BreakpointAddress bkpt) {
last_bkpt = bkpt;
last_bkpt_hit = true;
}
void ServeBreak() {
if (GDBStub::IsServerEnabled()) {
if (last_bkpt_hit) {
Reg[15] = last_bkpt.address;
}
Kernel::Thread* thread = Kernel::GetCurrentThread();
Core::CPU().SaveContext(thread->context);
if (last_bkpt_hit || GDBStub::GetCpuStepFlag()) {
last_bkpt_hit = false;
GDBStub::Break();
GDBStub::SendTrap(thread, 5);
}
}
}
std::array<u32, 16> Reg{}; // The current register file
std::array<u32, 2> Reg_usr{};
std::array<u32, 2> Reg_svc{}; // R13_SVC R14_SVC
@ -246,4 +268,7 @@ private:
u32 exclusive_tag; // The address for which the local monitor is in exclusive access mode
bool exclusive_state;
GDBStub::BreakpointAddress last_bkpt{};
bool last_bkpt_hit;
};

View File

@ -46,8 +46,7 @@ System::ResultStatus System::RunLoop(bool tight_loop) {
// execute. Otherwise, get out of the loop function.
if (GDBStub::GetCpuHaltFlag()) {
if (GDBStub::GetCpuStepFlag()) {
GDBStub::SetCpuStepFlag(false);
tight_loop = 1;
tight_loop = false;
} else {
return ResultStatus::Success;
}
@ -70,6 +69,10 @@ System::ResultStatus System::RunLoop(bool tight_loop) {
}
}
if (GDBStub::IsServerEnabled()) {
GDBStub::SetCpuStepFlag(false);
}
HW::Update();
Reschedule();

View File

@ -35,6 +35,7 @@
#include "core/arm/arm_interface.h"
#include "core/core.h"
#include "core/gdbstub/gdbstub.h"
#include "core/hle/kernel/process.h"
#include "core/loader/loader.h"
#include "core/memory.h"
@ -57,9 +58,12 @@ const u32 SIGTERM = 15;
const u32 MSG_WAITALL = 8;
#endif
const u32 R15_REGISTER = 15;
const u32 SP_REGISTER = 13;
const u32 LR_REGISTER = 14;
const u32 PC_REGISTER = 15;
const u32 CPSR_REGISTER = 25;
const u32 FPSCR_REGISTER = 58;
const u32 D0_REGISTER = 26;
const u32 FPSCR_REGISTER = 42;
// For sample XML files see the GDB source /gdb/features
// GDB also wants the l character at the start
@ -122,15 +126,17 @@ static u8 command_buffer[GDB_BUFFER_SIZE];
static u32 command_length;
static u32 latest_signal = 0;
static bool step_break = false;
static bool memory_break = false;
Kernel::Thread* current_thread = nullptr;
// Binding to a port within the reserved ports range (0-1023) requires root permissions,
// so default to a port outside of that range.
static u16 gdbstub_port = 24689;
static bool halt_loop = true;
static bool step_loop = false;
static bool send_trap = false;
// If set to false, the server will never be started and no
// gdbstub-related functions will be executed.
@ -144,12 +150,79 @@ struct Breakpoint {
bool active;
PAddr addr;
u32 len;
std::array<u8, 4> inst;
};
static std::map<u32, Breakpoint> breakpoints_execute;
static std::map<u32, Breakpoint> breakpoints_read;
static std::map<u32, Breakpoint> breakpoints_write;
static Kernel::Thread* FindThreadById(int id) {
const auto& threads = Kernel::GetThreadList();
for (auto& thread : threads) {
if (thread->GetThreadId() == static_cast<u32>(id)) {
return thread.get();
}
}
return nullptr;
}
static u32 RegRead(std::size_t id, Kernel::Thread* thread = nullptr) {
if (!thread) {
return 0;
}
if (id <= PC_REGISTER) {
return thread->context.get()->GetCpuRegister(id);
} else if (id == CPSR_REGISTER) {
return thread->context.get()->GetCpsr();
} else {
return 0;
}
}
static void RegWrite(std::size_t id, u32 val, Kernel::Thread* thread = nullptr) {
if (!thread) {
return;
}
if (id <= PC_REGISTER) {
return thread->context.get()->SetCpuRegister(id, val);
} else if (id == CPSR_REGISTER) {
return thread->context.get()->SetCpsr(val);
}
}
static u64 FpuRead(std::size_t id, Kernel::Thread* thread = nullptr) {
if (!thread) {
return 0;
}
if (id >= D0_REGISTER && id < FPSCR_REGISTER) {
u64 ret = thread->context.get()->GetFpuRegister(2 * (id - D0_REGISTER));
ret |= static_cast<u64>(thread->context.get()->GetFpuRegister(2 * (id - D0_REGISTER) + 1))
<< 32;
return ret;
} else if (id == FPSCR_REGISTER) {
return thread->context.get()->GetFpscr();
} else {
return 0;
}
}
static void FpuWrite(std::size_t id, u64 val, Kernel::Thread* thread = nullptr) {
if (!thread) {
return;
}
if (id >= D0_REGISTER && id < FPSCR_REGISTER) {
thread->context.get()->SetFpuRegister(2 * (id - D0_REGISTER), (u32)val);
thread->context.get()->SetFpuRegister(2 * (id - D0_REGISTER) + 1, val >> 32);
} else if (id == FPSCR_REGISTER) {
return thread->context.get()->SetFpscr(static_cast<u32>(val));
}
}
/**
* Turns hex string character into the equivalent byte.
*
@ -178,7 +251,7 @@ static u8 NibbleToHex(u8 n) {
if (n < 0xA) {
return '0' + n;
} else {
return 'A' + n - 0xA;
return 'a' + n - 0xA;
}
}
@ -255,6 +328,35 @@ static u32 GdbHexToInt(const u8* src) {
return output;
}
/**
* Convert a u64 into a gdb-formatted hex string.
*
* @param dest Pointer to buffer to store output hex string characters.
* @param v Value to convert.
*/
static void LongToGdbHex(u8* dest, u64 v) {
for (int i = 0; i < 16; i += 2) {
dest[i + 1] = NibbleToHex(static_cast<u8>(v >> (4 * i)));
dest[i] = NibbleToHex(static_cast<u8>(v >> (4 * (i + 1))));
}
}
/**
* Convert a gdb-formatted hex string into a u64.
*
* @param src Pointer to hex string.
*/
static u64 GdbHexToLong(const u8* src) {
u64 output = 0;
for (int i = 0; i < 16; i += 2) {
output = (output << 4) | HexCharToValue(src[15 - i - 1]);
output = (output << 4) | HexCharToValue(src[15 - i]);
}
return output;
}
/// Read a byte from the gdb client.
static u8 ReadByte() {
u8 c;
@ -303,6 +405,8 @@ static void RemoveBreakpoint(BreakpointType type, PAddr addr) {
if (bp != p.end()) {
LOG_DEBUG(Debug_GDBStub, "gdb: removed a breakpoint: {:08x} bytes at {:08x} of type {}\n",
bp->second.len, bp->second.addr, static_cast<int>(type));
Memory::WriteBlock(bp->second.addr, bp->second.inst.data(), bp->second.inst.size());
Core::CPU().ClearInstructionCache();
p.erase(addr);
}
}
@ -419,10 +523,32 @@ static void HandleQuery() {
SendReply("T0");
} else if (strncmp(query, "Supported", strlen("Supported")) == 0) {
// PacketSize needs to be large enough for target xml
SendReply("PacketSize=800;qXfer:features:read+");
SendReply("PacketSize=2000;qXfer:features:read+;qXfer:threads:read+");
} else if (strncmp(query, "Xfer:features:read:target.xml:",
strlen("Xfer:features:read:target.xml:")) == 0) {
SendReply(target_xml);
} else if (strncmp(query, "fThreadInfo", strlen("fThreadInfo")) == 0) {
std::string val = "m";
const auto& threads = Kernel::GetThreadList();
for (const auto& thread : threads) {
val += fmt::format("{:x}", thread->GetThreadId());
val += ",";
}
val.pop_back();
SendReply(val.c_str());
} else if (strncmp(query, "sThreadInfo", strlen("sThreadInfo")) == 0) {
SendReply("l");
} else if (strncmp(query, "Xfer:threads:read", strlen("Xfer:threads:read")) == 0) {
std::string buffer;
buffer += "l<?xml version=\"1.0\"?>";
buffer += "<threads>";
const auto& threads = Kernel::GetThreadList();
for (const auto& thread : threads) {
buffer += fmt::format(R"*(<thread id="{:x}" name="Thread {:x}"></thread>)*",
thread->GetThreadId(), thread->GetThreadId());
}
buffer += "</threads>";
SendReply(buffer.c_str());
} else {
SendReply("");
}
@ -430,11 +556,34 @@ static void HandleQuery() {
/// Handle set thread command from gdb client.
static void HandleSetThread() {
if (memcmp(command_buffer, "Hg0", 3) == 0 || memcmp(command_buffer, "Hc-1", 4) == 0 ||
memcmp(command_buffer, "Hc0", 4) == 0 || memcmp(command_buffer, "Hc1", 4) == 0) {
return SendReply("OK");
int thread_id = -1;
if (command_buffer[2] != '-') {
thread_id = static_cast<int>(HexToInt(command_buffer + 2, command_length - 2));
}
if (thread_id >= 1) {
current_thread = FindThreadById(thread_id);
}
if (!current_thread) {
thread_id = 1;
current_thread = FindThreadById(thread_id);
}
if (current_thread) {
SendReply("OK");
return;
}
SendReply("E01");
}
/// Handle thread alive command from gdb client.
static void HandleThreadAlive() {
int thread_id = static_cast<int>(HexToInt(command_buffer + 1, command_length - 1));
if (thread_id == 0) {
thread_id = 1;
}
if (FindThreadById(thread_id)) {
SendReply("OK");
return;
}
SendReply("E01");
}
@ -443,16 +592,31 @@ static void HandleSetThread() {
*
* @param signal Signal to be sent to client.
*/
static void SendSignal(u32 signal) {
static void SendSignal(Kernel::Thread* thread, u32 signal, bool full = true) {
if (gdbserver_socket == -1) {
return;
}
latest_signal = signal;
std::string buffer =
Common::StringFromFormat("T%02x%02x:%08x;%02x:%08x;", latest_signal, 15,
htonl(Core::CPU().GetPC()), 13, htonl(Core::CPU().GetReg(13)));
if (!thread) {
full = false;
}
std::string buffer;
if (full) {
buffer = Common::StringFromFormat("T%02x%02x:%08x;%02x:%08x;%02x:%08x", latest_signal,
PC_REGISTER, htonl(Core::CPU().GetPC()), SP_REGISTER,
htonl(Core::CPU().GetReg(SP_REGISTER)), LR_REGISTER,
htonl(Core::CPU().GetReg(LR_REGISTER)));
} else {
buffer = Common::StringFromFormat("T%02x", latest_signal);
}
if (thread) {
buffer += Common::StringFromFormat(";thread:%x;", thread->GetThreadId());
}
LOG_DEBUG(Debug_GDBStub, "Response: {}", buffer);
SendReply(buffer.c_str());
}
@ -469,7 +633,7 @@ static void ReadCommand() {
} else if (c == 0x03) {
LOG_INFO(Debug_GDBStub, "gdb: found break command\n");
halt_loop = true;
SendSignal(SIGTRAP);
SendSignal(current_thread, SIGTRAP);
return;
} else if (c != GDB_STUB_START) {
LOG_DEBUG(Debug_GDBStub, "gdb: read invalid byte {:02x}\n", c);
@ -539,17 +703,14 @@ static void ReadRegister() {
id |= HexCharToValue(command_buffer[2]);
}
if (id <= R15_REGISTER) {
IntToGdbHex(reply, Core::CPU().GetReg(id));
if (id <= PC_REGISTER) {
IntToGdbHex(reply, RegRead(id, current_thread));
} else if (id == CPSR_REGISTER) {
IntToGdbHex(reply, Core::CPU().GetCPSR());
} else if (id > CPSR_REGISTER && id < FPSCR_REGISTER) {
IntToGdbHex(reply, Core::CPU().GetVFPReg(
id - CPSR_REGISTER -
1)); // VFP registers should start at 26, so one after CSPR_REGISTER
IntToGdbHex(reply, RegRead(id, current_thread));
} else if (id >= D0_REGISTER && id < FPSCR_REGISTER) {
LongToGdbHex(reply, FpuRead(id, current_thread));
} else if (id == FPSCR_REGISTER) {
IntToGdbHex(reply, Core::CPU().GetVFPSystemReg(VFP_FPSCR)); // Get FPSCR
IntToGdbHex(reply + 8, 0);
IntToGdbHex(reply, static_cast<u32>(FpuRead(id, current_thread)));
} else {
return SendReply("E01");
}
@ -564,23 +725,23 @@ static void ReadRegisters() {
u8* bufptr = buffer;
for (u32 reg = 0; reg <= R15_REGISTER; reg++) {
IntToGdbHex(bufptr + reg * CHAR_BIT, Core::CPU().GetReg(reg));
for (u32 reg = 0; reg <= PC_REGISTER; reg++) {
IntToGdbHex(bufptr + reg * 8, RegRead(reg, current_thread));
}
bufptr += (16 * CHAR_BIT);
bufptr += 16 * 8;
IntToGdbHex(bufptr, Core::CPU().GetCPSR());
IntToGdbHex(bufptr, RegRead(CPSR_REGISTER, current_thread));
bufptr += CHAR_BIT;
bufptr += 8;
for (u32 reg = 0; reg <= 31; reg++) {
IntToGdbHex(bufptr + reg * CHAR_BIT, Core::CPU().GetVFPReg(reg));
for (u32 reg = D0_REGISTER; reg < FPSCR_REGISTER; reg++) {
LongToGdbHex(bufptr + reg * 16, FpuRead(reg, current_thread));
}
bufptr += (32 * CHAR_BIT);
bufptr += 16 * 16;
IntToGdbHex(bufptr, Core::CPU().GetVFPSystemReg(VFP_FPSCR));
IntToGdbHex(bufptr, static_cast<u32>(FpuRead(FPSCR_REGISTER, current_thread)));
SendReply(reinterpret_cast<char*>(buffer));
}
@ -596,18 +757,20 @@ static void WriteRegister() {
id |= HexCharToValue(command_buffer[2]);
}
if (id <= R15_REGISTER) {
Core::CPU().SetReg(id, GdbHexToInt(buffer_ptr));
if (id <= PC_REGISTER) {
RegWrite(id, GdbHexToInt(buffer_ptr), current_thread);
} else if (id == CPSR_REGISTER) {
Core::CPU().SetCPSR(GdbHexToInt(buffer_ptr));
} else if (id > CPSR_REGISTER && id < FPSCR_REGISTER) {
Core::CPU().SetVFPReg(id - CPSR_REGISTER - 1, GdbHexToInt(buffer_ptr));
RegWrite(id, GdbHexToInt(buffer_ptr), current_thread);
} else if (id >= D0_REGISTER && id < FPSCR_REGISTER) {
FpuWrite(id, GdbHexToLong(buffer_ptr), current_thread);
} else if (id == FPSCR_REGISTER) {
Core::CPU().SetVFPSystemReg(VFP_FPSCR, GdbHexToInt(buffer_ptr));
FpuWrite(id, GdbHexToInt(buffer_ptr), current_thread);
} else {
return SendReply("E01");
}
Core::CPU().LoadContext(current_thread->context);
SendReply("OK");
}
@ -619,23 +782,25 @@ static void WriteRegisters() {
return SendReply("E01");
for (u32 i = 0, reg = 0; reg <= FPSCR_REGISTER; i++, reg++) {
if (reg <= R15_REGISTER) {
Core::CPU().SetReg(reg, GdbHexToInt(buffer_ptr + i * CHAR_BIT));
if (reg <= PC_REGISTER) {
RegWrite(reg, GdbHexToInt(buffer_ptr + i * 8));
} else if (reg == CPSR_REGISTER) {
Core::CPU().SetCPSR(GdbHexToInt(buffer_ptr + i * CHAR_BIT));
RegWrite(reg, GdbHexToInt(buffer_ptr + i * 8));
} else if (reg == CPSR_REGISTER - 1) {
// Dummy FPA register, ignore
} else if (reg < CPSR_REGISTER) {
// Dummy FPA registers, ignore
i += 2;
} else if (reg > CPSR_REGISTER && reg < FPSCR_REGISTER) {
Core::CPU().SetVFPReg(reg - CPSR_REGISTER - 1, GdbHexToInt(buffer_ptr + i * CHAR_BIT));
} else if (reg >= D0_REGISTER && reg < FPSCR_REGISTER) {
FpuWrite(reg, GdbHexToLong(buffer_ptr + i * 16));
i++; // Skip padding
} else if (reg == FPSCR_REGISTER) {
Core::CPU().SetVFPSystemReg(VFP_FPSCR, GdbHexToInt(buffer_ptr + i * CHAR_BIT));
FpuWrite(reg, GdbHexToInt(buffer_ptr + i * 8));
}
}
Core::CPU().LoadContext(current_thread->context);
SendReply("OK");
}
@ -687,24 +852,26 @@ static void WriteMemory() {
GdbHexToMem(data.data(), len_pos + 1, len);
Memory::WriteBlock(addr, data.data(), len);
Core::CPU().ClearInstructionCache();
SendReply("OK");
}
void Break(bool is_memory_break) {
if (!halt_loop) {
halt_loop = true;
SendSignal(SIGTRAP);
}
send_trap = true;
memory_break = is_memory_break;
}
/// Tell the CPU that it should perform a single step.
static void Step() {
if (command_length > 1) {
RegWrite(PC_REGISTER, GdbHexToInt(command_buffer + 1), current_thread);
Core::CPU().LoadContext(current_thread->context);
}
step_loop = true;
halt_loop = true;
step_break = true;
SendSignal(SIGTRAP);
send_trap = true;
Core::CPU().ClearInstructionCache();
}
bool IsMemoryBreak() {
@ -718,9 +885,9 @@ bool IsMemoryBreak() {
/// Tell the CPU to continue executing.
static void Continue() {
memory_break = false;
step_break = false;
step_loop = false;
halt_loop = false;
Core::CPU().ClearInstructionCache();
}
/**
@ -737,6 +904,10 @@ static bool CommitBreakpoint(BreakpointType type, PAddr addr, u32 len) {
breakpoint.active = true;
breakpoint.addr = addr;
breakpoint.len = len;
Memory::ReadBlock(addr, breakpoint.inst.data(), breakpoint.inst.size());
static constexpr std::array<u8, 4> btrap{0x70, 0x00, 0x20, 0xe1};
Memory::WriteBlock(addr, btrap.data(), btrap.size());
Core::CPU().ClearInstructionCache();
p.insert({addr, breakpoint});
LOG_DEBUG(Debug_GDBStub, "gdb: added {} breakpoint: {:08x} bytes at {:08x}\n",
@ -857,7 +1028,7 @@ void HandlePacket() {
HandleSetThread();
break;
case '?':
SendSignal(latest_signal);
SendSignal(current_thread, latest_signal);
break;
case 'k':
Shutdown();
@ -894,6 +1065,9 @@ void HandlePacket() {
case 'Z':
AddBreakpoint();
break;
case 'T':
HandleThreadAlive();
break;
default:
SendReply("");
break;
@ -1038,4 +1212,15 @@ bool GetCpuStepFlag() {
void SetCpuStepFlag(bool is_step) {
step_loop = is_step;
}
void SendTrap(Kernel::Thread* thread, int trap) {
if (send_trap) {
if (!halt_loop || current_thread == thread) {
current_thread = thread;
SendSignal(thread, trap);
}
halt_loop = true;
send_trap = false;
}
}
}; // namespace GDBStub

View File

@ -7,6 +7,7 @@
#pragma once
#include "common/common_types.h"
#include "core/hle/kernel/thread.h"
namespace GDBStub {
@ -91,4 +92,12 @@ bool GetCpuStepFlag();
* @param is_step
*/
void SetCpuStepFlag(bool is_step);
/**
* Send trap signal from thread back to the gdbstub server.
*
* @param thread Sending thread.
* @param trap Trap no.
*/
void SendTrap(Kernel::Thread* thread, int trap);
} // namespace GDBStub