yuzu/src/core/hle/kernel/hle_ipc.cpp
mailwl a27befe456 IPC: skip empty buffer write
prevent yuzu crash, if games, like Axiom Verge, trying to read 0 bytes from file
2018-06-22 11:28:10 +03:00

364 lines
15 KiB
C++

// Copyright 2018 yuzu emulator team
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <boost/range/algorithm_ext/erase.hpp>
#include "common/assert.h"
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "core/hle/ipc_helpers.h"
#include "core/hle/kernel/event.h"
#include "core/hle/kernel/handle_table.h"
#include "core/hle/kernel/hle_ipc.h"
#include "core/hle/kernel/kernel.h"
#include "core/hle/kernel/process.h"
#include "core/hle/kernel/server_session.h"
#include "core/memory.h"
namespace Kernel {
void SessionRequestHandler::ClientConnected(SharedPtr<ServerSession> server_session) {
server_session->SetHleHandler(shared_from_this());
connected_sessions.push_back(server_session);
}
void SessionRequestHandler::ClientDisconnected(SharedPtr<ServerSession> server_session) {
server_session->SetHleHandler(nullptr);
boost::range::remove_erase(connected_sessions, server_session);
}
SharedPtr<Event> HLERequestContext::SleepClientThread(SharedPtr<Thread> thread,
const std::string& reason, u64 timeout,
WakeupCallback&& callback) {
// Put the client thread to sleep until the wait event is signaled or the timeout expires.
thread->wakeup_callback =
[context = *this, callback](ThreadWakeupReason reason, SharedPtr<Thread> thread,
SharedPtr<WaitObject> object, size_t index) mutable -> bool {
ASSERT(thread->status == THREADSTATUS_WAIT_HLE_EVENT);
callback(thread, context, reason);
context.WriteToOutgoingCommandBuffer(*thread);
return true;
};
auto event = Kernel::Event::Create(Kernel::ResetType::OneShot, "HLE Pause Event: " + reason);
thread->status = THREADSTATUS_WAIT_HLE_EVENT;
thread->wait_objects = {event};
event->AddWaitingThread(thread);
if (timeout > 0) {
thread->WakeAfterDelay(timeout);
}
return event;
}
HLERequestContext::HLERequestContext(SharedPtr<Kernel::ServerSession> server_session)
: server_session(std::move(server_session)) {
cmd_buf[0] = 0;
}
HLERequestContext::~HLERequestContext() = default;
void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
IPC::RequestParser rp(src_cmdbuf);
command_header = std::make_shared<IPC::CommandHeader>(rp.PopRaw<IPC::CommandHeader>());
if (command_header->type == IPC::CommandType::Close) {
// Close does not populate the rest of the IPC header
return;
}
// If handle descriptor is present, add size of it
if (command_header->enable_handle_descriptor) {
handle_descriptor_header =
std::make_shared<IPC::HandleDescriptorHeader>(rp.PopRaw<IPC::HandleDescriptorHeader>());
if (handle_descriptor_header->send_current_pid) {
rp.Skip(2, false);
}
if (incoming) {
// Populate the object lists with the data in the IPC request.
for (u32 handle = 0; handle < handle_descriptor_header->num_handles_to_copy; ++handle) {
copy_objects.push_back(Kernel::g_handle_table.GetGeneric(rp.Pop<Handle>()));
}
for (u32 handle = 0; handle < handle_descriptor_header->num_handles_to_move; ++handle) {
move_objects.push_back(Kernel::g_handle_table.GetGeneric(rp.Pop<Handle>()));
}
} else {
// For responses we just ignore the handles, they're empty and will be populated when
// translating the response.
rp.Skip(handle_descriptor_header->num_handles_to_copy, false);
rp.Skip(handle_descriptor_header->num_handles_to_move, false);
}
}
for (unsigned i = 0; i < command_header->num_buf_x_descriptors; ++i) {
buffer_x_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorX>());
}
for (unsigned i = 0; i < command_header->num_buf_a_descriptors; ++i) {
buffer_a_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
}
for (unsigned i = 0; i < command_header->num_buf_b_descriptors; ++i) {
buffer_b_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
}
for (unsigned i = 0; i < command_header->num_buf_w_descriptors; ++i) {
buffer_w_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorABW>());
}
buffer_c_offset = rp.GetCurrentOffset() + command_header->data_size;
// Padding to align to 16 bytes
rp.AlignWithPadding();
if (Session()->IsDomain() && ((command_header->type == IPC::CommandType::Request ||
command_header->type == IPC::CommandType::RequestWithContext) ||
!incoming)) {
// If this is an incoming message, only CommandType "Request" has a domain header
// All outgoing domain messages have the domain header, if only incoming has it
if (incoming || domain_message_header) {
domain_message_header =
std::make_shared<IPC::DomainMessageHeader>(rp.PopRaw<IPC::DomainMessageHeader>());
} else {
if (Session()->IsDomain())
NGLOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
}
}
data_payload_header =
std::make_shared<IPC::DataPayloadHeader>(rp.PopRaw<IPC::DataPayloadHeader>());
data_payload_offset = rp.GetCurrentOffset();
if (domain_message_header && domain_message_header->command ==
IPC::DomainMessageHeader::CommandType::CloseVirtualHandle) {
// CloseVirtualHandle command does not have SFC* or any data
return;
}
if (incoming) {
ASSERT(data_payload_header->magic == Common::MakeMagic('S', 'F', 'C', 'I'));
} else {
ASSERT(data_payload_header->magic == Common::MakeMagic('S', 'F', 'C', 'O'));
}
rp.SetCurrentOffset(buffer_c_offset);
// For Inline buffers, the response data is written directly to buffer_c_offset
// and in this case we don't have any BufferDescriptorC on the request.
if (command_header->buf_c_descriptor_flags >
IPC::CommandHeader::BufferDescriptorCFlag::InlineDescriptor) {
if (command_header->buf_c_descriptor_flags ==
IPC::CommandHeader::BufferDescriptorCFlag::OneDescriptor) {
buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
} else {
unsigned num_buf_c_descriptors =
static_cast<unsigned>(command_header->buf_c_descriptor_flags.Value()) - 2;
// This is used to detect possible underflows, in case something is broken
// with the two ifs above and the flags value is == 0 || == 1.
ASSERT(num_buf_c_descriptors < 14);
for (unsigned i = 0; i < num_buf_c_descriptors; ++i) {
buffer_c_desciptors.push_back(rp.PopRaw<IPC::BufferDescriptorC>());
}
}
}
rp.SetCurrentOffset(data_payload_offset);
command = rp.Pop<u32_le>();
rp.Skip(1, false); // The command is actually an u64, but we don't use the high part.
}
ResultCode HLERequestContext::PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf,
Process& src_process,
HandleTable& src_table) {
ParseCommandBuffer(src_cmdbuf, true);
if (command_header->type == IPC::CommandType::Close) {
// Close does not populate the rest of the IPC header
return RESULT_SUCCESS;
}
// The data_size already includes the payload header, the padding and the domain header.
size_t size = data_payload_offset + command_header->data_size -
sizeof(IPC::DataPayloadHeader) / sizeof(u32) - 4;
if (domain_message_header)
size -= sizeof(IPC::DomainMessageHeader) / sizeof(u32);
std::copy_n(src_cmdbuf, size, cmd_buf.begin());
return RESULT_SUCCESS;
}
ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
std::array<u32, IPC::COMMAND_BUFFER_LENGTH> dst_cmdbuf;
Memory::ReadBlock(*thread.owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
dst_cmdbuf.size() * sizeof(u32));
// The header was already built in the internal command buffer. Attempt to parse it to verify
// the integrity and then copy it over to the target command buffer.
ParseCommandBuffer(cmd_buf.data(), false);
// The data_size already includes the payload header, the padding and the domain header.
size_t size = data_payload_offset + command_header->data_size -
sizeof(IPC::DataPayloadHeader) / sizeof(u32) - 4;
if (domain_message_header)
size -= sizeof(IPC::DomainMessageHeader) / sizeof(u32);
std::copy_n(cmd_buf.begin(), size, dst_cmdbuf.data());
if (command_header->enable_handle_descriptor) {
ASSERT_MSG(!move_objects.empty() || !copy_objects.empty(),
"Handle descriptor bit set but no handles to translate");
// We write the translated handles at a specific offset in the command buffer, this space
// was already reserved when writing the header.
size_t current_offset =
(sizeof(IPC::CommandHeader) + sizeof(IPC::HandleDescriptorHeader)) / sizeof(u32);
ASSERT_MSG(!handle_descriptor_header->send_current_pid, "Sending PID is not implemented");
ASSERT_MSG(copy_objects.size() == handle_descriptor_header->num_handles_to_copy);
ASSERT_MSG(move_objects.size() == handle_descriptor_header->num_handles_to_move);
// We don't make a distinction between copy and move handles when translating since HLE
// services don't deal with handles directly. However, the guest applications might check
// for specific values in each of these descriptors.
for (auto& object : copy_objects) {
ASSERT(object != nullptr);
dst_cmdbuf[current_offset++] = Kernel::g_handle_table.Create(object).Unwrap();
}
for (auto& object : move_objects) {
ASSERT(object != nullptr);
dst_cmdbuf[current_offset++] = Kernel::g_handle_table.Create(object).Unwrap();
}
}
// TODO(Subv): Translate the X/A/B/W buffers.
if (Session()->IsDomain() && domain_message_header) {
ASSERT(domain_message_header->num_objects == domain_objects.size());
// Write the domain objects to the command buffer, these go after the raw untranslated data.
// TODO(Subv): This completely ignores C buffers.
size_t domain_offset = size - domain_message_header->num_objects;
auto& request_handlers = server_session->domain_request_handlers;
for (auto& object : domain_objects) {
request_handlers.emplace_back(object);
dst_cmdbuf[domain_offset++] = static_cast<u32_le>(request_handlers.size());
}
}
// Copy the translated command buffer back into the thread's command buffer area.
Memory::WriteBlock(*thread.owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
dst_cmdbuf.size() * sizeof(u32));
return RESULT_SUCCESS;
}
std::vector<u8> HLERequestContext::ReadBuffer(int buffer_index) const {
std::vector<u8> buffer;
const bool is_buffer_a{BufferDescriptorA().size() && BufferDescriptorA()[buffer_index].Size()};
if (is_buffer_a) {
buffer.resize(BufferDescriptorA()[buffer_index].Size());
Memory::ReadBlock(BufferDescriptorA()[buffer_index].Address(), buffer.data(),
buffer.size());
} else {
buffer.resize(BufferDescriptorX()[buffer_index].Size());
Memory::ReadBlock(BufferDescriptorX()[buffer_index].Address(), buffer.data(),
buffer.size());
}
return buffer;
}
size_t HLERequestContext::WriteBuffer(const void* buffer, size_t size, int buffer_index) const {
if (size == 0) {
NGLOG_WARNING(Core, "skip empty buffer write");
return 0;
}
const bool is_buffer_b{BufferDescriptorB().size() && BufferDescriptorB()[buffer_index].Size()};
const size_t buffer_size{GetWriteBufferSize(buffer_index)};
if (size > buffer_size) {
NGLOG_CRITICAL(Core, "size ({:016X}) is greater than buffer_size ({:016X})", size,
buffer_size);
size = buffer_size; // TODO(bunnei): This needs to be HW tested
}
if (is_buffer_b) {
Memory::WriteBlock(BufferDescriptorB()[buffer_index].Address(), buffer, size);
} else {
Memory::WriteBlock(BufferDescriptorC()[buffer_index].Address(), buffer, size);
}
return size;
}
size_t HLERequestContext::WriteBuffer(const std::vector<u8>& buffer, int buffer_index) const {
return WriteBuffer(buffer.data(), buffer.size());
}
size_t HLERequestContext::GetReadBufferSize(int buffer_index) const {
const bool is_buffer_a{BufferDescriptorA().size() && BufferDescriptorA()[buffer_index].Size()};
return is_buffer_a ? BufferDescriptorA()[buffer_index].Size()
: BufferDescriptorX()[buffer_index].Size();
}
size_t HLERequestContext::GetWriteBufferSize(int buffer_index) const {
const bool is_buffer_b{BufferDescriptorB().size() && BufferDescriptorB()[buffer_index].Size()};
return is_buffer_b ? BufferDescriptorB()[buffer_index].Size()
: BufferDescriptorC()[buffer_index].Size();
}
std::string HLERequestContext::Description() const {
if (!command_header) {
return "No command header available";
}
std::ostringstream s;
s << "IPC::CommandHeader: Type:" << static_cast<u32>(command_header->type.Value());
s << ", X(Pointer):" << command_header->num_buf_x_descriptors;
if (command_header->num_buf_x_descriptors) {
s << '[';
for (u64 i = 0; i < command_header->num_buf_x_descriptors; ++i) {
s << "0x" << std::hex << BufferDescriptorX()[i].Size();
if (i < command_header->num_buf_x_descriptors - 1)
s << ", ";
}
s << ']';
}
s << ", A(Send):" << command_header->num_buf_a_descriptors;
if (command_header->num_buf_a_descriptors) {
s << '[';
for (u64 i = 0; i < command_header->num_buf_a_descriptors; ++i) {
s << "0x" << std::hex << BufferDescriptorA()[i].Size();
if (i < command_header->num_buf_a_descriptors - 1)
s << ", ";
}
s << ']';
}
s << ", B(Receive):" << command_header->num_buf_b_descriptors;
if (command_header->num_buf_b_descriptors) {
s << '[';
for (u64 i = 0; i < command_header->num_buf_b_descriptors; ++i) {
s << "0x" << std::hex << BufferDescriptorB()[i].Size();
if (i < command_header->num_buf_b_descriptors - 1)
s << ", ";
}
s << ']';
}
s << ", C(ReceiveList):" << BufferDescriptorC().size();
if (!BufferDescriptorC().empty()) {
s << '[';
for (u64 i = 0; i < BufferDescriptorC().size(); ++i) {
s << "0x" << std::hex << BufferDescriptorC()[i].Size();
if (i < BufferDescriptorC().size() - 1)
s << ", ";
}
s << ']';
}
s << ", data_size:" << command_header->data_size.Value();
return s.str();
}
} // namespace Kernel