|
|
|
@ -19,26 +19,29 @@
|
|
|
|
|
namespace Service::Nvidia::Devices {
|
|
|
|
|
|
|
|
|
|
namespace {
|
|
|
|
|
// Splice vectors will copy count amount of type T from the input vector into the dst vector.
|
|
|
|
|
// Copies count amount of type T from the input vector into the dst vector.
|
|
|
|
|
// Returns the number of bytes written into dst.
|
|
|
|
|
template <typename T>
|
|
|
|
|
std::size_t SpliceVectors(const std::vector<u8>& input, std::vector<T>& dst, std::size_t count,
|
|
|
|
|
std::size_t SliceVectors(const std::vector<u8>& input, std::vector<T>& dst, std::size_t count,
|
|
|
|
|
std::size_t offset) {
|
|
|
|
|
if (!dst.empty()) {
|
|
|
|
|
std::memcpy(dst.data(), input.data() + offset, count * sizeof(T));
|
|
|
|
|
}
|
|
|
|
|
if (dst.empty()) {
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
const size_t bytes_copied = count * sizeof(T);
|
|
|
|
|
std::memcpy(dst.data(), input.data() + offset, bytes_copied);
|
|
|
|
|
return bytes_copied;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Write vectors will write data to the output buffer
|
|
|
|
|
// Writes the data in src to an offset into the dst vector. The offset is specified in bytes
|
|
|
|
|
// Returns the number of bytes written into dst.
|
|
|
|
|
template <typename T>
|
|
|
|
|
std::size_t WriteVectors(std::vector<u8>& dst, const std::vector<T>& src, std::size_t offset) {
|
|
|
|
|
if (src.empty()) {
|
|
|
|
|
return 0;
|
|
|
|
|
} else {
|
|
|
|
|
std::memcpy(dst.data() + offset, src.data(), src.size() * sizeof(T));
|
|
|
|
|
offset += src.size() * sizeof(T);
|
|
|
|
|
return offset;
|
|
|
|
|
}
|
|
|
|
|
const size_t bytes_copied = src.size() * sizeof(T);
|
|
|
|
|
std::memcpy(dst.data() + offset, src.data(), bytes_copied);
|
|
|
|
|
return bytes_copied;
|
|
|
|
|
}
|
|
|
|
|
} // Anonymous namespace
|
|
|
|
|
|
|
|
|
@ -62,7 +65,6 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
|
|
|
|
|
LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
|
|
|
|
|
|
|
|
|
|
// Instantiate param buffers
|
|
|
|
|
std::size_t offset = sizeof(IoctlSubmit);
|
|
|
|
|
std::vector<CommandBuffer> command_buffers(params.cmd_buffer_count);
|
|
|
|
|
std::vector<Reloc> relocs(params.relocation_count);
|
|
|
|
|
std::vector<u32> reloc_shifts(params.relocation_count);
|
|
|
|
@ -70,13 +72,14 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
|
|
|
|
|
std::vector<SyncptIncr> wait_checks(params.syncpoint_count);
|
|
|
|
|
std::vector<Fence> fences(params.fence_count);
|
|
|
|
|
|
|
|
|
|
// Splice input into their respective buffers
|
|
|
|
|
offset = SpliceVectors(input, command_buffers, params.cmd_buffer_count, offset);
|
|
|
|
|
offset = SpliceVectors(input, relocs, params.relocation_count, offset);
|
|
|
|
|
offset = SpliceVectors(input, reloc_shifts, params.relocation_count, offset);
|
|
|
|
|
offset = SpliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
|
|
|
|
|
offset = SpliceVectors(input, wait_checks, params.syncpoint_count, offset);
|
|
|
|
|
offset = SpliceVectors(input, fences, params.fence_count, offset);
|
|
|
|
|
// Slice input into their respective buffers
|
|
|
|
|
std::size_t offset = sizeof(IoctlSubmit);
|
|
|
|
|
offset += SliceVectors(input, command_buffers, params.cmd_buffer_count, offset);
|
|
|
|
|
offset += SliceVectors(input, relocs, params.relocation_count, offset);
|
|
|
|
|
offset += SliceVectors(input, reloc_shifts, params.relocation_count, offset);
|
|
|
|
|
offset += SliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
|
|
|
|
|
offset += SliceVectors(input, wait_checks, params.syncpoint_count, offset);
|
|
|
|
|
offset += SliceVectors(input, fences, params.fence_count, offset);
|
|
|
|
|
|
|
|
|
|
auto& gpu = system.GPU();
|
|
|
|
|
if (gpu.UseNvdec()) {
|
|
|
|
@ -88,35 +91,27 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
for (const auto& cmd_buffer : command_buffers) {
|
|
|
|
|
auto object = nvmap_dev->GetObject(cmd_buffer.memory_id);
|
|
|
|
|
const auto object = nvmap_dev->GetObject(cmd_buffer.memory_id);
|
|
|
|
|
ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
|
|
|
|
|
const auto map = FindBufferMap(object->dma_map_addr);
|
|
|
|
|
if (!map) {
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "Tried to submit an invalid offset 0x{:X} dma 0x{:X}",
|
|
|
|
|
object->addr, object->dma_map_addr);
|
|
|
|
|
return NvResult::Success;
|
|
|
|
|
}
|
|
|
|
|
Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
|
|
|
|
|
gpu.MemoryManager().ReadBlock(map->StartAddr() + cmd_buffer.offset, cmdlist.data(),
|
|
|
|
|
system.Memory().ReadBlock(object->addr + cmd_buffer.offset, cmdlist.data(),
|
|
|
|
|
cmdlist.size() * sizeof(u32));
|
|
|
|
|
gpu.PushCommandBuffer(cmdlist);
|
|
|
|
|
}
|
|
|
|
|
if (gpu.UseNvdec()) {
|
|
|
|
|
|
|
|
|
|
fences[0].value = syncpoint_manager.IncreaseSyncpoint(fences[0].id, 1);
|
|
|
|
|
|
|
|
|
|
Tegra::ChCommandHeaderList cmdlist{{(4 << 28) | fences[0].id}};
|
|
|
|
|
gpu.PushCommandBuffer(cmdlist);
|
|
|
|
|
}
|
|
|
|
|
std::memcpy(output.data(), ¶ms, sizeof(IoctlSubmit));
|
|
|
|
|
// Some games expect command_buffers to be written back
|
|
|
|
|
offset = sizeof(IoctlSubmit);
|
|
|
|
|
offset = WriteVectors(output, command_buffers, offset);
|
|
|
|
|
offset = WriteVectors(output, relocs, offset);
|
|
|
|
|
offset = WriteVectors(output, reloc_shifts, offset);
|
|
|
|
|
offset = WriteVectors(output, syncpt_increments, offset);
|
|
|
|
|
offset = WriteVectors(output, wait_checks, offset);
|
|
|
|
|
offset = WriteVectors(output, fences, offset);
|
|
|
|
|
offset += WriteVectors(output, command_buffers, offset);
|
|
|
|
|
offset += WriteVectors(output, relocs, offset);
|
|
|
|
|
offset += WriteVectors(output, reloc_shifts, offset);
|
|
|
|
|
offset += WriteVectors(output, syncpt_increments, offset);
|
|
|
|
|
offset += WriteVectors(output, wait_checks, offset);
|
|
|
|
|
offset += WriteVectors(output, fences, offset);
|
|
|
|
|
|
|
|
|
|
return NvResult::Success;
|
|
|
|
|
}
|
|
|
|
@ -148,14 +143,14 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
|
|
|
|
|
std::memcpy(¶ms, input.data(), sizeof(IoctlMapBuffer));
|
|
|
|
|
std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
|
|
|
|
|
|
|
|
|
|
SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
|
|
|
|
|
SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
|
|
|
|
|
|
|
|
|
|
auto& gpu = system.GPU();
|
|
|
|
|
|
|
|
|
|
for (auto& cmf_buff : cmd_buffer_handles) {
|
|
|
|
|
auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
|
|
|
|
|
for (auto& cmd_buffer : cmd_buffer_handles) {
|
|
|
|
|
auto object{nvmap_dev->GetObject(cmd_buffer.map_handle)};
|
|
|
|
|
if (!object) {
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmd_buffer.map_handle);
|
|
|
|
|
std::memcpy(output.data(), ¶ms, output.size());
|
|
|
|
|
return NvResult::InvalidState;
|
|
|
|
|
}
|
|
|
|
@ -170,7 +165,7 @@ NvResult nvhost_nvdec_common::MapBuffer(const std::vector<u8>& input, std::vecto
|
|
|
|
|
if (!object->dma_map_addr) {
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "failed to map size={}", object->size);
|
|
|
|
|
} else {
|
|
|
|
|
cmf_buff.map_address = object->dma_map_addr;
|
|
|
|
|
cmd_buffer.map_address = object->dma_map_addr;
|
|
|
|
|
AddBufferMap(object->dma_map_addr, object->size, object->addr,
|
|
|
|
|
object->status == nvmap::Object::Status::Allocated);
|
|
|
|
|
}
|
|
|
|
@ -186,14 +181,14 @@ NvResult nvhost_nvdec_common::UnmapBuffer(const std::vector<u8>& input, std::vec
|
|
|
|
|
IoctlMapBuffer params{};
|
|
|
|
|
std::memcpy(¶ms, input.data(), sizeof(IoctlMapBuffer));
|
|
|
|
|
std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
|
|
|
|
|
SpliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
|
|
|
|
|
SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
|
|
|
|
|
|
|
|
|
|
auto& gpu = system.GPU();
|
|
|
|
|
|
|
|
|
|
for (auto& cmf_buff : cmd_buffer_handles) {
|
|
|
|
|
const auto object{nvmap_dev->GetObject(cmf_buff.map_handle)};
|
|
|
|
|
for (auto& cmd_buffer : cmd_buffer_handles) {
|
|
|
|
|
const auto object{nvmap_dev->GetObject(cmd_buffer.map_handle)};
|
|
|
|
|
if (!object) {
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmf_buff.map_handle);
|
|
|
|
|
LOG_ERROR(Service_NVDRV, "invalid cmd_buffer nvmap_handle={:X}", cmd_buffer.map_handle);
|
|
|
|
|
std::memcpy(output.data(), ¶ms, output.size());
|
|
|
|
|
return NvResult::InvalidState;
|
|
|
|
|
}
|
|
|
|
|