nvdrv: rework to remove memcpy

src/core/hle/service/nvdrv/devices/ioctl_serialization.h

@@ -11,97 +11,149 @@
 
 namespace Service::Nvidia::Devices {
 
-struct Ioctl1Traits {
-    template <typename T, typename R, typename A>
-    static T GetClassImpl(R (T::*)(A));
-
-    template <typename T, typename R, typename A>
-    static A GetArgImpl(R (T::*)(A));
+struct IoctlOneArgTraits {
+    template <typename T, typename R, typename A, typename... B>
+    static A GetFirstArgImpl(R (T::*)(A, B...));
 };
 
-struct Ioctl23Traits {
-    template <typename T, typename R, typename A, typename B>
-    static T GetClassImpl(R (T::*)(A, B));
+struct IoctlTwoArgTraits {
+    template <typename T, typename R, typename A, typename B, typename... C>
+    static A GetFirstArgImpl(R (T::*)(A, B, C...));
 
-    template <typename T, typename R, typename A, typename B>
-    static A GetArgImpl(R (T::*)(A, B));
+    template <typename T, typename R, typename A, typename B, typename... C>
+    static B GetSecondArgImpl(R (T::*)(A, B, C...));
 };
 
-template <typename T>
-struct ContainerType {
-    using ValueType = T;
-};
+struct Null {};
 
-template <Common::IsContiguousContainer T>
-struct ContainerType<T> {
-    using ValueType = T::value_type;
-};
+// clang-format off
 
-template <typename InnerArg, typename F, typename Self, typename... Rest>
-NvResult Wrap(std::span<const u8> input, std::span<u8> output, Self* self, F&& callable,
-              Rest&&... rest) {
-    using Arg = ContainerType<InnerArg>::ValueType;
-    constexpr bool ArgumentIsContainer = Common::IsContiguousContainer<InnerArg>;
+template <typename FixedArg, typename VarArg, typename InlInVarArg, typename InlOutVarArg, typename F>
+NvResult WrapGeneric(F&& callable, std::span<const u8> input, std::span<const u8> inline_input, std::span<u8> output, std::span<u8> inline_output) {
+    constexpr bool HasFixedArg     = !std::is_same_v<FixedArg, Null>;
+    constexpr bool HasVarArg       = !std::is_same_v<VarArg, Null>;
+    constexpr bool HasInlInVarArg  = !std::is_same_v<InlInVarArg, Null>;
+    constexpr bool HasInlOutVarArg = !std::is_same_v<InlOutVarArg, Null>;
 
-    // Verify that the input and output sizes are valid.
-    const size_t in_params = input.size() / sizeof(Arg);
-    const size_t out_params = output.size() / sizeof(Arg);
-    if (in_params * sizeof(Arg) != input.size()) {
-        return NvResult::InvalidSize;
-    }
-    if (out_params * sizeof(Arg) != output.size()) {
-        return NvResult::InvalidSize;
-    }
-    if (in_params == 0 && out_params == 0 && !ArgumentIsContainer) {
-        return NvResult::InvalidSize;
+    // Declare the fixed-size input value.
+    FixedArg fixed{};
+    size_t var_offset = 0;
+
+    if constexpr (HasFixedArg) {
+        // Read the fixed-size input value.
+        var_offset = std::min(sizeof(FixedArg), input.size());
+        if (var_offset > 0) {
+            std::memcpy(&fixed, input.data(), var_offset);
+        }
     }
 
-    // Copy inputs, if needed.
-    std::vector<Arg> params(std::max(in_params, out_params));
-    if (in_params > 0) {
-        std::memcpy(params.data(), input.data(), input.size());
+    // Read the variable-sized inputs.
+    const size_t num_var_args = HasVarArg ? ((input.size() - var_offset) / sizeof(VarArg)) : 0;
+    std::vector<VarArg> var_args(num_var_args);
+    if constexpr (HasVarArg) {
+        if (num_var_args > 0) {
+            std::memcpy(var_args.data(), input.data() + var_offset, num_var_args * sizeof(VarArg));
+        }
     }
 
+    const size_t num_inl_in_var_args = HasInlInVarArg ? (inline_input.size() / sizeof(InlInVarArg)) : 0;
+    std::vector<InlInVarArg> inl_in_var_args(num_inl_in_var_args);
+    if constexpr (HasInlInVarArg) {
+        if (num_inl_in_var_args > 0) {
+            std::memcpy(inl_in_var_args.data(), inline_input.data(), num_inl_in_var_args * sizeof(InlInVarArg));
+        }
+    }
+
+    // Construct inline output data.
+    const size_t num_inl_out_var_args = HasInlOutVarArg ? (inline_output.size() / sizeof(InlOutVarArg)) : 0;
+    std::vector<InlOutVarArg> inl_out_var_args(num_inl_out_var_args);
+
     // Perform the call.
-    NvResult result;
-    if constexpr (ArgumentIsContainer) {
-        result = (self->*callable)(params, std::forward<Rest>(rest)...);
-    } else {
-        result = (self->*callable)(params.front(), std::forward<Rest>(rest)...);
+    NvResult result = callable(fixed, var_args, inl_in_var_args, inl_out_var_args);
+
+    // Copy outputs.
+    if constexpr (HasFixedArg) {
+        if (output.size() > 0) {
+            std::memcpy(output.data(), &fixed, std::min(output.size(), sizeof(FixedArg)));
+        }
     }
 
-    // Copy outputs, if needed.
-    if (out_params > 0) {
-        std::memcpy(output.data(), params.data(), output.size());
+    if constexpr (HasVarArg) {
+        if (num_var_args > 0 && output.size() > var_offset) {
+            const size_t max_var_size = output.size() - var_offset;
+            std::memcpy(output.data() + var_offset, var_args.data(), std::min(max_var_size, num_var_args * sizeof(VarArg)));
+        }
     }
 
+    // Copy inline outputs.
+    if constexpr (HasInlOutVarArg) {
+        if (num_inl_out_var_args > 0) {
+            std::memcpy(inline_output.data(), inl_out_var_args.data(), num_inl_out_var_args * sizeof(InlOutVarArg));
+        }
+    }
+
+    // We're done.
     return result;
 }
 
-template <typename F>
-NvResult nvdevice::Wrap1(F&& callable, std::span<const u8> input, std::span<u8> output) {
-    using Self = decltype(Ioctl1Traits::GetClassImpl(callable));
-    using InnerArg = std::remove_reference_t<decltype(Ioctl1Traits::GetArgImpl(callable))>;
+template <typename Self, typename F, typename... Rest>
+NvResult WrapFixed(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
+    using FixedArg = typename std::remove_reference_t<decltype(IoctlOneArgTraits::GetFirstArgImpl(callable))>;
 
-    return Wrap<InnerArg>(input, output, static_cast<Self*>(this), callable);
+    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
+        return (self->*callable)(fixed, std::forward<Rest>(rest)...);
+    };
+
+    return WrapGeneric<FixedArg, Null, Null, Null>(std::move(Callable), input, {}, output, {});
 }
 
-template <typename F>
-NvResult nvdevice::Wrap2(F&& callable, std::span<const u8> input, std::span<const u8> inline_input,
-                         std::span<u8> output) {
-    using Self = decltype(Ioctl23Traits::GetClassImpl(callable));
-    using InnerArg = std::remove_reference_t<decltype(Ioctl23Traits::GetArgImpl(callable))>;
+template <typename Self, typename F, typename... Rest>
+NvResult WrapFixedInlOut(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, std::span<u8> inline_output, Rest&&... rest) {
+    using FixedArg     = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
+    using InlOutVarArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
 
-    return Wrap<InnerArg>(input, output, static_cast<Self*>(this), callable, inline_input);
+    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
+        return (self->*callable)(fixed, inl_out, std::forward<Rest>(rest)...);
+    };
+
+    return WrapGeneric<FixedArg, Null, Null, InlOutVarArg>(std::move(Callable), input, {}, output, inline_output);
 }
 
-template <typename F>
-NvResult nvdevice::Wrap3(F&& callable, std::span<const u8> input, std::span<u8> output,
-                         std::span<u8> inline_output) {
-    using Self = decltype(Ioctl23Traits::GetClassImpl(callable));
-    using InnerArg = std::remove_reference_t<decltype(Ioctl23Traits::GetArgImpl(callable))>;
+template <typename Self, typename F, typename... Rest>
+NvResult WrapVariable(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
+    using VarArg = typename std::remove_reference_t<decltype(IoctlOneArgTraits::GetFirstArgImpl(callable))>::value_type;
 
-    return Wrap<InnerArg>(input, output, static_cast<Self*>(this), callable, inline_output);
+    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
+        return (self->*callable)(var, std::forward<Rest>(rest)...);
+    };
+
+    return WrapGeneric<Null, VarArg, Null, Null>(std::move(Callable), input, {}, output, {});
 }
 
+template <typename Self, typename F, typename... Rest>
+NvResult WrapFixedVariable(Self* self, F&& callable, std::span<const u8> input, std::span<u8> output, Rest&&... rest) {
+    using FixedArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
+    using VarArg   = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
+
+    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
+        return (self->*callable)(fixed, var, std::forward<Rest>(rest)...);
+    };
+
+    return WrapGeneric<FixedArg, VarArg, Null, Null>(std::move(Callable), input, {}, output, {});
+}
+
+template <typename Self, typename F, typename... Rest>
+NvResult WrapFixedInlIn(Self* self, F&& callable, std::span<const u8> input, std::span<const u8> inline_input, std::span<u8> output, Rest&&... rest) {
+    using FixedArg    = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetFirstArgImpl(callable))>;
+    using InlInVarArg = typename std::remove_reference_t<decltype(IoctlTwoArgTraits::GetSecondArgImpl(callable))>::value_type;
+
+    const auto Callable = [&](auto& fixed, auto& var, auto& inl_in, auto& inl_out) -> NvResult {
+        return (self->*callable)(fixed, inl_in, std::forward<Rest>(rest)...);
+    };
+
+    return WrapGeneric<FixedArg, Null, InlInVarArg, Null>(std::move(Callable), input, inline_input, output, {});
+}
+
+// clang-format on
+
 } // namespace Service::Nvidia::Devices

src/core/hle/service/nvdrv/devices/nvdevice.h

@@ -74,18 +74,6 @@ public:
         return nullptr;
     }
 
-protected:
-    template <typename F>
-    NvResult Wrap1(F&& callable, std::span<const u8> input, std::span<u8> output);
-
-    template <typename F>
-    NvResult Wrap2(F&& callable, std::span<const u8> input, std::span<const u8> inline_input,
-                   std::span<u8> output);
-
-    template <typename F>
-    NvResult Wrap3(F&& callable, std::span<const u8> input, std::span<u8> output,
-                   std::span<u8> inline_output);
-
 protected:
     Core::System& system;
 };

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -34,21 +34,21 @@ NvResult nvhost_as_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> i
     case 'A':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_as_gpu::BindChannel, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::BindChannel, input, output);
         case 0x2:
-            return Wrap1(&nvhost_as_gpu::AllocateSpace, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::AllocateSpace, input, output);
         case 0x3:
-            return Wrap1(&nvhost_as_gpu::FreeSpace, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::FreeSpace, input, output);
         case 0x5:
-            return Wrap1(&nvhost_as_gpu::UnmapBuffer, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::UnmapBuffer, input, output);
         case 0x6:
-            return Wrap1(&nvhost_as_gpu::MapBufferEx, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::MapBufferEx, input, output);
         case 0x8:
-            return Wrap1(&nvhost_as_gpu::GetVARegions1, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::GetVARegions1, input, output);
         case 0x9:
-            return Wrap1(&nvhost_as_gpu::AllocAsEx, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::AllocAsEx, input, output);
         case 0x14:
-            return Wrap1(&nvhost_as_gpu::Remap, input, output);
+            return WrapVariable(this, &nvhost_as_gpu::Remap, input, output);
         default:
             break;
         }
@@ -73,7 +73,8 @@ NvResult nvhost_as_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> i
     case 'A':
         switch (command.cmd) {
         case 0x8:
-            return Wrap3(&nvhost_as_gpu::GetVARegions3, input, output, inline_output);
+            return WrapFixedInlOut(this, &nvhost_as_gpu::GetVARegions3, input, output,
+                                   inline_output);
         default:
             break;
         }
@@ -482,7 +483,7 @@ NvResult nvhost_as_gpu::GetVARegions1(IoctlGetVaRegions& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::GetVARegions3(IoctlGetVaRegions& params, std::span<u8> inline_output) {
+NvResult nvhost_as_gpu::GetVARegions3(IoctlGetVaRegions& params, std::span<VaRegion> regions) {
     LOG_DEBUG(Service_NVDRV, "called, buf_addr={:X}, buf_size={:X}", params.buf_addr,
               params.buf_size);
 
@@ -494,7 +495,10 @@ NvResult nvhost_as_gpu::GetVARegions3(IoctlGetVaRegions& params, std::span<u8> i
 
     GetVARegionsImpl(params);
 
-    std::memcpy(inline_output.data(), params.regions.data(), 2 * sizeof(VaRegion));
+    const size_t num_regions = std::min(params.regions.size(), regions.size());
+    for (size_t i = 0; i < num_regions; i++) {
+        regions[i] = params.regions[i];
+    }
 
     return NvResult::Success;
 }

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h

@@ -149,7 +149,7 @@ private:
 
     void GetVARegionsImpl(IoctlGetVaRegions& params);
     NvResult GetVARegions1(IoctlGetVaRegions& params);
-    NvResult GetVARegions3(IoctlGetVaRegions& params, std::span<u8> inline_output);
+    NvResult GetVARegions3(IoctlGetVaRegions& params, std::span<VaRegion> regions);
 
     void FreeMappingLocked(u64 offset);
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp

@@ -41,19 +41,19 @@ NvResult nvhost_ctrl::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inp
     case 0x0:
         switch (command.cmd) {
         case 0x1b:
-            return Wrap1(&nvhost_ctrl::NvOsGetConfigU32, input, output);
+            return WrapFixed(this, &nvhost_ctrl::NvOsGetConfigU32, input, output);
         case 0x1c:
-            return Wrap1(&nvhost_ctrl::IocCtrlClearEventWait, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlClearEventWait, input, output);
         case 0x1d:
-            return Wrap1(&nvhost_ctrl::IocCtrlEventWaitWithAllocation, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventWait, input, output, true);
         case 0x1e:
-            return Wrap1(&nvhost_ctrl::IocCtrlEventWaitNotAllocation, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventWait, input, output, false);
         case 0x1f:
-            return Wrap1(&nvhost_ctrl::IocCtrlEventRegister, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventRegister, input, output);
         case 0x20:
-            return Wrap1(&nvhost_ctrl::IocCtrlEventUnregister, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventUnregister, input, output);
         case 0x21:
-            return Wrap1(&nvhost_ctrl::IocCtrlEventUnregisterBatch, input, output);
+            return WrapFixed(this, &nvhost_ctrl::IocCtrlEventUnregisterBatch, input, output);
         }
         break;
     default:
@@ -86,7 +86,7 @@ NvResult nvhost_ctrl::NvOsGetConfigU32(IocGetConfigParams& params) {
     return NvResult::ConfigVarNotFound; // Returns error on production mode
 }
 
-NvResult nvhost_ctrl::IocCtrlEventWaitImpl(IocCtrlEventWaitParams& params, bool is_allocation) {
+NvResult nvhost_ctrl::IocCtrlEventWait(IocCtrlEventWaitParams& params, bool is_allocation) {
     LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_allocation={}",
               params.fence.id, params.fence.value, params.timeout, is_allocation);
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl.h

@@ -190,20 +190,11 @@ private:
     NvResult IocCtrlEventRegister(IocCtrlEventRegisterParams& params);
     NvResult IocCtrlEventUnregister(IocCtrlEventUnregisterParams& params);
     NvResult IocCtrlEventUnregisterBatch(IocCtrlEventUnregisterBatchParams& params);
+    NvResult IocCtrlEventWait(IocCtrlEventWaitParams& params, bool is_allocation);
     NvResult IocCtrlClearEventWait(IocCtrlEventClearParams& params);
 
     NvResult FreeEvent(u32 slot);
 
-    // TODO: these are not the correct names
-    NvResult IocCtrlEventWaitNotAllocation(IocCtrlEventWaitParams& params) {
-        return this->IocCtrlEventWaitImpl(params, false);
-    }
-    NvResult IocCtrlEventWaitWithAllocation(IocCtrlEventWaitParams& params) {
-        return this->IocCtrlEventWaitImpl(params, true);
-    }
-
-    NvResult IocCtrlEventWaitImpl(IocCtrlEventWaitParams& params, bool is_allocation);
-
     EventInterface& events_interface;
     NvCore::Container& core;
     NvCore::SyncpointManager& syncpoint_manager;

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp

@@ -28,23 +28,23 @@ NvResult nvhost_ctrl_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8>
     case 'G':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_ctrl_gpu::ZCullGetCtxSize, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::ZCullGetCtxSize, input, output);
         case 0x2:
-            return Wrap1(&nvhost_ctrl_gpu::ZCullGetInfo, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::ZCullGetInfo, input, output);
         case 0x3:
-            return Wrap1(&nvhost_ctrl_gpu::ZBCSetTable, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::ZBCSetTable, input, output);
         case 0x4:
-            return Wrap1(&nvhost_ctrl_gpu::ZBCQueryTable, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::ZBCQueryTable, input, output);
         case 0x5:
-            return Wrap1(&nvhost_ctrl_gpu::GetCharacteristics1, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::GetCharacteristics1, input, output);
         case 0x6:
-            return Wrap1(&nvhost_ctrl_gpu::GetTPCMasks1, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::GetTPCMasks1, input, output);
         case 0x7:
-            return Wrap1(&nvhost_ctrl_gpu::FlushL2, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::FlushL2, input, output);
         case 0x14:
-            return Wrap1(&nvhost_ctrl_gpu::GetActiveSlotMask, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::GetActiveSlotMask, input, output);
         case 0x1c:
-            return Wrap1(&nvhost_ctrl_gpu::GetGpuTime, input, output);
+            return WrapFixed(this, &nvhost_ctrl_gpu::GetGpuTime, input, output);
         default:
             break;
         }
@@ -66,9 +66,11 @@ NvResult nvhost_ctrl_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8>
     case 'G':
         switch (command.cmd) {
         case 0x5:
-            return Wrap3(&nvhost_ctrl_gpu::GetCharacteristics3, input, output, inline_output);
+            return WrapFixedInlOut(this, &nvhost_ctrl_gpu::GetCharacteristics3, input, output,
+                                   inline_output);
         case 0x6:
-            return Wrap3(&nvhost_ctrl_gpu::GetTPCMasks3, input, output, inline_output);
+            return WrapFixedInlOut(this, &nvhost_ctrl_gpu::GetTPCMasks3, input, output,
+                                   inline_output);
         default:
             break;
         }
@@ -125,8 +127,8 @@ NvResult nvhost_ctrl_gpu::GetCharacteristics1(IoctlCharacteristics& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetCharacteristics3(IoctlCharacteristics& params,
-                                              std::span<u8> inline_output) {
+NvResult nvhost_ctrl_gpu::GetCharacteristics3(
+    IoctlCharacteristics& params, std::span<IoctlGpuCharacteristics> gpu_characteristics) {
     LOG_DEBUG(Service_NVDRV, "called");
 
     params.gc.arch = 0x120;
@@ -166,8 +168,9 @@ NvResult nvhost_ctrl_gpu::GetCharacteristics3(IoctlCharacteristics& params,
     params.gc.gr_compbit_store_base_hw = 0x0;
     params.gpu_characteristics_buf_size = 0xA0;
     params.gpu_characteristics_buf_addr = 0xdeadbeef; // Cannot be 0 (UNUSED)
-    std::memcpy(inline_output.data(), &params.gc,
-                std::min(sizeof(params.gc), inline_output.size()));
+    if (!gpu_characteristics.empty()) {
+        gpu_characteristics.front() = params.gc;
+    }
     return NvResult::Success;
 }
 
@@ -179,14 +182,14 @@ NvResult nvhost_ctrl_gpu::GetTPCMasks1(IoctlGpuGetTpcMasksArgs& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_ctrl_gpu::GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params,
-                                       std::span<u8> inline_output) {
+NvResult nvhost_ctrl_gpu::GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params, std::span<u32> tpc_mask) {
     LOG_DEBUG(Service_NVDRV, "called, mask_buffer_size=0x{:X}", params.mask_buffer_size);
     if (params.mask_buffer_size != 0) {
         params.tcp_mask = 3;
     }
-    std::memcpy(inline_output.data(), &params.tcp_mask,
-                std::min(sizeof(params.tcp_mask), inline_output.size()));
+    if (!tpc_mask.empty()) {
+        tpc_mask.front() = params.tcp_mask;
+    }
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h

@@ -152,10 +152,11 @@ private:
     static_assert(sizeof(IoctlGetGpuTime) == 0x10, "IoctlGetGpuTime is incorrect size");
 
     NvResult GetCharacteristics1(IoctlCharacteristics& params);
-    NvResult GetCharacteristics3(IoctlCharacteristics& params, std::span<u8> inline_output);
+    NvResult GetCharacteristics3(IoctlCharacteristics& params,
+                                 std::span<IoctlGpuCharacteristics> gpu_characteristics);
 
     NvResult GetTPCMasks1(IoctlGpuGetTpcMasksArgs& params);
-    NvResult GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params, std::span<u8> inline_output);
+    NvResult GetTPCMasks3(IoctlGpuGetTpcMasksArgs& params, std::span<u32> tpc_mask);
 
     NvResult GetActiveSlotMask(IoctlActiveSlotMask& params);
     NvResult ZCullGetCtxSize(IoctlZcullGetCtxSize& params);

src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp

@@ -53,7 +53,7 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 0x0:
         switch (command.cmd) {
         case 0x3:
-            return Wrap1(&nvhost_gpu::GetWaitbase, input, output);
+            return WrapFixed(this, &nvhost_gpu::GetWaitbase, input, output);
         default:
             break;
         }
@@ -61,25 +61,25 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_gpu::SetNVMAPfd, input, output);
+            return WrapFixed(this, &nvhost_gpu::SetNVMAPfd, input, output);
         case 0x3:
-            return Wrap1(&nvhost_gpu::ChannelSetTimeout, input, output);
+            return WrapFixed(this, &nvhost_gpu::ChannelSetTimeout, input, output);
         case 0x8:
-            return SubmitGPFIFOBase1(input, false);
+            return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, false);
         case 0x9:
-            return Wrap1(&nvhost_gpu::AllocateObjectContext, input, output);
+            return WrapFixed(this, &nvhost_gpu::AllocateObjectContext, input, output);
         case 0xb:
-            return Wrap1(&nvhost_gpu::ZCullBind, input, output);
+            return WrapFixed(this, &nvhost_gpu::ZCullBind, input, output);
         case 0xc:
-            return Wrap1(&nvhost_gpu::SetErrorNotifier, input, output);
+            return WrapFixed(this, &nvhost_gpu::SetErrorNotifier, input, output);
         case 0xd:
-            return Wrap1(&nvhost_gpu::SetChannelPriority, input, output);
+            return WrapFixed(this, &nvhost_gpu::SetChannelPriority, input, output);
         case 0x1a:
-            return Wrap1(&nvhost_gpu::AllocGPFIFOEx2, input, output);
+            return WrapFixed(this, &nvhost_gpu::AllocGPFIFOEx2, input, output);
         case 0x1b:
-            return SubmitGPFIFOBase1(input, true);
+            return WrapFixedVariable(this, &nvhost_gpu::SubmitGPFIFOBase1, input, output, true);
         case 0x1d:
-            return Wrap1(&nvhost_gpu::ChannelSetTimeslice, input, output);
+            return WrapFixed(this, &nvhost_gpu::ChannelSetTimeslice, input, output);
         default:
             break;
         }
@@ -87,9 +87,9 @@ NvResult nvhost_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'G':
         switch (command.cmd) {
         case 0x14:
-            return Wrap1(&nvhost_gpu::SetClientData, input, output);
+            return WrapFixed(this, &nvhost_gpu::SetClientData, input, output);
         case 0x15:
-            return Wrap1(&nvhost_gpu::GetClientData, input, output);
+            return WrapFixed(this, &nvhost_gpu::GetClientData, input, output);
         default:
             break;
         }
@@ -105,7 +105,8 @@ NvResult nvhost_gpu::Ioctl2(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1b:
-            return SubmitGPFIFOBase2(input, inline_input);
+            return WrapFixedInlIn(this, &nvhost_gpu::SubmitGPFIFOBase2, input, inline_input,
+                                  output);
         }
         break;
     }
@@ -271,36 +272,35 @@ NvResult nvhost_gpu::SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandL
     return NvResult::Success;
 }
 
-NvResult nvhost_gpu::SubmitGPFIFOBase1(std::span<const u8> input, bool kickoff) {
-    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
+NvResult nvhost_gpu::SubmitGPFIFOBase1(IoctlSubmitGpfifo& params,
+                                       std::span<Tegra::CommandListHeader> commands, bool kickoff) {
+    if (params.num_entries > commands.size()) {
         UNIMPLEMENTED();
         return NvResult::InvalidSize;
     }
-    IoctlSubmitGpfifo params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
-    Tegra::CommandList entries(params.num_entries);
 
+    Tegra::CommandList entries(params.num_entries);
     if (kickoff) {
         system.ApplicationMemory().ReadBlock(params.address, entries.command_lists.data(),
                                              params.num_entries * sizeof(Tegra::CommandListHeader));
     } else {
-        std::memcpy(entries.command_lists.data(), &input[sizeof(IoctlSubmitGpfifo)],
+        std::memcpy(entries.command_lists.data(), commands.data(),
                     params.num_entries * sizeof(Tegra::CommandListHeader));
     }
 
     return SubmitGPFIFOImpl(params, std::move(entries));
 }
 
-NvResult nvhost_gpu::SubmitGPFIFOBase2(std::span<const u8> input,
-                                       std::span<const u8> input_inline) {
-    if (input.size() < sizeof(IoctlSubmitGpfifo)) {
+NvResult nvhost_gpu::SubmitGPFIFOBase2(IoctlSubmitGpfifo& params,
+                                       std::span<const Tegra::CommandListHeader> commands) {
+    if (params.num_entries > commands.size()) {
         UNIMPLEMENTED();
         return NvResult::InvalidSize;
     }
-    IoctlSubmitGpfifo params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlSubmitGpfifo));
+
     Tegra::CommandList entries(params.num_entries);
-    std::memcpy(entries.command_lists.data(), input_inline.data(), input_inline.size());
+    std::memcpy(entries.command_lists.data(), commands.data(),
+                params.num_entries * sizeof(Tegra::CommandListHeader));
     return SubmitGPFIFOImpl(params, std::move(entries));
 }
 

src/core/hle/service/nvdrv/devices/nvhost_gpu.h

@@ -196,8 +196,10 @@ private:
     NvResult AllocateObjectContext(IoctlAllocObjCtx& params);
 
     NvResult SubmitGPFIFOImpl(IoctlSubmitGpfifo& params, Tegra::CommandList&& entries);
-    NvResult SubmitGPFIFOBase1(std::span<const u8> input, bool kickoff = false);
-    NvResult SubmitGPFIFOBase2(std::span<const u8> input, std::span<const u8> input_inline);
+    NvResult SubmitGPFIFOBase1(IoctlSubmitGpfifo& params,
+                               std::span<Tegra::CommandListHeader> commands, bool kickoff = false);
+    NvResult SubmitGPFIFOBase2(IoctlSubmitGpfifo& params,
+                               std::span<const Tegra::CommandListHeader> commands);
 
     NvResult GetWaitbase(IoctlGetWaitbase& params);
     NvResult ChannelSetTimeout(IoctlChannelSetTimeout& params);

src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp

@@ -26,18 +26,18 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
             if (!host1x_file.fd_to_id.contains(fd)) {
                 host1x_file.fd_to_id[fd] = host1x_file.nvdec_next_id++;
             }
-            return Submit(fd, input, output);
+            return WrapFixedVariable(this, &nvhost_nvdec::Submit, input, output, fd);
         }
         case 0x2:
-            return Wrap1(&nvhost_nvdec::GetSyncpoint, input, output);
+            return WrapFixed(this, &nvhost_nvdec::GetSyncpoint, input, output);
         case 0x3:
-            return Wrap1(&nvhost_nvdec::GetWaitbase, input, output);
+            return WrapFixed(this, &nvhost_nvdec::GetWaitbase, input, output);
         case 0x7:
-            return Wrap1(&nvhost_nvdec::SetSubmitTimeout, input, output);
+            return WrapFixed(this, &nvhost_nvdec::SetSubmitTimeout, input, output);
         case 0x9:
-            return MapBuffer(input, output);
+            return WrapFixedVariable(this, &nvhost_nvdec::MapBuffer, input, output);
         case 0xa:
-            return UnmapBuffer(input, output);
+            return WrapFixedVariable(this, &nvhost_nvdec::UnmapBuffer, input, output);
         default:
             break;
         }
@@ -45,7 +45,7 @@ NvResult nvhost_nvdec::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_nvdec::SetNVMAPfd, input, output);
+            return WrapFixed(this, &nvhost_nvdec::SetNVMAPfd, input, output);
         default:
             break;
         }

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp

@@ -76,13 +76,7 @@ NvResult nvhost_nvdec_common::SetNVMAPfd(IoctlSetNvmapFD& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::Submit(DeviceFD fd, std::span<const u8> input, std::span<u8> output) {
-    if (input.size() < sizeof(IoctlSubmit) || output.size() < sizeof(IoctlSubmit)) {
-        UNIMPLEMENTED();
-        return NvResult::InvalidSize;
-    }
-    IoctlSubmit params{};
-    std::memcpy(&params, input.data(), std::min(input.size(), sizeof(IoctlSubmit)));
+NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV, "called NVDEC Submit, cmd_buffer_count={}", params.cmd_buffer_count);
 
     // Instantiate param buffers
@@ -93,12 +87,12 @@ NvResult nvhost_nvdec_common::Submit(DeviceFD fd, std::span<const u8> input, std
     std::vector<u32> fence_thresholds(params.fence_count);
 
     // 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, fence_thresholds, params.fence_count, offset);
+    std::size_t offset = 0;
+    offset += SliceVectors(data, command_buffers, params.cmd_buffer_count, offset);
+    offset += SliceVectors(data, relocs, params.relocation_count, offset);
+    offset += SliceVectors(data, reloc_shifts, params.relocation_count, offset);
+    offset += SliceVectors(data, syncpt_increments, params.syncpoint_count, offset);
+    offset += SliceVectors(data, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
     if (gpu.UseNvdec()) {
@@ -116,14 +110,13 @@ NvResult nvhost_nvdec_common::Submit(DeviceFD fd, std::span<const u8> input, std
                                              cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(core.Host1xDeviceFile().fd_to_id[fd], cmdlist);
     }
-    std::memcpy(output.data(), &params, 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, fence_thresholds, offset);
+    offset = 0;
+    offset += WriteVectors(data, command_buffers, offset);
+    offset += WriteVectors(data, relocs, offset);
+    offset += WriteVectors(data, reloc_shifts, offset);
+    offset += WriteVectors(data, syncpt_increments, offset);
+    offset += WriteVectors(data, fence_thresholds, offset);
 
     return NvResult::Success;
 }
@@ -140,40 +133,24 @@ NvResult nvhost_nvdec_common::GetWaitbase(IoctlGetWaitbase& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::MapBuffer(std::span<const u8> input, std::span<u8> output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), sizeof(IoctlMapBuffer));
-    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
-
-    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
-
-    for (auto& cmd_buffer : cmd_buffer_handles) {
-        cmd_buffer.map_address = nvmap.PinHandle(cmd_buffer.map_handle);
+NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries) {
+    const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
+    for (size_t i = 0; i < num_entries; i++) {
+        entries[i].map_address = nvmap.PinHandle(entries[i].map_handle);
     }
 
-    if (output.size() <
-        sizeof(IoctlMapBuffer) + cmd_buffer_handles.size() * sizeof(MapBufferEntry)) {
-        return NvResult::InvalidSize;
-    }
-
-    std::memcpy(output.data(), &params, sizeof(IoctlMapBuffer));
-    std::memcpy(output.data() + sizeof(IoctlMapBuffer), cmd_buffer_handles.data(),
-                cmd_buffer_handles.size() * sizeof(MapBufferEntry));
-
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::UnmapBuffer(std::span<const u8> input, std::span<u8> output) {
-    IoctlMapBuffer params{};
-    std::memcpy(&params, input.data(), std::min(input.size(), sizeof(IoctlMapBuffer)));
-    std::vector<MapBufferEntry> cmd_buffer_handles(params.num_entries);
-
-    SliceVectors(input, cmd_buffer_handles, params.num_entries, sizeof(IoctlMapBuffer));
-    for (auto& cmd_buffer : cmd_buffer_handles) {
-        nvmap.UnpinHandle(cmd_buffer.map_handle);
+NvResult nvhost_nvdec_common::UnmapBuffer(IoctlMapBuffer& params,
+                                          std::span<MapBufferEntry> entries) {
+    const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
+    for (size_t i = 0; i < num_entries; i++) {
+        nvmap.UnpinHandle(entries[i].map_handle);
+        entries[i] = {};
     }
 
-    std::memset(output.data(), 0, output.size());
+    params = {};
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h

@@ -108,11 +108,11 @@ protected:
 
     /// Ioctl command implementations
     NvResult SetNVMAPfd(IoctlSetNvmapFD&);
-    NvResult Submit(DeviceFD fd, std::span<const u8> input, std::span<u8> output);
+    NvResult Submit(IoctlSubmit& params, std::span<u8> input, DeviceFD fd);
     NvResult GetSyncpoint(IoctlGetSyncpoint& params);
     NvResult GetWaitbase(IoctlGetWaitbase& params);
-    NvResult MapBuffer(std::span<const u8> input, std::span<u8> output);
-    NvResult UnmapBuffer(std::span<const u8> input, std::span<u8> output);
+    NvResult MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
+    NvResult UnmapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries);
     NvResult SetSubmitTimeout(u32 timeout);
 
     Kernel::KEvent* QueryEvent(u32 event_id) override;

src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp

@@ -19,7 +19,7 @@ NvResult nvhost_nvjpg::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> in
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_nvjpg::SetNVMAPfd, input, output);
+            return WrapFixed(this, &nvhost_nvjpg::SetNVMAPfd, input, output);
         default:
             break;
         }

src/core/hle/service/nvdrv/devices/nvhost_vic.cpp

@@ -26,16 +26,16 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
             if (!host1x_file.fd_to_id.contains(fd)) {
                 host1x_file.fd_to_id[fd] = host1x_file.vic_next_id++;
             }
-            return Submit(fd, input, output);
+            return WrapFixedVariable(this, &nvhost_vic::Submit, input, output, fd);
         }
         case 0x2:
-            return Wrap1(&nvhost_vic::GetSyncpoint, input, output);
+            return WrapFixed(this, &nvhost_vic::GetSyncpoint, input, output);
         case 0x3:
-            return Wrap1(&nvhost_vic::GetWaitbase, input, output);
+            return WrapFixed(this, &nvhost_vic::GetWaitbase, input, output);
         case 0x9:
-            return MapBuffer(input, output);
+            return WrapFixedVariable(this, &nvhost_vic::MapBuffer, input, output);
         case 0xa:
-            return UnmapBuffer(input, output);
+            return WrapFixedVariable(this, &nvhost_vic::UnmapBuffer, input, output);
         default:
             break;
         }
@@ -43,7 +43,7 @@ NvResult nvhost_vic::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> inpu
     case 'H':
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvhost_vic::SetNVMAPfd, input, output);
+            return WrapFixed(this, &nvhost_vic::SetNVMAPfd, input, output);
         default:
             break;
         }

src/core/hle/service/nvdrv/devices/nvmap.cpp

@@ -32,17 +32,17 @@ NvResult nvmap::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> input,
     case 0x1:
         switch (command.cmd) {
         case 0x1:
-            return Wrap1(&nvmap::IocCreate, input, output);
+            return WrapFixed(this, &nvmap::IocCreate, input, output);
         case 0x3:
-            return Wrap1(&nvmap::IocFromId, input, output);
+            return WrapFixed(this, &nvmap::IocFromId, input, output);
         case 0x4:
-            return Wrap1(&nvmap::IocAlloc, input, output);
+            return WrapFixed(this, &nvmap::IocAlloc, input, output);
         case 0x5:
-            return Wrap1(&nvmap::IocFree, input, output);
+            return WrapFixed(this, &nvmap::IocFree, input, output);
         case 0x9:
-            return Wrap1(&nvmap::IocParam, input, output);
+            return WrapFixed(this, &nvmap::IocParam, input, output);
         case 0xe:
-            return Wrap1(&nvmap::IocGetId, input, output);
+            return WrapFixed(this, &nvmap::IocGetId, input, output);
         default:
             break;
         }