NVDEC implementation using FFmpeg (#443)

* Initial nvdec implementation using FFmpeg

* Fix swapped channels on the video decoder and the G8R8 texture format

* Fix texture samplers not being set properly (regression)

* Rebased

* Remove unused code introduced on the rebase

* Add support for RGBA8 output format on the video image composer

* Correct spacing

* Some fixes for rebase and other tweaks

* Allow size mismatch on frame copy

* Get rid of GetHostAddress calls on VDec

9 files changed, 214 insertions, 44 deletions

diff --git a/Ryujinx.HLE/HOS/Services/Mm/IRequest.cs b/Ryujinx.HLE/HOS/Services/Mm/IRequest.cs
index 88cd57cf..72067d71 100644
--- a/Ryujinx.HLE/HOS/Services/Mm/IRequest.cs
+++ b/Ryujinx.HLE/HOS/Services/Mm/IRequest.cs
@@ -16,6 +16,7 @@ namespace Ryujinx.HLE.HOS.Services.Mm
             {
                 { 1, InitializeOld },
                 { 4, Initialize    },
+                { 5, Finalize      },
                 { 6, SetAndWait    },
                 { 7, Get           }
             };
@@ -40,6 +41,15 @@ namespace Ryujinx.HLE.HOS.Services.Mm
             return 0;
         }
 
+        public long Finalize(ServiceCtx Context)
+        {
+            Context.Device.Gpu.UninitializeVideoDecoder();
+
+            Logger.PrintStub(LogClass.ServiceMm, "Stubbed.");
+
+            return 0;
+        }
+
         public long SetAndWait(ServiceCtx Context)
         {
             Logger.PrintStub(LogClass.ServiceMm, "Stubbed.");
diff --git a/Ryujinx.HLE/HOS/Services/Nv/INvDrvServices.cs b/Ryujinx.HLE/HOS/Services/Nv/INvDrvServices.cs
index 6786d0e2..1b034bfa 100644
--- a/Ryujinx.HLE/HOS/Services/Nv/INvDrvServices.cs
+++ b/Ryujinx.HLE/HOS/Services/Nv/INvDrvServices.cs
@@ -23,11 +23,13 @@ namespace Ryujinx.HLE.HOS.Services.Nv
         private static Dictionary<string, IoctlProcessor> IoctlProcessors =
                    new Dictionary<string, IoctlProcessor>()
         {
-            { "/dev/nvhost-as-gpu",   ProcessIoctlNvGpuAS    },
-            { "/dev/nvhost-ctrl",     ProcessIoctlNvHostCtrl },
-            { "/dev/nvhost-ctrl-gpu", ProcessIoctlNvGpuGpu   },
-            { "/dev/nvhost-gpu",      ProcessIoctlNvHostGpu  },
-            { "/dev/nvmap",           ProcessIoctlNvMap      }
+            { "/dev/nvhost-as-gpu",   ProcessIoctlNvGpuAS       },
+            { "/dev/nvhost-ctrl",     ProcessIoctlNvHostCtrl    },
+            { "/dev/nvhost-ctrl-gpu", ProcessIoctlNvGpuGpu      },
+            { "/dev/nvhost-gpu",      ProcessIoctlNvHostChannel },
+            { "/dev/nvhost-nvdec",    ProcessIoctlNvHostChannel },
+            { "/dev/nvhost-vic",      ProcessIoctlNvHostChannel },
+            { "/dev/nvmap",           ProcessIoctlNvMap         }
         };
 
         public static GlobalStateTable Fds { get; private set; }
@@ -166,9 +168,9 @@ namespace Ryujinx.HLE.HOS.Services.Nv
             return ProcessIoctl(Context, Cmd, NvGpuGpuIoctl.ProcessIoctl);
         }
 
-        private static int ProcessIoctlNvHostGpu(ServiceCtx Context, int Cmd)
+        private static int ProcessIoctlNvHostChannel(ServiceCtx Context, int Cmd)
         {
-            return ProcessIoctl(Context, Cmd, NvHostChannelIoctl.ProcessIoctlGpu);
+            return ProcessIoctl(Context, Cmd, NvHostChannelIoctl.ProcessIoctl);
         }
 
         private static int ProcessIoctlNvMap(ServiceCtx Context, int Cmd)
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvChannelName.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvChannelName.cs
deleted file mode 100644
index a2b5ea4c..00000000
--- a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvChannelName.cs
+++ /dev/null
@@ -1,7 +0,0 @@
-namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
-{
-    enum NvChannelName
-    {
-        Gpu
-    }
-}
\ No newline at end of file
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelCmdBuf.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelCmdBuf.cs
new file mode 100644
index 00000000..44949597
--- /dev/null
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelCmdBuf.cs
@@ -0,0 +1,12 @@
+using System.Runtime.InteropServices;
+
+namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
+{
+    [StructLayout(LayoutKind.Sequential, Size = 8, Pack = 4)]
+    struct NvHostChannelCmdBuf
+    {
+        public int MemoryId;
+        public int Offset;
+        public int WordsCount;
+    }
+}
\ No newline at end of file
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelGetParamArg.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelGetParamArg.cs
new file mode 100644
index 00000000..5c04e5d4
--- /dev/null
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelGetParamArg.cs
@@ -0,0 +1,11 @@
+using System.Runtime.InteropServices;
+
+namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
+{
+    [StructLayout(LayoutKind.Sequential, Size = 8, Pack = 4)]
+    struct NvHostChannelGetParamArg
+    {
+        public int Param;
+        public int Value;
+    }
+}
\ No newline at end of file
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelIoctl.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelIoctl.cs
index d9f5602b..466f3e9b 100644
--- a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelIoctl.cs
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelIoctl.cs
@@ -3,6 +3,7 @@ using Ryujinx.Common.Logging;
 using Ryujinx.Graphics.Memory;
 using Ryujinx.HLE.HOS.Kernel;
 using Ryujinx.HLE.HOS.Services.Nv.NvGpuAS;
+using Ryujinx.HLE.HOS.Services.Nv.NvMap;
 using System;
 using System.Collections.Concurrent;
 
@@ -10,37 +11,25 @@ namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
 {
     class NvHostChannelIoctl
     {
-        private class ChannelsPerProcess
-        {
-            public ConcurrentDictionary<NvChannelName, NvChannel> Channels { get; private set; }
-
-            public ChannelsPerProcess()
-            {
-                Channels = new ConcurrentDictionary<NvChannelName, NvChannel>();
-
-                Channels.TryAdd(NvChannelName.Gpu, new NvChannel());
-            }
-        }
-
-        private static ConcurrentDictionary<KProcess, ChannelsPerProcess> Channels;
+        private static ConcurrentDictionary<KProcess, NvChannel> Channels;
 
         static NvHostChannelIoctl()
         {
-            Channels = new ConcurrentDictionary<KProcess, ChannelsPerProcess>();
+            Channels = new ConcurrentDictionary<KProcess, NvChannel>();
         }
 
-        public static int ProcessIoctlGpu(ServiceCtx Context, int Cmd)
-        {
-            return ProcessIoctl(Context, NvChannelName.Gpu, Cmd);
-        }
-
-        public static int ProcessIoctl(ServiceCtx Context, NvChannelName Channel, int Cmd)
+        public static int ProcessIoctl(ServiceCtx Context, int Cmd)
         {
             switch (Cmd & 0xffff)
             {
+                case 0x0001: return Submit           (Context);
+                case 0x0002: return GetSyncpoint     (Context);
+                case 0x0003: return GetWaitBase      (Context);
+                case 0x0009: return MapBuffer        (Context);
+                case 0x000a: return UnmapBuffer      (Context);
                 case 0x4714: return SetUserData      (Context);
                 case 0x4801: return SetNvMap         (Context);
-                case 0x4803: return SetTimeout       (Context, Channel);
+                case 0x4803: return SetTimeout       (Context);
                 case 0x4808: return SubmitGpfifo     (Context);
                 case 0x4809: return AllocObjCtx      (Context);
                 case 0x480b: return ZcullBind        (Context);
@@ -53,6 +42,138 @@ namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
             throw new NotImplementedException(Cmd.ToString("x8"));
         }
 
+        private static int Submit(ServiceCtx Context)
+        {
+            long InputPosition  = Context.Request.GetBufferType0x21().Position;
+            long OutputPosition = Context.Request.GetBufferType0x22().Position;
+
+            NvHostChannelSubmit Args = MemoryHelper.Read<NvHostChannelSubmit>(Context.Memory, InputPosition);
+
+            NvGpuVmm Vmm = NvGpuASIoctl.GetASCtx(Context).Vmm;
+
+            for (int Index = 0; Index < Args.CmdBufsCount; Index++)
+            {
+                long CmdBufOffset = InputPosition + 0x10 + Index * 0xc;
+
+                NvHostChannelCmdBuf CmdBuf = MemoryHelper.Read<NvHostChannelCmdBuf>(Context.Memory, CmdBufOffset);
+
+                NvMapHandle Map = NvMapIoctl.GetNvMap(Context, CmdBuf.MemoryId);
+
+                int[] CmdBufData = new int[CmdBuf.WordsCount];
+
+                for (int Offset = 0; Offset < CmdBufData.Length; Offset++)
+                {
+                    CmdBufData[Offset] = Context.Memory.ReadInt32(Map.Address + CmdBuf.Offset + Offset * 4);
+                }
+
+                Context.Device.Gpu.PushCommandBuffer(Vmm, CmdBufData);
+            }
+
+            //TODO: Relocation, waitchecks, etc.
+
+            return NvResult.Success;
+        }
+
+        private static int GetSyncpoint(ServiceCtx Context)
+        {
+            //TODO
+            long InputPosition  = Context.Request.GetBufferType0x21().Position;
+            long OutputPosition = Context.Request.GetBufferType0x22().Position;
+
+            NvHostChannelGetParamArg Args = MemoryHelper.Read<NvHostChannelGetParamArg>(Context.Memory, InputPosition);
+
+            Args.Value = 0;
+
+            MemoryHelper.Write(Context.Memory, OutputPosition, Args);
+
+            return NvResult.Success;
+        }
+
+        private static int GetWaitBase(ServiceCtx Context)
+        {
+            long InputPosition  = Context.Request.GetBufferType0x21().Position;
+            long OutputPosition = Context.Request.GetBufferType0x22().Position;
+
+            NvHostChannelGetParamArg Args = MemoryHelper.Read<NvHostChannelGetParamArg>(Context.Memory, InputPosition);
+
+            Args.Value = 0;
+
+            MemoryHelper.Write(Context.Memory, OutputPosition, Args);
+
+            return NvResult.Success;
+        }
+
+        private static int MapBuffer(ServiceCtx Context)
+        {
+            long InputPosition  = Context.Request.GetBufferType0x21().Position;
+            long OutputPosition = Context.Request.GetBufferType0x22().Position;
+
+            NvHostChannelMapBuffer Args = MemoryHelper.Read<NvHostChannelMapBuffer>(Context.Memory, InputPosition);
+
+            NvGpuVmm Vmm = NvGpuASIoctl.GetASCtx(Context).Vmm;
+
+            for (int Index = 0; Index < Args.NumEntries; Index++)
+            {
+                int Handle = Context.Memory.ReadInt32(InputPosition + 0xc + Index * 8);
+
+                NvMapHandle Map = NvMapIoctl.GetNvMap(Context, Handle);
+
+                if (Map == null)
+                {
+                    Logger.PrintWarning(LogClass.ServiceNv, $"Invalid handle 0x{Handle:x8}!");
+
+                    return NvResult.InvalidInput;
+                }
+
+                lock (Map)
+                {
+                    if (Map.DmaMapAddress == 0)
+                    {
+                        Map.DmaMapAddress = Vmm.MapLow(Map.Address, Map.Size);
+                    }
+
+                    Context.Memory.WriteInt32(OutputPosition + 0xc + 4 + Index * 8, (int)Map.DmaMapAddress);
+                }
+            }
+
+            return NvResult.Success;
+        }
+
+        private static int UnmapBuffer(ServiceCtx Context)
+        {
+            long InputPosition  = Context.Request.GetBufferType0x21().Position;
+
+            NvHostChannelMapBuffer Args = MemoryHelper.Read<NvHostChannelMapBuffer>(Context.Memory, InputPosition);
+
+            NvGpuVmm Vmm = NvGpuASIoctl.GetASCtx(Context).Vmm;
+
+            for (int Index = 0; Index < Args.NumEntries; Index++)
+            {
+                int Handle = Context.Memory.ReadInt32(InputPosition + 0xc + Index * 8);
+
+                NvMapHandle Map = NvMapIoctl.GetNvMap(Context, Handle);
+
+                if (Map == null)
+                {
+                    Logger.PrintWarning(LogClass.ServiceNv, $"Invalid handle 0x{Handle:x8}!");
+
+                    return NvResult.InvalidInput;
+                }
+
+                lock (Map)
+                {
+                    if (Map.DmaMapAddress != 0)
+                    {
+                        Vmm.Free(Map.DmaMapAddress, Map.Size);
+
+                        Map.DmaMapAddress = 0;
+                    }
+                }
+            }
+
+            return NvResult.Success;
+        }
+
         private static int SetUserData(ServiceCtx Context)
         {
             long InputPosition  = Context.Request.GetBufferType0x21().Position;
@@ -73,11 +194,11 @@ namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
             return NvResult.Success;
         }
 
-        private static int SetTimeout(ServiceCtx Context, NvChannelName Channel)
+        private static int SetTimeout(ServiceCtx Context)
         {
             long InputPosition = Context.Request.GetBufferType0x21().Position;
 
-            GetChannel(Context, Channel).Timeout = Context.Memory.ReadInt32(InputPosition);
+            GetChannel(Context).Timeout = Context.Memory.ReadInt32(InputPosition);
 
             return NvResult.Success;
         }
@@ -185,14 +306,9 @@ namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
             Context.Device.Gpu.Pusher.Push(Vmm, Gpfifo);
         }
 
-        public static NvChannel GetChannel(ServiceCtx Context, NvChannelName Channel)
+        public static NvChannel GetChannel(ServiceCtx Context)
         {
-            ChannelsPerProcess Cpp = Channels.GetOrAdd(Context.Process, (Key) =>
-            {
-                return new ChannelsPerProcess();
-            });
-
-            return Cpp.Channels[Channel];
+            return Channels.GetOrAdd(Context.Process, (Key) => new NvChannel());
         }
 
         public static void UnloadProcess(KProcess Process)
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelMapBuffer.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelMapBuffer.cs
new file mode 100644
index 00000000..1dfedf2d
--- /dev/null
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelMapBuffer.cs
@@ -0,0 +1,12 @@
+using System.Runtime.InteropServices;
+
+namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
+{
+    [StructLayout(LayoutKind.Sequential, Size = 0xc, Pack = 4)]
+    struct NvHostChannelMapBuffer
+    {
+        public int  NumEntries;
+        public int  DataAddress; //Ignored by the driver.
+        public bool AttachHostChDas;
+    }
+}
\ No newline at end of file
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelSubmit.cs b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelSubmit.cs
new file mode 100644
index 00000000..f776ad87
--- /dev/null
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvHostChannel/NvHostChannelSubmit.cs
@@ -0,0 +1,13 @@
+using System.Runtime.InteropServices;
+
+namespace Ryujinx.HLE.HOS.Services.Nv.NvHostChannel
+{
+    [StructLayout(LayoutKind.Sequential, Size = 8, Pack = 4)]
+    struct NvHostChannelSubmit
+    {
+        public int CmdBufsCount;
+        public int RelocsCount;
+        public int SyncptIncrsCount;
+        public int WaitchecksCount;
+    }
+}
\ No newline at end of file
diff --git a/Ryujinx.HLE/HOS/Services/Nv/NvMap/NvMapHandle.cs b/Ryujinx.HLE/HOS/Services/Nv/NvMap/NvMapHandle.cs
index 3f8a1517..e97e4ff4 100644
--- a/Ryujinx.HLE/HOS/Services/Nv/NvMap/NvMapHandle.cs
+++ b/Ryujinx.HLE/HOS/Services/Nv/NvMap/NvMapHandle.cs
@@ -11,6 +11,7 @@ namespace Ryujinx.HLE.HOS.Services.Nv.NvMap
         public int  Kind;
         public long Address;
         public bool Allocated;
+        public long DmaMapAddress;
 
         private long Dupes;