nx2: support mapping modify so NVN2 places compressible textures

NVN2 gives a texture in a compressible memory pool its GPU address only
after NvRmGpuMappingModify switched the pool's mapping to the texture's
kind. NvRm fails that call unless the GPU characteristics report
NVGPU_GPU_FLAGS_SUPPORT_MAPPING_MODIFY, so such textures kept address 0
and draws into them went nowhere: title 1's cup-opening movie stayed on
its first frame (also its 640x360 and 512x512 passes).

Report the flag and answer the request nn::gpu::sdk then sends, gpu
sub-session 212 (NormalModifyMemoryKind4K), with the wanted kind and a
signalled job slot. The movie now plays on screen and the race follows.

2 files changed, 21 insertions, 0 deletions

diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
index df2421cd13..5db984b53d 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
@@ -202,6 +202,11 @@ void ApplyNx2Characteristics(GC& gc) {
     gc.flags |= 0x100;             // NVGPU_GPU_FLAGS_SUPPORT_TSG: NvRm binds channels to a TSG
     gc.flags |= 1ULL << 22;        // NVGPU_GPU_FLAGS_SUPPORT_TSG_SUBCONTEXTS (see below)
     gc.flags |= 1ULL << 26;        // NVGPU_GPU_FLAGS_SUPPORT_SYNCPOINT_ADDRESS: syncpoint shim
+    // NVGPU_GPU_FLAGS_SUPPORT_MAPPING_MODIFY: NVN2 gives a texture in a compressible pool its
+    // address only after NvRmGpuMappingModify switched the pool's mapping to the texture's kind.
+    // Without the flag NvRm fails that call and the texture stays at GPU address 0 (title 1's
+    // movie and the passes rendered into compressible pools never showed).
+    gc.flags |= 1ULL << 34;
 
     // The Switch 2 layout keeps nine s16 interface versions at 0x70..0x80 (like nvgpu's
     // *_ioctl_nr_last); its NvRm requires e.g. [0x78] >= 6 and enables features above 0xa, 0x1f
diff --git a/src/core/hle/service/nx2_gsv.h b/src/core/hle/service/nx2_gsv.h
index cfef3f1229..5eab9d420f 100644
--- a/src/core/hle/service/nx2_gsv.h
+++ b/src/core/hle/service/nx2_gsv.h
@@ -309,6 +309,7 @@ public:
             {101, &INx2GpuSession::SubRegisterMemory, "CreateGpuAccessibleMemory"},
             {102, &INx2GpuSession::SubDestroyMemory, "DestroyGpuAccessibleMemory (guessed)"},
             {211, &INx2GpuSession::SubSubmitJob, "NormalMapGpuMemory4K"},
+            {212, &INx2GpuSession::SubModifyKind, "NormalModifyMemoryKind4K"},
             {213, &INx2GpuSession::SubUnmapJob, "NormalUnmapGpuMemory4K"},
             {214, &INx2GpuSession::SubCreateGpuOnlyRegion, "CreateSparseRegion"},
             {215, &INx2GpuSession::SubDestroySparseRegion, "DestroySparseRegion"},
@@ -457,6 +458,21 @@ private:
         rb.Push(ResultSuccess);
     }
 
+    // Sub 212 (nn::gpu::sdk::NormalModifyMemoryKind4K, behind NvRmGpuMappingModify): word 2 the
+    // address space, 3-4 first GPU address, 5-6 last, 7 the wanted (compressible) kind, 8 the
+    // fallback kind, 9 a sync flag, 10 the job slot. The reply's word 3 is the kind applied. Kinds
+    // do not matter to the emulated GPU, so keep the mapping and report the wanted kind.
+    void SubModifyKind(HLERequestContext& ctx) {
+        const auto* w = ctx.CommandBuffer();
+        const u64 first = u64{w[3]} | (u64{w[4]} << 32);
+        const u64 last = u64{w[5]} | (u64{w[6]} << 32);
+        const u32 kind = w[7] & 0xff;
+        LOG_DEBUG(Service, "nx2:gpu 212 modify [{:#x}, {:#x}] kind={:#x} fallback={:#x} slot={}",
+                  first, last, kind, w[8] & 0xff, w[10]);
+        EventFor(w[10])->Signal(system.Kernel());
+        TipcReplyWords(ctx, {kind});
+    }
+
     // Sub 213: words 3-4 first GPU address, 5-6 last, job slot in word 8.
     void SubUnmapJob(HLERequestContext& ctx) {
         const auto* w = ctx.CommandBuffer();