vgsp-disp: host-side EVO display engine (channels, push buffer, notifier, raster)

The guest nvidia-modeset drives the display through per-channel NV-DMA
push buffers kicked via the BAR0 UDISP control windows.  Emulate that
engine contract host-side, unmodified guest:

- channel registration: display-channel RM objects (core 0xc57d, window
  0xc57e, cursor 0xc57b, 0xc57a) are noted from the alloc stream, and the
  internal control NV2080_CTRL_CMD_INTERNAL_DISPLAY_CHANNEL_PUSHBUFFER
  (0x20800a58) records each push buffer (guest PA, size, class,
  instance, validity);
- PUT polling + GET publication in the channel control pages (core
  0x680000, window 0x690000+n*0x1000, cursor 0x6D8000+n*0x1000);
- NV-DMA method parser (METHOD/ONE_INC/NONINC/JUMP/SET_SUBDEVICE_MASK)
  with the wrap-JUMP follow;
- completion: SET_CONTEXT_DMA_NOTIFIER / SET_NOTIFIER_CONTROL tracking,
  exact notifier-surface resolution through the display instance hash
  table (no heuristic writes), NV_DISP_NOTIFIER FINISHED on UPDATE
  (0x200), and the channel completion event post;
- free-running 1080p60 raster line for NVC57E_GET_LINE.

All effects are behind independent knobs (disp_engine, disp_finish,
disp_vblank, disp_events, disp_notifier_pa) with verbose diagnostics
(disp_dbg, disp_dump), so the previous behaviour stays one load away.

1 files changed, 18 insertions, 0 deletions

diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c
index 6f0103b..fb75f33 100644
--- a/kernel/gsp_sema_service.c
+++ b/kernel/gsp_sema_service.c
@@ -53,6 +53,7 @@
 #include "bar_emulation.h"
 #include "gsp_sema_service.h"
 #include "gsp_rm_emulation.h"
+#include "gsp_display_engine.h"
 #include "gsp_phys_booter.h"
 #include "gsp_phys_gpu.h"
 #include "gsp_cuda_bridge.h"
@@ -44686,6 +44687,16 @@ static int svc_thread(void *arg)
 				pumped = true;
 
 			/*
+			 * r1621: display-engine progress.  The guest's
+			 * nvidia-modeset kicks its EVO channels through the
+			 * BAR0 UDISP control windows; consume the push buffers,
+			 * publish GET and complete each UPDATE's core notifier.
+			 * The boot display-init and every Xorg mode set wait on
+			 * that completion.
+			 */
+			gsp_disp_engine_pump(vgpu);
+
+			/*
 			 * Host CUDA bridge: marshal one queued launch into a
 			 * daemon job and resolve completions. Inert until a
 			 * launch is seen.
@@ -44876,8 +44887,11 @@ int gsp_sema_service_start(void)
 	 * while gsp_phys_work_forward is enabled; the software path needs no
 	 * workqueue at all.
 	 */
+	if (gsp_disp_engine_init())
+		pr_warn("sema-svc: display engine init failed\n");
 	svc.task = kthread_run(svc_thread, NULL, "gsp-sema-svc");
 	if (IS_ERR(svc.task)) {
+		gsp_disp_engine_fini();
 		int ret = PTR_ERR(svc.task);
 
 		svc.task = NULL;
@@ -44921,6 +44935,7 @@ void gsp_sema_service_stop(void)
 		svc_fwd_wq = NULL;
 	}
 	/* Thread is gone: release the bridge's pending jobs and debugfs files. */
+	gsp_disp_engine_fini();
 	gsp_cuda_bridge_fini();
 	svc_setup_demand_debugfs_fini();
 	/* A fresh session must not inherit shadowed bytes from the last one. */
@@ -44957,6 +44972,9 @@ void gsp_sema_service_vgpu_cleanup(struct nv_vgpu_device *vgpu)
 	if (!vgpu)
 		return;
 
+	/* r1621: drop the guest's display channels with the rest of its state. */
+	gsp_disp_engine_vgpu_cleanup(vgpu);
+
 	/*
 	 * The forward work item runs on its own kthread and may be using the
 	 * request's vgpu; quiesce waits for it (bounded by the submit's