r578 diag: event-function reachability probe (gsp_rm_ev_func)

The r577 A/B (sequence, bNotifyList, hClient) all yielded nv_post_event=0
with every record consumed by the guest: the drop is content-independent.
This knob posts event records with a caller-chosen RPC function instead of
POST_EVENT 0x1003; posting an event whose handler has an unmistakable guest
side effect (UCODE_LIBOS_PRINT 0x100C trips NV_ASSERT_FAILED on an
unsupported ucode; OS_ERROR_LOG 0x1006 logs an errString) proves whether
the guest dispatches our status-queue event records at all - dispatch vs
consumed-without-dispatch decides the next probe (inside _kgspRpcPostEvent
or earlier in the drain).

3 files changed, 25 insertions, 5 deletions

diff --git a/kernel/gsp_rm_emulation.c b/kernel/gsp_rm_emulation.c
index f44c01b..357997c 100644
--- a/kernel/gsp_rm_emulation.c
+++ b/kernel/gsp_rm_emulation.c
@@ -770,6 +770,19 @@ static void gsp_rm_evlog_post(struct nv_vgpu_device *vgpu,
 unsigned gsp_rm_ev_seq;
 unsigned gsp_rm_ev_nlist;
 unsigned gsp_rm_ev_cl;
+/*
+ * r578: reachability probe for the guest-RM event drop. When nonzero, event
+ * records carry this RPC function instead of POST_EVENT (0x1003). Posting a
+ * function with an unmistakable guest-side side effect (e.g.
+ * NV_VGPU_MSG_EVENT_UCODE_LIBOS_PRINT 0x100C with an unsupported ucodeEngDesc
+ * trips NV_ASSERT_FAILED in _kgspRpcUcodeLibosPrint, or
+ * NV_VGPU_MSG_EVENT_OS_ERROR_LOG 0x1006 logs an errString) proves whether
+ * the guest ever DISPATCHES our status-queue event records: visible side
+ * effect => the dispatch works and POST_EVENT is dropped inside
+ * _kgspRpcPostEvent; no side effect => the records are consumed without
+ * dispatch.
+ */
+unsigned gsp_rm_ev_func;
 
 static void gsp_rm_fill_ev_payload(const struct gsp_rm_emu *rm, u32 i,
 				   u8 *payload)
@@ -793,6 +806,11 @@ static u32 gsp_rm_ev_sequence(const struct gsp_rm_emu *rm)
 	return gsp_rm_ev_seq ? rm->stat_seq : 0;
 }
 
+static u32 gsp_rm_ev_function(void)
+{
+	return gsp_rm_ev_func ? gsp_rm_ev_func : GSPRM_RPC_EVENT_POST_EVENT;
+}
+
 /*
  * Post the GPU event that tells the guest's CPU-side RM a channel's
  * submitted work is complete. A real GPU raises an interrupt, the host GSP
@@ -827,8 +845,7 @@ int gsp_rm_post_channel_event(struct nv_vgpu_device *vgpu, u32 h_src)
 
 		gsp_rm_fill_ev_payload(rm, i, payload);
 		{
-			int rc = gsp_rm_send_flags(ctx,
-						   GSPRM_RPC_EVENT_POST_EVENT,
+			int rc = gsp_rm_send_flags(ctx, gsp_rm_ev_function(),
 						   gsp_rm_ev_sequence(rm), 0,
 						   payload, sizeof(payload),
 						   true);
@@ -854,7 +871,7 @@ int gsp_rm_post_channel_event(struct nv_vgpu_device *vgpu, u32 h_src)
 			gsp_rm_fill_ev_payload(rm, i, payload);
 			{
 				int rc = gsp_rm_send_flags(ctx,
-							   GSPRM_RPC_EVENT_POST_EVENT,
+							   gsp_rm_ev_function(),
 							   gsp_rm_ev_sequence(rm),
 							   0, payload,
 							   sizeof(payload),
@@ -896,7 +913,7 @@ static int gsp_rm_post_event_idx(struct gsp_emu_context *ctx, u32 idx)
 
 	gsp_rm_fill_ev_payload(rm, idx, payload);
 
-	rc = gsp_rm_send(ctx, GSPRM_RPC_EVENT_POST_EVENT,
+	rc = gsp_rm_send(ctx, gsp_rm_ev_function(),
 			 gsp_rm_ev_sequence(rm), 0, payload, sizeof(payload));
 	if (!rc)
 		rm->events[idx].last_post = jiffies;
@@ -977,7 +994,7 @@ int gsp_rm_post_all_events(struct nv_vgpu_device *vgpu)
 
 		gsp_rm_fill_ev_payload(rm, i, payload);
 		{
-			int rc = gsp_rm_send(ctx, GSPRM_RPC_EVENT_POST_EVENT,
+			int rc = gsp_rm_send(ctx, gsp_rm_ev_function(),
 					     gsp_rm_ev_sequence(rm), 0,
 					     payload, sizeof(payload));
 
diff --git a/kernel/gsp_rm_emulation.h b/kernel/gsp_rm_emulation.h
index 85ee4d9..a0bd4c7 100644
--- a/kernel/gsp_rm_emulation.h
+++ b/kernel/gsp_rm_emulation.h
@@ -224,6 +224,7 @@ extern unsigned gsp_rm_ev_log_max;
 extern unsigned gsp_rm_ev_seq;
 extern unsigned gsp_rm_ev_nlist;
 extern unsigned gsp_rm_ev_cl;
+extern unsigned gsp_rm_ev_func;
 
 int  gsp_rm_emu_init(struct nv_vgpu_device *vgpu, struct gsp_emu_context *ctx);
 void gsp_rm_emu_fini(struct gsp_emu_context *ctx);
diff --git a/kernel/main.c b/kernel/main.c
index 8d02e5b..d821762 100644
--- a/kernel/main.c
+++ b/kernel/main.c
@@ -2822,6 +2822,8 @@ module_param_named(gsp_rm_ev_nlist, gsp_rm_ev_nlist, uint, 0644);
 MODULE_PARM_DESC(gsp_rm_ev_nlist, "set bNotifyList on event records (default 0: specific-hEvent dispatch)");
 module_param_named(gsp_rm_ev_cl, gsp_rm_ev_cl, uint, 0644);
 MODULE_PARM_DESC(gsp_rm_ev_cl, "override event record hClient (default 0: the allocating client)");
+module_param_named(gsp_rm_ev_func, gsp_rm_ev_func, uint, 0644);
+MODULE_PARM_DESC(gsp_rm_ev_func, "reachability probe: event records carry this RPC function instead of POST_EVENT 0x1003 (default 0)");
 
 module_param_named(gsp_sema_walk_trace, gsp_sema_walk_trace, uint, 0644);
 MODULE_PARM_DESC(gsp_sema_walk_trace, "Trace GMMU walks for unresolved high-VA operands (default 0: off)");