vgsp: add gsp_phys_rmcall_channel to force one GR channel variant

Lets a bisection pair any promote shape with any channel shape, in
particular V6 (SKIP_CTXBUFFER_ALLOC, KMD-managed) with the server-managed
VA-only promote: with skip-ctxbuffer the server's kgrctxMapCtxBuffers does
not pre-book the engine context vaList, so the VA branch's
vaListSetManaged(NV_FALSE) gate should pass and the per-buffer commit can
program the engine context VA. Default 0 keeps the mode-selected walk
byte-identical. Not hardware-tested (GPU wedged; see Handoff ADDENDUM 51).

1 files changed, 19 insertions, 0 deletions

diff --git a/kernel/gsp_phys_gpu.c b/kernel/gsp_phys_gpu.c
index 94d6652..54fa84a 100644
--- a/kernel/gsp_phys_gpu.c
+++ b/kernel/gsp_phys_gpu.c
@@ -201,6 +201,25 @@ bool gsp_phys_srv_ctxcommit_value(void)
 EXPORT_SYMBOL_GPL(gsp_phys_srv_ctxcommit_value);
 
 /*
+ * Force one channel-alloc variant id on the GR ladder (default 0 = the
+ * promote-mode-selected default). 5 = V5 (client memdescs, no skip-ctxbuffer;
+ * the server books its managed engine context), 6 = V6 (SKIP_CTXBUFFER_ALLOC,
+ * KMD-managed: the server skips the ctx-buffer map/book, so a client VA-only
+ * promote can add and commit the engine context VA).
+ */
+static int gsp_phys_rmcall_channel_param;
+module_param_named(gsp_phys_rmcall_channel, gsp_phys_rmcall_channel_param,
+		   int, 0644);
+MODULE_PARM_DESC(gsp_phys_rmcall_channel,
+		 "Force one GR channel-alloc variant id (1/3/4/5/6); 0 = mode-selected default walk");
+
+int gsp_phys_rmcall_channel_value(void)
+{
+	return gsp_phys_rmcall_channel_param;
+}
+EXPORT_SYMBOL_GPL(gsp_phys_rmcall_channel_value);
+
+/*
  * P5 wave-1 shadow work forwarding (default 0). When set, the sema service
  * requests one real CE copy on the debugfs ladder's channel for each decoded
  * guest CE push, using only the ladder's module-owned buffers; the software