sema-svc: gate per-push/per-doorbell/RPC tracing behind default-off sema_dbg

A healthy run left dmesg dominated by per-operation lines (sema-push,
sema-lda, sema-chan rel-apply, sema-grce, DOORBELL/BAR writes, GSPRM
per-RPC lines, the teardown garbage-element stream). Introduce the
module parameter sema_dbg (default 0) and route every high-frequency
diagnostic through it; the finer-grained watch knobs (sema_ops_log,
sema_copy_watch, sema_copy_pat, sema_copy_dst, sema_copy_dst2) only
take effect while it is on. Boot stats, channel registrations, pin
events and state changes stay unconditional.

Also carries the per-side translation provenance (op_tr_src, printed
as prov= in the destination-window watch) and the 16 KB destination
window used to observe the guest's page-table writes.

Verified: fresh boot + full cdt2 smoke test (64/64, ALL OK) leaves a
520-line dmesg with zero per-push/per-doorbell/per-RPC lines.

1 files changed, 28 insertions, 5 deletions

diff --git a/kernel/main.c b/kernel/main.c
index d6e6c9d..dd38402 100644
--- a/kernel/main.c
+++ b/kernel/main.c
@@ -717,7 +717,9 @@ static ssize_t nv_vgpu_vfio_dev_read(struct vfio_device *vdev, char __user *buf,
 	struct nv_vgpu_device *vgpu = container_of(vdev, struct nv_vgpu_device, vfio_dev);
 	loff_t pos = *ppos;
 	
-	pr_info("gsp-resman-vfio: VFIO device read count=%zu pos=%lld (0x%llx)\n", count, pos, pos);
+	if (gsp_sema_dbg)
+		pr_info("gsp-resman-vfio: VFIO device read count=%zu pos=%lld (0x%llx)\n",
+			count, pos, pos);
 
 	/* Config space region: VFIO_PCI_INDEX_TO_OFFSET(VFIO_PCI_CONFIG_REGION_INDEX) = 0x70000000000 */
 	if (pos >= 0x70000000000ULL && pos < 0x70000000000ULL + 4096) {
@@ -772,7 +774,9 @@ static ssize_t nv_vgpu_vfio_dev_write(struct vfio_device *vdev, const char __use
 	struct nv_vgpu_device *vgpu = container_of(vdev, struct nv_vgpu_device, vfio_dev);
 	loff_t pos = *ppos;
 	
-	pr_info_ratelimited("gsp-resman-vfio: VFIO device write count=%zu pos=%lld (0x%llx)\n", count, pos, pos);
+	if (gsp_sema_dbg)
+		pr_info_ratelimited("gsp-resman-vfio: VFIO device write count=%zu pos=%lld (0x%llx)\n",
+				    count, pos, pos);
 
 	/* Config space region - route through the per-register filter so the
 	 * guest can only modify writable registers (not vendor/device IDs,
@@ -802,14 +806,14 @@ static ssize_t nv_vgpu_vfio_dev_write(struct vfio_device *vdev, const char __use
 		if (copy_from_user(tmp, buf, count))
 			return -EFAULT;
 
-		if (bar0_log < 64 || (u64)pos == 0x88704) {
+		if ((u64)pos == 0x88704 || (gsp_sema_dbg && bar0_log < 64)) {
 			static u32 db_log;
 			u64 v = 0;
 
 			memcpy(&v, tmp, count);
 
 			if ((u64)pos == 0x88704) {
-				if (db_log < 512) {
+				if (gsp_sema_dbg && db_log < 512) {
 					pr_info("gsp-resman-vfio: DOORBELL write pos=0x%llx size=%zu val=0x%llx\n",
 						(u64)pos, count, v);
 					db_log++;
@@ -837,7 +841,7 @@ static ssize_t nv_vgpu_vfio_dev_write(struct vfio_device *vdev, const char __use
 		u64 bar1_off = (u64)pos - (1ULL << 40);
 		static u32 bar1_log;
 
-		if (bar1_log < 64) {
+		if (gsp_sema_dbg && bar1_log < 64) {
 			u64 v = 0;
 
 			memcpy(&v, buf, count);
@@ -2450,6 +2454,17 @@ module_param_named(sema_skip_ops, gsp_sema_skip_ops, bool, 0644);
 MODULE_PARM_DESC(sema_skip_ops, "Apply releases without executing pushed memsets (default 0)");
 
 /*
+ * Master diagnostic switch for the per-push/per-doorbell/per-op tracing
+ * (sema-push, sema-lda, sema-chan rel-apply, sema-grce, DOORBELL, the
+ * GSPRM per-RPC lines, ...). Off by default so a healthy boot leaves the
+ * kernel log quiet; turn it on when instrumenting a cycle. The watch knobs
+ * below (ops_log, copy_watch, copy_pat, copy_dst) only take effect while
+ * this is on.
+ */
+module_param_named(sema_dbg, gsp_sema_dbg, bool, 0644);
+MODULE_PARM_DESC(sema_dbg, "Verbose per-push/per-RPC emulation tracing (default 0)");
+
+/*
  * Executed-op log cap: how many CE fills/copies are logged with full
  * operands. Raise it (e.g. 4096) to audit a data path; the default keeps a
  * run quiet.
@@ -2484,6 +2499,14 @@ module_param_named(sema_copy_dst, gsp_sema_copy_dst, ullong, 0644);
 MODULE_PARM_DESC(sema_copy_dst, "Dump source bytes of copies landing in this page (default 0)");
 
 /*
+ * Same watch, second 16 KB-aligned window (0 = off). Two windows let one
+ * run cover both the page-directory table and the page-table page it points
+ * into, e.g. 0x208000 and 0x230000.
+ */
+module_param_named(sema_copy_dst2, gsp_sema_copy_dst2, ullong, 0644);
+MODULE_PARM_DESC(sema_copy_dst2, "Second destination window for sema_copy_dst (default 0)");
+
+/*
  * Grace window (ms) an untranslatable CE entry stays pending before the
  * fallback acknowledges its release and consumes it. Its work must run before
  * the release the client synchronizes on; the translation usually lands within