vgsp: m85 ext - fill-map image probe over 12 offsets across the whole context image

The probe now copies the header, VEID0 subctx header and ten pages spread from
0x1000 to 0xf0000, printing a nonzero-word count and first words per page, so
one ladder run shows whether the golden image is actually populated across its
0xf2d00 span (the GPC slices load from it).

1 files changed, 35 insertions, 17 deletions

diff --git a/kernel/gsp_phys_booter.c b/kernel/gsp_phys_booter.c
index bc0156f..d7ba43d 100644
--- a/kernel/gsp_phys_booter.c
+++ b/kernel/gsp_phys_booter.c
@@ -14429,7 +14429,16 @@ static void gsp_phys_rmcall_ctx_image_probe(void __iomem *bar0)
 	struct gsp_phys_boot_chunk *dst =
 		&boot_stage.rmcall_chunks[GSP_PHYS_RMCALL_BUF_DST];
 	struct gsp_phys_rmcall_ce_submit sub;
-	u64 regions[2];
+	/*
+	 * m86 fill-map update: the header, the VEID0 subctx header, then pages
+	 * spread across the whole image; each page's nonzero-word count shows
+	 * whether the golden image is actually populated at that offset.
+	 */
+	static const u64 offs[] = {
+		0x0ULL, 0xf3000ULL, 0x1000ULL, 0x2000ULL, 0x4000ULL,
+		0x8000ULL, 0x10000ULL, 0x20000ULL, 0x40000ULL, 0x80000ULL,
+		0xc0000ULL, 0xf0000ULL,
+	};
 	u32 bind_params;
 	int r;
 
@@ -14449,20 +14458,20 @@ static void gsp_phys_rmcall_ctx_image_probe(void __iomem *bar0)
 			status, rpc_result, r);
 	}
 
-	regions[0] = GSP_PHYS_RMCALL_VA_GR_CTX;		/* image header */
-	regions[1] = GSP_PHYS_RMCALL_VA_GR_CTX + 0xf3000ULL; /* subctx hdr VEID0 */
 	if (!r) {
 		u32 k;
 
-		for (k = 0; k < 2; k++) {
+		for (k = 0; k < ARRAY_SIZE(offs); k++) {
+			u64 va = GSP_PHYS_RMCALL_VA_GR_CTX + offs[k];
 			u32 rpc_result = 0, status = 0, poll = 0;
 			unsigned long deadline;
 			u8 *d;
+			u32 i, nz = 0;
 
 			WRITE_ONCE(*(u32 *)gsp_phys_chunk_page_va(sem, 0), 0);
 			wmb();
-			r = gsp_phys_rmcall_ce_submit_locked(bar0,
-					regions[k], GSP_PHYS_RMCALL_VA_DST,
+			r = gsp_phys_rmcall_ce_submit_locked(bar0, va,
+					GSP_PHYS_RMCALL_VA_DST,
 					GSP_PHYS_RMCALL_VA_SEM,
 					GSP_PHYS_PAGE_SIZE,
 					GSP_PHYS_RMCALL_SEM_PAYLOAD, false,
@@ -14470,8 +14479,8 @@ static void gsp_phys_rmcall_ctx_image_probe(void __iomem *bar0)
 			(void)rpc_result;
 			(void)status;
 			if (r) {
-				pr_err("vgsp-boot: rm-call: m84 image probe: CE submit for region %u refused (%d)\n",
-				       k, r);
+				pr_err("vgsp-boot: rm-call: m84 image probe: CE submit for offset 0x%llx refused (%d)\n",
+				       offs[k], r);
 				break;
 			}
 			deadline = jiffies +
@@ -14481,8 +14490,8 @@ static void gsp_phys_rmcall_ctx_image_probe(void __iomem *bar0)
 				if (poll == GSP_PHYS_RMCALL_SEM_PAYLOAD)
 					break;
 				if (time_after_eq(jiffies, deadline)) {
-					pr_err("vgsp-boot: rm-call: m84 image probe: region %u CE copy TIMEOUT (sem=0x%08x) - an engine-side walk of VA 0x%llx failed or the CE did not run\n",
-					       k, poll, regions[k]);
+					pr_err("vgsp-boot: rm-call: m84 image probe: offset 0x%llx CE copy TIMEOUT (sem=0x%08x)\n",
+					       offs[k], poll);
 					break;
 				}
 				msleep(GSP_PHYS_RMCALL_POLL_MS);
@@ -14492,14 +14501,23 @@ static void gsp_phys_rmcall_ctx_image_probe(void __iomem *bar0)
 			d = gsp_phys_chunk_page_va(dst, 0);
 			if (!d)
 				break;
-			pr_info("vgsp-boot: rm-call: m84 image probe: region %u VA=0x%llx copied OK\n",
-				k, regions[k]);
-			{
-				u32 i;
-
+			for (i = 0; i < GSP_PHYS_PAGE_SIZE / 4; i++)
+				if (get_unaligned_le32(d + i * 4))
+					nz++;
+			pr_info("vgsp-ctximg: offset 0x%06llx nonzero=%u/%u first=%08x %08x %08x %08x %08x %08x %08x %08x\n",
+				offs[k], nz, GSP_PHYS_PAGE_SIZE / 4,
+				get_unaligned_le32(d),
+				get_unaligned_le32(d + 4),
+				get_unaligned_le32(d + 8),
+				get_unaligned_le32(d + 12),
+				get_unaligned_le32(d + 16),
+				get_unaligned_le32(d + 20),
+				get_unaligned_le32(d + 24),
+				get_unaligned_le32(d + 28));
+			if (offs[k] == 0ULL || offs[k] == 0xf3000ULL) {
 				for (i = 0; i < 0x80; i += 8)
-					pr_info("vgsp-ctximg[%u] [%03x] = %08x %08x %08x %08x %08x %08x %08x %08x\n",
-						k, i * 4,
+					pr_info("vgsp-ctximg[%03llx] = %08x %08x %08x %08x %08x %08x %08x %08x\n",
+						offs[k] + i * 4,
 						get_unaligned_le32(d + i * 4),
 						get_unaligned_le32(d + (i + 1) * 4),
 						get_unaligned_le32(d + (i + 2) * 4),