From 1ae11a19a3bc1dae2f33479416c0df0628f1c381 Mon Sep 17 00:00:00 2001 From: Arthur Hanson Rasmusson Date: Mon, 16 Nov 2026 04:15:26 -0500 Subject: vgsp: eng83 - decode the 0x3810 payload shape: base is d[4], end is d[2] c836 proved the client's VERTEX_STREAM CALL(2) payload is [stream, end_hi, end_lo, base_hi, base_lo] (MME-VSTREAM: end=0x1e02100 base=0x1e001f4, MME-VSTREAM-LINE base=0x450c1f4 end=0x450e100 pages=3 first8=...3f800000 3e4ccccd...). eng79 wrongly took d[2] (the END 0x450e100, a zero page) as the stream base; the base d[4]=0x450c1f4 carries the float2 vertices. The Doris slot-2 sim with the full 5-word payload emits LOCATION 0x1c08<-base and SIZE 0x604<-end-base. - service: svc_gl_vs_call[5] (getter out[5]) - capture v0..v4, log 5 words; eng82 page-set collection adds the end (v2) and base (v4). - booter eng79: slot-2 base = vcall[4] (require vcall[1]==0, vcall[3]==0, aligned), end = vcall[2]; size pair = end-base; receipt base=.. end=..; VS-BLOB2 dumps at the base. - booter eng77: emit the FULL 5-word 0x3810 call (op=5 cnt=5) even when the window carries its own call (the slot-34 reorder moves the draw ahead of it); the 0x3810 MME-VSTREAM patch maps [base,end] into one contiguous shadow line. --- kernel/gsp_sema_service.c | 29 +++++++++++++++++++++-------- 1 file changed, 21 insertions(+), 8 deletions(-) (limited to 'kernel/gsp_sema_service.c') diff --git a/kernel/gsp_sema_service.c b/kernel/gsp_sema_service.c index aeddb7f..4cd2eca 100644 --- a/kernel/gsp_sema_service.c +++ b/kernel/gsp_sema_service.c @@ -9808,7 +9808,7 @@ static u32 svc_gl_vs_log; * call; capture the last observed payload (3 words) so the booter can * replay it before the DRAW_IAB. The census scans every beef push. */ -static u32 svc_gl_vs_call[3]; +static u32 svc_gl_vs_call[5]; static u8 svc_gl_vs_call_ok; static u32 svc_gl_vs_call_log; @@ -9819,6 +9819,8 @@ bool gsp_sema_gl_vs_call(u32 out[3]) out[0] = READ_ONCE(svc_gl_vs_call[0]); out[1] = READ_ONCE(svc_gl_vs_call[1]); out[2] = READ_ONCE(svc_gl_vs_call[2]); + out[3] = READ_ONCE(svc_gl_vs_call[3]); + out[4] = READ_ONCE(svc_gl_vs_call[4]); return true; } @@ -10971,11 +10973,16 @@ static void svc_gl_field(u32 addr, u32 count, const u8 *data, u32 dlen) if (addr == 0x3810u) { u32 k; - for (k = 1; k < count && k < 4u; k++) { + /* + * [stream, end_hi, end_lo, base_hi, base_lo]: collect the + * 32-bit end/base words (v[2], v[4]) - the base page carries + * the vertex payload. + */ + for (k = 2; k < count && k < 5u; k++) { u32 w = v[k]; if (w >= 0x100000u && w < 0x1000000000ULL && - !(w & 3u)) + !(w & 3u) && (k == 2u || k == 4u)) svc_gl_add_ptr((u64)w); } return; @@ -18362,7 +18369,7 @@ static void svc_mme_census(const u8 *buf, u32 len, u64 push_va, u32 h_object) (addr >= 0x1c00u && addr <= 0x1cf8u) || (addr >= 0x3400u && addr <= 0x34fcu) || (addr >= 0x3800u && addr <= 0x39fcu)) { - u32 v0 = 0, v1 = 0, v2 = 0; + u32 v0 = 0, v1 = 0, v2 = 0, v3 = 0, v4 = 0; if (off + 8 <= len) memcpy(&v0, buf + off + 4, 4); @@ -18370,23 +18377,29 @@ static void svc_mme_census(const u8 *buf, u32 len, u64 push_va, u32 h_object) memcpy(&v1, buf + off + 8, 4); if (off + 16 <= len) memcpy(&v2, buf + off + 12, 4); + if (off + 20 <= len) + memcpy(&v3, buf + off + 16, 4); + if (off + 24 <= len) + memcpy(&v4, buf + off + 20, 4); /* * eng77: remember the VERTEX_STREAM_TURING * CALL_MME_MACRO(2) payload (the VTG stream state * the DRAW window slices omit). */ - if (addr == 0x3810u && off + 16 <= len) { + if (addr == 0x3810u && off + 24 <= len) { bool changed = !READ_ONCE(svc_gl_vs_call_ok) || - READ_ONCE(svc_gl_vs_call[2]) != v2; + READ_ONCE(svc_gl_vs_call[4]) != v4; WRITE_ONCE(svc_gl_vs_call[0], v0); WRITE_ONCE(svc_gl_vs_call[1], v1); WRITE_ONCE(svc_gl_vs_call[2], v2); + WRITE_ONCE(svc_gl_vs_call[3], v3); + WRITE_ONCE(svc_gl_vs_call[4], v4); WRITE_ONCE(svc_gl_vs_call_ok, 1); if (changed && svc_gl_vs_call_log < 8u) { svc_gl_vs_call_log++; - pr_info("sema-gl: VS-CALL 0x3810 payload=%08x,%08x,%08x\n", - v0, v1, v2); + pr_info("sema-gl: VS-CALL 0x3810 payload=%08x,%08x,%08x,%08x,%08x\n", + v0, v1, v2, v3, v4); } } lc += scnprintf(line + lc, sizeof(line) - lc, -- cgit v1.2.3