vgsp+diag: attr/pipeline prelude (families 0x1100-0x1200/0x2000-0x2060, dedup, 512w, program-pair VA record; emitted via patch_window relocation, run-aligned after MME prelude), mature-prelude gate (>=64 0x114 code words, sema_gr_prelude_gate_ms=60000 busy-defer), pin_gr_stream 0.75s pass, trap regs in sked sampler - first clean DS+MME (ds_esr=80000000 mme_esr=80000000, a2040a00 gone); 0x1168 CLASS ERROR persists as MME-MACRO-GENERATED WRITE (no 0x1168 header in emitted stream; macro(2) args [0,0,600000,0,610800]); next: disassemble macro (2) at pc 0xa + trapped-addr read pre-latch - SESSION 153 r1095-r1100

1 files changed, 22 insertions, 0 deletions

diff --git a/kernel/gsp_sema_service.h b/kernel/gsp_sema_service.h
index e47e2e8..7a9feb7 100644
--- a/kernel/gsp_sema_service.h
+++ b/kernel/gsp_sema_service.h
@@ -464,6 +464,28 @@ int gsp_sema_gr_reloc_resolve(u64 va, struct gsp_sema_gr_reloc *out);
 const u32 *gsp_sema_mme_prelude_get(u32 *n, u32 *gen);
 
 /*
+ * SESSION 153-CONTINUED-57: is the captured MME prelude mature enough to
+ * launch against - i.e. does it hold the client's macro program (>= 64
+ * captured 0x114 code words)? The r1094 premature-emission poison
+ * (MME ILLEGAL_OPCODE at +5 ms, capture still at 9 scratch words) is
+ * gated on this: the sema forward path withholds beef launches while it
+ * reads false, bounded by sema_gr_prelude_gate_ms.
+ */
+bool gsp_sema_mme_prelude_mature(void);
+
+/*
+ * SESSION 153-CONTINUED-57: attr/pipeline state prelude accessor (C597
+ * families 0x1100..0x11fc and 0x2000..0x205c, verbatim, first-encounter
+ * order, deduped). Same contract as gsp_sema_mme_prelude_get: preplay once
+ * per generation ahead of the first 3D window replay. The runs carry the
+ * pipeline program binds whose SPH the draw shader validates, so the
+ * builder also registers the program VAs (gsp_sema_attr_prelude_ptrs) in
+ * the launch page set to make the address relocation land.
+ */
+const u32 *gsp_sema_attr_prelude_get(u32 *n, u32 *gen);
+const u64 *gsp_sema_attr_prelude_ptrs(u32 *n);
+
+/*
  * r378 write guards. Both re-validate a guest-memory write target that was
  * resolved earlier (ledger/pin/walk/counter registration, or the GR submit's
  * per-page resolution) against the CURRENT resolution before the bytes land,