From 770d6b4af7f44acc104c6149b9c473de1db607e7 Mon Sep 17 00:00:00 2001 From: Arthur Hanson Rasmusson Date: Thu, 12 Nov 2026 23:25:00 -0500 Subject: vgsp: flip-event delivery defaults to assoc+strict (clears Flip event timeout); DDX latch reset clauses Wall (b) 'Flip event timeout on head 0' fix (validated c171): NVKMS/kernel- callback events register hSrcResource=0, so every display UPDATE completion rode the all-events shotgun (6-7 POST_EVENT records each), the guest's status-queue read cursor fell 20-44 slots behind (rc=-11 ring-full bursts) and flip records arrived past the 3 s NvDrm window. Default the r1765 pair ON - gsp_rm_ev_assoc (hSrc==0 -> alloc-parent channel) and gsp_disp_events_strict (armed UPDATE posts only to that channel's events, no fallback) - the exact combination r1771 measured as '0 flip WARNs'. Validating cycle c171: registrations carry real sources (hEvent=0x1007a hSrc=0x1001b etc.), UPDATE posts are exact 1:1 (posts=27), WARN count 1 for the boot, serial timeouts=0 (c160 pre-fix: 34). Both stay runtime-writable. Also stage the gsp_sema_service.c candidates for wall (a) (the later-DMA 0x51 is a consequence of 'Failed to idle DMA' on a recycled channel): the served-refusal in svc_userd_ddx_reprobe now exempts a latch reset to put==0, and the put-reset latch keep extends to the DDX family (was kick-only). These clauses target c171's observed reset/junk-adopt sequence; validation pending on a roll where wall (a) is otherwise survivable. --- kernel/gsp_rm_emulation.c | 24 +++++++++++++++++------- 1 file changed, 17 insertions(+), 7 deletions(-) (limited to 'kernel/gsp_rm_emulation.c') diff --git a/kernel/gsp_rm_emulation.c b/kernel/gsp_rm_emulation.c index 0cef9e7..957729b 100644 --- a/kernel/gsp_rm_emulation.c +++ b/kernel/gsp_rm_emulation.c @@ -818,16 +818,26 @@ unsigned gsp_rm_ev_cl; */ unsigned gsp_rm_ev_func; /* - * r1767: gate the r1765 flip-event association (hSrcResource == 0 -> - * alloc-header hParent). Default 0 keeps the pre-r1766 event semantics; - * the RCA exonerated the association from the r1766 soft-locks but not - * from the r1766 ISR crash class, so it stays off until validated - * end-to-end. + * r1765 flip-event association (hSrcResource == 0 -> alloc-header hParent). + * Default was 0 (pre-r1766 semantics) pending through r1771 A/B validation; + * r1771 measured on the live stack: assoc=1 alone -> flip WARNs + burst + * storms, assoc=1 + disp_events_strict=1 -> 0 flip WARNs (the r1771 X death + * was the separate DDX channel-idle wall, since fixed by 13a0c5d6/acb51d53). + * r1780 rationale (c160 receipts, eng5 session): NVKMS/kernel-callback + * events register hSrc=0, so every completion rode the all-events shotgun + * (6-7 POST_EVENT records per completion). The guest stat-queue read + * cursor sat 20-44 slots behind our write pointer, rc=-11 (ring full) + * refusals burst during X starts, and flip-event records reached the guest + * long after the 3 s NvDrm flip window -> "Flip event timeout on head 0" + * x7 per X start with zero nvidia-modeset kthread delivery. With the + * association the display UPDATE post (h_object=0x1001b) exact-matches only + * the NVKMS flip events registered on that channel: 1:1 delivery, no flood. + * Still runtime-writable for A/B. */ -static bool gsp_rm_ev_assoc; /* default 0: pre-r1766 event semantics */ +static bool gsp_rm_ev_assoc = true; /* r1780: default 1 (validated r1771+r1780) */ module_param_named(ev_assoc, gsp_rm_ev_assoc, bool, 0644); MODULE_PARM_DESC(ev_assoc, - "associate hSrcResource==0 events with their alloc parent (r1765 flip-event fix; default 0)"); + "associate hSrcResource==0 events with their alloc parent (r1765 flip-event fix; default 1)"); static void gsp_rm_fill_ev_payload(const struct gsp_rm_emu *rm, u32 i, u8 *payload) -- cgit v1.2.3