diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-11-16 01:14:56 -0500 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-11-16 01:14:56 -0500 |
| commit | 70e4e560b55e7155e6103601f078945c1d13c615 (patch) | |
| tree | eb387670ccfb729161b8557acbdbf1a7aa8c58a8 /kernel/gsp_phys_booter.c | |
| parent | 9d925787218af080316a81c85da4ea91884ecb21 (diff) | |
| download | open-iov-70e4e560b55e7155e6103601f078945c1d13c615.tar.gz open-iov-70e4e560b55e7155e6103601f078945c1d13c615.zip | |
vgsp: eng77 - capture and replay the client's VERTEX_STREAM_TURING CALL(2)
The beef 728-byte push carries CALL_MME_MACRO(2) (method 0x3810,
golden slot 2 = VERTEX_STREAM_TURING) with payload [0,0,0x0450e100]:
the Mesa mme_tu104_sim of golden slot 2 with those args emits
SET_VERTEX_STREAM_A_LOCATION 0x1c04<-0, 0x1c08<-0, and the Turing size
pair 0x600<-0, 0x604<-<the VA> - the VTG stream state the per-frame
DRAW window slices omit. The census (eng76, 0x100..0x1ff) finds it at
0x148 of the 728 push; the 247-word window carries the relocated call
at 0x2e0 (d2=0x1e22100), the 151-word DRAW window does not.
- service: sticky svc_gl_vs_call[3] from every 0x3810 run (VS-CALL
receipt), getter gsp_sema_gl_vs_call().
- booter release_save: when the window has no 0x3810 run, prepend the
captured ONE_INCR run (0x3810<-payload) ahead of the DRAW window;
the normal patch relocates the VA payload (sema_gl_mme_reloc).
- vs_pre buffer 12->24 words. Census range stays 0x100..0x1ff (eng76).
Diffstat (limited to 'kernel/gsp_phys_booter.c')
| -rw-r--r-- | kernel/gsp_phys_booter.c | 50 |
1 files changed, 49 insertions, 1 deletions
diff --git a/kernel/gsp_phys_booter.c b/kernel/gsp_phys_booter.c index 10fe9c3..3a8bb55 100644 --- a/kernel/gsp_phys_booter.c +++ b/kernel/gsp_phys_booter.c @@ -19295,6 +19295,28 @@ static u32 gsp_phys_gr_vs_scan(const struct gsp_phys_gr_submit *req, return 0; } +/* Does the window carry the VERTEX_STREAM_TURING CALL(2) run (0x3810)? */ +static bool gsp_phys_gr_mme_window_has_call2(const u32 *win, u32 n) +{ + u32 r = 0; + + while (r < n) { + u32 h = win[r]; + u32 op = (h >> 29) & 7u; + u32 cnt = (h >> 16) & 0x1fffu; + u32 span = 1u; + u32 m0 = (h & 0xfffu) << 2; + + if ((op == 1u || op == 3u || op == 5u) && cnt && + r + 1u + cnt <= n) + span = 1u + cnt; + if (m0 == 0x3810u) + return true; + r += span; + } + return false; +} + /* Does the window already carry the Turing stream-size run (0x600)? */ static bool gsp_phys_gr_mme_window_has_vs_size(const u32 *win, u32 n) { @@ -19425,7 +19447,7 @@ static void gsp_phys_gr_mme_release_save(const struct gsp_phys_gr_submit *req, bool draw_ia, keep = false; const char *prefer = NULL; u32 old_h = 0, old_win = 0; - u32 vs_pre[12]; + u32 vs_pre[24]; u32 vs_n = 0; static u32 prefer_log; @@ -19567,6 +19589,32 @@ static void gsp_phys_gr_mme_release_save(const struct gsp_phys_gr_submit *req, pr_info("vgsp-gr: MME-VS-SIZE synth j=0 size=0x100000 (window has VS) h=0x%x push=0x%llx\n", req->h_object, req->push_va); } + /* + * eng77: the client's VERTEX_STREAM_TURING macro + * (CALL_MME_MACRO(2) = 0x3810) payload carries the + * VTG stream size/base (the 0x0450e100-family VA). + * The DRAW-window slices omit it; replay the last + * captured call before the DRAW_IAB. + */ + { + u32 vcall[3]; + + if (!gsp_phys_gr_mme_window_has_call2( + req->stream, req->stream_n) && + gsp_sema_gl_vs_call(vcall) && + vs_n + 4u <= ARRAY_SIZE(vs_pre)) { + vs_pre[vs_n++] = (5u << 29) | + (3u << 16) | + ((0x3810u >> 2) & + 0xfffu); + vs_pre[vs_n++] = vcall[0]; + vs_pre[vs_n++] = vcall[1]; + vs_pre[vs_n++] = vcall[2]; + pr_info("vgsp-gr: MME-VS-CALL synth 0x3810 payload=%08x,%08x,%08x h=0x%x push=0x%llx\n", + vcall[0], vcall[1], vcall[2], + req->h_object, req->push_va); + } + } } if (req->stream_n + vs_n > GSP_PHYS_GR_MME_RELEASE_WIN_MAX) vs_n = 0; |
