From 5eac589d9a7af8f8b321b00398575070b9a88513 Mon Sep 17 00:00:00 2001 From: Arthur Hanson Rasmusson Date: Tue, 22 Sep 2026 02:16:18 -0400 Subject: kernel: floorsweeping masks + interrupt tree + WPR2 race + fn-10 handler - STATIC_KGR_GET_FLOORSWEEPING_MASKS(0x20800a26): TU104 gpcMask=0x3f (6 GPCs), 4 TPC/GPC. numGpcs==0 killed krcWatchdogInit -> RmInitAdapter bailed ('numGpcs > 0' assert at kernel_graphics_object.c:191). - CPU interrupt tree (osVerifyInterrupts): driver accesses VF window at virtualRegPhysOffset=NV_VIRTUAL_FUNCTION_FULL_PHYS_OFFSET(0xB80000)+off, NOT 0x1640 (was watching wrong offset). On LEAF_TRIGGER write, set LEAF/TOP pending bits + inject MSI-X so the sanity-test ISR runs - osVerifySystemEnvironment now passes. - WPR2 race: clear boot state BEFORE CPUCTL=HALTED on Booter-unload kick so WPR2 is down when driver checks kgspIsWpr2Up ('WPR2 is still up'). - Dedicated GSP_RM_FREE(10) handler: echo NVOS00_PARAMETERS + status=NV_OK (proper free_v03_00 form). fn-10 timeout persists despite this - characterized as driver-side reply-matching issue, not content. --- kernel/gsp_emulation.c | 41 ++++++++++++++++++++++++++++++++++++++--- 1 file changed, 38 insertions(+), 3 deletions(-) (limited to 'kernel/gsp_emulation.c') diff --git a/kernel/gsp_emulation.c b/kernel/gsp_emulation.c index c8f8041..94af0d2 100644 --- a/kernel/gsp_emulation.c +++ b/kernel/gsp_emulation.c @@ -179,6 +179,39 @@ static void gsp_emu_poll_boot(struct gsp_emu_context *ctx) u32 sec2_cpuctl = gsp_emu_shadow_read32(ctx, GSP_EMU_REG_SEC2_CPUCTL); u32 engine = gsp_emu_shadow_read32(ctx, GSP_EMU_REG_GSP_ENGINE); u32 mmu_inv = gsp_emu_shadow_read32(ctx, GSP_EMU_REG_MMU_INVALIDATE); + u32 sw_vector; + + /* Software-interrupt trigger: the driver writes a vector to + * CPU_INTR_LEAF_TRIGGER to fire it (the osVerifyInterrupts sanity + * test uses the doorbell vector). Set the pending bits in the LEAF + * and TOP interrupt-tree registers and inject MSI-X so the guest's + * ISR runs - without this the sanity test times out and + * osVerifySystemEnvironment fails, bailing RmInitAdapter. */ + sw_vector = gsp_emu_shadow_read32(ctx, GSP_EMU_REG_CPU_INTR_LEAF_TRIGGER); + if (sw_vector) { + u32 leaf_idx = sw_vector / 32; + u32 leaf_bit = sw_vector % 32; + u32 subtree = sw_vector / 64; + u32 top_idx = subtree / 32; + u32 top_bit = subtree % 32; + u32 leaf_val = gsp_emu_shadow_read32(ctx, + GSP_EMU_REG_CPU_INTR_LEAF_BASE + leaf_idx * 4); + u32 top_val = gsp_emu_shadow_read32(ctx, + GSP_EMU_REG_CPU_INTR_TOP_BASE + top_idx * 4); + + gsp_emu_shadow_write32(ctx, + GSP_EMU_REG_CPU_INTR_LEAF_BASE + leaf_idx * 4, + leaf_val | (1U << leaf_bit)); + gsp_emu_shadow_write32(ctx, + GSP_EMU_REG_CPU_INTR_TOP_BASE + top_idx * 4, + top_val | (1U << top_bit)); + gsp_emu_shadow_write32(ctx, GSP_EMU_REG_CPU_INTR_LEAF_TRIGGER, 0); + + nv_vgpu_info(ctx->vgpu, "GSP-BOOT: SW interrupt vector %u -> LEAF[%u]|=0x%x TOP[%u]|=0x%x, MSI-X", + sw_vector, leaf_idx, 1U << leaf_bit, top_idx, + 1U << top_bit); + gsp_emu_inject_completion_interrupt(ctx, 0); + } /* TLB invalidate auto-complete: the driver sets TRIGGER then spins * polling for it to clear (kgmmuCheckPendingInvalidates). Report @@ -267,16 +300,18 @@ static void gsp_emu_poll_boot(struct gsp_emu_context *ctx) * driver reads a stale/garbage error code (0xff) and the unload * fails, killing the subsequent re-init. A SEC2 kick after a * completed boot is a GSP shutdown/reload: drop the boot state - * (GSP no longer active, queue dead). */ + * (GSP no longer active, queue dead). Clear boot state FIRST so + * WPR2 is already down when the driver sees CPUCTL=HALTED and + * checks kgspIsWpr2Up (avoids "Booter Unload: WPR2 is still up"). */ if (gsp_cpuctl & GSP_EMU_CPUCTL_STARTCPU) gsp_emu_shadow_write32(ctx, GSP_EMU_REG_GSP_CPUCTL, GSP_EMU_CPUCTL_HALTED); if (sec2_cpuctl & GSP_EMU_CPUCTL_STARTCPU) { + if (ctx->boot_stage == GSP_BOOT_STAGE_BOOTER_DONE) + gsp_emu_clear_boot_state(ctx, "Booter unload/reload kick"); gsp_emu_shadow_write32(ctx, GSP_EMU_REG_SEC2_MAILBOX0, 0); gsp_emu_shadow_write32(ctx, GSP_EMU_REG_SEC2_CPUCTL, GSP_EMU_CPUCTL_HALTED); - if (ctx->boot_stage == GSP_BOOT_STAGE_BOOTER_DONE) - gsp_emu_clear_boot_state(ctx, "Booter unload/reload kick"); } } -- cgit v1.2.3