From: Dave Airlie Date: Tue, 17 Mar 2026 01:27:01 +0000 (+1000) Subject: Merge tag 'drm-intel-next-2026-03-16' of https://gitlab.freedesktop.org/drm/i915... X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=d93f8ea0e5ad41d661496d205fac3e2fbd9358c0;p=thirdparty%2Fkernel%2Flinux.git Merge tag 'drm-intel-next-2026-03-16' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-next [airlied: fixed conflict with xe tree] drm/i915 feature pull for v7.1: Features and functionality: - C10/C20/LT PHY PLL divider verification (Mika) - Use trans push mechanism to generate PSR frame change event on LNL+ (Jouni) - Account for DSC bubble overhead for horizontal slices (Ankit, Chaitanya) Refactoring and cleanups: - Refactor DP DSC slice config computation (Imre) - Use GVT versions of register helper macros for GVT MMIO table (Ankit) - C10/C20/LT PHY PLL computation refactoring (Mika) - VGA decode refactoring and related fixes/cleanups (Ville) - Move DSB buffer buffer implementation to display parent interface (Jani) - Move error interrupt capture to display irq snapshot (Jani) - Move pcode calls to display parent interface (Jani) - Reduce GVT dependency on display headers (Jani) - Compute config and mode valid refactoring for DSC (Ankit) - Stop using i915 core register headers in display (Uma) - Refactor DPT, move i915 parts to display parent interface (Jani) - Refactor gen2-4 overlay, move to display parent interface (Ville) - Refactor masked field register macro helpers, move to shared headers (Jani) - Convert a number of workaround checks to the new workaround framework (Luca) - Refactor and move frontbuffer calls to display parent interface (Jani) - Add VMA calls to display parent interface (Jani) - Refactor stolen memory allocation decisions (Vinod, Ville) - Clean up and unify workqueue usage (Marco Crivellari) - Preparation for UHBR DP tunnels (Imre) - Allow DSC passthrough modes during DP MST mode validation (Imre) - Move framebuffer bo interface to display parent interface (Jani) Fixes: - Plenty of DP SST HPD IRQ handling fixes (Imre) - DP AUX backlight and luminance control fixes (Suraj) - Respect VBT pipe joiner disable for eDP (Ankit) - Do not use CASF with joiner (Nemesa) - Clear C10/C20 PHY response read and error bit to avoid PHY hangs (Suraj) - Xe3p_LPD DMG clock gating, CDCLK, port sync workarounds (Suraj, Gustavo, Mitul) - Fix GVT error path (Michał) - Handle errors on DP DSC receiver cap reads (Suraj) - DSS clock gating workaround on MTL+ to avoid DSC corruption (Mika) - Skip state verification for LT PHY in TBT mode (Suraj) - Fix NULL pointer dereference on suspend when uc firmware not loaded (Rahul Bukte) - Fix an unlikely DMC state related NULL pointer dereference at probe (Imre) - Handle error returns from vga_get_uninterruptible() (Simon Richter) - Increase C10/C20/LT PHY timeouts to include SOC/OS turnaround (Arun) - Fix BIOS FB vs. stolen memory size check (Ville) - Fix LOBF to use computed guardband and set context latency (Ankit) - Handle modeset WW mutex lock failures due to contention properly (Imre) - Fix pipe BPP clamping due to HDR (Imre) - Fix stale state usage in DSC state computation (Imre) - Take HDCP 1.4 vs 2.x into account during link check (Suraj) - Fix forced link retrain handling in MST HPD IRQ handler (Imre) - Remove redundant warning on vcpi < 0 (Jonathan) Core changes: - iopoll: fix function parameter names in read_poll_timeout_atomic() (Randy Dunlap) Merges: - Backmerge drm-next for v7.0-rc1 (Jani) Signed-off-by: Dave Airlie From: Jani Nikula Link: https://patch.msgid.link/b14bb0f297b1750816cf5f342bde608e435655fa@intel.com --- d93f8ea0e5ad41d661496d205fac3e2fbd9358c0 diff --cc drivers/gpu/drm/xe/xe_lrc.c index 73a503d88217e,1e4357c6ed606..aa26c71ae34f5 --- a/drivers/gpu/drm/xe/xe_lrc.c +++ b/drivers/gpu/drm/xe/xe_lrc.c @@@ -1548,13 -1575,6 +1548,13 @@@ static int xe_lrc_ctx_init(struct xe_lr arb_enable = MI_ARB_ON_OFF | MI_ARB_ENABLE; xe_lrc_write_ring(lrc, &arb_enable, sizeof(arb_enable)); + if (init_flags & XE_LRC_DISABLE_STATE_CACHE_PERF_FIX) { + state_cache_perf_fix[0] = MI_LOAD_REGISTER_IMM | MI_LRI_NUM_REGS(1); + state_cache_perf_fix[1] = COMMON_SLICE_CHICKEN3.addr; - state_cache_perf_fix[2] = _MASKED_BIT_ENABLE(DISABLE_STATE_CACHE_PERF_FIX); ++ state_cache_perf_fix[2] = REG_MASKED_FIELD_ENABLE(DISABLE_STATE_CACHE_PERF_FIX); + xe_lrc_write_ring(lrc, state_cache_perf_fix, sizeof(state_cache_perf_fix)); + } + map = __xe_lrc_seqno_map(lrc); xe_map_write32(lrc_to_xe(lrc), &map, lrc->fence_ctx.next_seqno - 1);