When big joiner is enabled, it reserves the adjacent pipe as the secondary pipe. This happens without the user space knowing, and subsequent attempts at using the CRTC with that pipe will fail. If the user space does not have a coping mechanism, i.e. trying another CRTC, this leads to a black screen.
Try to reduce the impact of the problem on discrete platforms by mapping the CRTCs to pipes in order A, C, B, and D. If the user space reserves CRTCs in order, this should trick it to using pipes that are more likely to be available for and after joining. Limit this to discrete platforms, which have four pipes, and no eDP, a combination that should benefit the most with least drawbacks. Signed-off-by: Jani Nikula <[email protected]> --- v2: Also remove WARN_ON() v3: Limit to discrete v4: Revamp There are a number of issues in IGT with assuming CRTC index == pipe, at least with CRC and vblank waits. With them being used a lot in tests, we won't get enough test coverage until they're fixed. --- drivers/gpu/drm/i915/display/intel_crtc.c | 29 ++++++++++++++++++-- drivers/gpu/drm/i915/display/intel_display.c | 2 ++ 2 files changed, 29 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c b/drivers/gpu/drm/i915/display/intel_crtc.c index ed3c6c4ce025..c06b06cb1db7 100644 --- a/drivers/gpu/drm/i915/display/intel_crtc.c +++ b/drivers/gpu/drm/i915/display/intel_crtc.c @@ -393,8 +393,6 @@ static int __intel_crtc_init(struct intel_display *display, enum pipe pipe) cpu_latency_qos_add_request(&crtc->vblank_pm_qos, PM_QOS_DEFAULT_VALUE); - drm_WARN_ON(display->drm, drm_crtc_index(&crtc->base) != crtc->pipe); - if (HAS_CASF(display) && crtc->num_scalers >= 2) drm_crtc_create_sharpness_strength_property(&crtc->base); @@ -406,6 +404,31 @@ static int __intel_crtc_init(struct intel_display *display, enum pipe pipe) return ret; } +#define HAS_PIPE(display, pipe) (DISPLAY_RUNTIME_INFO(display)->pipe_mask & BIT(pipe)) + +/* + * Expose the pipes in order A, C, B, D on discrete platforms to trick user + * space into using pipes that are more likely to be available for both a) user + * space if pipe B has been reserved for the joiner, and b) the joiner if pipe A + * doesn't need the joiner. + * + * Swap pipes B and C only if both are available i.e. not fused off. + */ +static enum pipe reorder_pipe(struct intel_display *display, enum pipe pipe) +{ + if (!display->platform.dgfx || !HAS_PIPE(display, PIPE_B) || !HAS_PIPE(display, PIPE_C)) + return pipe; + + switch (pipe) { + case PIPE_B: + return PIPE_C; + case PIPE_C: + return PIPE_B; + default: + return pipe; + } +} + int intel_crtc_init(struct intel_display *display) { enum pipe pipe; @@ -415,6 +438,8 @@ int intel_crtc_init(struct intel_display *display) INTEL_NUM_PIPES(display), str_plural(INTEL_NUM_PIPES(display))); for_each_pipe(display, pipe) { + pipe = reorder_pipe(display, pipe); + ret = __intel_crtc_init(display, pipe); if (ret) return ret; diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index 295f14416be7..2a7a9c5639ee 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -5959,6 +5959,8 @@ static int intel_atomic_check_joiner(struct intel_atomic_state *state, * This works because the crtcs are created in pipe order, * and the hardware requires primary pipe < secondary pipe as well. * Should that change we need to rethink the logic. + * + * FIXME: What about with reordered pipes? */ if (WARN_ON(drm_crtc_index(&primary_crtc->base) > drm_crtc_index(&secondary_crtc->base))) -- 2.47.3
