Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] drm/tegra: unlocked drm_bridge_get_current_state() call in tegra_rgb_encoder_enable() --
commit 743a0ce150c28976e990ee80b4eaf7239b83fcc3 Author: Svyatoslav Ryhel <[email protected]> drm/tegra: rgb: Parameterize configuration based on bus flags This commit updates tegra_rgb_encoder_enable() to read bus flags from the bridge state and configure pixel clock polarity, display enable polarity, and data alignment accordingly. > diff --git a/drivers/gpu/drm/tegra/rgb.c b/drivers/gpu/drm/tegra/rgb.c > index b6d09a8d8e5ea..da9e37fe36689 100644 > --- a/drivers/gpu/drm/tegra/rgb.c > +++ b/drivers/gpu/drm/tegra/rgb.c [ ... ] > @@ -104,14 +105,24 @@ static void tegra_rgb_encoder_enable(struct drm_encoder > *encoder) > struct drm_display_mode *mode = &encoder->crtc->state->adjusted_mode; > struct tegra_output *output = encoder_to_output(encoder); > struct tegra_rgb *rgb = to_rgb(output); > - u32 value; > + struct drm_bridge_state *bridge_state; > + u32 bus_flags, value; > + > + /* Get but flags from the bridge state. */ > + bridge_state = drm_bridge_get_current_state(output->bridge); > + if (bridge_state) > + bus_flags = bridge_state->input_bus_cfg.flags; > + else > + bus_flags = DRM_BUS_FLAG_PIXDATA_SAMPLE_NEGEDGE | > + DRM_BUS_FLAG_DE_HIGH | > + DRM_BUS_FLAG_DATA_MSB_TO_LSB; [Severity: High] Is it safe to call drm_bridge_get_current_state() from tegra_rgb_encoder_enable() during a non-blocking atomic commit? In include/drm/drm_bridge.h, drm_bridge_get_current_state() asserts that bridge->base.lock is held before reading bridge->base.state: static inline struct drm_bridge_state * drm_bridge_get_current_state(struct drm_bridge *bridge) { if (!bridge) return NULL; drm_modeset_lock_assert_held(&bridge->base.lock); if (!bridge->base.state) return NULL; return drm_priv_to_bridge_state(bridge->base.state); } During a non-blocking atomic modeset commit in drivers/gpu/drm/drm_atomic_helper.c, drm_atomic_helper_commit() queues state->commit_work to run commit_tail() asynchronously on system_dfl_wq, and drm_mode_atomic_ioctl() drops all modeset locks before the worker runs: drm_mode_atomic_ioctl() -> drm_atomic_nonblocking_commit() -> drm_atomic_helper_commit() queue_work(system_dfl_wq, &state->commit_work); drm_modeset_drop_locks() When the worker executes commit_tail() -> tegra_atomic_commit_tail() -> drm_atomic_helper_commit_tail_rpm() -> drm_atomic_helper_commit_modeset_enables() -> drm_atomic_helper_commit_encoder_bridge_enable() -> tegra_rgb_encoder_enable(), no drm_modeset_lock is held. Calling drm_bridge_get_current_state(output->bridge) here will trigger the drm_modeset_lock_assert_held(&bridge->base.lock) lockdep warning and perform an unlocked read of bridge->base.state that can race with concurrent atomic commits swapping or freeing the state. In addition, if an attached bridge does not implement .atomic_duplicate_state, drm_atomic_add_encoder_bridges() skips acquiring bridge->base.lock even during blocking modeset commits, while drm_bridge_attach() unconditionally initializes bridge->base.lock via drm_atomic_private_obj_init(). Would it be better to switch tegra_rgb_encoder_helper_funcs to use the .atomic_enable callback and look up the state from the drm_atomic_state via drm_atomic_get_new_bridge_state(state, output->bridge)? -- Sashiko AI review ยท https://sashiko.dev/#/patchset/[email protected]?part=6
