- 08 May, 2015 40 commits
-
-
Matt Roper authored
With the recent modeset internal rework, we wind up setting crtc_state->enable to false, but leave crtc_state->active as true, which is incorrect. This mismatch gets caught by drm_atomic_crtc_check() and causes subsequent atomic operations (such as plane updates while the CRTC is disabled) to fail. Bisect points to commit dad9a7d6d96630182fb52aae7c3856e9e7285e13 Author: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com> Date: Tue Apr 21 17:13:19 2015 +0300 drm/i915: Use atomic helpers for computing changed flags as the commit that actually triggers the regression. v2: Update to alter in-flight state rather than already-committed state (first version was accidentally based on a midpoint of Ander's modeset rework series, before his final patches that add proper state swapping to the legacy modeset path). Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Testcase: igt/kms_universal_plane Signed-off-by:Matt Roper <matthew.d.roper@intel.com> Reviewed-by:
Ander Conselvan de Oliveira <conselvan2@gmail.com> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch>
-
Damien Lespiau authored
A part of this function was indented with 2 tabs and 1 space instead of just 2 tabs. We're going to touch that code, so start by re-indenting it. Signed-off-by:
Damien Lespiau <damien.lespiau@intel.com> Signed-off-by:
-
Joonas Lahtinen authored
Use partial view for huge BOs (bigger than half the mappable aperture) in fault handler so that they can be accessed withough trying to make room for them by evicting other objects. v2: - Only use partial views in the case where early rejection was previously done. - Account variable type changes from previous reroll. v3: - Add a comment about overwriting existing page entries. (Tvrtko Ursulin) - Whitespace fixes. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Signed-off-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Signed-off-by:
-
Joonas Lahtinen authored
Partial view type allows manipulating parts of huge BOs through the GGTT, which was not previously possible due to constraint that whole object had to be mapped for any access to it through GGTT. v2: - Retain error value from sg_alloc_table (Tvrtko Ursulin) - Do not zero already zeroed variable (Tvrtko Ursulin) - Use more common variable types for page size/offset (Tvrtko Ursulin) v3: - Only compare additional view parameters when need to (Tvrtko Ursulin) v4: - Do zero out the variable that needs to be (bug introduced in v2). Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Signed-off-by:
-
Joonas Lahtinen authored
Do not skip special GGTT views when considering whether an object is pinned or not. Wrong behaviour was introduced in; commit ec7adb6e Author: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Date: Mon Mar 16 14:11:13 2015 +0200 drm/i915: Do not use ggtt_view with (aliasing) PPGTT Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Signed-off-by:
-
Joonas Lahtinen authored
GGTT VMA sizes might be smaller than the whole object size due to different GGTT views. v2: - Separate GGTT view constraint calculations from normal view constraint calculations (Chris Wilson) v3: - Do not bother with debug wording. (Tvrtko Ursulin) v4: - Clearer logic for calculating map_and_fenceable (Tvrtko Ursulin) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
Nick Hoath <nicholas.hoath@intel.com> Reviewed-by:
Imre Deak <imre.deak@intel.com> Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Nick Hoath authored
Signed-off-by:
-
Chris Wilson authored
Reading a single value from the object, the locking only provides futile protection against userspace races. The locking is useless so remove it. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by:
-
Sonika Jindal authored
Renaming gen9_rates to skl_rates because other platforms may have different supported rates. Signed-off-by:
Sonika Jindal <sonika.jindal@intel.com> Reviewed-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
-
Ander Conselvan de Oliveira authored
Now that we do proper state swaps, we don't depend on this function anymore to keep the state in sync. Signed-off-by:
Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com> Reviewed-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Signed-off-by:
-
Ander Conselvan de Oliveira authored
Replace the commit output state function with a simple swap of states. Note that we still need to reconcile the legacy state after the swap, since there are still code that relies on those. Also note that even though changes to the state of a crtc different than the one passed as an argument to __intel_set_mode() will be saved, the modeset logic still deals with only one crtc. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Use lower level calls to better integrate with the modeset code and allow a full state swap in a follow up patch. Signed-off-by:
-
Ander Conselvan de Oliveira authored
When a new pipe_config is calculated, the fields related to shared dplls are reset, under the assumption that they will be recalculated as part of the modeset, which is true with the current state of the code. As we convert to atomic, however, it will be possible to calculate a new pipe_config and skip the modeset. In that case, after the state swap we still want the shared dplls to be preserved. Signed-off-by:
-
Ander Conselvan de Oliveira authored
To match the behavior of ->atomic_commit(). Signed-off-by:
-
Ander Conselvan de Oliveira authored
Replace the drivers own logic for computing mode_changed, active_changed and planes_changed flags with the check_modeset() atomic helper. Since that function needs to compare the crtc's new mode with the current, this patch also moves the set up of crtc_state->mode earlier in the call chain. Note that for the call to check_plane() to work properly, we need to check new plane state against new crtc state. But since we still use the plane update helper, which doesn't have a full atomic state, we need to hack around that in intel_plane_atomic_check(). Signed-off-by:
-
Ander Conselvan de Oliveira authored
So the i915 driver can use the same logic for setting mode and active changed flags, without having to implement encoder helpers and the mode_fixup() callback. Cc: dri-devel@lists.freedestkop.org Signed-off-by:
Dave Airlie <airlied@redhat.com> Signed-off-by:
-
Ander Conselvan de Oliveira authored
In a follow up patch the function that computes mode changes will be replaced with the one from the atomic helpers. To preserve the behavior of legacy modeset forcing DPMS on, that function will need to detect a change in the active state of the crtc, so that has to be kept up to date. Signed-off-by:
-
Ander Conselvan de Oliveira authored
This is no longer necessary since we only update the staged config on successfull modeset. The new configuration is stored in an atomic state struct which is freed in case of failure. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The logic that stages the state before the modeset was still updating first the old staged config and then populating the atomic state based on that. Change this to use only the atomic state. Note that now the staged config is updated in the function intel_modeset_commit_output_state(). This is done so that the modeset check and the force restore path in the hw state read out code continue to work. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Add a helper function to make the code slightly more readable. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Call intel_set_mode() uncondionally from intel_crtc_set_config(), since the former function is now properly wired to ignore all the modesets if the mode_changed and active_changed flags are false in crtc_state. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Use the atomic state instead. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Use the similar fields in crtc_state instead, so that this code can be moved to our future implementation of atomic_check(). Signed-off-by:
-
Ander Conselvan de Oliveira authored
We don't need to pass it down the call chain anymore now that the plane state is set up properly. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Add the primary plane state to the legacy modeset atomic state and use it when configuring the primary plane in __intel_set_mode(). This is a first step towards merging the flip path in intel_crtc_set_config() and __intel_set_mode(). v2: Set crtc to NULL if fb is NULL. (Maarten) Signed-off-by:
-
Ander Conselvan de Oliveira authored
The modeset code is now properly divided in two phases, so that it only changes hardware state if it succeeds, so there's no ill-effect that needs to be undone on failure anymore. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The remaining parts of the failure path could only be reached if the allocation of crtc_state_copy would fail. In that case, there is nothing to undo, so just get rid of the label for error handling and return an error code immediately. We also always allocate a pipe_config, even if the pipe is being disabled, so the remaining part of what was the error/done case can be simplified a little too. v2: Ignore return value from drm_plane_helper_update(). (Ander) Signed-off-by:
-
Ander Conselvan de Oliveira authored
The first function calls done in that function can still cause changes to the atomic state and may fail. This should eventually be part of our atomic check function, while the rest of the code in __intel_set_mode() is the commit hook. So this makes the legacy mode set more atomic-y. Signed-off-by:
-
Ander Conselvan de Oliveira authored
There's no way that function can fail after it sets crtc->mode anymore, so there's no need to save the old mode for the failure case. Signed-off-by:
-
Ander Conselvan de Oliveira authored
Set the mode_changed field on the crtc_states and use that instead. Note that even though this patch doesn't completely replace the logic in intel_modeset_affected_pipes(), that logic was never fully used to its full extent. Since the commit mentioned below, modeset_pipes and prepare_pipes would only contain at most the pipe for which the set_crtc ioctl was called. We can grow back that logic when the time comes. commit b6c5164d Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Fri Apr 12 18:48:43 2013 +0200 drm/i915: Fixup Oops in the pipe config computation v2: Don't set mode_changed unconditionally for modeset_crtc. (Ander) Check for needs_modeset() before trying to allocate a PLL. (Ander) Only call .crtc_enable() for pipes that were disabled. (Maarten) Signed-off-by:
-
Ander Conselvan de Oliveira authored
With the current implementation of intel_modeset_affected_pipes(), if a pipe will be enabled then it is in modeset_pipes. We'll remove that mask in a follow up patch, but want to preserve this behavior, so just make that explicit. Signed-off-by:
-
Ander Conselvan de Oliveira authored
The code in intel_modeset_pipe_config() still needs changes before it can calculate more than just one pipe_config, and pretending it can will only make those changes more difficult. Signed-off-by:
-
