[Intel-gfx] [PATCH v3 00/22] drm/i915: Convert to atomic, part 2.

[Matt Roper]

On Wed, May 20, 2015 at 03:38:05PM +0200, Maarten Lankhorst wrote:
> The goal of this patch series is to implement hardware readout using
> atomic state, and restore sw state with a single call to intel_set_mode.
> 
> After that's done intel_crtc_control can be safely converted to
> atomic modeset, because nothing relies on transitional state any
> more.
> 
> This patch series slightly reduces some user of crtc->config and
> crtc->active, but doesn't remove it completely.

I've reviewed all the patches here and responded to a few of the
messages; the ones I didn't explicitly respond to looked good and can be
considered to have my r-b tag.

Recap of the patches that I did reply directly to:
 * #8 causes a frontbuffer tracking warning as I mentioned, but that
   warning does go away a on a subsequent patch (I forget exactly which
   one now).  It would be nice to eliminate the warning here for bisect
   purposes.
 * #12 was giving me 'vblank not available on crtc #' warnings, but
   Maarten already identified a fix for that in a different
   email thread; with that fix applied, this patch has my r-b too.
 * #13 just needs some more details in the commit message.
 * #14 is just a cosmetic suggestion; feel free to ignore it if you
   disagree.
 * #15 is just a clarification question; I think we're still behaving
   the same way the original code did, so nothing that would block us
   merging this as-is.
 * #16 might warrant a follow-up patch to actually handle error
   return codes.
 * #20 is just a note about an outdated detail of the commit message.

Anything I didn't explicitly respond to is

        Reviewed-by: Matt Roper <matthew.d.roper at intel.com>

The comments on #13-16 and #20 are pretty minor and non-blocking, so if
you disagree with the suggestions or if we're in a rush to get these
merged, you can still consider them to have my r-b tag.


Matt

> 
> Ander Conselvan de Oliveira (6):
>   drm/i915: Make __intel_set_mode() take only atomic state as argument
>   drm/i915: Set mode_changed for audio in intel_modeset_pipe_config()
>   drm/i915: Support modeset across multiple pipes
>   drm/i915: Use global atomic state for staged pll config, v2.
>   drm/i915: Move cdclk and pll setup to intel_modeset_compute_config()
>   drm/i915: Read hw state into an atomic state struct
> 
> Maarten Lankhorst (16):
>   drm/i915: get rid of put_shared_dpll
>   drm/i915: get rid of intel_crtc_disable and related code, v2
>   drm/i915: use intel_crtc_control everywhere, v2.
>   drm/i915: Use drm_atomic_helper_update_legacy_modeset_state, v2.
>   drm/i915: Use crtc_state->active instead of crtc_state->enable
>   drm/i915: Zap call to drm_plane_helper_disable.
>   drm/i915: calculate primary visibility changes instead of calling from
>     set_config
>   drm/i915: Use drm_atomic_helper_swap_state in intel_atomic_commit.
>   drm/i915: Swap planes on each crtc separately.
>   drm/i915: Implement intel_crtc_control using atomic state, v3
>   drm/i915: move swap state to the right place
>   drm/i915: Use crtc->hwmode for vblanks, v2.
>   drm/i915: Remove use of crtc->config from i915_debugfs.c
>   drm/i915: Calculate haswell plane workaround, v3.
>   drm/i915: Use atomic state for calculating DVO_2X_MODE on i830.
>   drm/i915: use calculated state for vblank evasion
> 
>  drivers/gpu/drm/i915/i915_debugfs.c  |   50 +-
>  drivers/gpu/drm/i915/i915_drv.h      |    2 -
>  drivers/gpu/drm/i915/i915_irq.c      |   13 +-
>  drivers/gpu/drm/i915/intel_atomic.c  |   96 +--
>  drivers/gpu/drm/i915/intel_display.c | 1154 ++++++++++++++++------------------
>  drivers/gpu/drm/i915/intel_drv.h     |   32 +-
>  6 files changed, 664 insertions(+), 683 deletions(-)
> 
> -- 
> 2.1.0
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Matt Roper
Graphics Software Engineer
IoTG Platform Enabling & Development
Intel Corporation
(916) 356-2795