public inbox for intel-gfx@lists.freedesktop.org
 help / color / mirror / Atom feed
From: "Hogander, Jouni" <jouni.hogander@intel.com>
To: "Kandpal, Suraj" <suraj.kandpal@intel.com>,
	"intel-gfx@lists.freedesktop.org"
	<intel-gfx@lists.freedesktop.org>
Cc: "Shankar, Uma" <uma.shankar@intel.com>
Subject: Re: [PATCH] drm/i915/psr: Implement WA to help reach PC10
Date: Fri, 20 Sep 2024 11:45:54 +0000	[thread overview]
Message-ID: <adae5b5d1afde8d8aed7ec0662a7682cc8ec6aae.camel@intel.com> (raw)
In-Reply-To: <20240920091203.1043308-2-suraj.kandpal@intel.com>

On Fri, 2024-09-20 at 14:42 +0530, Suraj Kandpal wrote:
> To reach PC10 when PKG_C_LATENCY is configure we must do the
> following
> things
> 1) Enter PSR1 only when delayed_vblank < 6 lines and DC5 can be
> entered
> 2) Allow PSR2 deep sleep when DC5 can be entered
> 3) DC5 can be entered when all transocoder have either PSR1, PSR2 or
> eDP 1.5 PR ALPM enabled and VBI is disabled and flips and pushes are
> not happening.

One comment below related to PantherLake. Otherwise patch looks ok to
me.


> 
> --v2
> -Switch condition and do an early return [Jani]
> -Do some checks in compute_config [Jani]
> -Do not use register reads as a method of checking states for
> DPKGC or delayed vblank [Jani]
> -Use another way to see is vblank interrupts are disabled or not
> [Jani]
> 
> --v3
> -Use has_psr to check if psr can be enabled or not for dc5_entry cond
> [Uma]
> -Move the dc5 entry computation to psr_compute_config [Jouni]
> -No need to change sequence of enabled and activate,
> so dont make hsw_psr1_activate return anything [Jouni]
> -Use has_psr to stop psr1 activation [Jouni]
> -Use lineage no. in WA
> -Add the display ver restrictions for WA
> 
> --v4
> -use more appropriate name for check_vblank_limit() [Jouni]
> -Cover the case for idle frames when dpkgc is not configured [Jouni]
> -Check psr only for edp [Jouni]
> 
> --v5
> -move psr1 handling to plane update [Jouni]
> -add todo for cases when vblank is enabled when psr enabled [Jouni]
> -use intel_display instead of drm_i915_private
> 
> --v6
> -check target_dc_state [Jouni]
> -fix condition in pre/post plane update [Jouni]
> 
> --v7
> -fix has_psr condition [Uma]
> -fix typo in commit subject [Uma]
> -put psr1_wa check in its own helper [Uma]
> -fix the dc_entry check [Jouni]
> -use HAS_PSR() to cover two edp one with psr and one nonpsr [Jouni]
> 
> WA: 22019444797
> Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
> ---
>  .../drm/i915/display/intel_display_types.h    |   3 +
>  drivers/gpu/drm/i915/display/intel_psr.c      | 119
> +++++++++++++++++-
>  2 files changed, 121 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h
> b/drivers/gpu/drm/i915/display/intel_display_types.h
> index 3e694c1204db..2d790abee76e 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> @@ -1577,6 +1577,9 @@ struct intel_psr {
>  #define I915_PSR_DEBUG_PANEL_REPLAY_DISABLE    0x40
>  
>         u32 debug;
> +       bool is_dpkgc_configured;
> +       bool is_dc5_entry_possible;
> +       bool is_wa_delayed_vblank_limit;
>         bool sink_support;
>         bool source_support;
>         bool enabled;
> diff --git a/drivers/gpu/drm/i915/display/intel_psr.c
> b/drivers/gpu/drm/i915/display/intel_psr.c
> index 5b355d0a3565..b882ff25fb92 100644
> --- a/drivers/gpu/drm/i915/display/intel_psr.c
> +++ b/drivers/gpu/drm/i915/display/intel_psr.c
> @@ -26,6 +26,7 @@
>  #include <drm/drm_atomic_helper.h>
>  #include <drm/drm_damage_helper.h>
>  #include <drm/drm_debugfs.h>
> +#include <drm/drm_vblank.h>
>  
>  #include "i915_drv.h"
>  #include "i915_reg.h"
> @@ -895,6 +896,89 @@ static u8 psr_compute_idle_frames(struct
> intel_dp *intel_dp)
>         return idle_frames;
>  }
>  
> +static bool
> +intel_psr_check_wa_delayed_vblank(const struct drm_display_mode
> *adjusted_mode)
> +{
> +       return (adjusted_mode->crtc_vblank_start - adjusted_mode-
> >crtc_vdisplay) >= 6;
> +}
> +
> +/*
> + * PKG_C_LATENCY is configured only when DISPLAY_VER >= 20 and
> + * VRR is not enabled
> + */
> +static bool intel_psr_is_dpkgc_configured(struct intel_display
> *display,
> +                                         struct intel_atomic_state
> *state)
> +{
> +       struct intel_crtc *intel_crtc;
> +       struct intel_crtc_state *crtc_state;
> +       int i;
> +
> +       if (DISPLAY_VER(display) < 20)
> +               return false;
> +
> +       for_each_new_intel_crtc_in_state(state, intel_crtc,
> crtc_state, i) {
> +               if (!intel_crtc->active)
> +                       continue;
> +
> +               if (crtc_state->vrr.enable)
> +                       return false;
> +       }
> +
> +       return true;
> +}
> +
> +static bool wa_22019444797_psr1_check(const struct intel_crtc_state
> *crtc_state,
> +                                     struct intel_psr *psr)
> +{
> +       struct intel_display *display = to_intel_display(crtc_state);
> +
> +       return DISPLAY_VER(display) == 20 && psr->is_dpkgc_configured
> &&
> +               (psr->is_wa_delayed_vblank_limit || !psr-
> >is_dc5_entry_possible) &&
> +               !crtc_state->has_sel_update && !crtc_state-
> >has_panel_replay;
> +}
> +
> +/*
> + * DC5 entry is only possible if vblank interrupt is disabled
> + * and either psr1, psr2, edp 1.5 pr alpm is enabled on all
> + * enabled encoders.
> + */
> +static bool
> +intel_psr_is_dc5_entry_possible(struct intel_display *display,
> +                               struct intel_atomic_state *state)
> +{
> +       struct intel_crtc *intel_crtc;
> +       struct intel_crtc_state *crtc_state;
> +       int i;
> +
> +       if ((display->power.domains.target_dc_state &
> +            DC_STATE_EN_UPTO_DC5_DC6_MASK) == 0)
> +               return false;
> +
> +       for_each_new_intel_crtc_in_state(state, intel_crtc,
> crtc_state, i) {
> +               struct drm_crtc *crtc = &intel_crtc->base;
> +               struct drm_vblank_crtc *vblank;
> +               struct intel_encoder *encoder;
> +
> +               if (!intel_crtc->active)
> +                       continue;
> +
> +               vblank = drm_crtc_vblank_crtc(crtc);
> +
> +               if (vblank->enabled)
> +                       return false;
> +
> +               if (!crtc_state->has_psr)
> +                       return false;
> +
> +               for_each_encoder_on_crtc(display->drm, crtc, encoder)
> +                       if (encoder->type != INTEL_OUTPUT_EDP ||
> +                           !CAN_PSR(enc_to_intel_dp(encoder)))
> +                               return false;
> +       }
> +
> +       return true;
> +}
> +
>  static void hsw_activate_psr1(struct intel_dp *intel_dp)
>  {
>         struct intel_display *display = to_intel_display(intel_dp);
> @@ -1007,7 +1091,15 @@ static void hsw_activate_psr2(struct intel_dp
> *intel_dp)
>         u32 val = EDP_PSR2_ENABLE;
>         u32 psr_val = 0;
>  
> -       val |=
> EDP_PSR2_IDLE_FRAMES(psr_compute_idle_frames(intel_dp));
> +       /*
> +        * Wa_22019444797
> +        * TODO: Disable idle frames when vblank gets enabled while
> +        * PSR2 is enabled
> +        */
> +       if (DISPLAY_VER(dev_priv) != 20 ||


I think same Workaround is needed by PantherLake but only if stepping
is A0. You could have this here:

(DISPLAY_VER(display) != 20 && !IS_DISPLAY_VER_STEP(display, IP_VER(30,
0), STEP_A0, STEP_B0))

BR,

Jouni Högander


> +           !intel_dp->psr.is_dpkgc_configured ||
> +           intel_dp->psr.is_dc5_entry_possible)
> +               val |=
> EDP_PSR2_IDLE_FRAMES(psr_compute_idle_frames(intel_dp));
>  
>         if (DISPLAY_VER(display) < 14 && !IS_ALDERLAKE_P(dev_priv))
>                 val |= EDP_SU_TRACK_ENABLE;
> @@ -2692,10 +2784,20 @@ void intel_psr_pre_plane_update(struct
> intel_atomic_state *state,
>         const struct intel_crtc_state *new_crtc_state =
>                 intel_atomic_get_new_crtc_state(state, crtc);
>         struct intel_encoder *encoder;
> +       bool dpkgc_configured = false, dc5_entry_possible = false;
> +       bool wa_delayed_vblank_limit = false;
>  
>         if (!HAS_PSR(display))
>                 return;
>  
> +       if (DISPLAY_VER(display) == 20) {
> +               dpkgc_configured =
> intel_psr_is_dpkgc_configured(display, state);
> +               dc5_entry_possible =
> +                       intel_psr_is_dc5_entry_possible(display,
> state);
> +               wa_delayed_vblank_limit =
> +                       intel_psr_check_wa_delayed_vblank(&new_crtc_s
> tate->hw.adjusted_mode);
> +       }
> +
>         for_each_intel_encoder_mask_with_psr(state->base.dev,
> encoder,
>                                              old_crtc_state-
> >uapi.encoder_mask) {
>                 struct intel_dp *intel_dp = enc_to_intel_dp(encoder);
> @@ -2704,6 +2806,12 @@ void intel_psr_pre_plane_update(struct
> intel_atomic_state *state,
>  
>                 mutex_lock(&psr->lock);
>  
> +               if (DISPLAY_VER(i915) == 20) {
> +                       psr->is_dpkgc_configured = dpkgc_configured;
> +                       psr->is_dc5_entry_possible =
> dc5_entry_possible;
> +                       psr->is_wa_delayed_vblank_limit =
> wa_delayed_vblank_limit;
> +               }
> +
>                 /*
>                  * Reasons to disable:
>                  * - PSR disabled in new state
> @@ -2711,6 +2819,7 @@ void intel_psr_pre_plane_update(struct
> intel_atomic_state *state,
>                  * - Changing between PSR versions
>                  * - Region Early Transport changing
>                  * - Display WA #1136: skl, bxt
> +                * - Display WA_22019444797
>                  */
>                 needs_to_disable |=
> intel_crtc_needs_modeset(new_crtc_state);
>                 needs_to_disable |= !new_crtc_state->has_psr;
> @@ -2720,6 +2829,8 @@ void intel_psr_pre_plane_update(struct
> intel_atomic_state *state,
>                         psr->su_region_et_enabled;
>                 needs_to_disable |= DISPLAY_VER(i915) < 11 &&
>                         new_crtc_state->wm_level_disabled;
> +               /* TODO: Disable PSR1 when vblank gets enabled while
> PSR1 is enabled */
> +               needs_to_disable |=
> wa_22019444797_psr1_check(new_crtc_state, psr);
>  
>                 if (psr->enabled && needs_to_disable)
>                         intel_psr_disable_locked(intel_dp);
> @@ -2760,6 +2871,12 @@ void intel_psr_post_plane_update(struct
> intel_atomic_state *state,
>                 keep_disabled |= DISPLAY_VER(display) < 11 &&
>                         crtc_state->wm_level_disabled;
>  
> +               /*
> +                * Wa_22019444797
> +                * TODO: Disable PSR1 when vblank gets enabled while
> PSR1 is enabled
> +                */
> +               keep_disabled |=
> wa_22019444797_psr1_check(crtc_state, psr);
> +
>                 if (!psr->enabled && !keep_disabled)
>                         intel_psr_enable_locked(intel_dp,
> crtc_state);
>                 else if (psr->enabled && !crtc_state-
> >wm_level_disabled)


  reply	other threads:[~2024-09-20 11:46 UTC|newest]

Thread overview: 47+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-02  5:02 [PATCH] drm/i915/psr: Implment WA to help reach PC10 Suraj Kandpal
2024-09-02  6:13 ` ✗ Fi.CI.BAT: failure for " Patchwork
2024-09-02  9:37 ` [PATCH] " Hogander, Jouni
2024-09-02 10:01   ` Kandpal, Suraj
2024-09-03  7:42     ` Hogander, Jouni
2024-09-03  7:46       ` Kandpal, Suraj
2024-09-02 10:02   ` Hogander, Jouni
2024-09-02 10:14     ` Kandpal, Suraj
2024-09-02 11:02       ` Hogander, Jouni
2024-09-02 11:07         ` Kandpal, Suraj
2024-09-02 11:13           ` Hogander, Jouni
2024-09-02 11:16             ` Kandpal, Suraj
2024-09-03  8:24 ` Suraj Kandpal
2024-09-05  4:41   ` Suraj Kandpal
2024-09-09  4:54     ` Suraj Kandpal
2024-09-09  6:32       ` Suraj Kandpal
2024-09-19 12:14         ` Shankar, Uma
2024-09-19 12:44           ` Kandpal, Suraj
2024-09-19 12:55             ` Hogander, Jouni
2024-09-20  5:36           ` Hogander, Jouni
2024-09-20  6:29             ` Kandpal, Suraj
2024-09-20  6:41               ` Hogander, Jouni
2024-09-20  6:46                 ` Kandpal, Suraj
2024-09-20  9:12         ` [PATCH] drm/i915/psr: Implement " Suraj Kandpal
2024-09-20 11:45           ` Hogander, Jouni [this message]
2024-09-23  2:54             ` Kandpal, Suraj
2024-09-23 10:23               ` Shankar, Uma
2024-09-09 14:12   ` [PATCH] drm/i915/psr: Implment " Ville Syrjälä
2024-09-10  4:15     ` Kandpal, Suraj
2024-09-03 10:25 ` ✓ Fi.CI.BAT: success for drm/i915/psr: Implment WA to help reach PC10 (rev2) Patchwork
2024-09-04 13:03 ` ✓ Fi.CI.IGT: " Patchwork
2024-09-05  5:18 ` ✓ Fi.CI.BAT: success for drm/i915/psr: Implment WA to help reach PC10 (rev3) Patchwork
2024-09-06 15:26 ` ✓ Fi.CI.IGT: " Patchwork
2024-09-09  5:26 ` ✗ Fi.CI.BUILD: failure for drm/i915/psr: Implment WA to help reach PC10 (rev4) Patchwork
2024-09-09  7:28 ` ✓ Fi.CI.BAT: success for drm/i915/psr: Implment WA to help reach PC10 (rev5) Patchwork
2024-09-10 17:15 ` ✗ Fi.CI.IGT: failure " Patchwork
2024-09-20 10:57 ` ✓ Fi.CI.BAT: success for drm/i915/psr: Implment WA to help reach PC10 (rev6) Patchwork
2024-09-21 11:54 ` ✗ Fi.CI.IGT: failure " Patchwork
2024-09-27 18:54 ` ✓ Fi.CI.IGT: success " Patchwork
2024-09-28 16:12 ` Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2024-10-03 14:53 [PATCH] drm/i915/psr: Implement WA to help reach PC10 Suraj Kandpal
2024-10-06 16:53 ` Suraj Kandpal
2024-10-17  7:54 ` Suraj Kandpal
2024-11-04  9:12 Suraj Kandpal
2024-11-04 14:18 ` Jani Nikula
2024-11-06  8:18   ` Kandpal, Suraj
2024-11-06  9:00 Suraj Kandpal

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=adae5b5d1afde8d8aed7ec0662a7682cc8ec6aae.camel@intel.com \
    --to=jouni.hogander@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=suraj.kandpal@intel.com \
    --cc=uma.shankar@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox