From: Sean Paul <sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>
To: Jeykumar Sankaran <jsanka-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
Cc: Sean Paul <sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>,
dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org,
robdclark-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org,
Sean Paul <seanpaul-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org>,
linux-arm-msm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
freedreno-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org
Subject: Re: [PATCH 6/8] drm/msm: dpu: Separate crtc assignment from vblank enable
Date: Wed, 14 Nov 2018 10:16:15 -0500 [thread overview]
Message-ID: <20181114151615.GO154160@art_vandelay> (raw)
In-Reply-To: <a8d1ef336746c930e8274153c82ebf99-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
On Tue, Nov 13, 2018 at 04:48:12PM -0800, Jeykumar Sankaran wrote:
> On 2018-11-13 12:52, Sean Paul wrote:
> > From: Sean Paul <seanpaul@chromium.org>
> >
> > Instead of assigning/clearing the crtc on vblank enable/disable, we can
> > just assign and clear the crtc on modeset. That allows us to just toggle
> > the encoder's vblank interrupts on vblank_enable.
> >
> > So why is this important? Previously the driver was using the legacy
> > pointers to assign/clear the crtc. Legacy pointers are cleared _after_
> Which pointers are you referring here as legacy pointers? CRTC?
encoder->crtc in this case. The loop in vblank_enable_no_lock relies on
enc->crtc == crtc
> > disabling the hardware, so the legacy pointer was valid during
> > vblank_disable, but that's not something we should rely on.
> >
> > Instead of relying on the core ordering the legacy pointer assignments
> > just so, we'll assign the crtc in dpu_crtc enable/disable. This is the
> > only place that mapping can change, so we're covered there.
> >
> > We're also taking advantage of drm_crtc_vblank_on/off. By using this, we
> > ensure that vblank_enable/disable can never be called while the crtc is
> > off (which means the assigned crtc will always be valid). As such, we
>
> What about the drm_vblank_enable/disable triggered by drm_vblank_get when
> crtc
> is disabled? What is the expected behavior there? the vblank_requested
> variable removed by the next patch was introduced to cache the request.
That case is covered by the modeset locks held when calling disable().
>
> > don't need to use modeset locks or the crtc_lock in the
> > vblank_enable/disable routine to be sure state is consistent.
> >
> > ...I think.
> >
> > Signed-off-by: Sean Paul <seanpaul@chromium.org>
> > ---
> > drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c | 77 +++++++++------------
> > drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 27 +++++---
> > drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h | 10 +++
> > 3 files changed, 59 insertions(+), 55 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > index 4b7f98a6ab60..59e823281fdf 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > @@ -757,43 +757,6 @@ void dpu_crtc_commit_kickoff(struct drm_crtc *crtc)
> > DPU_ATRACE_END("crtc_commit");
> > }
> >
> > -/**
> > - * _dpu_crtc_vblank_enable_no_lock - update power resource and vblank
> > request
> > - * @dpu_crtc: Pointer to dpu crtc structure
> > - * @enable: Whether to enable/disable vblanks
> > - */
> > -static void _dpu_crtc_vblank_enable_no_lock(
> > - struct dpu_crtc *dpu_crtc, bool enable)
> > -{
> > - struct drm_crtc *crtc = &dpu_crtc->base;
> > - struct drm_device *dev = crtc->dev;
> > - struct drm_encoder *enc;
> > -
> > - if (enable) {
> > - list_for_each_entry(enc, &dev->mode_config.encoder_list,
> > head) {
> > - if (enc->crtc != crtc)
> > - continue;
> > -
> > -
> > trace_dpu_crtc_vblank_enable(DRMID(&dpu_crtc->base),
> > - DRMID(enc), enable,
> > - dpu_crtc);
> > -
> > - dpu_encoder_assign_crtc(enc, crtc);
> > - }
> > - } else {
> > - list_for_each_entry(enc, &dev->mode_config.encoder_list,
> > head) {
> > - if (enc->crtc != crtc)
> > - continue;
> > -
> > -
> > trace_dpu_crtc_vblank_enable(DRMID(&dpu_crtc->base),
> > - DRMID(enc), enable,
> > - dpu_crtc);
> > -
> > - dpu_encoder_assign_crtc(enc, NULL);
> > - }
> > - }
> > -}
> > -
> > /**
> > * dpu_crtc_duplicate_state - state duplicate hook
> > * @crtc: Pointer to drm crtc structure
> > @@ -847,6 +810,10 @@ static void dpu_crtc_disable(struct drm_crtc *crtc,
> > /* Disable/save vblank irq handling */
> > drm_crtc_vblank_off(crtc);
> >
> > + drm_for_each_encoder_mask(encoder, crtc->dev,
> > + old_crtc_state->encoder_mask)
> > + dpu_encoder_assign_crtc(encoder, NULL);
> > +
> > mutex_lock(&dpu_crtc->crtc_lock);
> >
> > /* wait for frame_event_done completion */
> > @@ -856,9 +823,6 @@ static void dpu_crtc_disable(struct drm_crtc *crtc,
> > atomic_read(&dpu_crtc->frame_pending));
> >
> > trace_dpu_crtc_disable(DRMID(crtc), false, dpu_crtc);
> > - if (dpu_crtc->enabled && dpu_crtc->vblank_requested) {
> > - _dpu_crtc_vblank_enable_no_lock(dpu_crtc, false);
> > - }
> > dpu_crtc->enabled = false;
> >
> > if (atomic_read(&dpu_crtc->frame_pending)) {
> > @@ -922,13 +886,13 @@ static void dpu_crtc_enable(struct drm_crtc *crtc,
> >
> > mutex_lock(&dpu_crtc->crtc_lock);
> > trace_dpu_crtc_enable(DRMID(crtc), true, dpu_crtc);
> > - if (!dpu_crtc->enabled && dpu_crtc->vblank_requested) {
> > - _dpu_crtc_vblank_enable_no_lock(dpu_crtc, true);
> > - }
> > dpu_crtc->enabled = true;
> >
> > mutex_unlock(&dpu_crtc->crtc_lock);
> >
> > + drm_for_each_encoder_mask(encoder, crtc->dev,
> > crtc->state->encoder_mask)
> > + dpu_encoder_assign_crtc(encoder, crtc);
> > +
> > /* Enable/restore vblank irq handling */
> > drm_crtc_vblank_on(crtc);
> > }
> > @@ -1173,10 +1137,33 @@ static int dpu_crtc_atomic_check(struct drm_crtc
> > *crtc,
> > int dpu_crtc_vblank(struct drm_crtc *crtc, bool en)
> > {
> > struct dpu_crtc *dpu_crtc = to_dpu_crtc(crtc);
> > + struct drm_encoder *enc;
> >
> > - mutex_lock(&dpu_crtc->crtc_lock);
> > trace_dpu_crtc_vblank(DRMID(&dpu_crtc->base), en, dpu_crtc);
> > - _dpu_crtc_vblank_enable_no_lock(dpu_crtc, en);
> > +
> > + /*
> > + * Normally we would iterate through encoder_mask in crtc state to
> > find
> > + * attached encoders. In this case, we might be disabling vblank
> > _after_
> > + * encoder_mask has been cleared.
> > + *
> > + * Instead, we "assign" a crtc to the encoder in enable and clear
> > it in
> > + * disable (which is also after encoder_mask is cleared). So
> > instead of
> > + * using encoder mask, we'll ask the encoder to toggle itself iff
> > it's
> > + * currently assigned to our crtc.
> > + *
> > + * Note also that this function cannot be called while crtc is
> > disabled
> > + * since we use drm_crtc_vblank_on/off. So we don't need to worry
> > + * about the assigned crtcs being inconsistent with the current
> > state
> > + * (which means no need to worry about modeset locks).
> > + */
> > + list_for_each_entry(enc, &crtc->dev->mode_config.encoder_list,
> > head) {
> > + trace_dpu_crtc_vblank_enable(DRMID(crtc), DRMID(enc), en,
> > + dpu_crtc);
> > +
> > + dpu_encoder_toggle_vblank_for_crtc(enc, crtc, en);
> > + }
> > +
> > + mutex_lock(&dpu_crtc->crtc_lock);
> > dpu_crtc->vblank_requested = en;
> > mutex_unlock(&dpu_crtc->crtc_lock);
> >
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> > b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> > index fd6514f681ae..5914ae70572c 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> > @@ -1267,22 +1267,29 @@ void dpu_encoder_assign_crtc(struct drm_encoder
> > *drm_enc, struct drm_crtc *crtc)
> > {
> > struct dpu_encoder_virt *dpu_enc = to_dpu_encoder_virt(drm_enc);
> > unsigned long lock_flags;
> > - bool enable;
> > - int i;
> > -
> > - enable = crtc ? true : false;
> > -
> > - if (!drm_enc) {
> > - DPU_ERROR("invalid encoder\n");
> > - return;
> > - }
> > - trace_dpu_enc_vblank_cb(DRMID(drm_enc), enable);
> >
> > spin_lock_irqsave(&dpu_enc->enc_spinlock, lock_flags);
> > /* crtc should always be cleared before re-assigning */
> > WARN_ON(crtc && dpu_enc->crtc);
> > dpu_enc->crtc = crtc;
> > spin_unlock_irqrestore(&dpu_enc->enc_spinlock, lock_flags);
> > +}
> > +
> > +void dpu_encoder_toggle_vblank_for_crtc(struct drm_encoder *drm_enc,
> > + struct drm_crtc *crtc, bool
> > enable)
> > +{
> > + struct dpu_encoder_virt *dpu_enc = to_dpu_encoder_virt(drm_enc);
> > + unsigned long lock_flags;
> > + int i;
> > +
> > + trace_dpu_enc_vblank_cb(DRMID(drm_enc), enable);
> > +
> > + spin_lock_irqsave(&dpu_enc->enc_spinlock, lock_flags);
> > + if (dpu_enc->crtc == crtc) {
> > + spin_unlock_irqrestore(&dpu_enc->enc_spinlock,
> > lock_flags);
> > + return;
> > + }
> Why are you returning when the crtc's are same?
> Won't they be same all the time between modesets?
>
> for both enable/disable crtc will be the same and you still need to call the
> for loop below to enable/disable the phys encoders.
>
> > + spin_unlock_irqrestore(&dpu_enc->enc_spinlock, lock_flags);
> >
> > for (i = 0; i < dpu_enc->num_phys_encs; i++) {
> > struct dpu_encoder_phys *phys = dpu_enc->phys_encs[i];
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h
> > b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h
> > index be1d80867834..6896ea2ab854 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h
> > @@ -62,6 +62,16 @@ void dpu_encoder_get_hw_resources(struct drm_encoder
> > *encoder,
> > void dpu_encoder_assign_crtc(struct drm_encoder *encoder,
> > struct drm_crtc *crtc);
> >
> > +/**
> > + * dpu_encoder_toggle_vblank_for_crtc - Toggles vblank interrupts on or
> > off if
> > + * the encoder is assigned to the given crtc
> > + * @encoder: encoder pointer
> > + * @crtc: crtc pointer
> > + * @enable: true if vblank should be enabled
> > + */
> > +void dpu_encoder_toggle_vblank_for_crtc(struct drm_encoder *encoder,
> > + struct drm_crtc *crtc, bool
> > enable);
> > +
> > /**
> > * dpu_encoder_register_frame_event_callback - provide callback to
> > encoder that
> > * will be called after the request is complete, or other events.
>
> --
> Jeykumar S
--
Sean Paul, Software Engineer, Google / Chromium OS
_______________________________________________
Freedreno mailing list
Freedreno@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/freedreno
next prev parent reply other threads:[~2018-11-14 15:16 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-13 20:52 [PATCH 1/8] drm/msm: dpu: Move pm_runtime_(get|put) from vblank_enable Sean Paul
2018-11-13 20:52 ` [PATCH 3/8] drm/msm: dpu: Remove vblank_callback from encoder Sean Paul
[not found] ` <20181113205257.170707-3-sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>
2018-11-14 0:47 ` Jeykumar Sankaran
[not found] ` <6d9307aae86d14ecbf70a39012b2d9f2-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
2018-11-14 15:13 ` Sean Paul
2018-11-14 19:33 ` Jeykumar Sankaran
[not found] ` <7e02a42c4a0aebc635cf6d798d8a667a-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
2018-11-14 19:43 ` Sean Paul
2018-11-16 18:28 ` Sean Paul
2018-11-13 20:52 ` [PATCH 4/8] drm/msm: dpu: Use atomic_disable for dpu_crtc_disable Sean Paul
2018-11-13 20:52 ` [PATCH 6/8] drm/msm: dpu: Separate crtc assignment from vblank enable Sean Paul
[not found] ` <20181113205257.170707-6-sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>
2018-11-14 0:48 ` Jeykumar Sankaran
[not found] ` <a8d1ef336746c930e8274153c82ebf99-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
2018-11-14 15:16 ` Sean Paul [this message]
2018-11-14 19:43 ` Jeykumar Sankaran
[not found] ` <25b009f90b6513f788fcc6748ae26e63-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
2018-11-14 19:53 ` Sean Paul
2018-11-14 19:57 ` Ville Syrjälä
[not found] ` <20181114195704.GV9144-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
2018-11-14 20:46 ` Jeykumar Sankaran
2018-11-14 20:52 ` Sean Paul
2018-11-14 21:50 ` Jeykumar Sankaran
2018-11-13 20:52 ` [PATCH 7/8] drm/msm: dpu: Remove vblank_requested flag from dpu_crtc Sean Paul
[not found] ` <20181113205257.170707-7-sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>
2018-11-14 0:47 ` Jeykumar Sankaran
[not found] ` <fd31b1c96143132b95bc937dfa201e85-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>
2018-11-14 15:14 ` Sean Paul
[not found] ` <20181113205257.170707-1-sean-p7yTbzM4H96eqtR555YLDQ@public.gmane.org>
2018-11-13 20:52 ` [PATCH 2/8] drm/msm: dpu: Remove crtc_lock from setup_mixers Sean Paul
2018-11-13 20:52 ` [PATCH 5/8] drm/msm: dpu: Don't bother checking ->enabled in dpu_crtc_vblank Sean Paul
2018-11-13 20:52 ` [PATCH 8/8] drm/msm: dpu: Remove crtc_lock Sean Paul
2018-11-13 20:57 ` [PATCH 1/8] drm/msm: dpu: Move pm_runtime_(get|put) from vblank_enable Sean Paul
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=20181114151615.GO154160@art_vandelay \
--to=sean-p7ytbzm4h96eqtr555yldq@public.gmane.org \
--cc=dri-devel-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
--cc=freedreno-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
--cc=jsanka-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org \
--cc=linux-arm-msm-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=robdclark-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
--cc=seanpaul-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org \
/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