[PATCH v2] drm/xe: remove GuC reload in D3Hot path

Rodrigo Vivi rodrigo.vivi at intel.com
Fri Jun 28 01:00:23 UTC 2024


On Mon, Jun 24, 2024 at 10:10:48AM +0200, Michal Wajdeczko wrote:
> Hi Riana,
> 
> 
> On 24.06.2024 09:27, Riana Tauro wrote:
> > Currently GuC is reloaded for both runtime suspend
> > and system suspend. For D3hot <-> D0 transitions
> > no power is lost on suspend and GuC reload is not
> > necessary.
> 
> quite aggressive line wrap at ~50, try wrap a ~72
> (only commit titles should be around ~50)
> 
> > 
> > remove GuC reload from D3Hot path and only enable/disable
> > CTB communications
> 
> "Remove .... communication."
> 
> > 
> > v2: rebase
> > 
> > Signed-off-by: Riana Tauro <riana.tauro at intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_gt.c     | 19 ++++++++++---
> >  drivers/gpu/drm/xe/xe_gt.h     |  4 +--
> >  drivers/gpu/drm/xe/xe_guc.c    | 52 +++++++++++++++++++++++++++++++++-
> >  drivers/gpu/drm/xe/xe_guc.h    |  2 ++
> >  drivers/gpu/drm/xe/xe_guc_ct.c | 35 ++++++++++++++++++-----
> >  drivers/gpu/drm/xe/xe_guc_ct.h |  3 +-
> >  drivers/gpu/drm/xe/xe_pm.c     |  8 +++---
> >  drivers/gpu/drm/xe/xe_uc.c     | 30 ++++++++++++++++++++
> >  drivers/gpu/drm/xe/xe_uc.h     |  2 ++
> >  9 files changed, 136 insertions(+), 19 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
> > index 759634cff1d8..5b48e4db9776 100644
> > --- a/drivers/gpu/drm/xe/xe_gt.c
> > +++ b/drivers/gpu/drm/xe/xe_gt.c
> > @@ -776,8 +776,9 @@ void xe_gt_suspend_prepare(struct xe_gt *gt)
> >  	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
> >  }
> >  
> > -int xe_gt_suspend(struct xe_gt *gt)
> > +int xe_gt_suspend(struct xe_gt *gt, bool runtime)
> 
> as you are changing function signature, please add full kernel-doc for
> this public function
> 
> >  {
> > +	struct xe_device *xe = gt_to_xe(gt);
> >  	int err;
> >  
> >  	xe_gt_dbg(gt, "suspending\n");
> > @@ -787,7 +788,11 @@ int xe_gt_suspend(struct xe_gt *gt)
> >  	if (err)
> >  		goto err_msg;
> >  
> > -	err = xe_uc_suspend(&gt->uc);
> > +	if (runtime && !xe->d3cold.allowed)
> 
> nit: maybe instead of just looking at some internal flag, it would be
> better to define helpers:
> 
> static inline bool xe_device_d3cold_allowed(xe) { .. }
> static inline bool xe_device_d3hot_allowed(xe) { .. }
> 
> > +		err = xe_uc_runtime_suspend(&gt->uc);
> > +	else
> > +		err = xe_uc_suspend(&gt->uc);
> > +
> >  	if (err)
> >  		goto err_force_wake;
> >  
> > @@ -825,8 +830,9 @@ int xe_gt_sanitize_freq(struct xe_gt *gt)
> >  	return ret;
> >  }
> >  
> > -int xe_gt_resume(struct xe_gt *gt)
> > +int xe_gt_resume(struct xe_gt *gt, bool runtime)
> 
> add kernel-doc
> 
> >  {
> > +	struct xe_device *xe = gt_to_xe(gt);
> >  	int err;
> >  
> >  	xe_gt_dbg(gt, "resuming\n");
> > @@ -834,7 +840,12 @@ int xe_gt_resume(struct xe_gt *gt)
> >  	if (err)
> >  		goto err_msg;
> >  
> > -	err = do_gt_restart(gt);
> > +	/* Do not reload GuC at runtime resume from D3Hot to D0 */
> 
> maybe better to explain "why" than "what":
> 
> /* GuC is still alive at D3hot, no need to reload it */
> 
> > +	if (runtime && !xe->d3cold.allowed)
> > +		xe_uc_runtime_resume(&gt->uc);
> > +	else
> > +		err = do_gt_restart(gt);
> > +
> >  	if (err)
> >  		goto err_force_wake;
> >  
> > diff --git a/drivers/gpu/drm/xe/xe_gt.h b/drivers/gpu/drm/xe/xe_gt.h
> > index 1123fdfc4ebc..4c942e4716ad 100644
> > --- a/drivers/gpu/drm/xe/xe_gt.h
> > +++ b/drivers/gpu/drm/xe/xe_gt.h
> > @@ -52,8 +52,8 @@ int xe_gt_record_default_lrcs(struct xe_gt *gt);
> >  void xe_gt_record_user_engines(struct xe_gt *gt);
> >  
> >  void xe_gt_suspend_prepare(struct xe_gt *gt);
> > -int xe_gt_suspend(struct xe_gt *gt);
> > -int xe_gt_resume(struct xe_gt *gt);
> > +int xe_gt_suspend(struct xe_gt *gt, bool runtime);
> > +int xe_gt_resume(struct xe_gt *gt, bool runtime);
> >  void xe_gt_reset_async(struct xe_gt *gt);
> >  void xe_gt_sanitize(struct xe_gt *gt);
> >  int xe_gt_sanitize_freq(struct xe_gt *gt);
> > diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
> > index 172b65a50e31..1ea1d00f93f3 100644
> > --- a/drivers/gpu/drm/xe/xe_guc.c
> > +++ b/drivers/gpu/drm/xe/xe_guc.c
> > @@ -870,7 +870,7 @@ int xe_guc_enable_communication(struct xe_guc *guc)
> >  		guc_enable_irq(guc);
> >  	}
> >  
> > -	err = xe_guc_ct_enable(&guc->ct);
> > +	err = xe_guc_ct_register(&guc->ct);
> >  	if (err)
> >  		return err;
> >  
> > @@ -1101,6 +1101,56 @@ void xe_guc_sanitize(struct xe_guc *guc)
> >  	guc->submission_state.enabled = false;
> >  }
> >  
> > +/**
> > + * xe_guc_runtime_suspend - GuC runtime suspend
> > + * @guc: GuC object
> > + *
> > + * This function waits for any outstanding CTB
> > + * and disables CTB communication before entering
> > + * D3Hot
> > + *
> > + * Return: 0 on success, negative error code on error.
> > + */
> > +int xe_guc_runtime_suspend(struct xe_guc *guc)
> > +{
> > +	struct xe_guc_ct *ct = &guc->ct;
> > +
> > +	/*
> > +	 * Wait for any outstanding CTB before tearing down communication
> > +	 * with GuC.
> > +	 */
> > +	if (!wait_event_timeout(ct->wq, !ct->g2h_outstanding, (HZ / 5)))
> > +		return -ETIMEDOUT;
> > +
> > +	/* Disable CTB */
> > +	xe_guc_ct_disable(&guc->ct);
> > +
> > +	return 0;
> > +}
> > +
> > +/**
> > + * xe_guc_runtime_resume - GuC runtime resume
> > + * @guc: GuC object
> > + *
> > + * This function enables CTB communication
> > + * and interrupts.
> > + */
> > +void xe_guc_runtime_resume(struct xe_guc *guc)
> > +{
> > +	/*
> > +	 * Power is not lost when resuming from D3Hot,
> > +	 * hence it is not necessary to reload GuC
> > +	 * everytime. Only enable interrupts and
> > +	 * CTB communication during resume
> > +	 */
> > +	guc_enable_irq(guc);
> 
> this will break VFs
> 
> can't you just use xe_guc_enable_communication() as it was ?
> it did the same steps what you are trying in this function
> 
> > +
> > +	xe_mmio_rmw32(guc_to_gt(guc), PMINTRMSK,
> > +		      ARAT_EXPIRED_INTRMSK, 0);
> 
> if power not lost we likely don't need this (as this is a default value
> and we already set it once as part of __xe_guc_upload()
> 
> > +
> > +	xe_guc_ct_enable(&guc->ct);
> > +}
> > +
> >  int xe_guc_reset_prepare(struct xe_guc *guc)
> >  {
> >  	return xe_guc_submit_reset_prepare(guc);
> > diff --git a/drivers/gpu/drm/xe/xe_guc.h b/drivers/gpu/drm/xe/xe_guc.h
> > index af59c9545753..a01999bb4a8e 100644
> > --- a/drivers/gpu/drm/xe/xe_guc.h
> > +++ b/drivers/gpu/drm/xe/xe_guc.h
> > @@ -22,6 +22,8 @@ int xe_guc_upload(struct xe_guc *guc);
> >  int xe_guc_min_load_for_hwconfig(struct xe_guc *guc);
> >  int xe_guc_enable_communication(struct xe_guc *guc);
> >  int xe_guc_suspend(struct xe_guc *guc);
> > +int xe_guc_runtime_suspend(struct xe_guc *guc);
> > +void xe_guc_runtime_resume(struct xe_guc *guc);
> >  void xe_guc_notify(struct xe_guc *guc);
> >  int xe_guc_auth_huc(struct xe_guc *guc, u32 rsa_addr);
> >  int xe_guc_mmio_send(struct xe_guc *guc, const u32 *request, u32 len);
> > diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
> > index b4137fe195a4..400680c429fd 100644
> > --- a/drivers/gpu/drm/xe/xe_guc_ct.c
> > +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
> > @@ -360,7 +360,16 @@ static void ct_exit_safe_mode(struct xe_guc_ct *ct)
> >  		xe_gt_dbg(ct_to_gt(ct), "GuC CT safe-mode disabled\n");
> >  }
> >  
> > -int xe_guc_ct_enable(struct xe_guc_ct *ct)
> > +/**
> > + * xe_guc_ct_register - register GuC CTB
> > + * @ct: GuC CT object
> > + *
> > + * register GuC CTB's and enable GuC CT
> > + * communication channel.
> > + *
> > + * Return: 0 on success, negative error code on error.
> > + */
> > +int xe_guc_ct_register(struct xe_guc_ct *ct)
> 
> hmm, can we have common naming convention across all driver components?
> 
> for the initialization there was pattern like:
> 
> 	- init_early
> 	- init
> 	- enable
> 	- disable
> 	- fini
> 
> and the "_register" suffix was mostly used by functions that registers
> the sysfs/debugfs attributes
> 
> now CTB is splitting 'enable' step into 'register' and 'enable' where
> the latter can't be used alone and be even more misleading
> 
> + Rodrigo

yeap, I fully agree with this and all the other points from Michal on
this review.

> 
> >  {
> >  	struct xe_device *xe = ct_to_xe(ct);
> >  	struct xe_gt *gt = ct_to_gt(ct);
> > @@ -383,11 +392,7 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct)
> >  	if (err)
> >  		goto err_out;
> >  
> > -	xe_guc_ct_set_state(ct, XE_GUC_CT_STATE_ENABLED);
> > -
> > -	smp_mb();
> > -	wake_up_all(&ct->wq);
> > -	xe_gt_dbg(gt, "GuC CT communication channel enabled\n");
> > +	xe_guc_ct_enable(ct);
> >  
> >  	if (ct_needs_safe_mode(ct))
> >  		ct_enter_safe_mode(ct);
> > @@ -396,10 +401,26 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct)
> >  
> >  err_out:
> >  	xe_gt_err(gt, "Failed to enable GuC CT (%pe)\n", ERR_PTR(err));
> > -
> >  	return err;
> >  }
> >  
> > +/**
> > + * xe_guc_ct_enable - Set GuC CT to enabled state
> > + * @ct: GuC CT object
> > + *
> > + * Set GuC CT to enabled state
> > + */
> > +void xe_guc_ct_enable(struct xe_guc_ct *ct)
> > +{
> > +	struct xe_gt *gt = ct_to_gt(ct);
> > +
> > +	xe_guc_ct_set_state(ct, XE_GUC_CT_STATE_ENABLED);
> > +
> > +	smp_mb();
> > +	wake_up_all(&ct->wq);
> > +	xe_gt_dbg(gt, "GuC CT communication channel enabled\n");
> 
> this seems wrong, as CTB was enabled and fully functional once you
> called guc_ct_control_toggle(true), decoupling reporting of what have
> already happened in the previous steps is IMO just misleading
> 
> + Matt
> 
> > +}
> > +
> >  static void stop_g2h_handler(struct xe_guc_ct *ct)
> >  {
> >  	cancel_work_sync(&ct->g2h_worker);
> > diff --git a/drivers/gpu/drm/xe/xe_guc_ct.h b/drivers/gpu/drm/xe/xe_guc_ct.h
> > index 105bb8e99a8d..f786b3ec2b68 100644
> > --- a/drivers/gpu/drm/xe/xe_guc_ct.h
> > +++ b/drivers/gpu/drm/xe/xe_guc_ct.h
> > @@ -11,7 +11,8 @@
> >  struct drm_printer;
> >  
> >  int xe_guc_ct_init(struct xe_guc_ct *ct);
> > -int xe_guc_ct_enable(struct xe_guc_ct *ct);
> > +int xe_guc_ct_register(struct xe_guc_ct *ct);
> > +void xe_guc_ct_enable(struct xe_guc_ct *ct);
> >  void xe_guc_ct_disable(struct xe_guc_ct *ct);
> >  void xe_guc_ct_stop(struct xe_guc_ct *ct);
> >  void xe_guc_ct_fast_path(struct xe_guc_ct *ct);
> > diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
> > index de3b5df65e48..070d28d193c7 100644
> > --- a/drivers/gpu/drm/xe/xe_pm.c
> > +++ b/drivers/gpu/drm/xe/xe_pm.c
> > @@ -99,7 +99,7 @@ int xe_pm_suspend(struct xe_device *xe)
> >  	xe_display_pm_suspend(xe, false);
> >  
> >  	for_each_gt(gt, xe, id) {
> > -		err = xe_gt_suspend(gt);
> > +		err = xe_gt_suspend(gt, false);
> >  		if (err) {
> >  			xe_display_pm_resume(xe, false);
> >  			goto err;
> > @@ -154,7 +154,7 @@ int xe_pm_resume(struct xe_device *xe)
> >  	xe_display_pm_resume(xe, false);
> >  
> >  	for_each_gt(gt, xe, id)
> > -		xe_gt_resume(gt);
> > +		xe_gt_resume(gt, false);
> >  
> >  	err = xe_bo_restore_user(xe);
> >  	if (err)
> > @@ -370,7 +370,7 @@ int xe_pm_runtime_suspend(struct xe_device *xe)
> >  	}
> >  
> >  	for_each_gt(gt, xe, id) {
> > -		err = xe_gt_suspend(gt);
> > +		err = xe_gt_suspend(gt, true);
> >  		if (err)
> >  			goto out;
> >  	}
> > @@ -423,7 +423,7 @@ int xe_pm_runtime_resume(struct xe_device *xe)
> >  	xe_irq_resume(xe);
> >  
> >  	for_each_gt(gt, xe, id)
> > -		xe_gt_resume(gt);
> > +		xe_gt_resume(gt, true);
> >  
> >  	if (xe->d3cold.allowed) {
> >  		xe_display_pm_resume(xe, true);
> > diff --git a/drivers/gpu/drm/xe/xe_uc.c b/drivers/gpu/drm/xe/xe_uc.c
> > index 0f240534fb72..997ab4e82931 100644
> > --- a/drivers/gpu/drm/xe/xe_uc.c
> > +++ b/drivers/gpu/drm/xe/xe_uc.c
> > @@ -288,6 +288,36 @@ int xe_uc_suspend(struct xe_uc *uc)
> >  	return xe_guc_suspend(&uc->guc);
> >  }
> >  
> > +/**
> > + * xe_uc_runtime_suspend - uC runtime suspend
> > + * @uc: uC object
> > + *
> > + * Return: 0 on success, negative error code on error.
> > + */
> > +int xe_uc_runtime_suspend(struct xe_uc *uc)
> > +{
> > +	/* GuC submission not enabled, nothing to do */
> 
> nit: this comment doesn't bring too much new info and likely can be
> dropped without losing any clarity of the code
> 
> > +	if (!xe_device_uc_enabled(uc_to_xe(uc)))
> > +		return 0;
> > +
> > +	return xe_guc_runtime_suspend(&uc->guc);
> > +}
> > +
> > +/**
> > + * xe_uc_runtime_resume - uC runtime resume
> > + * @uc: uC object
> > + *
> > + * Called while resuming from D3Hot
> > + */
> > +void xe_uc_runtime_resume(struct xe_uc *uc)
> > +{
> > +	/* GuC submission not enabled, nothing to do */
> > +	if (!xe_device_uc_enabled(uc_to_xe(uc)))
> > +		return;
> > +
> > +	xe_guc_runtime_resume(&uc->guc);
> > +}
> > +
> >  /**
> >   * xe_uc_remove() - Clean up the UC structures before driver removal
> >   * @uc: the UC object
> > diff --git a/drivers/gpu/drm/xe/xe_uc.h b/drivers/gpu/drm/xe/xe_uc.h
> > index 11856f24e6f9..dd483ea43b9b 100644
> > --- a/drivers/gpu/drm/xe/xe_uc.h
> > +++ b/drivers/gpu/drm/xe/xe_uc.h
> > @@ -15,6 +15,8 @@ int xe_uc_init_hw(struct xe_uc *uc);
> >  int xe_uc_fini_hw(struct xe_uc *uc);
> >  void xe_uc_gucrc_disable(struct xe_uc *uc);
> >  int xe_uc_reset_prepare(struct xe_uc *uc);
> > +int xe_uc_runtime_suspend(struct xe_uc *uc);
> > +void xe_uc_runtime_resume(struct xe_uc *uc);
> >  void xe_uc_stop_prepare(struct xe_uc *uc);
> >  void xe_uc_stop(struct xe_uc *uc);
> >  int xe_uc_start(struct xe_uc *uc);


More information about the Intel-xe mailing list