Re: [Intel-xe] [PATCH 1/2] RFC drm/xe: Disable ctb communication

[Michal Wajdeczko <michal.wajdeczko@intel.com>]

On 23.08.2023 07:09, Riana Tauro wrote:
> During Runtime suspend, GuC is reset for both D0->D3hot/D3Cold
> transistions. It is not necessary for GuC to reset for D0 -> D3hot,

typo

> only enable/disable ctb communication.
> 
> Modify the existing ct disable function to send a request
> to disable ct communication if firmware is running.
> Add a function to disable guc communication.

s/ctb/CTB
s/ct/CT
s/guc/GuC

> 
> Signed-off-by: Riana Tauro <riana.tauro@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_guc.c    | 15 +++++++++++++++
>  drivers/gpu/drm/xe/xe_guc.h    |  1 +
>  drivers/gpu/drm/xe/xe_guc_ct.c | 11 ++++++++++-
>  drivers/gpu/drm/xe/xe_guc_ct.h |  2 +-
>  4 files changed, 27 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
> index 2493c5859948..3e4c00bee6bf 100644
> --- a/drivers/gpu/drm/xe/xe_guc.c
> +++ b/drivers/gpu/drm/xe/xe_guc.c
> @@ -557,6 +557,21 @@ int xe_guc_enable_communication(struct xe_guc *guc)
>  	return 0;
>  }
>  
> +int xe_guc_disable_communication(struct xe_guc *guc)
> +{
> +	int err;
> +	struct xe_guc_ct *ct = &guc->ct;
> +
> +	err = xe_guc_ct_disable(ct);
> +	if (err)
> +		return err;


> +
> +	/* Check for messages received during/after the CT disable */
> +	guc_handle_mmio_msg(guc);

are you sure that we need this ?

> +
> +	return 0;
> +}
> +
>  int xe_guc_suspend(struct xe_guc *guc)
>  {
>  	int ret;
> diff --git a/drivers/gpu/drm/xe/xe_guc.h b/drivers/gpu/drm/xe/xe_guc.h
> index 3addd8fc674a..18cf5d02eea7 100644
> --- a/drivers/gpu/drm/xe/xe_guc.h
> +++ b/drivers/gpu/drm/xe/xe_guc.h
> @@ -18,6 +18,7 @@ int xe_guc_post_load_init(struct xe_guc *guc);
>  int xe_guc_reset(struct xe_guc *guc);
>  int xe_guc_upload(struct xe_guc *guc);
>  int xe_guc_min_load_for_hwconfig(struct xe_guc *guc);
> +int xe_guc_disable_communication(struct xe_guc *guc);
>  int xe_guc_enable_communication(struct xe_guc *guc);
>  int xe_guc_suspend(struct xe_guc *guc);
>  void xe_guc_notify(struct xe_guc *guc);
> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
> index 59136b6a7c6f..64c2029b2752 100644
> --- a/drivers/gpu/drm/xe/xe_guc_ct.c
> +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
> @@ -21,6 +21,7 @@
>  #include "xe_map.h"
>  #include "xe_pm.h"
>  #include "xe_trace.h"
> +#include "xe_uc_fw.h"
>  
>  /* Used when a CT send wants to block and / or receive data */
>  struct g2h_fence {
> @@ -319,15 +320,23 @@ int xe_guc_ct_enable(struct xe_guc_ct *ct)
>  	return err;
>  }
>  
> -void xe_guc_ct_disable(struct xe_guc_ct *ct)
> +int xe_guc_ct_disable(struct xe_guc_ct *ct)
>  {
> +	struct xe_guc *guc = ct_to_guc(ct);
> +	int ret = 0;
> +
>  	mutex_lock(&ct->lock); /* Serialise dequeue_one_g2h() */
>  	spin_lock_irq(&ct->fast_lock); /* Serialise CT fast-path */
>  	ct->enabled = false; /* Finally disable CT communication */
>  	spin_unlock_irq(&ct->fast_lock);
>  	mutex_unlock(&ct->lock);
>  
> +	if (xe_uc_fw_is_running(&guc->fw))

nit: maybe there should be a wrapper for this like:

	bool xe_guc_is_running(xe_guc *guc) { ... }

> +		ret = guc_ct_control_toggle(ct, false);

do we have a plan what to do when this CTB toggle fails?
and what it really means for the caller if it receives an error?

> +
>  	xa_destroy(&ct->fence_lookup);
> +
> +	return ret;
>  }
>  
>  static bool h2g_has_room(struct xe_guc_ct *ct, u32 cmd_len)
> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.h b/drivers/gpu/drm/xe/xe_guc_ct.h
> index f15f8a4857e0..1ad150c19ae3 100644
> --- a/drivers/gpu/drm/xe/xe_guc_ct.h
> +++ b/drivers/gpu/drm/xe/xe_guc_ct.h
> @@ -12,7 +12,7 @@ struct drm_printer;
>  
>  int xe_guc_ct_init(struct xe_guc_ct *ct);
>  int xe_guc_ct_enable(struct xe_guc_ct *ct);
> -void xe_guc_ct_disable(struct xe_guc_ct *ct);
> +int xe_guc_ct_disable(struct xe_guc_ct *ct);
>  void xe_guc_ct_fast_path(struct xe_guc_ct *ct);
>  
>  struct xe_guc_ct_snapshot *