Re: [Intel-gfx] [PATCH] drm/i915/pxp: don't start pxp without mei_pxp bind

[Ceraolo Spurio, Daniele]

On 8/9/2022 3:57 PM, Juston Li wrote:
pxp will not start correctly until after mei_pxp bind completes and
intel_pxp_init_hw() is called.

Signed-off-by: Juston Li <justo...@google.com>
---
  drivers/gpu/drm/i915/pxp/intel_pxp.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/pxp/intel_pxp.c 
b/drivers/gpu/drm/i915/pxp/intel_pxp.c
index 15311eaed848..3ef9e4e1870b 100644
--- a/drivers/gpu/drm/i915/pxp/intel_pxp.c
+++ b/drivers/gpu/drm/i915/pxp/intel_pxp.c
@@ -184,7 +184,7 @@ int intel_pxp_start(struct intel_pxp *pxp)
  {
        int ret = 0;
  
-	if (!intel_pxp_is_enabled(pxp))
+       if (!intel_pxp_is_enabled(pxp) || !pxp->pxp_component_added)

pxp_component_added being true only indicates that we've called 
component_add and not if the component is currently bound. For checking 
the current state of the component you can look at the 
pxp->pxp_component pointer, which is set/cleared on component 
bind/unbind. Note that pxp_component has to be accessed under 
pxp->tee_mutex.

Are you actually seeing a scenario where the user manages to submit 
before the bind is complete? the bind is async to i915, but I've always 
seen it complete before control was given to the user to start submitting.

Daniele

                return -ENODEV;
  
  	mutex_lock(&pxp->arb_mutex);