2141515_en-US

[#1]

s32g3 pfe - help reqired for implementing a driver running in secure trustzone

This is for the partner Provenrun, implementing the own OS ProvenCore that runs from the Secure World of the ARM TrustZone

As part of the integration of its formally Proven OS (ProvenCore) onto the TrustZone of the Cortex-A53 of the S32G3, ProvenRun developed a HIF driver for getting network access for its OS and Trusted Applications.

The HIF is assignment is as follows :

• HIF0 is the master HIF, belongs to the M7 core
• HIF1 is a slave assigned to ProvenCore (CortexA, S-World)
• HIF2 is a slave a QNX and VM guest (CortexA, NS-World)
• HIF3 is a slave assigned to ProvenCore (CortexA, S-World)

The PFE is setup as a L2 bridge. The initialization and loading of PFE FW are performed by the M7, before the Cortex-A are reset.

The current test setup is as it follows:

* S32G3 Evaluation board

* HIF 1 Standalone driver     

* HIF 1 driver communicate with EMAC 1

* Test application sending packets as fast as possible on top of the HIF driver.

Questions and requests:

1. Need clarification whether HIF1 standalone driver is considered as slave or not.

2. Can the HIF driver access PFE registers?

3. How to manage the LLTX? (Reverse engineering the Linux drivers is not an option.)

4. A debugging session with the PFE expert team.

Thanks

PFEPFE LinuxRe: s32g3 pfe - help reqired for implementing a driver running in secure trustzone

Closing thread as initial questions from customer are answered and clarified.

For further specific questions, Would create a new thread with reference to this case.

Regards,
Guru

Re: s32g3 pfe - help reqired for implementing a driver running in secure trustzone

Thank you Milan.

@Sridhar_FAE , FYI.

Re: s32g3 pfe - help reqired for implementing a driver running in secure trustzone

Hi Adrian, 

I have replied in e-mail thread, but to keep the update also here, I will try to answer the questions.

1.  In Multi-Instance setup with multiple HIF drivers instances, we separate them as Master - Slave drivers. Customer has multiple HIF drivers with HIF 1 standalone driver, which can be called Slave from our perspective.
2. Only Master driver can access PFE registers in our design, so no. HIF driver can access only HIF channel registers. - see PFE_LNX_DRV_S32G_UserManual.pdf (2.9.4 Memory access permissions)
3. LLTX is some kind of workaround for TX packets to avoid dropping in EMAC TMU. It is implemented by driver checking EMAC TMU queue level before sending data to HIF channel. It is described more in e-mail.
4. We had some engineering session with some PoC demo presentation shared. But it is not recommended to apply, as the customer use case might not be valid for LLTX.

Regards

Milan