SEOS-3232: Check compatibility of sel4 util libs pull request for i.MX6 ethernet driver

[FelixSchladt]

Check compatibility of sel4 util libs pull request for i.MX6 ethernet driver

Check compatibility of update pull request for [https://github.com/seL4/util_libs/pull/112] i.MX6 ethernet driver. This was something we have changed during our refactoring, as we e.g. introduced a custom config structure.

DoD:

• clarify if this causes any issue and create follow up tickets (/)
• Kent added a new field {{NIC_CONFIG_NO_CLOCK_SYS}} in the config. Ticket exists to update TRENTOS handbook and desrcibe this parameter -> (x) SEE COMMENTS BELOW
• Reconsider using the {{ethif_imx6_init}} interface, engage with seL4 community how this is related to the driver framework -> (x) SEE COMMENTS BELOW

Created by

axel heider
Created: Wednesday, November 17, 2021 01:50 AM UTC+01:00
Updated: Friday, December 17, 2021 10:21 AM UTC+01:00

Comments

Matthias Buhl

I would propose that we move the second point of the DoD to a separate ticket since the process should basically be:
Kent merges his changes to the master branch, our SDK referenced branch of the sel4_util_libs is updated to a branch that contains this change or is based on a newer master branch with this change and then we update the Handbook once we know that it will make its way into the next release.

created: Tuesday, December 07, 2021 05:23 PM UTC+01:00
updated: Tuesday, December 07, 2021 05:23 PM UTC+01:00

Matthias Buhl

From the analysis, it seems that the first commit that introduces the {{NIC_CONFIG_NO_CLOCK_SYS}} flag should not cause any big migration issues. We will mainly need to document this new flag in our handbook and decide what we set here as a default configuration in the {{plat_nic_imx6.camkes}}. As we probably would just like to keep the behavior, we would not need to change anything for our default configuration.

For the second commit that reintroduces the {{ethif_imx6_init }}function it also does not seem like we need to make any real adaptations here as the changes are all somewhat driver internal. The only topic we might need to verify is how the fact that a header include that was removed by Kent will impact our ethdriver.c module as this is making use of functions such as enet_mdio_read() that relied on this header being brought in by the eth_plat.h.

For now I would argue to close this ticket and possibly create a ticket for the last point on the DoD, as this seems more like longer running independent topic.
The second point should also result in a ticket but only once this change from Kent actually gets merged and we update our reference for the SDK to it. Currently it seems quite speculative to open a ticket for something that is not even merged yet.

created: Thursday, December 09, 2021 05:45 PM UTC+01:00
updated: Thursday, December 09, 2021 05:45 PM UTC+01:00

axel heider

I'm about to approve Kent's PR, so he can merge it. It's hard to hold my approval there back due to our internal reasons and we are finde to deal with the merge anyway.

I've asked Kent to keep the header in and created a dedicated issue to fix this include quirk: seL4/util_libs#116

created: Thursday, December 09, 2021 06:16 PM UTC+01:00
updated: Thursday, December 09, 2021 06:34 PM UTC+01:00

Matthias Buhl

Yes, I would also agree that you can approve Kents PR and we will then eventually deal with the changes once we switch to a more recent snapshot of the master branch.

created: Friday, December 10, 2021 08:34 AM UTC+01:00
updated: Friday, December 10, 2021 08:34 AM UTC+01:00

Matthias Buhl

I created  https://jira.hensoldt-cyber.systems/browse/SEOS-3309 to keep track of the handbook update that will be required at some point.

created: Friday, December 17, 2021 10:21 AM UTC+01:00
updated: Friday, December 17, 2021 10:21 AM UTC+01:00

Subtasks