Add support for Gigabit phy/speed

[raveslave]

Today, the speed is set to 100 Mbit/s here:

GenAVB_TSN/linux/scripts/avb.sh

Line 844 in 157d836

if [ "$AVB_MODE" -eq 1 ]; then

(easy to patch, but would be great if its configurable / driven by the phy mode selected)

In REALTIMEDEBUG_Rev2.5.pdf:

Attention: The PHY Delay Compensation Values in Table 64 are calibrated for 1 Gbps links. The i.MX AVB
endpoints are configured to run by default with 100 Mbps links. These compensation values should be enough to keep pDelay values under 800 ns (propagation time threshold), and therefore the port would still be declared as Capable.

Future releases shall have proper compensation values for each supported link speed.

would be great to see this changed in the next release

[raveslave]

@AlexandreHoward any updates here?

[AlexandreHoward]

Hi David,

Indeed the speed for the AVB use cases is forced to 100Mb/s.

The stack itself supports both 100Mb/s and 1Gb/s seamlessly and it detects link speed change automatically.
The 100 Mb/s rate was introduced to protect against specific corner cases : Mainly when the receive side of the endpoint is flooded with a high rate of small best effort packets on the FEC interface which can impact SRP/gPTP packets queued with the best effort ones.
I don't think there will be any plan to support the 1Gb/s rate for AVB use cases, unfortunately, as the chances of SRP/gPTP packets being lost amidst all the traffic on the same queue would be too high.

The compensation values have been updated in the latest version, you can find them in REALTIMEEDGEUG v2.8 (Table 77) if that can help

[raveslave]

it would be nice if this is a config or runtime param for scenarios where you might just run avdecc and higher level traffic on the link (Class-A traffic being offloaded/filtered in separate NIC's/HW), where the i.mx more acts as the controller

[raveslave]

checking in, this feel relevant on 118x series