Re: [PATCH net-next 3/3] net: stmmac: Add UltraRISC DP1000 GMAC support

Hi,

On 9/1/26 10:32, Jia Wang wrote:

On 2026-09-01 09:29 +0200, Maxime Chevallier wrote:

quoted

Hello,

On 9/1/26 03:27, Jia Wang wrote:

quoted

The DP1000 GMAC integration provides fixed TX and RX RGMII clock delays
that cannot be disabled. For boards without PCB delays, "rgmii-id"
describes the link, but passing that mode unchanged to the PHY would enable
a second set of delays.

Use phy_fix_phy_mode_for_mac_delays() to translate "rgmii-id" to "rgmii"
before attaching the PHY. Reject other RGMII modes because the fixed MAC
delays cannot be combined with them without duplicating or omitting a
delay.

[...]

Thanks. I enabled CONFIG_STMMAC_SELFTESTS and ran the test on the Milk-V
Titan with the new glue driver:

 1. MAC Loopback                 0
 2. MMC Counters                 0
 3. EEE                          -95
 4. Hash Filter MC               0
 5. Perfect Filter UC            -95
 6. MC Filter                    -95
 7. UC Filter                    -95
 8. Flow Control                 0
 9. RSS                          -95
10. VLAN Filtering               -110
11. VLAN Filtering (perf)        -110
12. Double VLAN Filter           -110
13. Double VLAN Filter (perf)    -110
14. Flexible RX Parser           -95
15. SA Insertion (desc)          0
16. SA Replacement (desc)        0
17. SA Insertion (reg)           0
18. SA Replacement (reg)         0
19. VLAN TX Insertion            -110
20. SVLAN TX Insertion           -110
21. L3 DA Filtering              -95
22. L3 SA Filtering              -95
23. L4 DA TCP Filtering          -95
24. L4 SA TCP Filtering          -95
25. L4 DA UDP Filtering          -95
26. L4 SA UDP Filtering          -95
27. ARP Offload                  -110
28. Jumbo Frame                  0
29. Multichannel Jumbo           -95
30. Split Header                 0
31. TBS (ETF Scheduler)          -95

Awesome, that's pretty-much what I expected, nothing too bad stands out :)

Code itself looks good, no obvious error, so

Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>

Maxime