| 1 Jun 2025 |
 emily | I don't know how to easily find out if it reached stable yet | 12:35:58 |
| emily | so it seems like inband/sgmii, inband/2500base-x, and phy/2500base-x are all things | 12:37:43 |
| emily | is inband the mode where the transceiver does the MAC and phy the mode where it just passes stuff through? | 12:37:58 |
 K900 | It did land yes | 12:38:03 |
| K900 | inband is just in-band negotiation | 12:38:20 |
| emily | which is different from "autoneg"? | 12:38:42 |
| emily | or the same? | 12:39:09 |
| emily | https://lore.kernel.org/all/Z1F1b8eh8s8T627j@shell.armlinux.org.uk/ looks like the relevant quirks for the PCS. I don't know if they landed or if they're even still necessary with the new series | 12:40:19 |
| K900 | That has landed yes | 12:40:58 |
| emily | hmm | 12:41:10 |
| emily | that surprises me because
Eric Woudstra reported that a PCS attached using 2500base-X does not
see link when phylink is using in-band mode, but autoneg is disabled,
despite there being a valid 2500base-X signal being received. We have
these settings:
| 12:41:13 |
| emily | seems to imply that it's meant to fix using inband rather than wroking around it | 12:41:19 |
| emily | * seems to imply that it's meant to fix using inband rather than working around it | 12:41:21 |
| emily | * seems to imply that it's meant to fix using inband rather than working around it (by disabling it) | 12:41:26 |
| K900 | https://docs.kernel.org/networking/sfp-phylink.html | 12:41:58 |
| emily | also this implies that in-band and autoneg aren't the same thing so I still feel like I have no idea what in-band is :) | 12:42:01 |
| K900 | This explains the modes kind of | 12:42:10 |
| emily | right. yeah I was looking at this | 12:42:13 |
| emily | it did not really make me understand the difference between autoneg and in-band, although I'm trying. | 12:42:27 |
| K900 | Autonegotiation is about negotiating with the other end of the cable | 12:43:33 |
| K900 | In-band mode is about what's between the MAC and the module | 12:43:49 |
| emily | ah | 12:43:54 |
| emily | so the tl;dr is that you have to explicitly specify 2500BASE-X on the host end for the PCS on the R4? | 12:44:20 |
| emily | to have it talk that to an SFP+? | 12:44:29 |
| emily | and then the other patchset hopefully makes that work? | 12:44:42 |
| K900 | In most cases it doesn't matter because you're either speaking GMII whatever to the module and then that's running at its own rate | 12:44:46 |
| K900 | But specifically for base2500-x it kinda matters because base2500-x is supposed to be dumb | 12:44:58 |
| K900 | EXCEPT WHEN IT'S NOT etc | 12:45:03 |
| emily | right… | 12:45:09 |
| K900 | Basically it depends on what your module is doing | 12:45:40 |
