F-tile ethernet hard ip in agilex7 - F-tile ethernet hard ip in agilex7 I generate 200G ethernet F-tile hard ip that base on agilex7, the IP configuration as follow: Our current test setup uses our custom development boards interconnected with Mellanox CX-8. The link fails to come up regardless of whether the CX-8 is configured in forced mode or auto-negotiation mode. Could you advise on how I should troubleshoot this issue next? Additionally, I would like to ask: Does the F-Tile Ethernet Hard IP support filtering packets smaller than 64 bytes? I cannot find the corresponding configuration options in the IP configuration GUI. Replies: Re: F-tile ethernet hard ip in agilex7 Hello, The recommended workaround is to re-apply the TX pause configuration in software after hot-plug/link recovery. In practice, after the optical module is reinserted and the link comes back up, rewrite TX_PAUSE_EN (tx_pause_en, offset 0x122C[8:0]) because that register controls the TX PAUSE / TX PFC enable bits and can end up cleared. Regards, Pavee Replies: Re: F-tile ethernet hard ip in agilex7 Thanks for the update. Can you please try the suggestion I provided earlier and do let me know if it helps. Regards, Pavee Replies: Re: F-tile ethernet hard ip in agilex7 During on-board debugging of the F-tile MAC Hard IP, we observed that the TX_PAUSE_EN register is automatically cleared to 0 after the optical module is hot-plugged (when the link goes down and then comes back up). What is the recommended workaround for this issue? Replies: Re: F-tile ethernet hard ip in agilex7 we use SFP112 and DAC cable, but it still fails to link up. Replies: Re: F-tile ethernet hard ip in agilex7 Hello, May I know what type of cable is being used in your setup? There is no Parameter Editor option to drop or filter undersized frames. The GUI only exposes maximum frame size (TX/RX maximum frame size, Enforce maximum frame size) which handles oversized frames, not undersized ones. On your MAC segmented interface, undersized frames are flagged via: o_rx_mac_fcs_error = asserted at EOP for FCS error, malformed, undersized, or truncated oversize o_rx_mac_error[1:0] = 2'd2 — explicitly “Undersized (frame shorter than 64b)” The MAC does not silently drop undersized frames by default, it forwards them to the client with error status set. To filter them, add logic in your client to discard frames where o_rx_mac_fcs_error or o_rx_mac_error == 2'd2 at EOP. Regards, Pavee