[BPI-R4] SFP works in openWRT 21.02 only?

[ericwoud (Eric W.)]

It is up, but network is not configured correctly. That is very different. Writing it like that suggests a totally different problem

[buz]

Fine but then what’s missing (and why does it work with older snapshots as it is)?

[ericwoud (Eric W.)]

Sadly I do not use openwrt… If this was a driver issue, that would not really matter, and I would also be curious as to what is wrong. But this is a userland issue and depends on the operating system of choice. That is, if others also can partly fix it with udhcpc.

[buz]

I agree it’s most likely a userland issue and the likely culprit is a major netifd upgrade in openwrt-snapshot in early August.

Also, FWIW, the DAC going to my switch works like it always did.

[Karasik]

I’m away from home until September, when I get there I’ll try this patch https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=a1a9572f43776c3fed46c4545e93fbbb25d923c2 Something was changed in kernel 6, so it works on 5.4.

[buz]

FWIW setting force_link worked for me (both on wan and wan6):

config interface 'wan'
        option device 'br-wan'
        option proto 'dhcp'
        option force_link '1'

[Thanks]

@Karasik

module ODI DFP-34X-2C3

I think the problem is that the module is detected as “DFP-34X-2C3” and not as “DFP-34X-2C2”, maybe the second is the only one supported in the linux kernel.

Try flashing the firmware of the “DFP-34X-2C2” model like these guys did, since both devices are the same and only the connectors are different (UPC vs APC):

Please update info that DFP-34X-2C2 and DFP-34X-2C3 is the same device but...

This is just a suggestion and I hope you can solve this problem.

[Stat_headcrabed (Shengyu Qu)]

@Karasik To change module connection speed, check this discussion: Right SFP EEPROM config for DFP-34X-2C2 to work automatically work at 2500base-X? · Anime4000/RTL960x · Discussion #250 · GitHub

(Though this shouldn’t have anything with your current problem)

[Karasik]

Stick is defined by the “HW ID”, the name is just a string in the EEPROM. Patches change link speed 1Gbps to 2.5Gbps without renaming 2c2 to 2c3. In kernel 5.4 it works without problems at a speed of 1 Gbps, in 6.6 it does not work completely. Apparently the reason is changes in 6.x kernels. When I get back, I’ll try this patch git.openwrt.org Git - openwrt/openwrt.git/commit

[Stat_headcrabed (Shengyu Qu)]

No need to use quirk to patch speed and mainline linux refuse to accept this patch due to possible lockup. The best way to change speed is still changing eeprom info.

Besides, this module boots very slow, so sometimes system won’t think it is working. Have you tried doing a soft reboot after plugging in the module for a while?

[Karasik]

Yes, I tried it, it doesn’t work in 6.6. I heard about flashing EEPROM, but didn’t understand how to upload the changed firmware.

[Stat_headcrabed (Shengyu Qu)]

@frank-w Do you have some actual hardware to test this? I see tons of compatibility issues on r4+different pon sticks. This is a severe issue

[Thanks]

Maybe this will help you:

• ODI DFP-34X-2C2 | Hack GPON
• ODI DFP-34X-2C2 official firmware repository by Anime4000
• ODI DFP-34X-2C2 custom firmware for HUAWEI OLT Brazil (Information: Here)
• ODI SFP XPON stick MAC_KEY generator
• Factory reset via commands for RTL960x
• Factory reset without command (physical) for RTL960x

[frank-w]

No, i have no pon sfp or similar…only the 3 sfp i added to my compatible list in wiki

[Karasik]

This will most likely help, but I don’t know how to upload the firmware. Somehow through the i2c utility.

Right SFP EEPROM config for DFP-34X-2C2 to work automatically work at...

[kuzevich1983 (Ruslan_Sabirov)]

Good day. I am also struggling with launching dfp-34x-2c2, I changed the eeprom values ​​as indicated on github, but now when calling the ethtool command, eth2 shows speeds from 1000 to 10000 and the optics connection does not work and it is impossible to connect to the sfp itself via the web interface. Did you manage to do something, if so, describe how. Thank you.

[Karasik]

Hi, what patches have you tried? My module will be back in a few days.

[kuzevich1983 (Ruslan_Sabirov)]

Hi, i not tried patches, on another topic on this forum i understand that some sfp patches a included on latest snapshot fit bpi-r4.

[Thanks]

Do you have something connected to the M.2 slot at the same time you connect the SFP module?

It appears that the BPI-R4 v1.1 has a design flaw and connecting something into the M.2 slot causes the SFP modules to not be detected or not work.

Source:

@dangowrt

I still suggest to not populate R228 and R230 in future runs of the board. See also my post:

I’ve just tried the M.2 Key-M socket on the back of the board with a cheap NVMe SSD (S3+ S3SSD240) and trying to read AT24 I2C EEPROM in the same bus which is also used as SMBus (MFG Data and MFG Clock) of the M.2 socket (see schematics excerpt below):

The problem here is that the presence of the SSD doesn’t only result in the at24 EEPROM (and probably also the RTC, I didn’t try) not being accessible any more, but also all other busses on the mux stop working and hence SFP modules can no longer be probed.

As MFG Data and MFG Clock signals are hardly ever needed, I suggest to not populate R228 and R230 in future versions.

[Karasik]

Maybe this will help + EEPROM mod For ODI: https://patchwork.kernel.org/project/netdevbpf/patch/AS1PR03MB81893D69344708C98EE2B470827F2@AS1PR03MB8189.eurprd03.prod.outlook.com/

For OEM: https://lore.kernel.org/lkml/[email protected]/T/

If not, then this: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=a1a9572f43776c3fed46c4545e93fbbb25d923c2

I can try it myself in a week.