This thread is part of a series on the VCT5 timeout issue (GitHub Issue #17077). I want to collect timing measurements from as many devices as possible so the patch can be properly dimensioned. We start with the MikroTik RB5009 - other devices will follow in separate threads.
Background:
ethtool --cable-test-tdr fails with a timeout on certain devices with a Marvell PHY chip. TimSmall has developed a fix patch, but the timing values are chip-dependent. On the Huawei AP5030DN the chip needs ~10ms. On my MikroTik RB5009UPr+S+IN I measured ~645ms - using my own extension of marvell_vct5_wait_complete() for timing measurements. The current patch is not sufficient for the RB5009. To change that we need measurements from more devices.
Option A - build it yourself:
Everything you need is here:
Option B - pre-built initramfs (for those without build experience):
I can provide a pre-built initramfs image on request. Testing is simple and completely safe:
Boot the initramfs (runs in RAM only)
Check your interface name first:
ip link show
Then run the test 3 times - the first measurement may differ, from the second onwards the values should be stable:
ethtool --cable-test-tdr
Any port works, a short patch cable is fine.
Retrieve dmesg - easiest via SSH from your host:
ssh root@192.168.1.1 'dmesg | grep -i vct' > my_measurement.txt
Post the file here in the thread
Pull the cable - everything gone, nothing changed permanently
No experience needed - everyone is welcome!
I'm counting on your help - it won't hurt and won't take long! Boot, measure, done - and you will have actively contributed to a kernel fix.
What's next?
Once we have enough measurements for the RB5009, Part 2/x will follow with other affected devices.