VF50 Network Link Down After Watchdog Reset

We are using VF50 modules with Colibri_VF_LinuxConsoleImageV2.5 and Kernel 4.1.15 on a custom board in a product.

Recently we have been testing use of the watchdog. We are finding some rather odd behavior when a watchdog reset is forced (by killing the watchdog toggle daemon).

After a forced watchdog reset, the unit correctly resets. However, on reboot, the network link does not come up. Linux reports the link as down and there is no link LED on the network connector/switch. Additionally, rebooting the device does not bring the network link back up.

The unit has to be power-cycled for the link to be restored.

Oddly, if the unit has been powered off for a while, the link DOES correctly come back up after a watchdog reset. If it is been powered up and on the network for an extended period, the link does NOT come back up after a watchdog reset.

Any help will be appreciated.

I am assuming you use the regular Vybrid internal Watchdog? It should cause a reset pulse which also resets the Ethernet PHY, so your findings are rather surprising.

It could be related to #16470. Can you observe the issue on multiple modules?

Yes, we are using the internal Vybrid watchdog and we see the problem on multiple VF50 modules.
The watchdog resets the module, as expected, however, networking does not come back up. There are no link or traffic LED’s on the ethernet connector and LINUX reports the link down.

Only powering off the device seems to restore the link.

It seems time related - initially after powerup, the module and ethernet resets on the watchdog correctly. However, when the module has been connected to the network for a longer period then the network link is not restored on a watchdog reset.

Is a hardware reset behaving differently? (Not power cycle, just reset button)?

Can’t easily test that one. The module is on a custom board - no reset button.

Don’t think the problem is related to #16470 since networking is fine until watchdog reset is forced.

could you try reboot using the reboot since you do not have a reset button?

Rebooting the device via the Linux command line does not bring the link back up. The link only seems to come back up from a power down.

thanks for the information. Could you say how much minimum time the module has to be connected to network to reproduce the issue?

I cannot provide a very precise answer to this. I found this out by having problems with the network connection all day after a watch dog reset, but next morning after power up it worked fine for a time.

Definitely less than 8 hours - I suspect 2 or 3.

You say you see the issue on multiple VF50. Can you reproduce the issue on every module or is it only really on some? If the latter, what percentage would you guess?

It still might be #16470. I assume you are using the 4.1 kernel which comes with the V2.5 BSP. Can you try backport this fix, and/or test with a Linux 4.4 kernel from the latest stable BSP?