Re: [Linuxptp-users] Regression - phc2sys fails with certain PHC char dev filenames

[Ed B. <br...@ar...>]

On 11/11/21 17:28, Ed Branch wrote:
> 
> Thanks for the quick reply. Understood, digging this out properly (ie. 
> finding the /sys/class/ptp/ptp*/dev file with a matching minor to that 
> of the device node) would add a lot of complexity. However, looking at 
> the alloc_chardev_region() call in the ptp driver 
> (drivers/ptp/ptp_clock.c) reveals the first minor in the requested range 
> is always 0, so it might be ok to assume the PHC index is just the 
> device node minor.
> 
> Of course your suggestion of just ignoring the failure in phc2sys would 
> be enough to fix my use case too.
> 
> Thanks again for looking at this,
> - Ed Branch

Reading a bit further into posix_clock_open(...), I see it already 
relied on existence of device nodes named "/dev/ptp#". If these have to 
exist anyway then maybe there's no point accounting for other names. My 
actual failure is because I access them through a symlink created by 
udev based on the clock_name sysfs attribute. So simply resolving 
symlinks prior to the existing logic would also work.

For now, my workaround is to rename the symlinks so they do not match 
the "/dev/ptp*" pattern.

- Ed Branch