stuck php processes after upgrade to 13.1-RELEASE

From: Martin_Waschb├╝sch <martin_at_waschbuesch.de>
Date: Mon, 06 Jun 2022 21:37:51 UTC
Hi there,

ever since I upgraded some of my webservers to FreeBSD 13.1-RELEASE, 
some cronjobs executing php scripts seem to end up stuck.
It does not always happen. I have not been able to pinpoint the 
circumstances. Nor have I managed to come up with a way to reproduce the 
issue.

With every processes that gets stuck that way, I can do

truss -p <pid>

and will get a neverending output of:

sched_yield()					 = 0 (0x0)

which to me looks like an attempt to let other threads get on with work, 
but being rescheduled immediately.

Scripts this has happened with were magento2 and shopware 5.7 
maintenance cronjobs running with PHP 7.4.29.

Also note, in case this makes a difference, that all of these servers 
are running on VMs provided using kvm (I use a proxmox cluster for 
virtualisation).
I noticed that the new kvmclock timer module is automatically used. On 
13.0 I had to install kvmclock-kmod for that. Don't know if the code 
changed now it is part of base.


Anyway, maybe someone else has had similar things happen on their systems?

I was loath to open a PR because at this point I don't rightly know how 
to categorize it. Is it a problem with FreeBSD? Or a formerly harmless 
issue with PHP that became more 'visible' on the newer OS version?
Is it something in conjunction with kvm, or a timing issue...

Any help appreciated.

Thanks,

Martin