I have a set of J1900 hosts running 11.0-RELEASE-p1 that experience
seemingly random panics. The panics are all basically the same:

Fatal trap 12: page fault while in kernel mode
fault code = supervisor read data, page not present

Adding workloads to the hosts seems to increase panic frequency, but the
panics have also occurred on completely idle hosts. Similarly, uptime
when panicking has been as low as minutes, and as high as ~620 days.

For reasons, it has not been possible to extract a coredump from these
hosts, nor practical to run memtest on them or upgrade them to a newer
release. About 1% of our hosts are affected each day, so we've just been
living with the problem.

However, while testing 12.0 on the same hardware, I encountered the same
panic and was able to capture the core dump. (See below.)

All of my Google-fu on this panic has turned up threads suggesting the
problem is hardware, but there are two problems with that idea...

One, memtest has turned up no errors on 12.0 host I witnessed the panic
on.

Two, a small number of systems on the same hardware are running
10.3-RELEASE, and have experienced no panics in their history. Panics
have only happened on 11s, and now 12.

kgdb output from the panic follows. (This particular host was in the
middle of rebooting when it panicked.)

Hoping someone here has some insight. My uninformed wild-ass guess is
something relating to spectre/meltdown fixes.

Thanks,


-Snow

I've been running 10.x, 11.x and 12.0 for a while on several J1900s, namely 
ASRock Q1900M and Q1900M Pro3 boards.

All of them are getting a good beating on occassion, running for example 
poudriere on top of GELI and ZFS software RAIDs attached to the onboard 2-port 
ACHI SATA controller and Marvel based PCIe 4-port SATA controllers.

I've outfitted all of them with 4-port Intel PRO/1000 PCIe driven by igb(4), 
and am not using the onboard re(4) NICs.

I can't recall ever seeing a panic like you described. Could you share a full 
dmesg and what mainboard(s) you are using ?

MfG CoCo

_______________________________________________
freebsd-stable@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"

Reply via email to