On Thu, 2021-06-03 at 08:34 -0400, Bruce Ashfield wrote:

Interesting.

The last two stalls/traps that you sent were both x86-64. Are we only
seeing this on x86 ?

Not sure. We have seen some builds fail but I haven't dived into properly
get down to the real cause. We run many more 64 builds than plain x86.

I'll have another look at the patches we carry, but there really
shouldn't be anything that could trigger something like this (famous
last words).

If these are happening regularly enough, we could change the
qemux86-64 branch to v5.10/base temporarily. That would be a pure
upstream -stable, and would help rule out the yocto patches if the
issues still occur.

They're happening regularly enough to be really annoying but not 
regularly enough to be sure any given fix fixes things :(

I thought we had resolved the ltp hang but see my other email, I think 
we just got lucky/unlucky with the kernel uprev and some good builds.
The traceback in mkdir is interesting and worrying.

I think these are related and we sometimes crash the kernel sometimes not.
There is something very wrong somewhere. I'm happy to try changing 
qemux86-64 and seeing if we continue to see issues or not but it will
take a week or so to be sure.

Cheers,

Richard