Re: new LOR tcphash, in6_ifaddr_lock

On Thu, 4 Dec 2025, Bjoern A. Zeeb wrote:

> On Thu, 4 Dec 2025, Jonathan T. Looney wrote:
>
>> On Wed, Dec 3, 2025 at 2:36 PM Jonathan T. Looney <[email protected]> wrote:
>> 
>>> This looks like a duplicate report to PR 289184 (
>>> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=289184).
>>> 
>>> If you can reproduce this, can you try the patch I included in the bug
>>> audit trail?
>>> 
>> 
>> 
>> Actually,
>> 
>> Given that this has popped up several times over the past several years, I
>> probably should just commit the patch to main so we have a better chance of
>> getting a backtrace for the call path that produces the true LOR. I’ll open
>> a review to do that.
>
> Sorry, I didn't have the capacity to try to reproduce this last night 
> anymore.
> I do beleive I know how I triggered it so I'll play around a bit to see if I 
> can
> again.  If I have a recipe I'll apply your patch and try again.
>
> I'll let you know &

When I cam back to the console of the (virtual) machine I had the LOR once
more there from over night;  I rebooted and was able to reproduce it just
running rtsol, bringing up vtnet, and running pkg update.

I then rebuilt the kernel with the patch from the PR applied.

I was no longer able to reproduce that (reverse) LOR after but got others *sigh*
(see net@).   I let the vmachine run over night with the same constant network
load as I had before but nothing happened.

I'll keep an eye on it.  I'd suggest if you commit the patch from the PR, that
you add a comment about the PR and that (for now) it is there to find the
opposite order callgraph.

/bz

-- 
Bjoern A. Zeeb                                                     r15:7