Crash Dump - 1st time, how or where should I share ?

ngr2001

I've never had PF crash and produce a dump before, how or where can I share the results.

No Major changes on my end expect for a few tweaks to my limiter / Codel setup, so this is very unexpected.

pfSense Plus
26.03-RELEASE (amd64)

ngr2001

textdump.tar.0

stephenw10

Backtrace:

db:1:pfs> bt
Tracing pid 9 tid 100229 td 0xfffff80004fad780
kdb_enter() at kdb_enter+0x33/frame 0xfffffe00fc4949c0
panic() at panic+0x43/frame 0xfffffe00fc494a20
trap_pfault() at trap_pfault+0x3cf/frame 0xfffffe00fc494a70
calltrap() at calltrap+0x8/frame 0xfffffe00fc494a70
--- trap 0xc, rip = 0xffffffff8124e8b9, rsp = 0xfffffe00fc494b40, rbp = 0xfffffe00fc494eb0 ---
vm_pageout_worker() at vm_pageout_worker+0xb09/frame 0xfffffe00fc494eb0
vm_pageout() at vm_pageout+0x1d7/frame 0xfffffe00fc494ef0
fork_exit() at fork_exit+0x7b/frame 0xfffffe00fc494f30
fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe00fc494f30
--- trap 0xa5a5a5a5, rip = 0xa5a5a5a5a5a5a5a5, rsp = 0xa5a5a5a5a5a5a5a5, rbp = 0xa5a5a5a5a5a5a5a5 ---

Pretty generic. Nothing specific shown.

But the message buffer is completely full of:

fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536
fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536
fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536
fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536
fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536
fq_codel_enqueue over limit
fq_codel_enqueue maxidx = 536

So it does look like some fq_codel issue. Something there is being continually overrun.

ngr2001

@stephenw10

I've been running the bufferbloat test from waveform a lot lately, just experimenting with different settings etc.

When running those test I've noticed the logs will often show fq_codel_enqueue over limit
fq_codel_enqueue maxidx = xyz

I assume that would be normal as the queues get full etc.. ?

That said, I did fairly recently change my queue management algorithm from tail drop to PIE. My goal is to have the lowest possible latency on my Cable (DOCSIS) connection. My current setting are below and they did seem to lower my latency 1-2ms giving me a more consistent A+ score where as before I would normally score A and seldom achieve A+

I made that setting change at least a week prior to this dump though, so far no other issues, I am open to any suggestions though.

stephenw10

Hmm, it could be just that testing. The dmesg logs have no time stamps so it can be some time before the panic if nothing else is logged in between.