NetBSD 9.2 (STABLE) kernel hangs without panic or ddb

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Hello,

My main server runs NetBSD 9.2 (STABLE). It has 16 GB of RAM, an i7 CPU 
(i7-4770), a lot of disks (wd0/wd1 configured as ccd [swap for diskless 
workstation], wd2/wd3 as raid level 1 [system], wd4/wd5/wd6 as raid 
level 5 [home] and an external NAS [iscsi target, bacula archives]). It 
runs a customized kernel as I have configured ALTQ. I have noticed that 
I cannot stop altqd or reload configuration without trouble : when altqd 
stops or restarts, it always takes 100% of a CPU...

Ethernet link used to access to NAS is configured on both sides with 
MTU=9000 (wm0) :

wm0: flags=0x8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 9000
        capabilities=7ff80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx>
        
capabilities=7ff80<TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx,TCP6CSUM_Rx>
        capabilities=7ff80<TCP6CSUM_Tx,UDP6CSUM_Rx,UDP6CSUM_Tx,TSO6>
        enabled=0
        ec_capabilities=17<VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,EEE>
        ec_enabled=2<VLAN_HWTAGGING>
        address: b4:96:91:92:77:6e
        media: Ethernet autoselect (1000baseT full-duplex)
        status: active
        inet 192.168.12.1/24 broadcast 192.168.12.255 flags 0x0
        inet6 fe80::b696:91ff:fe92:776e%wm0/64 flags 0x0 scopeid 0x1

I don't know when I have seen this issue the first time, but if I 
remember, 9.0 ran fine. Maybe 9.1 also. With 9.2, when bacula starts its 
monthly archive (around 50 files, 50 GB each on a ffs2 filesystem), 
kernel can crash, randomly. Since yesterday, I have seen two crashes.

System doesn't respond anymore, kernel doesn't enter in ddb even if ddb 
is set in sysctl, magic request doesn't do anything. Kernel doesn't 
panic. I have tried to access to serial console, no answer. No dump 
file. System only stops, maybe on a mutex or a spinlock.

If I umount iscsi target, system seems to be stable. I don't know it 
this issue is related to iscsi initiator or related to an interaction 
between iscsi and, maybe, ccd driver... And I cannot remove functions 
from this server to test.

I have to fix this issue, but why ? I don't have any usable 
information...

Best regards,

JB