Random segfaults on KVM if zfs snapshots were taken between qcow2 snapshot and blockcommit [not zfs related]

System information

Describe the problem you're observing

I use to replicate my zfs pools to a remote host via ssh. Those zfs pools contain Qemu KVM virtual machines. Randomly, one of my VMs stops and I get KVM segfaults while replication is ongoing. All other VMs continue to work, and I only get that segfault from one VM.

I've stopped replication via ZFS for 3 months, and didn't get any segfault. Today I've redone some replications, and got a segfault rightaway while replicating, so I'm pretty sure there's a link. Also, I have 3 other VMs which execute the same guest os and software, and none of them encountered any issues, while the VM that gets crashed by KVM segfault is the most IO intensive.

i use the following command (generated by bzfs tool )

'sh', '-c', ' zfs send --raw --compressed --large-block -I private_vm/prometheus@production_mar_2025-08-24_17:00:01_hourly private_vm/prometheus@production_mar_2025-08-24_19:04:02_5minutely |  LC_ALL=C pv --progress --timer --eta --fineta --rate --average-rate --bytes --interval=1 --width=120 --buffer-size=2M --force \'--name= 162MiB/     0B/ 162MiB\' --size=170315352 2>> /root/bzfs-logs/2025-08-24/zrun_2025-08-24_19:08:14-03jb_atn.pv | zstd -c -1 | mbuffer -s 2097152 -q -m 128M  | ssh -oBatchMode=yes -oServerAliveInterval=0 -x -T -S /root/.ssh/bzfs/s34113@1756055294040903359@135989939925@REMOTEIP@ REMOTEIP \'sh -c "mbuffer -s 2097152 -q -m 128M | zstd -dc | zfs receive -u -s private_vm/prometheus"\''

Describe how to reproduce the problem

I do understand that my report cannot be easily reproduced.

What I hope by creating this issue is for other people to perhaps find this issue and comment if they have encounter the same errors, and perhaps have found a workaround / resolution.

Any insight / idea here would be helpful. Willing to do some tests. I already encountered the same issue when I ran zfs-2.2.7 on that hypervisor.

Include any warning/errors/backtraces from the system logs

dmesg from the host hypervisor

[807710.528708] qemu-kvm[14491]: segfault at 7f3340ae6e10 ip 000055b3a7b19367 sp 00007ffe6b9b5f30 error 6 in qemu-kvm[55b3a7554000+77b000] likely on CPU 32 (core 8, socket 0)
[807710.529344] Code: 8d bc 24 e0 df ff ff 64 4c 89 38 49 8d 44 24 f8 48 89 44 24 08 64 48 8b 04 25 28 00 00 00 48 89 04 24 49 89 44 24 f8 0f 57 c0 <41> 0f 11 84 24 e0 df ff ff 4c 8d 6f 30 4d 8d b4 24 f0 df ff ff 4c

There are no errors in the guest logs, nor any other relevant logs on host than this.