Johannes Thumshirn

vmalloc of size 0 sounds suspiciously broken. I’ll look into it

So this get's interesting: ```C btrfs_get_dev_zone() unsigned int nr_zones = 1; `-> btrfs_get_dev_zones(/*...*/, &nr_zones); `-> blkdev_report_zones(/* ... */, *nr_zones, /* ... */); `-> disk->fops->report_zones(/* ... */, nr_zones, /* ... */);...

@Motophan can you run the following bcc python script while re-running the test and collect dmesg afterwards? ```python #!/usr/bin/env python3 from bcc import BPF from time import sleep bpf_program =...

So the trace doesn’t show anything unusual. Still no idea why we are OOM here.

@Motophan can you please apply this debug patch, so we can see why we're falling down to a vmalloc of 0. ```diff diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index 26af5ab7d7c1..7b94a1cabafd 100644 ---...

OK then let's try something different here. Can you re-test with the Nvidia kernel module unloaded? It is an out-of-tree module and we can't be sure it's not doing stupid...

This is where everything goes south: > Apr 28 18:53:14.763628 archlinux kernel: ata6.00: exception Emask 0x0 SAct 0x104000 SErr 0xc0000 action 0x6 frozen > Apr 28 18:53:14.763963 archlinux kernel: ata6:...

I've created #88 which fixes it for me on Fedora 38: QEMU emulator version 7.2.1 (qemu-7.2.1-2.fc38)