1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
|
<think>
Alright, I'm trying to classify this bug report from QEMU. Let me go through it step by step.
The user provided a bug report where the guest OS displays an error: "FATAL: Module scsi_wait_scan not found" followed by a call trace. The steps to reproduce include using QEMU with KVM acceleration and various devices, including virtio-blk-pci and virtio-net-pci.
First, I'll look at the categories provided: mistranslation, assembly, other, device, graphic, socket, network, KVM, boot, vnc, debug, files, permissions, performance, kernel, peripherals, hypervisor.
The error message mentions a missing module related to SCSI. However, in this case, the setup uses virtio-blk-pci for the disk, which is a VirtIO block device, not traditional SCSI. So maybe it's not directly about SCSI but something else.
Looking at the call trace, it shows panic and do_exit functions, which are part of the kernel's error handling. The EAX register has ffffffda, which often indicates a panic in the kernel with a specific reason code. The context suggests this is happening within the guest OS, possibly during boot.
The user suspects it's a QEMU regression due to a specific commit. The command line uses KVM for acceleration and VirtIO devices. The error occurs at boot time, so categories like 'boot' or 'hypervisor' come to mind.
But wait, the error is about a missing module in the guest OS. Since it's related to SCSI but using VirtIO block, maybe it's an issue with how VirtIO devices are handled. Alternatively, since KVM is involved, it could be a hypervisor-related problem.
The device specified in the command line is virtio-blk-pci and virtio-net-pci. However, the error message mentions "scsi_wait_scan," which might not directly relate to these VirtIO devices unless there's an underlying issue with how certain drivers are loaded or initialized in the guest.
Considering all this, I'm leaning towards 'hypervisor' because the issue is within the guest OS's boot process and relates to device handling under KVM. The problem seems to be a regression introduced by a specific QEMU commit, affecting how devices are managed during boot.
</think>
The bug report involves a fatal error during the guest OS boot, specifically a missing module related to SCSI scanning. Given that VirtIO devices are used and the issue occurs under KVM with a known bad commit in QEMU, this points towards a hypervisor-level problem affecting device initialization during the boot process.
hypervisor
|