bhyve - guest CentOS 7 - error in guest console - kernel:do_IRQ: 1.35 No irq handler for vector

Hi )

HOST
FreeBSD 12.1-RELEASE-p2 #0 r357852M
vm-bhyve-1.3.0

GUEST

Code:
loader="uefi"
cpu=2
cpu_sockets=1
cpu_cores=1
cpu_thread=2
memory=2G
network0_type="virtio-net"
network0_device="tap6"
network1_type="virtio-net"
network1_device="tap7"
disk0_type="ahci-hd"
disk0_name="centos7.hdd"
graphics="yes"
graphics_listen="192.168.240.5"
graphics_port="5905"
graphics_res="1280x720"
graphics_wait="no"

CentOS 7(7.7.1908)
Update to kernel 5.5.7 (kernel.org)
After reboot in console error and system working very very slow.

Code:
kernel:do_IRQ: 1.35 No irq handler for vector

When use kernel 5.2.9 - it's ok.

About this error message.
Initiator of this error - service irqbalance

Code:
GRUB_CMDLINE_LINUX_DEFAULT=".... pci=nomsi,noaer"
Doesn't effect

When stop service irqbalance, disable service irqbalance and reboot.
No error, system work, but work only one core.

May be anybody have this problem and solved, tell please ?
 
Last edited by a moderator:
Hello,
I do have the same configuration (centos on bhyve) as you. Also the same error (no irq handler) appears on my system. Thanks to your post I came to the conclusion that the problem lies something in an kernel update.

I performed the steps you described and it seems that it solved the error. The error
Code:
kernel:do_IRQ: 1.35 No irq handler for vector

didn't appear in the console anymore.

The steps I performed
  1. add quiet pci=nomsi,noaer to /etc/default/grub
  2. generate a new gub.cfg using grub2-mkconfig -o /boot/efi/EFI/fedora/grub.cfg
Did you generate a new grub.cfg using grub2-mkconfig -o /boot/efi/EFI/fedora/grub.cfg? This is required and added as a comment to grub.cfg.


My configuration

/etc/default/grub
Code:
GRUB_TIMEOUT=5
GRUB_DISTRIBUTOR="$(sed 's, release .*$,,g' /etc/system-release)"
GRUB_DEFAULT=saved
GRUB_DISABLE_SUBMENU=true
GRUB_TERMINAL_OUTPUT="console"
GRUB_CMDLINE_LINUX="resume=/dev/mapper/fedora-swap rd.lvm.lv=fedora/root rd.lvm.lv=fedora/swap rhgb quiet pci=nomsi,noaer"
GRUB_DISABLE_RECOVERY="true"
GRUB_ENABLE_BLSCFG=true

/usr/local/bhyve/fedoraserver1/fedoraserver1.conf
Code:
loader="uefi"
cpu="4"
memory="16G"
disk0_type="virtio-blk"
disk0_dev="zvol"
disk0_name="disk0"
network0_type="virtio-net"
network0_switch="custompublic"
network0_device="tap0"
graphics="yes"
graphics_port="5901"
graphics_listen="10.0.1.12"
graphics_res="1920x1080"
graphics_wait="no"
xhci_mouse="no"
uuid="xxx"
network0_mac="xx:xx:xx:xx:xx:xx"

/boot/efi/EFI/fedora/grub.cfg
Code:
# DO NOT EDIT THIS FILE
#
# It is automatically generated by grub2-mkconfig using templates
# from /etc/grub.d and settings from /etc/default/grub

Greetings
Alex
 
Well,
The information in my last post is wrong. The errors were just suppressed.

There was a FreeBSD update concerning bhyve. The following problem was fixed:

The solution which seemingly worked for my supermicro board was descibed in a centos forum. Enable the following settings in Bios helped

Code:
Advanced -> NB Configuration -> IOMMU (AMD) or VTd (Intel) (change to Enabled).
Advanced -> PCIe/PCI/PnP Configuration -> SR-IOV Support (change to Enabled).

https://forums.centos.org/search.php?author_id=116788&sr=posts&sid=7276f02be569d9542dc583fde4d2a968

Some background information concerning VTd or IOMMU.
https://us.informatiweb.net/tutorials/it/bios/enable-iommu-or-vt-d-in-your-bios.html

There also was a freebsd-update concerning bhyve. As the problem description is talking about OpenBSD as bhyve guest system, I am not sure if the freebsd-update had some impact on fixing my problem. But it might be hence I am sharing the information
Code:
II.  Problem Description

When an attempt is made to pass through a PCI device to a bhyve(8) VM (causing
initialization of IOMMU) on certain Intel chipsets using VT-d the PCI bus
stops working entirely resulting in a host crash.  This issue occurs at least
on the Intel Skylake series processors and those released later.

A device passed through to a guest VM running OpenBSD at least since version
6.4 on both AMD and Intel processors may not fully work in the guest.  OpenBSD
issues 4-byte PCI configuration-space register reads and writes to consecutive
2-byte fields, which were not handled correctly by bhyve(8).

https://www.mail-archive.com/freebsd-announce@freebsd.org/msg00971.html

Regards
Alex
 
Back
Top