Re: Dom0 panics when linux DomU tries to access to virtual disk

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

	I beg your pardon if this message was posted twice, but I think mailman
has refused this message due to screen capture in attachement.

Manuel Bouyer a écrit :
> On Mon, Oct 26, 2020 at 09:09:21PM +0100, BERTRAND Joël wrote:
>>> you'll have to tell me where to find devuan_beowulf_3.0.0_amd64-netinstall.iso.
>>
>> 	Here :
>>
>> https://mirror.serverion.com/devuan/devuan_beowulf/installer-iso/
>>
>> 	But I not sure this bug is related to this specific installer.
> 
> Not sure, but I run linux DomUs on a regular basis
> 
> Anyway I couldn't reproduce the panic with this installer either.
> Using the default choise, the domU installed and it booting from disk fine.
> 

	I try to reproduce this panic.

schwarz# sysctl -w ddb.onpanic=1
ddb.onpanic: 0 -> 1
schwarz# sysctl -w ddb.commandonenter=bt
ddb.commandonenter: show registers -> bt
schwarz# uname -a
NetBSD schwarz.systella.fr 9.1 NetBSD 9.1 (XEN3_DOM0_CUSTOM) #3: Sat Oct
24 14:07:08 CEST 2020
root%legendre.systella.fr@localhost:/usr/src/netbsd-9/obj/sys/arch/amd64/compile/XEN3_DOM0_CUSTOM
amd64

	dom0 kernel is customized as config contains :

config      netbsd  root on re0 type nfs

schwarz# df -h
Filesystem                    Size       Used      Avail %Cap Mounted on
192.168.10.128:/srv/schwarz   523G       244G       253G  49% /
192.168.10.128:/home          3.6T       2.4T       1.0T  70% /home
192.168.10.128:/usr/src       252G        38G       201G  15% /usr/src
kernfs                        1.0K       1.0K         0B 100% /kern
ptyfs                         1.0K       1.0K         0B 100% /dev/pts
procfs                        4.0K       4.0K         0B 100% /proc
tmpfs                         128M       4.0K       128M   0% /var/shm

	Please not that this XEN hypervisor is a diskless server.

	DomU's root filesystems are contained in files (/vms directory). If I
mount these files directly in NFS server with vndconfig, I can export
for example vnd0 as regular iSCSI device. In this configuration, domU
rootfs is not a file but a block device (/dev/sd1d in my case) and I can
install this domu until grub installation. Grub ends without error but
xen complains about a unbootable disk. I don't understand why disk
remains unbootable... but it's another question.

schwarz# xl list
Name                       ID   Mem VCPUs      State   Time(s)
Domain-0                   0   512     1     r-----       9.0
schwarz# cat devuan_1.conf
name = "Devuan"
memory = 4096
vcpus = 4
vif = [
        'mac=00:16:3E:6D:D1:01,bridge=bridge0'
]
disk = [
        '/vms/devuan_1.img,raw,xvda,rw',
# For te same disk image exported from server as
# iSCSI block device
#       '/dev/sd1d,,xvda,rw',
        '/vms/devuan_beowulf_3.0.0_amd64-netinstall.iso,,xvdc,cdrom'
]
vnclisten = "192.168.10.106"
keymap = "fr"
vnc = 1
type = "hvm"
firmware = "seabios"
boot = "d"

	Please note that this domU uses HVM virtualisation.

schwarz# xl create devuan_1.conf
Parsing config from devuan_1.conf
schwarz# xl list
Name                             ID   Mem VCPUs      State   Time(s)
Domain-0                         0   512     1     r-----      16.6
Devuan                           3  4096     1     -b----       0.5
schwarz#

schwarz# pkg_info xen*
Information for xenkernel413-4.13.1nb2:
...
Information for xentools413-4.13.1nb3:
...

	Information : when I start domU, it mounts disk image as vnd0 :
[   113.019337] xbd backend: attach device vnd0d (size 33554432) for
domain 1
[   114.839315] tap0: Ethernet address f2:0b:a4:58:5d:0e
[   120.769242] xvif1i0: Ethernet address 00:16:3e:6e:d1:01
schwarz# vndconfig -l
vnd0: dev 11,1 inode 7261449
vnd1: not in use
vnd2: not in use
vnd3: not in use

	I have reproduced this panic. System quickly reboots even if I
configure kernel to enter in ddb. In attachement, you will find a screen
capture. This panic seems to be related to nfs. Maybe you have to use a
disk image from a remote NFS server.

	Best regards,

	JKB