Re: kern/57002: Starting NetBSD in a HVM DomU crashes Dom0

[Brad Spencer <brad%anduin.eldar.org@localhost>]

The following reply was made to PR kern/57002; it has been noted by GNATS.

From: Brad Spencer <brad%anduin.eldar.org@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: kern-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost
Subject: Re: kern/57002: Starting NetBSD in a HVM DomU crashes Dom0
Date: Fri, 09 Sep 2022 15:50:45 -0400

 Hauke Fath <hf%spg.tu-darmstadt.de@localhost> writes:
 
 >>Number:         57002
 >>Category:       kern
 >>Synopsis:       Starting NetBSD in a HVM DomU crashes Dom0
 >>Confidential:   no
 >>Severity:       critical
 >>Priority:       high
 >>Responsible:    kern-bug-people
 >>State:          open
 >>Class:          sw-bug
 >>Submitter-Id:   net
 >>Arrival-Date:   Fri Sep 09 17:05:00 +0000 2022
 >>Originator:     Hauke Fath
 >>Release:        NetBSD 9.3
 >>Organization:
 > Technische Universitaet Darmstadt
 >>Environment:
 > 	
 > 	
 > System: NetBSD Treffauer 9.3 NetBSD 9.3 (FEUERKOGEL_DOM0-$Revision$) #0: Tue Aug 9 15:30:14 CEST 2022 hf@Hochstuhl:/var/obj/netbsd-builds/9/amd64/sys/arch/amd64/compile/FEUERKOGEL_DOM0 amd64
 > Architecture: x86_64
 > Machine: amd64
 > % pkg_info | grep xen
 > xenkernel413-4.13.4nb1 Xen 4.13.x Kernel
 > xentools413-4.13.4nb6 Userland Tools for Xen 4.13.x
 > %
 >
 >>Description:
 >
 > 	Starting a Xen HVM DomU from a NetBSD 9 install iso crashes
 > 	the Dom0 with
 >
 > [...]
 > [ 296.6180677] xbd backend: attach device vg0-hazel_root (size 20971520) for domain 2
 > [ 300.3380380] pmap_unwire: wiring for pmap 0xffffb48004fcf060 va 0x7c0c00107000 did not change!
 > [ 334.8277708] pmap_unwire: wiring for pmap 0xffffb48004fcf060 va 0x7c0c00151000 did not change!
 > [ 334.8277708] pmap_unwire: wiring for pmap 0xffffb48004fcf060 va 0x7c0c00152000 did not change!
 > [ 334.8277708] pmap_unwire: wiring for pmap 0xffffb48004fcf060 va 0x7c0c00153000 did not change!
 > [ 337.3877490] pmap_unwire: wiring for pmap 0xffffb48004fcf060 va 0x7c0bf5dfc000 did not change!
 > [ 337.3877490] panic: kernel diagnostic assertion "pg->wire_count != 0" failed: file "/u/netbsd-9/src/sys/uvm/uvm_page.c", line 1618 
 > [ 337.3877490] cpu0: Begin traceback...
 > [ 337.3977488] vpanic() at netbsd:vpanic+0x143
 > [ 337.3977488] kern_assert() at netbsd:kern_assert+0x48
 > [ 337.3977488] uvm_pageunwire() at netbsd:uvm_pageunwire+0x81
 > [ 337.3977488] uvm_fault_unwire_locked() at netbsd:uvm_fault_unwire_locked+0x135
 > [ 337.3977488] uvm_fault_unwire() at netbsd:uvm_fault_unwire+0x29
 > [ 337.3977488] physio_done() at netbsd:physio_done+0x77
 > [ 337.3977488] workqueue_worker() at netbsd:workqueue_worker+0xce
 > [ 337.3977488] cpu0: End traceback...
 >
 > [ 337.3977488] dumping to dev 18,1 (offset=33554431, size=0): not possible
 > [ 337.3977488] rebooting...
 > (XEN) Hardware Dom0 shutdown: rebooting machine
 > (XEN) Resetting with ACPI MEMORY or I/O RESET_REG.[-- hf@localhost detached -- Fri Sep  9 18:44:46 2022]
 >
 >
 > 	
 >>How-To-Repeat:
 > 	
 >
 > 	Set up a Xen DomU HVM configuration with an lvm disk device and a
 > 	NetBSD 9 iso image. Have it render to VNC.
 >
 > 	Watch the boot freeze while probing the disk device, then
 > 	watch the Xen host crash and reboot.
 >
 > 	The DomU configuration is
 >
 > type = "hvm"
 > firmware = "seabios"
 > kernel = "/usr/pkg/libexec/xen/boot/hvmloader"
 > memory = 1024
 > vcpus = 1
 > acpi = 1
 > name = "Hazel"
 > ioemu = [ 'mac=00:16:3e:NN:MM:OO, bridge=bridge1' ]
 > disk = [ 'backendtype=phy, vdev=hda, target=/dev/mapper/vg0-hazel_root',
 >          'vdev=hdb, devtype=cdrom, target=/var/tmp/NetBSD-9.0_STABLE-amd64.iso']
 > vga="cirrus"
 > vnc=1
 > vnclisten="127.0.0.1"
 > vncdisplay=10
 > serial = "pty"
 >
 >>Fix:
 > 	
 > 	Yes, please.
 >
 >>Unformatted:
 >  	
 >  	
 
 
 I saw this too with Xen 4.15 with a 9.2_STABLE DOM0 and mentioned this
 all in a post on port-xen.  The problem goes away for me if I use file
 backstore and not LVM LVs.  Further, I do not always get a immediate
 crash, but the DOM0 *WILL* panic at some point usually in the 24 hours
 or so after I messed with a HVM that trips this.
 
 
 
 -- 
 Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org