vbox VirtualBox: FreeBSD guest with UFS: disk I/O ceasing with PIIX4, PIIX3 and other virtual controllers

[grahamperrin]

Below, why are there log entries for PIIX3 when the controller is set to PIIX4?



I have been trying (amongst other things) to boot this 12.⋯ guest temporarily in safe mode. Neither of the commands below seem to be effective – note the multiple CPUs, above, in htop(8):

nextboot -o "safemode_enable"

nextboot -o safemode_enable="YES"

Seeing the logged resets, I wonder whether the OS is restarting non-gracefully before commands can take effect.

 

[grahamperrin]

I realised, the base system was not properly up-to-date (presumably the result of an unwanted reset during a prior OS upgrade routine):



Restarting, the system appeared to do nothing for an extraordinarily long time whilst waiting for bufdaemon to stop:



With the system properly upgraded to 12.3-RELEASE, I began upgrading packages, hoping for the system to not freeze (with another reset). Unfortunately:



– disk I/O ceased.

00:10:49.089256 VMMDev: vmmDevHeartbeatFlatlinedTimer: Guest seems to be unresponsive. Last heartbeat received 4 seconds ago

No known data errors for the pool where I store the .vhd:

root@mowa219-gjp4-8570p-freebsd:~ # zpool status -x
  pool: Transcend
 state: ONLINE
status: One or more devices has been taken offline by the administrator.
        Sufficient replicas exist for the pool to continue functioning in a
        degraded state.
action: Online the device using 'zpool online' or replace the device with
        'zpool replace'.
  scan: scrub repaired 0B in 01:32:30 with 0 errors on Sun Dec 12 16:25:35 2021
config:

        NAME                   STATE     READ WRITE CKSUM
        Transcend              ONLINE       0     0     0
          gpt/Transcend        ONLINE       0     0     0
        cache
          gpt/cache-transcend  OFFLINE      0     0     0

errors: No known data errors
root@mowa219-gjp4-8570p-freebsd:~ #

 

[grahamperrin]

Again, disk I/O ceased. This time, with PIIX3:



With this 12.3 guest, it's sometimes strangely difficult to get a non-interrupted run of fsck -y in single user mode.

More generally, I get a sense that FreeBSD guests are more likely (than a Windows 10 guest with the virtual AHCI controller) to get this problem with disk I/O ceasing.

For FreeBSD guests, which type of virtual controller might be least likely to have problems?

(I already tried AHCI with this guest. I'll retry.)

 

[grahamperrin]

Tentatively solved.

After various experiments and numerous failures, I achieved an uptime of over two hours, during which I reinstalled 1,029 packages ( pkg upgrade -f), with the following combination:

• synchronous for UFS in the guest
• AHCI virtual controller
• host I/O cache enabled
• EFI
• nested paging enabled
• PAE/NX disabled.

Output from mount in the guest:

/dev/gpt/rootfs on / (ufs, local, synchronous, journaled soft-updates)
devfs on /dev (devfs, local, multilabel)
fdescfs on /dev/fd (fdescfs)
procfs on /proc (procfs, local)
map -hosts on /net (autofs)

I can't imagine why a UFS-related change in the guest should be a factor, but it seemed that way.

Again, an extraordinarily long wait for bufdaemon to stop:



– but eventually, the stop did succeed.

---

From the opening post:

why are there log entries for PIIX3 when the controller is set to PIIX4?

For now, I'm happy for that to remain unanswered.

 

[grahamperrin]

I forgot about this problem, at some point (probably after moving guests from one hard disk drive to another) I changed settings, today the FreeBSD guest machine was terribly unreliable.

(A Windows 10 guest on the same drive is reliable.)

Visually stuck during a small download, disk activity continued long afterwards:



Eventually, disk activity ceased:

Tentatively solved. … the following combination: …

Unfortunately, that combination is no longer a workaround.

Stuck again (second screenshot):

 

[grahamperrin]

… a sense that FreeBSD guests are more likely (than a Windows 10 guest …) to get this problem …

Another death of the FreeBSD 12.3 guest:

… No known data errors for the pool where I store the .vhd: …

It's on a mobile hard disk drive, on USB.

hselasky@ and others with USB expertise: I'm clutching at straws here.

Can you think of any reason why FreeBSD as a guest in a virtual machine might be prone to dying, only when the VirtualBox host stores the VM data over USB?

% geom disk list da2
Geom name: da2
Providers:
1. Name: da2
   Mediasize: 500107862016 (466G)
   Sectorsize: 512
   Stripesize: 4096
   Stripeoffset: 0
   Mode: r1w1e3
   descr: StoreJet Transcend
   lunid: 5000000000000001
   ident: (null)
   rotationrate: unknown
   fwsectors: 63
   fwheads: 255

% sudo hw-probe -all -disable sysctl -upload
grahamperrin's password:
Probe for hardware ... Ok
Reading logs ... Ok
^C
% sudo hw-probe -all -upload
Probe for hardware ... Ok
Reading logs ... Ok
ERROR: the probe is larger than 500K

ERROR: failed to upload probe
% sudo hw-probe -all -disable sysctl -upload
Probe for hardware ... Ok
Reading logs ... Ok
Uploaded to DB, Thank you!

Probe URL: https://bsd-hardware.info/?probe=0c73871c49
%

<https://bsd-hardware.info/?probe=0c73871c49>