Re: Hard freeze during `pkg_add -u` on -current

On 2024-07-05, Ronald Dahlgren <ronald.dahlgren@gmail.com> wrote:
> --000000000000cbf9af061c80339e
> Content-Type: text/plain; charset="UTF-8"
> Content-Transfer-Encoding: quoted-printable
>
> Hello,
>
> On July 2nd, I updated a machine to the latest snapshot and rebooted. It
> came back without issue. I then issued `pkg_add -U`. This machine was last
> updated on June 6th, so not terribly long ago. Partway during the process,
> the disk indicated it was full (not true) and no commands were available
> (ls, cd, etc). Unable to do anything, I terminated my SSH session and
> attempted to reconnect. The machine failed to respond to pings. I had
> someone onsite reboot the machine. It then came back up. I did not try the
> `pkg_add -u` command again. Inspection showed that partitions had plenty of
> available space and inodes.
>
> The daily insecurity output that ran the following day, on Wednesday the
> 3rd, had this unusual snippet:
>
> ```
> vmm-firmware-1.16.3p0 firmware binary images for vmm(4) driver
> -xz-5.4.5 library and tools for XZ and LZMA compressed files
> +xz-5.6.2
> <FD>/??^L???.???<C5>/??<F4>???..??<FE>/??$???+DESC???<FF>/?????
> +CONTENTS????0??<C4>??^L+REQUIRED_BY??????????????????????????????????????=
> ???????????????????????????????????????????????????????????????????????????=
> ???????????????????????????????????????????????????????????????????????????=
> ???????????????????????????????????????????????????????????????????????????=
> ???????????????????????????????????????????????????????????????????????????=
> ??????????????????????????????????????????????????????????????????????
> zsh-5.9p0 Z shell, Bourne shell-compatible
> ```

The filesystem holding /var/db/pkg has some corruption.
I'd try running pkg_check and allow it to repair, reinstall xz
"pkg_add -r -D installed xz", and see how you get on.

> Given the package with the wacky description is `xz`, I'm more concerned
> than I would be otherwise.

The same could have happened to any package, there's nothing special
about xz there.

> I can see in `/var/log/messages` the snapshot update occurred without
> issue. Logs after the physical reboot show no core dump and only have
> complaints about filesystems not being properly unmounted - expected when
> the plug is pulled.
>
> Are there any other logs I can check and share to help get to the bottom of
> this? The impacted computer has been running current and humming along
> happily in a network closet for over a year.

Not sure about the disk full message (spurious seems unlikely - if space
is ok, is some filesystem tight on inodes? df -hi) or the hang.

--
Please keep replies on the mailing list.