[systemd-devel] /dev or /tmp disappears when program is running

[Lennart Poettering]

On Wed, 09.09.15 10:50, Steven Shiau (steven at nchc.org.tw) wrote:

> 
> On 9/8/2015 PM 10:05, Lennart Poettering wrote:
> >On Tue, 08.09.15 20:45, Steven Shiau (steven at nchc.org.tw) wrote:
> >
> >>Dear systemd developers,
> >>I encountered a very weird issue, i.e. the /dev or /tmp disappears when
> >>program is running.
> >>The client system is nfsroot (diskless), running on the Debian Sid with
> >>Linux kernel 4.1.0-2-amd64. The PXE/NFS server is a Debian Live system
> >>running NFS server by unfs3 [1], its mounting status is attached as
> >>"server-table.txt", and the client's initial mounting status is attached
> >>as "client-table.txt". You can see /dev/ is mounted as devtmpfs. This is
> >>quite normal.
> >>When a system deployment program (clonezilla [2]) is run in the diskless
> >>client machine, suddenly the /dev/ is gone, as shown in the attached
> >>file "client-table-dev-missing.txt". You can see "/dev/" is missing. To
> >>narrow down the issue, I have switched to sysvinit, but still use
> >>systemd-udevd on this Debian Sid system.
> >
> >What makes you think this is a systemd issue, and not a clonezilla issue?
> >
> >Lennart
> >
> BTW, attached please find the systemd-udevd debug info. I killed the
> original systemd-udevd then ran it again with --debug.
> You can find the problem occurred about in line 4859.
> Hope this helps.
> If you need more info, please let me know.
> Any comments would be highly appreciated.
> Thanks.

Well, you see that "bcache-probe" is invoked right before things go
bad? Whatever that is, maybe that's the problem? Have you tried
turning that rules file off?

Lennart

-- 
Lennart Poettering, Red Hat