[blfs-dev] ?==?utf-8?q? systemd+initrd=race condition
pierre.labastie at neuf.fr
Tue Apr 4 12:25:47 PDT 2017
On 04/04/2017 20:36, DJ Lucas - LFS wrote:
> On Tuesday, April 04, 2017 09:00 CDT, Pierre Labastie
> <pierre.labastie at neuf.fr> wrote:
> To break uniformity in my jhalfs tests, I decided I
> would build LFS-systemd. It happened it was on an LVM
> partition, so I needed an initrd, which I took from
> the book. After completing jhalfs, it booted OK, but
> then I installed Linux-PAM and could not login again.
> Each time I entered my username and hit <return>,
> there was a message telling:
> "System is booting up. |See pam_nologin(8)"
> Googling took me to , , and  took me to .
> Note that there is a thread leading to : , with a
> lot of details.
> Basically, systemd creates a /run/nologin file containing||:
> "System is booting up. See pam_nologin(8)"|
> |while it is booting, but if this occurs in the initrd, it
> may be never cleared because of a race condition (I do not
> understand the details, I am too new to systemd, if I am
> anything to systemd ;-).|
> This is after reinstalling systemd? The Linux-PAM page has a note to do so,
> and an instruction to temporarily remove /run/nologin.
I have read that page so many times, that I haven't seen that the note about
shadow now reads:" you should now reinstall the Shadow-4.4 *and Systemd-232*
packages". But I do not see anything about /run/nologin (and I think since
/run is a tmpfs, it would not be useful).
I'll try tomorrow and let you know.
More information about the blfs-dev