mirror of
git://git.gnupg.org/gnupg.git
synced 2024-12-22 10:19:57 +01:00
doc: Explain why socket activation is a problem
--
This commit is contained in:
parent
239c1fdc28
commit
431239b83d
30
README
30
README
@ -144,6 +144,13 @@
|
||||
gpg --import --import-options restore < allkeys.gpg
|
||||
gpgsm --import < allcerts.crt
|
||||
|
||||
In case the keyboxd is not able to startup due to a stale lockfile
|
||||
created by another host, the command
|
||||
|
||||
gpgconf --unlock pubring.db
|
||||
|
||||
can be used to remove the lock file.
|
||||
|
||||
** Socket directory
|
||||
|
||||
GnuPG uses Unix domain sockets to connect its components (on Windows
|
||||
@ -166,6 +173,29 @@
|
||||
fi
|
||||
done )
|
||||
|
||||
** Conflicts with systemd socket activation
|
||||
|
||||
Some Linux distribution use the meanwhile deprecated --supervised
|
||||
option with gpg-agent, dirmngr, and keyboxd. The idea is that the
|
||||
systemd process launches the daemons as soon as gpg or gpgsm try to
|
||||
access them. However, this creates a race condition with GnuPG's
|
||||
own on-demand launching of these daemon. It also conflicts with the
|
||||
remote use gpg-agent because the no-autostart feature on the remote
|
||||
site will not work as expected.
|
||||
|
||||
Thus the recommendation is not to use the --supervised option. All
|
||||
GnuPG components handle the startup of their daemons on their own.
|
||||
|
||||
The only problem is that for using GnuPG's ssh-agent protocol
|
||||
support, the gpg-agent must have been started before ssh. This can
|
||||
either be done with an ssh wrapper running
|
||||
|
||||
gpg-connect-agent updatestartuptty /bye
|
||||
|
||||
for each new tty or by using that command directly after login when
|
||||
the anyway required SSH_AUTH_SOCK envvar is set (see the example in
|
||||
the gpg-agent man page).
|
||||
|
||||
|
||||
* DOCUMENTATION
|
||||
|
||||
|
Loading…
x
Reference in New Issue
Block a user