On Tue, 12.05.15 18:04, Karel Zak (k...@redhat.com) wrote:
> Currently we have no way how to specify dependencies between fstab
> entries (or another units) in the /etc/fstab. It means that users are
> forced to bypass fstab and write .mount units manually.
>
> Years ago Lennart suggested to add
>
> x-systemd.after=<PATH>
>
> - to specify dependence an another mount (PATH is translated to <PATH>.mount)
>
> x-systemd.after=<UNIT>
>
> - to specify dependence on arbitrary UNIT
>
> The x-systemd.after= is implemented by After= and Requires=.
Hmm, I am pretty sure x-systemd.after= should only create After=
dependencies, not Requires=. I'd be open though to add
x-systemd.requires= which adds both After= and Requires=.
(So far we never augmented ordering deps by requirement deps, but
sometimes we do augment requirement deps with ordering deps, for
example in many cases for DefaultDependencies=yes logic. Hence we
should follow this here, too.)
Hence, please rename this option to "x-systemd.requires=". Otherwise I
like the concept.
> x-systemd.requires-mounts-for=<PATH ...>
>
> - to specify dependence on another paths, implemented by
> RequiresMountsFor=.
Sounds good.
> <varlistentry>
> + <term><option>x-systemd.after=</option></term>
> +
> + <listitem><para>Configures <varname>After=</varname> and
> <varname>Requires=</varname> dependence
s/dependence/dependency
> +
> + <listitem><para>Configures <varname>RequiresMountsFor=</varname>
> dependence between the mount and
> + another mount. The argument is a space-separated list of absolute
> paths. Note
> + that <filename>/etc/fstab</filename> format requires to escape space
> as \040.
> + See <varname>RequiresMountsFor=</varname> in
> +
> <citerefentry><refentrytitle>systemd.unit</refentrytitle><manvolnum>5</manvolnum></citerefentry>
> + for details.</para></listitem>
Hmm, I think it would be better to allow specifiying the option
multiple times rather than using spaces to seperate the values...
> + r = fstab_filter_options(opts, "x-systemd.after\0", NULL, &arg,
> NULL);
> + if (r < 0)
> + return log_warning_errno(r, "Failed to parse options: %m");
> + if (r == 0)
> + return 0;
> +
> + if (*arg == '/') {
> + r = unit_name_from_path(arg, ".mount", &unit);
I figure this should simply use unit_name_mangle_with_suffix(), which
will do the right thing for device nodes, normal paths and unit
names. No need to manually mangle this...
> + if (!automount && opts) {
> + write_requires_after(f, opts);
> + write_requires_mounts_for(f, opts);
> + }
> +
This should probably propagate errors.
> if (passno != 0) {
> r = generator_write_fsck_deps(f, arg_dest, what, where,
> fstype);
> if (r < 0)
> @@ -315,6 +359,11 @@ static int add_mount(
> "Before=%s\n",
> post);
>
> + if (opts) {
> + write_requires_after(f, opts);
> + write_requires_mounts_for(f, opts);
> + }
Same here.
Lennart
--
Lennart Poettering, Red Hat
_______________________________________________
systemd-devel mailing list
systemd-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/systemd-devel
