[systemd-devel] [PATCH] readme: CONFIG_FHANDLE is a requirement

[Lennart Poettering]

On Tue, 14.01.14 15:26, Umut Tezduyar Lindskog (umut.tezduyar at axis.com) wrote:

> From: Umut Tezduyar Lindskog <umuttl at axis.com>
> 
> ---
>  README |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/README b/README
> index a1058c5..6fcab4f 100644
> --- a/README
> +++ b/README
> @@ -63,7 +63,7 @@ REQUIREMENTS:
>          Some udev rules and virtualization detection relies on it:
>            CONFIG_DMIID
>  
> -        Mount and bind mount handling might require it:
> +        Mount and bind mount handling requires it:
>            CONFIG_FHANDLE
>  
>          Support for some SCSI devices serial number retrieval, to

Soooo. I don't really care too much whether we declare CONFIG_FHANLDE as
mandatory or not, but the original intention was to make this an
optional requirement, not a mandatory one.

We need the CONFIG_FHANLDE stuff only to detect within tmpfiles (for its
"aging" stuff) and suchlike whether we are descending into a mount point
that is of the same file system as the file system it is mounted
into. The usual checks with stat() on both fs and comparing st_dev won't
be reliable in this case. This is only a safety check, and in many
(especially embedded cases with a fixed set of software) it shouldn't
matter.

Now, I never ran the whole thing on a kernel lacking CONFIG_FHANDLE,
which is probably why the intended optional dep thing didn't work. It
should be fairly easy to fix that (probably just some checking for errno
== ENOTSUP or so somewhere). If there's interest to support kernels
without CONFIG_FHANDLE then I'd be happy to merge a patch that makes
this work and again declares CONFIG_FHANDLE optional. I'd suspect this
to be a three-line change or so, for somebody who's interested...

Lennart

-- 
Lennart Poettering, Red Hat