Bug#554120: aufs2 in Debian
 

Okajima-san,

We are including aufs2 in the Debian kernel for use on live CDs. We are
trying to build it for all Debian architectures, but found the following
error on alpha:

CC [M] fs/aufs/vdir.o
/build/buildd-linux-2.6_2.6.31-1-alpha-87a0KU/linux-2.6-2.6.31/debian/build/source_alpha_none/fs/aufs/vdir.c: In function 'calc_size':
/build/buildd-linux-2.6_2.6.31-1-alpha-87a0KU/linux-2.6-2.6.31/debian/build/source_alpha_none/fs/aufs/vdir.c:28: error: size of array 'type name' is negative

The code says:

BUILD_BUG_ON(sizeof(ino_t) != sizeof(long));
return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));

but on alpha, sizeof(ino_t) == 4 but sizeof(long) == 8.

Is it really necessary that these types have the same size? I can't see
why it would be necessary for the directory entries to be long-aligned.
If it is, perhaps we could use:

BUILD_BUG_ON(sizeof(ino_t) > sizeof(long));
return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(long));

Ben.

--
Ben Hutchings
The generation of random numbers is too important to be left to chance.
- Robert Coveyou

Bug#554120: aufs2 in Debian
 

Hello Ben,

Ben Hutchings:
> The code says:
>
> BUILD_BUG_ON(sizeof(ino_t) !=3D sizeof(long));
> return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
>
> but on alpha, sizeof(ino_t) =3D=3D 4 but sizeof(long) =3D=3D 8.
>
> Is it really necessary that these types have the same size? I can't see
> why it would be necessary for the directory entries to be long-aligned.
:::

It supposed to align the structure on memory, 8-byte alignment.
On some systems, accesses to incorrectly aligned member has to pay the
penalty which is a short-time delay.

As you wrote, the code is wrong.
I had made CONFIG_AUFS_INO_T_64 to support ino_t on alpha
system. Currently it is used by export.c only.
This problem was born when I switched aufs2 from aufs1.

I will fix it and release on next Monday.
Thank you for reporting.


J. R. Okajima



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Bug#554120: aufs2 in Debian
 

> Ben Hutchings:
> > The code says:
> >
> > BUILD_BUG_ON(sizeof(ino_t) !=3D sizeof(long));
> > return ALIGN(sizeof(struct au_vdir_de) + nlen, sizeof(ino_t));
> >
> > but on alpha, sizeof(ino_t) =3D=3D 4 but sizeof(long) =3D=3D 8.

I decided to remove the BUILD_BUG_ON() line simply. If you want to test
early, then try removing.


J. R. Okajima



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org