On Fri, Jun 24, 2016 at 09:55:37AM -0500, Eric Sandeen wrote:
> On 6/24/16 2:52 AM, Christoph Hellwig wrote:
> > Instead we always declare struct xfs_dir2_sf_hdr as packed. That's
> > the expected layout, and while most major architectures do the packing
> > by default the new structure size and offset checker showed that not
> > only the ARM old ABI got this wrong, but various minor embedded
> > architectures did as well.
> >
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> > ---
> > fs/xfs/libxfs/xfs_da_format.h | 2 +-
> > fs/xfs/xfs_linux.h | 7 -------
> > 2 files changed, 1 insertion(+), 8 deletions(-)
> >
> > diff --git a/fs/xfs/libxfs/xfs_da_format.h b/fs/xfs/libxfs/xfs_da_format.h
> > index f877bb1..685f23b 100644
> > --- a/fs/xfs/libxfs/xfs_da_format.h
> > +++ b/fs/xfs/libxfs/xfs_da_format.h
> > @@ -229,7 +229,7 @@ typedef struct xfs_dir2_sf_hdr {
> > __uint8_t count; /* count of entries */
> > __uint8_t i8count; /* count of 8-byte inode #s */
> > __uint8_t parent[8]; /* parent dir inode number */
> > -} __arch_pack xfs_dir2_sf_hdr_t;
> > +} __packed xfs_dir2_sf_hdr_t;
>
> The reason I did this in the first place was a vague notion that unconditional
> packing was harmful.
>
> http://digitalvampire.org/blog/index.php/2006/07/31/why-you-shouldnt-use-__attribute__packed/
>
> "However, it's actively harmful to add the attribute to a structure that's
> already going to be laid out with no padding."
> ...
> "gcc gets scared about unaligned accesses and generates six times as much code
> (96 bytes vs. 16 bytes)! sparc64 goes similarly crazy, bloating from 12 bytes
> to 52 bytes"
>
> I don't know if that's (still) correct or not, but that was the reason
> for the selective __pack application way back when. Might be worth
> investigating?
Christoph? The first two ptches are fine, but more info is needed
for this one...
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|