xfs
[Top] [All Lists]

Re: [PATCH] fix dir2 shortform structures on ARM old ABI

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: [PATCH] fix dir2 shortform structures on ARM old ABI
From: Timothy Shimmin <tes@xxxxxxx>
Date: Thu, 20 Mar 2008 11:35:42 +1100
Cc: Andre Draszik <xfs@xxxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <47E0A07D.5090803@xxxxxxxxxxx>
References: <b1b17e290803181631x60a3e921l836b115c3b026773@xxxxxxxxxxxxxx> <47E085F3.8030908@xxxxxxxxxxx> <47E0A07D.5090803@xxxxxxxxxxx>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: Thunderbird 2.0.0.9 (Macintosh/20071031)
Eric Sandeen wrote:
Eric Sandeen wrote:

I've helpfully provided structure layouts for the structures you mention
in the attached files, for your diffing pleasure.  I think you'll find
that it's not exactly as you described.

Ah hell the arm structs I attached were for oldabi.  It's what I get for
saving this fun work for late at night ;)

Attached are eabi structs; still only xfs_dir2_data_entry, xfs_dinode
and xfs_log_item seem to be affected by end-of-struct padding, of the
structures you mention.  And xfs_log_item isn't a disk structure...

which brings me back to, what specific failures do you see as a result
of end-of-struct padding on these structs?

Which reminds me when writing 122 that I noticed with xfs_dinode
but didn't think the end of struct padding would affect things -
remember chatting to Nathan at the time IIRC.
Actually, now that I think about it (Tim waking up a bit :-),
the xfs_dinode_t is of limited value, because a lot of those
union fields at the end aren't actually used directly and just
give an indication of what the layout is like.
We are in the variable literal area there.


--Tim


<Prev in Thread] Current Thread [Next in Thread>