xfs
[Top] [All Lists]

Re: [patch 01/11] Move compat ioctl structs & numbers into xfs_ioctl32.h

To: Eric Sandeen <sandeen@xxxxxxxxxxx>
Subject: Re: [patch 01/11] Move compat ioctl structs & numbers into xfs_ioctl32.h
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Thu, 20 Nov 2008 08:25:18 -0500
Cc: xfs@xxxxxxxxxxx
In-reply-to: <4924C36B.6090700@xxxxxxxxxxx>
References: <20081119044401.573365619@xxxxxxxxxxx> <20081119044907.776640320@xxxxxxxxxxx> <20081119145941.GA13050@xxxxxxxxxxxxx> <49242FAC.2060208@xxxxxxxxxxx> <4924C36B.6090700@xxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Wed, Nov 19, 2008 at 07:54:51PM -0600, Eric Sandeen wrote:
> So I went off to do it this way, and while I certainly can, (I'm almost
> done rewriting it this way) after a bit more thought I really don't see
> much wrong with just translating in the 32-bit args for any ioctl which
> is _IOW only... it doesn't seem that bad... any particular reason you're
> against it?

It's just generally more code and less readable, instead of directly
calling the underlying helper you return the mangled argument to the
main ioctl dispatcher, which then has to reset the 'cmd' value and then
you need to go through another round of dispathing in the native ioctl
handler.  It also means you have to do compat_alloc_user calls you
otherwise wouldn't have to do.

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