From: Martin Str|mberg <ams AT ludd DOT luth DOT se>
Message-Id: <200101132117.WAA26339@father.ludd.luth.se>
Subject: Re: FSEXTs and llseek()
In-Reply-To: <3A60A7A3.88C0F733@phekda.freeserve.co.uk> from Richard Dawe at "Jan 13, 2001 07:08:19 pm"
To: djgpp-workers AT delorie DOT com
Date: Sat, 13 Jan 2001 22:17:30 +0100 (MET)
X-Mailer: ELM [version 2.4ME+ PL54 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

According to Richard Dawe:
> Martin Str|mberg wrote:
> [snip]
> > > On Sat, 6 Jan 2001, Richard Dawe wrote:
> [snip]
> > > > I've just been reading about __FSEXT_llseek. 'offset_t' is
> > > > described as being a 'long long'. This is OK for the input
> > > > arguments to __FSEXT_llseek, but how do you return the offset
> > > > to the caller? The return value variable for the FSEXT handler
> > > > function is only an 'int'.
> [snip]
> > I'm not very good at the arithmetic in casting but wouldn't keeping
> > the return value as an int and casting it to long long in the llseek
> > hooker work and then convert anything < -1 to positive work?
> > 
> > If this would work we would only punish the llseek() hooker.
> 
> You can't really typecast the return value: 'int *' cast to 'long long *'
> will lead to memory corruption, when assigning a return value.

Of course you can't cast "int *" to "long long *" and expect anything
to work. But you can cast "int" to "long long" (not necessarily the
return value).

> The problem is returning a good (int-sized) offset to the FSEXT caller for
> __FSEXT_llseek. If you can't return a good offset, then it's broken.

There is a good (well, sufficiently non-broken at least) int-sized
return value: the range [-1, 2^32-2] which will be represented as the
range [-2^31, 2^31-1].


Right,

						MartinS