Re: p = kmalloc(sizeof(*p), )

From: Alan Cox <alan@lxorguk.ukuu.org.uk>
To: Andrew Morton <akpm@osdl.org>
Cc: rmk+lkml@arm.linux.org.uk, penberg@cs.Helsinki.FI,
	viro@ftp.linux.org.uk, linux-kernel@vger.kernel.org,
	torvalds@osdl.org
Subject: Re: p = kmalloc(sizeof(*p), )
Date: Tue, 20 Sep 2005 21:41:36 +0100	[thread overview]
Message-ID: <1127248896.7763.9.camel@localhost.localdomain> (raw)
In-Reply-To: <20050920105939.3c9c5e39.akpm@osdl.org>

On Maw, 2005-09-20 at 10:59 -0700, Andrew Morton wrote:
> umm, the three reasons which you deleted from the mail to which you're
> replying?

There were no reasons given in the mail I replied to. Perhaps I missed
another mail from you earlier.

I'm also puzzled by the one comemnt you made. You seem to imply that
seeing

foo = malloc(sizeof(*foo))

means it doesn't need checking. That is false on various grounds

1.	A lot of stuff is using void *, char * etc
2.	You've no idea that foo is the full object not a generic object with
stuff tacked on.

So thats very much false. You have to know what is really being
allocated in both cases. In the sizeof(*foo) case you also have to go
back, work out wtf foo really is and check that its not an array pointer
because sizeof char[40] is not the same as sizeof(* char *).

Thankfully Linus hates typedefs so that removes many of those