Date: Thu, 9 Oct 2008 12:22:23 -0700 (PDT)
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Miklos Szeredi <miklos@szeredi.hu>
cc: jens.axboe@oracle.com, linux-kernel@vger.kernel.org,
       linux-fsdevel@vger.kernel.org
Subject: Re: splice vs O_APPEND
In-Reply-To: <E1KnyOw-00044y-K4@pomaz-ex.szeredi.hu>
Message-ID: <alpine.LFD.2.00.0810091218300.3210@nehalem.linux-foundation.org>
References: <E1Knx26-0003ds-Ch@pomaz-ex.szeredi.hu> <alpine.LFD.2.00.0810090830170.3210@nehalem.linux-foundation.org> <E1Knxzm-0003y2-Jy@pomaz-ex.szeredi.hu> <alpine.LFD.2.00.0810090908200.3210@nehalem.linux-foundation.org>
 <E1KnyOw-00044y-K4@pomaz-ex.szeredi.hu>
User-Agent: Alpine 2.00 (LFD 1167 2008-08-23)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1140
Lines: 27


On Thu, 9 Oct 2008, Miklos Szeredi wrote:
> 
> We know that nobody is currently relying on O_APPEND semantics with
> splice, so this should be OK.

Having now realized that apparently you can't rely on O_APPEND for 
anything but plain write _anyway_ (ie pwrite shouldn't honor it), I'm 
going to drop this thing as "let's think about it more".

Maybe the right thing to do is to just say that O_APPEND (and IS_APPEND) 
is really not as reliable as people might expect, and just say that the 
only thing it affects is a plain write() system call. 

Of course, I think POSIX is crazy, and we probably _should_ always honor 
O_APPEND, and returning -EINVAL is the right thing for both pwrite and 
splice, but this is all a murkier issue than it looked like originally, 
and any possible "security" implications are dubious in that you cannot 
really depend on O_APPEND/IS_APPEND anyway.

		Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/