From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeff Liu <jeff.liu@oracle.com>
Subject: Re: [PATCH] vfs: Return EINVAL for default SEEK_HOLE, SEEK_DATA implementation
Date: Mon, 06 Jan 2014 17:32:41 +0800
Message-ID: <52CA7839.7070200@oracle.com>
References: <1388745326-13226-1-git-send-email-jack@suse.cz> <19710447-69E6-47A2-A679-68552967A750@dilger.ca> <52C8A8FC.4080201@fb.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: Al Viro <viro@ZenIV.linux.org.uk>, "Theodore Ts'o" <tytso@mit.edu>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>
To: Josef Bacik <jbacik@fb.com>, Andreas Dilger <adilger@dilger.ca>,
	Jan Kara <jack@suse.cz>
Return-path: <linux-fsdevel-owner@vger.kernel.org>
Received: from userp1040.oracle.com ([156.151.31.81]:18136 "EHLO
	userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1754410AbaAFJdE (ORCPT
	<rfc822;linux-fsdevel@vger.kernel.org>);
	Mon, 6 Jan 2014 04:33:04 -0500
In-Reply-To: <52C8A8FC.4080201@fb.com>
Sender: linux-fsdevel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

On 01/05 2014 08:36 AM, Josef Bacik wrote:
> 
> On 01/03/2014 05:17 PM, Andreas Dilger wrote:
>> Shouldn't this be -EOPNOTSUPP, since -EINVAL is for incorrect
>> arguments, which isn't the case here.
>>
> I think we want to be consistent with previous kernels which would
> return -EINVAL if whence wasn't set properly, hopefully breaking as few
> userspace stuff as possible.  Thanks,

It seems like that this fix would have a bit influence to man page of lseek(2)
as it suppose to:

       EINVAL whence is not valid.  Or: the resulting  file  offset  would  be
              negative, or beyond the end of a seekable device.

Where probably need an update to reflect this change regardless of whether
return -EINVAL or -EOPNOTSUPP IMHO.

Moreover, Solaris/FreeBSD use {f}pathconf(2) syscalls to detect if a file
system is support SEEK_HOLE function or not, maybe Glibc/GNULib folks would
like to couple lseek(..., SEEK_HOLE) to {f}pathconf(3) in the future, which
could make the userspace stuff a bit more flexible across different OS.

So, I gone through their man pages and in this case, seems both of them will
return -EINVAL if call to {f}pathconf(2) with _PC_MIN_HOLE_SIZE is failed:

Solaris: http://docs.oracle.com/cd/E23824_01/html/821-1463/pathconf-2.html
FreeBSD: http://www.freebsd.org/cgi/man.cgi?query=pathconf&sektion=2


Thanks,
-Jeff