From mboxrd@z Thu Jan 1 00:00:00 1970 From: "J. Bruce Fields" Subject: Re: v9fs (9p): syscall setxattr inside kernel 3.14-rc1 returns size of set xattr Date: Tue, 18 Feb 2014 13:04:50 -0500 Message-ID: <20140218180450.GG4319@fieldses.org> References: <5303666F.7000706@cea.fr> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-fsdevel To: DENIEL Philippe Return-path: Received: from fieldses.org ([174.143.236.118]:50331 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751178AbaBRSEv (ORCPT ); Tue, 18 Feb 2014 13:04:51 -0500 Content-Disposition: inline In-Reply-To: <5303666F.7000706@cea.fr> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Tue, Feb 18, 2014 at 02:55:59PM +0100, DENIEL Philippe wrote: > I run v9fs as a client on a F20, in front of my Ganesha server (see > http://github.com/nfs-ganesha for details), using 9p.2000L > My acl non-regression test showed errors when I installed a recent > 3.14-rc1 kernel (I got it from kernel.org) on my F20 box. > Investigation showed that the setfacl command line got messy because > setxattr() (called from acl_set_modify() in libattr.so) return a > non-zero value when successful. Further investigation showed that > this behavior seems to come from v9fs_fid_xattr_set() inside > fs/9p/xattr.c in the kernel's source. > > It seems like setxattr syscall does now return the size of the set > xattr, and that seems to be the root cause of my problem. I do not > believe that this change in setxattr is no bug, but a new feature. > So I guess I should patch my libattr and/or glibc to use xattr/acl > with kernel 3.14-rc1. > Question is : where could I get the right version of libattr source > treee (eventually with libacl if needed). New kernel features shouldn't break old libraries--sounds like a bug. --b.