From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail.pbcl.net ([88.198.119.4] helo=hetzner.pbcl.net) by linuxtogo.org with esmtp (Exim 4.72) (envelope-from ) id 1TJMeA-00058N-HD for openembedded-core@lists.openembedded.org; Wed, 03 Oct 2012 12:57:46 +0200 Received: from elite.brightsigndigital.co.uk ([81.142.160.137] helo=[172.30.1.145]) by hetzner.pbcl.net with esmtpsa (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.72) (envelope-from ) id 1TJMRd-0004rM-73; Wed, 03 Oct 2012 12:44:49 +0200 From: Phil Blundell To: Martin Jansa Date: Wed, 03 Oct 2012 11:44:48 +0100 In-Reply-To: <20121003103141.GA21043@jama.jama.net> References: <1349259854.32611.106.camel@phil-desktop> <20121003103141.GA21043@jama.jama.net> X-Mailer: Evolution 3.0.2- Message-ID: <1349261089.32611.115.camel@phil-desktop> Mime-Version: 1.0 Cc: oe-core Subject: Re: [PATCH] insane: detect and warn about relocations in .text X-BeenThere: openembedded-core@lists.openembedded.org X-Mailman-Version: 2.1.11 Precedence: list List-Id: Patches and discussions about the oe-core layer List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 03 Oct 2012 10:57:46 -0000 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit On Wed, 2012-10-03 at 12:31 +0200, Martin Jansa wrote: > On Wed, Oct 03, 2012 at 11:24:12AM +0100, Phil Blundell wrote: > > Signed-off-by: Phil Blundell > > Can you add a bit longer description of possible issues with relocations > in .text? So that people seeing this issue will know how dangerous it is > for them? > > From my understanding (after reading > http://www.gentoo.org/proj/en/hardened/pic-fix-guide.xml) it's mostly > performance issue? Yes, that's correct. It basically falls into the same sort of category as useless-rpaths; the binary will still work, but there will be some adverse impact on performance and memory usage. Historically, the most common cause of DT_TEXTREL was accidentally linking non-PIC code into a DSO. Recent versions of the linker will flatly refuse to do this on at least some architectures, though, so hopefully this problem will just go away over time. p.