From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S272280AbTGYUCA (ORCPT ); Fri, 25 Jul 2003 16:02:00 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S272281AbTGYUCA (ORCPT ); Fri, 25 Jul 2003 16:02:00 -0400 Received: from newweb.speedscript.com ([66.139.78.154]:52382 "EHLO mail.speedscript.com") by vger.kernel.org with ESMTP id S272280AbTGYUB7 (ORCPT ); Fri, 25 Jul 2003 16:01:59 -0400 Subject: Possible sparc64 signal handling issue From: Hal Duston To: linux-kernel@vger.kernel.org Content-Type: text/plain Message-Id: <1059164229.3926.16.camel@ulysseus> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.3 Date: 25 Jul 2003 15:17:10 -0500 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org In arch/sparc64/kernel/traps.c, data_access_exception() is defined to receive 3 parameters, (and uses all three of them). In arch/sparc64/kernel/unaligned.c it is only passed a single parameter. This is causing me difficulty when handling segfault signals. On most Linux platforms I have access to the siginfo_t.si_addr isn't populated at all anyway, and I use the ucontext_t third parameter given to the handler. Under sparc64 though, the third parameter doesn't point to anything that I can discern, and due to the aforementioned three versus one parameter issue, the value of si_addr in siginfo_t is incorrect as well. This issue appears to be present in both 2.4, and 2.6. -- Hal Duston hduston@speedscript.com