From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 9 Oct 2002 08:29:54 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 9 Oct 2002 08:29:54 -0400 Received: from d12lmsgate-3.de.ibm.com ([194.196.100.236]:47306 "EHLO d12lmsgate-3.de.ibm.com") by vger.kernel.org with ESMTP id convert rfc822-to-8bit; Wed, 9 Oct 2002 08:29:53 -0400 Content-Type: text/plain; charset="us-ascii" From: Martin Schwidefsky Organization: IBM Deutschland GmbH To: linux-kernel@vger.kernel.org, torvalds@transmeta.com Subject: [PATCH] 2.5.41 s390. Date: Wed, 9 Oct 2002 14:33:30 +0200 X-Mailer: KMail [version 1.4] MIME-Version: 1.0 Content-Transfer-Encoding: 8BIT Message-Id: <200210091425.36762.schwidefsky@de.ibm.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hi Linus, the latest s/390 changes for 2.5.41. Most of it is adaptions to recent changes, only the last three contain "real" changes: 01: Use the generic method to generate asm-offsets.h, fix some inclues and fix signal dequeueing in the 31 bit emulation code. 02: Remove last traces of tq_struct from the s390 drivers. 03: Work queues can't be use for 3215 and 3270 at the moment because they get initialized to late for the console. Therefore I replaced them by tasklets. This works but if the tasklets are supposed to go away then I have a problem. 04: A typo in vmlinus.lds.S 05: A superfluous memset. 06: A fix for strace. We noticed that some svcs didn't show up correctly in the strace output. Heiko debugged this an found that executed svcs are not handled properly. The execute instruction is an odd one, it points to another instruction to be executed in place of the execute instruction. In this case strace does not see an svc instruction but an execute at the current ip. Unluckily the two instruction have a different length. strace can't reliably distinguish between these two cases because the execute instruction could end just like an svc instruction. So the kernel has to provide the svc number to strace. 07: 3270 update from Richard Hitt. 08: A patch to remove duplicated code in the 31 bit emulation and use the generic code in kernel/uid16.c instead. This patch might be a bit problematic because it adds the __UID16 define to common code files which is ugly. Just for info: hwc, lcs and tape are currently rewritten. So don't worry about uglyness in their current implementation. At the moment its enough that they compile. blue skies, Martin.