From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753147Ab0EMGEp (ORCPT ); Thu, 13 May 2010 02:04:45 -0400 Received: from e23smtp04.au.ibm.com ([202.81.31.146]:34299 "EHLO e23smtp04.au.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752939Ab0EMGEn (ORCPT ); Thu, 13 May 2010 02:04:43 -0400 From: "Ian Munsie" To: linux-kernel@vger.kernel.org Cc: Peter Zijlstra , Paul Mackerras , Ingo Molnar , Arnaldo Carvalho de Melo , Frederic Weisbecker , Steven Rostedt Subject: perf trace: Fix array handling & value truncation Date: Thu, 13 May 2010 16:03:45 +1000 Message-Id: <1273730632-21008-1-git-send-email-imunsie@au1.ibm.com> X-Mailer: git-send-email 1.7.1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org This patch series fixes several issues with perf trace relating to the size of a long. The first issue fixed by patch #2 is an assumption that tracepoints with arrays as part of their output (for instance, raw_syscalls:sys_enter) that the size of each element within the array is a long size. Due to further problems fixed by patches #3 through #5, the long_size was not even being set in the trace-event-parse.c file, and as a result the code was previously treating the array elements as 0 size. Patch #6 fixes issues where printing the values of pointers and longs from 64bit kernels were being truncated if perf was compiled as 32bit. This obviously effects tracepoints with a "%p" or "%ld" in their output format, but also affects the printout of the location of kprobes. Patches #1 and #7 apply good software engineering practice to help spot future regressions of these issues. Thanks, -Ian