From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751399Ab1IZNkq (ORCPT ); Mon, 26 Sep 2011 09:40:46 -0400 Received: from hrndva-omtalb.mail.rr.com ([71.74.56.124]:48999 "EHLO hrndva-omtalb.mail.rr.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751145Ab1IZNkp (ORCPT ); Mon, 26 Sep 2011 09:40:45 -0400 X-Authority-Analysis: v=1.1 cv=XWD5/VRj2HUJOhsR8cgmvPBlhMACpZXxseY1Kn/ehQI= c=1 sm=0 a=gtOWHpAUYCwA:10 a=5SG0PmZfjMsA:10 a=Q9fys5e9bTEA:10 a=17wjrS5wAhQaEczCPkpxpQ==:17 a=pGLkceISAAAA:8 a=JUDzVgYUmskaudF2YpsA:9 a=PUjeQqilurYA:10 a=MSl-tDqOz04A:10 a=17wjrS5wAhQaEczCPkpxpQ==:117 X-Cloudmark-Score: 0 X-Originating-IP: 74.67.83.30 Subject: Re: [RFC PATCH 1/2] PM/runtime: introduce trace points for tracing rpm_* functions From: Steven Rostedt To: tom.leiming@gmail.com Cc: rjw@sisk.pl, stern@rowland.harvard.edu, linux-pm@lists.linux-foundation.org, linux-kernel@vger.kernel.org, mingo@redhat.com, fweisbec@gmail.com Date: Mon, 26 Sep 2011 09:40:43 -0400 In-Reply-To: <1316664701-15273-1-git-send-email-tom.leiming@gmail.com> References: <1316664701-15273-1-git-send-email-tom.leiming@gmail.com> Content-Type: text/plain; charset="ISO-8859-15" X-Mailer: Evolution 3.0.3- Content-Transfer-Encoding: 7bit Message-ID: <1317044445.26514.5.camel@gandalf.stny.rr.com> Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 2011-09-22 at 12:11 +0800, tom.leiming@gmail.com wrote: > From: Ming Lei > > This patch introduces 3 trace points to prepare for tracing > rpm_idle/rpm_suspend/rpm_resume functions, so we can use these > trace points to replace the current dev_dbg(). > > Signed-off-by: Ming Lei > --- > include/trace/events/rpm.h | 98 ++++++++++++++++++++++++++++++++++++++++++++ > kernel/trace/Makefile | 1 + > kernel/trace/rpm-traces.c | 21 +++++++++ > 3 files changed, 120 insertions(+), 0 deletions(-) > create mode 100644 include/trace/events/rpm.h > create mode 100644 kernel/trace/rpm-traces.c > > diff --git a/include/trace/events/rpm.h b/include/trace/events/rpm.h > new file mode 100644 > index 0000000..ecd40bb > --- /dev/null > +++ b/include/trace/events/rpm.h > @@ -0,0 +1,98 @@ > + > +#undef TRACE_SYSTEM > +#define TRACE_SYSTEM rpm > + > +#if !defined(_TRACE_RUNTIME_POWER_H) || defined(TRACE_HEADER_MULTI_READ) > +#define _TRACE_RUNTIME_POWER_H > + > +#include > +#include > +#include > + > +/* > + * The rpm_internal events are used for tracing some important > + * runtime pm internal functions. > + */ > +DECLARE_EVENT_CLASS(rpm_internal, > + > + TP_PROTO(struct device *dev, int flags), > + > + TP_ARGS(dev, flags), > + > + TP_STRUCT__entry( > + __field( const char *, name ) > + __field( int, flags ) > + __field( int , usage_count ) > + __field( int , disable_depth ) > + __field( int , runtime_auto ) > + __field( int , request_pending ) > + __field( int , irq_safe ) > + __field( int , child_count ) > + ), > + > + TP_fast_assign( > + __entry->name = dev_name(dev); Don't use pointers, you must copy the name here. Otherwise, if the trace is read after the device has left, you will be referencing a invalid pointer and possible crash the kernel. We have ways to express dynamic string fields. Use them. > + __entry->flags = flags; > + __entry->usage_count = atomic_read( > + &dev->power.usage_count); > + __entry->disable_depth = dev->power.disable_depth; > + __entry->runtime_auto = dev->power.runtime_auto; > + __entry->request_pending = dev->power.request_pending; > + __entry->irq_safe = dev->power.irq_safe; > + __entry->child_count = atomic_read( > + &dev->power.child_count); > + ), > + > + TP_printk("%s flags-%x cnt-%-2d dep-%-2d auto-%-1d p-%-1d" > + " irq-%-1d child-%d", > + __entry->name, __entry->flags, > + __entry->usage_count, > + __entry->disable_depth, > + __entry->runtime_auto, > + __entry->request_pending, > + __entry->irq_safe, > + __entry->child_count > + ) > +); > +DEFINE_EVENT(rpm_internal, rpm_suspend, > + > + TP_PROTO(struct device *dev, int flags), > + > + TP_ARGS(dev, flags) > +); > +DEFINE_EVENT(rpm_internal, rpm_resume, > + > + TP_PROTO(struct device *dev, int flags), > + > + TP_ARGS(dev, flags) > +); > +DEFINE_EVENT(rpm_internal, rpm_idle, > + > + TP_PROTO(struct device *dev, int flags), > + > + TP_ARGS(dev, flags) > +); > + > +TRACE_EVENT(rpm_return_int, > + TP_PROTO(struct device *dev, const char *func, int ret), > + TP_ARGS(dev, func, ret), > + > + TP_STRUCT__entry( > + __field( const char *, name ) > + __field( const char *, func ) > + __field( int, ret ) > + ), > + > + TP_fast_assign( > + __entry->name = dev_name(dev); Same here. > + __entry->ret = ret; > + __entry->func = func; And here. But as this is just __func__, you could use %pS and save the pointer to the function. This is acceptable as kallsyms will will used to find the function, and instead pass in __THIS_IP__ to the tracepoint. -- Steve > + ), > + > + TP_printk("%s:%s ret=%d", __entry->func, __entry->name, __entry->ret) > +); > + > +#endif /* _TRACE_RUNTIME_POWER_H */ > + > +/* This part must be outside protection */ > +#include > diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile > index 761c510..56bdab5 100644 > --- a/kernel/trace/Makefile > +++ b/kernel/trace/Makefile > @@ -53,6 +53,7 @@ endif > obj-$(CONFIG_EVENT_TRACING) += trace_events_filter.o > obj-$(CONFIG_KPROBE_EVENT) += trace_kprobe.o > obj-$(CONFIG_TRACEPOINTS) += power-traces.o > +obj-$(CONFIG_TRACEPOINTS) += rpm-traces.o > ifeq ($(CONFIG_TRACING),y) > obj-$(CONFIG_KGDB_KDB) += trace_kdb.o > endif > diff --git a/kernel/trace/rpm-traces.c b/kernel/trace/rpm-traces.c > new file mode 100644 > index 0000000..998949e > --- /dev/null > +++ b/kernel/trace/rpm-traces.c > @@ -0,0 +1,21 @@ > +/* > + * Power trace points > + * > + * Copyright (C) 2009 Ming Lei > + */ > + > +#include > +#include > +#include > +#include > +#include > +#include > + > +#define CREATE_TRACE_POINTS > +#include > + > +EXPORT_TRACEPOINT_SYMBOL_GPL(rpm_return_int); > +EXPORT_TRACEPOINT_SYMBOL_GPL(rpm_idle); > +EXPORT_TRACEPOINT_SYMBOL_GPL(rpm_suspend); > +EXPORT_TRACEPOINT_SYMBOL_GPL(rpm_resume); > +