linux-kernel - Re: [PATCH v2 10/15] perf ftrace: add support for trace option sleep-time

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Wed, 8 Jul 2020 20:59:20 +0800
From:   Changbin Du <changbin.du@...il.com>
To:     Arnaldo Carvalho de Melo <arnaldo.melo@...il.com>
Cc:     Changbin Du <changbin.du@...il.com>,
        Namhyung Kim <namhyung@...nel.org>,
        Jiri Olsa <jolsa@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 10/15] perf ftrace: add support for trace option
 sleep-time

On Tue, Jul 07, 2020 at 01:09:43PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Tue, Jul 07, 2020 at 11:35:14PM +0800, Changbin Du escreveu:
> > On Fri, Jul 03, 2020 at 03:40:59PM +0900, Namhyung Kim wrote:
> > > On Sat, Jun 27, 2020 at 10:38 PM Changbin Du <changbin.du@...il.com> wrote:
> > > > This adds an option '--graph-nosleep-time' which allow us only to measure
> > > > on-CPU time. This option is function_graph tracer only.
> 
> > > I'd suggest --graph-sleep-time instead and set it by default.
> > > Then we can have --no-graph-sleep-time as well.
> 
> > If so can we make --graph-sleep-time as default. Seems there is no something
> > like OPT_BOOLEAN_DEFAULT().
> 
> > > By the way,  didn't we agree to have something like --graph-opts instead?
> 
> > As I mentioned in previous version, --graph-opts make the parsing a little
> > complex.
> > 	--graph-opts depth=<n>,nosleep_time,noirqs,no_tail,verbose
> 
> > While the exsiting style is '--graph-depth <n>'. The cons is we must type more
> > characters. The pros is don't need to reimplement parsing and align with old
> > behaviour.
> > 	--graph-depth <n>
> > 	--graph-nosleep-time
> > 	--graph-noirqs
> > 	--graph-notail
> > 	--graph-verbose
> 
> This achieves the same result, yes, but it is more cumbersome, to use
> all we would have to do:
> 
>     perf ftrace --graph-depth 10 --graph-nosleep-time --graph-noirqs --graph-notail --graph-verbose ...
> 
> When we could have a more compact:
> 
>     perf ftrace --graph-opts depth=10,nosleep-time,noirqs,notail,verbose ...
> 
> I.e., instead of a series of:
> 
>         OPT_INTEGER(0, "graph-depth", &ftrace.graph_depth, "Max depth for function graph tracer"),
> 	OPT_BOOLEAN(0, "sleep-time", &ftrace.sleep_time, "explanation"),
> 	OPT_BOOLEAN(0, "irqs" &ftrace.irqs, "explanation"),
> 	OPT_BOOLEAN(0, "tail" &ftrace.tail, "explanation"),
> 	OPT_BOOLEAN(0, "verbose" &ftrace.verbose, "explanation"),
> 
> We'd have a:
> 
> 	OPT_CALLBACK(0, "graph-opts", &ftrace, "options", "graph options", parse_graph_opts),
> 
> And parse_graph_opts() would just have a simple strchr(opt, ',') + if
> strcmp()/else loop, i.e., it would be easier for users, slightly more
> difficult for perf developers :-)
>
And we also have to parse key-value pairs for 'graph-depth' and move option
'--graph-depth' to 'graph-opts'.

There is an existing similr option '--debug'. It can be factored out as a
general interface.
	--debug --debug verbose=2,stderr

> - Arnaldo

-- 
Cheers,
Changbin Du