linux-kernel - Re: [RFC PATCH 1/2] Use pause-on-trace with the latency tracers

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

Message-ID: <20210126094539.4969c357@gandalf.local.home>
Date:   Tue, 26 Jan 2021 09:45:39 -0500
From:   Steven Rostedt <rostedt@...dmis.org>
To:     Viktor Rosendahl <Viktor.Rosendahl@....de>
Cc:     Joel Fernandes <joel@...lfernandes.org>,
        <linux-kernel@...r.kernel.org>, Ingo Molnar <mingo@...hat.com>
Subject: Re: [RFC PATCH 1/2] Use pause-on-trace with the latency tracers


Sorry for the late reply.

On Tue, 19 Jan 2021 17:43:43 +0100
Viktor Rosendahl <Viktor.Rosendahl@....de> wrote:

> Eaerlier, tracing was disabled when reading the trace file. This behavior
> was changed with:
> 
> commit 06e0a548bad0 ("tracing: Do not disable tracing when reading the
> trace file").
> 
> This doesn't seem to work with the latency tracers.
> 
> The above mentioned commit dit not only change the behavior but also added
> an option to emulate the old behavior. The idea with this patch is to
> enable this pause-on-trace option when the latency tracers are used.
> 
> This is a workaround, perhaps it would be better to make the latency
> tracers work without pausing but I am not sure how to do that, or even
> how feasible it is without significant rework.

I think this is a fine workaround. The latency tracers are rather "special"
and some options just don't make sense for them. There's some things I wish
I had done different with them, but we can't fix because of backward
compatibility reasons.

I'll take a look at these today. Thanks.

-- Steve

> 
> Signed-off-by: Viktor Rosendahl <Viktor.Rosendahl@....de>
> ---
>  kernel/trace/trace_irqsoff.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
> index d06aab4dcbb8..6756379b661f 100644
> --- a/kernel/trace/trace_irqsoff.c
> +++ b/kernel/trace/trace_irqsoff.c
> @@ -562,6 +562,8 @@ static int __irqsoff_tracer_init(struct trace_array *tr)
>  	/* non overwrite screws up the latency tracers */
>  	set_tracer_flag(tr, TRACE_ITER_OVERWRITE, 1);
>  	set_tracer_flag(tr, TRACE_ITER_LATENCY_FMT, 1);
> +	/* without pause, we will produce garbage if another latency occurs */
> +	set_tracer_flag(tr, TRACE_ITER_PAUSE_ON_TRACE, 1);
>  
>  	tr->max_latency = 0;
>  	irqsoff_trace = tr;
> @@ -583,11 +585,13 @@ static void __irqsoff_tracer_reset(struct trace_array *tr)
>  {
>  	int lat_flag = save_flags & TRACE_ITER_LATENCY_FMT;
>  	int overwrite_flag = save_flags & TRACE_ITER_OVERWRITE;
> +	int pause_flag = save_flags & TRACE_ITER_PAUSE_ON_TRACE;
>  
>  	stop_irqsoff_tracer(tr, is_graph(tr));
>  
>  	set_tracer_flag(tr, TRACE_ITER_LATENCY_FMT, lat_flag);
>  	set_tracer_flag(tr, TRACE_ITER_OVERWRITE, overwrite_flag);
> +	set_tracer_flag(tr, TRACE_ITER_PAUSE_ON_TRACE, pause_flag);
>  	ftrace_reset_array_ops(tr);
>  
>  	irqsoff_busy = false;