linux-kernel - Re: [PATCH] dynamic debug: allow printing to trace event

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]

Date:   Thu, 23 Jul 2020 16:08:44 +0206
From:   John Ogness <john.ogness@...utronix.de>
To:     Steven Rostedt <rostedt@...dmis.org>
Cc:     Vincent Whitchurch <vincent.whitchurch@...s.com>,
        jbaron@...mai.com, mingo@...hat.com, kernel@...s.com,
        corbet@....net, linux-kernel@...r.kernel.org,
        Petr Mladek <pmladek@...e.com>,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>
Subject: Re: [PATCH] dynamic debug: allow printing to trace event

On 2020-07-22, Steven Rostedt <rostedt@...dmis.org> wrote:
>>> +static void dynamic_printk(unsigned int flags, const char *fmt, ...)
>>> +{
>>> +	if (flags & _DPRINTK_FLAGS_TRACE) {
>>> +		va_list args;
>>> +
>>> +		va_start(args, fmt);
>>> +		/*
>>> +		 * All callers include the KERN_DEBUG prefix to keep the
>>> +		 * vprintk case simple; strip it out for tracing.
>>> +		 */
>>> +		dynamic_trace(fmt + strlen(KERN_DEBUG), args);  
>> 
>> Do we really need a separate tracing event for this? Why not just:
>> 
>>                 ftrace_vprintk(fmt + strlen(KERN_DEBUG), args);
>
> It must be an event, one that can be enabled or disabled separately
> from trace_printk().
>
> If you are asking if this could be something like trace_printk(),
> which ftrace_vprintk() is. The reason for that nasty banner when
> people use trace_printk() is to keep developers from using it as their
> personal debugging tool in production.
>
> A trace_printk() can not be discretely disabled. It's either totally
> on, or totally off. And since it is used for debugging, if there's
> trace_printk()s all over the kernel, you will have to deal with the
> noise of everyone else's trace_printk(), making trace_printk()
> useless.

I understand and agree with your concern about trace_printk(). But it
seems to me that trace_printk() via pr_debug() should be OK because
there is discrete control per message implemented. Yes, more code is
necessary to distinguish between the two, such as letting dynamic_printk
use an internal function that does not trigger a splat. But I think that
is reasonable.

For me a trace event represents a specific point in the kernel code. But
this new printk trace event, instead, represents general log
redirection. I do not oppose it, but it feels like a hack to me. In
contrast, simply setting a dynamic printk flag to write the message
string to the trace buffer (without also activating some pseudo trace
event) feels more natural.

Just sharing my thoughts, as requested.

John Ogness