linux-kernel - Re: Why have another variable deciding a tracepoint?

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: <5285D8A5.6030903@nvidia.com>
Date:	Fri, 15 Nov 2013 10:17:41 +0200
From:	Terje Bergström <tbergstrom@...dia.com>
To:	Steven Rostedt <rostedt@...dmis.org>,
	LKML <linux-kernel@...r.kernel.org>
CC:	Arto Merilainen <amerilainen@...dia.com>,
	Thierry Reding <thierry.reding@...onic-design.de>,
	Erik Faye-Lund <kusmabite@...il.com>
Subject: Re: Why have another variable deciding a tracepoint?

On 15.11.2013 06:48, Steven Rostedt wrote:
> I've been reviewing different users of tracepoints and I stumbled
> across this:
> 
> drivers/gpu/host1x/cdma.c: host1x_cdma_push()
> 
> 	if (host1x_debug_trace_cmdbuf)
> 		trace_host1x_cdma_push(dev_name(cdma_to_channel(cdma)->dev),
> 				       op1, op2);
> 
> That host1x_debug_trace_cmdbuf is a variable that gets set by another
> debugfs file "trace_cmdbuf" that is custom to this driver.
> 
> Why?

This is because it takes a lot of time to prepare for dumping the cmdbuf
data, like mapping buffer and unmapping after tracing. We want to avoid
all that preparation time.

> The tracepoint host1x_cdma_push is already controlled by either ftrace
> or perf. If it gets enabled by perf or ftrace, it still wont be traced
> unless we also enable this trace_cmdbuf. Is there some reason for this?
> I can't figure it out from the change log: 6236451d83a720 ("gpu:
> host1x: Add debug support").
> 
> As tracepoints uses jump labels, there is no branch cost associated
> with them. That is, they are either a direct jump, or a nop (in most
> cases a nop). But here you added the overhead of a conditional branch
> depending on this variable.
> 
> If this is truly needed, then use TRACE_EVENT_CONDITION() for that
> tracepoint.

TRACE_EVENT_CONDITION() isn't documented, so I don't know how that would
help.

Terje
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/