linux-kernel - Re: Unused trace event in erofs

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: <20250613081728.6212a554@gandalf.local.home>
Date: Fri, 13 Jun 2025 08:17:28 -0400
From: Steven Rostedt <rostedt@...dmis.org>
To: Gao Xiang <hsiangkao@...ux.alibaba.com>
Cc: LKML <linux-kernel@...r.kernel.org>, linux-erofs@...ts.ozlabs.org, Gao
 Xiang <xiang@...nel.org>, Chao Yu <chao@...nel.org>
Subject: Re: Unused trace event in erofs

On Fri, 13 Jun 2025 14:08:32 +0800
Gao Xiang <hsiangkao@...ux.alibaba.com> wrote:

> Hi Steven,
> 
> On 2025/6/13 10:49, Steven Rostedt wrote:
> > I have code that will trigger a warning if a trace event is defined but
> > not used[1]. It gives a list of unused events. Here's what I have for
> > erofs:
> > 
> > warning: tracepoint 'erofs_destroy_inode' is unused.  
> 
> I'm fine to remove it, also I wonder if it's possible to disable
> erofs tracepoints (rather than disable all tracepoints) in some
> embedded use cases because erofs tracepoints might not be useful for
> them and it can save memory (and .ko size) as you said below.

You can add #ifdef around them.

Note, the "up to around 5K" means it can add up to that much depending on
what you have configured. The TRACE_EVENT() macro (and more specifically
the DECLARE_EVENT_CLASS() which TRACE_EVENT() has), is where all the bloat
is. I generates unique code for each trace event that prints it, parses it,
records it, the event fields, and has code specific for perf, ftrace and BPF.

The DEFINE_EVENT() which can be used to make several events that are
similar use the same DECLARE_EVENT_CLASS() only takes up around 250 bytes.
One reason I tell people to use DECLARE_EVENT_CLASS() when you have similar events.

There's also a DEFINE_EVENT_PRINT() that can use an existing
DECLARE_EVENT_CLASS() but update the "printk" section. That adds some more
code (the creation of the print function) but still much smaller than the
DECLARE_EVENT_CLASS(). But this requires the tracepoint function (what the
code calls) must have the same prototype.

> 
> > 
> > Each trace event can take up to around 5K in memory regardless if they
> > are used or not. Soon there will be warnings when they are defined but
> > not used. Please remove any unused trace event or at least hide it
> > under an #ifdef if they are used within configs. I'm planning on adding
> > these warning in the next merge window.  
> 
> If you don't have some interest to submit a removal patch, I will post
> a patch later.

Please make the patch. There's too many for me to do them all.

Thanks!

-- Steve