linux-kernel - Re: [PATCH 00/14] xfs: Remove unused trace events

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: <20250616105819.4d37b83a@batman.local.home>
Date: Mon, 16 Jun 2025 10:58:19 -0400
From: Steven Rostedt <rostedt@...dmis.org>
To: Christoph Hellwig <hch@....de>
Cc: "Darrick J. Wong" <djwong@...nel.org>, linux-kernel@...r.kernel.org,
 linux-trace-kernel@...r.kernel.org, linux-xfs@...r.kernel.org, Masami
 Hiramatsu <mhiramat@...nel.org>, Mathieu Desnoyers
 <mathieu.desnoyers@...icios.com>, Andrew Morton
 <akpm@...ux-foundation.org>, Carlos Maiolino <cem@...nel.org>
Subject: Re: [PATCH 00/14] xfs: Remove unused trace events

On Mon, 16 Jun 2025 07:31:19 +0200
Christoph Hellwig <hch@....de> wrote:

> > I just did an analysis of this:
> > 
> >   https://lore.kernel.org/lkml/20250613104240.509ff13c@batman.local.home/T/#md81abade0df19ba9062fd51ced4458161f885ac3
> > 
> > A TRACE_EVENT() is about 5K, and each DEFINE_EVENT() is about 1K.  
> 
> That's really quite expensive.  And you only measured the tezt/data/bss

Yes. This is something I've spent a bit of time over the years trying
to address. With moving a bunch of code into trace_event.c with the
added expense that trace events do function calls.

It looks like it's still growing as the last time I checked it was just
under 5K (something around 4800 bytes) and now it's over 5K, and the
tracepoint code grew 4x. I'll start looking into "why" later when I
have more time to deal with this. My time budget for removing unused
events has pretty much dried up.

> overhead and not even the dynamic memory overhead, which is probably
> a lot more.

Yes, and this is another area I look to make better. It was the
motivation for eventfs which saved over 20 megs of memory by having
trace event files dynamically created instead of being permanent.

-- Steve