Re: [PATCH v2] printk: Add kernel parameter to disable writes to /dev/kmsg

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 25 Apr 2016 21:24:35 +0200 Peter Zijlstra <peterz@infradead.org> wrote:

> On Mon, Apr 25, 2016 at 12:18:15PM -0700, Greg Kroah-Hartman wrote:
> > On Mon, Apr 25, 2016 at 09:06:51PM +0200, Peter Zijlstra wrote:
> > > On Mon, Apr 25, 2016 at 02:56:06PM -0400, Steven Rostedt wrote:
> > > > +static bool __read_mostly devkmsg_disabled;
> > > > +static int __init disable_devkmsg(char *str)
> > > > +{
> > > > +	devkmsg_disabled = true;
> > > > +	return 0;
> > > > +}
> > > > +__setup("printk.disable_kmsg_write", disable_devkmsg);
> > > 
> > > Again, please default enable and use an easier name to toggle this.
> > > Userspace flooding this with junk is really insane.
> > 
> > If you default enable, you break working systems today that want to log
> > stuff through this interface, sorry.
> 
> Oh, you mean those 'working' systems that livelocked my box because I
> had console to serial enabled which could not keep up with the endless
> flood of shite?
> 
> This systemd exposure has seriously eroded your sanity.

I'm trying to remember why we added /dev/kmsg in the first place - it
seems a quite stupid idea.  Why the heck should userspace be able to
diddle with the kernel->userspace messaging stream?  Sigh.

This:

commit e1b19d6e06a98874018c66990465dcc69b69c4bf
Author: Andrew Morton <akpm@zip.com.au>
Date:   Tue Aug 13 06:12:54 2002 -0700

    [PATCH] printk from userspace
    
    The patch allows userspace to issue printk's, via sys_syslog().
    
    The main use of this is within hpa's klibc - initial userspace needs a
    way of logging information and this API allows that information to be
    captured into the printk ringbuffer.  It ends up in /var/log/messages.
    
    Messages are truncated at 1024 characters by printk's vsprintf().
    
    Requires CAP_SYS_ADMIN.


This later got reverted and was replaced with /dev/kmsg and /proc/kmsg.


I guess that's a legitimate enough reason but in retrospect we should
have found a way of just shutting it off after booting has completed.