On Wed, 2012-06-06 at 03:10 +0200, Kay Sievers wrote:

On Wed, Jun 6, 2012 at 2:46 AM, Andrew Morton akpm@linux-foundation.org wrote:

...

On Tue, 05 Jun 2012 17:40:05 -0700 Joe Perches joe@perches.com wrote:

...

On Tue, 2012-06-05 at 17:37 -0700, Andrew Morton wrote:

...

On Tue, 05 Jun 2012 17:07:27 -0700 Joe Perches joe@perches.com wrote:

...

On Tue, 2012-06-05 at 16:58 -0700, Andrew Morton wrote:

...

echo "\0014Hello Joe" > /dev/kmsg

# echo -e "\x014Hello Me" > /dev/kmsg gives: 12,778,4057982669;Hello Me

That's changed behavior.

Which is an improvement too.

No it isn't. It exposes internal kernel implementation details in random weird inexplicable ways. It doesn't seem at all important though.

...

I very much doubt a single app will change because of this.

I doubt it as well.

Yeah, the value of injecting such binary data is kind of questionable. :)

Joe, maybe you can change printk_emit() to skip the prefix detection/stripping if a prefix is already passed to the function?

Sure, it's pretty trivial.

Perhaps all binary data data should be elided. Maybe print . for all non-printable ascii chars.