Re: [PATCH 4.4 24/28] can: dev: prevent potential information leak in can_fill_info()

[Pavel Machek]

On Tue 2021-02-02 22:51:01, Dan Carpenter wrote:
> On Tue, Feb 02, 2021 at 10:05:39PM +0300, Dan Carpenter wrote:
> > On Tue, Feb 02, 2021 at 07:53:17PM +0100, Pavel Machek wrote:
> > > Hi!
> > > 
> > > > From: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > > > 
> > > > [ Upstream commit b552766c872f5b0d90323b24e4c9e8fa67486dd5 ]
> > > > 
> > > > The "bec" struct isn't necessarily always initialized. For example, the
> > > > mcp251xfd_get_berr_counter() function doesn't initialize anything if the
> > > > interface is down.
> > > 
> > > Well, yes... and = {} does not neccessarily initialize all of the
> > > structure... for example padding.
> > > 
> > > It is really simple
> > > 
> > > struct can_berr_counter {
> > > 	__u16 txerr;
> > > 	__u16 rxerr;
> > > };
> > > 
> > > but maybe something like alpha uses padding in such case, and memset
> > > would be better?
> > 
> > I'm pretty sure nothing uses padding in this situation.  If it does then
> > we need to re-work a bunch of code.
> 
> Not necessarily related but in theory a "= {};" assignment is a GCC
> extension and it is supposed to zero out struct holes.  If the code
> does "= {0};" then that's standard C, and will not necessarily fill
> struct holes but I think GCC tries to.  The other complication is that
> some GCC versions have bugs related to this?  We had a long thread about
> this last August.
> 
> https://lore.kernel.org/lkml/20200801144030.GM24045@xxxxxxxx/
> 
> Anyway, this code has no holes so it's not affected.

Thanks for pointers. I remembered "just do memset", but there are
clearly more nuances to this.

Best regards,
								Pavel
-- 
http://www.livejournal.com/~pavelmachek

Attachment: signature.asc
Description: PGP signature