Re: Potential out-of-bounds in ftrace_regex_release

[Steven Rostedt <rostedt@goodmis.org>]

On Mon, 14 Oct 2013 12:29:13 +0400
Andrey Konovalov <andreyknvl@google.com> wrote:

> Testing now with your patch.
> I've seen this report only twice, so it will be difficult to say if
> it's not happening any more or just not triggered.

Can I assume that this is fixed? I'll put it in for 3.12 and mark  it
for stable too.

-- Steve

> 
> On Thu, Oct 10, 2013 at 6:23 AM, Steven Rostedt <rostedt@goodmis.org> wrote:
> > On Wed, 9 Oct 2013 14:05:26 +0400
> > Andrey Konovalov <andreyknvl@google.com> wrote:
> >time cat trace > /tmp/trace
> >> So I still think that the bug is in 'trage_get_user()':
> >> Checking that 'parser->idx < parser->size - 1' is not performed in 'if
> >> (isspace(ch))' section, so 'parser->idx' becomes equal to
> >> 'parser->size' after 'parser->buffer[parser->idx++] = ch;'.
> >> (However in 'while (cnt && !isspace(ch))' loop checking is performed.)
> >
> > Yep, you are correct. I put in some printk's and did same writing to
> > the set_events file and was able to prove that I could get that "0"
> > written into the +1 overflow boundary.
> >
> > Can you try this patch to see if it fixes it for you.
> >
> > Thanks,
> >
> > -- Steve
> >
> > diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> > index d5f7c4d..063a92b 100644
> > --- a/kernel/trace/trace.c
> > +++ b/kernel/trace/trace.c
> > @@ -843,9 +843,12 @@ int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
> >         if (isspace(ch)) {
> >                 parser->buffer[parser->idx] = 0;
> >                 parser->cont = false;
> > -       } else {
> > +       } else if (parser->idx < parser->size - 1) {
> >                 parser->cont = true;
> >                 parser->buffer[parser->idx++] = ch;
> > +       } else {
> > +               ret = -EINVAL;
> > +               goto out;
> >         }
> >
> >         *ppos += read;
> >