Re: [PATCH] tracing: Replace deprecated strncpy() with memcpy() for stack_trace_filter_buf

[Mathieu Desnoyers]

On 2025-04-04 08:28, Devaansh Kumar wrote:
> On Fri, 4 Apr 2025 at 01:05, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> > On Fri,  4 Apr 2025 00:43:40 +0530
> > Devaansh Kumar <devaanshk840@xxxxxxxxx> wrote:
> >
> > > @@ -537,14 +538,16 @@ stack_trace_sysctl(struct ctl_table *table, int write, void *buffer,
> > >        return ret;
> > >   }
> > >
> > > -static char stack_trace_filter_buf[COMMAND_LINE_SIZE+1] __initdata;
> > > +static char stack_trace_filter_buf[COMMAND_LINE_SIZE+1] __initdata __nonstring;
> > >
> > >   static __init int enable_stacktrace(char *str)
> > >   {
> > >        int len;
> > >
> > > -     if ((len = str_has_prefix(str, "_filter=")))
> > > -             strncpy(stack_trace_filter_buf, str + len, COMMAND_LINE_SIZE);
> > > +     len = str_has_prefix(str, "_filter=");
> > > +
> > > +     if (len)
> > > +             memcpy(stack_trace_filter_buf, str + len, sizeof(stack_trace_filter_buf));
> >
> > Hmm, this location looks like it can just use strscpy().
>
> Yes strscpy() also works. But since stack_trace_filter_buf is length
> bounded, shouldn't memcpy be the right choice?

It's not only about the destination, but also about the source length.

AFAIU, turning a strncpy into a memcpy here will overflow reading the
input @str if the input string is smaller than
sizeof(stack_trace_filter_buf) + len.

This can trigger page faults or make KASAN unhappy.

Thanks,

Mathieu


--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com