Re: [RFC] Recursive printk

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

On Sat, 6 Dec 2008 08:33:05 +0100 Willy Tarreau <w@1wt.eu> wrote:

> On Fri, Dec 05, 2008 at 11:20:16PM -0800, Andrew Morton wrote:
> > > void snd_verbose_printk(const char *file, int line, const char *format, ...)
> > > {
> > >         va_list args;
> > > 
> > >         if (format[0] == '<' && format[1] >= '0' && format[1] <= '7' && format[2] == '>') {
> > >                 char tmp[] = "<0>";
> > >                 tmp[1] = format[1];
> > >                 printk("%sALSA %s:%d: ", tmp, file, line);
> > >                 format += 3;
> > 
> > That's racy.  Two threads can fight over tmp[1].  It should do:
> > 
> > 		printk("<%c>ALSA %s:%d: ", format[1], tmp, file, line);
> > 
> > (I didn't know that you can even modify literal strings - shouldn't
> > they be in read-only storage?)
> 
> no Andrew, this tmp[] is declared on the stack, and gcc emits code to
> copy the constant "<0>" onto the stack every time this code is called

Good heavens, so it does.  It is unusual for C to cater to such stupid
code by generating such inefficient code.  A bit of a trap, really.