Re: [PATCH] Allow times and time system calls to return smallnegative values

[Joakim Tjernlund]

On Thu, 2008-12-04 at 16:40 +0100, Ingo Molnar wrote:
> * Paul Mackerras <paulus@xxxxxxxxx> wrote:
> 
> > Ingo Molnar writes:
> > 
> > > > +	force_successful_syscall_return();
> > > >  	return compat_jiffies_to_clock_t(jiffies);
> > > 
> > > just curious: what code does force_successful_syscall_return() actually 
> > > run in the powerpc case - those bits are missing from this patch. I 
> > > suspect it sets some sort of flag?
> > 
> > It sets the TIF_NOERROR thread flag, which is tested in the syscall 
> > exit path along with various other thread flags.
> > 
> > force_successful_syscall_return() is defined for powerpc in 
> > arch/powerpc/include/asm/ptrace.h.  It's nothing new, it has existed 
> > for ages and is used in a few other places already.  It has existing 
> > non-null definitions on alpha, ia64, powerpc, and sparc64.
> 
> ok, was just curious. To make this code nicer, we could keep the time 
> syscalls from returning small negative numbers.
> 
> the core issue is probably:
> 
>  jiffies.h:#define INITIAL_JIFFIES ((unsigned long)(unsigned int) (-300*HZ))
> 
> the thing is, we did INITIAL_JIFFIES to find _in kernel_ jiffies 
> wraparoudn bugs (hung timers, broken drivers, etc.) - we did not want to 
> confuse userspace with it.
> 
> So how about adding INITIAL_JIFFIES to the sys_times() return value?

Na, that will just hide the bug for some 497 days(or 49 days on alpha).
In user space on have to use this workaround(and I think some variation
is in glibc already):

static unsigned long quagga_times(void)
{
#if defined(GNU_LINUX)
  unsigned long ret;

  errno = 0;
  ret = times(NULL); /* Linux can handle NULL */
  /* Workaround broken syscall impl.
   * A bugfix exists for the kernel, hopefully
   * it will make it into 2.6.29
   */
  if (errno)
    ret = (unsigned long) (-errno);
  return ret;
#else
  struct tms dummy; /* Only return value is used */

  return times(&dummy);
#endif
}


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/