Re: [PATCH RFC 2/8] clock device: convert clock_gettime

[john stultz <johnstul@us.ibm.com>]

On Mon, 2010-11-08 at 07:56 +0100, Arnd Bergmann wrote:
> On Thursday 04 November 2010, Richard Cochran wrote:
> > diff --git a/include/linux/posix-timers.h b/include/linux/posix-timers.h
> > index 3e23844..70f40e6 100644
> > --- a/include/linux/posix-timers.h
> > +++ b/include/linux/posix-timers.h
> > @@ -17,10 +17,22 @@ struct cpu_timer_list {
> >  	int firing;
> >  };
> >  
> > +/* Bit fields within a clockid:
> > + *
> > + * The most significant 29 bits hold either a pid or a file descriptor.
> > + *
> > + * Bit 2 indicates whether a cpu clock refers to a thread or a process.
> > + *
> > + * Bits 1 and 0 give the type: PROF=0, VIRT=1, SCHED=2, or FD=3.
> > + *
> > + * A clockid is invalid if bits 2, 1, and 0 all set (see also CLOCK_INVALID
> > + * in include/linux/time.h)
> > + */
> > +
> >  #define 	(clock)		((pid_t) ~((clock) >> 3))
> >  #define CPUCLOCK_PERTHREAD(clock) \
> >  	(((clock) & (clockid_t) CPUCLOCK_PERTHREAD_MASK) != 0)
> > -#define CPUCLOCK_PID_MASK	7
> > +
> >  #define CPUCLOCK_PERTHREAD_MASK	4
> >  #define CPUCLOCK_WHICH(clock)	((clock) & (clockid_t) CPUCLOCK_CLOCK_MASK)
> >  #define CPUCLOCK_CLOCK_MASK	3
> > @@ -28,12 +40,17 @@ struct cpu_timer_list {
> >  #define CPUCLOCK_VIRT		1
> >  #define CPUCLOCK_SCHED		2
> >  #define CPUCLOCK_MAX		3
> > +#define CLOCKFD			CPUCLOCK_MAX
> > +#define CLOCKFD_MASK		(CPUCLOCK_PERTHREAD_MASK|CPUCLOCK_CLOCK_MASK)
> 
> It looks like you are turning a kernel internal interface into a user ABI,
> which I think is highly questionable. Using the bits like this internally is
> ok, but making it part of the syscall ABI means that we can never change this
> in the future.
> 
> Maybe I was misunderstanding your patch though.


As Richard already mentioned, the cpuclocks has already exported this as
ABI and its used in pthread_getcpuclockid().

I wonder thought if it would be worth having a syscall to convert from
fd to clock_id so it could be more flexible in the future. But it may
not be worth it, as we're probably already limited by the cpuclock
implementation.

thanks
-john