Messages in this thread | | | Date | Mon, 28 Apr 2008 19:31:31 +0200 | From | "Bert Wesarg" <> | Subject | Re: [PATCH] Fix calculus of bitmap_scnprintf_len() |
| |
On Mon, Apr 28, 2008 at 7:09 PM, Mike Travis <travis@sgi.com> wrote: > Paul Jackson wrote: > > Bert wrote: > >> The function bitmap_scnprintf_len() is currently not used > > > > Paul Jackson wrote: > >> How about we just remove that function? > > > > WANG Cong wrote: > >> I am afraid no. See: > >> > >> include/linux/cpumask.h:292: return bitmap_scnprintf_len(len); > > > > Good point. > > > > > > Then how about we also remove from cpumask.h: > > > > #define cpumask_scnprintf_len(len) \ > > __cpumask_scnprintf_len((len)) > > static inline int __cpumask_scnprintf_len(int len) > > { > > return bitmap_scnprintf_len(len); > > } > > > > That's fine with me. A later version of the patch did have > the function removed but it didn't get picked up. The other > changes there were to use function pointers instead of the > flag variable to select list or mask output format, and the > addition of mask variants for the cpu/{present,possible, > online,system} map outputs. I'm fine with this too. I did a hasty audit of cpumask_scnprintf() users, and no one can use these functions.
But one last note to the cpumask_scnprintf_len() macro: this macro should really not have an argument, it should be forced to NR_CPUS. Else a user could have a too small buffer for the call to cpumask_scnprintf(), which always calls bitmap_scnprintf() with NR_CPUS nbits.
> > I'll dig that one back up and resubmit it. Fine.
Regards. Bert > > Thanks, > Mike >
| |