Messages in this thread | | | Date | Mon, 14 Feb 2011 13:54:00 -0500 | From | Jason Baron <> | Subject | Re: [PATCH 0/2] jump label: 2.6.38 updates |
| |
On Mon, Feb 14, 2011 at 11:43:43AM -0500, Mathieu Desnoyers wrote: > * Peter Zijlstra (peterz@infradead.org) wrote: > > On Mon, 2011-02-14 at 11:29 -0500, Jason Baron wrote: > > > On Mon, Feb 14, 2011 at 05:25:54PM +0100, Peter Zijlstra wrote: > > > > > > > > > > I remember that atomic_t is defined in types.h now rather than atomic.h. > > > > > Any reason why you should keep including atomic.h from jump_label.h ? > > > > > > > > Ooh, shiny.. we could probably move the few atomic_{read,inc,dec} users > > > > in jump_label.h into out of line functions and have this sorted. > > > > > > > > > > inc and dec sure, but atomic_read() for the disabled case needs to be > > > inline.... > > > > D'0h yes of course, I was thinking about jump_label_enabled(), but > > there's still the static_branch() implementation to consider. > > > > We could of course cheat implement our own version of atomic_read() in > > order to avoid the whole header mess, but that's not pretty at all > > > > OK, so the other way around then : why does kernel.h need to include > dynamic_debug.h (which includes jump_label.h) ? >
well, its used to dynamically enable/disable pr_debug() statements which actually have now moved to linux/printk.h, which is included by kernel.h.
I don't need an atomic_read() in the disabled case for dynamic debug, and I would be ok, #ifdef CONFIG_JUMP_LABEL, in dynamic_debug.h. Its not the prettiest solution. But I can certainly live with it for now, so that we can sort out the atomic_read() issue independently.
Peter, Mathieu, are you guys ok with this?
-Jason
| |