Messages in this thread | | | Date | Mon, 30 Nov 2009 23:37:16 +0100 (CET) | From | Thomas Gleixner <> | Subject | Re: [rfc] "fair" rw spinlocks |
| |
On Mon, 30 Nov 2009, Linus Torvalds wrote: > On Mon, 30 Nov 2009, Thomas Gleixner wrote: > > > > Yeah, forgot to mention sched.c, but that's solvable > > It should be fairly easy to add a few 'spin_lock(&tasklist_lock)' around > stuff that really depended on exclusion from writers. That should > _hopefully_ be the rare case. > > The biggest problem is that there will almost inevitably be things that > get missed, and any races exposed by lacking locking will be _very_ hard > to debug and trigger. So what I'd be worried about is not getting to a > "practically working" state, but any really subtle cases that nobody > really hits in practice.
I'm aware of that. The number of places where we read_lock tasklist_lock is 79 in 36 files right now. That's not a horrible task to go through them one by one and do a case by case conversion with a proper changelog. That would only leave the write_lock sites.
We can then either do the rw_lock to spin_lock conversion or keep the rw_lock which has no readers anymore and behaves like a spinlock for a transition time so reverts of one of the read_lock -> rcu patches could be done to debug stuff.
Thanks,
tglx
| |