Re: PROBLEM: 3.0-rc kernels unbootable since -rc3

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Sun, Jul 10, 2011 at 01:45:35AM +0100, julie Sullivan wrote:
> >Very good.  I have created a new jms.2011.07.07a branch in the -rcu
> >git tree at:
> 
> I get my own branch! :-) (thanks for all your trouble with this.)
> 
> >Hello, Julie,
> 
> >Any news?
> 
> Sorry for the delay (again), I have to work very late Fridays and Saturdays.
> Anyway, I have tested the six commits on my v3.0-rc4 branch (with
> f8b7fc6b514f34a5 and a46e0899eec7a3069bc reverted) and can confirm
> that they _all_ produce a bootable kernel at my end.

I am not yet sure whether this is good news or bad news, but thank
you very much for all the testing!

> Btw, you may have noticed that my config has RCU_BOOST=n?
> If I build a simple v3.0-rc4 kernel (ie one which includes the four
> original commits (f8b7fc6b514, a46e0899eec, 09223371dea and
> 9a432736904)) but with RCU_BOOST=y it _will_ boot. I didn't know if
> this was helpful or not.

OK, that certainly tells me what the next patch in the series needs
to be.  The reverse of the usual experience, but so it goes.

> If it is helpful, you probably know this, but in order to get
> RCU_BOOST=y I have to change a couple of other configs - PREEMPT_RCU=y
> (previously n) and to get PREEMPT_RCU=y, my preemption model has to be
> 'Preemptible Kernel (Low-Latency Desktop)' - PREEMPT=y (instead of
> 'Voluntary Kernel Preemption (Desktop)' - PREEMPT=n which it was
> originally).

Agreed, to get RCU_BOOST, you need either TREE_PREEMPT_RCU or
TINY_PREEMPT_RCU, which requires PREEMPT.

> It's my day off tomorrow so I'll do some more testing (on -rc5 and
> -rc6) (past 1.30am here in the UK now).

I added another patch onto the jms.2011.07.07a in the -rcu git tree.
If you get a chance to test it, please let me know how it goes.

							Thanx, Paul