Re: [patch 01/11] Introducing generic hardware breakpoint handler interfaces

[Ingo Molnar <mingo@elte.hu>]

* Alan Stern <stern@rowland.harvard.edu> wrote:

> On Tue, 10 Mar 2009, Ingo Molnar wrote:
> 
> > * prasad@linux.vnet.ibm.com <prasad@linux.vnet.ibm.com> wrote:
> > 
> > 
> > > +static u8			tprio[HB_NUM];	/* Thread bp max priorities */
> > > +LIST_HEAD(kernel_bps);			/* Kernel breakpoint list */
> > > +static LIST_HEAD(thread_list);			/* thread_hw_breakpoint list */
> > > +DEFINE_PER_CPU(struct cpu_hw_breakpoint, cpu_bp);
> 
> If nobody minds, I'll answer some of these questions on 
> Prasad's behalf because they address parts of the code that 
> were written before he took over the project.
> 
> > hm, why do we need the whole 'priority' mechanism? It seems 
> > very over-designed to me.
> 
> This was done at Roland McGrath's express request.  We should 
> see what he has to say about it.
> 
> > The likelyhood of both user-space and kernel-space to use 
> > hw-breakpoints is very low to begin with. And if they use 
> > them, the likelyhood of there being more than 4 debugregs 
> > required in the same context is even lower.
> 
> Not all architectures have 4 debug registers.  Most have only 
> one.
>
> > If that happens we shouldnt try to be too smart about them - 
> > just override user-space ones with kernel space ones and 
> > that's it. No explicit priorities are needed.
> 
> Roland really did not want it done this way.

Well i guess i'll have to wait for Roland's reply then.

	Ingo