Re: [RFC] GDB patches for hw watchpoints - revised

[Wu Zhou <woodzltc@cn.ibm.com>]

On Tue, 6 Dec 2005, Daniel Jacobowitz wrote:

> On Tue, Dec 06, 2005 at 10:25:01PM +0200, Eli Zaretskii wrote:
> > > Date: Tue, 6 Dec 2005 14:12:26 +0800 (CST)
> > > From: Wu Zhou <woodzltc@cn.ibm.com>
> > > cc: drow@false.org, eliz@gnu.org, mark.kettenis@xs4all.nl
> > > 
> > > 3. Eli ever expressed a concern that the PPC doesn't have a way to return 
> > > the data address that triggered the watchpoint?  As far as I think, the 
> > > reason is that PPC will only have one DABR (if it does have). So maybe we 
> > > don't need to have such a method.  
> > 
> > Sorry, I'm not following: no matter how many debug registers the PPC
> > has, it can still return to GDB the data address that triggered the
> > watchpoint.  I don't think the higher levels of GDB (breakpoint.c)
> > should know or assume anything about the target capabilities or
> > resources.  Let's try keeping the code clean of such peculiarities.
> 
> Wu, you're using GETSIGINFO to check for a watchpoint.  Then isn't the
> faulting address stored in the siginfo anyway?

Currently what stored in the siginfo.si_addr is the address of next
instruction. Anton sent me a patch to change that to the data address 
instead. I am now playing with the patched kernel.

BTW. It seems that to_stopped_data_address is only used in rwatch and 
awatch, which is not that frequently used as watch itself.  So may we 
postpone its implementation till some later time?

Regards
- Wu Zhou