Re: [PATCH] kprobes for s390 architecture

[Mike Grundy <grundym at us dot ibm dot com>]

On Wed, Jun 21, 2006 at 06:38:40PM +0200, Martin Schwidefsky wrote:
> On Tue, 2006-06-20 at 21:28 -0700, Mike Grundy wrote:
> > Hi Martin - This patch implements the suggestions from your review. There were
> > a couple points I wanted to go over:
> > > There are some more instructions missing that need fixup:
> > > "brxh" 0x84??????, "brxle" 0x85??????, "brc" 0xa7?4????,
> > > "brct" 0xa7?6????, "brctg" 0xa7?7????, "bctgr" 0xb946????,
> > > "brxhg" 0xec????????44 and "brxlg" 0xec??????45.
> > Since all of these are relative branches, and they don't save the psw, the
> > standard clean up of adjusting the original psw by the offset from the out of
> > line address after single step. Unless I'm just being dense :-) 
> 
> All of these are conditional branches, if the branch is not taken you
> have to do a cleanup.
The reason I have a special cleanup for the other branches is the easy way to 
tell if the branch wasn't taken is the pswa = orig pswa + instruction length.
The relative branches get cleaned up the same way if the branch was taken or
not, pswa = probe_addr + (out of line end psw - out of line start psw). These
are all relative branches and while they need cleanup, they don't get treated
differently based on the branch status.

> You misunderstood me here. I'm not talking about storing the same piece
> of data to memory on each processor. I'm talking about isolating all
> other cpus so that the initiating cpu can store the breakpoint to memory
Yep, I misunderstood that. The serialization is the point, not the replacement
of a word in memory.

-- 
Thanks
Mike