Re: [PATCH v4 2/2] powerpc: Uprobes port to powerpc

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2012-08-23 at 21:47 +0530, Srikar Dronamraju wrote:
> * Benjamin Herrenschmidt <benh@kernel.crashing.org> [2012-08-23 20:06:18]:
> 
> > On Thu, 2012-08-23 at 11:02 +0530, Srikar Dronamraju wrote:
> > > > 
> > > 
> > > insn is updated/accessed in the arch independent code. Size of
> > > uprobe_opcode_t could be different for different archs.
> > > uprobe_opcode_t
> > > represents the size of the smallest breakpoint instruction for an
> > > arch.
> > > 
> > > Hence u8 works out the best. I know we could still use uprobe_opcode_t
> > > and achieve the same. In which case, we would have to interpret
> > > MAX_UINSN_BYTES differently. Do you see any advantages of using
> > > uprobe_opcode_t instead of u8 across archs?
> > 
> > But don't you actively rely on the fact that on powerpc, unlike x86, you
> > -can- atomically replace an instruction with a single 32-bit store ?
> > 
> 
> We are not doing a replace here, we are only copying from the ->vm_file
> for the largest size instruction possible for that instruction. For
> powerpc, this is easy because of fixed size instructions.  
> 
> On other archs, at this point, we dont even know the length of the
> underlying instruction.
> 
> Now there are 3 ways to handle this:
> 1. use arch independent copy_insn() (current.) (handles if the
> instruction spreads across multiple pages on non fixed instruction
> archs). 
> 
> 2. make the copy_insn() arch specific, that would mean every arch will
> have to do read_mapping_page etc.
> 
> 3. have a arch specific hook in arch independent copy_insn code that
> either does a memcpy for non fixed instruction archs or does an
> assignment in archs like powerpc.
>
> I think you are suggesting option 3.
> But instead of adding another call that does the arch specific stuff, we
> are probably be better of doing a memcpy. Right?
> 
> For all powerpc references to insn we could refer to it as u32 as
> suggested by Oleg.

Ok, doens't matter much either way, it's just odd and inefficient.

Cheers,
Ben.