Re: new text patching for review

[Andi Kleen <ak@suse.de>]

> Ewwwwwwwwwww.... you plan to run this in SMP ? So you actually go byte
> by byte changing pieces of instructions non atomically and doing
> non-Intel's errata friendly XMC. You are really looking for trouble
> there :) Two distinct errors can occur: 

In this case it is ok because this only happens when transitioning
from 1 CPU to 2 CPUs or vice versa and in both cases the other CPUs
are essentially stopped.

All the other manipulations currently are single byte.

I suppose for your immediate value patches something stronger is needed,
but we can worry about that post .23.

> What I don't like about this particular implementation is that it does
> not support "poking" more than 1 byte. In order to support this, you
> would have to deal with the case where the address range spans over more
> than one page.

I considered it, but the function would have been at least twice as big
to handle all the corner cases. And for the current callers it's all fine.

> Also, doing the copy in the same interface seems a bit awkward.

Splitting it would also seem quite awkward.
 
> 
> I would much prefer something like:
> 
> void *map_shadow_write(void *addr, size_t len);
> (returns a pointer to the shadow writable pages, at the same page offset
> as "addr")
> 
> int unmap_shadow_write(void *shadow_addr, size_t len);
> (unmap the shadow pages)
> 
> Then, the in-kernel user is free to modify their pages as they like.
> Since we cannot foresee each modification pattern, I think that leaving
> this kind of flexibility is useful.

You could as well call vmap directly then; it's not that much
more complicated.  I don't really see much value in complicating
it right now.

-Andi