[Xen-devel] Re: user/hypervisor address space solution

[Keir Fraser <Keir.Fraser@xxxxxxxxxxxx>]

On 13 Feb 2006, at 22:56, Hollis Blanchard wrote:

> There are very few uses of get/put_user() in common code right now, and
> those can be trivially converted. However, there are some in arch code
> e.g. xen/arch/x86/domain.c) that cannot, and so the current put_user
> interface would need to be preserved as some arch-specific macro. In
> this case I'd call it "x86_put_user" to emphasize that only x86 arch
> code should be using it. Too many x86isms creep into places like
> grant_table.c...
I expect copy_to/from_user would also need to change. There are places 
where an array of structs are passed to a hcall, a pointer into that 
array is passed around in Xen, and individual array entries are copied 
in/out.
However, this approach seems little different on the Xen side from the 
xencomm allocator approach -- the pointers are still opaque handles 
from the p.o.v. of generic Xen code. So some interface changes would be 
required even if we went down that road.
This opacity should probably be represented in the public header file 
definitions and in the prototypes of uaccess functions, instead of 
pretending the handles are really 'foo *' pointers. That would help 
clean up all uses of the uaccess functions in Xen.
We'll have to see how this all pans out....

 -- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel