Re: [Xen-devel] Should we revert "mm: New XENMEM space, XENMAPSPACE_gmfn_range"?

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 01.08.12 at 19:55, Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx> 
>>> wrote:
> I was reading more about this commit because this patch breaks the ABI
> on ARM, when I realized that on x86 there is no standard that specifies
> the alignment of fields in a struct.

There is - the psABI supplements to the SVR4 ABI.

> As a consequence I don't think we can really be sure that between .domid
> and .space we always have 16 bits of padding.

It would be very strange for a modern ABI (other than perhaps
ones targeting exclusively embedded environments, where space
matters) to allow structure fields at mis-aligned offsets. Is that
really the case for ARM? This would make the compiled code
accessing such fields pretty ugly, since I seem to recall that loads
and stores are required to be aligned there.

> I am afraid that if a user compiles Linux or another guest kernel with a
> compiler other than gcc, this hypercall might break. In fact it already
> happened just switching from x86 to ARM.
> Also, considering that the memory.h interface is supposed to be ANSI C,
> isn't it wrong to assume compiler specific artifacts anyway?

This is not compiler specific, but platform defined. Compilers
merely need to conform to that specification; if they don't they
can't be used for building Xen interfacing code without manual
tweaking (perhaps re-creation) of the interface headers.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel