Re: [PATCH 3/3] xen/privcmd: add PRIVCMD_MMAPBATCH_V2 ioctl

[David Vrabel <david.vrabel@citrix.com>]

On 24/08/12 12:41, Konrad Rzeszutek Wilk wrote:
>>>> +struct privcmd_mmapbatch_v2 {
>>>> +	unsigned int num; /* number of pages to populate */
>>>
>>> unsigend int? Not 'u32'?
>>>> +	domid_t dom;      /* target domain */
>>>> +	__u64 addr;       /* virtual address */
>>>> +	const xen_pfn_t __user *arr; /* array of mfns */
>>>> +	int __user *err;  /* array of error codes */
>>>
>>> int? Not a specific type?
>>
>> It's an existing interface supported by classic Xen kernels and
>> currently being used by libxc.  So while I agree that it's not the best
>> interface, I don't think it can be changed.

It's also the same as struct privcmd_mmapbatch except for the extra
'err' field and 'arr' being const.

> How does it work with a 64-bit dom0 and 32-bit userspace? Is the libxc
> smart enough to figure out the size of the structure?

privcmd doesn't support compat ioctls because there there is nothing
doing the translation of the hypercalls from the 32-bit to 64-bit ABI --
the hypervisor won't do it as the hypercalls are called from the 64-bit
kernel.

64 bit Xen, 64 bit dom0, 32 bit tools has never worked for this reason.
 I think Ian Campbell had some hacky patches for this but he may not
want to admit to them. ;)

David