Re: [PATCH 1 of 3] KEXEC: Add new hypercall to fix 64/32bit truncation errors. v2

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 23/12/11 08:34, Jan Beulich wrote:
>>>> On 22.12.11 at 18:36, Andrew Cooper <andrew.cooper3@citrix.com> wrote:
>> Currently, KEXEC_CMD_kexec_get_range in compat mode, or with 32bit
>> Xen, will truncate 64bit pointers to 32bits, leading to problems on
>> machines with more than 4GiB of RAM.  As 32bit dom0 kernels tend to be
>> PAE capable, this is especially wasteful, as most structures currently
>> limited to <4GiB could easily be <64GiB instead.
>>
>> Therefore, introduce a new hypercall 'KEXEC_CMD_kexec64_get_range'
>> which passes similar information as KEXEC_CMD_kexec_get_range, but
>> which uses a structure with explicitly specified field widths, causing
>> it to be identical in the compat and regular case.  This new
>> structure, xen_kexec64_range_t, will be the same as xen_kexec_range_t
>> if Xen is compiled for x86_64, but will still use 64bit integers if
>> Xen is compiled for x86_32.
>>
>> To fix 32bit Xen which uses 32bit integers for addresses and sizes,
>> change the internals to use xen_kexec64_range_t which will use 64bit
>> integers instead.  This also involves changing several casts to
>> explicitly use uint64_ts rather than unsigned longs.
> Just for the record - I continue to be opposed to doing this for the
> 32-bit hypervisor. All relevant allocations are below 4G there, so
> there's simply no need to decrease code readability for no benefit.
>
> Jan

I don't mind reducing the scope to ignore the 32bit Xen case.  Any
object Keir?

<snip>

-- 
Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer
T: +44 (0)1223 225 900, http://www.citrix.com