Re: [PATCH 1/1] hv_balloon: Enable hot-add for memblock sizes > 128 Mbytes

[David Hildenbrand <david@redhat.com>]

On 29.04.24 17:30, Michael Kelley wrote:
> From: Michael Kelley <mhklinux@outlook.com> Sent: Friday, April 26, 2024 9:36 AM
>>>> @@ -505,8 +505,9 @@ enum hv_dm_state {
>>>>
>>>>    static __u8 recv_buffer[HV_HYP_PAGE_SIZE];
>>>>    static __u8 balloon_up_send_buffer[HV_HYP_PAGE_SIZE];
>>>> +static unsigned long ha_chunk_pgs;
>>>
>>> Why not stick to PAGES_IN_2M and call this
>>>
>>> ha_pages_in_chunk? Much easier to get than "pgs".
>>
>> OK.  I was trying to keep the new identifier short so that
>> mechanically substituting it for HA_CHUNK didn't blow up
>> the line length.
>>
>>>
>>> Apart from that looks good. Some helper macros to convert size to chunks
>>> etc. might make the code even more readable.
>>
>> I'll look at this.  Might help the line length problem too.
>>
> 
> I didn't see any particular complexity in converting size to chunks. But
> this slightly opaque sequence is repeated in three places:
> 
> 	new_inc = (residual / HA_CHUNK) * HA_CHUNK;
> 	if (residual % HA_CHUNK)
> 		new_inc += HA_CHUNK;
> 
> If HA_CHUNK (or the new memblock size based variable) is a
> power of 2, then these can become:
> 
> 	new_inc = ALIGN(residual, HA_CHUNK);
> 
> which is a lot better.  I'll make that change, and a couple of other
> changes where things are open coded that could be existing
> kernel macros.

Cool! Make sure to CC me on v2.

> 
> Question:  Is memblock size guaranteed to be a power of 2?  It looks
> to be so in the x86 code, but I can't tell on s390 and ppc.  For safety,
> I'll add a check in the Hyper-V balloon driver init code, as the
> communication with Hyper-V expects a power of 2.

Yes, in memory_dev_init() we make sure that the size is a power of 2, 
and if it is not, we would panic().

-- 
Cheers,

David / dhildenb