Re: [PATCH v7 11/12] zsmalloc: page migration support

[Vlastimil Babka <vbabka@suse.cz>]

On 06/02/2016 02:25 AM, Minchan Kim wrote:
> On Wed, Jun 01, 2016 at 04:09:26PM +0200, Vlastimil Babka wrote:
>> On 06/01/2016 01:21 AM, Minchan Kim wrote:
>>> +	reset_page(page);
>>> +	put_page(page);
>>> +	page = newpage;
>>> +
>>> +	ret = 0;
>>> +unpin_objects:
>>> +	for (addr = s_addr + offset; addr < s_addr + pos;
>>> +						addr += class->size) {
>>> +		head = obj_to_head(page, addr);
>>> +		if (head & OBJ_ALLOCATED_TAG) {
>>> +			handle = head & ~OBJ_ALLOCATED_TAG;
>>> +			if (!testpin_tag(handle))
>>> +				BUG();
>>> +			unpin_tag(handle);
>>> +		}
>>> +	}
>>> +	kunmap_atomic(s_addr);
>>
>> The above seems suspicious to me. In the success case, page points to
>> newpage, but s_addr is still the original one?
>
> s_addr is virtual adress of old page by kmap_atomic so page pointer of
> new page doesn't matter.

Hmm, I see. The value (head address/handle) it reads from the old page 
should be the same as the one in the newpage. And this value doesn't get 
changed in the process. So it works, it's just subtle :)