Re: [PATCHv2, RFC 13/30] thp, mm: implement grab_cache_huge_page_write_begin()

[Dave]

On 03/26/2013 03:48 AM, Kirill A. Shutemov wrote:
> Dave Hansen wrote:
>>> +repeat:
>>> +	page = find_lock_page(mapping, index);
>>> +	if (page) {
>>> +		if (!PageTransHuge(page)) {
>>> +			unlock_page(page);
>>> +			page_cache_release(page);
>>> +			return NULL;
>>> +		}
>>> +		goto found;
>>> +	}
>>> +
>>> +	page = alloc_pages(gfp_mask & ~gfp_notmask, HPAGE_PMD_ORDER);
>>
>> I alluded to this a second ago, but what's wrong with alloc_hugepage()?
> 
> It's defined only for !CONFIG_NUMA and only inside mm/huge_memory.c.

It's a short function, but you could easily pull it out from under the
#ifdef and export it.  I kinda like the idea of these things being
allocated in as few code paths possible.  But, it's not a big deal.

>>> +found:
>>> +	wait_on_page_writeback(page);
>>> +	return page;
>>> +}
>>> +#endif
>>
>> So, I diffed :
>>
>> -struct page *grab_cache_page_write_begin(struct address_space
>> vs.
>> +struct page *grab_cache_huge_page_write_begin(struct address_space
>>
>> They're just to similar to ignore.  Please consolidate them somehow.
> 
> Will do.
> 
>>> +found:
>>> +	wait_on_page_writeback(page);
>>> +	return page;
>>> +}
>>> +#endif
>>
>> In grab_cache_page_write_begin(), this "wait" is:
>>
>>         wait_for_stable_page(page);
>>
>> Why is it different here?
> 
> It was wait_on_page_writeback() in grab_cache_page_write_begin() when I forked
> it :(
> 
> See 1d1d1a7 mm: only enforce stable page writes if the backing device requires it
> 
> Consolidation will fix this.

Excellent.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/