Re: [PATCH 19/23] truncate: support huge pages

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 08/03/2013 07:17 PM, Kirill A. Shutemov wrote:
> If a huge page is only partly in the range we zero out the part,
> exactly like we do for partial small pages.

What's the logic behind this behaviour?  Seems like the kind of place
that we would really want to be splitting pages.

> +	if (partial_thp_start || lstart & ~PAGE_CACHE_MASK) {
> +		pgoff_t off;
> +		struct page *page;
> +		unsigned pstart, pend;
> +		void (*zero_segment)(struct page *page,
> +				unsigned start, unsigned len);
> +retry_partial_start:
> +		if (partial_thp_start) {
> +			zero_segment = zero_huge_user_segment;

That's a pretty hackish way to conditionally call a function, especially
since its done twice in one function. :)

I seem to recall zero_user_segment() vs. zero_huge_user_segment() being
something that caused some ugliness in the previous versions too.
What's the barrier to just having a smart zero_..._user_segment()
function that can conditionally perform huge or base page-zeroing?

> +		if (partial_thp_end) {
> +			zero_segment = zero_huge_user_segment;
> +			off = end & ~HPAGE_CACHE_INDEX_MASK;
> +			pend = (lend - 1) & ~HPAGE_PMD_MASK;
> +		} else {
> +			zero_segment = zero_user_segment;
> +			off = end;
> +			pend = (lend - 1) & ~PAGE_CACHE_MASK;
> +		}

We went though a similar exercise for the fault code (I think), but I
really think you need to refactor this.  Way too much of the code is in
the style:

	if (thp) {
		// new behavior
	} else {
		// old behavior
	}

To me, that's just a recipe that makes it hard to review, and I also bet
it'll make the thp much more prone to bitrot.  Maybe something like this?

	size_t page_cache_mask = PAGE_CACHE_MASK;
	unsigned long end_mask = 0UL;
	
	if (partial_thp_end) {
		page_cache_mask = HPAGE_PMD_MASK;
		end_mask = HPAGE_CACHE_INDEX_MASK;
	}
	...
	magic_zero_user_segment(...);
	off = end & ~end_mask;
	pend = (lend - 1) & ~page_cache_mask;

Like I said before, I somehow like to rewrite your code. :)

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]