HugeTLB¶

This section is to explain how HugeTLB Vmemmap Optimization (HVO) works.

Thestructpage structures are used to describe a physical page frame. Bydefault, there is a one-to-one mapping from a page frame to its correspondingstructpage.

HugeTLB pages consist of multiple base page size pages and is supported by manyarchitectures. SeeHugeTLB Pages for moredetails. On the x86-64 architecture, HugeTLB pages of size 2MB and 1GB arecurrently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB pageconsists of 512 base pages and a 1GB HugeTLB page consists of 262144 base pages.For each base page, there is a correspondingstructpage.

Within the HugeTLB subsystem, only the first 4structpage are used tocontain unique information about a HugeTLB page.__NR_USED_SUBPAGE providesthis upper limit. The only ‘useful’ information in the remainingstructpageis the compound_head field, and this field is the same for all tail pages.

By removing redundantstructpage for HugeTLB pages, memory can be returnedto the buddy allocator for other uses.

Different architectures support different HugeTLB pages. For example, thefollowing table is the HugeTLB page size supported by x86 and arm64architectures. Because arm64 supports 4k, 16k, and 64k base pages andsupports contiguous entries, so it supports many kinds of sizes of HugeTLBpage.

When the system boot up, every HugeTLB page has more than onestructpagestructs which size is (unit: pages):

struct_size = HugeTLB_Size / PAGE_SIZE * sizeof(struct page) / PAGE_SIZE

Where HugeTLB_Size is the size of the HugeTLB page. We know that the sizeof the HugeTLB page is always n times PAGE_SIZE. So we can get the followingrelationship:

HugeTLB_Size = n * PAGE_SIZE

Then:

struct_size = n * PAGE_SIZE / PAGE_SIZE * sizeof(struct page) / PAGE_SIZE            = n * sizeof(struct page) / PAGE_SIZE

We can use huge mapping at the pud/pmd level for the HugeTLB page.

For the HugeTLB page of the pmd level mapping, then:

struct_size = n * sizeof(struct page) / PAGE_SIZE            = PAGE_SIZE / sizeof(pte_t) * sizeof(struct page) / PAGE_SIZE            = sizeof(struct page) / sizeof(pte_t)            = 64 / 8            = 8 (pages)

Where n is how many pte entries which one page can contains. So the value ofn is (PAGE_SIZE / sizeof(pte_t)).

This optimization only supports 64-bit system, so the value of sizeof(pte_t)is 8. And this optimization also applicable only when the size ofstructpageis a power of two. In most cases, the size ofstructpage is 64 bytes (e.g.x86-64 and arm64). So if we use pmd level mapping for a HugeTLB page, thesize ofstructpage structs of it is 8 page frames which size depends on thesize of the base page.

For the HugeTLB page of the pud level mapping, then:

struct_size = PAGE_SIZE / sizeof(pmd_t) * struct_size(pmd)            = PAGE_SIZE / 8 * 8 (pages)            = PAGE_SIZE (pages)

Where the struct_size(pmd) is the size of thestructpage structs of aHugeTLB page of the pmd level mapping.

E.g.: A 2MB HugeTLB page on x86_64 consists in 8 page frames while 1GBHugeTLB page consists in 4096.

Next, we take the pmd level mapping of the HugeTLB page as an example toshow the internal implementation of this optimization. There are 8 pagesstructpage structs associated with a HugeTLB page which is pmd mapped.

Here is how things look before optimization:

   HugeTLB                  struct pages(8 pages)         page frame(8 pages)+-----------+ ---virt_to_page---> +-----------+   mapping to   +-----------+|           |                     |     0     | -------------> |     0     ||           |                     +-----------+                +-----------+|           |                     |     1     | -------------> |     1     ||           |                     +-----------+                +-----------+|           |                     |     2     | -------------> |     2     ||           |                     +-----------+                +-----------+|           |                     |     3     | -------------> |     3     ||           |                     +-----------+                +-----------+|           |                     |     4     | -------------> |     4     ||    PMD    |                     +-----------+                +-----------+|   level   |                     |     5     | -------------> |     5     ||  mapping  |                     +-----------+                +-----------+|           |                     |     6     | -------------> |     6     ||           |                     +-----------+                +-----------+|           |                     |     7     | -------------> |     7     ||           |                     +-----------+                +-----------+|           ||           ||           |+-----------+

The value of page->compound_head is the same for all tail pages. The firstpage ofstructpage (page 0) associated with the HugeTLB page contains the 4structpage necessary to describe the HugeTLB. The only use of the remainingpages ofstructpage (page 1 to page 7) is to point to page->compound_head.Therefore, we can remap pages 1 to 7 to page 0. Only 1 page ofstructpagewill be used for each HugeTLB page. This will allow us to free the remaining7 pages to the buddy allocator.

Here is how things look after remapping:

   HugeTLB                  struct pages(8 pages)         page frame(8 pages)+-----------+ ---virt_to_page---> +-----------+   mapping to   +-----------+|           |                     |     0     | -------------> |     0     ||           |                     +-----------+                +-----------+|           |                     |     1     | ---------------^ ^ ^ ^ ^ ^ ^|           |                     +-----------+                  | | | | | ||           |                     |     2     | -----------------+ | | | | ||           |                     +-----------+                    | | | | ||           |                     |     3     | -------------------+ | | | ||           |                     +-----------+                      | | | ||           |                     |     4     | ---------------------+ | | ||    PMD    |                     +-----------+                        | | ||   level   |                     |     5     | -----------------------+ | ||  mapping  |                     +-----------+                          | ||           |                     |     6     | -------------------------+ ||           |                     +-----------+                            ||           |                     |     7     | ---------------------------+|           |                     +-----------+|           ||           ||           |+-----------+

When a HugeTLB is freed to the buddy system, we should allocate 7 pages forvmemmap pages and restore the previous mapping relationship.

For the HugeTLB page of the pud level mapping. It is similar to the former.We also can use this approach to free (PAGE_SIZE - 1) vmemmap pages.

Apart from the HugeTLB page of the pmd/pud level mapping, some architectures(e.g. aarch64) provides a contiguous bit in the translation table entriesthat hints to the MMU to indicate that it is one of a contiguous set ofentries that can be cached in a single TLB entry.

The contiguous bit is used to increase the mapping size at the pmd and pte(last) level. So this type of HugeTLB page can be optimized only when itssize of thestructpage structs is greater than1 page.

Notice: The head vmemmap page is not freed to the buddy allocator and alltail vmemmap pages are mapped to the head vmemmap page frame. So we can seemore than onestructpagestructwithPG_head (e.g. 8 per 2 MB HugeTLBpage) associated with each HugeTLB page. Thecompound_head() can handlethis correctly. There is onlyone headstructpage, the tailstructpage withPG_head are fake headstructpage. We need anapproach to distinguish between those two different types ofstructpage sothatcompound_head() can return the real headstructpage when theparameter is the tailstructpage but withPG_head.

Device DAX¶

The device-dax interface uses the same tail deduplication technique explainedin the previous chapter, except when used with the vmemmap inthe device (altmap).

The following page sizes are supported in DAX: PAGE_SIZE (4K on x86_64),PMD_SIZE (2M on x86_64) and PUD_SIZE (1G on x86_64).For powerpc equivalent details seeDevice DAX

The differences with HugeTLB are relatively minor.

It only use 3structpage for storing all information as opposedto 4 on HugeTLB pages.

There’s no remapping of vmemmap given that device-dax memory is not part ofSystem RAM ranges initialized at boot. Thus the tail page deduplicationhappens at a later stage when we populate the sections. HugeTLB reuses thethe head vmemmap page representing, whereas device-dax reuses the tailvmemmap page. This results in only half of the savings compared to HugeTLB.

Deduplicated tail pages are not mapped read-only.

Here’s how things look like on device-dax after the sections are populated:

+-----------+ ---virt_to_page---> +-----------+   mapping to   +-----------+|           |                     |     0     | -------------> |     0     ||           |                     +-----------+                +-----------+|           |                     |     1     | -------------> |     1     ||           |                     +-----------+                +-----------+|           |                     |     2     | ----------------^ ^ ^ ^ ^ ^|           |                     +-----------+                   | | | | ||           |                     |     3     | ------------------+ | | | ||           |                     +-----------+                     | | | ||           |                     |     4     | --------------------+ | | ||    PMD    |                     +-----------+                       | | ||   level   |                     |     5     | ----------------------+ | ||  mapping  |                     +-----------+                         | ||           |                     |     6     | ------------------------+ ||           |                     +-----------+                           ||           |                     |     7     | --------------------------+|           |                     +-----------+|           ||           ||           |+-----------+