High Memory Handling

By: Peter Zijlstra <a.p.zijlstra@chello.nl>

What Is High Memory?

High memory (highmem) is used when the size of physical memory approaches orexceeds the maximum size of virtual memory. At that point it becomesimpossible for the kernel to keep all of the available physical memory mappedat all times. This means the kernel needs to start using temporary mappings ofthe pieces of physical memory that it wants to access.

The part of (physical) memory not covered by a permanent mapping is what werefer to as ‘highmem’. There are various architecture dependent constraints onwhere exactly that border lies.

In the i386 arch, for example, we choose to map the kernel into every process’sVM space so that we don’t have to pay the full TLB invalidation costs forkernel entry/exit. This means the available virtual memory space (4GiB oni386) has to be divided between user and kernel space.

The traditional split for architectures using this approach is 3:1, 3GiB foruserspace and the top 1GiB for kernel space:

+--------+ 0xffffffff| Kernel |+--------+ 0xc0000000|        || User   ||        |+--------+ 0x00000000

This means that the kernel can at most map 1GiB of physical memory at any onetime, but because we need virtual address space for other things - includingtemporary maps to access the rest of the physical memory - the actual directmap will typically be less (usually around ~896MiB).

Other architectures that have mm context tagged TLBs can have separate kerneland user maps. Some hardware (like some ARMs), however, have limited virtualspace when they use mm context tags.

Temporary Virtual Mappings

The kernel contains several ways of creating temporary mappings:

  • vmap(). This can be used to make a long duration mapping of multiplephysical pages into a contiguous virtual space. It needs globalsynchronization to unmap.

  • kmap(). This permits a short duration mapping of a single page. It needsglobal synchronization, but is amortized somewhat. It is also prone todeadlocks when using in a nested fashion, and so it is not recommended fornew code.

  • kmap_atomic(). This permits a very short duration mapping of a singlepage. Since the mapping is restricted to the CPU that issued it, itperforms well, but the issuing task is therefore required to stay on thatCPU until it has finished, lest some other task displace its mappings.

    kmap_atomic() may also be used by interrupt contexts, since it is does notsleep and the caller may not sleep until after kunmap_atomic() is called.

    It may be assumed that k[un]map_atomic() won’t fail.

Using kmap_atomic

When and where to use kmap_atomic() is straightforward. It is used when codewants to access the contents of a page that might be allocated from high memory(see __GFP_HIGHMEM), for example a page in the pagecache. The API has twofunctions, and they can be used in a manner similar to the following:

/* Find the page of interest. */struct page *page = find_get_page(mapping, offset);/* Gain access to the contents of that page. */void *vaddr = kmap_atomic(page);/* Do something to the contents of that page. */memset(vaddr, 0, PAGE_SIZE);/* Unmap that page. */kunmap_atomic(vaddr);

Note that the kunmap_atomic() call takes the result of the kmap_atomic() callnot the argument.

If you need to map two pages because you want to copy from one page toanother you need to keep the kmap_atomic calls strictly nested, like:

vaddr1 = kmap_atomic(page1);vaddr2 = kmap_atomic(page2);memcpy(vaddr1, vaddr2, PAGE_SIZE);kunmap_atomic(vaddr2);kunmap_atomic(vaddr1);

Cost of Temporary Mappings

The cost of creating temporary mappings can be quite high. The arch has tomanipulate the kernel’s page tables, the data TLB and/or the MMU’s registers.

If CONFIG_HIGHMEM is not set, then the kernel will try and create a mappingsimply with a bit of arithmetic that will convert the page struct address intoa pointer to the page contents rather than juggling mappings about. In such acase, the unmap operation may be a null operation.

If CONFIG_MMU is not set, then there can be no temporary mappings and nohighmem. In such a case, the arithmetic approach will also be used.

i386 PAE

The i386 arch, under some circumstances, will permit you to stick up to 64GiBof RAM into your 32-bit machine. This has a number of consequences:

  • Linux needs a page-frame structure for each page in the system and thepageframes need to live in the permanent mapping, which means:
  • you can have 896M/sizeof(struct page) page-frames at most; with structpage being 32-bytes that would end up being something in the order of 112Gworth of pages; the kernel, however, needs to store more than justpage-frames in that memory…
  • PAE makes your page tables larger - which slows the system down as moredata has to be accessed to traverse in TLB fills and the like. Oneadvantage is that PAE has more PTE bits and can provide advanced featureslike NX and PAT.

The general recommendation is that you don’t use more than 8GiB on a 32-bitmachine - although more might work for you and your workload, you’re prettymuch on your own - don’t expect kernel developers to really care much if thingscome apart.