Pages

The kernel treats physical pages as the basic unit of memory management. Although the processor's smallest addressable unit is usually a word (or even a byte), the memory management unit (MMU, the hardware that manages memory and performs virtual to physical address translations) typically deals in pages. Therefore, the MMU manages the system's page tables with page-sized granularity (hence their name). In terms of virtual memory, pages are the smallest unit that matters.

As you'll see in Chapter 19, "Portability," each architecture enforces its own page size. Many architectures even support multiple page sizes. Most 32-bit architectures have 4KB pages, whereas most 64-bit architectures have 8KB pages. This implies that on a machine with 4KB pages and 1GB of physical memory, physical memory is divided up into 262,144 distinct pages.

The kernel represents every physical page on the system with a struct page structure. This structure is defined in <linux/mm.h>:

struct page {
        page_flags_t          flags;
        atomic_t              _count;
        atomic_t              _mapcount;
        unsigned long         private;
        struct address_space  *mapping;
        pgoff_t               index;
        struct list_head      lru;
        void                  *virtual;
};

Let's look at the important fields. The flags field stores the status of the page. Such flags include whether the page is dirty or whether it is locked in memory. Bit flags represent the various values, so at least 32 different flags are simultaneously available. The flag values are defined in <linux/page-flags.h>.

The _count field stores the usage count of the pagethat is, how many references there are to this page. When this count reaches negative zero, no one is using the page, and it becomes available for use in a new allocation. Kernel code should not check this field directly but instead use the function page_count(), which takes a page structure as its sole parameter. Although internally _count is negative one when the page is free, page_count() returns zero to indicate free and a positive nonzero integer when the page is in use. A page may be used by the page cache (in which case the mapping field points to the address_space object that is associated with this page), as private data (pointed at by private), or as a mapping in a process's page table.

The virtual field is the page's virtual address. Normally, this is simply the address of the page in virtual memory. Some memory (called high memory) is not permanently mapped in the kernel's address space. In that case, this field is NULL and the page must be dynamically mapped when needed. We'll discuss high memory shortly.

The important point to understand is that the page structure is associated with physical pages, not virtual pages. Therefore, what the structure describes is transient at best. Even if the data contained in the page continues to exist, it might not always be associated with the same page structure because of swapping and so on. The kernel uses this data structure to describe whatever is stored in the associated physical page at that moment. The data structure's goal is to describe physical memory, not the data contained therein.

The kernel uses this structure to keep track of all the pages in the system, because the kernel needs to know whether a page is free (that is, if the page is not allocated). If a page is not free, the kernel needs to know who owns the page. Possible owners include user-space processes, dynamically allocated kernel data, static kernel code, the page cache, and so on.

Developers are often surprised that an instance of this structure is allocated for each physical page in the system. They think, "What a lot of memory used!" Let's look at just how bad (or good) the space consumption is from all these pages. Assume struct page consumes 40 bytes of memory, the system has 4KB physical pages, and the system has 128MB of physical memory. In that case, all the page structures in the system consume slightly more than 1MB of memorynot too high a cost for managing all of the sys tem's physical pages.