Linux Kernel Architecture

Chapter 4: Virtual Process Memory

4.5 Operations on Regions

The kernel provides various functions to manipulate the regions of a process. Creating and deleting

regions (and finding a suitable memory location for a new region) are standard operations needed when

setting up or removing a mapping. The kernel is also responsible for performing optimizations when

managing the data structures, as shown in Figure 4-10.

Existing region New region Region is being deleted

Figure 4-10: Operations on regions.

❑ When a new region is added immediately before orafter an existing region (and therefore also

between two existing regions), the kernel merges the data structures involved into a single

structure — but, of course,onlyif the access permissions for all the regions involved are identical

and contiguous data are mapped from the same backing store.

❑ If a deletion is made at the start or end of a region, the existing data structure must be truncated

accordingly.

❑ If a regionbetweentwo other regions is deleted, the existing data structure is reduced in size, and

a new data structure is created for the resultant new region.

A further important standard operation is the search for a region associated with a specific virtual address

in userspace. Before explaining the optimizations mentioned above, let’s discuss the helper function used

to do this.

4.5.1 Associating Virtual Addresses with a Region

By reference to a virtual address,find_vmafinds the first region in user address space whose end isafter

the given address and therefore satisfies theaddr < vm_area_struct->vm_endcondition. As parameters,

the function requires not only the virtual address (addr) but also a pointer to themm_structinstance of

the relevant process whose address space is to be scanned.

<mm/mmap.c>

struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr)

{

struct vm_area_struct *vma = NULL;

if (mm) {

/* Check the cache first. */