which is the page_directory for init.) After 3 GB, a Process has the Kernel Page

Tables entries.Infact it has the same entries as that of the init process, after

3 GB. The KERNEL_PGD_PTRS entries are copied when a Process is created, from

swapper_pg_dir,by doing a memcpy(task->mm->pgd + USER_PGD_PTRS,

swapper_pg_dir + USER_PGD_PTRS , sizeof(pgd_t) * KERNEL_PGD_PTRS); Hence the

last 256 entries (The mapping for 3 GB and up) is copied directly to the process

page tables. The links to the next vm_area_struct are reflected by vm_next

entry, and the links to the inode address_spaces shared and private mapping, are

reflected by vm_next_share and vm_pprev_share entries in the vm_area_struct.The

root of the AVL tree,is represented by mmap_avl, in the mm_struct of a process,

and the head of the vm_area_structs is the mmap field in the mm_struct,and there

is also a mmap_cache for the last vma struct found. AVL trees are used to speed

up vm_area_struct searches by having the vm_area_structs represented by a

Tree,which starts builing when the mmap count reaches AVL_MIN_MAP_COUNT.You can

check out the avl_tree code in mm/mmap_avl.c, which gets included in mm/mmap.c.

Each vm area also has pointers to functions that represent the

page_faults,no_page handlers,etc. for that area, in vm_operations_struct. If the

process has a file mapped to that area,the file can be traced by vm_file field.

Across a fork,not only does the vm_area gets copied (check fork.c ),but any

inode shares,that the parent has mapped, gets linked to the child thread,and the

area is opened.(vm_ops->open ) In order to get a first hand glimpse of

vm_area management,you should have a look at mmap.c, which has important

definitions for do_mmap_pgoff, do_unmap,do_brk,insert_vm_struct, which inserts a

vm struct, sorted by address,and also builds the mmap_avl list,on reaching

AVL_MIN_MAP_COUNT,by calling build_mmap_avl.There are also other important

routines which you need to take up.(like find_vma,find_vma_prev,vmtruncate_list

in memory.c,etc.etc). The do_mmap_pgoff gets called on a mmap syscall. A mmap

system call is used to map Physical pages to the tasks virtual address

space.Supposing that a task wants to mmap LEN bytes from offset OFFSET (Page

aligned), from a File to its virtual address space,it uses a mmap system call,to

map them to its address space. The task of do_mmap in a nutshell,apart from