Linux Memory Management

Today's Lecture

• Q&A
• Kernel Hints
• Memory Management

Q&A

• How is project 3?
  • Too easy?
  • Too hard?
  • Too little time?
• Any questions

Debugging UML with gdb

• Compile UML with gdb support
  • Configure UML with kernel symbols and gdb support
    • cp /projects/cs378.ygz/src/uml.config-2.4.9-debug .config
  • Compile as usual
• Run UML with debug option
  • ./linux ubd0=../root_fs_utcs debug
• An xterm will pop up with gdb running
  • Initially stopped at init/main.c:start_kernel()
  • Treat it like any other gdb session

More about gdb'ing UML

• Be careful with multi-thread/multi-process debugging
• An example gdb session with UML:
  • http://user-mode-linux.sourceforge.net/debug-session.html
• More hints on using gdb with UML:
  • http://user-mode-linux.sourceforge.net/debugging.html
  • Debugging modules
  • Using debuggers other than gdb

Linux Memory Management

• Demand Paged Virtual Memory
• 3-Level Page Table: pgd_t, pmd_t, pte_t
  • 10-bit address space for each table
• i386 Architecture
  • Page size: 4KB (12-bit)
  • Total addressable space: 4GB (32-bit):
  • User space: 3GB (0x00000000-0xbfffffff)
  • Kernel space: 1GB (0xc0000000-0xffffffff)

Page Frame Data Structure

• Data type (in include/linux/mm.h)

  struct page {
  	...
  	struct list_head list;	/* linked list of page frame */
   	atomic_t count;	/* usage count */
  	struct list_head lru;	/* LRU */
  	unsigned long age;	/* page age counter */
  	struct zone_struct *zone;	/* memory zone */
  	...
  } mem_map_t;
  

• Linked list of all pages (In mm/memory.c):
  • mem_map_t * mem_map;

Memory Zone

• PC specific: 3 Zones
  • DMA (<16M), Normal, HIGHMEM (>896M)
• Each zone maintains its own free areas
  • In include/linux/mmzone.h:

    typedef struct zone_struct {
    	...
    	free_area_t             free_area[MAX_ORDER];
    	...
    } zone_t;
        

Page Allocation

• The Buddy System Algorithm
  • Pages are allocated in blocks which are powers of 2 in size (1, 2, 4, 8, 16, 32, 64, 128, 256, 512)
• Each size with its own free list (called free area)
  • In include/linux/mmzone.h:

  #define MAX_ORDER 10 
  typedef struct free_area_struct {
          struct list_head        free_list;
          unsigned long           *map;
  } free_area_t;
  ...
  free_area_t             free_area[MAX_ORDER];
  

Page Management Summary

• Linked list of all pages
• Each page belongs to a zone
• Each zone maintains 10 free lists
• Allocation is based on size (powers of 2)
• De-allocation
  • If the adjacent buddy block is also free, combine to form a new free block for the next size block of pages

Memory Management

• For small memory use
• For kernel data objects that need to be frequently allocated and released
  • Examples: file descriptors, sockets, inodes, ...
• Slab Allocator
  • Group memory area (called cache) by kernel data object type
  • For each type, maintain memory cache (free list) for objects that are previously allocated then released
  • Interface with page frame allocator
  • Chapter 6 in ULK book

General Purpose Cache

• For general purpose object
• 13 general purpose cache in slab allocator
  • Power of 2: 32, 64, ......, 131072
• To allocate memory in this category:
  • In include/linux/slab.h
  • void *kmalloc(size_t size, init flags)
• To free memory
  • void kfree(const void *objp)

Noncontiguous Memory

• Noncontiguous page frames in contiguous linear address
  • Not all virtual memory maps to the contiguous page frames
  • Make sense for infrequent use
• To allocate (in include/linux/vmalloc.h)
  • void *vmalloc(unsigned long size)
• To release