1The Linux kernel supports the following overcommit handling modes
 2
 30	-	Heuristic overcommit handling. Obvious overcommits of
 4		address space are refused. Used for a typical system. It
 5		ensures a seriously wild allocation fails while allowing
 6		overcommit to reduce swap usage.  root is allowed to 
 7		allocate slightly more memory in this mode. This is the 
 8		default.
 9
101	-	Always overcommit. Appropriate for some scientific
11		applications. Classic example is code using sparse arrays
12		and just relying on the virtual memory consisting almost
13		entirely of zero pages.
14
152	-	Don't overcommit. The total address space commit
16		for the system is not permitted to exceed swap + a
17		configurable amount (default is 50%) of physical RAM.
18		Depending on the amount you use, in most situations
19		this means a process will not be killed while accessing
20		pages but will receive errors on memory allocation as
21		appropriate.
22
23		Useful for applications that want to guarantee their
24		memory allocations will be available in the future
25		without having to initialize every page.
26
27The overcommit policy is set via the sysctl `vm.overcommit_memory'.
28
29The overcommit amount can be set via `vm.overcommit_ratio' (percentage)
30or `vm.overcommit_kbytes' (absolute value).
31
32The current overcommit limit and amount committed are viewable in
33/proc/meminfo as CommitLimit and Committed_AS respectively.
34
35Gotchas
36-------
37
38The C language stack growth does an implicit mremap. If you want absolute
39guarantees and run close to the edge you MUST mmap your stack for the 
40largest size you think you will need. For typical stack usage this does
41not matter much but it's a corner case if you really really care
42
43In mode 2 the MAP_NORESERVE flag is ignored. 
44
45
46How It Works
47------------
48
49The overcommit is based on the following rules
50
51For a file backed map
52	SHARED or READ-only	-	0 cost (the file is the map not swap)
53	PRIVATE WRITABLE	-	size of mapping per instance
54
55For an anonymous or /dev/zero map
56	SHARED			-	size of mapping
57	PRIVATE READ-only	-	0 cost (but of little use)
58	PRIVATE WRITABLE	-	size of mapping per instance
59
60Additional accounting
61	Pages made writable copies by mmap
62	shmfs memory drawn from the same pool
63
64Status
65------
66
67o	We account mmap memory mappings
68o	We account mprotect changes in commit
69o	We account mremap changes in size
70o	We account brk
71o	We account munmap
72o	We report the commit status in /proc
73o	Account and check on fork
74o	Review stack handling/building on exec
75o	SHMfs accounting
76o	Implement actual limit enforcement
77
78To Do
79-----
80o	Account ptrace pages (this is hard)

 1The Linux kernel supports the following overcommit handling modes
 2
 30	-	Heuristic overcommit handling. Obvious overcommits of
 4		address space are refused. Used for a typical system. It
 5		ensures a seriously wild allocation fails while allowing
 6		overcommit to reduce swap usage.  root is allowed to 
 7		allocate slightly more memory in this mode. This is the 
 8		default.
 9
101	-	Always overcommit. Appropriate for some scientific
11		applications.
 
 
12
132	-	Don't overcommit. The total address space commit
14		for the system is not permitted to exceed swap + a
15		configurable percentage (default is 50) of physical RAM.
16		Depending on the percentage you use, in most situations
17		this means a process will not be killed while accessing
18		pages but will receive errors on memory allocation as
19		appropriate.
20
 
 
 
 
21The overcommit policy is set via the sysctl `vm.overcommit_memory'.
22
23The overcommit percentage is set via `vm.overcommit_ratio'.
 
24
25The current overcommit limit and amount committed are viewable in
26/proc/meminfo as CommitLimit and Committed_AS respectively.
27
28Gotchas
29-------
30
31The C language stack growth does an implicit mremap. If you want absolute
32guarantees and run close to the edge you MUST mmap your stack for the 
33largest size you think you will need. For typical stack usage this does
34not matter much but it's a corner case if you really really care
35
36In mode 2 the MAP_NORESERVE flag is ignored. 
37
38
39How It Works
40------------
41
42The overcommit is based on the following rules
43
44For a file backed map
45	SHARED or READ-only	-	0 cost (the file is the map not swap)
46	PRIVATE WRITABLE	-	size of mapping per instance
47
48For an anonymous or /dev/zero map
49	SHARED			-	size of mapping
50	PRIVATE READ-only	-	0 cost (but of little use)
51	PRIVATE WRITABLE	-	size of mapping per instance
52
53Additional accounting
54	Pages made writable copies by mmap
55	shmfs memory drawn from the same pool
56
57Status
58------
59
60o	We account mmap memory mappings
61o	We account mprotect changes in commit
62o	We account mremap changes in size
63o	We account brk
64o	We account munmap
65o	We report the commit status in /proc
66o	Account and check on fork
67o	Review stack handling/building on exec
68o	SHMfs accounting
69o	Implement actual limit enforcement
70
71To Do
72-----
73o	Account ptrace pages (this is hard)