leandrof/linux
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

docs/vm: overcommit-accounting: convert to ReST format

Browse Source 
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
This commit is contained in:
Mike Rapoport
2018-03-21 21:22:30 +02:00
committed by Jonathan Corbet
parent cb5e4376e5
commit 8d83d826c2
1 changed files with 56 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

105
Documentation/vm/overcommit-accounting
View File

@@ -1,80 +1,87 @@
.. _overcommit_accounting:
=====================
Overcommit Accounting
=====================
The Linux kernel supports the following overcommit handling modes The Linux kernel supports the following overcommit handling modes
0 - Heuristic overcommit handling. Obvious overcommits of 0
address space are refused. Used for a typical system. It Heuristic overcommit handling. Obvious overcommits of address
ensures a seriously wild allocation fails while allowing space are refused. Used for a typical system. It ensures a
overcommit to reduce swap usage. root is allowed to seriously wild allocation fails while allowing overcommit to
allocate slightly more memory in this mode. This is the reduce swap usage. root is allowed to allocate slightly more
default. memory in this mode. This is the default.
1 - Always overcommit. Appropriate for some scientific 1
applications. Classic example is code using sparse arrays Always overcommit. Appropriate for some scientific
and just relying on the virtual memory consisting almost applications. Classic example is code using sparse arrays and
entirely of zero pages. just relying on the virtual memory consisting almost entirely
of zero pages.
2 - Don't overcommit. The total address space commit 2
for the system is not permitted to exceed swap + a Don't overcommit. The total address space commit for the
configurable amount (default is 50%) of physical RAM. system is not permitted to exceed swap + a configurable amount
Depending on the amount you use, in most situations (default is 50%) of physical RAM. Depending on the amount you
this means a process will not be killed while accessing use, in most situations this means a process will not be
pages but will receive errors on memory allocation as killed while accessing pages but will receive errors on memory
appropriate. allocation as appropriate.
Useful for applications that want to guarantee their Useful for applications that want to guarantee their memory
memory allocations will be available in the future allocations will be available in the future without having to
without having to initialize every page. initialize every page.
The overcommit policy is set via the sysctl `vm.overcommit_memory'. The overcommit policy is set via the sysctl ``vm.overcommit_memory``.
The overcommit amount can be set via `vm.overcommit_ratio' (percentage) The overcommit amount can be set via ``vm.overcommit_ratio`` (percentage)
or `vm.overcommit_kbytes' (absolute value). or ``vm.overcommit_kbytes`` (absolute value).
The current overcommit limit and amount committed are viewable in The current overcommit limit and amount committed are viewable in
/proc/meminfo as CommitLimit and Committed_AS respectively. ``/proc/meminfo`` as CommitLimit and Committed_AS respectively.
Gotchas Gotchas
------- =======
The C language stack growth does an implicit mremap. If you want absolute The C language stack growth does an implicit mremap. If you want absolute
guarantees and run close to the edge you MUST mmap your stack for the guarantees and run close to the edge you MUST mmap your stack for the
largest size you think you will need. For typical stack usage this does largest size you think you will need. For typical stack usage this does
not matter much but it's a corner case if you really really care not matter much but it's a corner case if you really really care
In mode 2 the MAP_NORESERVE flag is ignored. In mode 2 the MAP_NORESERVE flag is ignored.
How It Works How It Works
------------ ============
The overcommit is based on the following rules The overcommit is based on the following rules
For a file backed map For a file backed map
SHARED or READ-only - 0 cost (the file is the map not swap) | SHARED or READ-only - 0 cost (the file is the map not swap)
PRIVATE WRITABLE - size of mapping per instance | PRIVATE WRITABLE - size of mapping per instance
For an anonymous or /dev/zero map For an anonymous or ``/dev/zero`` map
SHARED - size of mapping | SHARED - size of mapping
PRIVATE READ-only - 0 cost (but of little use) | PRIVATE READ-only - 0 cost (but of little use)
PRIVATE WRITABLE - size of mapping per instance | PRIVATE WRITABLE - size of mapping per instance
Additional accounting Additional accounting
Pages made writable copies by mmap | Pages made writable copies by mmap
shmfs memory drawn from the same pool | shmfs memory drawn from the same pool
Status Status
------ ======
o We account mmap memory mappings * We account mmap memory mappings
o We account mprotect changes in commit * We account mprotect changes in commit
o We account mremap changes in size * We account mremap changes in size
o We account brk * We account brk
o We account munmap * We account munmap
o We report the commit status in /proc * We report the commit status in /proc
o Account and check on fork * Account and check on fork
o Review stack handling/building on exec * Review stack handling/building on exec
o SHMfs accounting * SHMfs accounting
o Implement actual limit enforcement * Implement actual limit enforcement
To Do To Do
----- =====
o Account ptrace pages (this is hard) * Account ptrace pages (this is hard)