doc: fix incorrect formula to calculate CommitLimit value
authorPetr Oros <poros@redhat.com>
Thu, 22 May 2014 12:04:44 +0000 (14:04 +0200)
committerJiri Kosina <jkosina@suse.cz>
Thu, 22 May 2014 21:01:55 +0000 (23:01 +0200)
The formula to calculate "CommitLimit" value mentioned in kernel documentation is incorrect.
Right formula is: CommitLimit = ([total RAM pages] - [total huge TLB pages]) * overcommit_ratio / 100 + [total swap pages]

Signed-off-by: Petr Oros <poros@redhat.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Documentation/filesystems/proc.txt

index d4fa7f52034ee17b54dae57017fcc09b3cb2987a..324328d1846421874a2e1cc071f6ec7499c0ee82 100644 (file)
@@ -854,7 +854,8 @@ WritebackTmp: Memory used by FUSE for temporary writeback buffers
               if strict overcommit accounting is enabled (mode 2 in
               'vm.overcommit_memory').
               The CommitLimit is calculated with the following formula:
-              CommitLimit = ('vm.overcommit_ratio' * Physical RAM) + Swap
+              CommitLimit = ([total RAM pages] - [total huge TLB pages]) *
+                             overcommit_ratio / 100 + [total swap pages]
               For example, on a system with 1G of physical RAM and 7G
               of swap with a `vm.overcommit_ratio` of 30 it would
               yield a CommitLimit of 7.3G.