diff options
author | Hugh Dickins <hugh@veritas.com> | 2008-07-23 21:28:23 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2008-07-24 10:47:21 -0700 |
commit | 78ecba081224a2db5876b6b81cfed0b78f58adc7 (patch) | |
tree | db29cc9357ea95172f42e64e833a3a7d3fcfa8fb /include/linux | |
parent | 83d1674a946141c3c59d430e96c224f7937e6158 (diff) | |
download | op-kernel-dev-78ecba081224a2db5876b6b81cfed0b78f58adc7.zip op-kernel-dev-78ecba081224a2db5876b6b81cfed0b78f58adc7.tar.gz |
mm: fix ever-decreasing swap priority
Vegard Nossum has noticed the ever-decreasing negative priority in a
swapon /swapoff loop, which eventually would misprioritize when int wraps
positive. Not worth spending much code on, but probably better fixed.
It's easy to handle the swapping on and off of just one area, but there's
not much point if a pair or more still misbehave. To handle the general
case, swapoff should compact negative priorities, keeping them always from
-1 to -MAX_SWAPFILES. That's a change, but should cause no regression,
since these negative (unspecified) priorities are disjoint from the the
positive specified priorities 0 to 32767.
One small functional difference, which seems appropriate: when swapoff
fails to free all swap from a negative priority area, that area is now
reinserted at lowest priority, rather than at its original priority.
In moving down swapon's setting of priority, I notice that an area is
visible to /proc/swaps when it has swap_map set, yet that was being set
before all the visible fields were properly filled in: corrected.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reported-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'include/linux')
0 files changed, 0 insertions, 0 deletions