aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorYing Han <yinghan@google.com>2010-10-26 17:21:23 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-26 19:52:05 -0400
commit3d5992d2ac7dc09aed8ab537cba074589f0f0a52 (patch)
tree0fbb51026e8aa13ddd7a6516c04471ceb97bc277 /fs
parent0f4d208f1975f16f269134cee5f44c1f048581da (diff)
oom: add per-mm oom disable count
It's pointless to kill a task if another thread sharing its mm cannot be killed to allow future memory freeing. A subsequent patch will prevent kills in such cases, but first it's necessary to have a way to flag a task that shares memory with an OOM_DISABLE task that doesn't incur an additional tasklist scan, which would make select_bad_process() an O(n^2) function. This patch adds an atomic counter to struct mm_struct that follows how many threads attached to it have an oom_score_adj of OOM_SCORE_ADJ_MIN. They cannot be killed by the kernel, so their memory cannot be freed in oom conditions. This only requires task_lock() on the task that we're operating on, it does not require mm->mmap_sem since task_lock() pins the mm and the operation is atomic. [rientjes@google.com: changelog and sys_unshare() code] [rientjes@google.com: protect oom_disable_count with task_lock in fork] [rientjes@google.com: use old_mm for oom_disable_count in exec] Signed-off-by: Ying Han <yinghan@google.com> Signed-off-by: David Rientjes <rientjes@google.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'fs')
-rw-r--r--fs/exec.c5
-rw-r--r--fs/proc/base.c30
2 files changed, 35 insertions, 0 deletions
diff --git a/fs/exec.c b/fs/exec.c
index 6d2b6f936858..3aa75b8888a1 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -54,6 +54,7 @@
54#include <linux/fsnotify.h> 54#include <linux/fsnotify.h>
55#include <linux/fs_struct.h> 55#include <linux/fs_struct.h>
56#include <linux/pipe_fs_i.h> 56#include <linux/pipe_fs_i.h>
57#include <linux/oom.h>
57 58
58#include <asm/uaccess.h> 59#include <asm/uaccess.h>
59#include <asm/mmu_context.h> 60#include <asm/mmu_context.h>
@@ -759,6 +760,10 @@ static int exec_mmap(struct mm_struct *mm)
759 tsk->mm = mm; 760 tsk->mm = mm;
760 tsk->active_mm = mm; 761 tsk->active_mm = mm;
761 activate_mm(active_mm, mm); 762 activate_mm(active_mm, mm);
763 if (old_mm && tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) {
764 atomic_dec(&old_mm->oom_disable_count);
765 atomic_inc(&tsk->mm->oom_disable_count);
766 }
762 task_unlock(tsk); 767 task_unlock(tsk);
763 arch_pick_mmap_layout(mm); 768 arch_pick_mmap_layout(mm);
764 if (old_mm) { 769 if (old_mm) {
diff --git a/fs/proc/base.c b/fs/proc/base.c
index dc5d5f51f3fe..6e50c8e65513 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1047,6 +1047,21 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1047 return -EACCES; 1047 return -EACCES;
1048 } 1048 }
1049 1049
1050 task_lock(task);
1051 if (!task->mm) {
1052 task_unlock(task);
1053 unlock_task_sighand(task, &flags);
1054 put_task_struct(task);
1055 return -EINVAL;
1056 }
1057
1058 if (oom_adjust != task->signal->oom_adj) {
1059 if (oom_adjust == OOM_DISABLE)
1060 atomic_inc(&task->mm->oom_disable_count);
1061 if (task->signal->oom_adj == OOM_DISABLE)
1062 atomic_dec(&task->mm->oom_disable_count);
1063 }
1064
1050 /* 1065 /*
1051 * Warn that /proc/pid/oom_adj is deprecated, see 1066 * Warn that /proc/pid/oom_adj is deprecated, see
1052 * Documentation/feature-removal-schedule.txt. 1067 * Documentation/feature-removal-schedule.txt.
@@ -1065,6 +1080,7 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1065 else 1080 else
1066 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) / 1081 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1067 -OOM_DISABLE; 1082 -OOM_DISABLE;
1083 task_unlock(task);
1068 unlock_task_sighand(task, &flags); 1084 unlock_task_sighand(task, &flags);
1069 put_task_struct(task); 1085 put_task_struct(task);
1070 1086
@@ -1133,6 +1149,19 @@ static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1133 return -EACCES; 1149 return -EACCES;
1134 } 1150 }
1135 1151
1152 task_lock(task);
1153 if (!task->mm) {
1154 task_unlock(task);
1155 unlock_task_sighand(task, &flags);
1156 put_task_struct(task);
1157 return -EINVAL;
1158 }
1159 if (oom_score_adj != task->signal->oom_score_adj) {
1160 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1161 atomic_inc(&task->mm->oom_disable_count);
1162 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1163 atomic_dec(&task->mm->oom_disable_count);
1164 }
1136 task->signal->oom_score_adj = oom_score_adj; 1165 task->signal->oom_score_adj = oom_score_adj;
1137 /* 1166 /*
1138 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is 1167 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
@@ -1143,6 +1172,7 @@ static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1143 else 1172 else
1144 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) / 1173 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1145 OOM_SCORE_ADJ_MAX; 1174 OOM_SCORE_ADJ_MAX;
1175 task_unlock(task);
1146 unlock_task_sighand(task, &flags); 1176 unlock_task_sighand(task, &flags);
1147 put_task_struct(task); 1177 put_task_struct(task);
1148 return count; 1178 return count;