Skip to contentSkip to conversation

oom-killer disable for iscsi/lvm2/multipath userland critical sections

PATCH Andrea Arcangeli 2 participants 2 messages Apr 1, 2005 – Apr 2, 2005

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

Hello,

some private discussion (that was continuing some kernel-summit-discuss thread) ended in the below patch. I also liked a textual "disable" instead of value "-17" (internally to the kernel it could be represented the same way, but the /proc parsing would be more complicated). If you prefer textual "disable" we can change this of course.

Comments welcome.

From: Andrea Arcangeli <[email protected]>
Subject: oom killer protection

iscsi/lvm2/multipath needs guaranteed protection from the oom-killer.

fs/proc/base.c renamed from fs/proc/base.c.~1~ +1 -1
@@ -751,7 +751,7 @@ static ssize_t oom_adjust_write(struct f
 	if (copy_from_user(buffer, buf, count))
 		return -EFAULT;
 	oom_adjust = simple_strtol(buffer, &end, 0);
-	if (oom_adjust < -16 || oom_adjust > 15)
+	if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
 		return -EINVAL;
 	if (*end == '\n')
 		end++;
include/linux/mm.h renamed from include/linux/mm.h.~1~ +3 -0
@@ -856,5 +856,8 @@ int in_gate_area_no_task(unsigned long a
 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
 #endif	/* __HAVE_ARCH_GATE_AREA */
 
+/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
+#define OOM_DISABLE -17
+
 #endif /* __KERNEL__ */
 #endif /* _LINUX_MM_H */
mm/oom_kill.c renamed from mm/oom_kill.c.~1~ +1 -1
@@ -145,7 +145,7 @@ static struct task_struct * select_bad_p
 	do_posix_clock_monotonic_gettime(&uptime);
 	do_each_thread(g, p)
 		/* skip the init task with pid == 1 */
-		if (p->pid > 1) {
+		if (p->pid > 1 && p->oomkilladj != OOM_DISABLE) {
 			unsigned long points;
 
 			/*
#1 lore raw

Andrea,

I just successfully tested the patch on my environment. It actually resolved OOM-killer problem for my iscsid.

Important note: daemon's parent must be init.

In my test, OOM-killer killed everything around but iscsid, and iscsid successfully finished registration of new SCSI host in the middle of crazy OOM-killer :)

Thanks!
Dima

Quoted 55 lines · Andrea Arcangeli

Hello,

some private discussion (that was continuing some kernel-summit-discuss thread) ended in the below patch. I also liked a textual "disable" instead of value "-17" (internally to the kernel it could be represented the same way, but the /proc parsing would be more complicated). If you prefer textual "disable" we can change this of course.

Comments welcome.

From: Andrea Arcangeli <[email protected]>
Subject: oom killer protection

iscsi/lvm2/multipath needs guaranteed protection from the oom-killer.

fs/proc/base.c renamed from fs/proc/base.c.~1~ +1 -1
@@ -751,7 +751,7 @@ static ssize_t oom_adjust_write(struct f
 	if (copy_from_user(buffer, buf, count))
 		return -EFAULT;
 	oom_adjust = simple_strtol(buffer, &end, 0);
-	if (oom_adjust < -16 || oom_adjust > 15)
+	if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
 		return -EINVAL;
 	if (*end == '\n')
 		end++;
include/linux/mm.h renamed from include/linux/mm.h.~1~ +3 -0
@@ -856,5 +856,8 @@ int in_gate_area_no_task(unsigned long a
 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
 #endif	/* __HAVE_ARCH_GATE_AREA */
 
+/* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
+#define OOM_DISABLE -17
+
 #endif /* __KERNEL__ */
 #endif /* _LINUX_MM_H */
mm/oom_kill.c renamed from mm/oom_kill.c.~1~ +1 -2
@@ -145,7 +145,7 @@ static struct task_struct * select_bad_p
 	do_posix_clock_monotonic_gettime(&uptime);
 	do_each_thread(g, p)
 		/* skip the init task with pid == 1 */
-		if (p->pid > 1) {
+		if (p->pid > 1 && p->oomkilladj != OOM_DISABLE) {
 			unsigned long points;
 
 			/*
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/