mm: oom_kill: reap memory of a task that receives SIGKILL

Free the pages parallely for a task that receives SIGKILL using the
oom_reaper. This freeing of pages will help to give the pages to buddy
system well advance.
This reaps for the process which received SIGKILL through
either sys_kill from user or kill_pid from kernel and that sending
process has CAP_KILL capability.
Also sysctl interface, reap_mem_on_sigkill, is added to turn on/off this
feature.

Change-Id: I21adb95de5e380a80d7eb0b87d9b5b553f52e28a
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
[swatsrid@codeaurora.org: Fix merge conflicts]
Signed-off-by: Swathi Sridhar <swatsrid@codeaurora.org>
[isaacm@codeaurora.org: Fix merge conflicts]
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>

Documentation/admin-guide/sysctl/vm.rst

@@ -57,6 +57,7 @@ Currently, these files are in /proc/sys/vm:
 - nr_trim_pages         (only if CONFIG_MMU=n)
 - numa_zonelist_order
 - oom_dump_tasks
+- reap_mem_on_sigkill
 - oom_kill_allocating_task
 - overcommit_kbytes
 - overcommit_memory
@@ -669,6 +670,22 @@ OOM killer actually kills a memory-hogging task.
 
 The default value is 1 (enabled).
 
+reap_mem_on_sigkill
+===================
+
+This enables or disables the memory reaping for a SIGKILL received
+process and that the sending process must have the CAP_KILL capabilities.
+
+If this is set to 1, when a process receives SIGKILL from a process
+that has the capability, CAP_KILL, the process is added into the oom_reaper
+queue which can be picked up by the oom_reaper thread to reap the memory of
+that process. This reaps for the process which received SIGKILL through
+either sys_kill from user or kill_pid from kernel.
+
+If this is set to 0, we are not reaping memory of a SIGKILL, sent through
+either sys_kill from user or kill_pid from kernel, received process.
+
+The default value is 0 (disabled).
 
 oom_kill_allocating_task
 ========================

include/linux/oom.h

@@ -126,4 +126,8 @@ extern struct task_struct *find_lock_task_mm(struct task_struct *p);
 extern int sysctl_oom_dump_tasks;
 extern int sysctl_oom_kill_allocating_task;
 extern int sysctl_panic_on_oom;
+extern int sysctl_reap_mem_on_sigkill;
+
+/* calls for LMK reaper */
+extern void add_to_oom_reaper(struct task_struct *p);
 #endif /* _INCLUDE_LINUX_OOM_H */

kernel/signal.c

@@ -46,6 +46,7 @@
 #include <linux/livepatch.h>
 #include <linux/cgroup.h>
 #include <linux/audit.h>
+#include <linux/oom.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/signal.h>
@@ -1404,8 +1405,11 @@ int group_send_sig_info(int sig, struct kernel_siginfo *info,
 	ret = check_kill_permission(sig, info, p);
 	rcu_read_unlock();
 
-	if (!ret && sig)
+	if (!ret && sig) {
 		ret = do_send_sig_info(sig, info, p, type);
+		if (capable(CAP_KILL) && sig == SIGKILL)
+			add_to_oom_reaper(p);
+	}
 
 	return ret;
 }

kernel/sysctl.c

@@ -1327,6 +1327,13 @@ static struct ctl_table vm_table[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+	{
+		.procname       = "reap_mem_on_sigkill",
+		.data           = &sysctl_reap_mem_on_sigkill,
+		.maxlen         = sizeof(sysctl_reap_mem_on_sigkill),
+		.mode           = 0644,
+		.proc_handler   = proc_dointvec,
+	},
 	{
 		.procname	= "overcommit_ratio",
 		.data		= &sysctl_overcommit_ratio,

mm/oom_kill.c

@@ -54,6 +54,7 @@
 int sysctl_panic_on_oom;
 int sysctl_oom_kill_allocating_task;
 int sysctl_oom_dump_tasks = 1;
+int sysctl_reap_mem_on_sigkill;
 
 /*
  * Serializes oom killer invocations (out_of_memory()) from all contexts to
@@ -661,13 +662,22 @@ static int oom_reaper(void *unused)
 
 static void wake_oom_reaper(struct task_struct *tsk)
 {
+	/*
+	 * Move the lock here to avoid scenario of queuing
+	 * the same task by both OOM killer and any other SIGKILL
+	 * path.
+	 */
+	spin_lock(&oom_reaper_lock);
+
 	/* mm is already queued? */
-	if (test_and_set_bit(MMF_OOM_REAP_QUEUED, &tsk->signal->oom_mm->flags))
+	if (test_and_set_bit(MMF_OOM_REAP_QUEUED,
+			     &tsk->signal->oom_mm->flags)) {
+		spin_unlock(&oom_reaper_lock);
 		return;
+	}
 
 	get_task_struct(tsk);
 
-	spin_lock(&oom_reaper_lock);
 	tsk->oom_reaper_list = oom_reaper_list;
 	oom_reaper_list = tsk;
 	spin_unlock(&oom_reaper_lock);
@@ -687,6 +697,16 @@ static inline void wake_oom_reaper(struct task_struct *tsk)
 }
 #endif /* CONFIG_MMU */
 
+static void __mark_oom_victim(struct task_struct *tsk)
+{
+	struct mm_struct *mm = tsk->mm;
+
+	if (!cmpxchg(&tsk->signal->oom_mm, NULL, mm)) {
+		mmgrab(tsk->signal->oom_mm);
+		set_bit(MMF_OOM_VICTIM, &mm->flags);
+	}
+}
+
 /**
  * mark_oom_victim - mark the given task as OOM victim
  * @tsk: task to mark
@@ -699,18 +719,13 @@ static inline void wake_oom_reaper(struct task_struct *tsk)
  */
 static void mark_oom_victim(struct task_struct *tsk)
 {
-	struct mm_struct *mm = tsk->mm;
-
 	WARN_ON(oom_killer_disabled);
 	/* OOM killer might race with memcg OOM */
 	if (test_and_set_tsk_thread_flag(tsk, TIF_MEMDIE))
 		return;
 
 	/* oom_mm is bound to the signal struct life time. */
-	if (!cmpxchg(&tsk->signal->oom_mm, NULL, mm)) {
-		mmgrab(tsk->signal->oom_mm);
-		set_bit(MMF_OOM_VICTIM, &mm->flags);
-	}
+	__mark_oom_victim(tsk);
 
 	/*
 	 * Make sure that the task is woken up from uninterruptible sleep
@@ -1140,3 +1155,21 @@ void pagefault_out_of_memory(void)
 	out_of_memory(&oc);
 	mutex_unlock(&oom_lock);
 }
+
+void add_to_oom_reaper(struct task_struct *p)
+{
+	if (!sysctl_reap_mem_on_sigkill)
+		return;
+
+	p = find_lock_task_mm(p);
+	if (!p)
+		return;
+
+	get_task_struct(p);
+	if (task_will_free_mem(p)) {
+		__mark_oom_victim(p);
+		wake_oom_reaper(p);
+	}
+	task_unlock(p);
+	put_task_struct(p);
+}