[PATCH] sched/deadline: Impose global limits on sched_attr::sched_period

[Peter Zijlstra]

There are two DoS scenarios with SCHED_DEADLINE related to
sched_attr::sched_period:

 - since access-control only looks at utilization and density, a very
   large period can allow a very large runtime, which in turn can
   incur a very large latency to lower priority tasks.

 - for very short periods we can end up spending more time programming
   the hardware timer than actually running the task.

Mitigate these by imposing limits on the period.

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Signed-off-by: Alessio Balsini <balsini@xxxxxxxxxxx>
Cc: rostedt@xxxxxxxxxxx
Cc: mingo@xxxxxxxxxx
Cc: luca.abeni@xxxxxxxxxxxxxxx
Cc: bristot@xxxxxxxxxx
Cc: vpillai@xxxxxxxxxxxxxxxx
Cc: kernel-team@xxxxxxxxxxx
Cc: juri.lelli@xxxxxxxxxx
Cc: dietmar.eggemann@xxxxxxx
Cc: dvyukov@xxxxxxxxxx
Cc: tglx@xxxxxxxxxxxxx
---
 include/linux/sched/sysctl.h |  7 ++++
 kernel/sched/deadline.c      | 68 ++++++++++++++++++++++++++++++++++--
 kernel/sysctl.c              | 14 ++++++++
 3 files changed, 87 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index d4f6215ee03f7..7c8ef07e52133 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -56,6 +56,13 @@ int sched_proc_update_handler(struct ctl_table *table, int write,
 extern unsigned int sysctl_sched_rt_period;
 extern int sysctl_sched_rt_runtime;
 
+extern unsigned int sysctl_sched_dl_period_max;
+extern unsigned int sysctl_sched_dl_period_min;
+
+extern int sched_dl_period_handler(struct ctl_table *table, int write,
+		void __user *buffer, size_t *lenp,
+		loff_t *ppos);
+
 #ifdef CONFIG_UCLAMP_TASK
 extern unsigned int sysctl_sched_uclamp_util_min;
 extern unsigned int sysctl_sched_uclamp_util_max;
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 0b9cbfb2b1d4f..c4a6107e055c7 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -2640,6 +2640,59 @@ void __getparam_dl(struct task_struct *p, struct sched_attr *attr)
 	attr->sched_flags = dl_se->flags;
 }
 
+/*
+ * Default limits for DL period: on the top end we guard against small util
+ * tasks still getting ridiculous long effective runtimes, on the bottom end we
+ * guard against timer DoS.
+ */
+unsigned int sysctl_sched_dl_period_max = 1 << 22; /* ~4.2 seconds */
+unsigned int sysctl_sched_dl_period_min = 100;     /* 100 us */
+
+int sched_dl_period_handler(struct ctl_table *table, int write,
+			    void __user *buffer, size_t *lenp,
+			    loff_t *ppos)
+{
+	static DEFINE_MUTEX(mutex);
+	int ret;
+
+	mutex_lock(&mutex);
+	if (write) {
+		/*
+		 * Use a temporary data structure to store the value read from
+		 * userspace. The sysctl_sched_dl_period_{max,min} value will
+		 * be updated only if the data is consistent.
+		 */
+		struct ctl_table new_table = *table;
+		unsigned int max, min;
+
+		if (new_table.data == &sysctl_sched_dl_period_max) {
+			new_table.data = &max;
+			min = sysctl_sched_dl_period_min;
+		} else {
+			new_table.data = &min;
+			max = sysctl_sched_dl_period_max;
+		}
+
+		ret = proc_douintvec(&new_table, write, buffer, lenp, ppos);
+		if (!ret) {
+			if (min > max ||
+			    (u64)min * NSEC_PER_USEC < 1ULL << DL_SCALE) {
+				ret = -EINVAL;
+			} else {
+				unsigned int *src = new_table.data;
+				unsigned int *dst = table->data;
+
+				WRITE_ONCE(*dst, *src);
+			}
+		}
+	} else {
+		ret = proc_douintvec(table, write, buffer, lenp, ppos);
+	}
+	mutex_unlock(&mutex);
+
+	return ret;
+}
+
 /*
  * This function validates the new parameters of a -deadline task.
  * We ask for the deadline not being zero, and greater or equal
@@ -2652,6 +2705,8 @@ void __getparam_dl(struct task_struct *p, struct sched_attr *attr)
  */
 bool __checkparam_dl(const struct sched_attr *attr)
 {
+	u64 period, max, min;
+
 	/* special dl tasks don't actually use any parameter */
 	if (attr->sched_flags & SCHED_FLAG_SUGOV)
 		return true;
@@ -2675,12 +2730,21 @@ bool __checkparam_dl(const struct sched_attr *attr)
 	    attr->sched_period & (1ULL << 63))
 		return false;
 
+	period = attr->sched_period;
+	if (!period)
+		period = attr->sched_deadline;
+
 	/* runtime <= deadline <= period (if period != 0) */
-	if ((attr->sched_period != 0 &&
-	     attr->sched_period < attr->sched_deadline) ||
+	if (period < attr->sched_deadline ||
 	    attr->sched_deadline < attr->sched_runtime)
 		return false;
 
+	max = (u64)READ_ONCE(sysctl_sched_dl_period_max) * NSEC_PER_USEC;
+	min = (u64)READ_ONCE(sysctl_sched_dl_period_min) * NSEC_PER_USEC;
+
+	if (period < min || period > max)
+		return false;
+
 	return true;
 }
 
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 078950d9605ba..0d07e4707e9d2 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -442,6 +442,20 @@ static struct ctl_table kern_table[] = {
 		.mode		= 0644,
 		.proc_handler	= sched_rt_handler,
 	},
+	{
+		.procname	= "sched_deadline_period_max_us",
+		.data		= &sysctl_sched_dl_period_max,
+		.maxlen		= sizeof(unsigned int),
+		.mode		= 0644,
+		.proc_handler	= sched_dl_period_handler,
+	},
+	{
+		.procname	= "sched_deadline_period_min_us",
+		.data		= &sysctl_sched_dl_period_min,
+		.maxlen		= sizeof(unsigned int),
+		.mode		= 0644,
+		.proc_handler	= sched_dl_period_handler,
+	},
 	{
 		.procname	= "sched_rr_timeslice_ms",
 		.data		= &sysctl_sched_rr_timeslice,
-- 
2.23.0.187.g17f5b7556c-goog