Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar: "Various fixlets, mostly related to the (root-only) SCHED_DEADLINE policy, but also a hotplug bug fix and a fix for a NR_CPUS related overallocation bug causing a suspend/resume regression" * 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: sched: Fix hotplug vs. set_cpus_allowed_ptr() sched/cpupri: Replace NR_CPUS arrays sched/deadline: Replace NR_CPUS arrays sched/deadline: Restrict user params max value to 2^63 ns sched/deadline: Change sched_getparam() behaviour vs SCHED_DEADLINE sched: Disallow sched_attr::sched_policy < 0 sched: Make sched_setattr() correctly return -EFBIG

Merge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull scheduler fixes from Ingo Molnar: "Various fixlets, mostly related to the (root-only) SCHED_DEADLINE policy, but also a hotplug bug fix and a fix for a NR_CPUS related overallocation bug causing a suspend/resume regression" * 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: sched: Fix hotplug vs. set_cpus_allowed_ptr() sched/cpupri: Replace NR_CPUS arrays sched/deadline: Replace NR_CPUS arrays sched/deadline: Restrict user params max value to 2^63 ns sched/deadline: Change sched_getparam() behaviour vs SCHED_DEADLINE sched: Disallow sched_attr::sched_policy < 0 sched: Make sched_setattr() correctly return -EFBIG
Linus Torvalds
2 parents a4bf79eb6a 6acbfb9697
Showing 6 changed files Side-by-side Diff
kernel/cpu.c
kernel/sched/core.c
kernel/sched/cpudeadline.c
kernel/sched/cpudeadline.h
kernel/sched/cpupri.c
kernel/sched/cpupri.h
@@ -726,10 +726,12 @@
  
 void set_cpu_online(unsigned int cpu, bool online)
 {
-	if (online)
+	if (online) {
 		cpumask_set_cpu(cpu, to_cpumask(cpu_online_bits));
-	else
+		cpumask_set_cpu(cpu, to_cpumask(cpu_active_bits));
+	} else {
 		cpumask_clear_cpu(cpu, to_cpumask(cpu_online_bits));
+	}
 }
  
 void set_cpu_active(unsigned int cpu, bool active)
@@ -3195,17 +3195,40 @@
  * We ask for the deadline not being zero, and greater or equal
  * than the runtime, as well as the period of being zero or
  * greater than deadline. Furthermore, we have to be sure that
- * user parameters are above the internal resolution (1us); we
- * check sched_runtime only since it is always the smaller one.
+ * user parameters are above the internal resolution of 1us (we
+ * check sched_runtime only since it is always the smaller one) and
+ * below 2^63 ns (we have to check both sched_deadline and
+ * sched_period, as the latter can be zero).
  */
 static bool
 __checkparam_dl(const struct sched_attr *attr)
 {
-	return attr && attr->sched_deadline != 0 &&
-		(attr->sched_period == 0 ||
-		(s64)(attr->sched_period   - attr->sched_deadline) >= 0) &&
-		(s64)(attr->sched_deadline - attr->sched_runtime ) >= 0  &&
-		attr->sched_runtime >= (2 << (DL_SCALE - 1));
+	/* deadline != 0 */
+	if (attr->sched_deadline == 0)
+		return false;
+
+	/*
+	 * Since we truncate DL_SCALE bits, make sure we're at least
+	 * that big.
+	 */
+	if (attr->sched_runtime < (1ULL << DL_SCALE))
+		return false;
+
+	/*
+	 * Since we use the MSB for wrap-around and sign issues, make
+	 * sure it's not set (mind that period can be equal to zero).
+	 */
+	if (attr->sched_deadline & (1ULL << 63) ||
+	    attr->sched_period & (1ULL << 63))
+		return false;
+
+	/* runtime <= deadline <= period (if period != 0) */
+	if ((attr->sched_period != 0 &&
+	     attr->sched_period < attr->sched_deadline) ||
+	    attr->sched_deadline < attr->sched_runtime)
+		return false;
+
+	return true;
 }
  
 /*
  
@@ -3658,9 +3681,13 @@
 	if (!uattr || pid < 0 || flags)
 		return -EINVAL;
  
-	if (sched_copy_attr(uattr, &attr))
-		return -EFAULT;
+	retval = sched_copy_attr(uattr, &attr);
+	if (retval)
+		return retval;
  
+	if (attr.sched_policy < 0)
+		return -EINVAL;
+
 	rcu_read_lock();
 	retval = -ESRCH;
 	p = find_process_by_pid(pid);
@@ -3709,7 +3736,7 @@
  */
 SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
 {
-	struct sched_param lp;
+	struct sched_param lp = { .sched_priority = 0 };
 	struct task_struct *p;
 	int retval;
  
@@ -3726,11 +3753,8 @@
 	if (retval)
 		goto out_unlock;
  
-	if (task_has_dl_policy(p)) {
-		retval = -EINVAL;
-		goto out_unlock;
-	}
-	lp.sched_priority = p->rt_priority;
+	if (task_has_rt_policy(p))
+		lp.sched_priority = p->rt_priority;
 	rcu_read_unlock();
  
 	/*
@@ -5052,7 +5076,6 @@
 				      unsigned long action, void *hcpu)
 {
 	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_STARTING:
 	case CPU_DOWN_FAILED:
 		set_cpu_active((long)hcpu, true);
 		return NOTIFY_OK;
@@ -13,6 +13,7 @@
  
 #include <linux/gfp.h>
 #include <linux/kernel.h>
+#include <linux/slab.h>
 #include "cpudeadline.h"
  
 static inline int parent(int i)
@@ -39,8 +40,10 @@
 {
 	int cpu_a = cp->elements[a].cpu, cpu_b = cp->elements[b].cpu;
  
-	swap(cp->elements[a], cp->elements[b]);
-	swap(cp->cpu_to_idx[cpu_a], cp->cpu_to_idx[cpu_b]);
+	swap(cp->elements[a].cpu, cp->elements[b].cpu);
+	swap(cp->elements[a].dl , cp->elements[b].dl );
+
+	swap(cp->elements[cpu_a].idx, cp->elements[cpu_b].idx);
 }
  
 static void cpudl_heapify(struct cpudl *cp, int idx)
@@ -140,7 +143,7 @@
 	WARN_ON(!cpu_present(cpu));
  
 	raw_spin_lock_irqsave(&cp->lock, flags);
-	old_idx = cp->cpu_to_idx[cpu];
+	old_idx = cp->elements[cpu].idx;
 	if (!is_valid) {
 		/* remove item */
 		if (old_idx == IDX_INVALID) {
@@ -155,8 +158,8 @@
 		cp->elements[old_idx].dl = cp->elements[cp->size - 1].dl;
 		cp->elements[old_idx].cpu = new_cpu;
 		cp->size--;
-		cp->cpu_to_idx[new_cpu] = old_idx;
-		cp->cpu_to_idx[cpu] = IDX_INVALID;
+		cp->elements[new_cpu].idx = old_idx;
+		cp->elements[cpu].idx = IDX_INVALID;
 		while (old_idx > 0 && dl_time_before(
 				cp->elements[parent(old_idx)].dl,
 				cp->elements[old_idx].dl)) {
@@ -173,7 +176,7 @@
 		cp->size++;
 		cp->elements[cp->size - 1].dl = 0;
 		cp->elements[cp->size - 1].cpu = cpu;
-		cp->cpu_to_idx[cpu] = cp->size - 1;
+		cp->elements[cpu].idx = cp->size - 1;
 		cpudl_change_key(cp, cp->size - 1, dl);
 		cpumask_clear_cpu(cpu, cp->free_cpus);
 	} else {
  
@@ -195,10 +198,21 @@
 	memset(cp, 0, sizeof(*cp));
 	raw_spin_lock_init(&cp->lock);
 	cp->size = 0;
-	for (i = 0; i < NR_CPUS; i++)
-		cp->cpu_to_idx[i] = IDX_INVALID;
-	if (!alloc_cpumask_var(&cp->free_cpus, GFP_KERNEL))
+
+	cp->elements = kcalloc(nr_cpu_ids,
+			       sizeof(struct cpudl_item),
+			       GFP_KERNEL);
+	if (!cp->elements)
 		return -ENOMEM;
+
+	if (!alloc_cpumask_var(&cp->free_cpus, GFP_KERNEL)) {
+		kfree(cp->elements);
+		return -ENOMEM;
+	}
+
+	for_each_possible_cpu(i)
+		cp->elements[i].idx = IDX_INVALID;
+
 	cpumask_setall(cp->free_cpus);
  
 	return 0;
@@ -211,5 +225,6 @@
 void cpudl_cleanup(struct cpudl *cp)
 {
 	free_cpumask_var(cp->free_cpus);
+	kfree(cp->elements);
 }
@@ -5,17 +5,17 @@
  
 #define IDX_INVALID     -1
  
-struct array_item {
+struct cpudl_item {
 	u64 dl;
 	int cpu;
+	int idx;
 };
  
 struct cpudl {
 	raw_spinlock_t lock;
 	int size;
-	int cpu_to_idx[NR_CPUS];
-	struct array_item elements[NR_CPUS];
 	cpumask_var_t free_cpus;
+	struct cpudl_item *elements;
 };
  
  
@@ -30,6 +30,7 @@
 #include <linux/gfp.h>
 #include <linux/sched.h>
 #include <linux/sched/rt.h>
+#include <linux/slab.h>
 #include "cpupri.h"
  
 /* Convert between a 140 based task->prio, and our 102 based cpupri */
  
@@ -218,8 +219,13 @@
 			goto cleanup;
 	}
  
+	cp->cpu_to_pri = kcalloc(nr_cpu_ids, sizeof(int), GFP_KERNEL);
+	if (!cp->cpu_to_pri)
+		goto cleanup;
+
 	for_each_possible_cpu(i)
 		cp->cpu_to_pri[i] = CPUPRI_INVALID;
+
 	return 0;
  
 cleanup:
@@ -236,6 +242,7 @@
 {
 	int i;
  
+	kfree(cp->cpu_to_pri);
 	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++)
 		free_cpumask_var(cp->pri_to_cpu[i].mask);
 }
@@ -17,7 +17,7 @@
  
 struct cpupri {
 	struct cpupri_vec pri_to_cpu[CPUPRI_NR_PRIORITIES];
-	int               cpu_to_pri[NR_CPUS];
+	int *cpu_to_pri;
 };
  
 #ifdef CONFIG_SMP
...	...	@@ -726,10 +726,12 @@
726	726
727	727	void set_cpu_online(unsigned int cpu, bool online)
728	728	{
729		- if (online)
	729	+ if (online) {
730	730	cpumask_set_cpu(cpu, to_cpumask(cpu_online_bits));
731		- else
	731	+ cpumask_set_cpu(cpu, to_cpumask(cpu_active_bits));
	732	+ } else {
732	733	cpumask_clear_cpu(cpu, to_cpumask(cpu_online_bits));
	734	+ }
733	735	}
734	736
735	737	void set_cpu_active(unsigned int cpu, bool active)
...	...	@@ -3195,17 +3195,40 @@
3195	3195	* We ask for the deadline not being zero, and greater or equal
3196	3196	* than the runtime, as well as the period of being zero or
3197	3197	* greater than deadline. Furthermore, we have to be sure that
3198		- * user parameters are above the internal resolution (1us); we
3199		- * check sched_runtime only since it is always the smaller one.
	3198	+ * user parameters are above the internal resolution of 1us (we
	3199	+ * check sched_runtime only since it is always the smaller one) and
	3200	+ * below 2^63 ns (we have to check both sched_deadline and
	3201	+ * sched_period, as the latter can be zero).
3200	3202	*/
3201	3203	static bool
3202	3204	__checkparam_dl(const struct sched_attr *attr)
3203	3205	{
3204		- return attr && attr->sched_deadline != 0 &&
3205		- (attr->sched_period == 0 \|\|
3206		- (s64)(attr->sched_period - attr->sched_deadline) >= 0) &&
3207		- (s64)(attr->sched_deadline - attr->sched_runtime ) >= 0 &&
3208		- attr->sched_runtime >= (2 << (DL_SCALE - 1));
	3206	+ /* deadline != 0 */
	3207	+ if (attr->sched_deadline == 0)
	3208	+ return false;
	3209	+
	3210	+ /*
	3211	+ * Since we truncate DL_SCALE bits, make sure we're at least
	3212	+ * that big.
	3213	+ */
	3214	+ if (attr->sched_runtime < (1ULL << DL_SCALE))
	3215	+ return false;
	3216	+
	3217	+ /*
	3218	+ * Since we use the MSB for wrap-around and sign issues, make
	3219	+ * sure it's not set (mind that period can be equal to zero).
	3220	+ */
	3221	+ if (attr->sched_deadline & (1ULL << 63) \|\|
	3222	+ attr->sched_period & (1ULL << 63))
	3223	+ return false;
	3224	+
	3225	+ /* runtime <= deadline <= period (if period != 0) */
	3226	+ if ((attr->sched_period != 0 &&
	3227	+ attr->sched_period < attr->sched_deadline) \|\|
	3228	+ attr->sched_deadline < attr->sched_runtime)
	3229	+ return false;
	3230	+
	3231	+ return true;
3209	3232	}
3210	3233
3211	3234	/*
3212	3235
...	...	@@ -3658,9 +3681,13 @@
3658	3681	if (!uattr \|\| pid < 0 \|\| flags)
3659	3682	return -EINVAL;
3660	3683
3661		- if (sched_copy_attr(uattr, &attr))
3662		- return -EFAULT;
	3684	+ retval = sched_copy_attr(uattr, &attr);
	3685	+ if (retval)
	3686	+ return retval;
3663	3687
	3688	+ if (attr.sched_policy < 0)
	3689	+ return -EINVAL;
	3690	+
3664	3691	rcu_read_lock();
3665	3692	retval = -ESRCH;
3666	3693	p = find_process_by_pid(pid);
...	...	@@ -3709,7 +3736,7 @@
3709	3736	*/
3710	3737	SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
3711	3738	{
3712		- struct sched_param lp;
	3739	+ struct sched_param lp = { .sched_priority = 0 };
3713	3740	struct task_struct *p;
3714	3741	int retval;
3715	3742
...	...	@@ -3726,11 +3753,8 @@
3726	3753	if (retval)
3727	3754	goto out_unlock;
3728	3755
3729		- if (task_has_dl_policy(p)) {
3730		- retval = -EINVAL;
3731		- goto out_unlock;
3732		- }
3733		- lp.sched_priority = p->rt_priority;
	3756	+ if (task_has_rt_policy(p))
	3757	+ lp.sched_priority = p->rt_priority;
3734	3758	rcu_read_unlock();
3735	3759
3736	3760	/*
...	...	@@ -5052,7 +5076,6 @@
5052	5076	unsigned long action, void *hcpu)
5053	5077	{
5054	5078	switch (action & ~CPU_TASKS_FROZEN) {
5055		- case CPU_STARTING:
5056	5079	case CPU_DOWN_FAILED:
5057	5080	set_cpu_active((long)hcpu, true);
5058	5081	return NOTIFY_OK;
...	...	@@ -13,6 +13,7 @@
13	13
14	14	#include <linux/gfp.h>
15	15	#include <linux/kernel.h>
	16	+#include <linux/slab.h>
16	17	#include "cpudeadline.h"
17	18
18	19	static inline int parent(int i)
...	...	@@ -39,8 +40,10 @@
39	40	{
40	41	int cpu_a = cp->elements[a].cpu, cpu_b = cp->elements[b].cpu;
41	42
42		- swap(cp->elements[a], cp->elements[b]);
43		- swap(cp->cpu_to_idx[cpu_a], cp->cpu_to_idx[cpu_b]);
	43	+ swap(cp->elements[a].cpu, cp->elements[b].cpu);
	44	+ swap(cp->elements[a].dl , cp->elements[b].dl );
	45	+
	46	+ swap(cp->elements[cpu_a].idx, cp->elements[cpu_b].idx);
44	47	}
45	48
46	49	static void cpudl_heapify(struct cpudl *cp, int idx)
...	...	@@ -140,7 +143,7 @@
140	143	WARN_ON(!cpu_present(cpu));
141	144
142	145	raw_spin_lock_irqsave(&cp->lock, flags);
143		- old_idx = cp->cpu_to_idx[cpu];
	146	+ old_idx = cp->elements[cpu].idx;
144	147	if (!is_valid) {
145	148	/* remove item */
146	149	if (old_idx == IDX_INVALID) {
...	...	@@ -155,8 +158,8 @@
155	158	cp->elements[old_idx].dl = cp->elements[cp->size - 1].dl;
156	159	cp->elements[old_idx].cpu = new_cpu;
157	160	cp->size--;
158		- cp->cpu_to_idx[new_cpu] = old_idx;
159		- cp->cpu_to_idx[cpu] = IDX_INVALID;
	161	+ cp->elements[new_cpu].idx = old_idx;
	162	+ cp->elements[cpu].idx = IDX_INVALID;
160	163	while (old_idx > 0 && dl_time_before(
161	164	cp->elements[parent(old_idx)].dl,
162	165	cp->elements[old_idx].dl)) {
...	...	@@ -173,7 +176,7 @@
173	176	cp->size++;
174	177	cp->elements[cp->size - 1].dl = 0;
175	178	cp->elements[cp->size - 1].cpu = cpu;
176		- cp->cpu_to_idx[cpu] = cp->size - 1;
	179	+ cp->elements[cpu].idx = cp->size - 1;
177	180	cpudl_change_key(cp, cp->size - 1, dl);
178	181	cpumask_clear_cpu(cpu, cp->free_cpus);
179	182	} else {
180	183
...	...	@@ -195,10 +198,21 @@
195	198	memset(cp, 0, sizeof(*cp));
196	199	raw_spin_lock_init(&cp->lock);
197	200	cp->size = 0;
198		- for (i = 0; i < NR_CPUS; i++)
199		- cp->cpu_to_idx[i] = IDX_INVALID;
200		- if (!alloc_cpumask_var(&cp->free_cpus, GFP_KERNEL))
	201	+
	202	+ cp->elements = kcalloc(nr_cpu_ids,
	203	+ sizeof(struct cpudl_item),
	204	+ GFP_KERNEL);
	205	+ if (!cp->elements)
201	206	return -ENOMEM;
	207	+
	208	+ if (!alloc_cpumask_var(&cp->free_cpus, GFP_KERNEL)) {
	209	+ kfree(cp->elements);
	210	+ return -ENOMEM;
	211	+ }
	212	+
	213	+ for_each_possible_cpu(i)
	214	+ cp->elements[i].idx = IDX_INVALID;
	215	+
202	216	cpumask_setall(cp->free_cpus);
203	217
204	218	return 0;
...	...	@@ -211,5 +225,6 @@
211	225	void cpudl_cleanup(struct cpudl *cp)
212	226	{
213	227	free_cpumask_var(cp->free_cpus);
	228	+ kfree(cp->elements);
214	229	}
...	...	@@ -5,17 +5,17 @@
5	5
6	6	#define IDX_INVALID -1
7	7
8		-struct array_item {
	8	+struct cpudl_item {
9	9	u64 dl;
10	10	int cpu;
	11	+ int idx;
11	12	};
12	13
13	14	struct cpudl {
14	15	raw_spinlock_t lock;
15	16	int size;
16		- int cpu_to_idx[NR_CPUS];
17		- struct array_item elements[NR_CPUS];
18	17	cpumask_var_t free_cpus;
	18	+ struct cpudl_item *elements;
19	19	};
20	20
21	21
...	...	@@ -30,6 +30,7 @@
30	30	#include <linux/gfp.h>
31	31	#include <linux/sched.h>
32	32	#include <linux/sched/rt.h>
	33	+#include <linux/slab.h>
33	34	#include "cpupri.h"
34	35
35	36	/* Convert between a 140 based task->prio, and our 102 based cpupri */
36	37
...	...	@@ -218,8 +219,13 @@
218	219	goto cleanup;
219	220	}
220	221
	222	+ cp->cpu_to_pri = kcalloc(nr_cpu_ids, sizeof(int), GFP_KERNEL);
	223	+ if (!cp->cpu_to_pri)
	224	+ goto cleanup;
	225	+
221	226	for_each_possible_cpu(i)
222	227	cp->cpu_to_pri[i] = CPUPRI_INVALID;
	228	+
223	229	return 0;
224	230
225	231	cleanup:
...	...	@@ -236,6 +242,7 @@
236	242	{
237	243	int i;
238	244
	245	+ kfree(cp->cpu_to_pri);
239	246	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++)
240	247	free_cpumask_var(cp->pri_to_cpu[i].mask);
241	248	}
...	...	@@ -17,7 +17,7 @@
17	17
18	18	struct cpupri {
19	19	struct cpupri_vec pri_to_cpu[CPUPRI_NR_PRIORITIES];
20		- int cpu_to_pri[NR_CPUS];
	20	+ int *cpu_to_pri;
21	21	};
22	22
23	23	#ifdef CONFIG_SMP