[thirdparty/kernel/stable.git] / lib / test_vmalloc.c

// SPDX-License-Identifier: GPL-2.0

/*
 * Test module for stress and analyze performance of vmalloc allocator.
 * (C) 2018 Uladzislau Rezki (Sony) <urezki@gmail.com>
 */
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/vmalloc.h>
#include <linux/random.h>
#include <linux/kthread.h>
#include <linux/moduleparam.h>
#include <linux/completion.h>
#include <linux/delay.h>
#include <linux/rwsem.h>
#include <linux/mm.h>

#define __param(type, name, init, msg)		\
	static type name = init;				\
	module_param(name, type, 0444);			\
	MODULE_PARM_DESC(name, msg)				\

__param(bool, single_cpu_test, false,
	"Use single first online CPU to run tests");

__param(bool, sequential_test_order, false,
	"Use sequential stress tests order");

__param(int, test_repeat_count, 1,
	"Set test repeat counter");

__param(int, test_loop_count, 1000000,
	"Set test loop counter");

__param(int, run_test_mask, INT_MAX,
	"Set tests specified in the mask.\n\n"
		"\t\tid: 1,   name: fix_size_alloc_test\n"
		"\t\tid: 2,   name: full_fit_alloc_test\n"
		"\t\tid: 4,   name: long_busy_list_alloc_test\n"
		"\t\tid: 8,   name: random_size_alloc_test\n"
		"\t\tid: 16,  name: fix_align_alloc_test\n"
		"\t\tid: 32,  name: random_size_align_alloc_test\n"
		"\t\tid: 64,  name: align_shift_alloc_test\n"
		"\t\tid: 128, name: pcpu_alloc_test\n"
		/* Add a new test case description here. */
);

/*
 * Depends on single_cpu_test parameter. If it is true, then
 * use first online CPU to trigger a test on, otherwise go with
 * all online CPUs.
 */
static cpumask_t cpus_run_test_mask = CPU_MASK_NONE;

/*
 * Read write semaphore for synchronization of setup
 * phase that is done in main thread and workers.
 */
static DECLARE_RWSEM(prepare_for_test_rwsem);

/*
 * Completion tracking for worker threads.
 */
static DECLARE_COMPLETION(test_all_done_comp);
static atomic_t test_n_undone = ATOMIC_INIT(0);

static inline void
test_report_one_done(void)
{
	if (atomic_dec_and_test(&test_n_undone))
		complete(&test_all_done_comp);
}

static int random_size_align_alloc_test(void)
{
	unsigned long size, align, rnd;
	void *ptr;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		get_random_bytes(&rnd, sizeof(rnd));

		/*
		 * Maximum 1024 pages, if PAGE_SIZE is 4096.
		 */
		align = 1 << (rnd % 23);

		/*
		 * Maximum 10 pages.
		 */
		size = ((rnd % 10) + 1) * PAGE_SIZE;

		ptr = __vmalloc_node_range(size, align,
		   VMALLOC_START, VMALLOC_END,
		   GFP_KERNEL | __GFP_ZERO,
		   PAGE_KERNEL,
		   0, 0, __builtin_return_address(0));

		if (!ptr)
			return -1;

		vfree(ptr);
	}

	return 0;
}

/*
 * This test case is supposed to be failed.
 */
static int align_shift_alloc_test(void)
{
	unsigned long align;
	void *ptr;
	int i;

	for (i = 0; i < BITS_PER_LONG; i++) {
		align = ((unsigned long) 1) << i;

		ptr = __vmalloc_node_range(PAGE_SIZE, align,
			VMALLOC_START, VMALLOC_END,
			GFP_KERNEL | __GFP_ZERO,
			PAGE_KERNEL,
			0, 0, __builtin_return_address(0));

		if (!ptr)
			return -1;

		vfree(ptr);
	}

	return 0;
}

static int fix_align_alloc_test(void)
{
	void *ptr;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		ptr = __vmalloc_node_range(5 * PAGE_SIZE,
			THREAD_ALIGN << 1,
			VMALLOC_START, VMALLOC_END,
			GFP_KERNEL | __GFP_ZERO,
			PAGE_KERNEL,
			0, 0, __builtin_return_address(0));

		if (!ptr)
			return -1;

		vfree(ptr);
	}

	return 0;
}

static int random_size_alloc_test(void)
{
	unsigned int n;
	void *p;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		get_random_bytes(&n, sizeof(i));
		n = (n % 100) + 1;

		p = vmalloc(n * PAGE_SIZE);

		if (!p)
			return -1;

		*((__u8 *)p) = 1;
		vfree(p);
	}

	return 0;
}

static int long_busy_list_alloc_test(void)
{
	void *ptr_1, *ptr_2;
	void **ptr;
	int rv = -1;
	int i;

	ptr = vmalloc(sizeof(void *) * 15000);
	if (!ptr)
		return rv;

	for (i = 0; i < 15000; i++)
		ptr[i] = vmalloc(1 * PAGE_SIZE);

	for (i = 0; i < test_loop_count; i++) {
		ptr_1 = vmalloc(100 * PAGE_SIZE);
		if (!ptr_1)
			goto leave;

		ptr_2 = vmalloc(1 * PAGE_SIZE);
		if (!ptr_2) {
			vfree(ptr_1);
			goto leave;
		}

		*((__u8 *)ptr_1) = 0;
		*((__u8 *)ptr_2) = 1;

		vfree(ptr_1);
		vfree(ptr_2);
	}

	/*  Success */
	rv = 0;

leave:
	for (i = 0; i < 15000; i++)
		vfree(ptr[i]);

	vfree(ptr);
	return rv;
}

static int full_fit_alloc_test(void)
{
	void **ptr, **junk_ptr, *tmp;
	int junk_length;
	int rv = -1;
	int i;

	junk_length = fls(num_online_cpus());
	junk_length *= (32 * 1024 * 1024 / PAGE_SIZE);

	ptr = vmalloc(sizeof(void *) * junk_length);
	if (!ptr)
		return rv;

	junk_ptr = vmalloc(sizeof(void *) * junk_length);
	if (!junk_ptr) {
		vfree(ptr);
		return rv;
	}

	for (i = 0; i < junk_length; i++) {
		ptr[i] = vmalloc(1 * PAGE_SIZE);
		junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
	}

	for (i = 0; i < junk_length; i++)
		vfree(junk_ptr[i]);

	for (i = 0; i < test_loop_count; i++) {
		tmp = vmalloc(1 * PAGE_SIZE);

		if (!tmp)
			goto error;

		*((__u8 *)tmp) = 1;
		vfree(tmp);
	}

	/* Success */
	rv = 0;

error:
	for (i = 0; i < junk_length; i++)
		vfree(ptr[i]);

	vfree(ptr);
	vfree(junk_ptr);

	return rv;
}

static int fix_size_alloc_test(void)
{
	void *ptr;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		ptr = vmalloc(3 * PAGE_SIZE);

		if (!ptr)
			return -1;

		*((__u8 *)ptr) = 0;

		vfree(ptr);
	}

	return 0;
}

static int
pcpu_alloc_test(void)
{
	int rv = 0;
#ifndef CONFIG_NEED_PER_CPU_KM
	void __percpu **pcpu;
	size_t size, align;
	int i;

	pcpu = vmalloc(sizeof(void __percpu *) * 35000);
	if (!pcpu)
		return -1;

	for (i = 0; i < 35000; i++) {
		unsigned int r;

		get_random_bytes(&r, sizeof(i));
		size = (r % (PAGE_SIZE / 4)) + 1;

		/*
		 * Maximum PAGE_SIZE
		 */
		get_random_bytes(&r, sizeof(i));
		align = 1 << ((i % 11) + 1);

		pcpu[i] = __alloc_percpu(size, align);
		if (!pcpu[i])
			rv = -1;
	}

	for (i = 0; i < 35000; i++)
		free_percpu(pcpu[i]);

	vfree(pcpu);
#endif
	return rv;
}

struct test_case_desc {
	const char *test_name;
	int (*test_func)(void);
};

static struct test_case_desc test_case_array[] = {
	{ "fix_size_alloc_test", fix_size_alloc_test },
	{ "full_fit_alloc_test", full_fit_alloc_test },
	{ "long_busy_list_alloc_test", long_busy_list_alloc_test },
	{ "random_size_alloc_test", random_size_alloc_test },
	{ "fix_align_alloc_test", fix_align_alloc_test },
	{ "random_size_align_alloc_test", random_size_align_alloc_test },
	{ "align_shift_alloc_test", align_shift_alloc_test },
	{ "pcpu_alloc_test", pcpu_alloc_test },
	/* Add a new test case here. */
};

struct test_case_data {
	int test_failed;
	int test_passed;
	u64 time;
};

/* Split it to get rid of: WARNING: line over 80 characters */
static struct test_case_data
	per_cpu_test_data[NR_CPUS][ARRAY_SIZE(test_case_array)];

static struct test_driver {
	struct task_struct *task;
	unsigned long start;
	unsigned long stop;
	int cpu;
} per_cpu_test_driver[NR_CPUS];

static void shuffle_array(int *arr, int n)
{
	unsigned int rnd;
	int i, j, x;

	for (i = n - 1; i > 0; i--)  {
		get_random_bytes(&rnd, sizeof(rnd));

		/* Cut the range. */
		j = rnd % i;

		/* Swap indexes. */
		x = arr[i];
		arr[i] = arr[j];
		arr[j] = x;
	}
}

static int test_func(void *private)
{
	struct test_driver *t = private;
	cpumask_t newmask = CPU_MASK_NONE;
	int random_array[ARRAY_SIZE(test_case_array)];
	int index, i, j, ret;
	ktime_t kt;
	u64 delta;

	cpumask_set_cpu(t->cpu, &newmask);
	set_cpus_allowed_ptr(current, &newmask);

	for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
		random_array[i] = i;

	if (!sequential_test_order)
		shuffle_array(random_array, ARRAY_SIZE(test_case_array));

	/*
	 * Block until initialization is done.
	 */
	down_read(&prepare_for_test_rwsem);

	t->start = get_cycles();
	for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
		index = random_array[i];

		/*
		 * Skip tests if run_test_mask has been specified.
		 */
		if (!((run_test_mask & (1 << index)) >> index))
			continue;

		kt = ktime_get();
		for (j = 0; j < test_repeat_count; j++) {
			ret = test_case_array[index].test_func();
			if (!ret)
				per_cpu_test_data[t->cpu][index].test_passed++;
			else
				per_cpu_test_data[t->cpu][index].test_failed++;
		}

		/*
		 * Take an average time that test took.
		 */
		delta = (u64) ktime_us_delta(ktime_get(), kt);
		do_div(delta, (u32) test_repeat_count);

		per_cpu_test_data[t->cpu][index].time = delta;
	}
	t->stop = get_cycles();

	up_read(&prepare_for_test_rwsem);
	test_report_one_done();

	/*
	 * Wait for the kthread_stop() call.
	 */
	while (!kthread_should_stop())
		msleep(10);

	return 0;
}

static void
init_test_configurtion(void)
{
	/*
	 * Reset all data of all CPUs.
	 */
	memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));

	if (single_cpu_test)
		cpumask_set_cpu(cpumask_first(cpu_online_mask),
			&cpus_run_test_mask);
	else
		cpumask_and(&cpus_run_test_mask, cpu_online_mask,
			cpu_online_mask);

	if (test_repeat_count <= 0)
		test_repeat_count = 1;

	if (test_loop_count <= 0)
		test_loop_count = 1;
}

static void do_concurrent_test(void)
{
	int cpu, ret;

	/*
	 * Set some basic configurations plus sanity check.
	 */
	init_test_configurtion();

	/*
	 * Put on hold all workers.
	 */
	down_write(&prepare_for_test_rwsem);

	for_each_cpu(cpu, &cpus_run_test_mask) {
		struct test_driver *t = &per_cpu_test_driver[cpu];

		t->cpu = cpu;
		t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);

		if (!IS_ERR(t->task))
			/* Success. */
			atomic_inc(&test_n_undone);
		else
			pr_err("Failed to start kthread for %d CPU\n", cpu);
	}

	/*
	 * Now let the workers do their job.
	 */
	up_write(&prepare_for_test_rwsem);

	/*
	 * Sleep quiet until all workers are done with 1 second
	 * interval. Since the test can take a lot of time we
	 * can run into a stack trace of the hung task. That is
	 * why we go with completion_timeout and HZ value.
	 */
	do {
		ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
	} while (!ret);

	for_each_cpu(cpu, &cpus_run_test_mask) {
		struct test_driver *t = &per_cpu_test_driver[cpu];
		int i;

		if (!IS_ERR(t->task))
			kthread_stop(t->task);

		for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
			if (!((run_test_mask & (1 << i)) >> i))
				continue;

			pr_info(
				"Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
				test_case_array[i].test_name,
				per_cpu_test_data[cpu][i].test_passed,
				per_cpu_test_data[cpu][i].test_failed,
				test_repeat_count, test_loop_count,
				per_cpu_test_data[cpu][i].time);
		}

		pr_info("All test took CPU%d=%lu cycles\n",
			cpu, t->stop - t->start);
	}
}

static int vmalloc_test_init(void)
{
	do_concurrent_test();
	return -EAGAIN; /* Fail will directly unload the module */
}

static void vmalloc_test_exit(void)
{
}

module_init(vmalloc_test_init)
module_exit(vmalloc_test_exit)

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Uladzislau Rezki");
MODULE_DESCRIPTION("vmalloc test module");
Commit	Line	Data
3f21a6b7 URS	1	// SPDX-License-Identifier: GPL-2.0
	2
	3	/*
	4	* Test module for stress and analyze performance of vmalloc allocator.
	5	* (C) 2018 Uladzislau Rezki (Sony) <urezki@gmail.com>
	6	*/
	7	#include <linux/init.h>
	8	#include <linux/kernel.h>
	9	#include <linux/module.h>
	10	#include <linux/vmalloc.h>
	11	#include <linux/random.h>
	12	#include <linux/kthread.h>
	13	#include <linux/moduleparam.h>
	14	#include <linux/completion.h>
	15	#include <linux/delay.h>
	16	#include <linux/rwsem.h>
	17	#include <linux/mm.h>
	18
	19	#define __param(type, name, init, msg) \
	20	static type name = init; \
	21	module_param(name, type, 0444); \
	22	MODULE_PARM_DESC(name, msg) \
	23
	24	__param(bool, single_cpu_test, false,
	25	"Use single first online CPU to run tests");
	26
	27	__param(bool, sequential_test_order, false,
	28	"Use sequential stress tests order");
	29
	30	__param(int, test_repeat_count, 1,
	31	"Set test repeat counter");
	32
	33	__param(int, test_loop_count, 1000000,
	34	"Set test loop counter");
	35
	36	__param(int, run_test_mask, INT_MAX,
	37	"Set tests specified in the mask.\n\n"
	38	"\t\tid: 1, name: fix_size_alloc_test\n"
	39	"\t\tid: 2, name: full_fit_alloc_test\n"
	40	"\t\tid: 4, name: long_busy_list_alloc_test\n"
	41	"\t\tid: 8, name: random_size_alloc_test\n"
	42	"\t\tid: 16, name: fix_align_alloc_test\n"
	43	"\t\tid: 32, name: random_size_align_alloc_test\n"
	44	"\t\tid: 64, name: align_shift_alloc_test\n"
	45	"\t\tid: 128, name: pcpu_alloc_test\n"
	46	/* Add a new test case description here. */
	47	);
	48
	49	/*
	50	* Depends on single_cpu_test parameter. If it is true, then
	51	* use first online CPU to trigger a test on, otherwise go with
	52	* all online CPUs.
	53	*/
	54	static cpumask_t cpus_run_test_mask = CPU_MASK_NONE;
	55
	56	/*
	57	* Read write semaphore for synchronization of setup
	58	* phase that is done in main thread and workers.
	59	*/
	60	static DECLARE_RWSEM(prepare_for_test_rwsem);
	61
	62	/*
	63	* Completion tracking for worker threads.
	64	*/
65	static DECLARE_COMPLETION(test_all_done_comp);
66	static atomic_t test_n_undone = ATOMIC_INIT(0);
67
68	static inline void
69	test_report_one_done(void)
70	{
71	if (atomic_dec_and_test(&test_n_undone))
72	complete(&test_all_done_comp);
73	}
74
75	static int random_size_align_alloc_test(void)
76	{
77	unsigned long size, align, rnd;
78	void *ptr;
79	int i;
80
81	for (i = 0; i < test_loop_count; i++) {
82	get_random_bytes(&rnd, sizeof(rnd));
83
84	/*
85	* Maximum 1024 pages, if PAGE_SIZE is 4096.
86	*/
87	align = 1 << (rnd % 23);
88
89	/*
90	* Maximum 10 pages.
91	*/
92	size = ((rnd % 10) + 1) * PAGE_SIZE;
93
94	ptr = __vmalloc_node_range(size, align,
95	VMALLOC_START, VMALLOC_END,
96	GFP_KERNEL \| __GFP_ZERO,
97	PAGE_KERNEL,
98	0, 0, __builtin_return_address(0));
99
100	if (!ptr)
101	return -1;
102
103	vfree(ptr);
104	}
105
106	return 0;
107	}
108
109	/*
110	* This test case is supposed to be failed.
111	*/
112	static int align_shift_alloc_test(void)
113	{
114	unsigned long align;
115	void *ptr;
116	int i;
117
118	for (i = 0; i < BITS_PER_LONG; i++) {
119	align = ((unsigned long) 1) << i;
120
121	ptr = __vmalloc_node_range(PAGE_SIZE, align,
122	VMALLOC_START, VMALLOC_END,
123	GFP_KERNEL \| __GFP_ZERO,
124	PAGE_KERNEL,
125	0, 0, __builtin_return_address(0));
126
127	if (!ptr)
128	return -1;
129
130	vfree(ptr);
131	}
132
133	return 0;
134	}
135
136	static int fix_align_alloc_test(void)
137	{
138	void *ptr;
139	int i;
140
141	for (i = 0; i < test_loop_count; i++) {
142	ptr = __vmalloc_node_range(5 * PAGE_SIZE,
143	THREAD_ALIGN << 1,
144	VMALLOC_START, VMALLOC_END,
145	GFP_KERNEL \| __GFP_ZERO,
146	PAGE_KERNEL,
147	0, 0, __builtin_return_address(0));
148
149	if (!ptr)
150	return -1;
151
152	vfree(ptr);
153	}
154
155	return 0;
156	}
157
158	static int random_size_alloc_test(void)
159	{
160	unsigned int n;
161	void *p;
162	int i;
163
164	for (i = 0; i < test_loop_count; i++) {
165	get_random_bytes(&n, sizeof(i));
166	n = (n % 100) + 1;
167
168	p = vmalloc(n * PAGE_SIZE);
169
170	if (!p)
171	return -1;
172
173	((__u8 )p) = 1;
174	vfree(p);
175	}
176
177	return 0;
178	}
179
180	static int long_busy_list_alloc_test(void)
181	{
182	void ptr_1, ptr_2;
183	void **ptr;
184	int rv = -1;
185	int i;
186
187	ptr = vmalloc(sizeof(void ) 15000);
188	if (!ptr)
189	return rv;
190
191	for (i = 0; i < 15000; i++)
192	ptr[i] = vmalloc(1 * PAGE_SIZE);
193
194	for (i = 0; i < test_loop_count; i++) {
195	ptr_1 = vmalloc(100 * PAGE_SIZE);
196	if (!ptr_1)
197	goto leave;
198
199	ptr_2 = vmalloc(1 * PAGE_SIZE);
200	if (!ptr_2) {
201	vfree(ptr_1);
202	goto leave;
203	}
204
205	((__u8 )ptr_1) = 0;
206	((__u8 )ptr_2) = 1;
207
208	vfree(ptr_1);
209	vfree(ptr_2);
210	}
211
212	/* Success */
213	rv = 0;
214
215	leave:
216	for (i = 0; i < 15000; i++)
217	vfree(ptr[i]);
218
219	vfree(ptr);
220	return rv;
221	}
222
223	static int full_fit_alloc_test(void)
224	{
225	void ptr, junk_ptr, *tmp;
226	int junk_length;
227	int rv = -1;
228	int i;
229
230	junk_length = fls(num_online_cpus());
231	junk_length = (32 1024 * 1024 / PAGE_SIZE);
232
233	ptr = vmalloc(sizeof(void ) junk_length);
234	if (!ptr)
235	return rv;
236
237	junk_ptr = vmalloc(sizeof(void ) junk_length);
238	if (!junk_ptr) {
239	vfree(ptr);
240	return rv;
241	}
242
243	for (i = 0; i < junk_length; i++) {
244	ptr[i] = vmalloc(1 * PAGE_SIZE);
245	junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
246	}
247
248	for (i = 0; i < junk_length; i++)
249	vfree(junk_ptr[i]);
250
251	for (i = 0; i < test_loop_count; i++) {
252	tmp = vmalloc(1 * PAGE_SIZE);
253
254	if (!tmp)
255	goto error;
256
257	((__u8 )tmp) = 1;
258	vfree(tmp);
259	}
260
261	/* Success */
262	rv = 0;
263
264	error:
265	for (i = 0; i < junk_length; i++)
266	vfree(ptr[i]);
267
268	vfree(ptr);
269	vfree(junk_ptr);
270
271	return rv;
272	}
273
274	static int fix_size_alloc_test(void)
275	{
276	void *ptr;
277	int i;
278
279	for (i = 0; i < test_loop_count; i++) {
280	ptr = vmalloc(3 * PAGE_SIZE);
281
282	if (!ptr)
283	return -1;
284
285	((__u8 )ptr) = 0;
286
287	vfree(ptr);
288	}
289
290	return 0;
291	}
292
293	static int
294	pcpu_alloc_test(void)
295	{
296	int rv = 0;
297	#ifndef CONFIG_NEED_PER_CPU_KM
298	void __percpu **pcpu;
299	size_t size, align;
300	int i;
301
302	pcpu = vmalloc(sizeof(void __percpu ) 35000);
303	if (!pcpu)
304	return -1;
305
306	for (i = 0; i < 35000; i++) {
307	unsigned int r;
308
309	get_random_bytes(&r, sizeof(i));
310	size = (r % (PAGE_SIZE / 4)) + 1;
311
312	/*
313	* Maximum PAGE_SIZE
314	*/
315	get_random_bytes(&r, sizeof(i));
316	align = 1 << ((i % 11) + 1);
317
318	pcpu[i] = __alloc_percpu(size, align);
319	if (!pcpu[i])
320	rv = -1;
321	}
322
323	for (i = 0; i < 35000; i++)
324	free_percpu(pcpu[i]);
325
326	vfree(pcpu);
327	#endif
328	return rv;
329	}
330
331	struct test_case_desc {
332	const char *test_name;
333	int (*test_func)(void);
334	};
335
336	static struct test_case_desc test_case_array[] = {
337	{ "fix_size_alloc_test", fix_size_alloc_test },
338	{ "full_fit_alloc_test", full_fit_alloc_test },
339	{ "long_busy_list_alloc_test", long_busy_list_alloc_test },
340	{ "random_size_alloc_test", random_size_alloc_test },
341	{ "fix_align_alloc_test", fix_align_alloc_test },
342	{ "random_size_align_alloc_test", random_size_align_alloc_test },
343	{ "align_shift_alloc_test", align_shift_alloc_test },
344	{ "pcpu_alloc_test", pcpu_alloc_test },
345	/* Add a new test case here. */
346	};
347
348	struct test_case_data {
349	int test_failed;
350	int test_passed;
351	u64 time;
352	};
353
354	/* Split it to get rid of: WARNING: line over 80 characters */
355	static struct test_case_data
356	per_cpu_test_data[NR_CPUS][ARRAY_SIZE(test_case_array)];
357
358	static struct test_driver {
359	struct task_struct *task;
360	unsigned long start;
361	unsigned long stop;
362	int cpu;
363	} per_cpu_test_driver[NR_CPUS];
364
365	static void shuffle_array(int *arr, int n)
366	{
367	unsigned int rnd;
368	int i, j, x;
369
370	for (i = n - 1; i > 0; i--) {
371	get_random_bytes(&rnd, sizeof(rnd));
372
373	/* Cut the range. */
374	j = rnd % i;
375
376	/* Swap indexes. */
377	x = arr[i];
378	arr[i] = arr[j];
379	arr[j] = x;
380	}
381	}
382
383	static int test_func(void *private)
384	{
385	struct test_driver *t = private;
386	cpumask_t newmask = CPU_MASK_NONE;
387	int random_array[ARRAY_SIZE(test_case_array)];
388	int index, i, j, ret;
389	ktime_t kt;
390	u64 delta;
391
392	cpumask_set_cpu(t->cpu, &newmask);
393	set_cpus_allowed_ptr(current, &newmask);
394
395	for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
396	random_array[i] = i;
397
398	if (!sequential_test_order)
399	shuffle_array(random_array, ARRAY_SIZE(test_case_array));
400
401	/*
402	* Block until initialization is done.
403	*/
404	down_read(&prepare_for_test_rwsem);
405
406	t->start = get_cycles();
407	for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
408	index = random_array[i];
409
410	/*
411	* Skip tests if run_test_mask has been specified.
412	*/
413	if (!((run_test_mask & (1 << index)) >> index))
414	continue;
415
416	kt = ktime_get();
417	for (j = 0; j < test_repeat_count; j++) {
418	ret = test_case_array[index].test_func();
419	if (!ret)
420	per_cpu_test_data[t->cpu][index].test_passed++;
421	else
422	per_cpu_test_data[t->cpu][index].test_failed++;
423	}
424
425	/*
426	* Take an average time that test took.
427	*/
428	delta = (u64) ktime_us_delta(ktime_get(), kt);
429	do_div(delta, (u32) test_repeat_count);
430
431	per_cpu_test_data[t->cpu][index].time = delta;
432	}
433	t->stop = get_cycles();
434
435	up_read(&prepare_for_test_rwsem);
436	test_report_one_done();
437
438	/*
439	* Wait for the kthread_stop() call.
440	*/
441	while (!kthread_should_stop())
442	msleep(10);
443
444	return 0;
445	}
446
447	static void
448	init_test_configurtion(void)
449	{
450	/*
451	* Reset all data of all CPUs.
452	*/
453	memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));
454
455	if (single_cpu_test)
456	cpumask_set_cpu(cpumask_first(cpu_online_mask),
457	&cpus_run_test_mask);
458	else
459	cpumask_and(&cpus_run_test_mask, cpu_online_mask,
460	cpu_online_mask);
461
462	if (test_repeat_count <= 0)
463	test_repeat_count = 1;
464
465	if (test_loop_count <= 0)
466	test_loop_count = 1;
467	}
468
469	static void do_concurrent_test(void)
470	{
471	int cpu, ret;
472
473	/*
474	* Set some basic configurations plus sanity check.
475	*/
476	init_test_configurtion();
477
478	/*
479	* Put on hold all workers.
480	*/
481	down_write(&prepare_for_test_rwsem);
482
483	for_each_cpu(cpu, &cpus_run_test_mask) {
484	struct test_driver *t = &per_cpu_test_driver[cpu];
485
486	t->cpu = cpu;
487	t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);
488
489	if (!IS_ERR(t->task))
490	/* Success. */
491	atomic_inc(&test_n_undone);
492	else
493	pr_err("Failed to start kthread for %d CPU\n", cpu);
494	}
495
496	/*
497	* Now let the workers do their job.
498	*/
499	up_write(&prepare_for_test_rwsem);
500
501	/*
502	* Sleep quiet until all workers are done with 1 second
503	* interval. Since the test can take a lot of time we
504	* can run into a stack trace of the hung task. That is
505	* why we go with completion_timeout and HZ value.
506	*/
507	do {
508	ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
509	} while (!ret);
510
511	for_each_cpu(cpu, &cpus_run_test_mask) {
512	struct test_driver *t = &per_cpu_test_driver[cpu];
513	int i;
514
515	if (!IS_ERR(t->task))
516	kthread_stop(t->task);
517
518	for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
519	if (!((run_test_mask & (1 << i)) >> i))
520	continue;
521
522	pr_info(
523	"Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
524	test_case_array[i].test_name,
525	per_cpu_test_data[cpu][i].test_passed,
526	per_cpu_test_data[cpu][i].test_failed,
527	test_repeat_count, test_loop_count,
528	per_cpu_test_data[cpu][i].time);
529	}
530
531	pr_info("All test took CPU%d=%lu cycles\n",
532	cpu, t->stop - t->start);
533	}
534	}
535
536	static int vmalloc_test_init(void)
537	{
538	do_concurrent_test();
539	return -EAGAIN; /* Fail will directly unload the module */
540	}
541
542	static void vmalloc_test_exit(void)
543	{
544	}
545
546	module_init(vmalloc_test_init)
547	module_exit(vmalloc_test_exit)
548
549	MODULE_LICENSE("GPL");
550	MODULE_AUTHOR("Uladzislau Rezki");
551	MODULE_DESCRIPTION("vmalloc test module");