[linux-block.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>
#include <linux/rcupdate.h>
#include <linux/slab.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"
		"\t\tid: 256,  name: kvfree_rcu_1_arg_vmalloc_test\n"
		"\t\tid: 512,  name: kvfree_rcu_2_arg_vmalloc_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(size, align, GFP_KERNEL | __GFP_ZERO, 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(PAGE_SIZE, align, GFP_KERNEL|__GFP_ZERO, 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(5 * PAGE_SIZE, THREAD_ALIGN << 1,
				GFP_KERNEL | __GFP_ZERO, 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_kvfree_rcu {
	struct rcu_head rcu;
	unsigned char array[20];
};

static int
kvfree_rcu_1_arg_vmalloc_test(void)
{
	struct test_kvfree_rcu *p;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		p = vmalloc(1 * PAGE_SIZE);
		if (!p)
			return -1;

		p->array[0] = 'a';
		kvfree_rcu(p);
	}

	return 0;
}

static int
kvfree_rcu_2_arg_vmalloc_test(void)
{
	struct test_kvfree_rcu *p;
	int i;

	for (i = 0; i < test_loop_count; i++) {
		p = vmalloc(1 * PAGE_SIZE);
		if (!p)
			return -1;

		p->array[0] = 'a';
		kvfree_rcu(p, rcu);
	}

	return 0;
}

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 },
	{ "kvfree_rcu_1_arg_vmalloc_test", kvfree_rcu_1_arg_vmalloc_test },
	{ "kvfree_rcu_2_arg_vmalloc_test", kvfree_rcu_2_arg_vmalloc_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;
	int random_array[ARRAY_SIZE(test_case_array)];
	int index, i, j;
	ktime_t kt;
	u64 delta;

	if (set_cpus_allowed_ptr(current, cpumask_of(t->cpu)) < 0)
		pr_err("Failed to set affinity to %d CPU\n", t->cpu);

	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++) {
			if (!test_case_array[index].test_func())
				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>
da4fc00a URS	18	#include <linux/rcupdate.h>
da4fc00a URS	19	#include <linux/slab.h>
3f21a6b7 URS	20
	21	#define __param(type, name, init, msg) \
	22	static type name = init; \
	23	module_param(name, type, 0444); \
	24	MODULE_PARM_DESC(name, msg) \
	25
	26	__param(bool, single_cpu_test, false,
	27	"Use single first online CPU to run tests");
	28
	29	__param(bool, sequential_test_order, false,
	30	"Use sequential stress tests order");
	31
	32	__param(int, test_repeat_count, 1,
	33	"Set test repeat counter");
	34
	35	__param(int, test_loop_count, 1000000,
	36	"Set test loop counter");
	37
	38	__param(int, run_test_mask, INT_MAX,
	39	"Set tests specified in the mask.\n\n"
da4fc00a URS	40	"\t\tid: 1, name: fix_size_alloc_test\n"
	41	"\t\tid: 2, name: full_fit_alloc_test\n"
	42	"\t\tid: 4, name: long_busy_list_alloc_test\n"
	43	"\t\tid: 8, name: random_size_alloc_test\n"
	44	"\t\tid: 16, name: fix_align_alloc_test\n"
	45	"\t\tid: 32, name: random_size_align_alloc_test\n"
	46	"\t\tid: 64, name: align_shift_alloc_test\n"
	47	"\t\tid: 128, name: pcpu_alloc_test\n"
	48	"\t\tid: 256, name: kvfree_rcu_1_arg_vmalloc_test\n"
	49	"\t\tid: 512, name: kvfree_rcu_2_arg_vmalloc_test\n"
3f21a6b7 URS	50	/* Add a new test case description here. */
	51	);
	52
	53	/*
	54	* Depends on single_cpu_test parameter. If it is true, then
	55	* use first online CPU to trigger a test on, otherwise go with
	56	* all online CPUs.
	57	*/
	58	static cpumask_t cpus_run_test_mask = CPU_MASK_NONE;
	59
	60	/*
	61	* Read write semaphore for synchronization of setup
	62	* phase that is done in main thread and workers.
	63	*/
	64	static DECLARE_RWSEM(prepare_for_test_rwsem);
	65
	66	/*
	67	* Completion tracking for worker threads.
	68	*/
	69	static DECLARE_COMPLETION(test_all_done_comp);
	70	static atomic_t test_n_undone = ATOMIC_INIT(0);
	71
	72	static inline void
	73	test_report_one_done(void)
	74	{
	75	if (atomic_dec_and_test(&test_n_undone))
	76	complete(&test_all_done_comp);
	77	}
	78
	79	static int random_size_align_alloc_test(void)
	80	{
	81	unsigned long size, align, rnd;
	82	void *ptr;
	83	int i;
	84
	85	for (i = 0; i < test_loop_count; i++) {
	86	get_random_bytes(&rnd, sizeof(rnd));
	87
	88	/*
	89	* Maximum 1024 pages, if PAGE_SIZE is 4096.
	90	*/
	91	align = 1 << (rnd % 23);
	92
	93	/*
	94	* Maximum 10 pages.
	95	*/
	96	size = ((rnd % 10) + 1) * PAGE_SIZE;
	97
c3f896dc CH	98	ptr = __vmalloc_node(size, align, GFP_KERNEL \| __GFP_ZERO, 0,
c3f896dc CH	99	__builtin_return_address(0));
3f21a6b7 URS	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
c3f896dc CH	121	ptr = __vmalloc_node(PAGE_SIZE, align, GFP_KERNEL\|__GFP_ZERO, 0,
c3f896dc CH	122	__builtin_return_address(0));
3f21a6b7 URS	123	if (!ptr)
	124	return -1;
	125
	126	vfree(ptr);
	127	}
	128
	129	return 0;
	130	}
	131
	132	static int fix_align_alloc_test(void)
	133	{
	134	void *ptr;
	135	int i;
	136
	137	for (i = 0; i < test_loop_count; i++) {
c3f896dc CH	138	ptr = __vmalloc_node(5 * PAGE_SIZE, THREAD_ALIGN << 1,
	139	GFP_KERNEL \| __GFP_ZERO, 0,
	140	__builtin_return_address(0));
3f21a6b7 URS	141	if (!ptr)
	142	return -1;
	143
	144	vfree(ptr);
	145	}
	146
	147	return 0;
	148	}
	149
	150	static int random_size_alloc_test(void)
	151	{
	152	unsigned int n;
	153	void *p;
	154	int i;
	155
	156	for (i = 0; i < test_loop_count; i++) {
	157	get_random_bytes(&n, sizeof(i));
	158	n = (n % 100) + 1;
	159
	160	p = vmalloc(n * PAGE_SIZE);
	161
	162	if (!p)
	163	return -1;
	164
	165	((__u8 )p) = 1;
	166	vfree(p);
	167	}
	168
	169	return 0;
	170	}
	171
	172	static int long_busy_list_alloc_test(void)
	173	{
	174	void ptr_1, ptr_2;
	175	void **ptr;
	176	int rv = -1;
	177	int i;
	178
	179	ptr = vmalloc(sizeof(void ) 15000);
	180	if (!ptr)
	181	return rv;
	182
	183	for (i = 0; i < 15000; i++)
	184	ptr[i] = vmalloc(1 * PAGE_SIZE);
	185
	186	for (i = 0; i < test_loop_count; i++) {
	187	ptr_1 = vmalloc(100 * PAGE_SIZE);
	188	if (!ptr_1)
	189	goto leave;
	190
	191	ptr_2 = vmalloc(1 * PAGE_SIZE);
	192	if (!ptr_2) {
	193	vfree(ptr_1);
	194	goto leave;
	195	}
	196
	197	((__u8 )ptr_1) = 0;
	198	((__u8 )ptr_2) = 1;
	199
	200	vfree(ptr_1);
	201	vfree(ptr_2);
	202	}
	203
	204	/* Success */
205	rv = 0;
206
207	leave:
208	for (i = 0; i < 15000; i++)
209	vfree(ptr[i]);
210
211	vfree(ptr);
212	return rv;
213	}
214
215	static int full_fit_alloc_test(void)
216	{
217	void ptr, junk_ptr, *tmp;
218	int junk_length;
219	int rv = -1;
220	int i;
221
222	junk_length = fls(num_online_cpus());
223	junk_length = (32 1024 * 1024 / PAGE_SIZE);
224
225	ptr = vmalloc(sizeof(void ) junk_length);
226	if (!ptr)
227	return rv;
228
229	junk_ptr = vmalloc(sizeof(void ) junk_length);
230	if (!junk_ptr) {
231	vfree(ptr);
232	return rv;
233	}
234
235	for (i = 0; i < junk_length; i++) {
236	ptr[i] = vmalloc(1 * PAGE_SIZE);
237	junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
238	}
239
240	for (i = 0; i < junk_length; i++)
241	vfree(junk_ptr[i]);
242
243	for (i = 0; i < test_loop_count; i++) {
244	tmp = vmalloc(1 * PAGE_SIZE);
245
246	if (!tmp)
247	goto error;
248
249	((__u8 )tmp) = 1;
250	vfree(tmp);
251	}
252
253	/* Success */
254	rv = 0;
255
256	error:
257	for (i = 0; i < junk_length; i++)
258	vfree(ptr[i]);
259
260	vfree(ptr);
261	vfree(junk_ptr);
262
263	return rv;
264	}
265
266	static int fix_size_alloc_test(void)
267	{
268	void *ptr;
269	int i;
270
271	for (i = 0; i < test_loop_count; i++) {
272	ptr = vmalloc(3 * PAGE_SIZE);
273
274	if (!ptr)
275	return -1;
276
277	((__u8 )ptr) = 0;
278
279	vfree(ptr);
280	}
281
282	return 0;
283	}
284
285	static int
286	pcpu_alloc_test(void)
287	{
288	int rv = 0;
289	#ifndef CONFIG_NEED_PER_CPU_KM
290	void __percpu **pcpu;
291	size_t size, align;
292	int i;
293
294	pcpu = vmalloc(sizeof(void __percpu ) 35000);
295	if (!pcpu)
296	return -1;
297
298	for (i = 0; i < 35000; i++) {
299	unsigned int r;
300
301	get_random_bytes(&r, sizeof(i));
302	size = (r % (PAGE_SIZE / 4)) + 1;
303
304	/*
305	* Maximum PAGE_SIZE
306	*/
307	get_random_bytes(&r, sizeof(i));
308	align = 1 << ((i % 11) + 1);
309
310	pcpu[i] = __alloc_percpu(size, align);
311	if (!pcpu[i])
312	rv = -1;
313	}
314
315	for (i = 0; i < 35000; i++)
316	free_percpu(pcpu[i]);
317
318	vfree(pcpu);
319	#endif
320	return rv;
321	}
322
da4fc00a URS	323	struct test_kvfree_rcu {
	324	struct rcu_head rcu;
	325	unsigned char array[20];
	326	};
	327
	328	static int
	329	kvfree_rcu_1_arg_vmalloc_test(void)
	330	{
	331	struct test_kvfree_rcu *p;
	332	int i;
	333
	334	for (i = 0; i < test_loop_count; i++) {
	335	p = vmalloc(1 * PAGE_SIZE);
	336	if (!p)
	337	return -1;
	338
	339	p->array[0] = 'a';
	340	kvfree_rcu(p);
	341	}
	342
	343	return 0;
	344	}
	345
	346	static int
	347	kvfree_rcu_2_arg_vmalloc_test(void)
	348	{
	349	struct test_kvfree_rcu *p;
	350	int i;
	351
	352	for (i = 0; i < test_loop_count; i++) {
	353	p = vmalloc(1 * PAGE_SIZE);
	354	if (!p)
	355	return -1;
	356
	357	p->array[0] = 'a';
	358	kvfree_rcu(p, rcu);
	359	}
	360
	361	return 0;
	362	}
	363
3f21a6b7 URS	364	struct test_case_desc {
	365	const char *test_name;
	366	int (*test_func)(void);
	367	};
	368
	369	static struct test_case_desc test_case_array[] = {
	370	{ "fix_size_alloc_test", fix_size_alloc_test },
	371	{ "full_fit_alloc_test", full_fit_alloc_test },
	372	{ "long_busy_list_alloc_test", long_busy_list_alloc_test },
	373	{ "random_size_alloc_test", random_size_alloc_test },
	374	{ "fix_align_alloc_test", fix_align_alloc_test },
	375	{ "random_size_align_alloc_test", random_size_align_alloc_test },
	376	{ "align_shift_alloc_test", align_shift_alloc_test },
	377	{ "pcpu_alloc_test", pcpu_alloc_test },
da4fc00a URS	378	{ "kvfree_rcu_1_arg_vmalloc_test", kvfree_rcu_1_arg_vmalloc_test },
da4fc00a URS	379	{ "kvfree_rcu_2_arg_vmalloc_test", kvfree_rcu_2_arg_vmalloc_test },
3f21a6b7 URS	380	/* Add a new test case here. */
	381	};
	382
	383	struct test_case_data {
	384	int test_failed;
	385	int test_passed;
	386	u64 time;
	387	};
	388
	389	/* Split it to get rid of: WARNING: line over 80 characters */
	390	static struct test_case_data
	391	per_cpu_test_data[NR_CPUS][ARRAY_SIZE(test_case_array)];
	392
	393	static struct test_driver {
	394	struct task_struct *task;
	395	unsigned long start;
	396	unsigned long stop;
	397	int cpu;
	398	} per_cpu_test_driver[NR_CPUS];
	399
	400	static void shuffle_array(int *arr, int n)
	401	{
	402	unsigned int rnd;
	403	int i, j, x;
	404
	405	for (i = n - 1; i > 0; i--) {
	406	get_random_bytes(&rnd, sizeof(rnd));
	407
	408	/* Cut the range. */
	409	j = rnd % i;
	410
	411	/* Swap indexes. */
	412	x = arr[i];
	413	arr[i] = arr[j];
	414	arr[j] = x;
	415	}
	416	}
	417
	418	static int test_func(void *private)
	419	{
	420	struct test_driver *t = private;
3f21a6b7	421	int random_array[ARRAY_SIZE(test_case_array)];
7507c402	422	int index, i, j;
3f21a6b7 URS	423	ktime_t kt;
	424	u64 delta;
	425
7507c402	426	if (set_cpus_allowed_ptr(current, cpumask_of(t->cpu)) < 0)
e7898035	427	pr_err("Failed to set affinity to %d CPU\n", t->cpu);
3f21a6b7 URS	428
	429	for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
	430	random_array[i] = i;
	431
	432	if (!sequential_test_order)
	433	shuffle_array(random_array, ARRAY_SIZE(test_case_array));
	434
	435	/*
	436	* Block until initialization is done.
	437	*/
	438	down_read(&prepare_for_test_rwsem);
	439
	440	t->start = get_cycles();
	441	for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
	442	index = random_array[i];
	443
	444	/*
	445	* Skip tests if run_test_mask has been specified.
	446	*/
	447	if (!((run_test_mask & (1 << index)) >> index))
	448	continue;
	449
	450	kt = ktime_get();
	451	for (j = 0; j < test_repeat_count; j++) {
7507c402	452	if (!test_case_array[index].test_func())
3f21a6b7 URS	453	per_cpu_test_data[t->cpu][index].test_passed++;
	454	else
	455	per_cpu_test_data[t->cpu][index].test_failed++;
	456	}
	457
	458	/*
	459	* Take an average time that test took.
	460	*/
	461	delta = (u64) ktime_us_delta(ktime_get(), kt);
	462	do_div(delta, (u32) test_repeat_count);
	463
	464	per_cpu_test_data[t->cpu][index].time = delta;
	465	}
	466	t->stop = get_cycles();
	467
	468	up_read(&prepare_for_test_rwsem);
	469	test_report_one_done();
	470
	471	/*
	472	* Wait for the kthread_stop() call.
	473	*/
	474	while (!kthread_should_stop())
	475	msleep(10);
	476
	477	return 0;
	478	}
	479
	480	static void
	481	init_test_configurtion(void)
	482	{
	483	/*
	484	* Reset all data of all CPUs.
	485	*/
	486	memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));
	487
	488	if (single_cpu_test)
	489	cpumask_set_cpu(cpumask_first(cpu_online_mask),
	490	&cpus_run_test_mask);
	491	else
	492	cpumask_and(&cpus_run_test_mask, cpu_online_mask,
	493	cpu_online_mask);
	494
	495	if (test_repeat_count <= 0)
	496	test_repeat_count = 1;
	497
	498	if (test_loop_count <= 0)
	499	test_loop_count = 1;
	500	}
	501
	502	static void do_concurrent_test(void)
	503	{
	504	int cpu, ret;
	505
	506	/*
	507	* Set some basic configurations plus sanity check.
	508	*/
	509	init_test_configurtion();
	510
	511	/*
	512	* Put on hold all workers.
	513	*/
	514	down_write(&prepare_for_test_rwsem);
	515
	516	for_each_cpu(cpu, &cpus_run_test_mask) {
517	struct test_driver *t = &per_cpu_test_driver[cpu];
518
519	t->cpu = cpu;
520	t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);
521
522	if (!IS_ERR(t->task))
523	/* Success. */
524	atomic_inc(&test_n_undone);
525	else
526	pr_err("Failed to start kthread for %d CPU\n", cpu);
527	}
528
529	/*
530	* Now let the workers do their job.
531	*/
532	up_write(&prepare_for_test_rwsem);
533
534	/*
535	* Sleep quiet until all workers are done with 1 second
536	* interval. Since the test can take a lot of time we
537	* can run into a stack trace of the hung task. That is
538	* why we go with completion_timeout and HZ value.
539	*/
540	do {
541	ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
542	} while (!ret);
543
544	for_each_cpu(cpu, &cpus_run_test_mask) {
545	struct test_driver *t = &per_cpu_test_driver[cpu];
546	int i;
547
548	if (!IS_ERR(t->task))
549	kthread_stop(t->task);
550
551	for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
552	if (!((run_test_mask & (1 << i)) >> i))
553	continue;
554
555	pr_info(
556	"Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
557	test_case_array[i].test_name,
558	per_cpu_test_data[cpu][i].test_passed,
559	per_cpu_test_data[cpu][i].test_failed,
560	test_repeat_count, test_loop_count,
561	per_cpu_test_data[cpu][i].time);
562	}
563
564	pr_info("All test took CPU%d=%lu cycles\n",
565	cpu, t->stop - t->start);
566	}
567	}
568
569	static int vmalloc_test_init(void)
570	{
571	do_concurrent_test();
572	return -EAGAIN; /* Fail will directly unload the module */
573	}
574
575	static void vmalloc_test_exit(void)
576	{
577	}
578
579	module_init(vmalloc_test_init)
580	module_exit(vmalloc_test_exit)
581
582	MODULE_LICENSE("GPL");
583	MODULE_AUTHOR("Uladzislau Rezki");
584	MODULE_DESCRIPTION("vmalloc test module");