summaryrefslogtreecommitdiff
path: root/tools/testing/selftests/vm/hugetlb-madvise.c
blob: a634f47d1e56314c681ea69e3940f52951becdd9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
// SPDX-License-Identifier: GPL-2.0
/*
 * hugepage-madvise:
 *
 * Basic functional testing of madvise MADV_DONTNEED and MADV_REMOVE
 * on hugetlb mappings.
 *
 * Before running this test, make sure the administrator has pre-allocated
 * at least MIN_FREE_PAGES hugetlb pages and they are free.  In addition,
 * the test takes an argument that is the path to a file in a hugetlbfs
 * filesystem.  Therefore, a hugetlbfs filesystem must be mounted on some
 * directory.
 */

#define _GNU_SOURCE
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/mman.h>
#define __USE_GNU
#include <fcntl.h>

#define MIN_FREE_PAGES	20
#define NR_HUGE_PAGES	10	/* common number of pages to map/allocate */

#define validate_free_pages(exp_free)					\
	do {								\
		int fhp = get_free_hugepages();				\
		if (fhp != (exp_free)) {				\
			printf("Unexpected number of free huge "	\
				"pages line %d\n", __LINE__);		\
			exit(1);					\
		}							\
	} while (0)

unsigned long huge_page_size;
unsigned long base_page_size;

/*
 * default_huge_page_size copied from mlock2-tests.c
 */
unsigned long default_huge_page_size(void)
{
	unsigned long hps = 0;
	char *line = NULL;
	size_t linelen = 0;
	FILE *f = fopen("/proc/meminfo", "r");

	if (!f)
		return 0;
	while (getline(&line, &linelen, f) > 0) {
		if (sscanf(line, "Hugepagesize:       %lu kB", &hps) == 1) {
			hps <<= 10;
			break;
		}
	}

	free(line);
	fclose(f);
	return hps;
}

unsigned long get_free_hugepages(void)
{
	unsigned long fhp = 0;
	char *line = NULL;
	size_t linelen = 0;
	FILE *f = fopen("/proc/meminfo", "r");

	if (!f)
		return fhp;
	while (getline(&line, &linelen, f) > 0) {
		if (sscanf(line, "HugePages_Free:      %lu", &fhp) == 1)
			break;
	}

	free(line);
	fclose(f);
	return fhp;
}

void write_fault_pages(void *addr, unsigned long nr_pages)
{
	unsigned long i;

	for (i = 0; i < nr_pages; i++)
		*((unsigned long *)(addr + (i * huge_page_size))) = i;
}

void read_fault_pages(void *addr, unsigned long nr_pages)
{
	unsigned long dummy = 0;
	unsigned long i;

	for (i = 0; i < nr_pages; i++)
		dummy += *((unsigned long *)(addr + (i * huge_page_size)));
}

int main(int argc, char **argv)
{
	unsigned long free_hugepages;
	void *addr, *addr2;
	int fd;
	int ret;

	huge_page_size = default_huge_page_size();
	if (!huge_page_size) {
		printf("Unable to determine huge page size, exiting!\n");
		exit(1);
	}
	base_page_size = sysconf(_SC_PAGE_SIZE);
	if (!huge_page_size) {
		printf("Unable to determine base page size, exiting!\n");
		exit(1);
	}

	free_hugepages = get_free_hugepages();
	if (free_hugepages < MIN_FREE_PAGES) {
		printf("Not enough free huge pages to test, exiting!\n");
		exit(1);
	}

	fd = memfd_create(argv[0], MFD_HUGETLB);
	if (fd < 0) {
		perror("memfd_create() failed");
		exit(1);
	}

	/*
	 * Test validity of MADV_DONTNEED addr and length arguments.  mmap
	 * size is NR_HUGE_PAGES + 2.  One page at the beginning and end of
	 * the mapping will be unmapped so we KNOW there is nothing mapped
	 * there.
	 */
	addr = mmap(NULL, (NR_HUGE_PAGES + 2) * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB,
			-1, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}
	if (munmap(addr, huge_page_size) ||
			munmap(addr + (NR_HUGE_PAGES + 1) * huge_page_size,
				huge_page_size)) {
		perror("munmap");
		exit(1);
	}
	addr = addr + huge_page_size;

	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* addr before mapping should fail */
	ret = madvise(addr - base_page_size, NR_HUGE_PAGES * huge_page_size,
		MADV_DONTNEED);
	if (!ret) {
		printf("Unexpected success of madvise call with invalid addr line %d\n",
				__LINE__);
			exit(1);
	}

	/* addr + length after mapping should fail */
	ret = madvise(addr, (NR_HUGE_PAGES * huge_page_size) + base_page_size,
		MADV_DONTNEED);
	if (!ret) {
		printf("Unexpected success of madvise call with invalid length line %d\n",
				__LINE__);
			exit(1);
	}

	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);

	/*
	 * Test alignment of MADV_DONTNEED addr and length arguments
	 */
	addr = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB,
			-1, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* addr is not huge page size aligned and should fail */
	ret = madvise(addr + base_page_size,
			NR_HUGE_PAGES * huge_page_size - base_page_size,
			MADV_DONTNEED);
	if (!ret) {
		printf("Unexpected success of madvise call with unaligned start address %d\n",
				__LINE__);
			exit(1);
	}

	/* addr + length should be aligned down to huge page size */
	if (madvise(addr,
			((NR_HUGE_PAGES - 1) * huge_page_size) + base_page_size,
			MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}

	/* should free all but last page in mapping */
	validate_free_pages(free_hugepages - 1);

	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);
	validate_free_pages(free_hugepages);

	/*
	 * Test MADV_DONTNEED on anonymous private mapping
	 */
	addr = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB,
			-1, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}

	/* should free all pages in mapping */
	validate_free_pages(free_hugepages);

	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);

	/*
	 * Test MADV_DONTNEED on private mapping of hugetlb file
	 */
	if (fallocate(fd, 0, 0, NR_HUGE_PAGES * huge_page_size)) {
		perror("fallocate");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	addr = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_PRIVATE, fd, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	/* read should not consume any pages */
	read_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* madvise should not free any pages */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* writes should allocate private pages */
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - (2 * NR_HUGE_PAGES));

	/* madvise should free private pages */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* writes should allocate private pages */
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - (2 * NR_HUGE_PAGES));

	/*
	 * The fallocate below certainly should free the pages associated
	 * with the file.  However, pages in the private mapping are also
	 * freed.  This is not the 'correct' behavior, but is expected
	 * because this is how it has worked since the initial hugetlb
	 * implementation.
	 */
	if (fallocate(fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
					0, NR_HUGE_PAGES * huge_page_size)) {
		perror("fallocate");
		exit(1);
	}
	validate_free_pages(free_hugepages);

	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);

	/*
	 * Test MADV_DONTNEED on shared mapping of hugetlb file
	 */
	if (fallocate(fd, 0, 0, NR_HUGE_PAGES * huge_page_size)) {
		perror("fallocate");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	addr = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_SHARED, fd, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	/* write should not consume any pages */
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* madvise should not free any pages */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/*
	 * Test MADV_REMOVE on shared mapping of hugetlb file
	 *
	 * madvise is same as hole punch and should free all pages.
	 */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_REMOVE)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages);
	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);

	/*
	 * Test MADV_REMOVE on shared and private mapping of hugetlb file
	 */
	if (fallocate(fd, 0, 0, NR_HUGE_PAGES * huge_page_size)) {
		perror("fallocate");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	addr = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_SHARED, fd, 0);
	if (addr == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	/* shared write should not consume any additional pages */
	write_fault_pages(addr, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	addr2 = mmap(NULL, NR_HUGE_PAGES * huge_page_size,
			PROT_READ | PROT_WRITE,
			MAP_PRIVATE, fd, 0);
	if (addr2 == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	/* private read should not consume any pages */
	read_fault_pages(addr2, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* private write should consume additional pages */
	write_fault_pages(addr2, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - (2 * NR_HUGE_PAGES));

	/* madvise of shared mapping should not free any pages */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages - (2 * NR_HUGE_PAGES));

	/* madvise of private mapping should free private pages */
	if (madvise(addr2, NR_HUGE_PAGES * huge_page_size, MADV_DONTNEED)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages - NR_HUGE_PAGES);

	/* private write should consume additional pages again */
	write_fault_pages(addr2, NR_HUGE_PAGES);
	validate_free_pages(free_hugepages - (2 * NR_HUGE_PAGES));

	/*
	 * madvise should free both file and private pages although this is
	 * not correct.  private pages should not be freed, but this is
	 * expected.  See comment associated with FALLOC_FL_PUNCH_HOLE call.
	 */
	if (madvise(addr, NR_HUGE_PAGES * huge_page_size, MADV_REMOVE)) {
		perror("madvise");
		exit(1);
	}
	validate_free_pages(free_hugepages);

	(void)munmap(addr, NR_HUGE_PAGES * huge_page_size);
	(void)munmap(addr2, NR_HUGE_PAGES * huge_page_size);

	close(fd);
	return 0;
}