1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 2
3 /* 3 /*
4 * Test module for stress and analyze performa 4 * Test module for stress and analyze performance of vmalloc allocator.
5 * (C) 2018 Uladzislau Rezki (Sony) <urezki@gm 5 * (C) 2018 Uladzislau Rezki (Sony) <urezki@gmail.com>
6 */ 6 */
7 #include <linux/init.h> 7 #include <linux/init.h>
8 #include <linux/kernel.h> 8 #include <linux/kernel.h>
9 #include <linux/module.h> 9 #include <linux/module.h>
10 #include <linux/vmalloc.h> 10 #include <linux/vmalloc.h>
11 #include <linux/random.h> 11 #include <linux/random.h>
12 #include <linux/kthread.h> 12 #include <linux/kthread.h>
13 #include <linux/moduleparam.h> 13 #include <linux/moduleparam.h>
14 #include <linux/completion.h> 14 #include <linux/completion.h>
15 #include <linux/delay.h> 15 #include <linux/delay.h>
16 #include <linux/rwsem.h> 16 #include <linux/rwsem.h>
17 #include <linux/mm.h> 17 #include <linux/mm.h>
18 #include <linux/rcupdate.h> <<
19 #include <linux/slab.h> <<
20 18
21 #define __param(type, name, init, msg) 19 #define __param(type, name, init, msg) \
22 static type name = init; 20 static type name = init; \
23 module_param(name, type, 0444); 21 module_param(name, type, 0444); \
24 MODULE_PARM_DESC(name, msg) 22 MODULE_PARM_DESC(name, msg) \
25 23
26 __param(int, nr_threads, 0, !! 24 __param(bool, single_cpu_test, false,
27 "Number of workers to perform tests(mi !! 25 "Use single first online CPU to run tests");
28 26
29 __param(bool, sequential_test_order, false, 27 __param(bool, sequential_test_order, false,
30 "Use sequential stress tests order"); 28 "Use sequential stress tests order");
31 29
32 __param(int, test_repeat_count, 1, 30 __param(int, test_repeat_count, 1,
33 "Set test repeat counter"); 31 "Set test repeat counter");
34 32
35 __param(int, test_loop_count, 1000000, 33 __param(int, test_loop_count, 1000000,
36 "Set test loop counter"); 34 "Set test loop counter");
37 35
38 __param(int, nr_pages, 0, <<
39 "Set number of pages for fix_size_allo <<
40 <<
41 __param(bool, use_huge, false, <<
42 "Use vmalloc_huge in fix_size_alloc_te <<
43 <<
44 __param(int, run_test_mask, INT_MAX, 36 __param(int, run_test_mask, INT_MAX,
45 "Set tests specified in the mask.\n\n" 37 "Set tests specified in the mask.\n\n"
46 "\t\tid: 1, name: fix_size_ !! 38 "\t\tid: 1, name: fix_size_alloc_test\n"
47 "\t\tid: 2, name: full_fit_ !! 39 "\t\tid: 2, name: full_fit_alloc_test\n"
48 "\t\tid: 4, name: long_busy !! 40 "\t\tid: 4, name: long_busy_list_alloc_test\n"
49 "\t\tid: 8, name: random_si !! 41 "\t\tid: 8, name: random_size_alloc_test\n"
50 "\t\tid: 16, name: fix_align !! 42 "\t\tid: 16, name: fix_align_alloc_test\n"
51 "\t\tid: 32, name: random_si !! 43 "\t\tid: 32, name: random_size_align_alloc_test\n"
52 "\t\tid: 64, name: align_shi !! 44 "\t\tid: 64, name: align_shift_alloc_test\n"
53 "\t\tid: 128, name: pcpu_allo !! 45 "\t\tid: 128, name: pcpu_alloc_test\n"
54 "\t\tid: 256, name: kvfree_rc <<
55 "\t\tid: 512, name: kvfree_rc <<
56 "\t\tid: 1024, name: vm_map_ra <<
57 /* Add a new test case descrip 46 /* Add a new test case description here. */
58 ); 47 );
59 48
60 /* 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 /*
61 * Read write semaphore for synchronization of 57 * Read write semaphore for synchronization of setup
62 * phase that is done in main thread and worke 58 * phase that is done in main thread and workers.
63 */ 59 */
64 static DECLARE_RWSEM(prepare_for_test_rwsem); 60 static DECLARE_RWSEM(prepare_for_test_rwsem);
65 61
66 /* 62 /*
67 * Completion tracking for worker threads. 63 * Completion tracking for worker threads.
68 */ 64 */
69 static DECLARE_COMPLETION(test_all_done_comp); 65 static DECLARE_COMPLETION(test_all_done_comp);
70 static atomic_t test_n_undone = ATOMIC_INIT(0) 66 static atomic_t test_n_undone = ATOMIC_INIT(0);
71 67
72 static inline void 68 static inline void
73 test_report_one_done(void) 69 test_report_one_done(void)
74 { 70 {
75 if (atomic_dec_and_test(&test_n_undone 71 if (atomic_dec_and_test(&test_n_undone))
76 complete(&test_all_done_comp); 72 complete(&test_all_done_comp);
77 } 73 }
78 74
79 static int random_size_align_alloc_test(void) 75 static int random_size_align_alloc_test(void)
80 { 76 {
81 unsigned long size, align; !! 77 unsigned long size, align, rnd;
82 unsigned int rnd; <<
83 void *ptr; 78 void *ptr;
84 int i; 79 int i;
85 80
86 for (i = 0; i < test_loop_count; i++) 81 for (i = 0; i < test_loop_count; i++) {
87 rnd = get_random_u8(); !! 82 get_random_bytes(&rnd, sizeof(rnd));
88 83
89 /* 84 /*
90 * Maximum 1024 pages, if PAGE 85 * Maximum 1024 pages, if PAGE_SIZE is 4096.
91 */ 86 */
92 align = 1 << (rnd % 23); 87 align = 1 << (rnd % 23);
93 88
94 /* 89 /*
95 * Maximum 10 pages. 90 * Maximum 10 pages.
96 */ 91 */
97 size = ((rnd % 10) + 1) * PAGE 92 size = ((rnd % 10) + 1) * PAGE_SIZE;
98 93
99 ptr = __vmalloc_node(size, ali 94 ptr = __vmalloc_node(size, align, GFP_KERNEL | __GFP_ZERO, 0,
100 __builtin_retu 95 __builtin_return_address(0));
101 if (!ptr) 96 if (!ptr)
102 return -1; 97 return -1;
103 98
104 vfree(ptr); 99 vfree(ptr);
105 } 100 }
106 101
107 return 0; 102 return 0;
108 } 103 }
109 104
110 /* 105 /*
111 * This test case is supposed to be failed. 106 * This test case is supposed to be failed.
112 */ 107 */
113 static int align_shift_alloc_test(void) 108 static int align_shift_alloc_test(void)
114 { 109 {
115 unsigned long align; 110 unsigned long align;
116 void *ptr; 111 void *ptr;
117 int i; 112 int i;
118 113
119 for (i = 0; i < BITS_PER_LONG; i++) { 114 for (i = 0; i < BITS_PER_LONG; i++) {
120 align = 1UL << i; !! 115 align = ((unsigned long) 1) << i;
121 116
122 ptr = __vmalloc_node(PAGE_SIZE 117 ptr = __vmalloc_node(PAGE_SIZE, align, GFP_KERNEL|__GFP_ZERO, 0,
123 __builtin_retu 118 __builtin_return_address(0));
124 if (!ptr) 119 if (!ptr)
125 return -1; 120 return -1;
126 121
127 vfree(ptr); 122 vfree(ptr);
128 } 123 }
129 124
130 return 0; 125 return 0;
131 } 126 }
132 127
133 static int fix_align_alloc_test(void) 128 static int fix_align_alloc_test(void)
134 { 129 {
135 void *ptr; 130 void *ptr;
136 int i; 131 int i;
137 132
138 for (i = 0; i < test_loop_count; i++) 133 for (i = 0; i < test_loop_count; i++) {
139 ptr = __vmalloc_node(5 * PAGE_ 134 ptr = __vmalloc_node(5 * PAGE_SIZE, THREAD_ALIGN << 1,
140 GFP_KERNEL | _ 135 GFP_KERNEL | __GFP_ZERO, 0,
141 __builtin_retu 136 __builtin_return_address(0));
142 if (!ptr) 137 if (!ptr)
143 return -1; 138 return -1;
144 139
145 vfree(ptr); 140 vfree(ptr);
146 } 141 }
147 142
148 return 0; 143 return 0;
149 } 144 }
150 145
151 static int random_size_alloc_test(void) 146 static int random_size_alloc_test(void)
152 { 147 {
153 unsigned int n; 148 unsigned int n;
154 void *p; 149 void *p;
155 int i; 150 int i;
156 151
157 for (i = 0; i < test_loop_count; i++) 152 for (i = 0; i < test_loop_count; i++) {
158 n = get_random_u32_inclusive(1 !! 153 get_random_bytes(&n, sizeof(i));
>> 154 n = (n % 100) + 1;
>> 155
159 p = vmalloc(n * PAGE_SIZE); 156 p = vmalloc(n * PAGE_SIZE);
160 157
161 if (!p) 158 if (!p)
162 return -1; 159 return -1;
163 160
164 *((__u8 *)p) = 1; 161 *((__u8 *)p) = 1;
165 vfree(p); 162 vfree(p);
166 } 163 }
167 164
168 return 0; 165 return 0;
169 } 166 }
170 167
171 static int long_busy_list_alloc_test(void) 168 static int long_busy_list_alloc_test(void)
172 { 169 {
173 void *ptr_1, *ptr_2; 170 void *ptr_1, *ptr_2;
174 void **ptr; 171 void **ptr;
175 int rv = -1; 172 int rv = -1;
176 int i; 173 int i;
177 174
178 ptr = vmalloc(sizeof(void *) * 15000); 175 ptr = vmalloc(sizeof(void *) * 15000);
179 if (!ptr) 176 if (!ptr)
180 return rv; 177 return rv;
181 178
182 for (i = 0; i < 15000; i++) 179 for (i = 0; i < 15000; i++)
183 ptr[i] = vmalloc(1 * PAGE_SIZE 180 ptr[i] = vmalloc(1 * PAGE_SIZE);
184 181
185 for (i = 0; i < test_loop_count; i++) 182 for (i = 0; i < test_loop_count; i++) {
186 ptr_1 = vmalloc(100 * PAGE_SIZ 183 ptr_1 = vmalloc(100 * PAGE_SIZE);
187 if (!ptr_1) 184 if (!ptr_1)
188 goto leave; 185 goto leave;
189 186
190 ptr_2 = vmalloc(1 * PAGE_SIZE) 187 ptr_2 = vmalloc(1 * PAGE_SIZE);
191 if (!ptr_2) { 188 if (!ptr_2) {
192 vfree(ptr_1); 189 vfree(ptr_1);
193 goto leave; 190 goto leave;
194 } 191 }
195 192
196 *((__u8 *)ptr_1) = 0; 193 *((__u8 *)ptr_1) = 0;
197 *((__u8 *)ptr_2) = 1; 194 *((__u8 *)ptr_2) = 1;
198 195
199 vfree(ptr_1); 196 vfree(ptr_1);
200 vfree(ptr_2); 197 vfree(ptr_2);
201 } 198 }
202 199
203 /* Success */ 200 /* Success */
204 rv = 0; 201 rv = 0;
205 202
206 leave: 203 leave:
207 for (i = 0; i < 15000; i++) 204 for (i = 0; i < 15000; i++)
208 vfree(ptr[i]); 205 vfree(ptr[i]);
209 206
210 vfree(ptr); 207 vfree(ptr);
211 return rv; 208 return rv;
212 } 209 }
213 210
214 static int full_fit_alloc_test(void) 211 static int full_fit_alloc_test(void)
215 { 212 {
216 void **ptr, **junk_ptr, *tmp; 213 void **ptr, **junk_ptr, *tmp;
217 int junk_length; 214 int junk_length;
218 int rv = -1; 215 int rv = -1;
219 int i; 216 int i;
220 217
221 junk_length = fls(num_online_cpus()); 218 junk_length = fls(num_online_cpus());
222 junk_length *= (32 * 1024 * 1024 / PAG 219 junk_length *= (32 * 1024 * 1024 / PAGE_SIZE);
223 220
224 ptr = vmalloc(sizeof(void *) * junk_le 221 ptr = vmalloc(sizeof(void *) * junk_length);
225 if (!ptr) 222 if (!ptr)
226 return rv; 223 return rv;
227 224
228 junk_ptr = vmalloc(sizeof(void *) * ju 225 junk_ptr = vmalloc(sizeof(void *) * junk_length);
229 if (!junk_ptr) { 226 if (!junk_ptr) {
230 vfree(ptr); 227 vfree(ptr);
231 return rv; 228 return rv;
232 } 229 }
233 230
234 for (i = 0; i < junk_length; i++) { 231 for (i = 0; i < junk_length; i++) {
235 ptr[i] = vmalloc(1 * PAGE_SIZE 232 ptr[i] = vmalloc(1 * PAGE_SIZE);
236 junk_ptr[i] = vmalloc(1 * PAGE 233 junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
237 } 234 }
238 235
239 for (i = 0; i < junk_length; i++) 236 for (i = 0; i < junk_length; i++)
240 vfree(junk_ptr[i]); 237 vfree(junk_ptr[i]);
241 238
242 for (i = 0; i < test_loop_count; i++) 239 for (i = 0; i < test_loop_count; i++) {
243 tmp = vmalloc(1 * PAGE_SIZE); 240 tmp = vmalloc(1 * PAGE_SIZE);
244 241
245 if (!tmp) 242 if (!tmp)
246 goto error; 243 goto error;
247 244
248 *((__u8 *)tmp) = 1; 245 *((__u8 *)tmp) = 1;
249 vfree(tmp); 246 vfree(tmp);
250 } 247 }
251 248
252 /* Success */ 249 /* Success */
253 rv = 0; 250 rv = 0;
254 251
255 error: 252 error:
256 for (i = 0; i < junk_length; i++) 253 for (i = 0; i < junk_length; i++)
257 vfree(ptr[i]); 254 vfree(ptr[i]);
258 255
259 vfree(ptr); 256 vfree(ptr);
260 vfree(junk_ptr); 257 vfree(junk_ptr);
261 258
262 return rv; 259 return rv;
263 } 260 }
264 261
265 static int fix_size_alloc_test(void) 262 static int fix_size_alloc_test(void)
266 { 263 {
267 void *ptr; 264 void *ptr;
268 int i; 265 int i;
269 266
270 for (i = 0; i < test_loop_count; i++) 267 for (i = 0; i < test_loop_count; i++) {
271 if (use_huge) !! 268 ptr = vmalloc(3 * PAGE_SIZE);
272 ptr = vmalloc_huge((nr <<
273 else <<
274 ptr = vmalloc((nr_page <<
275 269
276 if (!ptr) 270 if (!ptr)
277 return -1; 271 return -1;
278 272
279 *((__u8 *)ptr) = 0; 273 *((__u8 *)ptr) = 0;
280 274
281 vfree(ptr); 275 vfree(ptr);
282 } 276 }
283 277
284 return 0; 278 return 0;
285 } 279 }
286 280
287 static int 281 static int
288 pcpu_alloc_test(void) 282 pcpu_alloc_test(void)
289 { 283 {
290 int rv = 0; 284 int rv = 0;
291 #ifndef CONFIG_NEED_PER_CPU_KM 285 #ifndef CONFIG_NEED_PER_CPU_KM
292 void __percpu **pcpu; 286 void __percpu **pcpu;
293 size_t size, align; 287 size_t size, align;
294 int i; 288 int i;
295 289
296 pcpu = vmalloc(sizeof(void __percpu *) 290 pcpu = vmalloc(sizeof(void __percpu *) * 35000);
297 if (!pcpu) 291 if (!pcpu)
298 return -1; 292 return -1;
299 293
300 for (i = 0; i < 35000; i++) { 294 for (i = 0; i < 35000; i++) {
301 size = get_random_u32_inclusiv !! 295 unsigned int r;
>> 296
>> 297 get_random_bytes(&r, sizeof(i));
>> 298 size = (r % (PAGE_SIZE / 4)) + 1;
302 299
303 /* 300 /*
304 * Maximum PAGE_SIZE 301 * Maximum PAGE_SIZE
305 */ 302 */
306 align = 1 << get_random_u32_in !! 303 get_random_bytes(&r, sizeof(i));
>> 304 align = 1 << ((i % 11) + 1);
307 305
308 pcpu[i] = __alloc_percpu(size, 306 pcpu[i] = __alloc_percpu(size, align);
309 if (!pcpu[i]) 307 if (!pcpu[i])
310 rv = -1; 308 rv = -1;
311 } 309 }
312 310
313 for (i = 0; i < 35000; i++) 311 for (i = 0; i < 35000; i++)
314 free_percpu(pcpu[i]); 312 free_percpu(pcpu[i]);
315 313
316 vfree(pcpu); 314 vfree(pcpu);
317 #endif 315 #endif
318 return rv; 316 return rv;
319 } 317 }
320 318
321 struct test_kvfree_rcu { <<
322 struct rcu_head rcu; <<
323 unsigned char array[20]; <<
324 }; <<
325 <<
326 static int <<
327 kvfree_rcu_1_arg_vmalloc_test(void) <<
328 { <<
329 struct test_kvfree_rcu *p; <<
330 int i; <<
331 <<
332 for (i = 0; i < test_loop_count; i++) <<
333 p = vmalloc(1 * PAGE_SIZE); <<
334 if (!p) <<
335 return -1; <<
336 <<
337 p->array[0] = 'a'; <<
338 kvfree_rcu_mightsleep(p); <<
339 } <<
340 <<
341 return 0; <<
342 } <<
343 <<
344 static int <<
345 kvfree_rcu_2_arg_vmalloc_test(void) <<
346 { <<
347 struct test_kvfree_rcu *p; <<
348 int i; <<
349 <<
350 for (i = 0; i < test_loop_count; i++) <<
351 p = vmalloc(1 * PAGE_SIZE); <<
352 if (!p) <<
353 return -1; <<
354 <<
355 p->array[0] = 'a'; <<
356 kvfree_rcu(p, rcu); <<
357 } <<
358 <<
359 return 0; <<
360 } <<
361 <<
362 static int <<
363 vm_map_ram_test(void) <<
364 { <<
365 unsigned long nr_allocated; <<
366 unsigned int map_nr_pages; <<
367 unsigned char *v_ptr; <<
368 struct page **pages; <<
369 int i; <<
370 <<
371 map_nr_pages = nr_pages > 0 ? nr_pages <<
372 pages = kcalloc(map_nr_pages, sizeof(s <<
373 if (!pages) <<
374 return -1; <<
375 <<
376 nr_allocated = alloc_pages_bulk_array( <<
377 if (nr_allocated != map_nr_pages) <<
378 goto cleanup; <<
379 <<
380 /* Run the test loop. */ <<
381 for (i = 0; i < test_loop_count; i++) <<
382 v_ptr = vm_map_ram(pages, map_ <<
383 *v_ptr = 'a'; <<
384 vm_unmap_ram(v_ptr, map_nr_pag <<
385 } <<
386 <<
387 cleanup: <<
388 for (i = 0; i < nr_allocated; i++) <<
389 __free_page(pages[i]); <<
390 <<
391 kfree(pages); <<
392 <<
393 /* 0 indicates success. */ <<
394 return nr_allocated != map_nr_pages; <<
395 } <<
396 <<
397 struct test_case_desc { 319 struct test_case_desc {
398 const char *test_name; 320 const char *test_name;
399 int (*test_func)(void); 321 int (*test_func)(void);
400 }; 322 };
401 323
402 static struct test_case_desc test_case_array[] 324 static struct test_case_desc test_case_array[] = {
403 { "fix_size_alloc_test", fix_size_allo 325 { "fix_size_alloc_test", fix_size_alloc_test },
404 { "full_fit_alloc_test", full_fit_allo 326 { "full_fit_alloc_test", full_fit_alloc_test },
405 { "long_busy_list_alloc_test", long_bu 327 { "long_busy_list_alloc_test", long_busy_list_alloc_test },
406 { "random_size_alloc_test", random_siz 328 { "random_size_alloc_test", random_size_alloc_test },
407 { "fix_align_alloc_test", fix_align_al 329 { "fix_align_alloc_test", fix_align_alloc_test },
408 { "random_size_align_alloc_test", rand 330 { "random_size_align_alloc_test", random_size_align_alloc_test },
409 { "align_shift_alloc_test", align_shif 331 { "align_shift_alloc_test", align_shift_alloc_test },
410 { "pcpu_alloc_test", pcpu_alloc_test } 332 { "pcpu_alloc_test", pcpu_alloc_test },
411 { "kvfree_rcu_1_arg_vmalloc_test", kvf <<
412 { "kvfree_rcu_2_arg_vmalloc_test", kvf <<
413 { "vm_map_ram_test", vm_map_ram_test } <<
414 /* Add a new test case here. */ 333 /* Add a new test case here. */
415 }; 334 };
416 335
417 struct test_case_data { 336 struct test_case_data {
418 int test_failed; 337 int test_failed;
419 int test_passed; 338 int test_passed;
420 u64 time; 339 u64 time;
421 }; 340 };
422 341
>> 342 /* Split it to get rid of: WARNING: line over 80 characters */
>> 343 static struct test_case_data
>> 344 per_cpu_test_data[NR_CPUS][ARRAY_SIZE(test_case_array)];
>> 345
423 static struct test_driver { 346 static struct test_driver {
424 struct task_struct *task; 347 struct task_struct *task;
425 struct test_case_data data[ARRAY_SIZE( <<
426 <<
427 unsigned long start; 348 unsigned long start;
428 unsigned long stop; 349 unsigned long stop;
429 } *tdriver; !! 350 int cpu;
>> 351 } per_cpu_test_driver[NR_CPUS];
430 352
431 static void shuffle_array(int *arr, int n) 353 static void shuffle_array(int *arr, int n)
432 { 354 {
433 int i, j; !! 355 unsigned int rnd;
>> 356 int i, j, x;
434 357
435 for (i = n - 1; i > 0; i--) { 358 for (i = n - 1; i > 0; i--) {
>> 359 get_random_bytes(&rnd, sizeof(rnd));
>> 360
436 /* Cut the range. */ 361 /* Cut the range. */
437 j = get_random_u32_below(i); !! 362 j = rnd % i;
438 363
439 /* Swap indexes. */ 364 /* Swap indexes. */
440 swap(arr[i], arr[j]); !! 365 x = arr[i];
>> 366 arr[i] = arr[j];
>> 367 arr[j] = x;
441 } 368 }
442 } 369 }
443 370
444 static int test_func(void *private) 371 static int test_func(void *private)
445 { 372 {
446 struct test_driver *t = private; 373 struct test_driver *t = private;
447 int random_array[ARRAY_SIZE(test_case_ 374 int random_array[ARRAY_SIZE(test_case_array)];
448 int index, i, j; 375 int index, i, j;
449 ktime_t kt; 376 ktime_t kt;
450 u64 delta; 377 u64 delta;
451 378
>> 379 if (set_cpus_allowed_ptr(current, cpumask_of(t->cpu)) < 0)
>> 380 pr_err("Failed to set affinity to %d CPU\n", t->cpu);
>> 381
452 for (i = 0; i < ARRAY_SIZE(test_case_a 382 for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
453 random_array[i] = i; 383 random_array[i] = i;
454 384
455 if (!sequential_test_order) 385 if (!sequential_test_order)
456 shuffle_array(random_array, AR 386 shuffle_array(random_array, ARRAY_SIZE(test_case_array));
457 387
458 /* 388 /*
459 * Block until initialization is done. 389 * Block until initialization is done.
460 */ 390 */
461 down_read(&prepare_for_test_rwsem); 391 down_read(&prepare_for_test_rwsem);
462 392
463 t->start = get_cycles(); 393 t->start = get_cycles();
464 for (i = 0; i < ARRAY_SIZE(test_case_a 394 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
465 index = random_array[i]; 395 index = random_array[i];
466 396
467 /* 397 /*
468 * Skip tests if run_test_mask 398 * Skip tests if run_test_mask has been specified.
469 */ 399 */
470 if (!((run_test_mask & (1 << i 400 if (!((run_test_mask & (1 << index)) >> index))
471 continue; 401 continue;
472 402
473 kt = ktime_get(); 403 kt = ktime_get();
474 for (j = 0; j < test_repeat_co 404 for (j = 0; j < test_repeat_count; j++) {
475 if (!test_case_array[i 405 if (!test_case_array[index].test_func())
476 t->data[index] !! 406 per_cpu_test_data[t->cpu][index].test_passed++;
477 else 407 else
478 t->data[index] !! 408 per_cpu_test_data[t->cpu][index].test_failed++;
479 } 409 }
480 410
481 /* 411 /*
482 * Take an average time that t 412 * Take an average time that test took.
483 */ 413 */
484 delta = (u64) ktime_us_delta(k 414 delta = (u64) ktime_us_delta(ktime_get(), kt);
485 do_div(delta, (u32) test_repea 415 do_div(delta, (u32) test_repeat_count);
486 416
487 t->data[index].time = delta; !! 417 per_cpu_test_data[t->cpu][index].time = delta;
488 } 418 }
489 t->stop = get_cycles(); 419 t->stop = get_cycles();
490 420
491 up_read(&prepare_for_test_rwsem); 421 up_read(&prepare_for_test_rwsem);
492 test_report_one_done(); 422 test_report_one_done();
493 423
494 /* 424 /*
495 * Wait for the kthread_stop() call. 425 * Wait for the kthread_stop() call.
496 */ 426 */
497 while (!kthread_should_stop()) 427 while (!kthread_should_stop())
498 msleep(10); 428 msleep(10);
499 429
500 return 0; 430 return 0;
501 } 431 }
502 432
503 static int !! 433 static void
504 init_test_configuration(void) !! 434 init_test_configurtion(void)
505 { 435 {
506 /* 436 /*
507 * A maximum number of workers is defi !! 437 * Reset all data of all CPUs.
508 * value and set to USHRT_MAX. We add <<
509 * case and for potential heavy stress <<
510 */ 438 */
511 nr_threads = clamp(nr_threads, 1, (int !! 439 memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));
512 440
513 /* Allocate the space for test instanc !! 441 if (single_cpu_test)
514 tdriver = kvcalloc(nr_threads, sizeof( !! 442 cpumask_set_cpu(cpumask_first(cpu_online_mask),
515 if (tdriver == NULL) !! 443 &cpus_run_test_mask);
516 return -1; !! 444 else
>> 445 cpumask_and(&cpus_run_test_mask, cpu_online_mask,
>> 446 cpu_online_mask);
517 447
518 if (test_repeat_count <= 0) 448 if (test_repeat_count <= 0)
519 test_repeat_count = 1; 449 test_repeat_count = 1;
520 450
521 if (test_loop_count <= 0) 451 if (test_loop_count <= 0)
522 test_loop_count = 1; 452 test_loop_count = 1;
523 <<
524 return 0; <<
525 } 453 }
526 454
527 static void do_concurrent_test(void) 455 static void do_concurrent_test(void)
528 { 456 {
529 int i, ret; !! 457 int cpu, ret;
530 458
531 /* 459 /*
532 * Set some basic configurations plus 460 * Set some basic configurations plus sanity check.
533 */ 461 */
534 ret = init_test_configuration(); !! 462 init_test_configurtion();
535 if (ret < 0) <<
536 return; <<
537 463
538 /* 464 /*
539 * Put on hold all workers. 465 * Put on hold all workers.
540 */ 466 */
541 down_write(&prepare_for_test_rwsem); 467 down_write(&prepare_for_test_rwsem);
542 468
543 for (i = 0; i < nr_threads; i++) { !! 469 for_each_cpu(cpu, &cpus_run_test_mask) {
544 struct test_driver *t = &tdriv !! 470 struct test_driver *t = &per_cpu_test_driver[cpu];
545 471
546 t->task = kthread_run(test_fun !! 472 t->cpu = cpu;
>> 473 t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);
547 474
548 if (!IS_ERR(t->task)) 475 if (!IS_ERR(t->task))
549 /* Success. */ 476 /* Success. */
550 atomic_inc(&test_n_und 477 atomic_inc(&test_n_undone);
551 else 478 else
552 pr_err("Failed to star !! 479 pr_err("Failed to start kthread for %d CPU\n", cpu);
553 } 480 }
554 481
555 /* 482 /*
556 * Now let the workers do their job. 483 * Now let the workers do their job.
557 */ 484 */
558 up_write(&prepare_for_test_rwsem); 485 up_write(&prepare_for_test_rwsem);
559 486
560 /* 487 /*
561 * Sleep quiet until all workers are d 488 * Sleep quiet until all workers are done with 1 second
562 * interval. Since the test can take a 489 * interval. Since the test can take a lot of time we
563 * can run into a stack trace of the h 490 * can run into a stack trace of the hung task. That is
564 * why we go with completion_timeout a 491 * why we go with completion_timeout and HZ value.
565 */ 492 */
566 do { 493 do {
567 ret = wait_for_completion_time 494 ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
568 } while (!ret); 495 } while (!ret);
569 496
570 for (i = 0; i < nr_threads; i++) { !! 497 for_each_cpu(cpu, &cpus_run_test_mask) {
571 struct test_driver *t = &tdriv !! 498 struct test_driver *t = &per_cpu_test_driver[cpu];
572 int j; !! 499 int i;
573 500
574 if (!IS_ERR(t->task)) 501 if (!IS_ERR(t->task))
575 kthread_stop(t->task); 502 kthread_stop(t->task);
576 503
577 for (j = 0; j < ARRAY_SIZE(tes !! 504 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
578 if (!((run_test_mask & !! 505 if (!((run_test_mask & (1 << i)) >> i))
579 continue; 506 continue;
580 507
581 pr_info( 508 pr_info(
582 "Summary: %s p 509 "Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
583 test_case_arra !! 510 test_case_array[i].test_name,
584 t->data[j].tes !! 511 per_cpu_test_data[cpu][i].test_passed,
585 t->data[j].tes !! 512 per_cpu_test_data[cpu][i].test_failed,
586 test_repeat_co 513 test_repeat_count, test_loop_count,
587 t->data[j].tim !! 514 per_cpu_test_data[cpu][i].time);
588 } 515 }
589 516
590 pr_info("All test took worker% !! 517 pr_info("All test took CPU%d=%lu cycles\n",
591 i, t->stop - t->start) !! 518 cpu, t->stop - t->start);
592 } 519 }
593 <<
594 kvfree(tdriver); <<
595 } 520 }
596 521
597 static int vmalloc_test_init(void) 522 static int vmalloc_test_init(void)
598 { 523 {
599 do_concurrent_test(); 524 do_concurrent_test();
600 return -EAGAIN; /* Fail will directly 525 return -EAGAIN; /* Fail will directly unload the module */
601 } 526 }
602 527
>> 528 static void vmalloc_test_exit(void)
>> 529 {
>> 530 }
>> 531
603 module_init(vmalloc_test_init) 532 module_init(vmalloc_test_init)
>> 533 module_exit(vmalloc_test_exit)
604 534
605 MODULE_LICENSE("GPL"); 535 MODULE_LICENSE("GPL");
606 MODULE_AUTHOR("Uladzislau Rezki"); 536 MODULE_AUTHOR("Uladzislau Rezki");
607 MODULE_DESCRIPTION("vmalloc test module"); 537 MODULE_DESCRIPTION("vmalloc test module");
608 538
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.