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_range(size, align,
100 __builtin_retu !! 95 VMALLOC_START, VMALLOC_END,
>> 96 GFP_KERNEL | __GFP_ZERO,
>> 97 PAGE_KERNEL,
>> 98 0, 0, __builtin_return_address(0));
>> 99
101 if (!ptr) 100 if (!ptr)
102 return -1; 101 return -1;
103 102
104 vfree(ptr); 103 vfree(ptr);
105 } 104 }
106 105
107 return 0; 106 return 0;
108 } 107 }
109 108
110 /* 109 /*
111 * This test case is supposed to be failed. 110 * This test case is supposed to be failed.
112 */ 111 */
113 static int align_shift_alloc_test(void) 112 static int align_shift_alloc_test(void)
114 { 113 {
115 unsigned long align; 114 unsigned long align;
116 void *ptr; 115 void *ptr;
117 int i; 116 int i;
118 117
119 for (i = 0; i < BITS_PER_LONG; i++) { 118 for (i = 0; i < BITS_PER_LONG; i++) {
120 align = 1UL << 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));
121 126
122 ptr = __vmalloc_node(PAGE_SIZE <<
123 __builtin_retu <<
124 if (!ptr) 127 if (!ptr)
125 return -1; 128 return -1;
126 129
127 vfree(ptr); 130 vfree(ptr);
128 } 131 }
129 132
130 return 0; 133 return 0;
131 } 134 }
132 135
133 static int fix_align_alloc_test(void) 136 static int fix_align_alloc_test(void)
134 { 137 {
135 void *ptr; 138 void *ptr;
136 int i; 139 int i;
137 140
138 for (i = 0; i < test_loop_count; i++) 141 for (i = 0; i < test_loop_count; i++) {
139 ptr = __vmalloc_node(5 * PAGE_ !! 142 ptr = __vmalloc_node_range(5 * PAGE_SIZE,
140 GFP_KERNEL | _ !! 143 THREAD_ALIGN << 1,
141 __builtin_retu !! 144 VMALLOC_START, VMALLOC_END,
>> 145 GFP_KERNEL | __GFP_ZERO,
>> 146 PAGE_KERNEL,
>> 147 0, 0, __builtin_return_address(0));
>> 148
142 if (!ptr) 149 if (!ptr)
143 return -1; 150 return -1;
144 151
145 vfree(ptr); 152 vfree(ptr);
146 } 153 }
147 154
148 return 0; 155 return 0;
149 } 156 }
150 157
151 static int random_size_alloc_test(void) 158 static int random_size_alloc_test(void)
152 { 159 {
153 unsigned int n; 160 unsigned int n;
154 void *p; 161 void *p;
155 int i; 162 int i;
156 163
157 for (i = 0; i < test_loop_count; i++) 164 for (i = 0; i < test_loop_count; i++) {
158 n = get_random_u32_inclusive(1 !! 165 get_random_bytes(&n, sizeof(i));
>> 166 n = (n % 100) + 1;
>> 167
159 p = vmalloc(n * PAGE_SIZE); 168 p = vmalloc(n * PAGE_SIZE);
160 169
161 if (!p) 170 if (!p)
162 return -1; 171 return -1;
163 172
164 *((__u8 *)p) = 1; 173 *((__u8 *)p) = 1;
165 vfree(p); 174 vfree(p);
166 } 175 }
167 176
168 return 0; 177 return 0;
169 } 178 }
170 179
171 static int long_busy_list_alloc_test(void) 180 static int long_busy_list_alloc_test(void)
172 { 181 {
173 void *ptr_1, *ptr_2; 182 void *ptr_1, *ptr_2;
174 void **ptr; 183 void **ptr;
175 int rv = -1; 184 int rv = -1;
176 int i; 185 int i;
177 186
178 ptr = vmalloc(sizeof(void *) * 15000); 187 ptr = vmalloc(sizeof(void *) * 15000);
179 if (!ptr) 188 if (!ptr)
180 return rv; 189 return rv;
181 190
182 for (i = 0; i < 15000; i++) 191 for (i = 0; i < 15000; i++)
183 ptr[i] = vmalloc(1 * PAGE_SIZE 192 ptr[i] = vmalloc(1 * PAGE_SIZE);
184 193
185 for (i = 0; i < test_loop_count; i++) 194 for (i = 0; i < test_loop_count; i++) {
186 ptr_1 = vmalloc(100 * PAGE_SIZ 195 ptr_1 = vmalloc(100 * PAGE_SIZE);
187 if (!ptr_1) 196 if (!ptr_1)
188 goto leave; 197 goto leave;
189 198
190 ptr_2 = vmalloc(1 * PAGE_SIZE) 199 ptr_2 = vmalloc(1 * PAGE_SIZE);
191 if (!ptr_2) { 200 if (!ptr_2) {
192 vfree(ptr_1); 201 vfree(ptr_1);
193 goto leave; 202 goto leave;
194 } 203 }
195 204
196 *((__u8 *)ptr_1) = 0; 205 *((__u8 *)ptr_1) = 0;
197 *((__u8 *)ptr_2) = 1; 206 *((__u8 *)ptr_2) = 1;
198 207
199 vfree(ptr_1); 208 vfree(ptr_1);
200 vfree(ptr_2); 209 vfree(ptr_2);
201 } 210 }
202 211
203 /* Success */ 212 /* Success */
204 rv = 0; 213 rv = 0;
205 214
206 leave: 215 leave:
207 for (i = 0; i < 15000; i++) 216 for (i = 0; i < 15000; i++)
208 vfree(ptr[i]); 217 vfree(ptr[i]);
209 218
210 vfree(ptr); 219 vfree(ptr);
211 return rv; 220 return rv;
212 } 221 }
213 222
214 static int full_fit_alloc_test(void) 223 static int full_fit_alloc_test(void)
215 { 224 {
216 void **ptr, **junk_ptr, *tmp; 225 void **ptr, **junk_ptr, *tmp;
217 int junk_length; 226 int junk_length;
218 int rv = -1; 227 int rv = -1;
219 int i; 228 int i;
220 229
221 junk_length = fls(num_online_cpus()); 230 junk_length = fls(num_online_cpus());
222 junk_length *= (32 * 1024 * 1024 / PAG 231 junk_length *= (32 * 1024 * 1024 / PAGE_SIZE);
223 232
224 ptr = vmalloc(sizeof(void *) * junk_le 233 ptr = vmalloc(sizeof(void *) * junk_length);
225 if (!ptr) 234 if (!ptr)
226 return rv; 235 return rv;
227 236
228 junk_ptr = vmalloc(sizeof(void *) * ju 237 junk_ptr = vmalloc(sizeof(void *) * junk_length);
229 if (!junk_ptr) { 238 if (!junk_ptr) {
230 vfree(ptr); 239 vfree(ptr);
231 return rv; 240 return rv;
232 } 241 }
233 242
234 for (i = 0; i < junk_length; i++) { 243 for (i = 0; i < junk_length; i++) {
235 ptr[i] = vmalloc(1 * PAGE_SIZE 244 ptr[i] = vmalloc(1 * PAGE_SIZE);
236 junk_ptr[i] = vmalloc(1 * PAGE 245 junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
237 } 246 }
238 247
239 for (i = 0; i < junk_length; i++) 248 for (i = 0; i < junk_length; i++)
240 vfree(junk_ptr[i]); 249 vfree(junk_ptr[i]);
241 250
242 for (i = 0; i < test_loop_count; i++) 251 for (i = 0; i < test_loop_count; i++) {
243 tmp = vmalloc(1 * PAGE_SIZE); 252 tmp = vmalloc(1 * PAGE_SIZE);
244 253
245 if (!tmp) 254 if (!tmp)
246 goto error; 255 goto error;
247 256
248 *((__u8 *)tmp) = 1; 257 *((__u8 *)tmp) = 1;
249 vfree(tmp); 258 vfree(tmp);
250 } 259 }
251 260
252 /* Success */ 261 /* Success */
253 rv = 0; 262 rv = 0;
254 263
255 error: 264 error:
256 for (i = 0; i < junk_length; i++) 265 for (i = 0; i < junk_length; i++)
257 vfree(ptr[i]); 266 vfree(ptr[i]);
258 267
259 vfree(ptr); 268 vfree(ptr);
260 vfree(junk_ptr); 269 vfree(junk_ptr);
261 270
262 return rv; 271 return rv;
263 } 272 }
264 273
265 static int fix_size_alloc_test(void) 274 static int fix_size_alloc_test(void)
266 { 275 {
267 void *ptr; 276 void *ptr;
268 int i; 277 int i;
269 278
270 for (i = 0; i < test_loop_count; i++) 279 for (i = 0; i < test_loop_count; i++) {
271 if (use_huge) !! 280 ptr = vmalloc(3 * PAGE_SIZE);
272 ptr = vmalloc_huge((nr <<
273 else <<
274 ptr = vmalloc((nr_page <<
275 281
276 if (!ptr) 282 if (!ptr)
277 return -1; 283 return -1;
278 284
279 *((__u8 *)ptr) = 0; 285 *((__u8 *)ptr) = 0;
280 286
281 vfree(ptr); 287 vfree(ptr);
282 } 288 }
283 289
284 return 0; 290 return 0;
285 } 291 }
286 292
287 static int 293 static int
288 pcpu_alloc_test(void) 294 pcpu_alloc_test(void)
289 { 295 {
290 int rv = 0; 296 int rv = 0;
291 #ifndef CONFIG_NEED_PER_CPU_KM 297 #ifndef CONFIG_NEED_PER_CPU_KM
292 void __percpu **pcpu; 298 void __percpu **pcpu;
293 size_t size, align; 299 size_t size, align;
294 int i; 300 int i;
295 301
296 pcpu = vmalloc(sizeof(void __percpu *) 302 pcpu = vmalloc(sizeof(void __percpu *) * 35000);
297 if (!pcpu) 303 if (!pcpu)
298 return -1; 304 return -1;
299 305
300 for (i = 0; i < 35000; i++) { 306 for (i = 0; i < 35000; i++) {
301 size = get_random_u32_inclusiv !! 307 unsigned int r;
>> 308
>> 309 get_random_bytes(&r, sizeof(i));
>> 310 size = (r % (PAGE_SIZE / 4)) + 1;
302 311
303 /* 312 /*
304 * Maximum PAGE_SIZE 313 * Maximum PAGE_SIZE
305 */ 314 */
306 align = 1 << get_random_u32_in !! 315 get_random_bytes(&r, sizeof(i));
>> 316 align = 1 << ((i % 11) + 1);
307 317
308 pcpu[i] = __alloc_percpu(size, 318 pcpu[i] = __alloc_percpu(size, align);
309 if (!pcpu[i]) 319 if (!pcpu[i])
310 rv = -1; 320 rv = -1;
311 } 321 }
312 322
313 for (i = 0; i < 35000; i++) 323 for (i = 0; i < 35000; i++)
314 free_percpu(pcpu[i]); 324 free_percpu(pcpu[i]);
315 325
316 vfree(pcpu); 326 vfree(pcpu);
317 #endif 327 #endif
318 return rv; 328 return rv;
319 } 329 }
320 330
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 { 331 struct test_case_desc {
398 const char *test_name; 332 const char *test_name;
399 int (*test_func)(void); 333 int (*test_func)(void);
400 }; 334 };
401 335
402 static struct test_case_desc test_case_array[] 336 static struct test_case_desc test_case_array[] = {
403 { "fix_size_alloc_test", fix_size_allo 337 { "fix_size_alloc_test", fix_size_alloc_test },
404 { "full_fit_alloc_test", full_fit_allo 338 { "full_fit_alloc_test", full_fit_alloc_test },
405 { "long_busy_list_alloc_test", long_bu 339 { "long_busy_list_alloc_test", long_busy_list_alloc_test },
406 { "random_size_alloc_test", random_siz 340 { "random_size_alloc_test", random_size_alloc_test },
407 { "fix_align_alloc_test", fix_align_al 341 { "fix_align_alloc_test", fix_align_alloc_test },
408 { "random_size_align_alloc_test", rand 342 { "random_size_align_alloc_test", random_size_align_alloc_test },
409 { "align_shift_alloc_test", align_shif 343 { "align_shift_alloc_test", align_shift_alloc_test },
410 { "pcpu_alloc_test", pcpu_alloc_test } 344 { "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. */ 345 /* Add a new test case here. */
415 }; 346 };
416 347
417 struct test_case_data { 348 struct test_case_data {
418 int test_failed; 349 int test_failed;
419 int test_passed; 350 int test_passed;
420 u64 time; 351 u64 time;
421 }; 352 };
422 353
>> 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
423 static struct test_driver { 358 static struct test_driver {
424 struct task_struct *task; 359 struct task_struct *task;
425 struct test_case_data data[ARRAY_SIZE( <<
426 <<
427 unsigned long start; 360 unsigned long start;
428 unsigned long stop; 361 unsigned long stop;
429 } *tdriver; !! 362 int cpu;
>> 363 } per_cpu_test_driver[NR_CPUS];
430 364
431 static void shuffle_array(int *arr, int n) 365 static void shuffle_array(int *arr, int n)
432 { 366 {
433 int i, j; !! 367 unsigned int rnd;
>> 368 int i, j, x;
434 369
435 for (i = n - 1; i > 0; i--) { 370 for (i = n - 1; i > 0; i--) {
>> 371 get_random_bytes(&rnd, sizeof(rnd));
>> 372
436 /* Cut the range. */ 373 /* Cut the range. */
437 j = get_random_u32_below(i); !! 374 j = rnd % i;
438 375
439 /* Swap indexes. */ 376 /* Swap indexes. */
440 swap(arr[i], arr[j]); !! 377 x = arr[i];
>> 378 arr[i] = arr[j];
>> 379 arr[j] = x;
441 } 380 }
442 } 381 }
443 382
444 static int test_func(void *private) 383 static int test_func(void *private)
445 { 384 {
446 struct test_driver *t = private; 385 struct test_driver *t = private;
447 int random_array[ARRAY_SIZE(test_case_ 386 int random_array[ARRAY_SIZE(test_case_array)];
448 int index, i, j; 387 int index, i, j;
449 ktime_t kt; 388 ktime_t kt;
450 u64 delta; 389 u64 delta;
451 390
>> 391 if (set_cpus_allowed_ptr(current, cpumask_of(t->cpu)) < 0)
>> 392 pr_err("Failed to set affinity to %d CPU\n", t->cpu);
>> 393
452 for (i = 0; i < ARRAY_SIZE(test_case_a 394 for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
453 random_array[i] = i; 395 random_array[i] = i;
454 396
455 if (!sequential_test_order) 397 if (!sequential_test_order)
456 shuffle_array(random_array, AR 398 shuffle_array(random_array, ARRAY_SIZE(test_case_array));
457 399
458 /* 400 /*
459 * Block until initialization is done. 401 * Block until initialization is done.
460 */ 402 */
461 down_read(&prepare_for_test_rwsem); 403 down_read(&prepare_for_test_rwsem);
462 404
463 t->start = get_cycles(); 405 t->start = get_cycles();
464 for (i = 0; i < ARRAY_SIZE(test_case_a 406 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
465 index = random_array[i]; 407 index = random_array[i];
466 408
467 /* 409 /*
468 * Skip tests if run_test_mask 410 * Skip tests if run_test_mask has been specified.
469 */ 411 */
470 if (!((run_test_mask & (1 << i 412 if (!((run_test_mask & (1 << index)) >> index))
471 continue; 413 continue;
472 414
473 kt = ktime_get(); 415 kt = ktime_get();
474 for (j = 0; j < test_repeat_co 416 for (j = 0; j < test_repeat_count; j++) {
475 if (!test_case_array[i 417 if (!test_case_array[index].test_func())
476 t->data[index] !! 418 per_cpu_test_data[t->cpu][index].test_passed++;
477 else 419 else
478 t->data[index] !! 420 per_cpu_test_data[t->cpu][index].test_failed++;
479 } 421 }
480 422
481 /* 423 /*
482 * Take an average time that t 424 * Take an average time that test took.
483 */ 425 */
484 delta = (u64) ktime_us_delta(k 426 delta = (u64) ktime_us_delta(ktime_get(), kt);
485 do_div(delta, (u32) test_repea 427 do_div(delta, (u32) test_repeat_count);
486 428
487 t->data[index].time = delta; !! 429 per_cpu_test_data[t->cpu][index].time = delta;
488 } 430 }
489 t->stop = get_cycles(); 431 t->stop = get_cycles();
490 432
491 up_read(&prepare_for_test_rwsem); 433 up_read(&prepare_for_test_rwsem);
492 test_report_one_done(); 434 test_report_one_done();
493 435
494 /* 436 /*
495 * Wait for the kthread_stop() call. 437 * Wait for the kthread_stop() call.
496 */ 438 */
497 while (!kthread_should_stop()) 439 while (!kthread_should_stop())
498 msleep(10); 440 msleep(10);
499 441
500 return 0; 442 return 0;
501 } 443 }
502 444
503 static int !! 445 static void
504 init_test_configuration(void) !! 446 init_test_configurtion(void)
505 { 447 {
506 /* 448 /*
507 * A maximum number of workers is defi !! 449 * Reset all data of all CPUs.
508 * value and set to USHRT_MAX. We add <<
509 * case and for potential heavy stress <<
510 */ 450 */
511 nr_threads = clamp(nr_threads, 1, (int !! 451 memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));
512 452
513 /* Allocate the space for test instanc !! 453 if (single_cpu_test)
514 tdriver = kvcalloc(nr_threads, sizeof( !! 454 cpumask_set_cpu(cpumask_first(cpu_online_mask),
515 if (tdriver == NULL) !! 455 &cpus_run_test_mask);
516 return -1; !! 456 else
>> 457 cpumask_and(&cpus_run_test_mask, cpu_online_mask,
>> 458 cpu_online_mask);
517 459
518 if (test_repeat_count <= 0) 460 if (test_repeat_count <= 0)
519 test_repeat_count = 1; 461 test_repeat_count = 1;
520 462
521 if (test_loop_count <= 0) 463 if (test_loop_count <= 0)
522 test_loop_count = 1; 464 test_loop_count = 1;
523 <<
524 return 0; <<
525 } 465 }
526 466
527 static void do_concurrent_test(void) 467 static void do_concurrent_test(void)
528 { 468 {
529 int i, ret; !! 469 int cpu, ret;
530 470
531 /* 471 /*
532 * Set some basic configurations plus 472 * Set some basic configurations plus sanity check.
533 */ 473 */
534 ret = init_test_configuration(); !! 474 init_test_configurtion();
535 if (ret < 0) <<
536 return; <<
537 475
538 /* 476 /*
539 * Put on hold all workers. 477 * Put on hold all workers.
540 */ 478 */
541 down_write(&prepare_for_test_rwsem); 479 down_write(&prepare_for_test_rwsem);
542 480
543 for (i = 0; i < nr_threads; i++) { !! 481 for_each_cpu(cpu, &cpus_run_test_mask) {
544 struct test_driver *t = &tdriv !! 482 struct test_driver *t = &per_cpu_test_driver[cpu];
545 483
546 t->task = kthread_run(test_fun !! 484 t->cpu = cpu;
>> 485 t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);
547 486
548 if (!IS_ERR(t->task)) 487 if (!IS_ERR(t->task))
549 /* Success. */ 488 /* Success. */
550 atomic_inc(&test_n_und 489 atomic_inc(&test_n_undone);
551 else 490 else
552 pr_err("Failed to star !! 491 pr_err("Failed to start kthread for %d CPU\n", cpu);
553 } 492 }
554 493
555 /* 494 /*
556 * Now let the workers do their job. 495 * Now let the workers do their job.
557 */ 496 */
558 up_write(&prepare_for_test_rwsem); 497 up_write(&prepare_for_test_rwsem);
559 498
560 /* 499 /*
561 * Sleep quiet until all workers are d 500 * Sleep quiet until all workers are done with 1 second
562 * interval. Since the test can take a 501 * interval. Since the test can take a lot of time we
563 * can run into a stack trace of the h 502 * can run into a stack trace of the hung task. That is
564 * why we go with completion_timeout a 503 * why we go with completion_timeout and HZ value.
565 */ 504 */
566 do { 505 do {
567 ret = wait_for_completion_time 506 ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
568 } while (!ret); 507 } while (!ret);
569 508
570 for (i = 0; i < nr_threads; i++) { !! 509 for_each_cpu(cpu, &cpus_run_test_mask) {
571 struct test_driver *t = &tdriv !! 510 struct test_driver *t = &per_cpu_test_driver[cpu];
572 int j; !! 511 int i;
573 512
574 if (!IS_ERR(t->task)) 513 if (!IS_ERR(t->task))
575 kthread_stop(t->task); 514 kthread_stop(t->task);
576 515
577 for (j = 0; j < ARRAY_SIZE(tes !! 516 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
578 if (!((run_test_mask & !! 517 if (!((run_test_mask & (1 << i)) >> i))
579 continue; 518 continue;
580 519
581 pr_info( 520 pr_info(
582 "Summary: %s p 521 "Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
583 test_case_arra !! 522 test_case_array[i].test_name,
584 t->data[j].tes !! 523 per_cpu_test_data[cpu][i].test_passed,
585 t->data[j].tes !! 524 per_cpu_test_data[cpu][i].test_failed,
586 test_repeat_co 525 test_repeat_count, test_loop_count,
587 t->data[j].tim !! 526 per_cpu_test_data[cpu][i].time);
588 } 527 }
589 528
590 pr_info("All test took worker% !! 529 pr_info("All test took CPU%d=%lu cycles\n",
591 i, t->stop - t->start) !! 530 cpu, t->stop - t->start);
592 } 531 }
593 <<
594 kvfree(tdriver); <<
595 } 532 }
596 533
597 static int vmalloc_test_init(void) 534 static int vmalloc_test_init(void)
598 { 535 {
599 do_concurrent_test(); 536 do_concurrent_test();
600 return -EAGAIN; /* Fail will directly 537 return -EAGAIN; /* Fail will directly unload the module */
601 } 538 }
602 539
>> 540 static void vmalloc_test_exit(void)
>> 541 {
>> 542 }
>> 543
603 module_init(vmalloc_test_init) 544 module_init(vmalloc_test_init)
>> 545 module_exit(vmalloc_test_exit)
604 546
605 MODULE_LICENSE("GPL"); 547 MODULE_LICENSE("GPL");
606 MODULE_AUTHOR("Uladzislau Rezki"); 548 MODULE_AUTHOR("Uladzislau Rezki");
607 MODULE_DESCRIPTION("vmalloc test module"); 549 MODULE_DESCRIPTION("vmalloc test module");
608 550
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.