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> 18 #include <linux/rcupdate.h>
19 #include <linux/slab.h> 19 #include <linux/slab.h>
20 20
21 #define __param(type, name, init, msg) 21 #define __param(type, name, init, msg) \
22 static type name = init; 22 static type name = init; \
23 module_param(name, type, 0444); 23 module_param(name, type, 0444); \
24 MODULE_PARM_DESC(name, msg) 24 MODULE_PARM_DESC(name, msg) \
25 25
26 __param(int, nr_threads, 0, !! 26 __param(bool, single_cpu_test, false,
27 "Number of workers to perform tests(mi !! 27 "Use single first online CPU to run tests");
28 28
29 __param(bool, sequential_test_order, false, 29 __param(bool, sequential_test_order, false,
30 "Use sequential stress tests order"); 30 "Use sequential stress tests order");
31 31
32 __param(int, test_repeat_count, 1, 32 __param(int, test_repeat_count, 1,
33 "Set test repeat counter"); 33 "Set test repeat counter");
34 34
35 __param(int, test_loop_count, 1000000, 35 __param(int, test_loop_count, 1000000,
36 "Set test loop counter"); 36 "Set test loop counter");
37 37
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, 38 __param(int, run_test_mask, INT_MAX,
45 "Set tests specified in the mask.\n\n" 39 "Set tests specified in the mask.\n\n"
46 "\t\tid: 1, name: fix_size_ 40 "\t\tid: 1, name: fix_size_alloc_test\n"
47 "\t\tid: 2, name: full_fit_ 41 "\t\tid: 2, name: full_fit_alloc_test\n"
48 "\t\tid: 4, name: long_busy 42 "\t\tid: 4, name: long_busy_list_alloc_test\n"
49 "\t\tid: 8, name: random_si 43 "\t\tid: 8, name: random_size_alloc_test\n"
50 "\t\tid: 16, name: fix_align 44 "\t\tid: 16, name: fix_align_alloc_test\n"
51 "\t\tid: 32, name: random_si 45 "\t\tid: 32, name: random_size_align_alloc_test\n"
52 "\t\tid: 64, name: align_shi 46 "\t\tid: 64, name: align_shift_alloc_test\n"
53 "\t\tid: 128, name: pcpu_allo 47 "\t\tid: 128, name: pcpu_alloc_test\n"
54 "\t\tid: 256, name: kvfree_rc 48 "\t\tid: 256, name: kvfree_rcu_1_arg_vmalloc_test\n"
55 "\t\tid: 512, name: kvfree_rc 49 "\t\tid: 512, name: kvfree_rcu_2_arg_vmalloc_test\n"
56 "\t\tid: 1024, name: vm_map_ra !! 50 "\t\tid: 1024, name: kvfree_rcu_1_arg_slab_test\n"
>> 51 "\t\tid: 2048, name: kvfree_rcu_2_arg_slab_test\n"
57 /* Add a new test case descrip 52 /* Add a new test case description here. */
58 ); 53 );
59 54
60 /* 55 /*
>> 56 * Depends on single_cpu_test parameter. If it is true, then
>> 57 * use first online CPU to trigger a test on, otherwise go with
>> 58 * all online CPUs.
>> 59 */
>> 60 static cpumask_t cpus_run_test_mask = CPU_MASK_NONE;
>> 61
>> 62 /*
61 * Read write semaphore for synchronization of 63 * Read write semaphore for synchronization of setup
62 * phase that is done in main thread and worke 64 * phase that is done in main thread and workers.
63 */ 65 */
64 static DECLARE_RWSEM(prepare_for_test_rwsem); 66 static DECLARE_RWSEM(prepare_for_test_rwsem);
65 67
66 /* 68 /*
67 * Completion tracking for worker threads. 69 * Completion tracking for worker threads.
68 */ 70 */
69 static DECLARE_COMPLETION(test_all_done_comp); 71 static DECLARE_COMPLETION(test_all_done_comp);
70 static atomic_t test_n_undone = ATOMIC_INIT(0) 72 static atomic_t test_n_undone = ATOMIC_INIT(0);
71 73
72 static inline void 74 static inline void
73 test_report_one_done(void) 75 test_report_one_done(void)
74 { 76 {
75 if (atomic_dec_and_test(&test_n_undone 77 if (atomic_dec_and_test(&test_n_undone))
76 complete(&test_all_done_comp); 78 complete(&test_all_done_comp);
77 } 79 }
78 80
79 static int random_size_align_alloc_test(void) 81 static int random_size_align_alloc_test(void)
80 { 82 {
81 unsigned long size, align; !! 83 unsigned long size, align, rnd;
82 unsigned int rnd; <<
83 void *ptr; 84 void *ptr;
84 int i; 85 int i;
85 86
86 for (i = 0; i < test_loop_count; i++) 87 for (i = 0; i < test_loop_count; i++) {
87 rnd = get_random_u8(); !! 88 get_random_bytes(&rnd, sizeof(rnd));
88 89
89 /* 90 /*
90 * Maximum 1024 pages, if PAGE 91 * Maximum 1024 pages, if PAGE_SIZE is 4096.
91 */ 92 */
92 align = 1 << (rnd % 23); 93 align = 1 << (rnd % 23);
93 94
94 /* 95 /*
95 * Maximum 10 pages. 96 * Maximum 10 pages.
96 */ 97 */
97 size = ((rnd % 10) + 1) * PAGE 98 size = ((rnd % 10) + 1) * PAGE_SIZE;
98 99
99 ptr = __vmalloc_node(size, ali 100 ptr = __vmalloc_node(size, align, GFP_KERNEL | __GFP_ZERO, 0,
100 __builtin_retu 101 __builtin_return_address(0));
101 if (!ptr) 102 if (!ptr)
102 return -1; 103 return -1;
103 104
104 vfree(ptr); 105 vfree(ptr);
105 } 106 }
106 107
107 return 0; 108 return 0;
108 } 109 }
109 110
110 /* 111 /*
111 * This test case is supposed to be failed. 112 * This test case is supposed to be failed.
112 */ 113 */
113 static int align_shift_alloc_test(void) 114 static int align_shift_alloc_test(void)
114 { 115 {
115 unsigned long align; 116 unsigned long align;
116 void *ptr; 117 void *ptr;
117 int i; 118 int i;
118 119
119 for (i = 0; i < BITS_PER_LONG; i++) { 120 for (i = 0; i < BITS_PER_LONG; i++) {
120 align = 1UL << i; !! 121 align = ((unsigned long) 1) << i;
121 122
122 ptr = __vmalloc_node(PAGE_SIZE 123 ptr = __vmalloc_node(PAGE_SIZE, align, GFP_KERNEL|__GFP_ZERO, 0,
123 __builtin_retu 124 __builtin_return_address(0));
124 if (!ptr) 125 if (!ptr)
125 return -1; 126 return -1;
126 127
127 vfree(ptr); 128 vfree(ptr);
128 } 129 }
129 130
130 return 0; 131 return 0;
131 } 132 }
132 133
133 static int fix_align_alloc_test(void) 134 static int fix_align_alloc_test(void)
134 { 135 {
135 void *ptr; 136 void *ptr;
136 int i; 137 int i;
137 138
138 for (i = 0; i < test_loop_count; i++) 139 for (i = 0; i < test_loop_count; i++) {
139 ptr = __vmalloc_node(5 * PAGE_ 140 ptr = __vmalloc_node(5 * PAGE_SIZE, THREAD_ALIGN << 1,
140 GFP_KERNEL | _ 141 GFP_KERNEL | __GFP_ZERO, 0,
141 __builtin_retu 142 __builtin_return_address(0));
142 if (!ptr) 143 if (!ptr)
143 return -1; 144 return -1;
144 145
145 vfree(ptr); 146 vfree(ptr);
146 } 147 }
147 148
148 return 0; 149 return 0;
149 } 150 }
150 151
151 static int random_size_alloc_test(void) 152 static int random_size_alloc_test(void)
152 { 153 {
153 unsigned int n; 154 unsigned int n;
154 void *p; 155 void *p;
155 int i; 156 int i;
156 157
157 for (i = 0; i < test_loop_count; i++) 158 for (i = 0; i < test_loop_count; i++) {
158 n = get_random_u32_inclusive(1 !! 159 get_random_bytes(&n, sizeof(i));
>> 160 n = (n % 100) + 1;
>> 161
159 p = vmalloc(n * PAGE_SIZE); 162 p = vmalloc(n * PAGE_SIZE);
160 163
161 if (!p) 164 if (!p)
162 return -1; 165 return -1;
163 166
164 *((__u8 *)p) = 1; 167 *((__u8 *)p) = 1;
165 vfree(p); 168 vfree(p);
166 } 169 }
167 170
168 return 0; 171 return 0;
169 } 172 }
170 173
171 static int long_busy_list_alloc_test(void) 174 static int long_busy_list_alloc_test(void)
172 { 175 {
173 void *ptr_1, *ptr_2; 176 void *ptr_1, *ptr_2;
174 void **ptr; 177 void **ptr;
175 int rv = -1; 178 int rv = -1;
176 int i; 179 int i;
177 180
178 ptr = vmalloc(sizeof(void *) * 15000); 181 ptr = vmalloc(sizeof(void *) * 15000);
179 if (!ptr) 182 if (!ptr)
180 return rv; 183 return rv;
181 184
182 for (i = 0; i < 15000; i++) 185 for (i = 0; i < 15000; i++)
183 ptr[i] = vmalloc(1 * PAGE_SIZE 186 ptr[i] = vmalloc(1 * PAGE_SIZE);
184 187
185 for (i = 0; i < test_loop_count; i++) 188 for (i = 0; i < test_loop_count; i++) {
186 ptr_1 = vmalloc(100 * PAGE_SIZ 189 ptr_1 = vmalloc(100 * PAGE_SIZE);
187 if (!ptr_1) 190 if (!ptr_1)
188 goto leave; 191 goto leave;
189 192
190 ptr_2 = vmalloc(1 * PAGE_SIZE) 193 ptr_2 = vmalloc(1 * PAGE_SIZE);
191 if (!ptr_2) { 194 if (!ptr_2) {
192 vfree(ptr_1); 195 vfree(ptr_1);
193 goto leave; 196 goto leave;
194 } 197 }
195 198
196 *((__u8 *)ptr_1) = 0; 199 *((__u8 *)ptr_1) = 0;
197 *((__u8 *)ptr_2) = 1; 200 *((__u8 *)ptr_2) = 1;
198 201
199 vfree(ptr_1); 202 vfree(ptr_1);
200 vfree(ptr_2); 203 vfree(ptr_2);
201 } 204 }
202 205
203 /* Success */ 206 /* Success */
204 rv = 0; 207 rv = 0;
205 208
206 leave: 209 leave:
207 for (i = 0; i < 15000; i++) 210 for (i = 0; i < 15000; i++)
208 vfree(ptr[i]); 211 vfree(ptr[i]);
209 212
210 vfree(ptr); 213 vfree(ptr);
211 return rv; 214 return rv;
212 } 215 }
213 216
214 static int full_fit_alloc_test(void) 217 static int full_fit_alloc_test(void)
215 { 218 {
216 void **ptr, **junk_ptr, *tmp; 219 void **ptr, **junk_ptr, *tmp;
217 int junk_length; 220 int junk_length;
218 int rv = -1; 221 int rv = -1;
219 int i; 222 int i;
220 223
221 junk_length = fls(num_online_cpus()); 224 junk_length = fls(num_online_cpus());
222 junk_length *= (32 * 1024 * 1024 / PAG 225 junk_length *= (32 * 1024 * 1024 / PAGE_SIZE);
223 226
224 ptr = vmalloc(sizeof(void *) * junk_le 227 ptr = vmalloc(sizeof(void *) * junk_length);
225 if (!ptr) 228 if (!ptr)
226 return rv; 229 return rv;
227 230
228 junk_ptr = vmalloc(sizeof(void *) * ju 231 junk_ptr = vmalloc(sizeof(void *) * junk_length);
229 if (!junk_ptr) { 232 if (!junk_ptr) {
230 vfree(ptr); 233 vfree(ptr);
231 return rv; 234 return rv;
232 } 235 }
233 236
234 for (i = 0; i < junk_length; i++) { 237 for (i = 0; i < junk_length; i++) {
235 ptr[i] = vmalloc(1 * PAGE_SIZE 238 ptr[i] = vmalloc(1 * PAGE_SIZE);
236 junk_ptr[i] = vmalloc(1 * PAGE 239 junk_ptr[i] = vmalloc(1 * PAGE_SIZE);
237 } 240 }
238 241
239 for (i = 0; i < junk_length; i++) 242 for (i = 0; i < junk_length; i++)
240 vfree(junk_ptr[i]); 243 vfree(junk_ptr[i]);
241 244
242 for (i = 0; i < test_loop_count; i++) 245 for (i = 0; i < test_loop_count; i++) {
243 tmp = vmalloc(1 * PAGE_SIZE); 246 tmp = vmalloc(1 * PAGE_SIZE);
244 247
245 if (!tmp) 248 if (!tmp)
246 goto error; 249 goto error;
247 250
248 *((__u8 *)tmp) = 1; 251 *((__u8 *)tmp) = 1;
249 vfree(tmp); 252 vfree(tmp);
250 } 253 }
251 254
252 /* Success */ 255 /* Success */
253 rv = 0; 256 rv = 0;
254 257
255 error: 258 error:
256 for (i = 0; i < junk_length; i++) 259 for (i = 0; i < junk_length; i++)
257 vfree(ptr[i]); 260 vfree(ptr[i]);
258 261
259 vfree(ptr); 262 vfree(ptr);
260 vfree(junk_ptr); 263 vfree(junk_ptr);
261 264
262 return rv; 265 return rv;
263 } 266 }
264 267
265 static int fix_size_alloc_test(void) 268 static int fix_size_alloc_test(void)
266 { 269 {
267 void *ptr; 270 void *ptr;
268 int i; 271 int i;
269 272
270 for (i = 0; i < test_loop_count; i++) 273 for (i = 0; i < test_loop_count; i++) {
271 if (use_huge) !! 274 ptr = vmalloc(3 * PAGE_SIZE);
272 ptr = vmalloc_huge((nr <<
273 else <<
274 ptr = vmalloc((nr_page <<
275 275
276 if (!ptr) 276 if (!ptr)
277 return -1; 277 return -1;
278 278
279 *((__u8 *)ptr) = 0; 279 *((__u8 *)ptr) = 0;
280 280
281 vfree(ptr); 281 vfree(ptr);
282 } 282 }
283 283
284 return 0; 284 return 0;
285 } 285 }
286 286
287 static int 287 static int
288 pcpu_alloc_test(void) 288 pcpu_alloc_test(void)
289 { 289 {
290 int rv = 0; 290 int rv = 0;
291 #ifndef CONFIG_NEED_PER_CPU_KM 291 #ifndef CONFIG_NEED_PER_CPU_KM
292 void __percpu **pcpu; 292 void __percpu **pcpu;
293 size_t size, align; 293 size_t size, align;
294 int i; 294 int i;
295 295
296 pcpu = vmalloc(sizeof(void __percpu *) 296 pcpu = vmalloc(sizeof(void __percpu *) * 35000);
297 if (!pcpu) 297 if (!pcpu)
298 return -1; 298 return -1;
299 299
300 for (i = 0; i < 35000; i++) { 300 for (i = 0; i < 35000; i++) {
301 size = get_random_u32_inclusiv !! 301 unsigned int r;
>> 302
>> 303 get_random_bytes(&r, sizeof(i));
>> 304 size = (r % (PAGE_SIZE / 4)) + 1;
302 305
303 /* 306 /*
304 * Maximum PAGE_SIZE 307 * Maximum PAGE_SIZE
305 */ 308 */
306 align = 1 << get_random_u32_in !! 309 get_random_bytes(&r, sizeof(i));
>> 310 align = 1 << ((i % 11) + 1);
307 311
308 pcpu[i] = __alloc_percpu(size, 312 pcpu[i] = __alloc_percpu(size, align);
309 if (!pcpu[i]) 313 if (!pcpu[i])
310 rv = -1; 314 rv = -1;
311 } 315 }
312 316
313 for (i = 0; i < 35000; i++) 317 for (i = 0; i < 35000; i++)
314 free_percpu(pcpu[i]); 318 free_percpu(pcpu[i]);
315 319
316 vfree(pcpu); 320 vfree(pcpu);
317 #endif 321 #endif
318 return rv; 322 return rv;
319 } 323 }
320 324
321 struct test_kvfree_rcu { 325 struct test_kvfree_rcu {
322 struct rcu_head rcu; 326 struct rcu_head rcu;
323 unsigned char array[20]; 327 unsigned char array[20];
324 }; 328 };
325 329
326 static int 330 static int
327 kvfree_rcu_1_arg_vmalloc_test(void) 331 kvfree_rcu_1_arg_vmalloc_test(void)
328 { 332 {
329 struct test_kvfree_rcu *p; 333 struct test_kvfree_rcu *p;
330 int i; 334 int i;
331 335
332 for (i = 0; i < test_loop_count; i++) 336 for (i = 0; i < test_loop_count; i++) {
333 p = vmalloc(1 * PAGE_SIZE); 337 p = vmalloc(1 * PAGE_SIZE);
334 if (!p) 338 if (!p)
335 return -1; 339 return -1;
336 340
337 p->array[0] = 'a'; 341 p->array[0] = 'a';
338 kvfree_rcu_mightsleep(p); !! 342 kvfree_rcu(p);
339 } 343 }
340 344
341 return 0; 345 return 0;
342 } 346 }
343 347
344 static int 348 static int
345 kvfree_rcu_2_arg_vmalloc_test(void) 349 kvfree_rcu_2_arg_vmalloc_test(void)
346 { 350 {
347 struct test_kvfree_rcu *p; 351 struct test_kvfree_rcu *p;
348 int i; 352 int i;
349 353
350 for (i = 0; i < test_loop_count; i++) 354 for (i = 0; i < test_loop_count; i++) {
351 p = vmalloc(1 * PAGE_SIZE); 355 p = vmalloc(1 * PAGE_SIZE);
352 if (!p) 356 if (!p)
353 return -1; 357 return -1;
354 358
355 p->array[0] = 'a'; 359 p->array[0] = 'a';
356 kvfree_rcu(p, rcu); 360 kvfree_rcu(p, rcu);
357 } 361 }
358 362
359 return 0; 363 return 0;
360 } 364 }
361 365
362 static int 366 static int
363 vm_map_ram_test(void) !! 367 kvfree_rcu_1_arg_slab_test(void)
364 { 368 {
365 unsigned long nr_allocated; !! 369 struct test_kvfree_rcu *p;
366 unsigned int map_nr_pages; <<
367 unsigned char *v_ptr; <<
368 struct page **pages; <<
369 int i; 370 int i;
370 371
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++) 372 for (i = 0; i < test_loop_count; i++) {
382 v_ptr = vm_map_ram(pages, map_ !! 373 p = kmalloc(sizeof(*p), GFP_KERNEL);
383 *v_ptr = 'a'; !! 374 if (!p)
384 vm_unmap_ram(v_ptr, map_nr_pag !! 375 return -1;
>> 376
>> 377 p->array[0] = 'a';
>> 378 kvfree_rcu(p);
385 } 379 }
386 380
387 cleanup: !! 381 return 0;
388 for (i = 0; i < nr_allocated; i++) !! 382 }
389 __free_page(pages[i]); <<
390 383
391 kfree(pages); !! 384 static int
>> 385 kvfree_rcu_2_arg_slab_test(void)
>> 386 {
>> 387 struct test_kvfree_rcu *p;
>> 388 int i;
392 389
393 /* 0 indicates success. */ !! 390 for (i = 0; i < test_loop_count; i++) {
394 return nr_allocated != map_nr_pages; !! 391 p = kmalloc(sizeof(*p), GFP_KERNEL);
>> 392 if (!p)
>> 393 return -1;
>> 394
>> 395 p->array[0] = 'a';
>> 396 kvfree_rcu(p, rcu);
>> 397 }
>> 398
>> 399 return 0;
395 } 400 }
396 401
397 struct test_case_desc { 402 struct test_case_desc {
398 const char *test_name; 403 const char *test_name;
399 int (*test_func)(void); 404 int (*test_func)(void);
400 }; 405 };
401 406
402 static struct test_case_desc test_case_array[] 407 static struct test_case_desc test_case_array[] = {
403 { "fix_size_alloc_test", fix_size_allo 408 { "fix_size_alloc_test", fix_size_alloc_test },
404 { "full_fit_alloc_test", full_fit_allo 409 { "full_fit_alloc_test", full_fit_alloc_test },
405 { "long_busy_list_alloc_test", long_bu 410 { "long_busy_list_alloc_test", long_busy_list_alloc_test },
406 { "random_size_alloc_test", random_siz 411 { "random_size_alloc_test", random_size_alloc_test },
407 { "fix_align_alloc_test", fix_align_al 412 { "fix_align_alloc_test", fix_align_alloc_test },
408 { "random_size_align_alloc_test", rand 413 { "random_size_align_alloc_test", random_size_align_alloc_test },
409 { "align_shift_alloc_test", align_shif 414 { "align_shift_alloc_test", align_shift_alloc_test },
410 { "pcpu_alloc_test", pcpu_alloc_test } 415 { "pcpu_alloc_test", pcpu_alloc_test },
411 { "kvfree_rcu_1_arg_vmalloc_test", kvf 416 { "kvfree_rcu_1_arg_vmalloc_test", kvfree_rcu_1_arg_vmalloc_test },
412 { "kvfree_rcu_2_arg_vmalloc_test", kvf 417 { "kvfree_rcu_2_arg_vmalloc_test", kvfree_rcu_2_arg_vmalloc_test },
413 { "vm_map_ram_test", vm_map_ram_test } !! 418 { "kvfree_rcu_1_arg_slab_test", kvfree_rcu_1_arg_slab_test },
>> 419 { "kvfree_rcu_2_arg_slab_test", kvfree_rcu_2_arg_slab_test },
414 /* Add a new test case here. */ 420 /* Add a new test case here. */
415 }; 421 };
416 422
417 struct test_case_data { 423 struct test_case_data {
418 int test_failed; 424 int test_failed;
419 int test_passed; 425 int test_passed;
420 u64 time; 426 u64 time;
421 }; 427 };
422 428
>> 429 /* Split it to get rid of: WARNING: line over 80 characters */
>> 430 static struct test_case_data
>> 431 per_cpu_test_data[NR_CPUS][ARRAY_SIZE(test_case_array)];
>> 432
423 static struct test_driver { 433 static struct test_driver {
424 struct task_struct *task; 434 struct task_struct *task;
425 struct test_case_data data[ARRAY_SIZE( <<
426 <<
427 unsigned long start; 435 unsigned long start;
428 unsigned long stop; 436 unsigned long stop;
429 } *tdriver; !! 437 int cpu;
>> 438 } per_cpu_test_driver[NR_CPUS];
430 439
431 static void shuffle_array(int *arr, int n) 440 static void shuffle_array(int *arr, int n)
432 { 441 {
433 int i, j; !! 442 unsigned int rnd;
>> 443 int i, j, x;
434 444
435 for (i = n - 1; i > 0; i--) { 445 for (i = n - 1; i > 0; i--) {
>> 446 get_random_bytes(&rnd, sizeof(rnd));
>> 447
436 /* Cut the range. */ 448 /* Cut the range. */
437 j = get_random_u32_below(i); !! 449 j = rnd % i;
438 450
439 /* Swap indexes. */ 451 /* Swap indexes. */
440 swap(arr[i], arr[j]); !! 452 x = arr[i];
>> 453 arr[i] = arr[j];
>> 454 arr[j] = x;
441 } 455 }
442 } 456 }
443 457
444 static int test_func(void *private) 458 static int test_func(void *private)
445 { 459 {
446 struct test_driver *t = private; 460 struct test_driver *t = private;
447 int random_array[ARRAY_SIZE(test_case_ 461 int random_array[ARRAY_SIZE(test_case_array)];
448 int index, i, j; 462 int index, i, j;
449 ktime_t kt; 463 ktime_t kt;
450 u64 delta; 464 u64 delta;
451 465
>> 466 if (set_cpus_allowed_ptr(current, cpumask_of(t->cpu)) < 0)
>> 467 pr_err("Failed to set affinity to %d CPU\n", t->cpu);
>> 468
452 for (i = 0; i < ARRAY_SIZE(test_case_a 469 for (i = 0; i < ARRAY_SIZE(test_case_array); i++)
453 random_array[i] = i; 470 random_array[i] = i;
454 471
455 if (!sequential_test_order) 472 if (!sequential_test_order)
456 shuffle_array(random_array, AR 473 shuffle_array(random_array, ARRAY_SIZE(test_case_array));
457 474
458 /* 475 /*
459 * Block until initialization is done. 476 * Block until initialization is done.
460 */ 477 */
461 down_read(&prepare_for_test_rwsem); 478 down_read(&prepare_for_test_rwsem);
462 479
463 t->start = get_cycles(); 480 t->start = get_cycles();
464 for (i = 0; i < ARRAY_SIZE(test_case_a 481 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
465 index = random_array[i]; 482 index = random_array[i];
466 483
467 /* 484 /*
468 * Skip tests if run_test_mask 485 * Skip tests if run_test_mask has been specified.
469 */ 486 */
470 if (!((run_test_mask & (1 << i 487 if (!((run_test_mask & (1 << index)) >> index))
471 continue; 488 continue;
472 489
473 kt = ktime_get(); 490 kt = ktime_get();
474 for (j = 0; j < test_repeat_co 491 for (j = 0; j < test_repeat_count; j++) {
475 if (!test_case_array[i 492 if (!test_case_array[index].test_func())
476 t->data[index] !! 493 per_cpu_test_data[t->cpu][index].test_passed++;
477 else 494 else
478 t->data[index] !! 495 per_cpu_test_data[t->cpu][index].test_failed++;
479 } 496 }
480 497
481 /* 498 /*
482 * Take an average time that t 499 * Take an average time that test took.
483 */ 500 */
484 delta = (u64) ktime_us_delta(k 501 delta = (u64) ktime_us_delta(ktime_get(), kt);
485 do_div(delta, (u32) test_repea 502 do_div(delta, (u32) test_repeat_count);
486 503
487 t->data[index].time = delta; !! 504 per_cpu_test_data[t->cpu][index].time = delta;
488 } 505 }
489 t->stop = get_cycles(); 506 t->stop = get_cycles();
490 507
491 up_read(&prepare_for_test_rwsem); 508 up_read(&prepare_for_test_rwsem);
492 test_report_one_done(); 509 test_report_one_done();
493 510
494 /* 511 /*
495 * Wait for the kthread_stop() call. 512 * Wait for the kthread_stop() call.
496 */ 513 */
497 while (!kthread_should_stop()) 514 while (!kthread_should_stop())
498 msleep(10); 515 msleep(10);
499 516
500 return 0; 517 return 0;
501 } 518 }
502 519
503 static int !! 520 static void
504 init_test_configuration(void) !! 521 init_test_configurtion(void)
505 { 522 {
506 /* 523 /*
507 * A maximum number of workers is defi !! 524 * Reset all data of all CPUs.
508 * value and set to USHRT_MAX. We add <<
509 * case and for potential heavy stress <<
510 */ 525 */
511 nr_threads = clamp(nr_threads, 1, (int !! 526 memset(per_cpu_test_data, 0, sizeof(per_cpu_test_data));
512 527
513 /* Allocate the space for test instanc !! 528 if (single_cpu_test)
514 tdriver = kvcalloc(nr_threads, sizeof( !! 529 cpumask_set_cpu(cpumask_first(cpu_online_mask),
515 if (tdriver == NULL) !! 530 &cpus_run_test_mask);
516 return -1; !! 531 else
>> 532 cpumask_and(&cpus_run_test_mask, cpu_online_mask,
>> 533 cpu_online_mask);
517 534
518 if (test_repeat_count <= 0) 535 if (test_repeat_count <= 0)
519 test_repeat_count = 1; 536 test_repeat_count = 1;
520 537
521 if (test_loop_count <= 0) 538 if (test_loop_count <= 0)
522 test_loop_count = 1; 539 test_loop_count = 1;
523 <<
524 return 0; <<
525 } 540 }
526 541
527 static void do_concurrent_test(void) 542 static void do_concurrent_test(void)
528 { 543 {
529 int i, ret; !! 544 int cpu, ret;
530 545
531 /* 546 /*
532 * Set some basic configurations plus 547 * Set some basic configurations plus sanity check.
533 */ 548 */
534 ret = init_test_configuration(); !! 549 init_test_configurtion();
535 if (ret < 0) <<
536 return; <<
537 550
538 /* 551 /*
539 * Put on hold all workers. 552 * Put on hold all workers.
540 */ 553 */
541 down_write(&prepare_for_test_rwsem); 554 down_write(&prepare_for_test_rwsem);
542 555
543 for (i = 0; i < nr_threads; i++) { !! 556 for_each_cpu(cpu, &cpus_run_test_mask) {
544 struct test_driver *t = &tdriv !! 557 struct test_driver *t = &per_cpu_test_driver[cpu];
545 558
546 t->task = kthread_run(test_fun !! 559 t->cpu = cpu;
>> 560 t->task = kthread_run(test_func, t, "vmalloc_test/%d", cpu);
547 561
548 if (!IS_ERR(t->task)) 562 if (!IS_ERR(t->task))
549 /* Success. */ 563 /* Success. */
550 atomic_inc(&test_n_und 564 atomic_inc(&test_n_undone);
551 else 565 else
552 pr_err("Failed to star !! 566 pr_err("Failed to start kthread for %d CPU\n", cpu);
553 } 567 }
554 568
555 /* 569 /*
556 * Now let the workers do their job. 570 * Now let the workers do their job.
557 */ 571 */
558 up_write(&prepare_for_test_rwsem); 572 up_write(&prepare_for_test_rwsem);
559 573
560 /* 574 /*
561 * Sleep quiet until all workers are d 575 * Sleep quiet until all workers are done with 1 second
562 * interval. Since the test can take a 576 * interval. Since the test can take a lot of time we
563 * can run into a stack trace of the h 577 * can run into a stack trace of the hung task. That is
564 * why we go with completion_timeout a 578 * why we go with completion_timeout and HZ value.
565 */ 579 */
566 do { 580 do {
567 ret = wait_for_completion_time 581 ret = wait_for_completion_timeout(&test_all_done_comp, HZ);
568 } while (!ret); 582 } while (!ret);
569 583
570 for (i = 0; i < nr_threads; i++) { !! 584 for_each_cpu(cpu, &cpus_run_test_mask) {
571 struct test_driver *t = &tdriv !! 585 struct test_driver *t = &per_cpu_test_driver[cpu];
572 int j; !! 586 int i;
573 587
574 if (!IS_ERR(t->task)) 588 if (!IS_ERR(t->task))
575 kthread_stop(t->task); 589 kthread_stop(t->task);
576 590
577 for (j = 0; j < ARRAY_SIZE(tes !! 591 for (i = 0; i < ARRAY_SIZE(test_case_array); i++) {
578 if (!((run_test_mask & !! 592 if (!((run_test_mask & (1 << i)) >> i))
579 continue; 593 continue;
580 594
581 pr_info( 595 pr_info(
582 "Summary: %s p 596 "Summary: %s passed: %d failed: %d repeat: %d loops: %d avg: %llu usec\n",
583 test_case_arra !! 597 test_case_array[i].test_name,
584 t->data[j].tes !! 598 per_cpu_test_data[cpu][i].test_passed,
585 t->data[j].tes !! 599 per_cpu_test_data[cpu][i].test_failed,
586 test_repeat_co 600 test_repeat_count, test_loop_count,
587 t->data[j].tim !! 601 per_cpu_test_data[cpu][i].time);
588 } 602 }
589 603
590 pr_info("All test took worker% !! 604 pr_info("All test took CPU%d=%lu cycles\n",
591 i, t->stop - t->start) !! 605 cpu, t->stop - t->start);
592 } 606 }
593 <<
594 kvfree(tdriver); <<
595 } 607 }
596 608
597 static int vmalloc_test_init(void) 609 static int vmalloc_test_init(void)
598 { 610 {
599 do_concurrent_test(); 611 do_concurrent_test();
600 return -EAGAIN; /* Fail will directly 612 return -EAGAIN; /* Fail will directly unload the module */
601 } 613 }
602 614
>> 615 static void vmalloc_test_exit(void)
>> 616 {
>> 617 }
>> 618
603 module_init(vmalloc_test_init) 619 module_init(vmalloc_test_init)
>> 620 module_exit(vmalloc_test_exit)
604 621
605 MODULE_LICENSE("GPL"); 622 MODULE_LICENSE("GPL");
606 MODULE_AUTHOR("Uladzislau Rezki"); 623 MODULE_AUTHOR("Uladzislau Rezki");
607 MODULE_DESCRIPTION("vmalloc test module"); 624 MODULE_DESCRIPTION("vmalloc test module");
608 625
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.