1LTP C Test API 2============== 3 4NOTE: See also 5 https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines[Test Writing Guidelines], 6 https://github.com/linux-test-project/ltp/wiki/C-Test-Case-Tutorial[C Test Case Tutorial], 7 https://github.com/linux-test-project/ltp/wiki/Shell-Test-API[Shell Test API]. 8 91 Writing a test in C 10--------------------- 11 121.1 Basic test structure 13~~~~~~~~~~~~~~~~~~~~~~~~ 14 15Let's start with an example, following code is a simple test for a 'getenv()'. 16 17[source,c] 18------------------------------------------------------------------------------- 19/*\ 20 * [Description] 21 * Tests basic functionality of getenv(). 22 * 23 * - create an env variable and verify that getenv() can get it 24 * - call getenv() with nonexisting variable name, check that it returns NULL 25 */ 26 27#include "tst_test.h" 28 29#define ENV1 "LTP_TEST_ENV" 30#define ENV2 "LTP_TEST_THIS_DOES_NOT_EXIST" 31#define ENV_VAL "val" 32 33static void setup(void) 34{ 35 if (setenv(ENV1, ENV_VAL, 1)) 36 tst_brk(TBROK | TERRNO, "setenv() failed"); 37} 38 39static void run(void) 40{ 41 char *ret; 42 43 ret = getenv(ENV1); 44 45 if (!ret) { 46 tst_res(TFAIL, "getenv(" ENV1 ") = NULL"); 47 goto next; 48 } 49 50 if (!strcmp(ret, ENV_VAL)) { 51 tst_res(TPASS, "getenv(" ENV1 ") = '"ENV_VAL "'"); 52 } else { 53 tst_res(TFAIL, "getenv(" ENV1 ") = '%s', expected '" 54 ENV_VAL "'", ret); 55 } 56 57next: 58 ret = getenv(ENV2); 59 60 if (ret) 61 tst_res(TFAIL, "getenv(" ENV2 ") = '%s'", ret); 62 else 63 tst_res(TPASS, "getenv(" ENV2 ") = NULL"); 64} 65 66static struct tst_test test = { 67 .test_all = run, 68 .setup = setup, 69}; 70------------------------------------------------------------------------------- 71 72Each test includes the 'tst_test.h' header and must define the 'struct 73tst_test test' structure. 74 75The overall test initialization is done in the 'setup()' function. 76 77The overall cleanup is done in a 'cleanup()' function. Here 'cleanup()' is 78omitted as the test does not have anything to clean up. If cleanup is set in 79the test structure it's called on test exit just before the test library 80cleanup. That especially means that cleanup can be called at any point in a 81test execution. For example even when a test setup step has failed, therefore 82the 'cleanup()' function must be able to cope with unfinished initialization, 83and so on. 84 85The test itself is done in the 'test()' function. The test function must work 86fine if called in a loop. 87 88There are two types of a test function pointers in the test structure. The 89first one is a '.test_all' pointer that is used when test is implemented as a 90single function. Then there is a '.test' function along with the number of 91tests '.tcnt' that allows for more detailed result reporting. If the '.test' 92pointer is set the function is called '.tcnt' times with an integer parameter 93in range of [0, '.tcnt' - 1]. 94 95IMPORTANT: Only one of '.test' and '.test_all' can be set at a time. 96 97Each test has a limit on how long it can run and the limit composes of two 98parts max_runtime and timeout. The max_runtime is a limit for how long can the 99'.test_all' or a set of '.test' functions take and the timeout is static part 100that should cover the duration of test setup and cleanup plus some safety. 101 102Any test that runs for more than a second or two has to make sure to: 103 104- set the runtime either by setting the '.max_runtime' in tst_test or by 105 calling 'tst_set_max_runtime()' in the test setup 106 107- monitor remaning runtime by regular calls to 'tst_remaining_runtime()' and 108 exit when runtime has been used up 109 110Test is free to exit before max_runtime has been used up for example when 111minimal number of iteration was finished. 112 113The limit is applied to a single call of the '.test_all' function that means 114that for example when '.test_variants' or '.all_filesystems' is set the whole 115test will be limited by 'variants * (max_runtime + timeout)' seconds and the 116test runtime will be likely close to 'variants * max_runtime' seconds. 117 118[source,c] 119------------------------------------------------------------------------------- 120/* 121 * Returns number of seconds or zero in case that runtime has been used up. 122 */ 123 124int tst_remaining_runtime(void); 125------------------------------------------------------------------------------- 126 127LAPI headers 128++++++++++++ 129 130Use our LAPI headers ('include "lapi/foo.h"') to keep compatibility with old 131distributions. LAPI header should always include original header. Older linux 132headers were problematic, therefore we preferred to use libc headers. There are 133still some bugs when combining certain glibc headers with linux headers, see 134https://sourceware.org/glibc/wiki/Synchronizing_Headers. 135 136A word about the cleanup() callback 137+++++++++++++++++++++++++++++++++++ 138 139There are a few rules that needs to be followed in order to write correct 140cleanup() callback. 141 1421. Free only resources that were initialized. Keep in mind that callback can 143 be executed at any point in the test run. 144 1452. Make sure to free resources in the reverse order they were 146 initialized. (Some of the steps may not depend on others and everything 147 will work if there were swapped but let's keep it in order.) 148 149The first rule may seem complicated at first however, on the contrary, it's 150quite easy. All you have to do is to keep track of what was already 151initialized. For example file descriptors needs to be closed only if they were 152assigned a valid file descriptor. For most of the things you need to create 153extra flag that is set right after successful initialization though. Consider, 154for example, test setup below. 155 156We also prefer cleaning up resources that would otherwise be released on the 157program exit. There are two main reasons for this decision. Resources such as 158file descriptors and mmaped memory could block umounting a block device in 159cases where the test library has mounted a filesystem for the test temporary 160directory. Not freeing allocated memory would upset static analysis and tools 161such as valgrind and produce false-positives when checking for leaks in the 162libc and other low level libraries. 163 164[source,c] 165------------------------------------------------------------------------------- 166static int fd0, fd1, mount_flag; 167 168#define MNTPOINT "mntpoint" 169#define FILE1 "mntpoint/file1" 170#define FILE2 "mntpoint/file2" 171 172static void setup(void) 173{ 174 SAFE_MKDIR(MNTPOINT, 0777); 175 SAFE_MKFS(tst_device->dev, tst_device->fs_type, NULL, NULL); 176 SAFE_MOUNT(tst_device->dev, MNTPOINT, tst_device->fs_type, 0, 0); 177 mount_flag = 1; 178 179 fd0 = SAFE_OPEN(cleanup, FILE1, O_CREAT | O_RDWR, 0666); 180 fd1 = SAFE_OPEN(cleanup, FILE2, O_CREAT | O_RDWR, 0666); 181} 182------------------------------------------------------------------------------- 183 184In this case the 'cleanup()' function may be invoked when any of the 'SAFE_*' 185macros has failed and therefore must be able to work with unfinished 186initialization as well. Since global variables are initialized to zero we can 187just check that fd > 0 before we attempt to close it. The mount function 188requires extra flag to be set after device was successfully mounted. 189 190[source,c] 191------------------------------------------------------------------------------- 192static void cleanup(void) 193{ 194 if (fd1 > 0) 195 SAFE_CLOSE(fd1); 196 197 if (fd0 > 0) 198 SAFE_CLOSE(fd0); 199 200 if (mount_flag && tst_umouont(MNTPOINT)) 201 tst_res(TWARN | TERRNO, "umount(%s)", MNTPOINT); 202} 203------------------------------------------------------------------------------- 204 205IMPORTANT: 'SAFE_MACROS()' used in cleanup *do not* exit the test. Failure 206 only produces a warning and the 'cleanup()' carries on. This is 207 intentional as we want to execute as much 'cleanup()' as possible. 208 209WARNING: Calling tst_brk() in test 'cleanup()' does not exit the test as well 210 and 'TBROK' is converted to 'TWARN'. 211 212NOTE: Creation and removal of the test temporary directory is handled in 213 the test library and the directory is removed recursively. Therefore 214 we do not have to remove files and directories in the test cleanup. 215 2161.2 Basic test interface 217~~~~~~~~~~~~~~~~~~~~~~~~ 218 219[source,c] 220------------------------------------------------------------------------------- 221void tst_res(int ttype, char *arg_fmt, ...); 222------------------------------------------------------------------------------- 223 224Printf-like function to report test result, it's mostly used with ttype: 225 226|============================== 227| 'TPASS' | Test has passed. 228| 'TFAIL' | Test has failed. 229| 'TINFO' | General message. 230| 'TDEBUG' | Debug message (new C API only, printed with '-D' or via 'LTP_ENABLE_DEBUG=1' or 'y' 231 environment variable), only for messages which would be too verbose for normal run. 232| 'TWARN' | Something went wrong but we decided to continue. Mostly used in cleanup functions. 233|============================== 234 235The 'ttype' can be combined bitwise with 'TERRNO' or 'TTERRNO' to print 236'errno', 'TST_ERR' respectively. 237 238[source,c] 239------------------------------------------------------------------------------- 240void tst_brk(int ttype, char *arg_fmt, ...); 241------------------------------------------------------------------------------- 242 243Printf-like function to report error and exit the test, it can be used with ttype: 244 245|============================================================ 246| 'TBROK' | Something has failed in test preparation phase. 247| 'TCONF' | Test is not appropriate for current configuration 248 (syscall not implemented, unsupported arch, ...) 249|============================================================ 250 251The 'ttype' can be combined bitwise with 'TERRNO' or 'TTERRNO' to print 252'errno', 'TST_ERR' respectively. 253 254There are also 'TST_EXP_*()' macros that can simplify syscall unit tests to a 255single line, use them whenever possible. These macros take a function call as 256the first parameter and a printf-like format string and parameters as well. 257These test macros then expand to a code that runs the call, checks the return 258value and errno and reports the test result. 259 260[source,c] 261------------------------------------------------------------------------------- 262static void run(void) 263{ 264 ... 265 TST_EXP_PASS(stat(fname, &statbuf), "stat(%s, ...)", fname); 266 267 if (!TST_PASS) 268 return; 269 ... 270} 271------------------------------------------------------------------------------- 272 273The 'TST_EXP_PASS()' can be used for calls that return -1 on failure and 0 on 274success. It will check for the return value and reports failure if the return 275value is not equal to 0. The call also sets the 'TST_PASS' variable to 1 if 276the call succeeeded. 277 278As seen above, this and similar macros take optional variadic arguments. These 279begin with a format string and then appropriate values to be formatted. 280 281[source,c] 282------------------------------------------------------------------------------- 283static void run(void) 284{ 285 ... 286 TST_EXP_FD(open(fname, O_RDONLY), "open(%s, O_RDONLY)", fname); 287 288 SAFE_CLOSE(TST_RET); 289 ... 290} 291------------------------------------------------------------------------------- 292 293The 'TST_EXP_FD()' is the same as 'TST_EXP_PASS()' the only difference is that 294the return value is expected to be a file descriptor so the call passes if 295positive integer is returned. 296 297[source,c] 298------------------------------------------------------------------------------- 299static void run(void) 300{ 301 ... 302 TST_EXP_FAIL(stat(fname, &statbuf), ENOENT, "stat(%s, ...)", fname); 303 ... 304} 305------------------------------------------------------------------------------- 306 307The 'TST_EXP_FAIL()' is similar to 'TST_EXP_PASS()' but it fails the test if 308the call haven't failed with -1 and 'errno' wasn't set to the expected one 309passed as the second argument. 310 311[source,c] 312------------------------------------------------------------------------------- 313static void run(void) 314{ 315 ... 316 TST_EXP_FAIL2(msgget(key, flags), EINVAL, "msgget(%i, %i)", key, flags); 317 ... 318} 319------------------------------------------------------------------------------- 320 321The 'TST_EXP_FAIL2()' is the same as 'TST_EXP_FAIL()' except the return value is 322expected to be non-negative integer if call passes. These macros build upon the 323+TEST()+ macro and associated variables. 324 325'TST_EXP_FAIL_SILENT()' and 'TST_EXP_FAIL2_SILENT()' variants are less verbose 326and do not print TPASS messages when SCALL fails as expected. 327 328[source,c] 329------------------------------------------------------------------------------- 330TEST(socket(AF_INET, SOCK_RAW, 1)); 331if (TST_RET > -1) { 332 tst_res(TFAIL, "Created raw socket"); 333 SAFE_CLOSE(TST_RET); 334} else if (TST_ERR != EPERM) { 335 tst_res(TFAIL | TTERRNO, 336 "Failed to create socket for wrong reason"); 337} else { 338 tst_res(TPASS | TTERRNO, "Didn't create raw socket"); 339} 340------------------------------------------------------------------------------- 341 342The +TEST+ macro sets +TST_RET+ to its argument's return value and +TST_ERR+ to 343+errno+. The +TTERNO+ flag can be used to print the error number's symbolic 344value. 345 346No LTP library function or macro, except those in 'tst_test_macros.h', will 347write to these variables (rule 'LTP-002'). So their values will not be changed 348unexpectedly. 349 350[source,c] 351------------------------------------------------------------------------------- 352TST_EXP_POSITIVE(wait(&status)); 353 354if (!TST_PASS) 355 return; 356------------------------------------------------------------------------------- 357 358If the return value of 'wait' is positive or zero, this macro will print a pass 359result and set +TST_PASS+ appropriately. If the return value is negative, then 360it will print fail. There are many similar macros to those shown here, please 361see 'tst_test_macros.h'. 362 363[source,c] 364------------------------------------------------------------------------------- 365TST_EXP_EQ_LI(val1, val2); 366TST_EXP_EQ_UI(val1, val2); 367TST_EXP_EQ_SZ(val1, val2); 368TST_EXP_EQ_SSZ(val1, val2); 369 370/* Use as */ 371TST_EXP_EQ_LI(sig_caught, SIGCHLD); 372------------------------------------------------------------------------------- 373 374Set of macros for different integer type comparsions. These macros print the 375variable names as well as values in both pass and fail scenarios. 376 377[source,c] 378------------------------------------------------------------------------------- 379const char *tst_strsig(int sig); 380------------------------------------------------------------------------------- 381 382Return the given signal number's corresponding string. 383 384[source,c] 385------------------------------------------------------------------------------- 386const char *tst_strerrno(int err); 387------------------------------------------------------------------------------- 388 389Return the given errno number's corresponding string. Using this function to 390translate 'errno' values to strings is preferred. You should not use the 391'strerror()' function in the testcases. 392 393[source,c] 394------------------------------------------------------------------------------- 395const char *tst_strstatus(int status); 396------------------------------------------------------------------------------- 397 398Returns string describing the status as returned by 'wait()'. 399 400WARNING: This function is not thread safe. 401 402[source,c] 403------------------------------------------------------------------------------- 404void tst_set_max_runtime(int max_runtime); 405------------------------------------------------------------------------------- 406 407Allows for setting max_runtime per test iteration dynamically in the test 'setup()', 408the timeout is specified in seconds. There are a few testcases whose runtime 409can vary arbitrarily, these can disable timeouts by setting it to 410TST_UNLIMITED_RUNTIME. 411 412[source,c] 413------------------------------------------------------------------------------- 414void tst_flush(void); 415------------------------------------------------------------------------------- 416 417Flush output streams, handling errors appropriately. 418 419This function is rarely needed when you have to flush the output streams 420before calling 'fork()' or 'clone()'. Note that the 'SAFE_FORK()' and 'SAFE_CLONE()' 421calls this function automatically. See 2.4 FILE buffers and fork() for explanation 422why is this needed. 423 4241.3 Test temporary directory 425~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 426 427If '.needs_tmpdir' is set to '1' in the 'struct tst_test' unique test 428temporary is created and it's set as the test working directory. Tests *MUST 429NOT* create temporary files outside that directory. The flag is not needed to 430be set when use these flags: '.all_filesystems', '.format_device', '.mntpoint', 431'.mount_device' '.needs_checkpoints', '.needs_device', '.resource_file' 432(these flags imply creating temporary directory). 433 434IMPORTANT: Close all file descriptors (that point to files in test temporary 435 directory, even the unlinked ones) either in the 'test()' function 436 or in the test 'cleanup()' otherwise the test may break temporary 437 directory removal on NFS (look for "NFS silly rename"). 438 4391.4 Safe macros 440~~~~~~~~~~~~~~~ 441 442Safe macros aim to simplify error checking in test preparation. Instead of 443calling system API functions, checking for their return value and aborting the 444test if the operation has failed, you just use corresponding safe macro. 445 446Use them whenever it's possible. 447 448Instead of writing: 449 450[source,c] 451------------------------------------------------------------------------------- 452fd = open("/dev/null", O_RDONLY); 453if (fd < 0) 454 tst_brk(TBROK | TERRNO, "opening /dev/null failed"); 455------------------------------------------------------------------------------- 456 457You write just: 458 459[source,c] 460------------------------------------------------------------------------------- 461fd = SAFE_OPEN("/dev/null", O_RDONLY); 462------------------------------------------------------------------------------- 463 464IMPORTANT: The 'SAFE_CLOSE()' function also sets the passed file descriptor to -1 465 after it's successfully closed. 466 467They can also simplify reading and writing of sysfs files, you can, for 468example, do: 469 470[source,c] 471------------------------------------------------------------------------------- 472SAFE_FILE_SCANF("/proc/sys/kernel/pid_max", "%lu", &pid_max); 473------------------------------------------------------------------------------- 474 475See 'include/tst_safe_macros.h', 'include/tst_safe_stdio.h' and 476'include/tst_safe_file_ops.h' and 'include/tst_safe_net.h' for a complete list. 477 4781.5 Test specific command line options 479~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 480 481[source,c] 482------------------------------------------------------------------------------- 483struct tst_option { 484 char *optstr; 485 char **arg; 486 char *help; 487}; 488------------------------------------------------------------------------------- 489 490Test specific command line parameters can be passed with the 'NULL' terminated 491array of 'struct tst_option'. The 'optstr' is the command line option i.e. "o" 492or "o:" if option has a parameter. Only short options are supported. The 'arg' 493is where 'optarg' is stored upon match. If option has no parameter it's set to 494non-'NULL' value if option was present. The 'help' is a short help string. 495 496NOTE: The test parameters must not collide with common test parameters defined 497 in the library the currently used ones are +-i+, +-I+, +-C+, and +-h+. 498 499[source,c] 500------------------------------------------------------------------------------- 501int tst_parse_int(const char *str, int *val, int min, int max); 502int tst_parse_long(const char *str, long *val, long min, long max); 503int tst_parse_float(const char *str, float *val, float min, float max); 504int tst_parse_filesize(const char *str, long long *val, long long min, long long max); 505------------------------------------------------------------------------------- 506 507Helpers for parsing the strings returned in the 'struct tst_option'. 508 509Helpers return zero on success and 'errno', mostly 'EINVAL' or 'ERANGE', on 510failure. 511 512Helpers functions are no-op if 'str' is 'NULL'. 513 514The valid range for result includes both 'min' and 'max'. 515 516In particular, 'tst_parse_filesize' function accepts prefix multiplies such as 517"k/K for kilobytes, "m/M" for megabytes and "g/G" for gigabytes. For example, 51810K are converted into 10240 bytes. 519 520[source,c] 521------------------------------------------------------------------------------- 522#include <limits.h> 523#include "tst_test.h" 524 525static char *str_threads; 526static int threads = 10; 527 528static void setup(void) 529{ 530 if (tst_parse_int(str_threads, &threads, 1, INT_MAX)) 531 tst_brk(TBROK, "Invalid number of threads '%s'", str_threads); 532 533 ... 534} 535 536static void test_threads(void) 537{ 538 ... 539 540 for (i = 0; i < threads; i++) { 541 ... 542 } 543 544 ... 545} 546 547static struct tst_test test = { 548 ... 549 .options = (struct tst_option[]) { 550 {"t:", &str_threads, "Number of threads (default 10)"}, 551 {}, 552 ... 553}; 554------------------------------------------------------------------------------- 555 556 5571.6 Runtime kernel version detection 558~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 559 560Testcases for newly added kernel functionality require kernel newer than a 561certain version to run. All you need to skip a test on older kernels is to 562set the '.min_kver' string in the 'struct tst_test' to a minimal required 563kernel version, e.g. '.min_kver = "4.10.0"'. 564 565For more complicated operations such as skipping a test for a certain range 566of kernel versions, following functions could be used: 567 568[source,c] 569------------------------------------------------------------------------------- 570int tst_kvercmp(int r1, int r2, int r3); 571 572struct tst_kern_exv { 573 char *dist_name; 574 char *extra_ver; 575}; 576 577int tst_kvercmp2(int r1, int r2, int r3, struct tst_kern_exv *vers); 578------------------------------------------------------------------------------- 579 580These two functions are intended for runtime kernel version detection. They 581parse the output from 'uname()' and compare it to the passed values. 582 583The return value is similar to the 'strcmp()' function, i.e. zero means equal, 584negative value means that the kernel is older than the expected value and 585positive means that it's newer. 586 587The second function 'tst_kvercmp2()' allows for specifying per-vendor table of 588kernel versions as vendors typically backport fixes to their kernels and the 589test may be relevant even if the kernel version does not suggests so. 590 591[source,c] 592------------------------------------------------------------------------------- 593if (tst_kvercmp(5, 19, 0) >= 0) 594 tst_res(TCONF, "Test valid only for kernel < 5.19"); 595 596static struct tst_kern_exv kvers[] = { 597 { "UBUNTU", "4.4.0-48.69" }, 598 { NULL, NULL}, 599}; 600 601if (tst_kvercmp2(4, 4, 27, kvers) < 0) 602 /* code for kernel < v4.4.27 or ubuntu kernel < 4.4.0-48.69 */ 603------------------------------------------------------------------------------- 604 605WARNING: The shell 'tst_kvercmp' maps the result into unsigned integer - the 606 process exit value. 607 608NOTE: See also LTP 609 https://github.com/linux-test-project/ltp/wiki/Supported-kernel,-libc,-toolchain-versions#13-minimal-supported-kernel-version[minimal supported kernel version]. 610 6111.7 Fork()-ing 612~~~~~~~~~~~~~~ 613 614Be wary that if the test forks and there were messages printed by the 615'tst_*()' interfaces, the data may still be in libc/kernel buffers and these 616*ARE NOT* flushed automatically. 617 618This happens when 'stdout' gets redirected to a file. In this case, the 619'stdout' is not line buffered, but block buffered. Hence after a fork content 620of the buffers will be printed by the parent and each of the children. 621 622To avoid that you should use 'SAFE_FORK()', 'SAFE_CLONE()' or 'tst_clone()'. 623 624IMPORTANT: You have to set the '.forks_child' flag in the test structure 625 if your testcase forks or calls 'SAFE_CLONE()'. 626 6271.8 Doing the test in the child process 628~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 629 630Results reported by 'tst_res()' are propagated to the parent test process via 631block of shared memory. 632 633Calling 'tst_brk()' causes child process to exit with non-zero exit value. 634Which means that it's safe to use 'SAFE_*()' macros in the child processes as 635well. 636 637Children that outlive the 'test()' function execution are waited for in the 638test library. Unclean child exit (killed by signal, non-zero exit value, etc.) 639will cause the main test process to exit with 'tst_brk()', which especially 640means that 'TBROK' propagated from a child process will cause the whole test 641to exit with 'TBROK'. 642 643If a test needs a child that segfaults or does anything else that cause it to 644exit uncleanly all you need to do is to wait for such children from the 645'test()' function so that it's reaped before the main test exits the 'test()' 646function. 647 648[source,c] 649------------------------------------------------------------------------------- 650#include "tst_test.h" 651 652void tst_reap_children(void); 653------------------------------------------------------------------------------- 654 655The 'tst_reap_children()' function makes the process wait for all of its 656children and exits with 'tst_brk(TBROK, ...)' if any of them returned 657a non zero exit code. 658 659When using 'SAFE_CLONE' or 'tst_clone', this may not work depending on 660the parameters passed to clone. The following call to 'SAFE_CLONE' is 661identical to 'fork()', so will work as expected. 662 663[source,c] 664-------------------------------------------------------------------------------- 665const struct tst_clone_args args = { 666 .exit_signal = SIGCHLD, 667}; 668 669SAFE_CLONE(&args); 670-------------------------------------------------------------------------------- 671 672If 'exit_signal' is set to something else, then this will break 673'tst_reap_children'. It's not expected that all parameters to clone will 674work with the LTP library unless specific action is taken by the test code. 675 676.Using 'tst_res()' from binaries started by 'exec()' 677[source,c] 678------------------------------------------------------------------------------- 679/* test.c */ 680#define _GNU_SOURCE 681#include <unistd.h> 682#include "tst_test.h" 683 684static void do_test(void) 685{ 686 char *const argv[] = {"test_exec_child", NULL}; 687 char path[4096]; 688 689 if (tst_get_path("test_exec_child", path, sizeof(path))) 690 tst_brk(TCONF, "Couldn't find test_exec_child in $PATH"); 691 692 execve(path, argv, environ); 693 694 tst_res(TFAIL | TERRNO, "EXEC!"); 695} 696 697static struct tst_test test = { 698 .test_all = do_test, 699 .child_needs_reinit = 1, 700}; 701 702/* test_exec_child.c */ 703#define TST_NO_DEFAULT_MAIN 704#include "tst_test.h" 705 706int main(void) 707{ 708 tst_reinit(); 709 tst_res(TPASS, "Child passed!"); 710 return 0; 711} 712------------------------------------------------------------------------------- 713 714The 'tst_res()' function can be also used from binaries started by 'exec()', 715the parent test process has to set the '.child_needs_reinit' flag so that the 716library prepares for it and has to make sure the 'LTP_IPC_PATH' environment 717variable is passed down, then the very first thing the program has to call in 718'main()' is 'tst_reinit()' that sets up the IPC. 719 7201.9 Fork() and Parent-child synchronization 721~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 722 723As LTP tests are written for Linux, most of the tests involve fork()-ing and 724parent-child process synchronization. LTP includes a checkpoint library that 725provides wait/wake futex based functions. 726 727In order to use checkpoints the '.needs_checkpoints' flag in the 'struct 728tst_test' must be set to '1', this causes the test library to initialize 729checkpoints before the 'test()' function is called. 730 731[source,c] 732------------------------------------------------------------------------------- 733#include "tst_test.h" 734 735TST_CHECKPOINT_WAIT(id) 736 737TST_CHECKPOINT_WAIT2(id, msec_timeout) 738 739TST_CHECKPOINT_WAKE(id) 740 741TST_CHECKPOINT_WAKE2(id, nr_wake) 742 743TST_CHECKPOINT_WAKE_AND_WAIT(id) 744------------------------------------------------------------------------------- 745 746The checkpoint interface provides pair of wake and wait functions. The 'id' is 747unsigned integer which specifies checkpoint to wake/wait for. As a matter of 748fact it's an index to an array stored in a shared memory, so it starts on 749'0' and there should be enough room for at least of hundred of them. 750 751The 'TST_CHECKPOINT_WAIT()' and 'TST_CHECKPOINT_WAIT2()' suspends process 752execution until it's woken up or until timeout is reached. 753 754The 'TST_CHECKPOINT_WAKE()' wakes one process waiting on the checkpoint. 755If no process is waiting the function retries until it success or until 756timeout is reached. 757 758If timeout has been reached process exits with appropriate error message (uses 759'tst_brk()'). 760 761The 'TST_CHECKPOINT_WAKE2()' does the same as 'TST_CHECKPOINT_WAKE()' but can 762be used to wake precisely 'nr_wake' processes. 763 764The 'TST_CHECKPOINT_WAKE_AND_WAIT()' is a shorthand for doing wake and then 765immediately waiting on the same checkpoint. 766 767Child processes created via 'SAFE_FORK()' are ready to use the checkpoint 768synchronization functions, as they inherited the mapped page automatically. 769 770Child processes started via 'exec()', or any other processes not forked from 771the test process must initialize the checkpoint by calling 'tst_reinit()'. 772 773For the details of the interface, look into the 'include/tst_checkpoint.h'. 774 775[source,c] 776------------------------------------------------------------------------------- 777#include "tst_test.h" 778 779/* 780 * Waits for process state change. 781 * 782 * The state is one of the following: 783 * 784 * R - process is running 785 * S - process is sleeping 786 * D - process sleeping uninterruptibly 787 * Z - zombie process 788 * T - process is traced 789 */ 790TST_PROCESS_STATE_WAIT(pid, state, msec_timeout) 791------------------------------------------------------------------------------- 792 793The 'TST_PROCESS_STATE_WAIT()' waits until process 'pid' is in requested 794'state' or timeout is reached. The call polls +/proc/pid/stat+ to get this 795information. A timeout of 0 will wait infinitely. 796 797On timeout -1 is returned and errno set to ETIMEDOUT. 798 799It's mostly used with state 'S' which means that process is sleeping in kernel 800for example in 'pause()' or any other blocking syscall. 801 8021.10 Signals and signal handlers 803~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 804 805If you need to use signal handlers, keep the code short and simple. Don't 806forget that the signal handler is called asynchronously and can interrupt the 807code execution at any place. 808 809This means that problems arise when global state is changed both from the test 810code and signal handler, which will occasionally lead to: 811 812* Data corruption (data gets into inconsistent state), this may happen, for 813 example, for any operations on 'FILE' objects. 814 815* Deadlock, this happens, for example, if you call 'malloc(2)', 'free(2)', 816 etc. from both the test code and the signal handler at the same time since 817 'malloc' has global lock for it's internal data structures. (Be wary that 818 'malloc(2)' is used by the libc functions internally too.) 819 820* Any other unreproducible and unexpected behavior. 821 822Quite common mistake is to call 'exit(3)' from a signal handler. Note that this 823function is not signal-async-safe as it flushes buffers, etc. If you need to 824exit a test immediately from a signal handler use '_exit(2)' instead. 825 826TIP: See 'man 7 signal' for the list of signal-async-safe functions. 827 828If a signal handler sets a variable, its declaration must be 'volatile', 829otherwise compiler may misoptimize the code. This is because the variable may 830not be changed in the compiler code flow analysis. There is 'sig_atomic_t' 831type defined in C99 but this one *DOES NOT* imply 'volatile' (it's just a 832'typedef' to 'int'). So the correct type for a flag that is changed from a 833signal handler is either 'volatile int' or 'volatile sig_atomic_t'. 834 835If a crash (e.g. triggered by signal SIGSEGV) is expected in testing, you 836can avoid creation of core files by calling 'tst_no_corefile()' function. 837This takes effect for process (and its children) which invoked it, unless 838they subsequently modify RLIMIT_CORE. 839 840Note that LTP library will reap any processes that test didn't reap itself, 841and report any non-zero exit code as failure. 842 8431.11 Kernel Modules 844~~~~~~~~~~~~~~~~~~~ 845 846There are certain cases where the test needs a kernel part and userspace part, 847happily, LTP can build a kernel module and then insert it to the kernel on test 848start for you. See 'testcases/kernel/device-drivers/block' for details. 849 8501.12 Useful macros 851~~~~~~~~~~~~~~~~~~ 852 853These macros are defined in 'include/tst_common.h'. 854 855[source,c] 856------------------------------------------------------------------------------- 857ARRAY_SIZE(arr) 858------------------------------------------------------------------------------- 859 860Returns the size of statically defined array, i.e. 861'(sizeof(arr) / sizeof(*arr))' 862 863[source,c] 864------------------------------------------------------------------------------- 865LTP_ALIGN(x, a) 866------------------------------------------------------------------------------- 867 868Aligns the x to be next multiple of a. The a must be power of 2. 869 870[source,c] 871------------------------------------------------------------------------------- 872TST_TO_STR(s) /* stringification */ 873TST_TO_STR_(s) /* macro expansion */ 874------------------------------------------------------------------------------- 875 876Macros for stringification. 877 8781.13 Filesystem type detection and skiplist 879~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 880 881Some tests are known to fail on certain filesystems (you cannot swap on TMPFS, 882there are unimplemented 'fcntl()' etc.). 883 884If your test needs to be skipped on certain filesystems use the 885'.skip_filesystems' field in the tst_test structure as follows: 886 887[source,c] 888------------------------------------------------------------------------------- 889#include "tst_test.h" 890 891static struct tst_test test = { 892 ... 893 .skip_filesystems = (const char *const []) { 894 "tmpfs", 895 "ramfs", 896 "nfs", 897 NULL 898 }, 899}; 900------------------------------------------------------------------------------- 901 902When the '.all_filesystems' flag is set the '.skip_filesystems' list is passed 903to the function that detects supported filesystems any listed filesystem is 904not included in the resulting list of supported filesystems. 905 906If test needs to adjust expectations based on filesystem type it's also 907possible to detect filesystem type at the runtime. This is preferably used 908when only subset of the test is not applicable for a given filesystem. 909 910NOTE: ext2, ext3 or ext4 in '.skip_filesystems' on tests which does *not* use 911 '.all_filesystems' needs to be defined as 'ext2/ext3/ext4'. The reason 912 is that it is hard to detect used filesystem due to overlapping the functionality. 913 OTOH tests which use '.skip_filesystems' *with* '.all_filesystems' can skip 914 only filesystems which are actually used in '.all_filesystems': ext2, ext3, 915 ext4, xfs, btrfs, vfat, exfat, ntfs, tmpfs (defined in 'fs_type_whitelist[]'). 916 It does not make sense to list other filesystems. 917 918 919[source,c] 920------------------------------------------------------------------------------- 921#include "tst_test.h" 922 923static void run(void) 924{ 925 ... 926 927 switch ((type = tst_fs_type("."))) { 928 case TST_NFS_MAGIC: 929 case TST_TMPFS_MAGIC: 930 case TST_RAMFS_MAGIC: 931 tst_brk(TCONF, "Subtest not supported on %s", 932 tst_fs_type_name(type)); 933 return; 934 break; 935 } 936 937 ... 938} 939------------------------------------------------------------------------------- 940 9411.14 Thread-safety in the LTP library 942~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 943 944It is safe to use library 'tst_res()' function in multi-threaded tests. 945 946Only the main thread must return from the 'test()' function to the test 947library and that must be done only after all threads that may call any library 948function has been terminated. That especially means that threads that may call 949'tst_brk()' must terminate before the execution of the 'test()' function 950returns to the library. This is usually done by the main thread joining all 951worker threads at the end of the 'test()' function. Note that the main thread 952will never get to the library code in a case that 'tst_brk()' was called from 953one of the threads since it will sleep at least in 'pthread_join()' on the 954thread that called the 'tst_brk()' till 'exit()' is called by 'tst_brk()'. 955 956The test-supplied cleanup function runs *concurrently* to the rest of the 957threads in a case that cleanup was entered from 'tst_brk()'. Subsequent 958threads entering 'tst_brk()' must be suspended or terminated at the start of 959the user supplied cleanup function. It may be necessary to stop or exit 960the rest of the threads before the test cleans up as well. For example threads 961that create new files should be stopped before temporary directory is be 962removed. 963 964Following code example shows thread safe cleanup function example using atomic 965increment as a guard. The library calls its cleanup after the execution returns 966from the user supplied cleanup and expects that only one thread returns from 967the user supplied cleanup to the test library. 968 969[source,c] 970------------------------------------------------------------------------------- 971#include "tst_test.h" 972 973static void cleanup(void) 974{ 975 static int flag; 976 977 if (tst_atomic_inc(&flag) != 1) 978 pthread_exit(NULL); 979 980 /* if needed stop the rest of the threads here */ 981 982 ... 983 984 /* then do cleanup work */ 985 986 ... 987 988 /* only one thread returns to the library */ 989} 990------------------------------------------------------------------------------- 991 992 9931.15 Testing with a block device 994~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 995 996Some tests needs a block device (inotify tests, syscall 'EROFS' failures, 997etc.). LTP library contains a code to prepare a testing device. 998 999If '.needs_device' flag in the 'struct tst_test' is set the 'tst_device' 1000structure is initialized with a path to a test device and default filesystem 1001to be used. 1002 1003You can also request minimal device size in megabytes by setting 1004'.dev_min_size' the device is guaranteed to have at least the requested size 1005then. 1006 1007If '.format_device' flag is set the device is formatted with a filesystem as 1008well. You can use '.dev_fs_type' to override the default filesystem type if 1009needed and pass additional options to mkfs via '.dev_fs_opts' and 1010'.dev_extra_opts' pointers. Note that '.format_device' implies '.needs_device' 1011there is no need to set both. 1012 1013If '.mount_device' is set, the device is mounted at '.mntpoint' which is used 1014to pass a directory name that will be created and used as mount destination. 1015You can pass additional flags and data to the mount command via '.mnt_flags' 1016and '.mnt_data' pointers. Note that '.mount_device' implies '.needs_device' 1017and '.format_device' so there is no need to set the later two. 1018 1019If '.needs_rofs' is set, read-only filesystem is mounted at '.mntpoint' this 1020one is supposed to be used for 'EROFS' tests. 1021 1022If '.all_filesystems' is set the test function is executed for all supported 1023filesystems. Supported filesystems are detected based on existence of the 1024'mkfs.$fs' helper and on kernel support to mount it. For each supported 1025filesystem the 'tst_device.fs_type' is set to the currently tested fs type, if 1026'.format_device' is set the device is formatted as well, if '.mount_device' is 1027set it's mounted at '.mntpoint'. Also the test timeout is reset for each 1028execution of the test function. This flag is expected to be used for filesystem 1029related syscalls that are at least partly implemented in the filesystem 1030specific code e.g. 'fallocate()'. 1031 1032[source,c] 1033------------------------------------------------------------------------------- 1034#include "tst_test.h" 1035 1036struct tst_device { 1037 const char *dev; 1038 const char *fs_type; 1039}; 1040 1041extern struct tst_device *tst_device; 1042 1043int tst_umount(const char *path); 1044------------------------------------------------------------------------------- 1045 1046In case that 'LTP_DEV' is passed to the test in an environment, the library 1047checks that the file exists and that it's a block device, if 1048'.device_min_size' is set the device size is checked as well. If 'LTP_DEV' 1049wasn't set or if size requirements were not met a temporary file is created 1050and attached to a free loop device. 1051 1052If there is no usable device and loop device couldn't be initialized the test 1053exits with 'TCONF'. 1054 1055The 'tst_umount()' function works exactly as 'umount(2)' but retries several 1056times on 'EBUSY'. This is because various desktop daemons (gvfsd-trash is known 1057for that) may be stupid enough to probe all newly mounted filesystem which 1058results in 'umount(2)' failing with 'EBUSY'. 1059 1060IMPORTANT: All testcases should use 'tst_umount()' instead of 'umount(2)' to 1061 umount filesystems. 1062 1063[source,c] 1064------------------------------------------------------------------------------- 1065#include "tst_test.h" 1066 1067int tst_find_free_loopdev(const char *path, size_t path_len); 1068------------------------------------------------------------------------------- 1069 1070This function finds a free loopdev and returns the free loopdev minor (-1 for no 1071free loopdev). If path is non-NULL, it will be filled with free loopdev path. 1072If you want to use a customized loop device, we can call 'tst_find_free_loopdev(NULL, 0)' 1073in tests to get a free minor number and then mknod. 1074 1075[source,c] 1076------------------------------------------------------------------------------- 1077#include "tst_test.h" 1078 1079unsigned long tst_dev_bytes_written(const char *dev); 1080------------------------------------------------------------------------------- 1081 1082This function reads test block device stat file ('/sys/block/<device>/stat') and 1083returns the bytes written since the last invocation of this function. To avoid 1084FS deferred IO metadata/cache interference, we suggest doing "syncfs" before the 1085tst_dev_bytes_written first invocation. And an inline function named 'tst_dev_sync()' 1086is created for that intention. 1087 1088[source,c] 1089------------------------------------------------------------------------------- 1090#include "tst_test.h" 1091 1092void tst_find_backing_dev(const char *path, char *dev, size_t dev_size); 1093------------------------------------------------------------------------------- 1094 1095This function finds the block dev that this path belongs to, using uevent in sysfs. 1096For Btrfs it uses '/sys/fs/btrfs/UUID/devices/DEV_NAME/uevent'; for other 1097filesystems it uses '/sys/dev/block/MAJOR:MINOR/uevent'. 1098 1099[source,c] 1100------------------------------------------------------------------------------- 1101#include "tst_test.h" 1102 1103uint64_t tst_get_device_size(const char *dev_path); 1104------------------------------------------------------------------------------- 1105 1106This function gets size of the given block device, it checks the 'dev_path' is 1107valid first, if yes, return the size in MB, otherwise return -1. 1108 1109[source,c] 1110------------------------------------------------------------------------------- 1111#include "tst_test.h" 1112 1113int tst_dev_block_size(const char *path); 1114------------------------------------------------------------------------------- 1115 1116This function returns the physical device block size for the specific `path`. 1117It finds the device where `path` is located and then uses `ioctl` (BLKSSZGET) 1118to get a physical device block size. 1119 11201.16 Formatting a device with a filesystem 1121~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1122 1123[source,c] 1124------------------------------------------------------------------------------- 1125#include "tst_test.h" 1126 1127static void setup(void) 1128{ 1129 ... 1130 SAFE_MKFS(tst_device->dev, tst_device->fs_type, NULL, NULL); 1131 ... 1132} 1133------------------------------------------------------------------------------- 1134 1135This function takes a path to a device, filesystem type and an array of extra 1136options passed to mkfs. 1137 1138The fs options 'fs_opts' should either be 'NULL' if there are none, or a 1139'NULL' terminated array of strings such as: 1140+const char *const opts[] = {"-b", "1024", NULL}+. 1141 1142The extra options 'extra_opts' should either be 'NULL' if there are none, or a 1143'NULL' terminated array of strings such as +{"102400", NULL}+; 'extra_opts' 1144will be passed after device name. e.g: +mkfs -t ext4 -b 1024 /dev/sda1 102400+ 1145in this case. 1146 1147Note that perfer to store the options which can be passed before or after device 1148name by 'fs_opts' array. 1149 11501.17 Verifying a filesystem's free space 1151~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1152 1153Some tests have size requirements for the filesystem's free space. If these 1154requirements are not satisfied, the tests should be skipped. 1155 1156[source,c] 1157------------------------------------------------------------------------------- 1158#include "tst_test.h" 1159 1160int tst_fs_has_free(const char *path, unsigned int size, unsigned int mult); 1161------------------------------------------------------------------------------- 1162 1163The 'tst_fs_has_free()' function returns 1 if there is enough space and 0 if 1164there is not. 1165 1166The 'path' is the pathname of any directory/file within a filesystem. 1167 1168The 'mult' is a multiplier, one of 'TST_BYTES', 'TST_KB', 'TST_MB' or 'TST_GB'. 1169 1170The required free space is calculated by 'size * mult', e.g. 1171'tst_fs_has_free("/tmp/testfile", 64, TST_MB)' will return 1 if the 1172filesystem, which '"/tmp/testfile"' is in, has 64MB free space at least, and 0 1173if not. 1174 11751.18 Files, directories and fs limits 1176~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1177 1178Some tests need to know the maximum count of links to a regular file or 1179directory, such as 'rename(2)' or 'linkat(2)' to test 'EMLINK' error. 1180 1181[source,c] 1182------------------------------------------------------------------------------- 1183#include "tst_test.h" 1184 1185int tst_fs_fill_hardlinks(const char *dir); 1186------------------------------------------------------------------------------- 1187 1188Try to get maximum count of hard links to a regular file inside the 'dir'. 1189 1190NOTE: This number depends on the filesystem 'dir' is on. 1191 1192This function uses 'link(2)' to create hard links to a single file until it 1193gets 'EMLINK' or creates 65535 links. If the limit is hit, the maximum number of 1194hardlinks is returned and the 'dir' is filled with hardlinks in format 1195"testfile%i", where i belongs to [0, limit) interval. If no limit is hit or if 1196'link(2)' failed with 'ENOSPC' or 'EDQUOT', zero is returned and previously 1197created files are removed. 1198 1199[source,c] 1200------------------------------------------------------------------------------- 1201#include "tst_test.h" 1202 1203int tst_fs_fill_subdirs(const char *dir); 1204------------------------------------------------------------------------------- 1205 1206Try to get maximum number of subdirectories in directory. 1207 1208NOTE: This number depends on the filesystem 'dir' is on. For current kernel, 1209subdir limit is not available for all filesystems (available for ext2, ext3, 1210minix, sysv and more). If the test runs on some other filesystems, like ramfs, 1211tmpfs, it will not even try to reach the limit and return 0. 1212 1213This function uses 'mkdir(2)' to create directories in 'dir' until it gets 1214'EMLINK' or creates 65535 directories. If the limit is hit, the maximum number 1215of subdirectories is returned and the 'dir' is filled with subdirectories in 1216format "testdir%i", where i belongs to [0, limit - 2) interval (because each 1217newly created dir has two links already - the '.' and the link from parent 1218dir). If no limit is hit or if 'mkdir(2)' failed with 'ENOSPC' or 'EDQUOT', 1219zero is returned and previously created directories are removed. 1220 1221[source,c] 1222------------------------------------------------------------------------------- 1223#include "tst_test.h" 1224 1225int tst_dir_is_empty(const char *dir, int verbose); 1226------------------------------------------------------------------------------- 1227 1228Returns non-zero if directory is empty and zero otherwise. 1229 1230Directory is considered empty if it contains only '.' and '..'. 1231 1232[source,c] 1233------------------------------------------------------------------------------- 1234#include "tst_test.h" 1235 1236void tst_purge_dir(const char *path); 1237------------------------------------------------------------------------------- 1238 1239Deletes the contents of given directory but keeps the directory itself. Useful 1240for cleaning up the temporary directory and mount points between test cases or 1241test iterations. Terminates the program with 'TBROK' on error. 1242 1243[source,c] 1244------------------------------------------------------------------------------- 1245#include "tst_test.h" 1246 1247int tst_fill_fd(int fd, char pattern, size_t bs, size_t bcount); 1248------------------------------------------------------------------------------- 1249 1250Fill a file with specified pattern using file descriptor. 1251 1252[source,c] 1253------------------------------------------------------------------------------- 1254#include "tst_test.h" 1255 1256int tst_prealloc_size_fd(int fd, size_t bs, size_t bcount); 1257------------------------------------------------------------------------------- 1258 1259Preallocate the specified amount of space using 'fallocate()'. Falls back to 1260'tst_fill_fd()' if 'fallocate()' fails. 1261 1262[source,c] 1263------------------------------------------------------------------------------- 1264#include "tst_test.h" 1265 1266int tst_fill_file(const char *path, char pattern, size_t bs, size_t bcount); 1267------------------------------------------------------------------------------- 1268 1269Creates/overwrites a file with specified pattern using file path. 1270 1271[source,c] 1272------------------------------------------------------------------------------- 1273#include "tst_test.h" 1274 1275int tst_prealloc_file(const char *path, size_t bs, size_t bcount); 1276------------------------------------------------------------------------------- 1277 1278Create/overwrite a file and preallocate the specified amount of space for it. 1279The allocated space will not be initialized to any particular content. 1280 12811.19 Getting an unused PID number 1282~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1283 1284Some tests require a 'PID', which is not used by the OS (does not belong to 1285any process within it). For example, kill(2) should set errno to 'ESRCH' if 1286it's passed such 'PID'. 1287 1288[source,c] 1289------------------------------------------------------------------------------- 1290#include "tst_test.h" 1291 1292pid_t tst_get_unused_pid(void); 1293------------------------------------------------------------------------------- 1294 1295Return a 'PID' value not used by the OS or any process within it. 1296 1297[source,c] 1298------------------------------------------------------------------------------- 1299#include "tst_test.h" 1300 1301int tst_get_free_pids(void); 1302------------------------------------------------------------------------------- 1303 1304Returns number of unused pids in the system. Note that this number may be 1305different once the call returns and should be used only for rough estimates. 1306 13071.20 Running executables 1308~~~~~~~~~~~~~~~~~~~~~~~~ 1309 1310[source,c] 1311------------------------------------------------------------------------------- 1312#include "tst_test.h" 1313 1314int tst_cmd(const char *const argv[], 1315 const char *stdout_path, 1316 const char *stderr_path, 1317 enum tst_cmd_flags flags); 1318------------------------------------------------------------------------------- 1319 1320'tst_cmd()' is a wrapper for 'vfork() + execvp()' which provides a way 1321to execute an external program. 1322 1323'argv[]' is a 'NULL' terminated array of strings starting with the program name 1324which is followed by optional arguments. 1325 1326'TST_CMD_PASS_RETVAL' enum 'tst_cmd_flags' makes 'tst_cmd()' 1327return the program exit code to the caller, otherwise 'tst_cmd()' exit the 1328tests on failure. 'TST_CMD_TCONF_ON_MISSING' check for program in '$PATH' and exit 1329with 'TCONF' if not found. 1330 1331In case that 'execvp()' has failed and the enum 'TST_CMD_PASS_RETVAL' flag was set, the 1332return value is '255' if 'execvp()' failed with 'ENOENT' and '254' otherwise. 1333 1334'stdout_path' and 'stderr_path' determine where to redirect the program 1335stdout and stderr I/O streams. 1336 1337The 'SAFE_CMD()' macro can be used automatic handling non-zero exits (exits 1338with 'TBROK') and 'ENOENT' (exits with 'TCONF'). 1339 1340.Example 1341[source,c] 1342------------------------------------------------------------------------------- 1343#include "tst_test.h" 1344 1345const char *const cmd[] = { "ls", "-l", NULL }; 1346 1347... 1348 /* Store output of 'ls -l' into log.txt */ 1349 tst_cmd(cmd, "log.txt", NULL, 0); 1350... 1351------------------------------------------------------------------------------- 1352 13531.21 Measuring elapsed time and helper functions 1354~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1355 1356[source,c] 1357------------------------------------------------------------------------------- 1358#include "tst_timer.h" 1359 1360void tst_timer_check(clockid_t clk_id); 1361 1362void tst_timer_start(clockid_t clk_id); 1363 1364void tst_timer_stop(void); 1365 1366struct timespec tst_timer_elapsed(void); 1367 1368long long tst_timer_elapsed_ms(void); 1369 1370long long tst_timer_elapsed_us(void); 1371 1372int tst_timer_expired_ms(long long ms); 1373------------------------------------------------------------------------------- 1374 1375The 'tst_timer_check()' function checks if specified 'clk_id' is supported and 1376exits the test with 'TCONF' otherwise. It's expected to be used in test 1377'setup()' before any resources that needs to be cleaned up are initialized, 1378hence it does not include a cleanup function parameter. 1379 1380The 'tst_timer_start()' marks start time and stores the 'clk_id' for further 1381use. 1382 1383The 'tst_timer_stop()' marks the stop time using the same 'clk_id' as last 1384call to 'tst_timer_start()'. 1385 1386The 'tst_timer_elapsed*()' returns time difference between the timer start and 1387last timer stop in several formats and units. 1388 1389The 'tst_timer_expired_ms()' function checks if the timer started by 1390'tst_timer_start()' has been running longer than ms milliseconds. The function 1391returns non-zero if timer has expired and zero otherwise. 1392 1393IMPORTANT: The timer functions use 'clock_gettime()' internally which needs to 1394 be linked with '-lrt' on older glibc. Please do not forget to add 1395 'LDLIBS+=-lrt' in Makefile. 1396 1397[source,c] 1398------------------------------------------------------------------------------- 1399#include "tst_test.h" 1400#include "tst_timer.h" 1401 1402static void setup(void) 1403{ 1404 ... 1405 tst_timer_check(CLOCK_MONOTONIC); 1406 ... 1407} 1408 1409static void run(void) 1410{ 1411 ... 1412 tst_timer_start(CLOCK_MONOTONIC); 1413 ... 1414 while (!tst_timer_expired_ms(5000)) { 1415 ... 1416 } 1417 ... 1418} 1419 1420struct tst_test test = { 1421 ... 1422 .setup = setup, 1423 .test_all = run, 1424 ... 1425}; 1426------------------------------------------------------------------------------- 1427 1428Expiration timer example usage. 1429 1430[source,c] 1431------------------------------------------------------------------------------- 1432long long tst_timespec_to_us(struct timespec t); 1433long long tst_timespec_to_ms(struct timespec t); 1434 1435struct timeval tst_us_to_timeval(long long us); 1436struct timeval tst_ms_to_timeval(long long ms); 1437 1438int tst_timespec_lt(struct timespec t1, struct timespec t2); 1439 1440struct timespec tst_timespec_add_us(struct timespec t, long long us); 1441 1442struct timespec tst_timespec_diff(struct timespec t1, struct timespec t2); 1443long long tst_timespec_diff_us(struct timespec t1, struct timespec t2); 1444long long tst_timespec_diff_ms(struct timespec t1, struct timespec t2); 1445 1446struct timespec tst_timespec_abs_diff(struct timespec t1, struct timespec t2); 1447long long tst_timespec_abs_diff_us(struct timespec t1, struct timespec t2); 1448long long tst_timespec_abs_diff_ms(struct timespec t1, struct timespec t2); 1449------------------------------------------------------------------------------- 1450 1451The first four functions are simple inline conversion functions. 1452 1453The 'tst_timespec_lt()' function returns non-zero if 't1' is earlier than 1454't2'. 1455 1456The 'tst_timespec_add_us()' function adds 'us' microseconds to the timespec 1457't'. The 'us' is expected to be positive. 1458 1459The 'tst_timespec_diff*()' functions returns difference between two times, the 1460't1' is expected to be later than 't2'. 1461 1462The 'tst_timespec_abs_diff*()' functions returns absolute value of difference 1463between two times. 1464 1465NOTE: All conversions to ms and us rounds the value. 1466 14671.22 Datafiles 1468~~~~~~~~~~~~~~ 1469 1470[source,c] 1471------------------------------------------------------------------------------- 1472#include "tst_test.h" 1473 1474static const char *const res_files[] = { 1475 "foo", 1476 "bar", 1477 NULL 1478}; 1479 1480static struct tst_test test = { 1481 ... 1482 .resource_files = res_files, 1483 ... 1484} 1485------------------------------------------------------------------------------- 1486 1487If the test needs additional files to be copied to the test temporary 1488directory all you need to do is to list their filenames in the 1489'NULL' terminated array '.resource_files' in the tst_test structure. 1490 1491When resource files is set test temporary directory is created automatically, 1492there is need to set '.needs_tmpdir' as well. 1493 1494The test library looks for datafiles first, these are either stored in a 1495directory called +datafiles+ in the +$PWD+ at the start of the test or in 1496+$LTPROOT/testcases/data/${test_binary_name}+. If the file is not found the 1497library looks into +$LTPROOT/testcases/bin/+ and to +$PWD+ at the start of the 1498test. This ensures that the testcases can copy the file(s) effortlessly both 1499when test is started from the directory it was compiled in as well as when LTP 1500was installed. 1501 1502The file(s) are copied to the newly created test temporary directory which is 1503set as the test working directory when the 'test()' functions is executed. 1504 15051.23 Code path tracing 1506~~~~~~~~~~~~~~~~~~~~~~ 1507 1508'tst_res' is a macro, so on when you define a function in one file: 1509 1510[source,c] 1511------------------------------------------------------------------------------- 1512int do_action(int arg) 1513{ 1514 ... 1515 1516 if (ok) { 1517 tst_res(TPASS, "check passed"); 1518 return 0; 1519 } else { 1520 tst_res(TFAIL, "check failed"); 1521 return -1; 1522 } 1523} 1524------------------------------------------------------------------------------- 1525 1526and call it from another file, the file and line reported by 'tst_res' in this 1527function will be from the former file. 1528 1529'TST_TRACE' can make the analysis of such situations easier. It's a macro which 1530inserts a call to 'tst_res(TINFO, ...)' in case its argument evaluates to 1531non-zero. In this call to 'tst_res(TINFO, ...)' the file and line will be 1532expanded using the actual location of 'TST_TRACE'. 1533 1534For example, if this another file contains: 1535 1536[source,c] 1537------------------------------------------------------------------------------- 1538#include "tst_test.h" 1539 1540if (TST_TRACE(do_action(arg))) { 1541 ... 1542} 1543------------------------------------------------------------------------------- 1544 1545the generated output may look similar to: 1546 1547------------------------------------------------------------------------------- 1548common.h:9: FAIL: check failed 1549test.c:8: INFO: do_action(arg) failed 1550------------------------------------------------------------------------------- 1551 15521.24 Tainted kernels 1553~~~~~~~~~~~~~~~~~~~~ 1554 1555If you need to detect whether a testcase triggers a kernel warning, bug or 1556oops, the following can be used to detect TAINT_W or TAINT_D: 1557 1558[source,c] 1559------------------------------------------------------------------------------- 1560#include "tst_test.h" 1561 1562static struct tst_test test = { 1563 ... 1564 .taint_check = TST_TAINT_W | TST_TAINT_D, 1565 ... 1566}; 1567 1568void run(void) 1569{ 1570 ... 1571 if (tst_taint_check() != 0) 1572 tst_res(TFAIL, "kernel has issues"); 1573 else 1574 tst_res(TPASS, "kernel seems to be fine"); 1575} 1576------------------------------------------------------------------------------- 1577 1578To initialize taint checks, you have to set the taint flags you want to test 1579for in the 'taint_check' attribute of the tst_test struct. LTP library will 1580then automatically call 'tst_taint_init()' during test setup. The function 1581will generate a 'TCONF' if the requested flags are not fully supported on the 1582running kernel, and 'TBROK' if the kernel is already tainted before executing 1583the test. 1584 1585LTP library will then automatically check kernel taint at the end of testing. 1586If '.all_filesystems' is set in struct tst_test, taint check will be performed 1587after each file system and taint will abort testing early with 'TFAIL'. You 1588can optionally also call 'tst_taint_check()' during 'run()', which returns 0 1589or the tainted flags set in '/proc/sys/kernel/tainted' as specified earlier. 1590 1591Depending on your kernel version, not all tainted-flags will be supported. 1592 1593For reference to tainted kernels, see kernel documentation: 1594Documentation/admin-guide/tainted-kernels.rst or 1595https://www.kernel.org/doc/html/latest/admin-guide/tainted-kernels.html 1596 15971.25 Checksums 1598~~~~~~~~~~~~~~ 1599 1600CRC32c checksum generation is supported by LTP. In order to use it, the 1601test should include 'tst_checksum.h' header, then can call 'tst_crc32c()'. 1602 16031.26 Checking kernel for the driver support 1604~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1605 1606Some tests may need specific kernel drivers, either compiled in, or built 1607as a module. If '.needs_drivers' points to a 'NULL' terminated array of kernel 1608module names these are all checked and the test exits with 'TCONF' on the 1609first missing driver. 1610 1611The detection is based on reading 'modules.dep' and 'modules.builtin' files 1612generated by kmod. The check is skipped on Android. 1613 16141.27 Saving & restoring /proc|sys values 1615~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1616 1617LTP library can be instructed to save and restore value of specified 1618(/proc|sys) files. This is achieved by initialized tst_test struct 1619field 'save_restore'. It is a NULL-terminated array of struct 1620'tst_path_val' where each tst_path_val.path represents a file, whose 1621value is saved at the beginning and restored at the end of the test. 1622If non-NULL string is passed in tst_path_val.val, it is written 1623to the respective file at the beginning of the test. Only the first line 1624of a specified file is saved and restored. 1625 1626By default, the test will end with TCONF if the file is read-only or 1627does not exist. If the optional write of new value fails, the test will end 1628with 'TBROK'. This behavior can be changed using tst_path_val.flags: 1629 1630* 'TST_SR_TBROK_MISSING' – End test with 'TBROK' if the file does not exist 1631* 'TST_SR_TCONF_MISSING' – End test with 'TCONF' if the file does not exist 1632* 'TST_SR_SKIP_MISSING' – Continue without saving the file if it does not exist 1633* 'TST_SR_TBROK_RO' – End test with 'TBROK' if the file is read-only 1634* 'TST_SR_TCONF_RO' – End test with 'TCONF' if the file is read-only 1635* 'TST_SR_SKIP_RO' – Continue without saving the file if it is read-only 1636* 'TST_SR_IGNORE_ERR' – Ignore errors when writing new value into the file 1637 1638Common flag combinations also have shortcuts: 1639 1640* 'TST_SR_TCONF' – Equivalent to 'TST_SR_TCONF_MISSING | TST_SR_TCONF_RO' 1641* 'TST_SR_TBROK' – Equivalent to 'TST_SR_TBROK_MISSING | TST_SR_TBROK_RO' 1642* 'TST_SR_SKIP' – Equivalent to 'TST_SR_SKIP_MISSING | TST_SR_SKIP_RO' 1643 1644'restore' is always strict and will TWARN if it encounters any error. 1645 1646[source,c] 1647------------------------------------------------------------------------------- 1648static struct tst_test test = { 1649 ... 1650 .setup = setup, 1651 .save_restore = (const struct tst_path_val[]) { 1652 {"/proc/sys/kernel/core_pattern", NULL, TST_SR_TCONF}, 1653 {"/proc/sys/user/max_user_namespaces", NULL, TST_SR_SKIP}, 1654 {"/sys/kernel/mm/ksm/run", "1", TST_SR_TBROK}, 1655 {} 1656 }, 1657}; 1658------------------------------------------------------------------------------- 1659 16601.28 Parsing kernel .config 1661~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1662 1663Generally testcases should attempt to autodetect as much kernel features as 1664possible based on the currently running kernel. We do have tst_check_driver() 1665to check if functionality that could be compiled as kernel module is present 1666on the system, disabled syscalls can be detected by checking for 'ENOSYS' 1667errno etc. 1668 1669However in rare cases core kernel features couldn't be detected based on the 1670kernel userspace API and we have to resort to parse the kernel .config. 1671 1672For this cases the test should set the 'NULL' terminated '.needs_kconfigs' 1673array of boolean expressions with constraints on the kconfig variables. The 1674boolean expression consits of variables, two binary operations '&' and '|', 1675negation '!' and correct sequence of parentesis '()'. Variables are expected 1676to be in a form of "CONFIG_FOO[=bar]". 1677 1678The test will continue to run if all expressions are evaluated to 'True'. 1679Missing variable is mapped to 'False' as well as variable with different than 1680specified value, e.g. 'CONFIG_FOO=bar' will evaluate to 'False' if the value 1681is anything else but 'bar'. If config variable is specified as plain 1682'CONFIG_FOO' it's evaluated to true it's set to any value (typically =y or =m). 1683 1684[source,c] 1685------------------------------------------------------------------------------- 1686#include "tst_test.h" 1687 1688static const char *kconfigs[] = { 1689 "CONFIG_X86_INTEL_UMIP | CONFIG_X86_UMIP", 1690 NULL 1691}; 1692 1693static struct tst_test test = { 1694 ... 1695 .needs_kconfigs = kconfigs, 1696 ... 1697}; 1698------------------------------------------------------------------------------- 1699 17001.29 Changing the Wall Clock Time during test execution 1701~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1702 1703There are some tests that, for different reasons, might need to change the 1704system-wide clock time. Whenever this happens, it is imperative that the clock 1705is restored, at the end of test's execution, taking in consideration the amount 1706of time elapsed during that test. 1707 1708In order for that to happen, struct tst_test has a variable called 1709"restore_wallclock" that should be set to "1" so LTP knows it should: (1) 1710initialize a monotonic clock during test setup phase and (2) use that monotonic 1711clock to fix the system-wide clock time at the test cleanup phase. 1712 1713[source,c] 1714------------------------------------------------------------------------------- 1715#include "tst_test.h" 1716 1717static void setup(void) 1718{ 1719 ... 1720} 1721 1722static void run(void) 1723{ 1724 ... 1725} 1726 1727struct tst_test test = { 1728 ... 1729 .setup = setup, 1730 .test_all = run, 1731 .restore_wallclock = 1, 1732 ... 1733}; 1734------------------------------------------------------------------------------- 1735 17361.30 Testing similar syscalls in one test 1737~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1738 1739In some cases kernel has several very similar syscalls that do either the same 1740or very similar job. This is most noticeable on i386 where we commonly have 1741two or three syscall versions. That is because i386 was first platform that 1742Linux was developed on and because of that most mistakes in API happened there 1743as well. However this is not limited to i386 at all, it's quite common that 1744version two syscall has added missing flags parameters or so. 1745 1746In such cases it does not make much sense to copy&paste the test code over and 1747over, rather than that the test library provides support for test variants. 1748The idea behind test variants is simple, we run the test several times each 1749time with different syscall variant. 1750 1751The implementation consist of test_variants integer that, if set, denotes number 1752of test variants. The test is then forked and executed test_variants times each 1753time with different value in global tst_variant variable. 1754 1755[source,c] 1756------------------------------------------------------------------------------- 1757#include "tst_test.h" 1758 1759static int do_foo(void) 1760{ 1761 switch (tst_variant) { 1762 case 0: 1763 return foo(); 1764 case 1: 1765 return syscall(__NR_foo); 1766 } 1767 1768 return -1; 1769} 1770 1771static void run(void) 1772{ 1773 ... 1774 1775 TEST(do_foo); 1776 1777 ... 1778} 1779 1780static void setup(void) 1781{ 1782 switch (tst_variant) { 1783 case 0: 1784 tst_res(TINFO, "Testing foo variant 1"); 1785 break; 1786 case 1: 1787 tst_res(TINFO, "Testing foo variant 2"); 1788 break; 1789 } 1790} 1791 1792struct tst_test test = { 1793 ... 1794 .setup = setup, 1795 .test_all = run, 1796 .test_variants = 2, 1797 ... 1798}; 1799------------------------------------------------------------------------------- 1800 18011.31 Guarded buffers 1802~~~~~~~~~~~~~~~~~~~~ 1803 1804The test library supports guarded buffers, which are buffers allocated so 1805that: 1806 1807* The end of the buffer is followed by a PROT_NONE page 1808 1809* The remainder of the page before the buffer is filled with random canary 1810 data 1811 1812Which means that the any access after the buffer will yield a Segmentation 1813fault or EFAULT depending on if the access happened in userspace or the kernel 1814respectively. The canary before the buffer will also catch any write access 1815outside of the buffer. 1816 1817The purpose of the patch is to catch off-by-one bugs which happens when 1818buffers and structures are passed to syscalls. New tests should allocate 1819guarded buffers for all data passed to the tested syscall which are passed by 1820a pointer. 1821 1822[source,c] 1823------------------------------------------------------------------------------- 1824#include "tst_test.h" 1825 1826static struct foo *foo_ptr; 1827static struct iovec *iov; 1828static void *buf_ptr; 1829static char *id; 1830... 1831 1832static void run(void) 1833{ 1834 ... 1835 1836 foo_ptr->bar = 1; 1837 foo_ptr->buf = buf_ptr; 1838 1839 ... 1840} 1841 1842static void setup(void) 1843{ 1844 ... 1845 1846 id = tst_strdup(string); 1847 1848 ... 1849} 1850 1851static struct tst_test test = { 1852 ... 1853 .bufs = (struct tst_buffers []) { 1854 {&foo_ptr, .size = sizeof(*foo_ptr)}, 1855 {&buf_ptr, .size = BUF_SIZE}, 1856 {&iov, .iov_sizes = (int[]){128, 32, -1}, 1857 {} 1858 } 1859}; 1860------------------------------------------------------------------------------- 1861 1862Guarded buffers can be allocated on runtime in a test setup() by a 1863'tst_alloc()' or by 'tst_strdup()' as well as by filling up the .bufs array in 1864the tst_test structure. 1865 1866So far the tst_test structure supports allocating either a plain buffer by 1867setting up the size or struct iovec, which is allocated recursively including 1868the individual buffers as described by an '-1' terminated array of buffer 1869sizes. 1870 18711.32 Adding and removing capabilities 1872~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1873 1874Some tests may require the presence or absence of particular 1875capabilities. Using the API provided by 'tst_capability.h' the test author can 1876try to ensure that some capabilities are either present or absent during the 1877test. 1878 1879For example; below we try to create a raw socket, which requires 1880CAP_NET_ADMIN. During setup we should be able to do it, then during run it 1881should be impossible. The LTP capability library will check before setup that 1882we have this capability, then after setup it will drop it. 1883 1884[source,c] 1885-------------------------------------------------------------------------------- 1886#include "tst_test.h" 1887#include "tst_capability.h" 1888#include "tst_safe_net.h" 1889 1890#include "lapi/socket.h" 1891 1892static void run(void) 1893{ 1894 TEST(socket(AF_INET, SOCK_RAW, 1)); 1895 if (TST_RET > -1) { 1896 tst_res(TFAIL, "Created raw socket"); 1897 } else if (TST_ERR != EPERM) { 1898 tst_res(TFAIL | TTERRNO, 1899 "Failed to create socket for wrong reason"); 1900 } else { 1901 tst_res(TPASS | TTERRNO, "Didn't create raw socket"); 1902 } 1903} 1904 1905static void setup(void) 1906{ 1907 TEST(socket(AF_INET, SOCK_RAW, 1)); 1908 if (TST_RET < 0) 1909 tst_brk(TCONF | TTERRNO, "We don't have CAP_NET_RAW to begin with"); 1910 1911 SAFE_CLOSE(TST_RET); 1912} 1913 1914static struct tst_test test = { 1915 .setup = setup, 1916 .test_all = run, 1917 .caps = (struct tst_cap []) { 1918 TST_CAP(TST_CAP_REQ, CAP_NET_RAW), 1919 TST_CAP(TST_CAP_DROP, CAP_NET_RAW), 1920 {} 1921 }, 1922}; 1923-------------------------------------------------------------------------------- 1924 1925Look at the test struct at the bottom. We have filled in the 'caps' field with 1926a 'NULL' terminated array containing two 'tst_cap' structs. 'TST_CAP_REQ' 1927actions are executed before setup and 'TST_CAP_DROP' are executed after 1928setup. This means it is possible to both request and drop a capability. 1929 1930[source,c] 1931-------------------------------------------------------------------------------- 1932static struct tst_test test = { 1933 .test_all = run, 1934 .caps = (struct tst_cap []) { 1935 TST_CAP(TST_CAP_REQ, CAP_NET_RAW), 1936 TST_CAP(TST_CAP_DROP, CAP_SYS_ADMIN), 1937 {} 1938 }, 1939}; 1940-------------------------------------------------------------------------------- 1941 1942Here we request 'CAP_NET_RAW', but drop 'CAP_SYS_ADMIN'. If the capability is 1943in the permitted set, but not the effective set, the library will try to 1944permit it. If it is not in the permitted set, then it will fail with 'TCONF'. 1945 1946This API does not require 'libcap' to be installed. However it has limited 1947features relative to 'libcap'. It only tries to add or remove capabilities 1948from the effective set. This means that tests which need to spawn child 1949processes may have difficulties ensuring the correct capabilities are 1950available to the children (see the capabilities (7) manual pages). 1951 1952However a lot of problems can be solved by using 'tst_cap_action(struct 1953tst_cap *cap)' directly which can be called at any time. This also helps if 1954you wish to drop a capability at the beginning of setup. 1955 19561.33 Reproducing race-conditions 1957~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1958 1959If a bug is caused by two tasks in the kernel racing and you wish to create a 1960regression test (or bug-fix validation test) then the 'tst_fuzzy_sync.h' 1961library should be used. 1962 1963It allows you to specify, in your code, two race windows. One window in each 1964thread's loop (triggering a race usually requires many iterations). These 1965windows show fuzzy-sync where the race can happen. They don't need to be 1966exact, hence the 'fuzzy' part. If the race condition is not immediately 1967triggered then the library will begin experimenting with different timings. 1968 1969[source,c] 1970-------------------------------------------------------------------------------- 1971#include "tst_fuzzy_sync.h" 1972 1973static struct tst_fzsync_pair fzsync_pair; 1974 1975static void setup(void) 1976{ 1977 tst_fzsync_pair_init(&fzsync_pair); 1978} 1979 1980static void cleanup(void) 1981{ 1982 tst_fzsync_pair_cleanup(&fzsync_pair); 1983} 1984 1985static void *thread_b(void *arg) 1986{ 1987 while (tst_fzsync_run_b(&fzsync_pair)) { 1988 1989 tst_fzsync_start_race_b(&fzsync_pair); 1990 1991 /* This is the race window for thread B */ 1992 1993 tst_fzsync_end_race_b(&fzsync_pair); 1994 } 1995 1996 return arg; 1997} 1998 1999static void thread_a(void) 2000{ 2001 tst_fzsync_pair_reset(&fzsync_pair, thread_b); 2002 2003 while (tst_fzsync_run_a(&fzsync_pair)) { 2004 2005 tst_fzsync_start_race_a(&fzsync_pair); 2006 2007 /* This is the race window for thread A */ 2008 2009 tst_fzsync_end_race_a(&fzsync_pair); 2010 } 2011} 2012 2013static struct tst_test test = { 2014 .test_all = thread_a, 2015 .setup = setup, 2016 .cleanup = cleanup, 2017}; 2018-------------------------------------------------------------------------------- 2019 2020Above is a minimal template for a test using fuzzy-sync. In a simple case, you 2021just need to put the bits you want to race inbetween 'start_race' and 2022'end_race'. Meanwhile, any setup you need to do per-iteration goes outside the 2023windows. 2024 2025Fuzzy sync synchronises 'run_a' and 'run_b', which act as barriers, so that 2026neither thread can progress until the other has caught up with it. There is 2027also the 'pair_wait' function which can be used to add barriers in other 2028locations. Of course 'start/end_race_a/b' are also a barriers. 2029 2030The library decides how long the test should run for based on the timeout 2031specified by the user plus some other heuristics. 2032 2033For full documentation see the comments in 'include/tst_fuzzy_sync.h'. 2034 20351.34 Reserving hugepages 2036~~~~~~~~~~~~~~~~~~~~~~~~ 2037 2038Many of the LTP tests need to use hugepage in their testing, this allows the 2039test can reserve hugepages from system via '.hugepages = {xx, TST_REQUEST}'. 2040 2041We achieved two policies for reserving hugepages: 2042 2043TST_REQUEST: 2044 It will try the best to reserve available huge pages and return the number 2045 of available hugepages in tst_hugepages, which may be 0 if hugepages are 2046 not supported at all. 2047 2048TST_NEEDS: 2049 This is an enforced requirement, LTP should strictly do hpages applying and 2050 guarantee the 'HugePages_Free' no less than pages which makes that test can 2051 use these specified numbers correctly. Otherwise, test exits with TCONF if 2052 the attempt to reserve hugepages fails or reserves less than requested. 2053 2054With success test stores the reserved hugepage number in 'tst_hugepages'. For 2055system without hugetlb supporting, variable 'tst_hugepages' will be set to 0. 2056If the hugepage number needs to be set to 0 on supported hugetlb system, please 2057use '.hugepages = {TST_NO_HUGEPAGES}'. 2058 2059Also, we do cleanup and restore work for the hpages resetting automatically. 2060 2061[source,c] 2062------------------------------------------------------------------------------- 2063#include "tst_test.h" 2064 2065static void run(void) 2066{ 2067 ... 2068 2069 if (tst_hugepages == test.hugepages.number) 2070 TEST(do_hpage_test); 2071 else 2072 ... 2073 ... 2074} 2075 2076struct tst_test test = { 2077 .test_all = run, 2078 .hugepages = {2, TST_REQUEST}, 2079 ... 2080}; 2081------------------------------------------------------------------------------- 2082 2083or, 2084 2085[source,c] 2086------------------------------------------------------------------------------- 2087#include "tst_test.h" 2088 2089static void run(void) 2090{ 2091 ... 2092} 2093 2094static void setup(void) 2095{ 2096 /* TST_NEEDS achieved this automatically in the library */ 2097 if (tst_hugepages != test.hugepages.number) 2098 tst_brk(TCONF, "..."); 2099} 2100 2101struct tst_test test = { 2102 .test_all = run, 2103 .hugepages = {2, TST_NEEDS}, 2104 ... 2105}; 2106------------------------------------------------------------------------------- 2107 21081.35 Checking for required commands 2109~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2110 2111Required commands can be checked with '.needs_cmds', which points to a 'NULL' 2112terminated array of strings such as: 2113 2114[source,c] 2115------------------------------------------------------------------------------- 2116.needs_cmds = (const char *const []) { 2117 "useradd", 2118 "userdel", 2119 NULL 2120}, 2121------------------------------------------------------------------------------- 2122 2123Also can check required command version whether is satisfied by using 'needs_cmds' 2124such as: 2125 2126[source,c] 2127------------------------------------------------------------------------------- 2128.needs_cmds = (const char *const []) { 2129 "mkfs.ext4 >= 1.43.0", 2130 NULL 2131}, 2132------------------------------------------------------------------------------- 2133 2134Currently, we only support mkfs.ext4 command version check. 2135If you want to support more commands, please fill your own .parser and .table_get 2136method in the version_parsers structure of lib/tst_cmd.c. 2137 21381.36 Assert sys or proc file value 2139~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2140Using TST_ASSERT_INT/STR(path, val) to assert that integer value or string stored in 2141the prefix field of file pointed by path equals to the value passed to this function. 2142 2143Also having a similar api pair TST_ASSERT_FILE_INT/STR(path, prefix, val) to assert 2144the field value of file. 2145 21461.37 Using Control Group 2147~~~~~~~~~~~~~~~~~~~~~~~~ 2148 2149Some LTP tests need specific Control Group configurations. 'tst_cgroup.h' 2150provides APIs to discover and use CGroups. There are many differences between 2151CGroups API V1 and V2. We encapsulate the details of configuring CGroups in 2152high-level functions which follow the V2 kernel API where possible. Allowing one 2153to write code that works on both V1 or V2. At least some of the time anyway; 2154often the behavioural differences between V1 and V2 are too great. In such cases 2155we revert to branching on the CGroup version. 2156 2157Also, the LTP library will automatically mount/umount and configure the CGroup 2158hierarchies if that is required (e.g. if you run the tests from init with no 2159system manager). 2160 2161[source,c] 2162------------------------------------------------------------------------------- 2163#include "tst_test.h" 2164 2165static void run(void) 2166{ 2167 ... 2168 // do test under cgroup 2169 ... 2170} 2171 2172static void setup(void) 2173{ 2174 SAFE_CG_PRINTF(tst_cg, "cgroup.procs", "%d", getpid()); 2175 SAFE_CG_PRINTF(tst_cg, "memory.max", "%lu", MEMSIZE); 2176 if (SAFE_CG_HAS(tst_cg, "memory.swap.max")) 2177 SAFE_CG_PRINTF(tst_cg, "memory.swap.max", "%zu", memsw); 2178} 2179 2180struct tst_test test = { 2181 .setup = setup, 2182 .test_all = run, 2183 .cleanup = cleanup, 2184 .needs_cgroup_ctrls = (const char *const []){ "memory", NULL }, 2185 ... 2186}; 2187------------------------------------------------------------------------------- 2188 2189Above, we first ensure the memory controller is available on the 2190test's CGroup with '.needs_cgroup_ctrls'. This populates a structure, 2191'tst_cg', which represents the test's CGroup. 2192 2193We then write the current processes PID into 'cgroup.procs', which 2194moves the current process into the test's CGroup. After which we set 2195the maximum memory size by writing to 'memory.max'. If the memory 2196controller is mounted on CGroups V1 then the library will actually 2197write to 'memory.limit_in_bytes'. As a general rule, if a file exists 2198on both CGroup versions, then we use the V2 naming. 2199 2200Some controller features, such as 'memory.swap', can be 2201disabled. Therefor we need to check if they exist before accessing 2202them. This can be done with 'SAFE_CG_HAS' which can be called on 2203any control file or feature. 2204 2205Most tests only require setting a few limits similar to the above. In 2206such cases the differences between V1 and V2 are hidden. Setup and 2207cleanup is also mostly hidden. However things can get much worse. 2208 2209[source,c] 2210------------------------------------------------------------------------------- 2211static struct tst_cg_group *cg_child; 2212 2213static void run(void) 2214{ 2215 char buf[BUFSIZ]; 2216 size_t mem = 0; 2217 2218 cg_child = tst_cg_group_mk(tst_cg, "child"); 2219 SAFE_CG_PRINTF(cg_child, "cgroup.procs", "%d", getpid()); 2220 2221 if (!TST_CG_VER_IS_V1(tst_cg, "memory")) 2222 SAFE_CG_PRINT(tst_cg, "cgroup.subtree_control", "+memory"); 2223 if (!TST_CG_VER_IS_V1(tst_cg, "cpuset")) 2224 SAFE_CG_PRINT(tst_cg, "cgroup.subtree_control", "+cpuset"); 2225 2226 if (!SAFE_FORK()) { 2227 SAFE_CG_PRINTF(cg_child, "cgroup.procs", "%d", getpid()); 2228 2229 if (SAFE_CG_HAS(cg_child, "memory.swap")) { 2230 SAFE_CG_SCANF(cg_child, 2231 "memory.swap.current", "%zu", &mem); 2232 } 2233 SAFE_CG_READ(cg_child, "cpuset.mems", buf, sizeof(buf)); 2234 2235 // Do something with cpuset.mems and memory.current values 2236 ... 2237 2238 exit(0); 2239 } 2240 2241 tst_reap_children(); 2242 SAFE_CG_PRINTF(tst_cg_drain, "cgroup.procs", "%d", getpid()); 2243 cg_child = tst_cg_group_rm(cg_child); 2244} 2245 2246static void cleanup(void) 2247{ 2248 if (cg_child) { 2249 SAFE_CG_PRINTF(tst_cg_drain, "cgroup.procs", "%d", getpid()); 2250 cg_child = tst_cg_group_rm(cg_child); 2251 } 2252} 2253 2254struct tst_test test = { 2255 .setup = setup, 2256 .test_all = run, 2257 .needs_cgroup_ctrls = (const char *const []){ 2258 "cpuset", 2259 "memory", 2260 NULL 2261 }, 2262 ... 2263}; 2264------------------------------------------------------------------------------- 2265 2266Starting with setup; we can see here that we fetch the 'drain' 2267CGroup. This is a shared group (between parallel tests) which may 2268contain processes from other tests. It should have default settings 2269and these should not be changed by the test. It can be used to remove 2270processes from other CGroups incase the hierarchy root is not 2271accessible. 2272 2273Note that 'tst_cg_get_drain_group' should not be called many times, 2274as it is allocated in a guarded buffer (See section 2.2.31). Therefor 2275it is best to call it once in 'setup' and not 'run' because 'run' may 2276be repeated with the '-i' option. 2277 2278In 'run', we first create a child CGroup with 'tst_cg_mk'. As we 2279create this CGroup in 'run' we should also remove it at the end of 2280run. We also need to check if it exists and remove it in cleanup as 2281well. Because there are 'SAFE_' functions which may jump to cleanup. 2282 2283We then move the main test process into the child CGroup. This is 2284important as it means that before we destroy the child CGroup we have 2285to move the main test process elsewhere. For that we use the 'drain' 2286group. 2287 2288Next we enable the memory and cpuset controller configuration on the 2289test CGroup's descendants (i.e. 'cg_child'). This allows each child to 2290have its own settings. The file 'cgroup.subtree_control' does not 2291exist on V1. Because it is possible to have both V1 and V2 active at 2292the same time. We can not simply check if 'subtree_control' exists 2293before writing to it. We have to check if a particular controller is 2294on V2 before trying to add it to 'subtree_control'. Trying to add a V1 2295controller will result in 'ENOENT'. 2296 2297We then fork a child process and add this to the child CGroup. Within 2298the child process we try to read 'memory.swap.current'. It is possible 2299that the memory controller was compiled without swap support, so it is 2300necessary to check if 'memory.swap' is enabled. That is unless the 2301test will never reach the point where 'memory.swap.*' are used without 2302swap support. 2303 2304The parent process waits for the child process to be reaped before 2305destroying the child CGroup. So there is no need to transfer the child 2306to drain. However the parent process must be moved otherwise we will 2307get 'EBUSY' when trying to remove the child CGroup. 2308 2309Another example of a behavioral difference between versions is shown below. 2310 2311[source,c] 2312------------------------------------------------------------------------------- 2313 if (TST_CG_VER_IS_V1(tst_cg, "memory")) 2314 SAFE_CG_PRINTF(tst_cg, "memory.swap.max", "%lu", ~0UL); 2315 else 2316 SAFE_CG_PRINT(tst_cg, "memory.swap.max", "max"); 2317------------------------------------------------------------------------------- 2318 2319CGroups V2 introduced a feature where 'memory[.swap].max' could be set to 2320"max". This does not appear to work on V1 'limit_in_bytes' however. For most 2321tests, simply using a large number is sufficient and there is no need to use 2322"max". Importantly though, one should be careful to read both the V1 and V2 2323kernel docs. Presently the LTP library does not attempt to handle most 2324differences in semantics. It does the minimal amount of work to make testing on 2325both V1 and V2 feasible. 2326 23271.38 Require minimum numbers of CPU for a testcase 2328~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2329 2330Some tests require more than specific number of CPU. It can be defined with 2331`.min_cpus = N`. 2332 23331.39 Require minimum memory or swap size for a testcase 2334~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2335 2336Some tests require at least size(MB) of free RAM or Swap. 2337 2338To make sure that test will run only on systems with more than minimal 2339required amount of RAM set `.min_mem_avail = N`. 2340 2341Similarily for tests that require certain amount of free Swap use 2342`.min_swap_avail = N`. 2343 23441.40 Test tags 2345~~~~~~~~~~~~~~ 2346 2347Test tags are name-value pairs that can hold any test metadata. 2348 2349We have additional support for CVE entries, git commit in mainline kernel, 2350stable kernel or glibc git repository. If a test is a regression test it 2351should include these tags. They are printed when test fails and exported 2352into documentation. 2353 2354CVE, mainline and stable kernel git commits in a regression test for a kernel bug: 2355[source,c] 2356------------------------------------------------------------------------------- 2357struct tst_test test = { 2358 ... 2359 .tags = (const struct tst_tag[]) { 2360 {"linux-git", "9392a27d88b9"}, 2361 {"linux-git", "ff002b30181d"}, 2362 {"known-fail", "ustat() is known to fail with EINVAL on Btrfs"}, 2363 {"linux-stable-git", "c4a23c852e80"}, 2364 {"CVE", "2020-29373"}, 2365 {} 2366 } 2367}; 2368------------------------------------------------------------------------------- 2369 2370NOTE: We don't track all backports to stable kernel but just those which are 2371 stable branch specific (unique), i.e. no commit in mainline. Example of 2372 commits: c4a23c852e80, cac68d12c531. 2373 2374Glibc and musl git commits in a regression test for glibc and musl bugs: 2375[source,c] 2376------------------------------------------------------------------------------- 2377struct tst_test test = { 2378 ... 2379 .tags = (const struct tst_tag[]) { 2380 {"glibc-git", "574500a108be"}, 2381 {"musl-git", "fa4a8abd06a4"}, 2382 {} 2383 } 2384}; 2385------------------------------------------------------------------------------- 2386 23871.41 Testing on the specific architecture 2388~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2389Testcases for specific arch should be limited on that only being supported 2390platform to run, we now involve a '.supported_archs' to achieve this feature 2391in LTP library. All you need to run a test on the expected arch is to set 2392the '.supported_archs' array in the 'struct tst_test' to choose the required 2393arch list. e.g. 2394 2395 .supported_archs = (const char *const []){"x86_64", "ppc64", NULL} 2396 2397This helps move the TCONF info from code to tst_test metadata as well. 2398 2399And, we also export a struct tst_arch to save the system architecture for 2400using in the whole test cases. 2401 2402 extern const struct tst_arch { 2403 char name[16]; 2404 enum tst_arch_type type; 2405 } tst_arch; 2406 2407[source,c] 2408------------------------------------------------------------------------------- 2409#include "tst_test.h" 2410 2411static struct tst_test test = { 2412 ... 2413 .setup = setup, 2414 .supported_archs = (const char *const []) { 2415 "x86_64", 2416 "ppc64", 2417 "s390x", 2418 NULL 2419 }, 2420}; 2421------------------------------------------------------------------------------- 2422 24231.42 Skipping test based on system state 2424~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2425Test can be skipped on various conditions: on enabled SecureBoot 2426('.skip_in_secureboot = 1'), lockdown ('.skip_in_lockdown = 1') or in 32-bit 2427compat mode ('.skip_in_compat = 1'). 2428 24292. Common problems 2430------------------ 2431 2432This chapter describes common problems/misuses and less obvious design patters 2433(quirks) in UNIX interfaces. Read it carefully :) 2434 24352.1 umask() 2436~~~~~~~~~~~ 2437 2438I've been hit by this one several times already... When you create files 2439with 'open()' or 'creat()' etc, the mode specified as the last parameter *is 2440not* the mode the file is created with. The mode depends on current 'umask()' 2441settings which may clear some of the bits. If your test depends on specific 2442file permissions you need either to change umask to 0 or 'chmod()' the file 2443afterwards or use 'SAFE_TOUCH()' that does the 'chmod()' for you. 2444 24452.2 access() 2446~~~~~~~~~~~~ 2447 2448If 'access(some_file, W_OK)' is executed by root, it will return success even 2449if the file doesn't have write permission bits set (the same holds for R_OK 2450too). For sysfs files you can use 'open()' as a workaround to check file 2451read/write permissions. It might not work for other filesystems, for these you 2452have to use 'stat()', 'lstat()' or 'fstat()'. 2453 24542.3 umount() EBUSY 2455~~~~~~~~~~~~~~~~~~ 2456 2457Various desktop daemons (gvfsd-trash is known for that) may be stupid enough 2458to probe all newly mounted filesystem which results in 'umount(2)' failing 2459with 'EBUSY'; use 'tst_umount()' described in 1.19 that retries in this case 2460instead of plain 'umount(2)'. 2461 24622.4 FILE buffers and fork() 2463~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2464 2465Be vary that if a process calls 'fork(2)' the child process inherits open 2466descriptors as well as copy of the parent memory so especially if there are 2467any open 'FILE' buffers with a data in them they may be written both by the 2468parent and children resulting in corrupted/duplicated data in the resulting 2469files. 2470 2471Also open 'FILE' streams are flushed and closed at 'exit(3)' so if your 2472program works with 'FILE' streams, does 'fork(2)', and the child may end up 2473calling 'exit(3)' you will likely end up with corrupted files. 2474 2475The solution to this problem is either simply call 'fflush(NULL)' that flushes 2476all open output 'FILE' streams just before doing 'fork(2)'. You may also use 2477'_exit(2)' in child processes which does not flush 'FILE' buffers and also 2478skips 'atexit(3)' callbacks. 2479