• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1C Test Case Tutorial
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-API[C Test API].
7
8This is a step-by-step tutorial on writing a simple C LTP test, where topics
9of the LTP and Linux kernel testing will be introduced gradually using a
10concrete example. Most sections will include exercises, some trivial and
11others not so much. If you find an exercise is leading you off at too much of
12a tangent, just leave it for later and move on.
13
14LTP tests can be written in C or Shell script. This tutorial is only for tests
15written in C using the new LTP test API. Note that while we go into some
16detail on using Git, this is not intended as a canonical or complete guide
17for Git.
18
190. Assumptions & Feedback
20-------------------------
21
22We assume the reader is familiar with C, Git and common Unix/Linux/GNU tools
23and has some general knowledge of Operating Systems. Experienced Linux
24developers may find it too verbose while people new to system level Linux
25development may find it overwhelming.
26
27Comments and feedback are welcome, please direct them to the mailing list (see
28+README+).
29
301. Getting Started
31------------------
32
33Git-clone the main LTP repository as described in the +README+ and change
34directory to the checked-out Git repository. We recommend installing the LTP
35and running one of the tests mentioned in the Quick guide (in the +README+) to
36ensure you are starting from a good state.
37
38We also recommended cloning the Linux kernel repository for reference, this
39guide will refer to files and directories within the mainline kernel 4.12.
40
41[source,shell]
42------------------------------------------------------------------------------
43$ git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
44------------------------------------------------------------------------------
45
46There are a number of other repositories which are useful for reference as
47well, including the GNU C library +glibc+ and the alternative C library
48+musl+. Some system calls are partially or even entirely implemented in user
49land as part of the standard C library. So in these cases, the C library is an
50important reference. +glibc+ is the most common C library for Linux, however
51+musl+ is generally easier to understand.
52
53How system calls are implemented varies from one architecture to another and
54across kernel and C library versions. To find out whether a system call is
55actually accessing the kernel (whether it is actually a system call) on any
56given machine you can use the +strace+ utility. This intercepts system calls
57made by an executable and prints them. We will use this later in the tutorial.
58
592. Choose a System Call to test
60-------------------------------
61
62We will use the +statx()+ system call, to provide a concrete example of a
63test. At the time of writing there is no test for this call which was
64introduced in Linux kernel version 4.11.
65
66Linux system call specific tests are primarily contained in
67+testcases/kernel/syscalls+, but you should also +git grep+ the entire LTP
68repository to check for any existing usages of a system call.
69
70One way to find a system call which is not currently tested by the LTP is to
71look at +include/linux/syscalls.h+ in the kernel tree.
72
73Something the LTP excels at is ensuring bug-fixes are back ported to
74maintenance releases, so targeting a specific regression is another
75option.
76
772.1. Find an untested System call
78~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
79
80Try to find an untested system call which has a manual page (i.e. +man
81syscall+ produces a result). It is a good idea to Git-clone the latest kernel
82man-pages repository.
83
84[source,shell]
85------------------------------------------------------------------------------
86$ git clone git://git.kernel.org/pub/scm/docs/man-pages/man-pages.git
87------------------------------------------------------------------------------
88
89At the time of writing the difference between the latest man-pages release and
90the +HEAD+ of the repository (usually the latest commit) is well over 100
91commits. This represents about 9 weeks of changes. If you are using a stable
92Linux distribution, your man-pages package may well be years old. So as with
93the kernel, it is best to have the Git repository as a reference.
94
95You could also find a system call with untested parameters or use whatever it
96is you are planning to use the LTP for.
97
983. Create the test skeleton
99---------------------------
100
101I shall call my test +statx01.c+, by the time you read this that file name
102will probably be taken, so increment the number in the file name as
103appropriate or replace +statx+ with the system call chosen in exercise 2.1.
104
105[source,shell]
106------------------------------------------------------------------------------
107$ mkdir testcases/kernel/syscalls/statx
108$ cd testcases/kernel/syscalls/statx
109$ echo statx >> .gitignore
110------------------------------------------------------------------------------
111
112Next open +statx01.c+ and add the following boilerplate. Make sure to change
113the copy right notice to your name/company, correct the test name and minimum
114kernel version if necessary. I will explain what the code does below.
115
116[source,c]
117------------------------------------------------------------------------------
118// SPDX-License-Identifier: GPL-2.0-or-later
119/*
120 * Copyright (c) 2017 Instruction Ignorer <"can't"@be.bothered.com>
121 */
122
123/*
124 * Test statx
125 *
126 * All tests should start with a description of _what_ we are testing.
127 * Non-trivial explanations of _how_ the code works should also go here.
128 * Include relevant links, Git commit hashes and CVE numbers.
129 * Inline comments should be avoided.
130 */
131
132#include "tst_test.h"
133
134static void run(void)
135{
136	tst_res(TPASS, "Doing hardly anything is easy");
137}
138
139static struct tst_test test = {
140	.test_all = run,
141	.min_kver = "4.11",
142};
143------------------------------------------------------------------------------
144
145Starting with the +#include+ statement we copy in the main LTP test library
146headers. This includes the most common test API functions and the test harness
147initialisation code. It is important to note that this is a completely
148ordinary, independent C program, however +main()+ is missing because it is
149implemented in +tst_test.h+.
150
151We specify what code we want to run as part of the test using the +tst_test
152test+ structure. Various callbacks can be set by the test writer, including
153+test.test_all+, which we have set to +run()+. The test harness will execute
154this callback in a separate process (using +fork()+), forcibly terminating it
155if it does not return after +test.timeout+ seconds.
156
157We have also set +test.min_kver+ to the kernel version where +statx+ was
158introduced. The test library will determine the kernel version at runtime. If
159the version is less than 4.11 then the test harness will return +TCONF+,
160indicating that this test is not suitable for the current system
161configuration.
162
163Occasionally features are back ported to older kernel versions, so +statx+ may
164exist on kernels with a lower version. However we don't need to worry about
165that unless there is evidence of it happening.
166
167As mentioned in the code itself, you should specify what you are testing and
168the expected outcome, even if it is relatively simple. If your program flow is
169necessarily complex and difficult to understand (which is often the case when
170trying to manipulate the kernel into doing something bad), then a detailed
171explanation of how the code works is welcome.
172
173What you should not do, is use inline comments or include the same level of
174explanation which is written here. As a general rule, if something is easy to
175document, then the code should also be easy to read. So don't document the easy
176stuff (except for the basic test specification).
177
178Before continuing we should compile this and check that the basics work. In
179order to compile the test we need a +Makefile+ in the same subdirectory. If
180one already exists, then nothing needs to be done, otherwise add one with the
181following contents.
182
183[source,make]
184------------------------------------------------------------------------------
185# SPDX-License-Identifier: GPL-2.0-or-later
186# Copyright (c) 2019 Linux Test Project
187
188top_srcdir		?= ../../../..
189
190include $(top_srcdir)/include/mk/testcases.mk
191
192include $(top_srcdir)/include/mk/generic_leaf_target.mk
193
194------------------------------------------------------------------------------
195
196This will automatically add +statx01.c+ as a build target producing a
197+statx01+ executable. Unless you have heavily deviated from the tutorial, and
198probably need to change +top_srcdir+, nothing else needs to be done.
199
200Normally, if you were starting a Makefile from scratch, then you would need to
201add +statx01+ as a build target. Specifying that you would like to run some
202program (e.g. +gcc+ or +clang+) to transform +statx01.c+ into +statx01+. Here
203we don't need to do that, but sometimes it is still necessary. For example, if
204we needed to link to the POSIX threading library, then we could add the
205following line after +testcases.mk+.
206
207[source,make]
208--------------------------------------------------------------------------------
209statx01: CFLAGS += -pthread
210--------------------------------------------------------------------------------
211
212Assuming you are in the test's subdirectory +testcases/kernel/syscalls/statx+,
213do
214
215[source,shell]
216--------------------------------------------------------------------------------
217$ make
218$ ./statx01
219--------------------------------------------------------------------------------
220
221This should build the test and then run it. However, even though the test is
222in the +syscalls+ directory it won't be automatically ran as part of the
223_syscalls_ test group (remember +./runltp -f syscalls+ from the +README+?). For
224this we need to add it to the +runtest+ file. So open +runtest/statx+ and add
225the lines starting with a +++.
226
227[source,diff]
228--------------------------------------------------------------------------------
229 statvfs01 statvfs01
230 statvfs02 statvfs02
231
232+statx01 statx01
233+
234 stime01 stime01
235 stime02 stime02
236
237--------------------------------------------------------------------------------
238
239The +runtest+ files are in a two column format. The first column is the test
240name, which is mainly used by test runners for reporting and filtering. It is
241just a single string of text with no spaces. The second column, which can
242contain spaces, is passed to the shell in order to execute the test. Often it
243is just the executable name, but some tests also take arguments (the LTP has a
244library for argument parsing, by the way).
245
246If you haven't done so already, we should add all these new files to Git. It
247is vitally important that you do not make changes to the master branch. If you
248do then pulling changes from upstream becomes a major issue. So first of all
249create a new branch.
250
251[source,shell]
252--------------------------------------------------------------------------------
253$ git checkout -b statx01 master
254--------------------------------------------------------------------------------
255
256Now we want to add the files we have created or modified, but before doing a
257commit make sure you have configured Git correctly. You need to at least set
258your Name and e-mail address in +~/.gitconfig+, but there are some other
259settings which come in handy too. My relatively simple configuration is similar to
260the below
261
262[source,conf]
263--------------------------------------------------------------------------------
264[user]
265	name = Sarah Jane
266	email = sjane@e-mail.address
267[core]
268	editor = emacs
269[sendemail]
270	smtpServer = smtp.server.address
271--------------------------------------------------------------------------------
272
273Obviously you need to at least change your name and e-mail. The SMTP server is
274useful for +git send-mail+, which we will discuss later. The editor value is
275used for things like writing commits (without the +-m+ option).
276
277[source,shell]
278--------------------------------------------------------------------------------
279$ git add -v :/testcases/kernel/syscalls/statx :/runtest/syscalls
280$ git commit -m "statx01: Add new test for statx syscall"
281--------------------------------------------------------------------------------
282
283This should add all the new files in the +statx+ directory and the +runtest+
284file. It is good practice to commit early and often. Later on we will do a
285Git-rebase, which allows us to clean up the commit history. So don't worry
286about how presentable your commit log is for now. Also don't hesitate to
287create a new branch when doing the exercises or experimenting. This will allow
288you to diverge from the tutorial and then easily come back again.
289
290I can't emphasize enough that Git makes things easy through branching and that
291things quickly get complicated if you don't do it. However if you do get into
292a mess, Git-reflog and Git-reset, will usually get you out of it. If you also
293mess that up then it may be possible to cherry pick 'dangling' commits out of
294the database into a branch.
295
2963.1 Report TCONF instead of TPASS
297~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
298
299Maybe the test should report "TCONF: Not implemented" instead or perhaps
300+TBROK+. Try changing it do so (see +doc/test-writing-guidelines.txt+ or
301https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines[the
302Wiki]).
303
3043.2 Check Git ignores the executable
305~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
306
307Is your +.gitignore+ correct?
308
3093.3 Run make check
310~~~~~~~~~~~~~~~~~~
311
312Check coding style with `make check`
313 (more in https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines#21-c-coding-style[C coding style])
314
3153.4 Install the LTP and run the test with runtest
316~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
317
318Run +statx01+ on its own; similar to the +madvise+ tests in the +README+.
319
3204. Call the system call
321-----------------------
322
323At the time of writing +statx+ has no +glibc+ wrapper. It is also fairly common
324for a distribution's C library version to be older than its kernel or it may use a
325cut down C library in comparison to the GNU one. So we must call +statx()+
326using the general +syscall()+ interface.
327
328The LTP contains a library for dealing with the +syscall+ interface, which is
329located in +include/lapi+. System call numbers are listed against the relevant
330call in the +*.in+ files (e.g. +x86_64.in+) which are used to generate
331+syscalls.h+, which is the header you should include. On rare occasions you
332may find the system call number is missing from the +*.in+ files and will need
333to add it (see +include/lapi/syscalls/strip_syscall.awk+).
334
335System call numbers vary between architectures, hence why there are multiple
336+*.in+ files for each architecture. You can find the various values for the
337+statx+ system call across a number of +uinstd.h+ files in the Linux kernel.
338
339Note that we don't use the system-call-identifier value available in
340+/usr/include/linux/uinstd.h+ because the kernel might be much newer than the
341user land development packages.
342
343For +statx+ we had to add +statx 332+ to +testcases/kernel/include/x86_64.in+,
344+statx 383+ to +testcases/kernel/include/powerpc.in+, etc.  Now lets look at
345the code, which I will explain in more detail further down.
346
347[source,c]
348--------------------------------------------------------------------------------
349/*
350 * Test statx
351 *
352 * Check if statx exists and what error code it returns when we give it dodgy
353 * data.
354 */
355
356#include <stdint.h>
357#include "tst_test.h"
358#include "lapi/syscalls.h"
359
360struct statx_timestamp {
361	int64_t	       tv_sec;
362	uint32_t       tv_nsec;
363	int32_t	       __reserved;
364};
365
366struct statx {
367	uint32_t	stx_mask;
368	uint32_t	stx_blksize;
369	uint64_t	stx_attributes;
370	uint32_t	stx_nlink;
371	uint32_t	stx_uid;
372	uint32_t	stx_gid;
373	uint16_t	stx_mode;
374	uint16_t	__spare0[1];
375	uint64_t	stx_ino;
376	uint64_t	stx_size;
377	uint64_t	stx_blocks;
378	uint64_t	stx_attributes_mask;
379	struct statx_timestamp	stx_atime;
380	struct statx_timestamp	stx_btime;
381	struct statx_timestamp	stx_ctime;
382	struct statx_timestamp	stx_mtime;
383	uint32_t	stx_rdev_major;
384	uint32_t	stx_rdev_minor;
385	uint32_t	stx_dev_major;
386	uint32_t	stx_dev_minor;
387	uint64_t	__spare2[14];
388};
389
390static int sys_statx(int dirfd, const char *pathname, int flags,
391		     unsigned int mask, struct statx *statxbuf)
392{
393	return tst_syscall(__NR_statx, dirfd, pathname, flags, mask, statxbuf);
394}
395
396...
397--------------------------------------------------------------------------------
398
399So the top part of the code is now boiler plate for calling +statx+. It is
400common for the kernel to be newer than the user land libraries and headers. So
401for new system calls like +statx+, we copy, with a few modifications, the
402relevant definitions into the LTP. This is somewhat like 'vendoring', although
403we are usually just copying headers required for interacting with the Kernel's
404ABI (Application Binary Interface), rather than internalising actual
405functionality.
406
407So from the top we include the +stdint.h+ library which gives us the standard
408+(u)int*_t+ type definitions. We use these in place of the Kernel type
409definitions such as +__u64+ in +linux/types.h+. We then have a couple of
410structure definitions which form part of the +statx+ API. These were copied
411from +include/uapi/linux/stat.h+ in the Kernel tree.
412
413After that, there is a wrapper function, which saves us from writing
414+tst_syscall(__NR_statx, ...+, every time we want to make a call to
415+statx+. This also provides a stub for when +statx+ is eventually integrated
416into the LTP library and also implemented by the C library. At that point we
417can switch to using the C library implementation if available or fallback to
418our own.
419
420The advantage of using the C library implementation is that it will often be
421better supported across multiple architectures. It will also mean we are using
422the system call in the same way most real programs would. Sometimes there are
423advantages to bypassing the C library, but in general it should not be our
424first choice.
425
426The final test should do a check during configuration (i.e. when we run
427+./configure+ before building) which checks if the +statx+ system call and
428associated structures exists. This requires writing an +m4+ file for use with
429+configure.ac+ which is processed during +make autotools+ and produces the
430configure script.
431
432For the time being though we shall just ignore this. All you need to know for
433now is that this is a problem which eventually needs to be dealt with and that
434there is a system in place to handle it.
435
436[source,c]
437--------------------------------------------------------------------------------
438...
439
440static void run(void)
441{
442	struct statx statxbuf = { 0 };
443
444	TEST(sys_statx(0, NULL, 0, 0, &statxbuf));
445
446	if (TST_RET == 0)
447		tst_res(TFAIL, "statx thinks it can stat NULL");
448	else if (TST_ERR == EFAULT)
449		tst_res(TPASS, "statx set errno to EFAULT as expected");
450	else
451		tst_res(TFAIL | TERRNO, "statx set errno to some unexpected value");
452}
453
454static struct tst_test test = {
455	.test_all = run,
456	.min_kver = "4.11",
457};
458--------------------------------------------------------------------------------
459
460The +TEST+ macro sets +TST_RET+ to the return value of +tst_statx()+ and
461+TST_ERR+ to the value of +errno+ immediately after the functions
462return. This is mainly just for convenience, although it potentially could
463have other uses.
464
465We check whether the return value indicates success and if it doesn't also
466check the value of +errno+. The last call to +tst_res+ includes +TERRNO+,
467which will print the current error number and associated description in
468addition to the message we have provided. Note that it uses the current value
469of +errno+ not +TST_ERR+.
470
471What we should have done in the example above is use +TTERRNO+ which takes the
472value of +TST_ERR+.
473
474If we try to run the test on a kernel where +statx+ does not exist, then
475+tst_syscall+ will cause it to fail gracefully with +TCONF+. Where +TCONF+
476indicates the test is not applicable to our configuration.
477
478The function +tst_syscall+ calls +tst_brk(TCONF,...)+ on failure. +tst_brk+
479causes the test to exit immediately, which prevents any further test code from
480being run.
481
4824.1 What are the differences between tst_brk and tst_res?
483~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
484
485See +include/tst_test.h+ and the
486https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines[test
487writing guide]. Also what do they have in common?
488
4894.2 What happens if you call tst_res(TINFO, ...) after sys_statx?
490~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
491
492Does the test still function correctly?
493
4944.3 Extend the test to handle other basic error conditions.
495~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
496
497For example, see if you can trigger +ENOENT+ instead. You shouldn't
498have to create any files, which is discussed in the next section.
499
5005. Setup, Cleanup and files
501---------------------------
502
503Some tests require resources to be allocated, or system settings to be
504changed, before the test begins. This 'setup' only has to be done once at the
505beginning and at the end of the test needs to be removed or reverted. The
506'cleanup' also has to be done regardless of whether the test breaks.
507
508Fortunately, like most test libraries, we have setup and cleanup (teardown)
509callbacks. +setup+ is called once before +run+ and +cleanup+ is called once
510afterwards. Note that +run+ itself can be called multiple times by the test
511harness, but that +setup+ and +cleanup+ are only called once.
512
513If either your code, a +SAFE_*+ macro or a library function such as
514+tst_syscall+ call +tst_brk+, then +run+ will exit immediately and the
515+cleanup+ function is then called. Once 'cleanup' is completed, the test
516executable will then exit altogether abandoning any remaining iterations of
517+run+.
518
519For +statx+ we would like to create some files or file like objects which we
520have control over. Deciding where to create the files is easy, we just create
521it in the current working directory and let the LTP test harness handle where
522that should be by setting +.needs_tmpdir = 1+.
523
524[source,c]
525--------------------------------------------------------------------------------
526/*
527 * Test statx
528 *
529 * Check if statx exists and what error code it returns when we give it dodgy
530 * data. Then stat a file and check it returns success.
531 */
532
533#include <stdint.h>
534#include "tst_test.h"
535#include "lapi/syscalls.h"
536#include "lapi/fcntl.h"
537
538#define FNAME "file_to_stat"
539#define STATX_BASIC_STATS 0x000007ffU
540
541/*************** statx structure and wrapper goes here ! ***************/
542
543...
544--------------------------------------------------------------------------------
545
546We have added an extra include +lapi/fcntl.h+ which wraps the system header by
547the same name (+#include <fcntl.h>+). This header ensures we have definitions
548for recently added macros such as +AT_FDCWD+ by providing fall backs if the
549system header does not have them. The +lapi+ directory contains a number of
550headers like this.
551
552At some point we may wish to add +lapi/stat.h+ to provide a fall back for
553macros such as +STATX_BASIC_STATS+. However for the time being we have just
554defined it in the test.
555
556[source,c]
557--------------------------------------------------------------------------------
558...
559
560static void setup(void)
561{
562	SAFE_TOUCH(FNAME, 0777, NULL);
563}
564
565static void run(void)
566{
567	struct statx statxbuf = { 0 };
568
569	TEST(sys_statx(0, NULL, 0, 0, &statxbuf));
570	if (TST_RET == 0)
571		tst_res(TFAIL, "statx thinks it can stat NULL");
572	else if (TST_ERR == EFAULT)
573		tst_res(TPASS, "statx set errno to EFAULT as expected");
574	else
575		tst_res(TFAIL | TERRNO, "statx set errno to some unexpected value");
576
577	TEST(sys_statx(AT_FDCWD, FNAME, 0, STATX_BASIC_STATS, &statxbuf));
578	if (TST_RET == 0)
579		tst_res(TPASS, "It returned zero so it must have worked!");
580	else
581		tst_res(TFAIL | TERRNO, "statx can not stat a basic file");
582}
583
584static struct tst_test test = {
585	.setup = setup,
586	.test_all = run,
587	.min_kver = "4.11",
588	.needs_tmpdir = 1
589};
590--------------------------------------------------------------------------------
591
592The +setup+ callback uses one of the LTP's +SAFE+ functions to create an empty
593file +file_to_stat+. Because we have set +.needs_tmpdir+, we can just create
594this file in the present working directory. We don't need to create a
595+cleanup+ callback yet because the LTP test harness will recursively delete
596the temporary directory and its contents.
597
598The +run+ function can be called multiple times by the test harness, however
599+setup+ and +cleanup+ callbacks will only be ran once.
600
601[WARNING]
602By this point you may have begun to explore the LTP library headers or older
603tests. In which case you will have come across functions from the old API such
604as +tst_brkm+. The old API is being phased out, so you should not use these
605functions.
606
607So far we haven't had to do any clean up. So our example doesn't answer the
608question "what happens if part of the clean up fails?". To answer this we are
609going to modify the test to ask the (highly contrived) question "What happens
610if I create and open a file, then create a hard-link to it, then call open
611again on the hard-link, then 'stat' the file".
612
613[source,c]
614--------------------------------------------------------------------------------
615#define LNAME "file_to_stat_link"
616
617...
618
619static void setup(void)
620{
621	fd = SAFE_OPEN(FNAME, O_CREAT, 0777);
622	SAFE_LINK(FNAME, LNAME);
623	lfd = SAFE_OPEN(LNAME, 0);
624}
625
626static void cleanup(void)
627{
628	if (lfd != 0)
629		SAFE_CLOSE(lfd);
630
631	if (fd != 0)
632		SAFE_CLOSE(fd);
633}
634
635static void run(void)
636{
637        ...
638
639	TEST(sys_statx(AT_FDCWD, LNAME, 0, STATX_BASIC_STATS, &statxbuf));
640	if (TST_RET == 0)
641		tst_res(TPASS, "It returned zero so it must have worked!");
642	else
643		tst_res(TFAIL | TERRNO, "statx can not stat a basic file");
644}
645
646static struct tst_test test = {
647	.setup = setup,
648	.cleanup = cleanup,
649	.test_all = run,
650	.tcnt = 2,
651	.min_kver = "4.11",
652	.needs_tmpdir = 1
653};
654--------------------------------------------------------------------------------
655
656Because we are now opening a file, we need a +cleanup+ function to close the
657file descriptors. We have to manually close the files to ensure the temporary
658directory is deleted by the test harness (see the
659https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines[test
660writing guidelines] for details).
661
662As a matter of good practice, the file descriptors are closed in reverse
663order. In some circumstances the order in which clean up is performed is
664significant. In that case resources created towards the end of 'setup' are
665dependent on ones near the beginning. So during 'cleanup' we remove the
666dependants before their dependencies.
667
668If, for some reason, the file descriptor +lfd+ became invalid during the test,
669but +fd+ was still open, we do not want +SAFE_CLOSE(lfd)+ to cause the
670+cleanup+ function to exit prematurely. If it did, then +fd+ would remain open
671which would cause problems on some file systems.
672
673Nor do we want to call +cleanup+ recursively. So during 'cleanup' +tst_brk+,
674and consequently the +SAFE+ functions, do not cause the test to exit with
675+TBROK+. Instead they just print an error message with +TWARN+.
676
677It is not entirely necessary to check if the file descriptors have a none zero
678value before attempting to close them. However it avoids a bunch of spurious
679warning messages if we fail to open +file_to_stat+. Test case failures can be
680difficult to interpret at the best of times, so avoid filling the log with
681noise.
682
6835.1 Check statx returns the correct number of hard links
684~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
685
686The field +statx.stx_nlink+ should be equal to 2, right?
687
6885.2 Git-branch
689~~~~~~~~~~~~~~
690
691We are about to make some organisational changes to the test, so now would be
692a good time to branch. Then we can switch between the old and new versions, to
693check the behavior has not been changed by accident.
694
6956. Split the test
696-----------------
697
698In our current test, we have essentially rolled two different test cases into
699one. Firstly we check if an error is returned when bad arguments are provided
700and secondly we check what happens when we stat an actual file. Quite often it
701makes sense to call +tst_res+ multiple times in a single test case because we
702are checking different properties of the same result, but here we are clearly
703testing two different scenarios.
704
705So we should split the test in two. One obvious way to do this is to create
706+statx02.c+, but that seems like overkill in order to separate two simple test
707cases. So, for now at least, we are going to do it a different way.
708
709[source,c]
710--------------------------------------------------------------------------------
711...
712
713static void run_stat_null(void)
714{
715	struct statx statxbuf = { 0 };
716
717	TEST(sys_statx(0, NULL, 0, 0, &statxbuf));
718	if (TST_RET == 0)
719		tst_res(TFAIL, "statx thinks it can stat NULL");
720	else if (TST_ERR == EFAULT)
721		tst_res(TPASS, "statx set errno to EFAULT as expected");
722	else
723		tst_res(TFAIL | TERRNO, "statx set errno to some unexpected value");
724}
725
726static void run_stat_symlink(void)
727{
728	struct statx statxbuf = { 0 };
729
730	TEST(sys_statx(AT_FDCWD, LNAME, 0, STATX_BASIC_STATS, &statxbuf));
731	if (TST_RET == 0)
732		tst_res(TPASS, "It returned zero so it must have worked!");
733	else
734		tst_res(TFAIL | TERRNO, "statx can not stat a basic file");
735}
736
737static void run(unsigned int i)
738{
739	switch(i) {
740	case 0: run_stat_null();
741	case 1: run_stat_symlink();
742	}
743}
744
745static struct tst_test test = {
746	.setup = setup,
747	.cleanup = cleanup,
748	.test = run,
749	.tcnt = 2,
750	.min_kver = "4.11",
751	.needs_tmpdir = 1
752};
753--------------------------------------------------------------------------------
754
755So we have used an alternative form of the +test+ or +run+ callback which
756accepts an index. Some tests use this index with an array of parameters and
757expected return values. Others do something similar to the above. The index
758can be used how you want so long as each iteration calls +tst_res+ in a
759meaningful way.
760
761If an iteration fails to return a result (i.e. call +tst_res+ with a value
762other than +TINFO+) then the test harness will report +TBROK+ and print the
763iteration which failed. This prevents a scenario in your test from silently
764failing due to some faulty logic.
765
7666.1 What is wrong with the switch statement?
767~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
768
769Were you paying attention? Also see the output of +make check+.
770
7716.2 Test a feature unique to statx
772~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
773
774So far we have not tested anything which is unique to +statx+. So, for
775example, you could check stx_btime is correct (possibly only to within a
776margin of error) and that it differs from +stx_mtime+ after writing to the
777file.
778
779Alternatively you could check that +stx_dev_major+ and +stx_dev_minor+ are set
780correctly. Note that the LTP has helper functions for creating devices and
781file systems (see
782https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines#2214-testing-with-a-block-device[section
7832.2.14] of the Test Writing Guidelines).
784
785This could be quite a challenging exercise. You may wish to tackle an
786altogether different test scenario instead. If you get stuck just move onto
787the next section and come back later.
788
7897. Submitting the test for review
790---------------------------------
791
792Ignoring the fact we should probably create +lapi/stat.h+ along with a bunch
793of fallback logic in the build system. We can now get our test ready for
794submission.
795
796The first thing you need to do before considering submitting your test is run
797+make check-statx01+ or + make check+ in the test's directory. Again, we use
798the kernel style guidelines where possible. Next you should create a new
799branch, this will allow you to reshape your commit history without fear.
800
801After that we have the pleasure of doing an interactive 'rebase' to clean up
802our commit history. In its current form the test only really needs a single
803commit, but if you have been using Git correctly then you should have
804many. The main reason we want to compress it to a single commit, is to make
805the LTP's Git-log readable. It also allows us to write a coherent description
806of the work as a whole in retrospective. Although, when adding a new test, the
807test description in the code will probably make the commit message redundant.
808
809Anyway, as an example, we shall look at my personal commit history from this
810tutorial and 'rebase' it. You should try following along with your own
811repository. First lets look at the commit history since we branched from
812master.
813
814[source,shell]
815--------------------------------------------------------------------------------
816$ git log -oneline master..HEAD
817152d39fe7 (HEAD -> tutorial-rebase2, tutorial-rebase) tutorial: Start Submitting patch section
81870f7ce7ce statx01: Stop checkpatch from complaining
819bb0332bd7 tutorial: Fix review problems
8206a87a084a statx01: Fix review problems
821d784b1e85 test-writing-guidelines: Remove old API argument
822c26e1be7a fixup! tutorial
8231e24a5fb5 (me/tutorial-rebase) fixup! tutorial
824568a3f7be fixup! tutorial
82509dd2c829 statx: stage 6
826bfeef7902 statx: stage 5b
82776e03d714 statx: stage 5a
82898f5bc7ac statx: stage 4
8296f8c16438 statx: stage 3 (Add statx01)
8305d93b84d8 Add statx and other syscall numbers
8315ca627b78 tutorial: Add a step-by-step C test tutorial
832--------------------------------------------------------------------------------
833
834So we have told git to show all the commits which don't exist in 'master', but
835are in +HEAD+, where +HEAD+ is the top of the current branch. The current
836branch is +tutorial-rebase2+ which I just created. I have already done one
837'rebase' and submitted a patch for review, so my original branch was just called
838+tutorial+.
839
840As usual my commit history is starting to look like a bit of mess! There is
841even a commit in there which should not be in the this branch (Remove old API
842argument), however it can be ignored for now and 'cherry picked' into a new branch
843later.
844
845For my patch I actually need at least two commits, one which contains the
846tutorial text and one which contains the test and associated files. So first
847of all I want to 'squash' (amalgamate) all the commits appended with
848+tutorial:+ into the bottom commit.
849
850[source,shell]
851--------------------------------------------------------------------------------
852$ git rebase -i 5ca627b78\^
853...
854--------------------------------------------------------------------------------
855
856This begins an interactive 'rebase' where commit 5ca6427b78 is the earliest
857commit we want to edit. The +^+ symbol after the commit hash, specifies the
858commit before this one. The interactive 'rebase' command takes the last commit
859we want to keep unaltered as it's argument (in other words it takes a
860non-inclusive range).
861
862Upon entering a similar command you will be presented with a text file
863similar to the following. The file should be displayed in your text editor of
864choice, if it doesn't, then you may change the editor variable in +.gitconfig+
865which was shown in section 3.
866
867[source,rebase]
868--------------------------------------------------------------------------------
869pick 5ca627b78 tutorial: Add a step-by-step C test tutorial
870pick 5d93b84d8 Add statx and other syscall numbers
871pick 6f8c16438 statx: stage 3 (Add statx01)
872pick 98f5bc7ac statx: stage 4
873pick 76e03d714 statx: stage 5a
874pick bfeef7902 statx: stage 5b
875pick 09dd2c829 statx: stage 6
876pick 568a3f7be fixup! tutorial
877pick 1e24a5fb5 fixup! tutorial
878pick c26e1be7a fixup! tutorial
879pick d784b1e85 test-writing-guidelines: Remove old API argument
880pick 6a87a084a statx01: Fix review problems
881pick bb0332bd7 tutorial: Fix review problems
882pick 70f7ce7ce statx01: Stop checkpatch from complaining
883pick 152d39fe7 tutorial: Start Submitting patch section
884--------------------------------------------------------------------------------
885
886The last commit from Git-log is shown at the top. The left hand column
887contains the commands we want to run on each commit. +pick+ just means we
888re-apply the commit as-is. We can reorder the lines to apply the commits in a
889different order, but we need to be careful when reordering commits to the same
890file. If your 'rebase' results in a merge conflict, then you have probably
891reordered some commits which contained changes to the same piece of code.
892
893Perhaps a better name for the interactive 'rebase' command would be 'replay'. As
894we pick a point in the commit history, undo all those commits before that
895point, then reapply them one at a time. During the replay we can reorder the
896commits, drop, merge, split and edit them, creating a new history.
897
898The commands I am going to use are +reword+ and +fixup+. The +reword+ command
899allows you to edit a single commit's message. The 'fixup' command 'squashes' a
900commit into the commit above/preceding it, merging the two commits into
901one. The commit which has +fixup+ applied has its commit message deleted. If
902you think a commit might have something useful in its message then you can use
903+squash+ instead.
904
905[source,rebase]
906--------------------------------------------------------------------------------
907reword 5ca627b78 tutorial: Add a step-by-step C test tutorial
908fixup 568a3f7be fixup! tutorial
909fixup 1e24a5fb5 fixup! tutorial
910fixup c26e1be7a fixup! tutorial
911fixup bb0332bd7 tutorial: Fix review problems
912fixup 152d39fe7 tutorial: Start Submitting patch section
913fixup 276edecab tutorial: Save changes before rebase
914pick 5d93b84d8 Add statx and other syscall numbers
915pick 6f8c16438 statx: stage 3 (Add statx01)
916pick 98f5bc7ac statx: stage 4
917pick 76e03d714 statx: stage 5a
918pick bfeef7902 statx: stage 5b
919pick 09dd2c829 statx: stage 6
920pick d784b1e85 test-writing-guidelines: Remove old API argument
921pick 6a87a084a statx01: Fix review problems
922--------------------------------------------------------------------------------
923
924So all the commits marked with +fixup+ will be re-played by Git immediately
925after 5ca62 at the top. A new commit will then be created with the amalgamated
926changes of all the commits and 5ca62's log message. It turns out that I didn't
927need to reword anything, but there is no harm in checking. It is easy to
928forget the +Signed-off-by:+ line.
929
930I could now do the same for the commits to the +statx+ test, making the commit
931message prefixes consistent. However I am not actually going to submit the
932test (yet).
933
934I won't attempt to show you this, but if you need to do the opposite and split
935apart a commit. It is also possible using Git-rebase by marking a line with
936+edit+. This will pause Git just after replaying the marked commit. You can
937then use a 'soft' Git-reset to bring the selected commit's changes back into
938the 'index' where you are then able to un-stage some parts before
939re-committing.
940
941You can also use +edit+ and +git commit --amend+ together to change a commit
942deep in your history, but without reseting the 'index'. The 'index' contains
943changes which you have staged with +git add+, but not yet committed.
944
945So now that the commit history has been cleaned up, we need to submit a patch
946to the mailing list or make a pull request on GitHub. The mailing list is the
947preferred place to make submissions and is more difficult for most people, so
948I will only cover that method.
949
950Just before we create the patch, we need to check that our changes will still
951apply to the master branch without problems. To do this we can use another
952type of 'rebase' and then try rebuilding and running the test.
953
954[source,shell]
955--------------------------------------------------------------------------------
956$ git checkout master
957$ git pull origin
958$ git checkout tutorial-rebase2
959$ git rebase master
960--------------------------------------------------------------------------------
961
962Above, I update the master branch and then replay our changes onto it using
963+git rebase master+. You may find that after the rebase there is a merge
964conflict. This will result in something which looks like the following (taken
965from a Makefile conflict which was caused by reordering commits in a 'rebase').
966
967[source,diff]
968--------------------------------------------------------------------------------
969<<<<<<< HEAD
970cve-2016-7117:	LDFLAGS += -lpthread
971=======
972cve-2014-0196:	LDFLAGS += -lpthread -lutil -lrt
973cve-2016-7117:	LDFLAGS += -lpthread -lrt
974>>>>>>> 4dbfb8e79... Add -lrt
975--------------------------------------------------------------------------------
976
977The first line tells us this is the beginning of a conflict. The third line
978separates the two conflicting pieces of content and the last line is the end
979of the conflict. Usually, all you need to do is remove the lines you don't
980want, stage the changes and continue the 'rebase' with +git rebase
981--continue+.
982
983In order to create a patch e-mail we use +git format-patch+, we can then send
984that e-mail using +git send-email+. It is also possible to import the patch
985(+mbox+) file into a number of e-mail programs.
986
987[source,shell]
988--------------------------------------------------------------------------------
989$ git format-patch -1 -v 2 -o output --to ltp@lists.linux.it fd3cc8596
990output/v2-0001-tutorial-Add-a-step-by-step-C-test-tutorial.patch
991--------------------------------------------------------------------------------
992
993The first argument +-1+ specifies we want one commit from fd3cc8596
994onwards. If we wanted this commit and the one after it we could specify +-2+
995instead.
996
997This is my second patch submission so I have used +-v 2+, which indicates this
998is the second version of a patch set. The +-o+ option specifies the output
999directory (literally called +output+). The +--to+ option adds the +To:+ e-mail
1000header, which I have set to the LTP mailing list.
1001
1002We can then send this patch with the following command sans +--dry-run+.
1003
1004[source,shell]
1005--------------------------------------------------------------------------------
1006$ git send-email --dry-run output/v2-0001-tutorial-Add-a-step-by-step-C-test-tutorial.patch
1007--------------------------------------------------------------------------------
1008
1009Git will ask some questions (which you an ignore) and then tell you what it
1010would do if this weren't a dry-run. In order for this to work you have to have
1011a valid SMTP server set in +.gitconfig+ and also be signed up to the LTP
1012mailing list under the same e-mail address you have configured in Git. You can
1013sign up at https://lists.linux.it/listinfo/ltp.
1014
10158. Doing code review
1016--------------------
1017
1018While waiting for your test to be reviewed, you are invited and encouraged to
1019review other contributors' code. This may seem bizarre when you are completely
1020new to the project, but there are two important ways in which you can
1021contribute here:
1022
1023A.   Point out logical errors in the code.
1024B.   Improve your own understanding
1025
1026It doesn't matter whether you know the canonical way of writing an LTP test in
1027C. An error of logic, when properly explained, is usually indisputable. These
1028are the most important errors to find as they always result in false test
1029results. Once someone points out such an error it is usually obvious to
1030everyone that it is a bug and needs to be fixed.
1031
1032Obviously testing the patch is one way of finding errors. You can apply
1033patches using +git am+. Then it is just a case of compiling and running the
1034tests.
1035
1036Finally, reading and attempting to comment on other peoples patches, gives
1037you a better understanding of the reviewers perspective. This is better for
1038the project and for you.
1039
1040Style and organisational issues are best left to after you have found logical
1041errors.
1042
10439. Final notes
1044--------------
1045
1046Hopefully you can now grasp the structure of an LTP test and have some idea of
1047what is available in the LTP test library. There are a vast number of library
1048functions available (mainly located in include and lib), some of which are
1049documented in the test writing guidelines and many of which are not.
1050
1051We have only scratched the surface of the immense technical complexity of
1052systems programming across multiple Kernel and C lib versions as well as
1053different hardware architectures. The important thing to take away from this
1054is that you have to be conscientious of what will happen on systems different
1055from yours. The LTP has a huge and varied user base, so situations you may
1056thing are unlikely can and do happen to somebody.
1057
1058Of course you don't want to spend time allowing for situations which may never
1059arise either, so you have to do your research and think about each situation
1060critically. The more systems you can test on before submitting your changes,
1061the better, although we understand not everyone has access to a lab.
1062
1063One important topic which has not been covered by this tutorial, is
1064multi-process or multi-threaded testing. The LTP library functions work inside
1065child processes and threads, but their semantics change slightly. There are
1066also various helper functions for synchronising and forking processes. For
1067more information see
1068https://github.com/linux-test-project/ltp/wiki/C-Test-API[C Test API],
1069in particular sections
1070https://github.com/linux-test-project/ltp/wiki/C-Test-API#17-fork-ing[1.7 Fork()-ing] to
1071https://github.com/linux-test-project/ltp/wiki/C-Test-API#110-signals-and-signal-handlers[1.10 Signals and signal handlers] and
1072https://github.com/linux-test-project/ltp/wiki/C-Test-API#114-thread-safety-in-the-ltp-library[1.14 Thread-safety in the LTP library].
1073
1074When it comes time to submit a test, the preferred way to do it is on the
1075mailing list although you can also use GitHub. The LTP follows similar rules
1076to the kernel for formatting and submitting patches. Generally speaking the
1077review cycle is easier for small patches, so try to make small changes or
1078additions where possible.
1079