1 _______________________
2 BUILDING THE FSQA SUITE
3 _______________________
5 - cd into the xfstests directory
6 - install prerequisite packages
7 For example, for Ubuntu:
8 sudo apt-get install xfslibs-dev uuid-dev libtool-bin \
9 e2fsprogs automake gcc libuuid1 quota attr make \
10 libacl1-dev libaio-dev xfsprogs libgdbm-dev gawk fio dbench \
11 uuid-runtime python sqlite3 liburing-dev libcap-dev
12 For Fedora, RHEL, or CentOS:
13 yum install acl attr automake bc dbench dump e2fsprogs fio \
14 gawk gcc indent libtool lvm2 make psmisc quota sed \
16 libacl-devel libaio-devel libuuid-devel \
17 xfsprogs-devel btrfs-progs-devel python sqlite liburing-devel \
19 (Older distributions may require xfsprogs-qa-devel as well.)
20 (Note that for RHEL and CentOS, you may need the EPEL repo.)
23 - create fsgqa test user ("sudo useradd -m fsgqa")
24 - create fsgqa group ("sudo groupadd fsgqa")
25 - create 123456-fsgqa test user ("sudo useradd 123456-fsgqa")
26 this 2nd user creation step can be safely skipped if your system
27 doesn't support names starting with digits, only a handful of tests
29 - create fsgqa2 test user ("sudo useradd fsgqa2")
31 ______________________
33 ______________________
35 Preparing system for tests:
37 - compile XFS into your kernel or load XFS modules
38 - install administrative tools specific to the filesystem you wish to test
39 - If you wish to run the udf components of the suite install
40 mkudffs. Also download and build the Philips UDF Verification Software
41 from https://www.lscdweb.com/registered/udf_verifier.html, then copy the udf_test
42 binary to xfstests/src/. If you wish to disable UDF verification test
43 set the environment variable DISABLE_UDF_TEST to 1.
45 - create one or two partitions to use for testing
47 - format as XFS, mount & optionally populate with
49 - one SCRATCH partition (optional)
50 - leave empty and expect this partition to be clobbered
51 by some tests. If this is not provided, many tests will
53 (SCRATCH and TEST must be two DIFFERENT partitions)
55 - for btrfs only: some btrfs test cases will need 3 or more independent
56 SCRATCH disks which should be set using SCRATCH_DEV_POOL (for eg:
57 SCRATCH_DEV_POOL="/dev/sda /dev/sdb /dev/sdc") with which
58 SCRATCH_DEV should be unused by the tester, and for the legacy
59 support SCRATCH_DEV will be set to the first disk of the
60 SCRATCH_DEV_POOL by xfstests script.
62 - setup your environment
64 - copy local.config.example to local.config and edit as needed
66 - setenv TEST_DEV "device containing TEST PARTITION"
67 - setenv TEST_DIR "mount point of TEST PARTITION"
69 - setenv SCRATCH_DEV "device containing SCRATCH PARTITION" OR
70 (btrfs only) setenv SCRATCH_DEV_POOL "to 3 or more SCRATCH disks for
71 testing btrfs raid concepts"
72 - setenv SCRATCH_MNT "mount point for SCRATCH PARTITION"
73 - setenv TAPE_DEV "tape device for testing xfsdump"
74 - setenv RMT_TAPE_DEV "remote tape device for testing xfsdump"
75 - setenv RMT_IRIXTAPE_DEV "remote IRIX tape device for testing xfsdump"
76 - setenv SCRATCH_LOGDEV "device for scratch-fs external log"
77 - setenv SCRATCH_RTDEV "device for scratch-fs realtime data"
78 - setenv TEST_LOGDEV "device for test-fs external log"
79 - setenv TEST_RTDEV "device for test-fs realtime data"
80 - if TEST_LOGDEV and/or TEST_RTDEV, these will always be used.
81 - if SCRATCH_LOGDEV and/or SCRATCH_RTDEV, the USE_EXTERNAL
82 environment variable set to "yes" will enable their use.
83 - setenv DIFF_LENGTH "number of diff lines to print from a failed test",
84 by default 10, set to 0 to print the full diff
85 - setenv FSTYP "the filesystem you want to test", the filesystem
86 type is devised from the TEST_DEV device, but you may want to
87 override it; if unset, the default is 'xfs'
88 - setenv FSSTRESS_AVOID and/or FSX_AVOID, which contain options
89 added to the end of fsstresss and fsx invocations, respectively,
90 in case you wish to exclude certain operational modes from these
92 - set TEST_XFS_REPAIR_REBUILD=1 to have _check_xfs_filesystem
93 run xfs_repair -n to check the filesystem; xfs_repair to rebuild
94 metadata indexes; and xfs_repair -n (a third time) to check the
95 results of the rebuilding.
96 - xfs_scrub, if present, will always check the test and scratch
97 filesystems if they are still online at the end of the test.
98 It is no longer necessary to set TEST_XFS_SCRUB.
99 - setenv LOGWRITES_DEV to a block device to use for power fail
101 - setenv PERF_CONFIGNAME to a arbitrary string to be used for
102 identifying the test setup for running perf tests. This should
103 be different for each type of performance test you wish to run so
104 that relevant results are compared. For example 'spinningrust'
105 for configurations that use spinning disks and 'nvme' for tests
107 - set USE_KMEMLEAK=yes to scan for memory leaks in the kernel
108 after every test, if the kernel supports kmemleak.
109 - set KEEP_DMESG=yes to keep dmesg log after test
110 - Set TEST_FS_MODULE_RELOAD=1 to unload the module and reload
111 it between test invocations. This assumes that the name of
112 the module is the same as FSTYP.
113 - Set DUMP_CORRUPT_FS=1 to record metadata dumps of XFS or ext*
114 filesystems if a filesystem check fails.
115 - Set DUMP_COMPRESSOR to a compression program to compress
116 metadumps of filesystems. This program must accept '-f' and the
117 name of a file to compress; and it must accept '-d -f -k' and
118 the name of a file to decompress. In other words, it must
120 - Set MIN_FSSIZE to specify the minimal size (bytes) of a
121 filesystem we can create. Setting this parameter will
122 skip the tests creating a filesystem less than
125 - or add a case to the switch in common/config assigning
126 these variables based on the hostname of your test
128 - or add these variables to a file called local.config and keep that
129 file in your workarea.
131 - if testing xfsdump, make sure the tape devices have a
132 tape which can be overwritten.
134 - make sure $TEST_DEV is a mounted XFS partition
135 - make sure that $SCRATCH_DEV or $SCRATCH_DEV_POOL contains nothing useful
140 - By default the tests suite will run all the tests in the auto group. These
141 are the tests that are expected to function correctly as regression tests,
142 and it excludes tests that exercise conditions known to cause machine
143 failures (i.e. the "dangerous" tests).
144 - ./check '*/001' '*/002' '*/003'
146 - Groups of tests maybe ran by: ./check -g [group(s)]
147 See the tests/*/group.list files after building xfstests to learn about
148 each test's group memberships.
149 - If you want to run all tests regardless of what group they are in
150 (including dangerous tests), use the "all" group: ./check -g all
151 - To randomize test order: ./check -r [test(s)]
152 - You can explicitly specify NFS/CIFS/OVERLAY, otherwise
153 the filesystem type will be autodetected from $TEST_DEV:
154 - for running nfs tests: ./check -nfs [test(s)]
155 - for running cifs/smb3 tests: ./check -cifs [test(s)]
156 - for overlay tests: ./check -overlay [test(s)]
157 The TEST and SCRATCH partitions should be pre-formatted
158 with another base fs, where the overlay dirs will be created
161 The check script tests the return value of each script, and
162 compares the output against the expected output. If the output
163 is not as expected, a diff will be output and an .out.bad file
164 will be produced for the failing test.
166 Unexpected console messages, crashes and hangs may be considered
167 to be failures but are not necessarily detected by the QA system.
169 __________________________
170 ADDING TO THE FSQA SUITE
171 __________________________
174 Creating new tests scripts:
176 Use the "new" script.
178 Test script environment:
180 When developing a new test script keep the following things in
181 mind. All of the environment variables and shell procedures are
182 available to the script once the "common/preamble" file has been
183 sourced and the "_begin_fstest" function has been called.
185 1. The tests are run from an arbitrary directory. If you want to
186 do operations on an XFS filesystem (good idea, eh?), then do
187 one of the following:
189 (a) Create directories and files at will in the directory
190 $TEST_DIR ... this is within an XFS filesystem and world
191 writeable. You should cleanup when your test is done,
192 e.g. use a _cleanup shell procedure in the trap ... see
193 001 for an example. If you need to know, the $TEST_DIR
194 directory is within the filesystem on the block device
197 (b) mkfs a new XFS filesystem on $SCRATCH_DEV, and mount this
198 on $SCRATCH_MNT. Call the the _require_scratch function
199 on startup if you require use of the scratch partition.
200 _require_scratch does some checks on $SCRATCH_DEV &
201 $SCRATCH_MNT and makes sure they're unmounted. You should
202 cleanup when your test is done, and in particular unmount
204 Tests can make use of $SCRATCH_LOGDEV and $SCRATCH_RTDEV
205 for testing external log and realtime volumes - however,
206 these tests need to simply "pass" (e.g. cat $seq.out; exit
207 - or default to an internal log) in the common case where
208 these variables are not set.
210 2. You can safely create temporary files that are not part of the
211 filesystem tests (e.g. to catch output, prepare lists of things
212 to do, etc.) in files named $tmp.<anything>. The standard test
213 script framework created by "new" will initialize $tmp and
216 3. By default, tests are run as the same uid as the person
217 executing the control script "check" that runs the test scripts.
219 4. Some other useful shell procedures:
221 _get_fqdn - echo the host's fully qualified
224 _get_pids_by_name - one argument is a process name, and
225 return all of the matching pids on
228 _within_tolerance - fancy numerical "close enough is good
229 enough" filter for deterministic
230 output ... see comments in
231 common/filter for an explanation
233 _filter_date - turn ctime(3) format dates into the
234 string DATE for deterministic
237 _cat_passwd, - dump the content of the password
238 _cat_group or group file (both the local file
239 and the content of the NIS database
240 if it is likely to be present)
242 5. General recommendations, usage conventions, etc.:
243 - When the content of the password or group file is
244 required, get it using the _cat_passwd and _cat_group
245 functions, to ensure NIS information is included if NIS
247 - When calling getfacl in a test, pass the "-n" argument so
248 that numeric rather than symbolic identifiers are used in
250 - When creating a new test, it is possible to enter a custom name
251 for the file. Filenames are in form NNN-custom-name, where NNN
252 is automatically added by the ./new script as an unique ID,
253 and "custom-name" is the optional string entered into a prompt
254 in the ./new script. It can contain only alphanumeric characters
255 and dash. Note the "NNN-" part is added automatically.
257 6. Test group membership: Each test can be associated with any number
258 of groups for convenient selection of subsets of tests. Group names
259 can be any sequence of non-whitespace characters. Test authors
260 associate a test with groups by passing the names of those groups as
261 arguments to the _begin_fstest function. For example, the code:
263 _begin_fstest auto quick subvol snapshot
265 associates the current test with the "auto", "quick", "subvol", and
266 "snapshot" groups. It is not necessary to specify the "all" group
267 in the list because that group is computed at run time.
269 The build process scans test files for _begin_fstest invocations and
270 compiles the group list from that information. In other words, test
271 files must call _begin_fstest or they will not be run.
275 Each test script has a name, e.g. 007, and an associated
276 verified output, e.g. 007.out.
278 It is important that the verified output is deterministic, and
279 part of the job of the test script is to filter the output to
280 make this so. Examples of the sort of things that need filtering:
287 - variable directory contents
288 - imprecise numbers, especially sizes and times
292 The script "check" may be used to run one or more tests.
294 Test number $seq is deemed to "pass" when:
295 (a) no "core" file is created,
296 (b) the file $seq.notrun is not created,
297 (c) the exit status is 0, and
298 (d) the output matches the verified output.
300 In the "not run" case (b), the $seq.notrun file should contain a
301 short one-line summary of why the test was not run. The standard
302 output is not checked, so this can be used for a more verbose
303 explanation and to provide feedback when the QA test is run
307 To force a non-zero exit status use:
313 won't have the desired effect because of the way the exit trap
316 The recent pass/fail history is maintained in the file "check.log".
317 The elapsed time for the most recent pass for each test is kept
320 The compare-failures script in tools/ may be used to compare failures
321 across multiple runs, given files containing stdout from those runs.
327 Send patches to the fstests mailing list at fstests@vger.kernel.org.