1.. Copyright 2010 Nicolas Palix <npalix@diku.dk>
2.. Copyright 2010 Julia Lawall <julia@diku.dk>
3.. Copyright 2010 Gilles Muller <Gilles.Muller@lip6.fr>
4
5.. highlight:: none
6
7.. _devtools_coccinelle:
8
9Coccinelle
10==========
11
12Coccinelle is a tool for pattern matching and text transformation that has
13many uses in kernel development, including the application of complex,
14tree-wide patches and detection of problematic programming patterns.
15
16Getting Coccinelle
17------------------
18
19The semantic patches included in the kernel use features and options
20which are provided by Coccinelle version 1.0.0-rc11 and above.
21Using earlier versions will fail as the option names used by
22the Coccinelle files and coccicheck have been updated.
23
24Coccinelle is available through the package manager
25of many distributions, e.g. :
26
27 - Debian
28 - Fedora
29 - Ubuntu
30 - OpenSUSE
31 - Arch Linux
32 - NetBSD
33 - FreeBSD
34
35Some distribution packages are obsolete and it is recommended
36to use the latest version released from the Coccinelle homepage at
37http://coccinelle.lip6.fr/
38
39Or from Github at:
40
41https://github.com/coccinelle/coccinelle
42
43Once you have it, run the following commands::
44
45        ./autogen
46        ./configure
47        make
48
49as a regular user, and install it with::
50
51        sudo make install
52
53More detailed installation instructions to build from source can be
54found at:
55
56https://github.com/coccinelle/coccinelle/blob/master/install.txt
57
58Supplemental documentation
59--------------------------
60
61For supplemental documentation refer to the wiki:
62
63https://bottest.wiki.kernel.org/coccicheck
64
65The wiki documentation always refers to the linux-next version of the script.
66
67For Semantic Patch Language(SmPL) grammar documentation refer to:
68
69http://coccinelle.lip6.fr/documentation.php
70
71Using Coccinelle on the Linux kernel
72------------------------------------
73
74A Coccinelle-specific target is defined in the top level
75Makefile. This target is named ``coccicheck`` and calls the ``coccicheck``
76front-end in the ``scripts`` directory.
77
78Four basic modes are defined: ``patch``, ``report``, ``context``, and
79``org``. The mode to use is specified by setting the MODE variable with
80``MODE=<mode>``.
81
82- ``patch`` proposes a fix, when possible.
83
84- ``report`` generates a list in the following format:
85  file:line:column-column: message
86
87- ``context`` highlights lines of interest and their context in a
88  diff-like style. Lines of interest are indicated with ``-``.
89
90- ``org`` generates a report in the Org mode format of Emacs.
91
92Note that not all semantic patches implement all modes. For easy use
93of Coccinelle, the default mode is "report".
94
95Two other modes provide some common combinations of these modes.
96
97- ``chain`` tries the previous modes in the order above until one succeeds.
98
99- ``rep+ctxt`` runs successively the report mode and the context mode.
100  It should be used with the C option (described later)
101  which checks the code on a file basis.
102
103Examples
104~~~~~~~~
105
106To make a report for every semantic patch, run the following command::
107
108		make coccicheck MODE=report
109
110To produce patches, run::
111
112		make coccicheck MODE=patch
113
114
115The coccicheck target applies every semantic patch available in the
116sub-directories of ``scripts/coccinelle`` to the entire Linux kernel.
117
118For each semantic patch, a commit message is proposed.  It gives a
119description of the problem being checked by the semantic patch, and
120includes a reference to Coccinelle.
121
122As with any static code analyzer, Coccinelle produces false
123positives. Thus, reports must be carefully checked, and patches
124reviewed.
125
126To enable verbose messages set the V= variable, for example::
127
128   make coccicheck MODE=report V=1
129
130Coccinelle parallelization
131--------------------------
132
133By default, coccicheck tries to run as parallel as possible. To change
134the parallelism, set the J= variable. For example, to run across 4 CPUs::
135
136   make coccicheck MODE=report J=4
137
138As of Coccinelle 1.0.2 Coccinelle uses Ocaml parmap for parallelization;
139if support for this is detected you will benefit from parmap parallelization.
140
141When parmap is enabled coccicheck will enable dynamic load balancing by using
142``--chunksize 1`` argument. This ensures we keep feeding threads with work
143one by one, so that we avoid the situation where most work gets done by only
144a few threads. With dynamic load balancing, if a thread finishes early we keep
145feeding it more work.
146
147When parmap is enabled, if an error occurs in Coccinelle, this error
148value is propagated back, and the return value of the ``make coccicheck``
149command captures this return value.
150
151Using Coccinelle with a single semantic patch
152---------------------------------------------
153
154The optional make variable COCCI can be used to check a single
155semantic patch. In that case, the variable must be initialized with
156the name of the semantic patch to apply.
157
158For instance::
159
160	make coccicheck COCCI=<my_SP.cocci> MODE=patch
161
162or::
163
164	make coccicheck COCCI=<my_SP.cocci> MODE=report
165
166
167Controlling Which Files are Processed by Coccinelle
168---------------------------------------------------
169
170By default the entire kernel source tree is checked.
171
172To apply Coccinelle to a specific directory, ``M=`` can be used.
173For example, to check drivers/net/wireless/ one may write::
174
175    make coccicheck M=drivers/net/wireless/
176
177To apply Coccinelle on a file basis, instead of a directory basis, the
178C variable is used by the makefile to select which files to work with.
179This variable can be used to run scripts for the entire kernel, a
180specific directory, or for a single file.
181
182For example, to check drivers/bluetooth/bfusb.c, the value 1 is
183passed to the C variable to check files that make considers
184need to be compiled.::
185
186    make C=1 CHECK=scripts/coccicheck drivers/bluetooth/bfusb.o
187
188The value 2 is passed to the C variable to check files regardless of
189whether they need to be compiled or not.::
190
191    make C=2 CHECK=scripts/coccicheck drivers/bluetooth/bfusb.o
192
193In these modes, which work on a file basis, there is no information
194about semantic patches displayed, and no commit message proposed.
195
196This runs every semantic patch in scripts/coccinelle by default. The
197COCCI variable may additionally be used to only apply a single
198semantic patch as shown in the previous section.
199
200The "report" mode is the default. You can select another one with the
201MODE variable explained above.
202
203Debugging Coccinelle SmPL patches
204---------------------------------
205
206Using coccicheck is best as it provides in the spatch command line
207include options matching the options used when we compile the kernel.
208You can learn what these options are by using V=1; you could then
209manually run Coccinelle with debug options added.
210
211Alternatively you can debug running Coccinelle against SmPL patches
212by asking for stderr to be redirected to stderr. By default stderr
213is redirected to /dev/null; if you'd like to capture stderr you
214can specify the ``DEBUG_FILE="file.txt"`` option to coccicheck. For
215instance::
216
217    rm -f cocci.err
218    make coccicheck COCCI=scripts/coccinelle/free/kfree.cocci MODE=report DEBUG_FILE=cocci.err
219    cat cocci.err
220
221You can use SPFLAGS to add debugging flags; for instance you may want to
222add both --profile --show-trying to SPFLAGS when debugging. For example
223you may want to use::
224
225    rm -f err.log
226    export COCCI=scripts/coccinelle/misc/irqf_oneshot.cocci
227    make coccicheck DEBUG_FILE="err.log" MODE=report SPFLAGS="--profile --show-trying" M=./drivers/mfd
228
229err.log will now have the profiling information, while stdout will
230provide some progress information as Coccinelle moves forward with
231work.
232
233NOTE:
234
235DEBUG_FILE support is only supported when using coccinelle >= 1.0.2.
236
237Currently, DEBUG_FILE support is only available to check folders, and
238not single files. This is because checking a single file requires spatch
239to be called twice leading to DEBUG_FILE being set both times to the same value,
240giving rise to an error.
241
242.cocciconfig support
243--------------------
244
245Coccinelle supports reading .cocciconfig for default Coccinelle options that
246should be used every time spatch is spawned. The order of precedence for
247variables for .cocciconfig is as follows:
248
249- Your current user's home directory is processed first
250- Your directory from which spatch is called is processed next
251- The directory provided with the --dir option is processed last, if used
252
253Since coccicheck runs through make, it naturally runs from the kernel
254proper dir; as such the second rule above would be implied for picking up a
255.cocciconfig when using ``make coccicheck``.
256
257``make coccicheck`` also supports using M= targets. If you do not supply
258any M= target, it is assumed you want to target the entire kernel.
259The kernel coccicheck script has::
260
261    if [ "$KBUILD_EXTMOD" = "" ] ; then
262        OPTIONS="--dir $srctree $COCCIINCLUDE"
263    else
264        OPTIONS="--dir $KBUILD_EXTMOD $COCCIINCLUDE"
265    fi
266
267KBUILD_EXTMOD is set when an explicit target with M= is used. For both cases
268the spatch --dir argument is used, as such third rule applies when whether M=
269is used or not, and when M= is used the target directory can have its own
270.cocciconfig file. When M= is not passed as an argument to coccicheck the
271target directory is the same as the directory from where spatch was called.
272
273If not using the kernel's coccicheck target, keep the above precedence
274order logic of .cocciconfig reading. If using the kernel's coccicheck target,
275override any of the kernel's .coccicheck's settings using SPFLAGS.
276
277We help Coccinelle when used against Linux with a set of sensible default
278options for Linux with our own Linux .cocciconfig. This hints to coccinelle
279that git can be used for ``git grep`` queries over coccigrep. A timeout of 200
280seconds should suffice for now.
281
282The options picked up by coccinelle when reading a .cocciconfig do not appear
283as arguments to spatch processes running on your system. To confirm what
284options will be used by Coccinelle run::
285
286      spatch --print-options-only
287
288You can override with your own preferred index option by using SPFLAGS. Take
289note that when there are conflicting options Coccinelle takes precedence for
290the last options passed. Using .cocciconfig is possible to use idutils, however
291given the order of precedence followed by Coccinelle, since the kernel now
292carries its own .cocciconfig, you will need to use SPFLAGS to use idutils if
293desired. See below section "Additional flags" for more details on how to use
294idutils.
295
296Additional flags
297----------------
298
299Additional flags can be passed to spatch through the SPFLAGS
300variable. This works as Coccinelle respects the last flags
301given to it when options are in conflict. ::
302
303    make SPFLAGS=--use-glimpse coccicheck
304
305Coccinelle supports idutils as well but requires coccinelle >= 1.0.6.
306When no ID file is specified coccinelle assumes your ID database file
307is in the file .id-utils.index on the top level of the kernel. Coccinelle
308carries a script scripts/idutils_index.sh which creates the database with::
309
310    mkid -i C --output .id-utils.index
311
312If you have another database filename you can also just symlink with this
313name. ::
314
315    make SPFLAGS=--use-idutils coccicheck
316
317Alternatively you can specify the database filename explicitly, for
318instance::
319
320    make SPFLAGS="--use-idutils /full-path/to/ID" coccicheck
321
322See ``spatch --help`` to learn more about spatch options.
323
324Note that the ``--use-glimpse`` and ``--use-idutils`` options
325require external tools for indexing the code. None of them is
326thus active by default. However, by indexing the code with
327one of these tools, and according to the cocci file used,
328spatch could proceed the entire code base more quickly.
329
330SmPL patch specific options
331---------------------------
332
333SmPL patches can have their own requirements for options passed
334to Coccinelle. SmPL patch-specific options can be provided by
335providing them at the top of the SmPL patch, for instance::
336
337	// Options: --no-includes --include-headers
338
339SmPL patch Coccinelle requirements
340----------------------------------
341
342As Coccinelle features get added some more advanced SmPL patches
343may require newer versions of Coccinelle. If an SmPL patch requires
344a minimum version of Coccinelle, this can be specified as follows,
345as an example if requiring at least Coccinelle >= 1.0.5::
346
347	// Requires: 1.0.5
348
349Proposing new semantic patches
350------------------------------
351
352New semantic patches can be proposed and submitted by kernel
353developers. For sake of clarity, they should be organized in the
354sub-directories of ``scripts/coccinelle/``.
355
356
357Detailed description of the ``report`` mode
358-------------------------------------------
359
360``report`` generates a list in the following format::
361
362  file:line:column-column: message
363
364Example
365~~~~~~~
366
367Running::
368
369	make coccicheck MODE=report COCCI=scripts/coccinelle/api/err_cast.cocci
370
371will execute the following part of the SmPL script::
372
373   <smpl>
374   @r depends on !context && !patch && (org || report)@
375   expression x;
376   position p;
377   @@
378
379     ERR_PTR@p(PTR_ERR(x))
380
381   @script:python depends on report@
382   p << r.p;
383   x << r.x;
384   @@
385
386   msg="ERR_CAST can be used with %s" % (x)
387   coccilib.report.print_report(p[0], msg)
388   </smpl>
389
390This SmPL excerpt generates entries on the standard output, as
391illustrated below::
392
393    /home/user/linux/crypto/ctr.c:188:9-16: ERR_CAST can be used with alg
394    /home/user/linux/crypto/authenc.c:619:9-16: ERR_CAST can be used with auth
395    /home/user/linux/crypto/xts.c:227:9-16: ERR_CAST can be used with alg
396
397
398Detailed description of the ``patch`` mode
399------------------------------------------
400
401When the ``patch`` mode is available, it proposes a fix for each problem
402identified.
403
404Example
405~~~~~~~
406
407Running::
408
409	make coccicheck MODE=patch COCCI=scripts/coccinelle/api/err_cast.cocci
410
411will execute the following part of the SmPL script::
412
413    <smpl>
414    @ depends on !context && patch && !org && !report @
415    expression x;
416    @@
417
418    - ERR_PTR(PTR_ERR(x))
419    + ERR_CAST(x)
420    </smpl>
421
422This SmPL excerpt generates patch hunks on the standard output, as
423illustrated below::
424
425    diff -u -p a/crypto/ctr.c b/crypto/ctr.c
426    --- a/crypto/ctr.c 2010-05-26 10:49:38.000000000 +0200
427    +++ b/crypto/ctr.c 2010-06-03 23:44:49.000000000 +0200
428    @@ -185,7 +185,7 @@ static struct crypto_instance *crypto_ct
429 	alg = crypto_attr_alg(tb[1], CRYPTO_ALG_TYPE_CIPHER,
430 				  CRYPTO_ALG_TYPE_MASK);
431 	if (IS_ERR(alg))
432    -		return ERR_PTR(PTR_ERR(alg));
433    +		return ERR_CAST(alg);
434
435 	/* Block size must be >= 4 bytes. */
436 	err = -EINVAL;
437
438Detailed description of the ``context`` mode
439--------------------------------------------
440
441``context`` highlights lines of interest and their context
442in a diff-like style.
443
444      **NOTE**: The diff-like output generated is NOT an applicable patch. The
445      intent of the ``context`` mode is to highlight the important lines
446      (annotated with minus, ``-``) and gives some surrounding context
447      lines around. This output can be used with the diff mode of
448      Emacs to review the code.
449
450Example
451~~~~~~~
452
453Running::
454
455	make coccicheck MODE=context COCCI=scripts/coccinelle/api/err_cast.cocci
456
457will execute the following part of the SmPL script::
458
459    <smpl>
460    @ depends on context && !patch && !org && !report@
461    expression x;
462    @@
463
464    * ERR_PTR(PTR_ERR(x))
465    </smpl>
466
467This SmPL excerpt generates diff hunks on the standard output, as
468illustrated below::
469
470    diff -u -p /home/user/linux/crypto/ctr.c /tmp/nothing
471    --- /home/user/linux/crypto/ctr.c	2010-05-26 10:49:38.000000000 +0200
472    +++ /tmp/nothing
473    @@ -185,7 +185,6 @@ static struct crypto_instance *crypto_ct
474 	alg = crypto_attr_alg(tb[1], CRYPTO_ALG_TYPE_CIPHER,
475 				  CRYPTO_ALG_TYPE_MASK);
476 	if (IS_ERR(alg))
477    -		return ERR_PTR(PTR_ERR(alg));
478
479 	/* Block size must be >= 4 bytes. */
480 	err = -EINVAL;
481
482Detailed description of the ``org`` mode
483----------------------------------------
484
485``org`` generates a report in the Org mode format of Emacs.
486
487Example
488~~~~~~~
489
490Running::
491
492	make coccicheck MODE=org COCCI=scripts/coccinelle/api/err_cast.cocci
493
494will execute the following part of the SmPL script::
495
496    <smpl>
497    @r depends on !context && !patch && (org || report)@
498    expression x;
499    position p;
500    @@
501
502      ERR_PTR@p(PTR_ERR(x))
503
504    @script:python depends on org@
505    p << r.p;
506    x << r.x;
507    @@
508
509    msg="ERR_CAST can be used with %s" % (x)
510    msg_safe=msg.replace("[","@(").replace("]",")")
511    coccilib.org.print_todo(p[0], msg_safe)
512    </smpl>
513
514This SmPL excerpt generates Org entries on the standard output, as
515illustrated below::
516
517    * TODO [[view:/home/user/linux/crypto/ctr.c::face=ovl-face1::linb=188::colb=9::cole=16][ERR_CAST can be used with alg]]
518    * TODO [[view:/home/user/linux/crypto/authenc.c::face=ovl-face1::linb=619::colb=9::cole=16][ERR_CAST can be used with auth]]
519    * TODO [[view:/home/user/linux/crypto/xts.c::face=ovl-face1::linb=227::colb=9::cole=16][ERR_CAST can be used with alg]]
520