1.. Copyright 2004 Linus Torvalds
2.. Copyright 2004 Pavel Machek <pavel@ucw.cz>
3.. Copyright 2006 Bob Copeland <me@bobcopeland.com>
4
5Sparse
6======
7
8Sparse is a semantic checker for C programs; it can be used to find a
9number of potential problems with kernel code.  See
10https://lwn.net/Articles/689907/ for an overview of sparse; this document
11contains some kernel-specific sparse information.
12
13
14Using sparse for typechecking
15-----------------------------
16
17"__bitwise" is a type attribute, so you have to do something like this::
18
19        typedef int __bitwise pm_request_t;
20
21        enum pm_request {
22                PM_SUSPEND = (__force pm_request_t) 1,
23                PM_RESUME = (__force pm_request_t) 2
24        };
25
26which makes PM_SUSPEND and PM_RESUME "bitwise" integers (the "__force" is
27there because sparse will complain about casting to/from a bitwise type,
28but in this case we really _do_ want to force the conversion). And because
29the enum values are all the same type, now "enum pm_request" will be that
30type too.
31
32And with gcc, all the "__bitwise"/"__force stuff" goes away, and it all
33ends up looking just like integers to gcc.
34
35Quite frankly, you don't need the enum there. The above all really just
36boils down to one special "int __bitwise" type.
37
38So the simpler way is to just do::
39
40        typedef int __bitwise pm_request_t;
41
42        #define PM_SUSPEND ((__force pm_request_t) 1)
43        #define PM_RESUME ((__force pm_request_t) 2)
44
45and you now have all the infrastructure needed for strict typechecking.
46
47One small note: the constant integer "0" is special. You can use a
48constant zero as a bitwise integer type without sparse ever complaining.
49This is because "bitwise" (as the name implies) was designed for making
50sure that bitwise types don't get mixed up (little-endian vs big-endian
51vs cpu-endian vs whatever), and there the constant "0" really _is_
52special.
53
54Using sparse for lock checking
55------------------------------
56
57The following macros are undefined for gcc and defined during a sparse
58run to use the "context" tracking feature of sparse, applied to
59locking.  These annotations tell sparse when a lock is held, with
60regard to the annotated function's entry and exit.
61
62__must_hold - The specified lock is held on function entry and exit.
63
64__acquires - The specified lock is held on function exit, but not entry.
65
66__releases - The specified lock is held on function entry, but not exit.
67
68If the function enters and exits without the lock held, acquiring and
69releasing the lock inside the function in a balanced way, no
70annotation is needed.  The tree annotations above are for cases where
71sparse would otherwise report a context imbalance.
72
73Getting sparse
74--------------
75
76You can get latest released versions from the Sparse homepage at
77https://sparse.wiki.kernel.org/index.php/Main_Page
78
79Alternatively, you can get snapshots of the latest development version
80of sparse using git to clone::
81
82        git://git.kernel.org/pub/scm/devel/sparse/sparse.git
83
84DaveJ has hourly generated tarballs of the git tree available at::
85
86        http://www.codemonkey.org.uk/projects/git-snapshots/sparse/
87
88
89Once you have it, just do::
90
91        make
92        make install
93
94as a regular user, and it will install sparse in your ~/bin directory.
95
96Using sparse
97------------
98
99Do a kernel make with "make C=1" to run sparse on all the C files that get
100recompiled, or use "make C=2" to run sparse on the files whether they need to
101be recompiled or not.  The latter is a fast way to check the whole tree if you
102have already built it.
103
104The optional make variable CF can be used to pass arguments to sparse.  The
105build system passes -Wbitwise to sparse automatically.
106