kernel: fix a type bug handling err_cast()

[smatch.git] / sparse.1

.\" Sparse manpage by Josh Triplett
.TH sparse "1"
.
.SH NAME
sparse \- Semantic Parser for C
.
.SH SYNOPSIS
.B sparse
[\fIWARNING OPTIONS\fR]... \fIfile.c\fR
.
.SH DESCRIPTION
Sparse parses C source and looks for errors, producing warnings on standard
error.
.P
Sparse accepts options controlling the set of warnings to generate.  To turn
on warnings Sparse does not issue by default, use the corresponding warning
option \fB\-Wsomething\fR.  Sparse issues some warnings by default; to turn
off those warnings, pass the negation of the associated warning option,
\fB\-Wno\-something\fR.
.
.SH WARNING OPTIONS
.TP
.B \-fmax-warnings=COUNT
Set the maximum number of displayed warnings to COUNT, which should be
a numerical value or 'unlimited'.
The default limit is 100.
.
.TP
.B \-Wsparse\-all
Turn on all sparse warnings, except for those explicitly disabled via
\fB\-Wno\-something\fR.
.TP
.B \-Wsparse\-error
Turn all sparse warnings into errors.
.TP
.B \-Waddress\-space
Warn about code which mixes pointers to different address spaces.

Sparse allows an extended attribute
.BI __attribute__((address_space( id )))
on pointers, which designates a pointer target in address space \fIid\fR (an
identifier or a constant integer).
With \fB\-Waddress\-space\fR, Sparse treats pointers with
identical target types but different address spaces as distinct types and
will warn accordingly.

Sparse will also warn on casts which remove the address space (casts to an
integer type or to a plain pointer type). An exception to this is when the
destination type is \fBuintptr_t\fR (or \fBunsigned long\fR) since such casts
are often used to "get a pointer value representation in an integer type" and
such values are independent of the address space.

To override these warnings, use a type that includes \fB__attribute__((force))\fR.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-address\-space\fR.
.
.TP
.B \-Wbitwise
Warn about unsupported operations or type mismatches with restricted integer
types.

Sparse supports an extended attribute, \fB__attribute__((bitwise))\fR, which
creates a new restricted integer type from a base integer type, distinct from
the base integer type and from any other restricted integer type not declared
in the same declaration or \fBtypedef\fR.  For example, this allows programs
to create \fBtypedef\fRs for integer types with specific endianness.  With
\fB-Wbitwise\fR, Sparse will warn on any use of a restricted type in
arithmetic operations other than bitwise operations, and on any conversion of
one restricted type into another, except via a cast that includes
\fB__attribute__((force))\fR.

__bitwise ends up being a "stronger integer separation", one that
doesn't allow you to mix with non-bitwise integers, so now it's much
harder to lose the type by mistake.

__bitwise is for *unique types* that cannot be mixed with other
types, and that you'd never want to just use as a random integer (the
integer 0 is special, though, and gets silently accepted iirc - it's
kind of like "NULL" for pointers). So "gfp_t" or the "safe endianness"
types would be __bitwise: you can only operate on them by doing
specific operations that know about *that* particular type.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-bitwise\fR.
.
.TP
.B \-Wbitwise\-pointer
Same as \fB\-Wbitwise\fR but for casts to or from pointers to bitwise types.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wcast\-from\-as
Warn about casts which remove an address space from a pointer type.

This is similar to \fB\-Waddress\-space\fR but will also warn
on casts to \fBunsigned long\fR.

Sparse does not issues these warnings by default.
.
.TP
.B \-Wcast\-to\-as
Warn about casts which add an address space to a pointer type.

A cast that includes \fB__attribute__((force))\fR will suppress this warning.
No warning is generated if the original type is \fBuintptr_t\fR
(or \fBunsigned long\fR).

Sparse does not issue these warnings by default.
.
.TP
.B \-Wcast\-truncate
Warn about casts that truncate constant values.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-cast\-truncate\fR.
.
.TP
.B \-Wconstant\-suffix
Warn if an integer constant is larger than the maximum representable value
of the type indicated by its type suffix (if any). For example, on a
system where ints are 32-bit and longs 64-bit, the constant \fB0x100000000U\fR
is larger than can be represented by an \fBunsigned int\fR but fits in an
\fBunsigned long\fR. So its type is \fBunsigned long\fR but this is not
indicated by its suffix. In this case, the warning could be suppressed by
using the suffix \fBUL\fR: \fB0x100000000UL\fR.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wconstexpr-not-const
Warn if a non-constant expression is encountered when really expecting a
constant expression instead.
Currently, this warns when initializing an object of static storage duration
with an initializer which is not a constant expression.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wcontext
Warn about potential errors in synchronization or other delimited contexts.

Sparse supports several means of designating functions or statements that
delimit contexts, such as synchronization.  Functions with the extended
attribute
.BI __attribute__((context( expression , in_context , out_context ))
require the context \fIexpression\fR (for instance, a lock) to have the value
\fIin_context\fR (a constant nonnegative integer) when called, and return with
the value \fIout_context\fR (a constant nonnegative integer).  For APIs
defined via macros, use the statement form
.BI __context__( expression , in_value , out_value )
in the body of the macro.

With \fB-Wcontext\fR Sparse will warn when it sees a function change the
context without indicating this with a \fBcontext\fR attribute, either by
decreasing a context below zero (such as by releasing a lock without acquiring
it), or returning with a changed context (such as by acquiring a lock without
releasing it).  Sparse will also warn about blocks of code which may
potentially execute with different contexts.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-context\fR.
.
.TP
.B \-Wdecl
Warn about any non-\fBstatic\fR variable or function definition that has no
previous declaration.

Private symbols (functions and variables) internal to a given source file
should use \fBstatic\fR, to allow additional compiler optimizations, allow
detection of unused symbols, and prevent other code from relying on these
internal symbols.  Public symbols used by other source files will need
declarations visible to those other source files, such as in a header file.
All declarations should fall into one of these two categories.  Thus, with
\fB-Wdecl\fR, Sparse warns about any symbol definition with neither
\fBstatic\fR nor a declaration.  To fix this warning, declare private symbols
\fBstatic\fR, and ensure that the files defining public symbols have the
symbol declarations available first (such as by including the appropriate
header file).

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-decl\fR.
.
.TP
.B \-Wdeclaration-after-statement
Warn about declarations that are not at the start of a block.

These declarations are permitted in C99 but not in C89.

Sparse issues these warnings by default only when the C dialect is
C89 (i.e. -ansi or -std=c89).  To turn them off, use
\fB\-Wno\-declaration\-after\-statement\fR.
.
.TP
.B \-Wdefault\-bitfield\-sign
Warn about any bitfield with no explicit signedness.

Bitfields have no standard-specified default signedness. (C99 6.7.2) A
bitfield without an explicit \fBsigned\fR or \fBunsigned\fR creates a
portability problem for software that relies on the available range of values.
To fix this, specify the bitfield type as \fBsigned\fR or \fBunsigned\fR
explicitly.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wdesignated\-init
Warn about positional initialization of structs marked as requiring designated
initializers.

Sparse allows an attribute
.BI __attribute__((designated_init))
which marks a struct as requiring designated initializers.  Sparse will warn
about positional initialization of a struct variable or struct literal of a
type that has this attribute.

Requiring designated initializers for a particular struct type will insulate
code using that struct type from changes to the layout of the type, avoiding
the need to change initializers for that type unless they initialize a removed
or incompatibly changed field.

Common examples of this type of struct include collections of function pointers
for the implementations of a class of related operations, for which the default
NULL for an unmentioned field in a designated initializer will correctly
indicate the absence of that operation.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-designated\-init\fR.
.
.TP
.B \-Wdo\-while
Warn about do-while loops that do not delimit the loop body with braces.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wenum\-mismatch
Warn about the use of an expression of an incorrect \fBenum\fR type when
initializing another \fBenum\fR type, assigning to another \fBenum\fR type, or
passing an argument to a function which expects another \fBenum\fR type.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-enum\-mismatch\fR.
.
.TP
.B \-Wexternal\-function\-has\-definition
Warn about function definitions that are declared with external linkage.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-external\-function\-has\-definition\fR.
.
.TP
.B \-Winit\-cstring
Warn about initialization of a char array with a too long constant C string.

If the size of the char array and the length of the string are the same,
there is no space for the last nul char of the string in the array:

.nf
char s[3] = "abc";
.fi

If the array is used as a byte array, not as C string, this
warning is just noise. However, if the array is passed to functions
dealing with C string like printf(%s) and strcmp, it may cause a
trouble.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wmemcpy\-max\-count
Warn about call of \fBmemcpy()\fR, \fBmemset()\fR, \fBcopy_from_user()\fR, or
\fBcopy_to_user()\fR with a large compile-time byte count.

Sparse issues these warnings by default. To turn them off, use
\fB\-Wno\-memcpy\-max\-count\fR.

The limit can be changed with \fB\-fmemcpy\-max\-count=COUNT\fR,
the default being \fB100000\fR.
.
.TP
.B \-Wnon\-pointer\-null
Warn about the use of 0 as a NULL pointer.

0 has integer type.  NULL has pointer type.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-non\-pointer\-null\fR.
.
.TP
.B \-Wold\-initializer
Warn about the use of the pre-C99 GCC syntax for designated initializers.

C99 provides a standard syntax for designated fields in \fBstruct\fR or
\fBunion\fR initializers:

.nf
struct structname var = { .field = value };
.fi

GCC also has an old, non-standard syntax for designated initializers which
predates C99:

.nf
struct structname var = { field: value };
.fi

Sparse will warn about the use of GCC's non-standard syntax for designated
initializers.  To fix this warning, convert designated initializers to use the
standard C99 syntax.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-old\-initializer\fR.
.
.TP
.B \-Wone\-bit\-signed\-bitfield
Warn about any one-bit \fBsigned\fR bitfields.

A one-bit \fBsigned\fR bitfield can only have the values 0 and -1, or with
some compilers only 0; this results in unexpected behavior for programs which
expected the ability to store 0 and 1.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-one\-bit\-signed\-bitfield\fR.
.
.TP
.B \-Wparen\-string
Warn about the use of a parenthesized string to initialize an array.

Standard C syntax does not permit a parenthesized string as an array
initializer.  GCC allows this syntax as an extension.  With
\fB\-Wparen\-string\fR, Sparse will warn about this syntax.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wpointer\-arith
Warn about anything that depends on the \fBsizeof\fR a void or function type.

C99 does not allow the \fBsizeof\fR operator to be applied to function types
or to incomplete types such as void. GCC allows \fBsizeof\fR to be applied to
these types as an extension and assigns these types a size of \fI1\fR. With
\fB\-pointer\-arith\fR, Sparse will warn about pointer arithmetic on void
or function pointers, as well as expressions which directly apply the
\fBsizeof\fR operator to void or function types.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wptr\-subtraction\-blows
Warn when subtracting two pointers to a type with a non-power-of-two size.

Subtracting two pointers to a given type gives a difference in terms of the
number of items of that type.  To generate this value, compilers will usually
need to divide the difference by the size of the type, an potentially
expensive operation for sizes other than powers of two.

Code written using pointer subtraction can often use another approach instead,
such as array indexing with an explicit array index variable, which may allow
compilers to generate more efficient code.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wreturn\-void
Warn if a function with return type void returns a void expression.

C99 permits this, and in some cases this allows for more generic code in
macros that use typeof or take a type as a macro argument.  However, some
programs consider this poor style, and those programs can use
\fB\-Wreturn\-void\fR to get warnings about it.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wshadow
Warn when declaring a symbol which shadows a declaration with the same name in
an outer scope.

Such declarations can lead to error-prone code.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wshift-count-negative
Warn if a shift count is negative.

Sparse issues these warnings by default.
.
.TP
.B \-Wshift-count-overflow
Warn if a shift count is bigger than the operand's width.

Sparse issues these warnings by default.
.
.TP
.B \-Wsizeof-bool
Warn when checking the sizeof a _Bool.

C99 does not specify the size of a _Bool. GCC, by default, uses \fI1\fR.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wtransparent\-union
Warn about any declaration using the GCC extension
\fB__attribute__((transparent_union))\fR.

Sparse issues these warnings by default.  To turn them off, use
\fB\-Wno\-transparent\-union\fR.
.
.TP
.B \-Wtypesign
Warn when converting a pointer to an integer type into a pointer to an integer
type with different signedness.

Sparse does not issue these warnings by default.
.
.TP
.B \-Wundef
Warn about preprocessor conditionals that use the value of an undefined
preprocessor symbol.

Standard C (C99 6.10.1) permits using the value of an undefined preprocessor
symbol in preprocessor conditionals, and specifies it has a value of 0.
However, this behavior can lead to subtle errors.

Sparse does not issue these warnings by default.
.
.SH MISC OPTIONS
.TP
.B \-gcc-base-dir \fIdir\fR
Look for compiler-provided system headers in \fIdir\fR/include/ and \fIdir\fR/include-fixed/.
.
.TP
.B \-multiarch-dir \fIdir\fR
Look for system headers in the multiarch subdirectory \fIdir\fR.
The \fIdir\fR name would normally take the form of the target's
normalized GNU triplet. (e.g. i386-linux-gnu).
.
.SH DEBUG OPTIONS
.TP
.B \-fmem-report
Report some statistics about memory allocation used by the tool.
.
.SH OTHER OPTIONS
.TP
.B \-fdiagnostic-prefix[=PREFIX]
Prefix all diagnostics by the given PREFIX, followed by ": ".
If no one is given "sparse" is used.
The default is to not use a prefix at all.
.
.TP
.B \-fmemcpy-max-count=COUNT
Set the limit for the warnings given by \fB-Wmemcpy-max-count\fR.
A COUNT of 'unlimited' or '0' will effectively disable the warning.
The default limit is 100000.
.
.TP
.B \-ftabstop=WIDTH
Set the distance between tab stops.  This helps sparse report correct
column numbers in warnings or errors.  If the value is less than 1 or
greater than 100, the option is ignored.  The default is 8.
.
.TP
.B \-f[no-]unsigned-char, \-f[no-]signed-char
Let plain 'char' be unsigned or signed.
By default chars are signed.
.
.SH SEE ALSO
.BR cgcc (1)
.
.SH HOMEPAGE
http://www.kernel.org/pub/software/devel/sparse/
.
.SH MAILING LIST
linux-sparse@vger.kernel.org
.
.SH CONTRIBUTINGS AND REPORTING BUGS
Submission of patches and reporting of bugs, as well as discussions
related to Sparse, should be done via the mailing list (linux-sparse@vger.kernel.org)
where the development and maintenance is primarily done.
You do not have to be subscribed to the list to send a message there.

Bugs can also be reported and tracked via the Linux kernel's bugzilla:
http://bugzilla.kernel.org/enter_bug.cgi?component=Sparse&product=Tools .
.
.SH AUTHORS
Sparse was started by Linus Torvalds.
The complete list of contributors can be find at
https://www.openhub.net/p/sparse/contributors .

Luc Van Oostenryck is Sparse's current maintainer.