Utilities

This page contains reference documentation for <sys/util.h>, which provides miscellaneous utility functions and macros.

group sys-util

Defines

POINTER_TO_UINT(x)

Cast x, a pointer, to an unsigned integer.

UINT_TO_POINTER(x)

Cast x, an unsigned integer, to a void*.

POINTER_TO_INT(x)

Cast x, a pointer, to a signed integer.

INT_TO_POINTER(x)

Cast x, a signed integer, to a void*.

BITS_PER_LONG

Number of bits in a long int.

GENMASK(h, l)

Create a contiguous bitmask starting at bit position l and ending at position h.

ZERO_OR_COMPILE_ERROR(cond)

0 if cond is true-ish; causes a compile error otherwise.

IS_ARRAY(array)

Zero if array has an array type, a compile error otherwise.

This macro is available only from C, not C++.

ARRAY_SIZE(array)

Number of elements in the given array.

In C++, due to language limitations, this will accept as array any type that implements operator[]. The results may not be particulary meaningful in this case.

In C, passing a pointer as array causes a compile error.

PART_OF_ARRAY(array, ptr)

Check if a pointer ptr lies within array.

In C but not C++, this causes a compile error if array is not an array (e.g. if ptr and array are mixed up).

Return

1 if ptr is part of array, 0 otherwise

Parameters
  • ptr: a pointer

  • array: an array

CONTAINER_OF(ptr, type, field)

Get a pointer to a container structure from an element.

Example:

 struct foo {
    int bar;
 };

 struct foo my_foo;
 int *ptr = &my_foo.bar;

 struct foo *container = CONTAINER_OF(ptr, struct foo, bar);

Above, container points at my_foo.

Return

a pointer to the structure that contains ptr

Parameters
  • ptr: pointer to a structure element

  • type: name of the type that ptr is an element of

  • field: the name of the field within the struct ptr points to

ROUND_UP(x, align)

Value of x rounded up to the next multiple of align, which must be a power of 2.

ROUND_DOWN(x, align)

Value of x rounded down to the previous multiple of align, which must be a power of 2.

WB_UP(x)

Value of x rounded up to the next word boundary.

WB_DN(x)

Value of x rounded down to the previous word boundary.

ceiling_fraction(numerator, divider)

Ceiling function applied to numerator / divider as a fraction.

MAX(a, b)

The larger value between a and b.

Note

Arguments are evaluated twice.

MIN(a, b)

The smaller value between a and b.

Note

Arguments are evaluated twice.

CLAMP(val, low, high)

Clamp a value to a given range.

Note

Arguments are evaluated multiple times.

KB(x)

Number of bytes in x kibibytes.

MB(x)

Number of bytes in x mebibytes.

GB(x)

Number of bytes in x gibibytes.

KHZ(x)

Number of Hz in x kHz.

MHZ(x)

Number of Hz in x MHz.

BIT(n)

Unsigned integer with bit position n set (signed in assembly language).

BIT64(_n)

64-bit unsigned integer with bit position _n set.

WRITE_BIT(var, bit, set)

Set or clear a bit depending on a boolean value.

The argument var is a variable whose value is written to as a side effect.

Parameters
  • var: Variable to be altered

  • bit: Bit number

  • set: if 0, clears bit in var; any other value sets bit

BIT_MASK(n)

Bit mask with bits 0 through n-1 (inclusive) set, or 0 if n is 0.

BIT64_MASK(n)

64-bit bit mask with bits 0 through n-1 (inclusive) set, or 0 if n is 0.

IS_ENABLED(config_macro)

Check for macro definition in compiler-visible expressions.

This trick was pioneered in Linux as the config_enabled() macro. It has the effect of taking a macro value that may be defined to “1” or may not be defined at all and turning it into a literal expression that can be handled by the C compiler instead of just the preprocessor. It is often used with a CONFIG_FOO macro which may be defined to 1 via Kconfig, or left undefined.

That is, it works similarly to #if defined(CONFIG_FOO) except that its expansion is a C expression. Thus, much #ifdef usage can be replaced with equivalents like:

if (IS_ENABLED(CONFIG_FOO)) {
        do_something_with_foo
}

This is cleaner since the compiler can generate errors and warnings for do_something_with_foo even when CONFIG_FOO is undefined.

Return

1 if config_macro is defined to 1, 0 otherwise (including if config_macro is not defined)

Parameters
  • config_macro: Macro to check

COND_CODE_1(_flag, _if_1_code, _else_code)

Insert code depending on whether _flag expands to 1 or not.

This relies on similar tricks as IS_ENABLED(), but as the result of _flag expansion, results in either _if_1_code or _else_code is expanded.

To prevent the preprocessor from treating commas as argument separators, the _if_1_code and _else_code expressions must be inside brackets/parentheses: (). These are stripped away during macro expansion.

Example:

COND_CODE_1(CONFIG_FLAG, (uint32_t x;), (there_is_no_flag();))

If CONFIG_FLAG is defined to 1, this expands to:

uint32_t x;

It expands to there_is_no_flag(); otherwise.

This could be used as an alternative to:

#if defined(CONFIG_FLAG) && (CONFIG_FLAG == 1)
#define MAYBE_DECLARE(x) uint32_t x
#else
#define MAYBE_DECLARE(x) there_is_no_flag()
#endif

MAYBE_DECLARE(x);

However, the advantage of COND_CODE_1() is that code is resolved in place where it is used, while the #if method defines MAYBE_DECLARE on two lines and requires it to be invoked again on a separate line. This makes COND_CODE_1() more concise and also sometimes more useful when used within another macro’s expansion.

Note

_flag can be the result of preprocessor expansion, e.g. an expression involving NUM_VA_ARGS_LESS_1(…). However, _if_1_code is only expanded if _flag expands to the integer literal 1. Integer expressions that evaluate to 1, e.g. after doing some arithmetic, will not work.

Parameters
  • _flag: evaluated flag

  • _if_1_code: result if _flag expands to 1; must be in parentheses

  • _else_code: result otherwise; must be in parentheses

COND_CODE_0(_flag, _if_0_code, _else_code)

Like COND_CODE_1() except tests if _flag is 0.

This is like COND_CODE_1(), except that it tests whether _flag expands to the integer literal 0. It expands to _if_0_code if so, and _else_code otherwise; both of these must be enclosed in parentheses.

See

COND_CODE_1()

Parameters
  • _flag: evaluated flag

  • _if_0_code: result if _flag expands to 0; must be in parentheses

  • _else_code: result otherwise; must be in parentheses

IF_ENABLED(_flag, _code)

Insert code if _flag is defined and equals 1.

Like COND_CODE_1(), this expands to _code if _flag is defined to 1; it expands to nothing otherwise.

Example:

IF_ENABLED(CONFIG_FLAG, (uint32_t foo;))

If CONFIG_FLAG is defined to 1, this expands to:

uint32_t foo;

and to nothing otherwise.

It can be considered as a more compact alternative to:

#if defined(CONFIG_FLAG) && (CONFIG_FLAG == 1)
uint32_t foo;
#endif

Parameters
  • _flag: evaluated flag

  • _code: result if _flag expands to 1; must be in parentheses

IS_EMPTY(a)

Check if a macro has a replacement expression.

If a is a macro defined to a nonempty value, this will return true, otherwise it will return false. It only works with defined macros, so an additional #ifdef test may be needed in some cases.

This macro may be used with COND_CODE_1() and COND_CODE_0() while processing __VA_ARGS__ to avoid processing empty arguments.

Note that this macro is intended to check macro names that evaluate to replacement lists being empty or containing numbers or macro name like tokens.

Example:

 #define EMPTY
 #define NON_EMPTY  1
 #undef  UNDEFINED
 IS_EMPTY(EMPTY)
 IS_EMPTY(NON_EMPTY)
 IS_EMPTY(UNDEFINED)
 #if defined(EMPTY) && IS_EMPTY(EMPTY) == true
 some_conditional_code
 #endif
Note

Not all arguments are accepted by this macro and compilation will fail if argument cannot be concatenated with literal constant. That will happen if argument does not start with letter or number. Example arguments that will fail during compilation: .arg, (arg), “arg”, {arg}.

In above examples, the invocations of IS_EMPTY(…) return true, false, and true; some_conditional_code is included.

Parameters
  • a: macro to check for emptiness

LIST_DROP_EMPTY(...)

Remove empty arguments from list.

During macro expansion, __VA_ARGS__ and other preprocessor generated lists may contain empty elements, e.g.:

 #define LIST ,a,b,,d,

Using EMPTY to show each empty element, LIST contains:

 EMPTY, a, b, EMPTY, d

When processing such lists, e.g. using FOR_EACH(), all empty elements will be processed, and may require filtering out. To make that process easier, it is enough to invoke LIST_DROP_EMPTY which will remove all empty elements.

Example:

 LIST_DROP_EMPTY(LIST)

expands to:

 a, b, d

Parameters
  • ...: list to be processed

EMPTY

Macro with an empty expansion.

This trivial definition is provided for readability when a macro should expand to an empty result, which e.g. is sometimes needed to silence checkpatch.

Example:

 #define LIST_ITEM(n) , item##n

The above would cause checkpatch to complain, but:

 #define LIST_ITEM(n) EMPTY, item##n

would not.

IDENTITY(V)

Macro that expands to its argument.

This is useful in macros like FOR_EACH() when there is no transformation required on the list elements.

Parameters
  • V: any value

GET_ARG_N(N, ...)

Get nth argument from argument list.

Return

Nth argument.

Parameters
  • N: Argument index to fetch. Counter from 1.

  • ...: Variable list of argments from which one argument is returned.

GET_ARGS_LESS_N(N, ...)

Strips n first arguments from the argument list.

Return

argument list without N first arguments.

Parameters
  • N: Number of arguments to discard.

  • ...: Variable list of argments.

GET_ARG1(...)

Expands to the first argument.

GET_ARG2(...)

Expands to the second argument.

GET_ARGS_LESS_1(...)

Expands to all arguments except the first one.

UTIL_OR(a, b)

Like a || b, but does evaluation and short-circuiting at C preprocessor time.

This is not the same as the binary || operator; in particular, a should expand to an integer literal 0 or 1. However, b can be any value.

This can be useful when b is an expression that would cause a build error when a is 1.

UTIL_AND(a, b)

Like a && b, but does evaluation and short-circuiting at C preprocessor time.

This is not the same as the binary &&, however; in particular, a should expand to an integer literal 0 or 1. However, b can be any value.

This can be useful when b is an expression that would cause a build error when a is 0.

UTIL_LISTIFY(LEN, F, ...)

Generates a sequence of code.

Example:

#define FOO(i, _) MY_PWM ## i ,
{ UTIL_LISTIFY(PWM_COUNT, FOO) }

The above two lines expand to:

{ MY_PWM0 , MY_PWM1 , }

Note

Calling UTIL_LISTIFY with undefined arguments has undefined behavior.

Parameters
  • LEN: The length of the sequence. Must be an integer literal less than 255.

  • F: A macro function that accepts at least two arguments: F(i, ...). F is called repeatedly in the expansion. Its first argument i is the index in the sequence, and the variable list of arguments passed to UTIL_LISTIFY are passed through to F.

FOR_EACH(F, sep, ...)

Call a macro F on each provided argument with a given separator between each call.

Example:

#define F(x) int a##x
FOR_EACH(F, (;), 4, 5, 6);

This expands to:

int a4;
int a5;
int a6;

Parameters
  • F: Macro to invoke

  • sep: Separator (e.g. comma or semicolon). Must be in parentheses; this is required to enable providing a comma as separator.

  • ...: Variable argument list. The macro F is invoked as F(element) for each element in the list.

FOR_EACH_NONEMPTY_TERM(F, term, ...)

Like FOR_EACH(), but with a terminator instead of a separator, and drops empty elements from the argument list.

The sep argument to FOR_EACH(F, (sep), a, b) is a separator which is placed between calls to F, like this:

FOR_EACH(F, (sep), a, b) // F(a) sep F(b)
                         //               ^^^ no sep here!

By contrast, the term argument to FOR_EACH_NONEMPTY_TERM(F, (term), a, b) is added after each time F appears in the expansion:

FOR_EACH_NONEMPTY_TERM(F, (term), a, b) // F(a) term F(b) term
                                        //                ^^^^

Further, any empty elements are dropped:

FOR_EACH_NONEMPTY_TERM(F, (term), a, EMPTY, b) // F(a) term F(b) term

This is more convenient in some cases, because FOR_EACH_NONEMPTY_TERM() expands to nothing when given an empty argument list, and it’s often cumbersome to write a macro F that does the right thing even when given an empty argument.

One example is when __VA_ARGS__ may or may not be empty, and the results are embedded in a larger initializer:

#define SQUARE(x) ((x)*(x))

int my_array[] = {
        FOR_EACH_NONEMPTY_TERM(SQUARE, (,), FOO(...))
        FOR_EACH_NONEMPTY_TERM(SQUARE, (,), BAR(...))
        FOR_EACH_NONEMPTY_TERM(SQUARE, (,), BAZ(...))
};

This is more convenient than:

  1. figuring out whether the FOO, BAR, and BAZ expansions are empty and adding a comma manually (or not) between FOR_EACH() calls

  2. rewriting SQUARE so it reacts appropriately when “x” is empty (which would be necessary if e.g. FOO expands to nothing)

Parameters
  • F: Macro to invoke on each nonempty element of the variable arguments

  • term: Terminator (e.g. comma or semicolon) placed after each invocation of F. Must be in parentheses; this is required to enable providing a comma as separator.

  • ...: Variable argument list. The macro F is invoked as F(element) for each nonempty element in the list.

FOR_EACH_IDX(F, sep, ...)

Call macro F on each provided argument, with the argument’s index as an additional parameter.

This is like FOR_EACH(), except F should be a macro which takes two arguments: F(index, variable_arg).

Example:

#define F(idx, x) int a##idx = x
FOR_EACH_IDX(F, (;), 4, 5, 6);

This expands to:

int a0 = 4;
int a1 = 5;
int a2 = 6;

Parameters
  • F: Macro to invoke

  • sep: Separator (e.g. comma or semicolon). Must be in parentheses; this is required to enable providing a comma as separator.

  • ...: Variable argument list. The macro F is invoked as F(index, element) for each element in the list.

FOR_EACH_FIXED_ARG(F, sep, fixed_arg, ...)

Call macro F on each provided argument, with an additional fixed argument as a parameter.

This is like FOR_EACH(), except F should be a macro which takes two arguments: F(variable_arg, fixed_arg).

Example:

static void func(int val, void *dev);
FOR_EACH_FIXED_ARG(func, (;), dev, 4, 5, 6);

This expands to:

func(4, dev);
func(5, dev);
func(6, dev);

Parameters
  • F: Macro to invoke

  • sep: Separator (e.g. comma or semicolon). Must be in parentheses; this is required to enable providing a comma as separator.

  • fixed_arg: Fixed argument passed to F as the second macro parameter.

  • ...: Variable argument list. The macro F is invoked as F(element, fixed_arg) for each element in the list.

FOR_EACH_IDX_FIXED_ARG(F, sep, fixed_arg, ...)

Calls macro F for each variable argument with an index and fixed argument.

This is like the combination of FOR_EACH_IDX() with FOR_EACH_FIXED_ARG().

Example:

#define F(idx, x, fixed_arg) int fixed_arg##idx = x
FOR_EACH_IDX_FIXED_ARG(F, (;), a, 4, 5, 6);

This expands to:

int a0 = 4;
int a1 = 5;
int a2 = 6;

Parameters
  • F: Macro to invoke

  • sep: Separator (e.g. comma or semicolon). Must be in parentheses; This is required to enable providing a comma as separator.

  • fixed_arg: Fixed argument passed to F as the third macro parameter.

  • ...: Variable list of arguments. The macro F is invoked as F(index, element, fixed_arg) for each element in the list.

NUM_VA_ARGS_LESS_1(...)

Number of arguments in the variable arguments list minus one.

Return

Number of variadic arguments in the argument list, minus one

Parameters
  • ...: List of arguments

MACRO_MAP_CAT(...)

Mapping macro that pastes results together.

This is similar to FOR_EACH() in that it invokes a macro repeatedly on each element of __VA_ARGS__. However, unlike FOR_EACH(), MACRO_MAP_CAT() pastes the results together into a single token.

For example, with this macro FOO:

#define FOO(x) item_##x##_

MACRO_MAP_CAT(FOO, a, b, c), expands to the token:

item_a_item_b_item_c_

Return

The results of expanding the macro on each argument, all pasted together

Parameters
  • ...: Macro to expand on each argument, followed by its arguments. (The macro should take exactly one argument.)

MACRO_MAP_CAT_N(N, ...)

Mapping macro that pastes a fixed number of results together.

Similar to MACRO_MAP_CAT(), but expects a fixed number of arguments. If more arguments are given than are expected, the rest are ignored.

Return

The results of expanding the macro on each argument, all pasted together

Parameters
  • N: Number of arguments to map

  • ...: Macro to expand on each argument, followed by its arguments. (The macro should take exactly one argument.)

Functions

bool is_power_of_two(unsigned int x)

Is x a power of two?

Return

true if x is a power of two, false otherwise

Parameters
  • x: value to check

int64_t arithmetic_shift_right(int64_t value, uint8_t shift)

Arithmetic shift right.

Return

value shifted right by shift; opened bit positions are filled with the sign bit

Parameters
  • value: value to shift

  • shift: number of bits to shift

int char2hex(char c, uint8_t *x)

Convert a single character into a hexadecimal nibble.

Return

Zero on success or (negative) error code otherwise.

Parameters
  • c: The character to convert

  • x: The address of storage for the converted number.

int hex2char(uint8_t x, char *c)

Convert a single hexadecimal nibble into a character.

Return

Zero on success or (negative) error code otherwise.

Parameters
  • c: The number to convert

  • x: The address of storage for the converted character.

size_t bin2hex(const uint8_t *buf, size_t buflen, char *hex, size_t hexlen)

Convert a binary array into string representation.

Return

The length of the converted string, or 0 if an error occurred.

Parameters
  • buf: The binary array to convert

  • buflen: The length of the binary array to convert

  • hex: Address of where to store the string representation.

  • hexlen: Size of the storage area for string representation.

size_t hex2bin(const char *hex, size_t hexlen, uint8_t *buf, size_t buflen)

Convert a hexadecimal string into a binary array.

Return

The length of the binary array, or 0 if an error occurred.

Parameters
  • hex: The hexadecimal string to convert

  • hexlen: The length of the hexadecimal string to convert.

  • buf: Address of where to store the binary data

  • buflen: Size of the storage area for binary data

uint8_t u8_to_dec(char *buf, uint8_t buflen, uint8_t value)

Convert a uint8_t into a decimal string representation.

Convert a uint8_t value into its ASCII decimal string representation. The string is terminated if there is enough space in buf.

Return

The length of the converted string (excluding terminator if any), or 0 if an error occurred.

Parameters
  • buf: Address of where to store the string representation.

  • buflen: Size of the storage area for string representation.

  • value: The value to convert to decimal string