Mirror of :pserver:cvs@cvs.fefe.de:/cvs libowfat https://www.fefe.de/libowfat/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

172 lines
6.7 KiB

/* this header file comes from libowfat, http://www.fefe.de/libowfat/ */
#include <inttypes.h>
#include <stddef.h>
#ifdef __cplusplus
extern "C" {
/* We are trying to achieve that gcc has to inline the function and we
* don't want it to emit a copy of the function. This can be done with
* static inline or with extern inline. static inline tells gcc to not
* emit a copy unless someone is using & to take a pointer, which nobody
* is ever supposed to do. extern inline tells gcc to not ever emit a
* copy.
* Unfortunately, the C99 standard defines extern inline to mean "always
* emit a copy for external reference", so this causes duplicate symbol
* linking errors. gcc signals C99 inline expansion mode by defining
* __GNUC_STDC_INLINE__ and it then has an attribute gnu_inline to
* switch back to GNU behavior. So that's what we are doing. Taking
* the address of one of these functions is considered a user error.
* We are so anal about inlining here because these checks can in most
* cases be optimized away. In particular, if you call this function
* often, gcc can see that some of the basic checks are done repeatedly
* and not do them again. But this only works if the function is
* inlined. */
#if defined(__GNUC_STDC_INLINE__)
#define __gnuinline __attribute__((gnu_inline))
#define __gnuinline
#if defined(__GNUC__) && !defined(__NO_INLINE__) && !defined(__clang__)
#define __static extern
#define __static static
#if !defined(__GNUC__) || (__GNUC__ < 3)
#define __builtin_expect(foo,bar) (foo)
#define __expect(foo,bar) (foo)
#define __expect(foo,bar) __builtin_expect((long)(foo),bar)
#if defined(__GNUC__) && !defined(__likely)
#define __likely(foo) __expect((foo),1)
#define __unlikely(foo) __expect((foo),0)
/* return 0 for range error / overflow, 1 for ok */
/* we assume the normal case is that the checked value is in range */
/* does ptr point to one of buf[0], buf[1], ... buf[len-1]? */
__static inline __gnuinline int range_ptrinbuf(const void* buf,size_t len,const void* ptr) {
register const char* c=(const char*)buf; /* no pointer arithmetic on void* */
return __likely(c && /* is buf non-NULL? */
((uintptr_t)c)+len>(uintptr_t)c && /* gcc 4.1 miscompiles without (uintptr_t) */
/* catch integer overflows and fail if buffer is 0 bytes long */
/* because then ptr can't point _in_ the buffer */
(uintptr_t)((const char*)ptr-c)<len); /* this one is a little tricky.
"ptr-c" checks the offset of ptr in the buffer is inside the buffer size.
Now, ptr-c can underflow; say it is -1. When we cast it to uintptr_t, it becomes
a very large number. */
/* same thing, but the buffer is specified by a pointer to the first
* byte (Min) and a pointer after the last byte (Max). */
__static inline __gnuinline int range_ptrinbuf2(const void* Min,const void* Max,const void* ptr) {
return __likely(Min && ptr>=Min && ptr<Max);
/* Min <= Max is implicitly checked here */
/* Is this a plausible buffer?
* Check whether buf is NULL, and whether buf+len overflows.
* Does NOT check whether buf has a non-zero length! */
__static inline __gnuinline int range_validbuf(const void* buf,size_t len) {
return __likely(buf && (uintptr_t)buf+len>=(uintptr_t)buf);
/* same thing but buffer is given as pointer to first byte (Min) and
* pointer beyond last byte (Max). Again, an 0-size buffer is valid. */
__static inline __gnuinline int range_validbuf2(const void* Min,const void* Max) {
return __likely(Min && Max>=Min);
/* is buf2[0..len2-1] inside buf1[0..len-1]? */
__static inline __gnuinline int range_bufinbuf(const void* buf1,size_t len1,const void* buf2,size_t len2) {
return range_validbuf(buf1,len1) &&
range_validbuf(buf2,len2) &&
__likely(buf1<=buf2 &&
/* does an array of "elements" members of size "membersize" starting at
* "arraystart" lie inside buf1[0..len-1]? */
int range_arrayinbuf(const void* buf,size_t len,
const void* arraystart,size_t elements,size_t membersize);
/* does an ASCIIZ string starting at "ptr" lie in buf[0..len-1]? */
int range_strinbuf(const void* buf,size_t len,const void* stringstart);
/* does an UTF-16 string starting at "ptr" lie in buf[0..len-1]? */
int range_str2inbuf(const void* buf,size_t len,const void* stringstart);
/* does an UTF-32 string starting at "ptr" lie in buf[0..len-1]? */
int range_str4inbuf(const void* buf,size_t len,const void* stringstart);
/* I originally omitted addition and substraction because it appeared
* trivial. You could just add the two numbers and see if it was
* smaller than either of them. This always worked for me because I
* only cared about unsigned arithmetic, but for signed arithmetic,
* adding two numbers is undefined if the result does not fit in the
* int. gcc has started to actually use this undefinedness to screw
* you. The following code illustrates this:
* int a=INT_MAX,b=a+5;
* if (b<a) abort(); // whole statement optimized away by gcc 4.1
* // at this point, b<a
* So I decided to add some integer overflow protection functionality
* here for addition and subtraction, too. */
/* usage:
* if (add_of(dest,a,b)) return EINVAL; // dest=a+b;
* if (sub_of(dest,a,b)) return EINVAL; // dest=a-b;
* if (assign(dest,some_int)) return EINVAL; // dest=some_int;
/* two important assumptions:
* 1. the platform is using two's complement
* 2. there are 8 bits in a byte
#define __HALF_MAX_SIGNED(type) ((type)1 << (sizeof(type)*8-2))
#define __MAX_SIGNED(type) (__HALF_MAX_SIGNED(type) - 1 + __HALF_MAX_SIGNED(type))
#define __MIN_SIGNED(type) (-1 - __MAX_SIGNED(type))
/* we use <1 and not <0 to avoid a gcc warning */
#define __MIN(type) ((type)-1 < 1?__MIN_SIGNED(type):(type)0)
#define __MAX(type) ((type)~__MIN(type))
#define assign(dest,src) ({ typeof(src) __x=(src); typeof(dest) __y=__x; (__x==__y && ((__x<1) == (__y<1))?(void)((dest)=__y),0:1); })
/* gcc 5 now has nice builtins we can use instead */
#if defined(__GNUC__) && (__GNUC__ >= 5)
#define add_of(c,a,b) __builtin_add_overflow(a,b,&c)
#define sub_of(c,a,b) __builtin_sub_overflow(a,b,&c)
/* if a+b is defined and does not have an integer overflow, do c=a+b and
* return 0. Otherwise, return 1. */
#define add_of(c,a,b) ({ typeof(a) __a=a; typeof(b) __b=b; (__b)<1?((__MIN(typeof(a+b))-(__b)<=(__a))?assign(c,__a+__b):1) : ((__MAX(typeof(c))-(__b)>=(__a))?assign(c,__a+__b):1); })
#define sub_of(c,a,b) ({ typeof(a) __a=a; typeof(b) __b=b; (__b)<1?((__MAX(typeof(a+b))+__b>=__a)?assign(c,__a-__b):1) : ((__MIN(typeof(c))+__b<=__a)?assign(c,__a-__b):1); })
#undef __static
#ifdef __cplusplus