C test if variable is in read-only section - c

I'd like to write a low-level logging function that would look like:
DO_DBG("some string", val1, val2)
What I want it to do, is to store the pointer to the string rather than a copy of the string, for performance reasons. This assumes that the string is a read-only literal. To prevent people having to debug the debugger, it would be nice if the compiler could complain if the first parameter of DO_DBG was in a writable section of code vs text, etc. I'm wondering if a mechanism to do that exists. (I'm using gcc 4.9.1, ld 2.24).

You could use the automatic literal string concatenation to your advantage:
#define DO_DBG(s, a, b) _DO_DBG("" s, a, b)
And implement your real macro as _DO_DBG.

You can set up a SIGSEGV handler, and then try to do:
s[0] = s[0];
If the handler is triggered, it means that the variable is in a read-only segment, so you can save the pointer. If not, you can complain (or just make a copy of the string).
You'll need to declare the string volatile so the compiler doesn't optimize the assignment away.

Alternatively you could use the stringify macro construct:
#define DO_DBG(s, a, b) DO_DBG_INTERNAL(# s, a, b)
and call your function DO_DGB_INTERNAL
so people would do:
DO_DBG(some string, val1, val2)
PS, quite possibly a variadic function and macro would be a good idea here.

Your function gets a pointer to the start of the string. If your operating system allows it, and the compiler sets it up that way, a constant string (like in your example) could be allocated in a read-only memory area. If your compiler is smart, it will store one copy of a string constant that appears several times. It might even figure out that some strings are never used, and just not store them at all.
Other than the above, unless your program relies on undefined behaviour (getting away with writing on a constant string), you should not be able to figure out if it is in read-only memory or not. You could compare addresses of strings to see if they are only one copy, but that is it.
In any reasonable C implementation (i.e., one that doesn't go out of its's way just to screw you over), you will see no (or at most an extremely small) performance difference if the string is read-only, read-write, one copy or several.
If you write you function with a char * parameter, and store said pointer away (not a copy to the string pointed to), you should always get the same string back (unless your environment is really strange).

Here's a SIGSEGV-handler-based test to check if pointed-at memory is mapped read-only:
#define _GNU_SOURCE
#include <stdio.h>
#include <signal.h>
#include <string.h>
#include <setjmp.h>
static sigjmp_buf label;
static void segv_hndlr(int Sig) { siglongjmp(label,1); } //safe in this ctx
_Bool is_ro(char volatile *X)
{
_Bool r=0;
struct sigaction old;
X[0]; //fault now if the target isn't writable
if(sigsetjmp(label,1))
{ r=1; goto out; }
sigaction(SIGSEGV,&(struct sigaction){.sa_handler=segv_hndlr}, &old);
X[0]=X[0];
out: sigaction(SIGSEGV,&old, NULL);
return r;
}
//example:
int main()
{
#define PR_BOOL(X) printf(#X"=%d\n", X)
char bar[]="bar";
char *baz = strdup("baz");
static int const static_ro_int = 42;
int const auto_ro_int = 42;
PR_BOOL(is_ro("foo")); //1
PR_BOOL(is_ro(bar)); //0
PR_BOOL(is_ro(baz)); //0
PR_BOOL(is_ro((void*)&static_ro_int)); //1
PR_BOOL(is_ro((void*)&auto_ro_int)); //0
}

Related

Can I store pointer to __PRETTY_FUNCTION__ for later use or do I need to copy the string immediately?

What is the lifetime of the string __PRETTY_FUNCTION__ produces?
I know __func__ generates equivalent of static const char __func__[] = "function-name";, Thanks to that static, if I want to create a record of something that happened in the function somewhere for referencing later, I can just store the value of __func__ in a const char* field in the record and retrieve it whenever I want, long after I left the scope of the function.
Does __PRETTY_FUNCTION__ behave the same way, or does the string it generates behave as a local, temporary value? Say, will this work reliably, or can it fail, fname pointing to a location where the function name used to be, and I need to do make a buffer and do strncpy() or similar instead?
#include <stdio.h>
const char* fname = NULL;
void foo()
{
fname = __PRETTY_FUNCTION__;
}
int main()
{
foo();
printf(fname);
return 0;
}
All three constants, __func__, __FUNCTION__ and __PRETTY_FUNCTION__, behave the same way. The GCC Manual states:
GCC provides three magic constants that hold the name of the current function as a string. In C++11 and later modes, all three are treated as constant expressions and can be used in constexpr contexts.

Is it legal to use memcpy with a destination structure with constant members?

For example, is the following function legal:
struct two_int {
const int a, b;
}
void copy_two(const two_int *src, two_int *dest) {
memcpy(dest, src, sizeof(two_int));
}
It seems like at least some types of modifications of constant-defined values is not allowed, but it is not clear to me if this qualifies.
If the answer is "it is not allowed, in general", I'm also wondering about the special case where dest is newly allocated memory with malloc (and hence hasn't yet been assigned any value), such as:
two_int s = {.a = 1, .b = 2};
two_int *d = malloc(sizeof(two_int));
copy_two(&s, d);
Update: It seems like the latter question seems to answered in the affirmative (it's OK) for the case of a newly malloc'd structure, but the original, more general question still stands, I think.
Use of memcpy for such purposes would have defined behavior only if the actual destination object does not have static or automatic duration.
Given the code:
struct foo { double const x; };
void outsideFunction(struct foo *p);
double test(void)
{
struct foo s1 = {0.12345678};
outsideFunction(&s1);
return 0.12345678;
}
a compiler would be entitled to optimize the function to:
double test(void)
{
static struct foo const s1 = {0.12345678};
outsideFunction(&s1);
return s1.x;
}
On many processors, the only way to load a double with an arbitrary constant is to load its value from object holding that constant, and in this case the compiler would conveniently know an object (s1.x) which must hold the constant 0.12345678. The net effect would be that code which used memcpy to write to s1.x could corrupt other uses of the numeric constant 0.12345678. As the saying goes, "variables won't; constants aren't". Nasty stuff.
The problem would not exist for objects of allocated duration because memcpy requires a compiler to "forget" everything about what had previously been stored in the destination storage, including any "effective type". The declared types of static and automatic objects exist independent of anything being stored into them, and cannot be erased by "memcpy" nor any other means, but allocated-duration objects only have an "effective type" which would get erased.
First note, that almost anything is allowed, in the wider sense of the word: It might simply have undefined behavior. You can take an arbitrary number, reinterpret_cast<> it as an address, and write to that address - perfectly "legal" C++. As for what it does - usually, no guarantees. Here be dragons.
However, as indicated in an answer to this question:
behavior of const_cast in C++
if the structure you're pointing to was not originally defined as const, and simply got const'ified on the way to your code, then const_cast'ing the reference or pointer to it and setting the pointed-to values is required to succeed as you might expect.

C Aligning string literals for a specific use case

I'm trying to align string literals in a specific way, as how I'm using it in my code is fairly specific. I don't want to have to assign it to a variable, for instance many of my functions are using it as a direct argument. And I want it to work both in local scope or global scope.
Usage example:
char *str = ALIGNED_STRING("blah"); //what I want
foo(ALIGNED_STRING("blah")); //what I want
_Alignas(16) char str[] = "blah"; //not what I want (but would correctly align the string)
The ideal solution would be (_Alignas(16) char[]){ "blah" } or a worser case using the GCC/Clang compiler extensions for alignment (__attribute__((alignment(16))) char[]){ "blah" }, but neither works (they're ignored and the default alignment for the type is used).
So my next thought was to align it myself, and then my functions that use the string could then fix it up correctly. e.g. #define ALIGNED_STRING(str) (char*)(((uintptr_t)(char[]){ "xxxxxxxxxxxxxxx" str } + 16 - 1) & ~(16 - 1)) (where the string containing 'x' would represent data needed to understand where the real string can be found, that's easy but just for the example assume the 'x' is fine). Now that works fine in local scope, but fails in the global scope. Since the compiler complains about it not being a compile-time constant (error: initializer element is not a compile-time constant); I would've thought it would work but it seems only addition and subtraction are valid operations on the pointer at compile-time.
So I'm wondering if there's anyway to achieve what I want to do? At the moment I'm just using the latter example (padding and manually aligning) and avoiding to use it in the global scope (but I would really want to). And the best solution would avoid needing to make runtime adjustments (like using the alignment qualifier would), but that doesn't seem possible unless I apply it to a variable (but as mentioned that's not what I want to do).
Was able to get close to OP's need with a compound literal. (C99)
#include <stdio.h>
#include <stddef.h>
void bar(const char *s) {
printf("%p %s\n", (void*)s, s);
}
// v-- compound literal --------------------------v
#define ALIGNED_STRING(S) (struct { _Alignas(16) char s[sizeof S]; }){ S }.s
int main() {
char s[] = "12";
bar(s);
char t[] = "34";
bar(t);
bar(ALIGNED_STRING("asdfas"));
char *u = ALIGNED_STRING("agsdas");
bar(u);
}
Output
0x28cc2d 12
0x28cc2a 34
0x28cc30 asdfas // 16 Aligned
0x28cc20 agsdas // 16 Aligned

Know if const qualifier is used

Is there any way in C to find if a variable has the const qualifier? Or if it's stored in the .rodata section?
For example, if I have this function:
void foo(char* myString) {...}
different actions should be taken in these two different function calls:
char str[] = "abc";
foo(str);
foo("def");
In the first case I can modify the string, in the second one no.
Not in standard C, i.e. not portably.
myString is just a char* in foo, all other information is lost. Whatever you feed into the function is automatically converted to char*.
And C does not know about ".rodata".
Depending on your platform you could check the address in myString (if you know your address ranges).
You can't differ them using the language alone. In other words, this is not possible without recurring to features specific to the compiler you're using, which is likely not to be portable. A few important remarks though:
In the first case you COULD modify the string, but you MUST NOT. If you want a mutable string, use initialization instead of assignment.
char *str1 = "abc"; // NOT OK, should be const char *
const char *str2 = "abc"; // OK, but not mutable
char str3[] = "abc"; // OK, using initialization, you can change its contents
#include<stdio.h>
void foo(char *mystr)
{
int a;
/*code goes here*/
#ifdef CHECK
int local_var;
printf(" strings address %p\n",mystr);
printf("local variables address %p \n",&local_var);
puts("");
puts("");
#endif
return;
}
int main()
{
char a[]="hello";
char *b="hello";
foo(a);
foo(b);
foo("hello");
}
On compiling with gcc -DCHECK prog_name.c and executing on my linux machine the following output comes...
strings address 0xbfdcacf6
local variables address 0xbfdcacc8
strings address 0x8048583
local variables address 0xbfdcacc8
strings address 0x8048583
local variables address 0xbfdcacc8
for first case when string is defined and initialized in the "proper c way for mutable strings" the difference between the addresses is 0x2E.(5 bytes).
in the second case when string is defined as char *p="hello" the differences in addresses is
0xB7D82745.Thats bigger than the size of my stack.so i am pretty sure the string is not on the stack.Hence the only place where you can find it is .rodata section.
The third one is similar case
PS:As mentioned above this isn't portable but the original question hardly leaves any scope for portability by mentioning .rodata :)
GCC provides the __builtin_constant_p builtin function, which enables you to determine whether an expression is constant or not at compile-time:
Built-in Function: int __builtin_constant_p (exp)
You can use the built-in function __builtin_constant_p to determine if a value is known to be constant at compile-time and hence that GCC can perform constant-folding on expressions involving that value. The argument of the function is the value to test. The function returns the integer 1 if the argument is known to be a compile-time constant and 0 if it is not known to be a compile-time constant. A return of 0 does not indicate that the value is not a constant, but merely that GCC cannot prove it is a constant with the specified value of the `-O' option.
So I guess you should rewrite your foo function as a macro in such a case:
#define foo(x) \
(__builtin_constant_p(x) ? foo_on_const(x) : foo_on_var(x))
foo("abc") would expand to foo_on_const("abc") and foo(str) would expand to foo_on_var(str).

Pointer to literal value

Suppose I have a constant defined in a header file
#define THIS_CONST 'A'
I want to write this constant to a stream. I do something like:
char c = THIS_CONST;
write(fd, &c, sizeof(c))
However, what I would like to do for brevity and clarity is:
write(fd, &THIS_CONST, sizeof(char)); // error
// lvalue required as unary ‘&’ operand
Does anyone know of any macro/other trick for obtaining a pointer to a literal? I would like something which can be used like this:
write(fd, PTR_TO(THIS_CONST), sizeof(char))
Note: I realise I could declare my constants as static const variables, but then I can't use them in switch/case statements. i.e.
static const char THIS_CONST = 'A'
...
switch(c) {
case THIS_CONST: // error - case label does not reduce to an integer constant
...
}
Unless there is a way to use a const variable in a case label?
There is no way to do this directly in C89. You would have to use a set of macros to create such an expression.
In C99, it is allowed to declare struct-or-union literals, and initializers to scalars can be written using a similar syntax. Therefore, there is one way to achieve the desired effect:
#include <stdio.h>
void f(const int *i) {
printf("%i\n", *i);
}
int main(void) {
f(&(int){1});
return 0;
}
These answers are all outdated, and apart from a comment nobody refers to recent language updates.
On a C99-C11-C17 compiler using a compound literal, http://en.cppreference.com/w/c/language/compound_literal, is possible to create
a pointer to a nameless constant, as in:
int *p = &((int){10});
The only way you can obtain a pointer is by putting the literal into a variable (your first code example). You can then use the variable with write() and the literal in switch.
C simply does not allow the address of character literals like 'A'. For what it's worth, the type of character literals in C is int (char in C++ but this question is tagged C). 'A' would have an implementation defined value (such as 65 on ASCII systems). Taking the address of a value doesn't make any sense and is not possible.
Now, of course you may take the address of other kinds of literals such as string literals, for example the following is okay:
write(fd, "potato", sizeof "potato");
This is because the string literal "potato" is an array, and its value is a pointer to the 'p' at the start.
To elaborate/clarify, you may only take the address of objects. ie, the & (address-of) operator requires an object, not a value.
And to answer the other question that I missed, C doesn't allow non-constant case labels, and this includes variables declared const.
Since calling write() to write a single character to a file descriptor is almost certainly a performance killer, you probably want to just do fputc( THIS_CONST, stream ).
#define THIS_CONST 'a'
Is just a macro. The compiler basically just inserts 'a' everywhere you use THIS_CONST.
You could try:
const char THIS_CONST = 'a';
But I suspect that will not work wither (don't have a c-compiler handy to try it out on, and it has been quite a few years since I've written c code).
just use a string constant, which is a pointer to a character, and then only write 1 byte:
#define MY_CONST_STRING "A"
write(fd, MY_CONST_STRING, 1);
Note that the '\0' byte at the end of the string is not written.
You can do this for all sorts of constant values, just use the appropriate hex code string, e.g.
#define MY_CONST_STRING "\x41"
will give also the character 'A'. For multiple-byte stuff, take care that you use the correct endianness.
Let's say you want to have a pointer to a INT_MAX, which is e.g. 0x7FFFFFFF on a 32 bit system. Then you can do the following:
#define PTR_TO_INT_MAX "\xFF\xFF\xFF\x7F"
You can see that this works by passing it as a dereferenced pointer to printf:
printf ("max int value = %d\n", *(int*)PTR_TO_INT_MAX);
which should print 2147483647.
For chars, you may use extra global static variables. Maybe something like:
#define THIS_CONST 'a'
static char tmp;
#define PTR_TO(X) ((tmp = X),&tmp)
write(fd,PTR_TO(THIS_CONST),sizeof(char));
There's no reason the compiler has to put the literal into any memory location, so your question doesn't make sense. For example a statement like
int a;
a = 10;
would probably just be directly translated into "put a value ten into a register". In the assembly language output of the compiler, the value ten itself never even exists as something in memory which could be pointed at, except as part of the actual program text.
You can't take a pointer to it.
If you really want a macro to get a pointer,
#include <stdio.h>
static char getapointer[1];
#define GETAPOINTER(x) &getapointer, getapointer[0] = x
int main ()
{
printf ("%d\n",GETAPOINTER('A'));
}
I can see what you're trying to do here, but you're trying to use two fundamentally different things here. The crux of the matter is that case statements need to use values which are present at compile time, but pointers to data in memory are available only at run time.
When you do this:
#define THIS_CONST 'A'
char c = THIS_CONST;
write(fd, &c, sizeof(c))
you are doing two things. You are making the macro THIS_CONST available to the rest of the code at compile time, and you are creating a new char at runtime which is initialised to this value. At the point at which the line write(fd, &c, sizeof(c)) is executed, the concept of THIS_CONST no longer exists, so you have correctly identified that you can create a pointer to c, but not a pointer to THIS_CONST.
Now, when you do this:
static const char THIS_CONST = 'A';
switch(c) {
case THIS_CONST: // error - case label does not reduce to an integer constant
...
}
you are writing code where the value of the case statement needs to be evaluated at compile time. However, in this case, you have specified THIS_CONST in a way where it is a variable, and therefore its value is available only at runtime despite you "knowing" that it is going to have a particular value. Certainly, other languages allow different things to happen with case statements, but those are the rules with C.
Here's what I'd suggest:
1) Don't call a variable THIS_CONST. There's no technical reason not to, but convention suggests that this is a compile-time macro, and you don't want to confuse your readers.
2) If you want the same values to be available at compile time and runtime, find a suitable way of mapping compile-time macros into run-time variables. This may well be as simple as:
#define CONST_STAR '*'
#define CONST_NEWLINE '\n'
static const char c_star CONST_STAR;
static const char c_newline CONST_NEWLINE;
Then you can do:
switch(c) {
case CONST_STAR:
...
write(fd, &c_star, sizeof(c_star))
...
}
(Note also that sizeof(char) is always one, by definition. You may know that already, but it's not as widely appreciated as perhaps it should be.)
Here is another way to solve this old but still relevant question
#define THIS_CONST 'A'
//I place this in a header file
static inline size_t write1(int fd, char byte)
{
return write(fd, &byte, 1);
}
//sample usage
int main(int argc, char * argv[])
{
char s[] = "Hello World!\r\n";
write(0, s, sizeof(s));
write1(0, THIS_CONST);
return 0;
}
Ok, I've come up with a bit of a hack, for chars only - but I'll put it here to see if it inspires any better solutions from anyone else
static const char *charptr(char c) {
static char val[UCHAR_MAX + 1];
val[(unsigned char)c] = c;
return &val[(unsigned char)c];
}
...
write(fd, charptr(THIS_CONST), sizeof(char));

Resources