System calls with struct parameters (Linux) - linux-kernel

How is it that certain System calls take pointers to structs as arguments? If these structs are defined in the kernel, then how can user programs create instances of them?

There is no magic here. The struct types used in syscalls and meant to be user-createable are declared in header files, just as the syscalls themselves are. Take stat(2):
int stat(const char *path, struct stat *buf);
You get the declaration of struct stat (on Linux) by including sys/stat.h.
Some types are not meant to be directly declared by client code, however. In comments you mentioned semaphores, and sem_t is an example of such. The user header provides only an incomplete declaration, so you can't create an instance directly. This is intentional. In those cases there will be a call that creates an instance and returns a pointer to it, for example:
sem_t *sem_open(const char *name, int oflag);
You are expected to provide that same pointer as an argument to subsequent syscalls, even though you can't dereference it yourself (because its declaration is incomplete). The distinction between structs and struct pointers is extremely important here.

Every time , when you create a new structure inside the kernel , it can be exported to userspace by executing "make headers_install".
So if the user space binary is built in the same machine , it would be having identical copy of the header files(typically in /usr/include).Hence system calls can specify pointers to structures as parameters.

Related

How do compilers handle records and unions? [duplicate]

This question already has answers here:
Closed 12 years ago.
Possible Duplicate:
How does a compiled C++ class look like?
Hi all,
bash$cat struct.c
struct test
{
int i;
float f;
};
bash$gcc -c struct.c
The object file struct.o is of elf format. I am trying to understand what does this object file contain. The source code is just a definition of a struct. There is nothing executable here so there should be nothing in text, and there is no data really either.
So where does the definition of struct go really?
I tried using;
readelf -a struct.o
objdump -s struct.o
but don't quite understand this.
Thanks,
Jagrati
So where does the definition of struct
go really?
Struct definition usually goes to /dev/null. C does not have any introspection features, so struct definition is not needed at run time. During compilation, calls to struct fields are converted to numeric offsets, eg. x->f would be compiled to equivalent of *((void*)x + sizeof(int)). That's why you need to include headers every time you use struct.
There is nothing. It does not exist. You have created nothing and used nothing.
The definition of the struct is used at compile time. That definition would normally be placed in a non-compiled header file. It is when a struct is used that some code is generated. The definition affects what the compiler produces at that point.
This, among other reasons, is why compiling against one version of a library and then using another version at runtime can crash programs.
structs are not compiled, they are declared. Functions get compiled though.
I'm not an expert and I can't actually answer the question... But I thought of this.
Memory is memory: if you use 1 byte as integer or char, it is still one byte. The results depends only on the compiler.
So, why can't be the same for structs? I mean, the compiler probably will calculate the memory to allocate (as your computer probably will allocate WORDS of memory, not bytes, if your struct is 1 byte long, probably 3 bytes will be added allowing the allocation of 4 bytes word), and then struct will just be a "reference" for you when accessing data.
I think that there is no need to actually HAVE something underneath: it's sufficient for the compiler to know that, in compile time, if you refer to field "name" of your struct, it shall treat is as an array of chars of length X.
As I said, I'm not expert in such internals, but as I see it, there is no need for a struct to be converted in "real code"... It's just an annotation for the compiler, which can be destroyed after the compilation is done.

Can register name be passed into assembly template in GCC inline assembly [duplicate]

I have recently started learning how to use the inline assembly in C Code and came across an interesting feature where you can specify registers for local variables (https://gcc.gnu.org/onlinedocs/gcc/Local-Register-Variables.html#Local-Register-Variables).
The usage of this feature is as follows:
register int *foo asm ("r12");
Then I started to wonder whether it was possible to insert a char pointer such as
const char d[4] = "r12";
register int *foo asm (d);
but got the error: expected string literal before ‘d’ (as expected)
I can understand why this would be a bad practice, but is there any possible way to achieve a similar effect where I can use a char pointer to access the register? If not, is there any particular reason why this is not allowed besides the potential security issues?
Additionally, I read this StackOverflow question: String literals: pointer vs. char array
Thank you.
The syntax to initialize the variable would be register char *foo asm ("r12") = d; to point an asm-register variable at a string. You can't use a runtime-variable string as the register name; register choices have to get assembled into machine code at compile time.
If that's what you're trying to do, you're misunderstanding something fundamental about assembly language and/or how ahead-of-time compiled languages compile into machine code. GCC won't make self-modifying code (and even if it wanted to, doing that safely would require redoing register allocation done by the ahead-of-time optimizer), or code that re-JITs itself based on a string.
(The first time I looked at your question, I didn't understand what you were even trying to do, because I was only considering things that are possible. #FelixG's comment was the clue I needed to make sense of the question.)
(Also note that registers aren't indexable; even in asm you can't use a single instruction to read a register number selected by an integer in another register. You could branch on it, or store all the registers in memory and index that like variadic functions do for their incoming register args.)
And if you do want a compile-time constant string literal, just use it with the normal syntax. Use a CPP macro if you want the same string to initialize a char array.

What To Do With cdev_init When Converting file_operations To proc_ops?

According to this question Passing argument 4 of ‘proc_create’ from incompatible pointer type
You have to use proc_ops instead of file_operations structs on newer kernels.
How should we handle initializing the cdev with cdev_init when it uses the old file_operations structs?
I have looked through some examples on newer linux kernels (https://elixir.bootlin.com/linux/latest/source/drivers/char/pcmcia/scr24x_cs.c#L216) but they still use the old file_operations.
Thanks,
-Special K
Signature of cdev_init() function still uses argument pointed to struct file_operations object, so you needn't to change this call when adapt your module for newer kernels.
Signature for proc_create() function has been changed because it works with files under proc filesystem (procfs), and this filesystem has been re-designed in the newer kernels.
cdev_init function creates (initializes) a character device. While character devices are usually represented by a file (which could also be located under procfs), this file is of special type, and operations of this files are not the same as operations of character device itself. Because cdev_init is only responsible for character device (not a file), it is unaffected by changings in procfs design.

Can I define C functions that accept native Go types through CGo?

For the work I'm doing to integrate with an existing library, I ended up needing to write some additional C code to provide an interface that was usable through CGo.
In order to avoid redundant data copies, I would like to be able to pass some standard Go types (e.g. Go strings) to these C adapter functions.
I can see that there are GoString and GoInterface types defined in the header CGo generates for use by exported Go functions, but is there any way to use these types in my own function prototypes that CGo will recognise?
At the moment, I've ended up using void * in the C prototypes and passing unsafe.Pointer(&value) on the Go side. This is less clean than I'd like though (for one thing, it gives the C code the ability to write to the value).
Update:
Just to be clear, I do know the difference between Go's native string type and C char *. My point is that since I will be copying the string data passed into my C function anyway, it doesn't make sense to have the code on the Go side make its own copy.
I also understand that the string layout could change in a future version of Go, and its size may differ by platform. But CGo is already exposing type definitions that match the current platform to me via the documented _cgo_export.h header it generates for me, so it seems a bit odd to talk of it being unspecified:
typedef struct { char *p; int n; } GoString;
But there doesn't seem to be a way to use this definition in prototypes visible to CGo. I'm not overly worried about binary compatibility, since the code making use of this definition would be part of my Go package, so source level compatibility would be enough (and it wouldn't be that big a deal to update the package if that wasn't the case).
Not really. You cannot safely mix, for example Go strings (string) and C "strings" (*char) code without using the provided helpers for that, ie. GoString and CString. The reason is that to conform to the language specs a full copy of the string's content between the Go and C worlds must be made. Not only that, the garbage collector must know what to consider (Go strings) and what to ignore (C strings). And there are even more things to do about this, but let me keep it simple here.
Similar and/or other restrictions/problems apply to other Go "magical" types, like map or interface{} types. In the interface types case (but not only it), it's important to realize that the inner implementation of an interface{} (again not only this type), is not specified and is implementation specific.
That's not only about the possible differences between, say gc and gccgo. It also means that your code will break at any time the compiler developers decide to change some detail of the (unspecified and thus non guaranteed) implementation.
Additionally, even though Go doesn't (now) use a compacting garbage collector, it may change and without some pinning mechanism, any code accessing Go run time stuff directly will be again doomed.
Conclusion: Pass only simple entities as arguments to C functions. POD structs with simple fields are safe as well (pointer fields generally not). From the complex Go types, use the provided helpers for Go strings, they exists for a (very good) reason.
Passing a Go string to C is harder than it should be. There is no really good way to do it today. See https://golang.org/issue/6907.
The best approach I know of today is
// typedef struct { const char *p; ptrdiff_t n; } gostring;
// extern CFunc(gostring s);
import "C"
func GoFunc(s string) {
C.CFunc(*(*C.gostring)(unsafe.Pointer(&s)))
}
This of course assumes that Go representation of a string value will not change, which is not guaranteed.

What does this line of C/C++ preprocessor mean?

This is Line 519 of WinNT.h (BUILD Version: 0091)
#define DECLARE_HANDLE(name) struct name##__{int unused;}; typedef struct name##__ *name
Why do we need a pointer to an struct with a single int member with a weird name called unused?
And will we ever need to use a line of code like this one?
HINSTANCE hInstance = new HINSTANCE__;
Overall declaring different data types with the same structures, doesn't make sense to me. What's the idea behind this?
DECLARE_HANDLE(HRGN);
DECLARE_HANDLE(HRSRC);
DECLARE_HANDLE(HSPRITE);
DECLARE_HANDLE(HLSURF);
DECLARE_HANDLE(HSTR);
DECLARE_HANDLE(HTASK);
DECLARE_HANDLE(HWINSTA);
DECLARE_HANDLE(HKL);
The point is for the different handles to have different types so that, for example, a HINSTANCE isn't assignable to a HANDLE. If they were all defined as "void*", then there are classes of errors that the compiler could not detect.
And will we ever need to use a line of code like this one?
HINSTANCE hInstance = new HINSTANCE__;
You usually use a HINSTANCE value returned by a Windows system call; I have never seen code executing a line like that.
They don't actually point to anything to memory; they are just used to refer to objects (files, resource, semaphores, windows) when making calls to the Windows API. While they're nothing more than just indexes into kernel's object tables, the developers decided that they make it a pointer to an unused structure which would make them "opaque" and cause less confusion between other types. The DECLARE_HANDLE is a function macro that does just that - declaring opaque types for handles.

Resources