How are shared libraries referenced by various programs? - c

I understand that shared libraries are loaded into memory and used by various programs.
How can a program know where in memory the library is?

When a shared library is used, there are two parts to the linkage process. At compile time, the linker program, ld in Linux, links against the shared library in order to learn which symbols are defined by it. However, none of the code or data initializers from the shared library are actually included in the ultimate a.out file. Instead, ld just records which dynamic libraries were linked against and the information is placed into an auxiliary section of the a.out file.
The second phase takes placed at execution time, before main gets invoked. The kernel loads a small helper program, ld.so, into the address space and this gets executed. Therefore, the start address of the program is not main or even _start (if you have heard of it). Rather, it is actually the start address of the dynamic library loader.
In Linux, the kernel maps the ld.so loader code into a convenient place in the precess address space and sets up the stack so that the list of required shared libraries (and other necessary info) is present. The dynamic loader finds each of the required libraries by looking at a sequence of directories which are often point in the LD_LIBRARY_PATH environment variable. There is also a pre-defined list which is hard-coded into ld.so (and additional search places can be hard-coded into the a.out during link time). For each of the libraries, the dynamic loader reads its header and then uses mmap to create memory regions for the library.
Now for the fun part.
Since the actual libraries used at run-time to satisfy the requirements are not known at link-time, we need to figure out a way to access functions defined in the shared library and global variables that are exported by the shared library (this practice is deprecated since exporting global variables is not thread-safe, but it is still something we try to handle).
Global variables are assigned a statics address at link time and are then accessed by absolute memory address.
For functions exported by the library, the user of the library is going to emit a series of call assembly instructions, which reference an absolute memory address. But, the exact absolute memory address of the referenced function is not known at link time. How do we deal with this?
Well, the linker creates what is known as a Procedure Linkage Table, which is a series of jmp (assembly jump) instructions. The target of the jump is filled in at run time.
Now, when dealing with the dynamic portions of the code (i.e. the .o files that have been compiled with -fpic), there are no absolute memory references whatsoever. In order to access global variables which are also visible to the static portion of the code, another table called the Global Offset Table is used. This table is an array of pointers. At link time, since the absolute memory addresses of the global variables are known, the linker populates this table. Then, at run time, dynamic code is able to access the global variables by first finding the Global Offset Table, then loading the address of the correct variable from the appropriate slot in the table, and finally dereferencing the pointer.

Related

Does linker add the library function into the source code?

Linker basic function is to link the object code with other object code(it can be standard library code).
#include<stdio.h>
int main()
{
printf("hello");
}
I want to know will linker replace the printf() function with its definition (like an inline function in c++). Or it will paste the printf()
function outside the main() function and pass "hello" as argument to that function.
For printf("hello");, the compiler generates an instruction to call a subroutine. It leaves the address of the subroutine not completely filled in. The object module the compiler generates has some notes about what routine’s address should be filled in there.
The linker may work in different ways. For static linking, the linker will find the implementation of printf in a library and copy the object module for it from the library into the executable file it is building. Depending on certain characteristics of the link, the linker might then complete the call instruction with the final address of the printf routine or it might leave notes in the executable file about the relationship between the call instruction and the printf routine. Later, when the program is being loaded into memory, the program loader will complete the address in the instruction.
For dynamic linking, the linker will find the implementation of printf in a library (or in a file with sufficient information about the library). It will not copy the printf function’s object module into the executable file, but it will include notes about the relationship between the call instruction and the printf routine and its library in the executable file. Later, the program loader will copy the printf function’s object module into the memory of the process. (This might be done by mapping part of the process’ virtual address space to physical memory that already contains the object module from the library and that is shared by other processes on the system. This sharing reduces the load on the system and makes dynamic loading more favorable in this regard.) And the loader will complete the address in the call instruction.
Some dynamic loading is not done as soon as the program is loaded. When a process is started, the loader might load just the program entry point and some essential parts. Some call instructions might be left incomplete. They will have been filled in with the addresses of special subroutines of the program loader (or dynamic library loader). When one of these subroutines is called, it will then load the desired routine, change the address in the call instruction (or otherwise arrange for future calls to call the desired routine), and then jump to the desired routine. This is beneficial because routines that are not used by your program in a particular run do not have to be loaded into memory at all. For example, if your program has a lot of code and data to log errors and inform the user when certain errors occur, that code and data does not have to be loaded into memory of those errors do not occur in a particular session.

Modify Relocations Conditionally at Runtime

TL;DR
I want to overwrite .got, .got.plt,... to point to the correct addresses, because the linker makes wrong decisions.
I need to use two different dynamic allocation functions (i.e., malloc(),...) in the code. The appropriate one will be selected based on some condition during program execution. Therefore, I provided two glibc instances and used the LD_PRELOAD trick. The LD_PRELOAD value is something like the following:
LD_PRELOAD=multiplexer_library.so:glibc1.so:glibc2.so
where, multiplexer_library.so chooses the correct library. glibc1.so is accessed using dlsym(RTLD_NEXT, "malloc") and glibc2.so is accessed using dlopen() followed by dlsym(). The same goes for calloc(),...
The problem is that the second malloc will interfere with the first one. This happens because all dynamic relocations of the latter will be mapped to the former by the linker. For example, when glibc2.so calls the global function pointer morecore it will be mapped to the morecore and its target which is the __default_morecore() function in glibc1.so. The relocation entry for this global variable in glibc2.so is as follows:
0000003addc0 085600000006 R_X86_64_GLOB_DAT 00000000003af4d8 __morecore##GLIBC_2.2.5 + 0
I traced the execution in Pin. 125 out of over 1370 relocation entries, were accessed in my code during dynamic allocations. For example, an imporant entry is the global variable __curbrk which determines the brk boundary for dynamic allocations (NOTE THAT I provided an isolated brk region for each library at the system call level). This will obviously corrupt the allocations, because both allocators use the same __curbrk. The relocation entry for __curbrk in glibc1.so is shown below:
0000003adeb8 044400000006 R_X86_64_GLOB_DAT 00000000003b10b8 __curbrk##GLIBC_2.2.5 + 0
I tried to rename these conflicting names but 125 is a huge number and the code is hard to grasp. Because it is full of nested macros which makes the manual rename solution, practically, infeasible.
IIUC, for each relocation entry, there exists a memory address (e.g., somewhere in .got,...) where the linker will put the target relocated address and this address is exclusive to each shared library. I will call that address the TARGET HOLDER. For example, in the __curbrk case, the linker placed the runtime address for the __curbrk variable of glibc1.so in the target holder of __curbrk in both glibc1.so and glibc2.so. If this is right, at runtime, I will have to update the value in the target holder of __curbrk in glibc2.so to hold the runtime address of __curbrk variable of glibc2.so. And to completely resolve the problem, this should be done for all of the 125 relocation entries accessed by malloc(),.... Is it possible?
Any help is appreciated!
Therefore, I provided two glibc instances and used the LD_PRELOAD trick.
These answers explain why using LD_PRELOAD=glibc.so can not work (at least not reliably).
I decided to use two separate glibc libraries and use each of them for a dedicated region. The region is isolated in each library.
This can't possibly work, because designers of GLIBC don't support this approach. You'll need to do something else.

Why aren't LIBs and DLLs interchangeable?

LIB files are static libraries that must be included at compile time, whereas DLL files can be "dynamically" accessed by a program during runtime. (DLLs must be linked, however, either implicitly before runtime with an import library (LIB) or explicitly via LoadLibrary).
My question is: why differentiate between these file types at all? Why can't LIB files be treated as DLLs and vice versa? It seems like an unnecessary distinction.
Some related questions:
DLL and LIB files - what and why?
Why are LIB files beasts of such a duplicitous nature?
DLL and LIB files
What exactly are DLL files, and how do they work?
You must differentiate between shareable objects and static libraries simply because they are really different objects.
A shareable object file, as a DLL or a SO, contains structures used by the loader to allow the dynamic link to and from other executable images (i.e. export tables).
A DLL is at all effects an executable image that, as an executable, can be loaded in memory and relocated (if not position independent code), but not only import symbols, as the executable do, but also expose exported symbols.
Exported symbols can be used by the loader to interlink different executable modules in memory.
A static library, on the other hand, is simply a collection of object modules to be linked in a single executable, or even a DLL.
An object module contains instruction bytecode and placeholder for external symbols which are referenced through the relocation table.
The linker collect object modules one by one each time they are referenced, i.e. a function call, and add the object to the linking code stream, than examine the relocation table of the object module, and replace the occurrence of each external symbol, with a displacement of the symbol inside the linked code. Eventually adding more object modules as new references are discovered. This is a recursive process that will end when no more undefined references remain.
At the end of linking process you have an image of your executable code in the memory. This image will be read and placed in memory by the loader, an OS component, that will fix some minor references and fills the import table with the addresses of symbols imported from DLL's.
Moreover if it is true that you can extract each single object module you need from an archive (library file), you can't extract single parts from a DLL because it is a merge of all modules without any reference for the start and the end of each.
It should be clear now that while an object module, the .obj file, or a collection of them, .lib file, is quite different from a DLL. Raw code the first, a fully linked and 'ready to run' piece of code the second.
The very reason for existence of shareable objects and static libraries is related to efficiency and resource rationalization.
When you statically link library modules you replicate same code for each executable you create using that static library, implying larger executable files that will take longer time to load wasting kernel execution time and memory space.
When you use shareable objects you load the code only the first time, then for all subsequent executables you need only to map the space where the DLL code lays in the new process memory space and create a new data segment (this must be unique for each process to avoid conflicts), effectively optimizing memory and system usage (for lighter loader workload).
So how we have to choose between the two?
Static linking is convenient when your code is used by a limited number of programs, in which case the effort to load a separate DLL module isn't worth.
Static linking also allows to easily reference to process defined global variables or other process local data. This is not possible, or not so easy, with a DLL because being a complete executable can't have undefined references so you must define any global inside the DLL, and this reference will be common for all processes accessing the DLL code.
Dynamic linking is convenient when the code is used by many programs making more efficient the loader work, and reducing the memory usage. Example of this are the system libraries, which are used by almost all programs, or compiler runtime.

When is dynamic linking between a program and a shared library performed?

In C, when is dynamic linking between a program and a shared library performed:
Once loading of the program into the memory, but before executing the main() of the program, or
After executing the main() of the program, when the first call to a routine from the library is executed? Will dynamic linking happen again when a second or third or... call to a routine from the library is executed?
I was thinking the first, until I read the following quote, and now I am not sure.
Not sure if OS matters, I am using Linux.
From Operating System Concepts:
With dynamic linking, a stub is included in the image for each
library- routine reference. The stub is a small piece of code that
indicates how to locate the appropriate memory-resident library
routine or how to load the library if the routine is not already
present.
When the stub is executed, it checks to see whether the needed routine is already in memory. If it is not, the program loads the
routine into memory. Either way, the stub replaces itself with the
address of the routine and executes the routine. Thus, the next time
that particular code segment is reached, the library routine is
executed directly, incurring no cost for dynamic linking. Under this
scheme, all processes that use a language library execute only one
copy of the library code.
I was thinking the first, until I read the following quote, and now I am not sure.
It's complicated (and depends on exactly what you call "dynamic linking").
The Linux kernel loads a.out into memory. It then examines PT_INTERP segment (if any).
If that segment is not present, the binary is statically linked and the kernel transfers control to the Elf{32,64}Ehdr.e_entry (usually the _start routine).
If the PT_INTERP segment is present, the kernel loads it into memory, and transfers control to it's .e_entry. It is here that the dynamic linking begins.
The dynamic loader relocates itself, then looks in a.outs PT_DYNAMIC segment for instructions on what else is necessary.
For example, it will usually find one or more DT_NEEDED entries -- shared libraries that a.out was directly linked against. The loader loads any such libraries, initializes them, and resolves any data references between them.
IF a.outs PT_DYNAMIC has a DT_FLAGS entry, and IF that entry contains DF_BIND_NOW flag, then function references from a.out will also be resolved. Otherwise (and assuming that LD_BIND_NOW is not set in the environment), lazy PLT resolution will be performed (resolving functions as part of first call to any given function). Details here.
When the stub is executed, it checks to see whether the needed routine is already in memory. If it is not, the program loads the routine into memory.
I don't know which book you are quoting from, but no current UNIX OS works that way.
The OS (and compiler, etc.) certainly matters: the language itself has nothing to say about dynamic libraries (and very little about linking in general). Even if we know that dynamic linking is occurring, a strictly-conforming program cannot observe any effect from timing among its translation units (since non-local initialization cannot have side effects).
That said, the common toolchains on Linux do support automatic initialization upon loading a dynamic library (for implementing C++, among other things). Executables and the dynamic libraries on which they depend (usually specified with -l) are loaded and initialized recursively to allow initialization in each module to (successfully) use functions from its dependencies. (There is an unfortunate choice of order in some cases.) Of course, dlopen(3) can be used to load and initialize more libraries later.

How to get memory locations of library functions?

I am compiling a C program with the SPARC RTEMS C compiler.
Using the Xlinker -M option, I am able to get a large memory map with a lot of things I don't recognize.
I have also tried using the RCC nm utility, which returns a slightly more readable symbol table. I assume that the location given by this utility for, say, printf, is the location where printf is in memory and that every program that calls printf will reach that location during execution. Is this a valid assumption?
Is there any way to get a list of locations for all the library/system functions? Also, when the linking is done, does it link just the functions that the executable calls, or is it all functions in the library? It seems to me to be the latter, given the number of things I found in the symbol table and memory map. Can I make it link only the required functions?
Thanks for your help.
Most often, when using a dynamic library, the nm utility will not be able to give you the exact answer. Binaries these days use what is known as relocatable addresses. These addresses change when they are mapped to the process' address space.
Using the Xlinker -M option, I am able to get a large memory map with a lot of things I don't recognize.
The linker map will usually have all symbols -- yours, the standard libraries, runtime hooks etc.
Is there any way to get a list of locations for all the library/system functions?
The headers are a good place to look.
Also, when the linking is done, does it link just the functions that the executable calls, or is it all functions in the library?
Linking does not necessarily mean that all symbols will be resolved (i.e. given an address). It depends on the type of binary you are creating.
Some compilers like gcc however, does allow you whether to create a non-relocatable binary or not. (For gcc you may check out exp files, dlltool etc.) Check with the appropriate documentation.
With dynamic linking,
1. your executable has a special place for all external calls (PLT table).
2. your executable has a list of libraries it depends on
These two things are independent. It is impossible to say which external function lives in which library.
When a program does an external function call, what actually happens it calls an entry in the PLT table, which does a jump into the dynamic loader. The dynamic loader looks which function was called (via PLT), looks its name (via symbol table in the executable) and looks up that name in ALL libraries that are mapped (all that given executable is dependant on). Once the name is found, the address of the corresponding function is written back to the PLT, so next time the call is made directly bypassing the dynamic linker.
To answer your question, you should do the same job as dynamic linker does: get a list of dependent libs, and lookup all names in them. This could be done using 'nm' or 'readelf' utility.
As for static linkage, I think all symbols in given object file within libXXX.a get linked in. For example, static library libXXX.a consists of object files a.o, b.o and c.o. If you need a function foo(), and it resides in a.o, then a.o will be linked to your app - together with function foo() and all other data defined in it. This is the reason why for example C library functions are split per file.
If you want to dynamically link you use dlopen/dlsym to resolve UNIX .so shared library entry points.
http://www.opengroup.org/onlinepubs/009695399/functions/dlsym.html
Assuming you know the names of the functions you want to call, and which .so they are in. It is fairly simple.
void *handle;
int *iptr, (*fptr)(int);
/* open the needed object */
handle = dlopen("/usr/home/me/libfoo.so", RTLD_LOCAL | RTLD_LAZY);
/* find the address of function and data objects */
*(void **)(&fptr) = dlsym(handle, "my_function");
iptr = (int *)dlsym(handle, "my_object");
/* invoke function, passing value of integer as a parameter */
(*fptr)(*iptr);
If you want to get a list of all dynamic symbols, objdump -T file.so is your best bet. (objdump -t file.a if your looking for statically bound functions). Objdump is cross platform, part of binutils, so in a pinch, you can copy your binary files to another system and interrorgate them with objdump on a different platform.
If you want dynamic linking to be optimal, you should take a look at your ld.so.conf, which specifie's the search order for the ld.so.cache (so.cache right ;).

Resources