The "Libc" library provides wrappers for most system
calls and the "glue" code for the dynamic linker and
userland process initialization. 

1. Thread-local storage

How thread-local variables are addressed and stored:
http://www.technovelty.org/linux/debugging-__thead-variables-from-coredumps.html 
This post operates with core dumps for a reason: it is very
hard to obtain the current values of segment descriptors (GDT)
of a running process (see 
http://stackoverflow.com/questions/10354063/how-to-use-a-logical-address-in-gdb
http://stackoverflow.com/questions/4006686/how-to-resolve-segmentoffset-adres-in-gdb). Note that this would be a great project.

2. Dynamic linking

A very quick overview of structures involved in dynamic linking:
http://thexploit.com/secdev/how-is-glibc-loaded-at-runtime/

A detailed analysis of dynamic linking:
"Cheating the ELF" by the grugq, 
http://althing.cs.dartmouth.edu/local/subversiveld.pdf

3. Program startup

An excellent step-by-step:
"Linux x86 Program Start Up
or - How the heck do we get to main()?" by Patrick Horgan
http://dbp-consulting.com/tutorials/debugging/linuxProgramStartup.html

A old summary from a different angle:
http://asm.sourceforge.net/articles/startup.html
 
4. Linker scripts and initial memory layout

This example is from an embedded system with almost 
no kernel -- hence a very simple linker map
(Listing 13):
http://www.bravegnu.org/gnu-eprog/c-startup.html

This is a chapter from http://www.bravegnu.org/gnu-eprog/index.html,
a good book on programming tiny embedded computers (ARM-based
"gumstix", literally the size of a pack of gum). Chapters 6 & 7
given very simple examples of linker operations: resolution,
relocation.