I don't know where you can check it on a specific Linux system. I just know that 3GB is typical and is the value on every 32 bit Linux I use. I'm pretty sure it can and sometimes is adjusted as a build time option.
32 bit mode has a 4GB limit per process (including kernel) in the hardware architecture. An OS design must divide that between kernel and process. Windows by default divides it 2GB/2GB but can be set to divide it 1GB/3GB. Linux typically divides it 1GB/3GB, but can be set to 2GB/2GB or to (inefficiently elsewhere)/4GB.
The current hardware architecture for 64 bit has a limit (per process including kernel virtual) of 262144GB, and is designed to make an even division between Kernel and process most practical. So an OS could easily have a limit of 131072GB virtual for each process.
However, there are some performance advantages to designing the OS with a lower limit on per process virtual memory. I think I read somewhere that XP64 set that to 320768GB per process virtual. I think there would be a significant simplification and performance advantage at a limit of 1024GB process virtual (but that would actually constrain some plausible uses, so it wouldn't be a good idea). I don't recall what if anything I've ever read about what Linux has done for that.
People asking these impractical questions are usually referred to
http://linux-mm.org/LinuxMM
as an introduction to where you might look to find it yourself. Just now, I followed a few links there to see if your answer was given. I didn't find it.
BTW, the x86_64 architecture (unlike the 32bit architecture) is designed to allow a future expansion far beyond 262144GB involving a change to both CPU chip design and OS kernel source code, but NOT requiring any user code to be changed nor even recompiled.