What is the difference between fork() and vfork()?

Question

What is the difference between fork() and vfork()? Does vfork() return like fork().

Answer 1

The intent of vfork was to eliminate the overhead of copying the whole process image if you only want to do an exec* in the child. Because exec* replaces the whole image of the child process, there is no point in copying the image of the parent.

if ((pid = vfork()) == 0) {   execl(..., NULL); /* after a successful execl the parent should be resumed */   _exit(127); /* terminate the child in case execl fails */ } 

For other kinds of uses, vfork is dangerous and unpredictable.

With most current kernels, however, including Linux, the primary benefit of vfork has disappeared because of the way fork is implemented. Rather than copying the whole image when fork is executed, copy-on-write techniques are used.

Answer 2

As already stated, the vfork man page is clear about the differences. This topic gives a good description of fork, vfork, clone and exec.

Below are some often overlooked differences between fork and vfork I experienced on some Linux 2.6.3x embedded systems I worked with.

Even with copy-on-write techniques, fork fails if you don't have enough memory to duplicate the memory used by the parent process. For example, if the parent process uses 2 GB of resident memory (ie, memory that is used and not just allocated), fork fails if you have less than 2 GB of free memory left. That's frustrating when you just want to exec a simple program and therefore will never need that huge parent address space!

vfork doesn't have this memory issue, as it doesn't duplicate the parent address space. The child process acts more like a thread in which you are able to call exec* or _exit without hurting your parent process.

Because memory page tables are not duplicated, vfork is much faster than fork and vfork's execution time is not affected by the amount of memory the parent process uses, as pointed out here: http://blog.famzah.net/2009/11/20/fork-gets-slower-as-parent-process-use-more-memory/

In situations where performance is critical and/or memory limited, vfork + exec* can therefore be a good alternative to fork + exec*. The problem is that it is less safe and the man page says vfork is likely to become deprecated in the future.

A safer and more portable solution may be to look at the posix_spawn function, which is higher level and offers more options. It safely uses vfork when possible, depending on the options you pass it. I have been able to use posix_spawn successfully and overcome that annoying "double memory checking issue" that fork + exec was giving me.

A really good page on this topic, with links to some posix_spawn examples.