Suggest using devtmpfs for /dev

[ghost]

1. You can use devtmpfs for /dev instead of tmpfs. This allows mdev to be eliminated from busybox.
2. There is a delay of around 25 seconds if you boot the CPU stress image from QEMU.
3. Show Victor Chew this: http://blueponies666.github.io/ubcd-advice.html
4. You don't need to make /tmp a mountpoint. It is already handled by initramfs root filesystem.
5. I wouldn't suggest removing functions from eglibc. It can cause unresolved symbol errors that can be hard to track down.

[Explorer09]

1. Thanks. In fact I was planning to replace mdev with devtmpfs. I have analyzed the code sizes and the behaviors between the two, and concluded that devtmpfs is the winner. However, I wish to document clearly the analysis and the rationale for switching, and unfortunately I don't have much time these days to maintain this. (BTW, I'm sort of a perfectionist in this project and I wish to configure a new kernel with as minimal size as possible, in order to bring down the memory requirements for older machines.)
2. QEMU and other virtualizations are not in my support scope. Although I do test this image in VirtualBox before trying in my real laptop. If you can pin down the problem then I may try fix that, but don't expect anything.
3. Why aren't you bothered to put your UBCD advice on the UBCD forum? (http://www.ultimatebootcd.com/forums/index.php) I can't help forward that for you.
4. Removing the /tmp mountpoint might be a good idea one I have devtmpfs used. I'll take that, thanks.
5. Removing function from eglibc is my decision in order to keep the size down. I do track symbol usages when I remove things in order not to break anything. See https://github.com/Explorer09/cpustress-ubcd/blob/master/cpustress/build/compile-notes/eglibc/option-groups.config where I commented every change.