FS, is it worth it?

[Dario]

Hi,

I planed to create my own very simple file system, but now when I see how much work is needed to be done in the rest of the operating system, I'm not convinced that it's really worth it....
So, what do you think?
Have you made your FS and how long did it took?
Is it painful?
If you think it's worth it, should I go and create VFS layer or should it be hard coded in the system?

Thank you!

[NickJohnson]

Probably not. At least when implementing your first filesystem driver, it's best to use something common like FAT or ext2. The obvious advantage is that you can read and write to the filesystem from your host system without modification. In addition, any filesystem you design will probably not make any advances unless you know how other filesystems work, so you should wait until you've implemented a few to decide to make a new one.

[rJah]

Depends on what you intend to use it for. I developed a simple boot FS(read only) I use for my ramdisk, so I only have to load a single boot module with GRUB. I also use it to test my VFS implementation.

[Dario]

Agree...I will implement FAT12/ext2 just to get things work and get to know FS. Later, I might try to design my own.

Thanks!

[NickJohnson]

@rjah: I use a simple tarball as a ramdisk "filesystem": it has lots of good tools for every platform and works as well as any readonly filesystem (with proper caching.)

[bewing]

Yes, I have some things I consider important new ideas for filesystems. I created a very efficient FS of my own and implemented it (Bang!FS). I think it's much better than any of the ext's (or FATs). It took me a year of solid work. During that year, I accomplished almost nothing on my OS.

Unless you have a year to waste, don't bother doing it yourself. The overview of FS creation is not hard, but there are a million details that take forever to get straight. There is not a lot to learn by just trying it as a learning experience.

[Dario]

bewing, my goal is to have all the functional parts of todays modern operating systems, but made simple. Now I don't tend to work on it for couple of months, but years, slowly creating/upgrading the system. I have already learned a lot, even dough I've began just 3 months ago.
I won't avoid FS, it will be developed, but right now it would take too much of my time.

Thank you for advice!

[Combuster]

bewing, my goal is to have all the functional parts of todays modern operating systems, but made simple. (...)
I won't avoid FS, it will be developed, but right now it would take too much of my time.

I thought the simplest solution was to reuse existing code rather than write something from scratch

[Thomas]

Hi Dario,

my goal is to have all the functional parts of todays modern operating systems, but made simple.

It might useful for you to checkout pintos, it is developed by Stanford university I guess.It has similar goals . See : http://www.scs.stanford.edu/10wi-cs140/ .You can read through papers and the additional exercises Hope it helps

--Thomas

[Dario]

bewing, my goal is to have all the functional parts of todays modern operating systems, but made simple. (...)
I won't avoid FS, it will be developed, but right now it would take too much of my time.

I thought the simplest solution was to reuse existing code rather than write something from scratch

Hello everybody out there using Linux -

I'm doing a (commercial) operating system (just a hobby, won't be small and
amateur like Windows) for x86, arm, alpha, sparc, mips, ia64, powerpc...and that stuff in your washing machine...

Thank you Thomas for the link!

[Thomas]

bewing, my goal is to have all the functional parts of todays modern operating systems, but made simple. (...)
I won't avoid FS, it will be developed, but right now it would take too much of my time.

I thought the simplest solution was to reuse existing code rather than write something from scratch

Hello everybody out there using Linux -

I'm doing a (commercial) operating system (just a hobby, won't be small and
amateur like Windows) for x86, arm, alpha, sparc, mips, ia64, powerpc...and that stuff in your washing machine...



Thank you Thomas for the link!

lol ! . I work on a commercial operating system, but then teams of people are involved dedicated to smaller subsystems.I takes lot of effort to create a commercial OS.Understand that I wont call ver0.01 of linux or v.1 of linux a commercial OS by any stretch of imagination.It became commercially viable when teams of people where involved and companies started supporting it.I am just telling you because I want you to set realistic goals and I don't want you to get disappointed in the end .

--Thomas

[Dario]

lol ! . I work on a commercial operating system, but then teams of people are involved dedicated to smaller subsystems.I takes lot of effort to create a commercial OS.Understand that I wont call ver0.01 of linux or v.1 of linux a commercial OS by any stretch of imagination.It became commercially viable when teams of people where involved and companies started supporting it.I am just telling you because I want you to set realistic goals and I don't want you to get disappointed in the end .

Naah Thomas...that was just a parody on Linus's first introduction to Linux on Minix mailing lists in 1992.
http://www.cs.cmu.edu/~awb/linux.history.html
I should be more careful dough...since there are people with such illusions and might get me serious.

[Combuster]

You already know something's wrong when you get a in response to a joke...

[Dario]

Don't get it twisted. It was a joke in response to a joke.
[joke]Next time I'll put "[joke][/joke]" tags.[/joke]

Cheers

[rJah]

@rjah: I use a simple tarball as a ramdisk "filesystem": it has lots of good tools for every platform and works as well as any readonly filesystem (with proper caching.)

hmm.... i'll have to look into that in the near future, but i'm trying to keep complexity at the lowest levels at a minimum, so we'll see.