On Thu, Nov 20, 2008 at 4:59 PM, Santiago <santi@xxxxxx> wrote: > Hello, > > thank you for your attention! > > @Peter Teoh: >> > But I'm a bit confused about your statement using fuse. >> > Well, I have to confess, that I don't really understand what fuse offers, >> > but >> >> FUSE just enable you to create any filesystem at userspace.....thus >> doing away with the need for virtual machine like what you are doing >> now. > > Ok - I gonna dig into the docs. > >> A good example of multi-fs working together is hostfs + >> any-filesystem > > the problem is, I can't find any intros for real beginners like me. > You have underestimated the power of Google....:-). http://www.google.com/search?q=simple+filesystem+linux+kernel&btnG=Google+Search And from the above I found this: http://shfs.sourceforge.net/ http://www.geocities.com/ravikiran_uvs/articles/rkfs-old.html====>on how to write a simple filesystem. These are very simple filesystem, which u can easily "adapt" (not "copy") for your own use. >> simplified function is in fs/libfs.c (not the same libfs earlier). > > So then I did not find libfs. > > I only found libfs.c in the kernel tree - on my debian box libfs stands for > font-services and is related to X. I am talking about: http://www.linuxhq.com/kernel/v2.6/28-rc5/fs/libfs.c > > @Roberto A. Foglietta: >> Why unionfs is not the answer to your needs? > > Hm, I only scratched the docs of unionfs and aufs a bit - the latter a bit > deeper - don't know whether I got it right: > > I want to "manage" several other filesystems (like unionfs and aufs). The main > difference is, I want to change the filesystem during a read or write > operation. Change in the meaning of mount/unmount - not just use another > pointer. > Don't know whether unionfs or aufs support such an operation. I found no hint > on that. > > @Peter Teoh: >> not sure my understanding is correct...but the difference is that >> Unionfs merge all FS into different directory hierarchy...but Santiago >> wanted something secret....just like that of ext3 and >> journalling...the file called ".journal" is in fact not visible to >> anyone but the ext3 itself. > > Like in OO - information hiding would be nice, but is not my major interest. > But as my teacher used to say: first make it work, ... > The usage should be easy and without misunderstanding. Hiding the > private/local filesystem could help. > > My major interest are: > - no merging of the used filesystems > - change the accessed filesystem during certain operations, or better said > accessing several filesystems during read, compareable to md systems. > > But as I stated in a previous mail - I don't wonna care about physical devices > (like md does). > > I don't know, wether it is possible to realize my idea ... > > So may be fuse is right for me - I'll gonna give it a try. > > kind regards > > Santiago > Overall....I am too confused on your requirements.....but there one major hurdle u have to clear too....is the set of userspace tools needed to manipulate the filesystem u created. eg, mkfs to make fs. -- Regards, Peter Teoh -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
