Re: [Osso-develop] Back step about MPS!!! and news about vfs
Status: Abandoned
Brought to you by:
iurlano
|
From: Stefano R. <st...@in...> - 2001-07-04 20:01:58
|
> ... > Just one thing: > If you plan to use modules pool for balancing and others, each request to > a driver should pass through mps. Actually it should pass through mps just > the first time, then the user process remembers the pid. > If you want to do serious raid or net load balancing, each process have to > request a module pid at each operation. MPS can act as a request redirector or, better, the pool can be an external process that address the request to the right module: the pid is the pool one (in this way we make pool transparent to processes) > And, what do you mean by swap files? Explicitating the parallelism between > different swap files/partitions ? Yes! > By the way, I've started to implement vfs system, It is just a few lines, > but I've started to write mount.c too. > The status is : > vfs starts and demonizes itself :)) > ide starts and wait for vfs, then looks for its partitions and registers > them. Actually those are wrong values. > then fat starts and looks directly for ide WITHOUT passing through vfs and > works as before. > The mount utility simply ask vfs to print all the partitions. > > The reason I've not implemente fat to use vfs is that by now the osso disk > has no partition. We've got to find a way to create a partitioned osso > image. > This is what I've discovered : > fdisk can work with a disk image file, specifying the geometry (same as > ..mtoolsrc) so partitions can be created. > the .mtoolsrc (mtools.config) can be configured with a value > "partition" that specifies which partition of the disk to use and this > works too. > > So, by now I've been able to create a disk image composed of a correct > partition table and a formatted filesystem at partition 1 with osso copied > into. > > We still need the bootloader, or better, two bootloaders. > One is the osso bootloader that has to be put into the first sector of the > partition, so we'll have to modify the installer program. > The second has to be a system bootloader as lilo or grub which should be > able to load and execute the osso bootloader. > > Any idea, comments, suggestion ? I don't understand the problem... On my 486 I've installed osso using /dev/hdb3 as target, then I installed lilo specifing that partition as a boot partition (at boot I can start Linux, Windows or OSSO!!!). I refer to that partitions as o: in mtools. The problem for me is the reverse: I can't use FAT on 486 because partitions are not managed! A problem I found is this: how can I know if a disk have partitions or not?!? (For example first Zip drive has a unique partition (no MBR), instead my SyQuest drive have normal hd partitions). In other words, how can ide understand if it is managing a disk with or without partitions?!? > P.S. Ste, I've put partitions into vfs, do you agree or you'd prefer > another way ? Sincerely, I've started coding without thinking about it too > much cause I wanted to see mount working :)) > So if you don't like it we can rearrange it another way. I'm ok... and then I want to see a working mount too... Bye, Stefano Rosellini E-MAIL: st...@in... HOME PAGE: web.tiscali.it/stefano_rosellini |