From: gary ng <gar...@ya...> - 2006-12-16 10:51:17
|
Hi, I am wondering how difficult it is to add this. With the 0.8 build(udev), I can boot and run debian's etch installer(netinstall initrd) without any special procedure other than having the proper pcap driver(so colinux can access the network). However, the debian installer's partition manager would try to partition the block device and failed. At the moment, I need to switch to another virtual terminal and do the "mke2fs /dev/cobd0" then re-run the partitioner so it would not try to partition it again. With partition table support, it would be a smooth install which would be even more impressive. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com |
From: Henry N. <Henry.Ne@Arcor.de> - 2006-12-21 22:28:25
|
gary ng wrote: > I am wondering how difficult it is to add this. > > With the 0.8 build(udev), I can boot and run debian's > etch installer(netinstall initrd) without any special > procedure other than having the proper pcap driver(so > colinux can access the network). > > However, the debian installer's partition manager > would try to partition the block device and failed. At > the moment, I need to switch to another virtual > terminal and do the "mke2fs /dev/cobd0" then re-run > the partitioner so it would not try to partition it > again. > > With partition table support, it would be a smooth > install which would be even more impressive. It's in our todo/wish list. The problem under colinux is, after using partition tables, all block devices (hda1, hda2, hda3, ... ) are relative from the sector of partition table. In other words a using partition tables need to define a file as whole complete disk, not as single partition. Or we need a transformer for every disk access, that would be transform the offset into a filename and a difference offsets. Currently, the disk access to a partition is a direct access to the offset in the file (without recalculation the prameters). -- Henry Nestler |
From: David H. L. Jr. <dh...@dl...> - 2006-12-22 09:40:39
|
Henry Nestler wrote: > It's in our todo/wish list. > The problem under colinux is, after using partition tables, all block > devices (hda1, hda2, hda3, ... ) are relative from the sector of > partition table. In other words a using partition tables need to define > a file as whole complete disk, not as single partition. > > Or we need a transformer for every disk access, that would be transform > the offset into a filename and a difference offsets. > > Currently, the disk access to a partition is a direct access to the > offset in the file (without recalculation the prameters). > > Option 1: Make each virtual disk the equivalent to a virtual physical drive. I have not looked at the colinux disk driver, But this just means mapping cobd to a single physical file.I should think that the existing Linux code should take care of partition tables and properly calculating offsets. If you want multiple files to correspond to different files put one partition/file and have the equivalent of hda,hdb.hdc.... The problem with this approach it breaks existing colinux disk images and systems. Option 2: Leave the cobd1,cobd2,.... stuff alone but create a virtual partition table that is read/written when you access /dev/cobd the save as if you access /dev/hda. It is more complex. While you may wish to allow writes to the virtual partition table, you probably do not really want to use the information that is written - it would not make much sense to write it in a context where each file is a different partition. -- Dave Lynch DLA Systems Software Development: Embedded Linux 717.627.3770 dh...@dl... http://www.dlasys.net fax: 1.253.369.9244 Cell: 1.717.587.7774 Over 25 years' experience in platforms, languages, and technologies too numerous to list. "Any intelligent fool can make things bigger and more complex... It takes a touch of genius - and a lot of courage to move in the opposite direction." Albert Einstein |