"Gerard W. Patterson" wrote:
>
> On Tue, Jul 10, 2001 at 04:58:03PM -0600, D. Stimits wrote:
> >
> > What I'd find more interesting (since floppy images can't work with some
> > of the possible kernel options...size gets too large) would be an ISO CD
> > image that is equivalent to Tom's Root Boot (rescue with everything
> > under the sun), but XFS based (any rescue or repair tool would be nice,
> > even if it is for other fs's like reiser or ext3). Such could be done
> > with probably a 25 MB download (10 to 15 MB with gzip -9), and provide
> > the ability to format, repair, etc. I mention this because I am still
> > working on my own custom CD rescue (I recently discovered that one big
> > problem was that my CD was not a "multiread" version, and could not
> > handle CD-RW's, only CD-R's...duh). Add to that every test I do
> > requiring upload/download through 56k/33k modems...I'm slowly getting
> > there, but it is a lesson in humility.
> >
> > D. Stimits, stimits@xxxxxxxxxx
>
> Interestingly enough, this is exactly what I am working on for my current
> project. It is currently based on a 2.4.5 kernel with the xfs
> patch applied. (I started this a week and a bit ago and we are already at
> 2.4.6! jeezz). I have succeeded in getting a boot floppy image (I needed
> support for systems that didn't have CDROM drives) that contains the kernel
> (1015k) and a small ( <170k ) root file system. Go busy-box and uClibc!
> With this image, I can mount any large drive (magneto-optical) in this case
> and grab the rest of the system that I need (including binaries to replace
> the temporary busy-box ones. --vi is not vim and msh is not bash.) This image
> is also suitable for use to combine the floppy and the large drive onto a
> CDR for a complete, boot-able system.
>
> I plan on updating to a newer kernel and utilities once I am closer to
> release. (probably go for some type of auto-build). If you would like to
> collaborate let me know.
>
> Regards,
>
> Gerry
>
> --
> Gerard W. Patterson, B.Sc | Computalog Ltd.
> Software Engineering | Edmonton, AB
> <gerry.patterson@xxxxxxxxxxxxxx> | Canada
If you use a 2.4.5 kernel, or 2.4.6-pre1, be sure to manually add one
patch...without this patch there are some serious problems under various
circumstances, all of which are fatal. In kernel tree, fs/block_dev.c,
approximately line 596, is this function (the line starting with "+"
should be added, don't literally put in the "+"):
int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long
arg)
{
kdev_t rdev = to_kdev_t(bdev->bd_dev);
struct inode inode_fake;
int res;
mm_segment_t old_fs = get_fs();
if (!bdev->bd_op->ioctl)
return -EINVAL;
inode_fake.i_rdev=rdev;
+ inode_fake.i_bdev=bdev;
init_waitqueue_head(&inode_fake.i_wait);
set_fs(KERNEL_DS);
res = bdev->bd_op->ioctl(&inode_fake, NULL, cmd, arg);
set_fs(old_fs);
return res;
}
Use of loop devices is extremely risky without this. The first patches
occured in Alan Cox's 2.4.5-ac3, and it also appeared at 2.4.6-pre2.
Prior versions lack the fix.
D. Stimits, stimits@xxxxxxxxxx
|