Re: [PATCH] biosdisk / open_device() messing up offsets

[Pavel Roskin <proski@gnu.org>]

On Sun, 2008-06-08 at 15:11 +0800, Bean wrote:
> On Sun, Jun 8, 2008 at 2:52 PM, Pavel Roskin <proski@gnu.org> wrote:
> >>
> >> The reason I add device support for hexdump is to debug the nand
> >> driver. I need to go around the disk cache and call the underlying
> >> disk driver directly, so I use disk->dev->read. For (nand), there is
> >> just one partition, so I didn't notice the problem then.
> >
> > Here's the patch.  Everything seems to be OK.  "--skip=N" is not
> > recognized, but it's something in the option parsing code.  "-s N" is
> > working.
> >
> > Please feel free to apply.
> 
> Perhaps we should keep the low level api, just fix the offset. Hexdump
> is normally used for debugging, and the disk cache is quite annoying.

OK, it's your choice.

> btw, please try the patch, it dumps the journal information. There
> could be quite some output, you can use grub-fstest to capture it.

It breaks installation:

# grub-install /dev/sda
grub-mkimage: error: cannot stat /usr/local/lib/grub/i386-pc/header:.mod

"header:" must be from the patch.

I cannot generate a short image easily because loopback devices are not
supported.  Perhaps I'll try making a real partition initialized with
zeroes or something like that.

-- 
Regards,
Pavel Roskin