[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Repost: Working patch to boot an iso image file from Gujin signature (ia32)


> sorry for the late response, it hit the time where i had no stable
> internet for about 6 weeks due to mooving. anyhow..

No need to be sorry, I also have a TODO list which is only growing.

> the patch is huge and i don't think all of it is necessary[0].

The high level logic is that, before searching all filesystems to
find a FROMISO= filename, you just check the memory at the beginning
of the BIOS IRQ 0x13 real mode handler for a signature.
If that signature is not found then do not modify anything.
I am not sure how the functions to find the iso file from partition
and disk size, IDE addresses and EBIOS disk serial number can be re-used
in the standard case because then you just have the filename of the iso.

> but first of all, rather than fetching gujin from the sf page, it
> should come from a debian package.
> any plans to upload a gujin package to debian?

I did everything I could to simplify the job of someone else adding
that package to the official repository, you even have a .deb package
in sourceforge, but I cannot commit the time to learn how to do that.
My TODO list is already too big, I cannot even find the time to do another
release of the source code which is needed.

> [0] basically you're replacing much of the live-boot own logic
> completely, which is not a problem per se, but in order to review and
> apply this, we should seperate things to logical pieces, means, first
> make the modifications/adjustments/fixes to the generic functions, and
> then add gujiin specificas on top of it.

The patch do not use the generic functions because you know a lot more
of the iso file so there is no need to mount each filesystem - and that
is basically what the generic function do.

> that way we can ensure we're
> not breaking other stuff and don't add yet another dublication layer to
> live-boot, which is already way too messy in its current state anyway :(

Note that Gujin handle that iso booting (when you select
"filename.iso:noemul") at a lower layer than other bootloader do, so
that the El-Torito bootloader of the CDROM is executed at a later stage.
Other bootloaders only handle fetching the kernel/initrd inside the iso
and run directly this kernel, which is a different problem (and Gujin
can also do that when you select filename.iso:{vmlinuz,initrd})

About not breaking other stuff, I would have like a kind of 10 other
messages from 10 other people saying if it worked for them, there is
so many cases and so many broken BIOSes...

Thanks for your answer,

Reply to: