There is a patch for the kernel that make it able to preload modules
before the acutal booting.
I wonder if this feature will be included in the official linux kernel.
The patch can be found at
http://home.t-online.de/home/ChristianK./patches/ .
thanks;-)
On Wed, Aug 13, 2003 at 01:36:52AM +0300, Nufarul Alb wrote:
> There is a patch for the kernel that make it able to preload modules
> before the acutal booting.
> I wonder if this feature will be included in the official linux kernel.
> The patch can be found at
> http://home.t-online.de/home/ChristianK./patches/ .
> thanks;-)
No idea. It might help if someone (this means you) started maintaining
it and sending it in. =)
-- wli
On Wed, Aug 13, 2003 at 01:36:52AM +0300, Nufarul Alb wrote:
> There is a patch for the kernel that make it able to preload modules
> before the acutal booting.
>
> I wonder if this feature will be included in the official linux kernel.
Mutliboot support would be nice, not sure about the module loading thing.
But there's a bunch of issues with the paches:
(1) please port to 2.6 first because
(a) there's not chance this will get into 2.4
(b) 2.6 has the inkernel module loader so you don't have to duplicate
so much loader code.
(2) please convert from GNU to Linux style
(3) please use the predefined __ASSEMBLY__ instead of ASM
On 13 August 2003 03:29, William Lee Irwin III wrote:
> On Wed, Aug 13, 2003 at 01:36:52AM +0300, Nufarul Alb wrote:
> > There is a patch for the kernel that make it able to preload modules
> > before the acutal booting.
> > I wonder if this feature will be included in the official linux kernel.
> > The patch can be found at
> > http://home.t-online.de/home/ChristianK./patches/ .
> > thanks;-)
>
> No idea. It might help if someone (this means you) started maintaining
> it and sending it in. =)
Do we want to stuff every imaginable early userspace stuff into kernel?
<sarcasm>
I vote for iwconfig and cipe tunnels, because I mount my root filesystem
over them!
</sarcasm>
--
vda
Christoph Hellwig wrote:
>On Wed, Aug 13, 2003 at 01:36:52AM +0300, Nufarul Alb wrote:
>
>
>>There is a patch for the kernel that make it able to preload modules
>>before the acutal booting.
>>
>>I wonder if this feature will be included in the official linux kernel.
>>
>>
>
>Mutliboot support would be nice, not sure about the module loading thing.
>
>But there's a bunch of issues with the paches:
>
>(1) please port to 2.6 first because
> (a) there's not chance this will get into 2.4
> (b) 2.6 has the inkernel module loader so you don't have to duplicate
> so much loader code.
>(2) please convert from GNU to Linux style
>(3) please use the predefined __ASSEMBLY__ instead of ASM
>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to [email protected]
>More majordomo info at http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at http://www.tux.org/lkml/
>
>
>
Those are not my patches. They belong to a certain Christian Koenig. He
doesn't mantain them any more and I'm searching for someone to mantain
this project.
I know very little about kernel programming, but I liked the idea of
having such a feature in the kernel. It gives more freedom in compiling
the kernel. Maybe one day we would have the main piece of the kernel as
a standard and only have to recompile modules. THIS HAS NOTHING TO DO
WITH MICROKERNELS. Multibooting is a different stuff.
SO, if somebody knows who might want to update this patches, it would be
great.
Thanks!!
> I know very little about kernel programming, but I liked the idea of
> having such a feature in the kernel. It gives more freedom in compiling
> the kernel. Maybe one day we would have the main piece of the kernel as
> a standard and only have to recompile modules. THIS HAS NOTHING TO DO
> WITH MICROKERNELS. Multibooting is a different stuff.
I swear, that sounds like a solaris kernel =) Even the CPU is a module.
--
Lab tests show that use of micro$oft causes cancer in lab animals
Jim Carter wrote:
>On Wed, 13 Aug 2003, Nufarul Alb wrote:
>
>
>>>>There is a patch for the kernel that make it able to preload modules
>>>>before the acutal booting.
>>>>
>>>>
>
>Perhaps I'm misunderstanding the issue, but this sounds like a job for
>initrd. Someone jokingly mentioned booting wirelessly through a
>cipe pipe, so you would stick iwconfig and cipe (and your RAID driver and
>all the other wierd stuff) in the initrd. A lot of distros do this for you
>automatically, including SuSE.
>
>So how are the kernel and initrd images read? Grub has drivers for a fair
>number of devices, and can read i86 BIOS-supported discs, conventional
>Ethernet, some RAID (I think), etc. You can't load modules from your
>regular filesystem until you can read your filesystem, meaning that the
>module used to read your filesystem has to be provided by some other means.
>Like initrd.
>
>Hope this helps rather than smokes up the issue.
>
>James F. Carter Voice 310 825 2897 FAX 310 206 6673
>UCLA-Mathnet; 6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA 90095-1555
>Email: [email protected] http://www.math.ucla.edu/~jimc (q.v. for PGP key)
>
>
>
the use of initrd is a real pain in the butt. with multibooting GRUB
loads the modules into memory and the kernel can take them from there.
In this way the main kernel can be smaller, you won't need built-in
filesystem drivers (they will be preloaded by GRUB), you will not have
to recompile the entire kernel all the time, but only the modules.
The MAIN THING that a multiboot kernel will solve is the problem of
module portability. I think that the reason why hardware producers are
not making drivers for linux is that they have to make them open source
in order to be compiled by every user. With multibooting it become
possible to make a standard type of the main image of the kernel and all
the modules will be built for it. So, the hardware vendors won't have to
make public their sources and they will be encouraged to write drivers.
AND besides all that you can do many operations before mounting the
actual root like, for example, make a module that encapsulates a nice
boot animation, a progress bar, show the services start in a nice
graphical way on top of the framebuffer.Of course this can be done with
the current kernel, but you have to rebuild the entire kernel each time,
errors will appear all the time.
Nufarul Alb wrote:
> the use of initrd is a real pain in the butt.
It's mainly a packaging issue. So far, it seems that nobody has
written a useful set of generic tools to build an initrd. But
the ingredients are there.
> with multibooting GRUB
> loads the modules into memory and the kernel can take them from there.
A solution that depends on GRUB and that only works for
modules ? That's not so nice. What if a driver also needs some
setup script, a user-space demon, or has to download a non-GPL
firmware binary ?
If you envision a use where someone would load a set of drivers
from removable media to bring up the system, it would be more
flexible to allow for multiple initrds/initramfs, which then
contain whatever code or data is needed.
And please don't stop there, but also specify what these
additional file systems should contain (e.g. a script
"initializeme" in the top-level directory, or such).
> The MAIN THING that a multiboot kernel will solve is the problem of
> module portability. I think that the reason why hardware producers are
> not making drivers for linux is that they have to make them open source
> in order to be compiled by every user.
So we're just hallucinating all those binary-only drivers ? :-)
> With multibooting it become
> possible to make a standard type of the main image of the kernel and all
> the modules will be built for it.
A developer's dream - the immutable standard kernel with
internal data structures frozen for all eternity ;-))
> AND besides all that you can do many operations before mounting the
> actual root like, for example, make a module that encapsulates a nice
> boot animation,
Once kexec makes into the mainstream kernel, I'm sure we won't
have to wait long for such things to show up.
> a progress bar, show the services start in a nice
> graphical way on top of the framebuffer.
That's entirely a user-space issue.
- Werner
--
_________________________________________________________________________
/ Werner Almesberger, Buenos Aires, Argentina [email protected] /
/_http://www.almesberger.net/____________________________________________/
On 13 August 2003 03:29, William Lee Irwin III wrote:
>> No idea. It might help if someone (this means you) started maintaining
>> it and sending it in. =)
On Wed, Aug 13, 2003 at 10:18:27AM +0300, Denis Vlasenko wrote:
> Do we want to stuff every imaginable early userspace stuff into kernel?
> <sarcasm>
> I vote for iwconfig and cipe tunnels, because I mount my root filesystem
> over them!
> </sarcasm>
I had more in mind the notion that it generally helps to hack on things
and send things in yourself as opposed to asking for some random piece
of wildly out-of-date code to get merged.
-- wli
Followup to: <[email protected]>
By author: Nufarul Alb <[email protected]>
In newsgroup: linux.dev.kernel
>
> There is a patch for the kernel that make it able to preload modules
> before the acutal booting.
>
> I wonder if this feature will be included in the official linux kernel.
>
Use initramfs.
-hpa
--
<[email protected]> at work, <[email protected]> in private!
If you send me mail in HTML format I will assume it's spam.
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64