Re: [uml-devel] Re: [uml-user] UML (unofficial) patches against vanilla 2.6.2

[BlaisorBlade <blaisorblade_spam@yahoo.it>]

Alle 06:23, venerdì 13 febbraio 2004, Jeff Dike ha scritto:
> blaisorblade_spam@yahoo.it said:
> > I've updated my patches for modules and so on to 2.6.2. Since the main
> > Uml  patch against 2.6.1 needs some easy-but-not-trivial changes to
> > work onto  2.6.2
>
> OK, I'm finishing merging this in.
>
> I don't understand what you're doing with G-02-LinkScriptCleanup.patch. 
> The stuff you're adding to vmlinux.lds.S is only needed for the final
> binary, which isn't built by this script.

I understand your point, but the "kallsyms" step is done on the vmlinux 
binary, not on the "linux" one, and it thinks that text symbols are between 
_stext and _etext; so you need at least to set correctly _stext and _etext, 
so the definition of the first must be added and the section .text must be 
put in the middle. If you look at the preprocessed vmlinux.lds.s, you'll see 
that it means _etext = 0. This creates problem for /proc/kallsyms, for the 
kernel oopses decoding, and possibly other stuff; anyway, I don't see why 
avoiding the fix.

> The next three hunks are also
> somewhat obscure.

Just cleanups and resyncs with the original linker scripts used by ld (i.e. 
setting the empty space to 0x90= NOP)

> The final one is merged.

> I don't like H-01-Pcap_link_fix.patch.  As much as that fixes pcap, we
> really need to find a fix that's local to arch/um/drivers.  I'll merge this
> only as a last-ditch thing if it becomes clear that there is no other way.

*Note*: in that patch, I forgot to un-uncomment pcap-objs in arch/um/drivers 
(I then fixed that, I don't know which version is there).
For this, I hope I can build a more general mechanism to do the same thing.

However that is not easy, since sub makefiles cannot set variables for the 
main Makefile (and in the kbuild process they are used after). Changing this 
would be much more crappy.

> H-02-Fixdep-improve.patch is similar, except it's not clear to me what
> you're fixing.  Here also, I don't want to munge the global kbuild with
> UML special cases.

If you change one setting in the .config, every file including 
arch/um/uml-config.h is marked as needing rebuild (maybe that  even means the 
whole kernel); since that is like autoconfig.h, with that patch it's treated 
in the same way. Since uml-config.h basically cannot go away,

> In H-04-Kbuild_cleanup.patch, the skas/Makefile patch isn't obviously an
> improvement over what's already there.
A resync against main Makefile (and more coherency in the output); sadly, that 
utility is not put inside Makefile.lib, so I had to duplicate it.

> In arch/um/Makefile, what's that @:
> doing?  A quiet true?
I wanted to specify an empty command list, so I did it as for sys_prepare 
target.
>  I left out the commented stuff.
That's perfectly ok. 
>  Send it back in
> when you're sure what you want there.

> As for I-01-Debian_workaround.patch, I had the impression that this was
> caused by a Debian bug.

It's a name change in the kernel headers, in the same way as for hostfs_user.c 
(and even the fix is the same one).

> If so, they should fix it, and I shouldn't have
> to put workarounds in UML for it.
>
> 				Jeff


-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id\x1356&alloc_id438&opÌk
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel