Re: Paul's devel branch on Lombard?

Re: Paul's devel branch on Lombard?

[Hugh Caley]

I have not had this problem with building Paul's 2.2.12 kernel.  I did not use
a config file from a previous version, which sounds like the major difference
between us.

Or:  which version of gcc are you using?  I'm using the gcc from the fsirl
directory on dev.linuxppc.org, and the glibc and related files from the same
site.

Hugh

Date: Wed, 8 Sep 1999 15:02:20 -0500
From: Shaw Terwilliger <sterwill@io.nu>
Subject: Paul's devel branch on Lombard?

I'm using Pauls' stable branch (2.2.12) from his rsync archives
on my Lombard PowerBook.  I've wanted to use the power management
features in the new kernels, but I have yet to build a kernel that
fully boots.

As of today's kernel sources, it boots but pulls a Machine Check in
one of my init scripts.  Each time I boot it, it does in a different place.
The attached dump is failing on the "exim" script ("S20exim").

I'm using the same kernel configuration I'm using for 2.2.12 (which
works great from the SCSI to the modem to the framebuffer to USB),
but 2.2.12 doesn't offer sleep mode.  :)

A hand-typed dump appears below.  I'll send my .config if it would
help.

Machine check in kernel mode.
Caused by (From msr): regs c75dbd10 Unknown values in msr
NIP: C000BDCC XER: 00000000 LR: C0011DD0 REGS: c75dbd10 TRAP: 0200
MSR: 00009030 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = c75da000[111] 'S20exim' Last syscall: 90
last math c75da000
GPR00: B6337806 C75DBDC0 C75DA000 BB04D805 01743000 0007FFC0 C037BCC0 C037BCE0

GPR08: 00000008 00000001 00000000 00000001 24842224 018CF840 00000000 01808294

GPR16: 00000000 30056500 7FFFE408 30056530 00009032 075DBE80 00000000 C7D4CB00

GPR24: 01745000 01742000 01745000 00003000 01745000 C02CD220 FE8BCFFF 01743000

vector: 200 at pc = c000bdcc, msr = 9030, sp = c75dbdc0 [c75dbd10]
current = c75da000, pid = 111, comm = S20exim

- --
Shaw Terwilliger (sterwill@io.nu)



--
Hugh Caley, Unix Administrator
Babcock & Brown, San Francisco
510-524-1672
hughc@babcockbrown.com


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Shaw Terwilliger]

Hugh Caley wrote:
> I have not had this problem with building Paul's 2.2.12 kernel.  I did not use
> a config file from a previous version, which sounds like the major difference
> between us.

The problem is not with the 2.2.12 kernel, but with the 2.3.16 branch in
his "devel" archive.  The "stable" branch is working like a champ for me.
I'm sorry if I didn't make that very clear in my message (I had just
spent three minutes writing down all those addresses).  :)

> Or:  which version of gcc are you using?  I'm using the gcc from the fsirl
> directory on dev.linuxppc.org, and the glibc and related files from the same
> site.

If it matters, I'm using GCC from the Debian Potato branch (which I have
installed).  I believe it's just EGCS 1.1.2 release; but it's building 
2.2.12 great, 2.3.16 just isn't building.  I guess it's possible there's
a 2.3 kernel problem with the compiler, although I don't see any problems
in the builds.

Has anyone built kernels (either 2.2 or 2.3) with GCC 2.95?  I'd build a new
compiler (even just to try) if I could build the newest one.  :P


-- 
Shaw Terwilliger (sterwill@io.nu)

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Hugh Caley]

Shaw Terwilliger wrote:

> Has anyone built kernels (either 2.2 or 2.3) with GCC 2.95?  I'd build a new
> compiler (even just to try) if I could build the newest one.  :P
>
> --
> Shaw Terwilliger (sterwill@io.nu)

I build mine with 2.95, from the fsirl directory on dev.linuxppc.org.

Hugh

--
Hugh Caley, Unix Administrator
Babcock & Brown, San Francisco
510-524-1672
hughc@babcockbrown.com


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Martin Costabel]

Hugh Caley wrote:
> 
> Shaw Terwilliger wrote:
> 
> > Has anyone built kernels (either 2.2 or 2.3) with GCC 2.95?  I'd build a new
> > compiler (even just to try) if I could build the newest one.  :P
> >
> > --
> > Shaw Terwilliger (sterwill@io.nu)
> 
> I build mine with 2.95, from the fsirl directory on dev.linuxppc.org.

Here is my /proc/version:

root[432]#cat /proc/version
Linux version 2.3.16 (root@chezmoi) (gcc version 2.95.1 19990816
(release)) #63 Sun Sep 5 17:09:44 CEST 1999
root[433]#uptime
 10:57pm  up 4 days,  3:48,  5 users,  load average: 0.26, 0.14, 0.10

Kernel 2.3.16 is from cvs@openprojects, with 2 little patches to make it
compile and to recognize my ethernet card. I'll reboot in a minute to
try 2.3.17.

Compiler is a mixture of F. Sirl's patch and spec file with the
"official" 2.95.1 tar.gz file.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Shaw Terwilliger]

Martin Costabel wrote:
> Kernel 2.3.16 is from cvs@openprojects, with 2 little patches to make it
> compile and to recognize my ethernet card. I'll reboot in a minute to
> try 2.3.17.

I tried 2.3.16 last week (from cvs from openprojects), and had similar
machine check problems on boot.  It _did_ seem to recognize all my
hardware (perhaps the BMAC did not work, but I couldn't tell, it wouldn't
stay up that long).  

If I use the CVS kernel, will I _need_ 2.95, or will EGCS 1.1.2 work?

-- 
Shaw Terwilliger (sterwill@io.nu)

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Martin Costabel]

Shaw Terwilliger wrote:
> 
> Martin Costabel wrote:
> > Kernel 2.3.16 is from cvs@openprojects, with 2 little patches to make it
> > compile and to recognize my ethernet card. I'll reboot in a minute to
> > try 2.3.17.
> 
> I tried 2.3.16 last week (from cvs from openprojects), and had similar
> machine check problems on boot.  It _did_ seem to recognize all my
> hardware (perhaps the BMAC did not work, but I couldn't tell, it wouldn't
> stay up that long).
> 
> If I use the CVS kernel, will I _need_ 2.95, or will EGCS 1.1.2 work?

I think there is even some explicit warning somewhere that
gcc-2.95-compiled kernels are bad for your health or something. If you
ask Linus (or read Documentation/Changes), you are told to use
gcc-2.7.2.3, no kidding.

2.3.17 is now running, although I had to recompile it once; make
oldconfig had eaten my CONFIG_ADBMOUSE. Someone is playing games with
the config options. In make menuconfig, the "APPLE DESKTOP MOUSE" is
marked as "NEW". Mine doesn't look very new anymore :-)

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[David D. Kilzer]

Martin Costabel wrote:

>2.3.17 is now running, although I had to recompile it once; make
>oldconfig had eaten my CONFIG_ADBMOUSE. Someone is playing games with
>the config options. In make menuconfig, the "APPLE DESKTOP MOUSE" is
>marked as "NEW". Mine doesn't look very new anymore :-)

It probably changed to CONFIG_ADB_MOUSE.  This may be a side-effect of the
ADB merger between Linux/m68k and LinuxPPC.

Dave


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Paul Mackerras]

> It probably changed to CONFIG_ADB_MOUSE.  This may be a side-effect of the
> ADB merger between Linux/m68k and LinuxPPC.

Yes it did, and yes it is.  I have the merged ADB stuff in my
linux-pmac-devel tree now.

Paul.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Geert Uytterhoeven]

On Thu, 9 Sep 1999, David D. Kilzer wrote:
> Martin Costabel wrote:
> 
> >2.3.17 is now running, although I had to recompile it once; make
> >oldconfig had eaten my CONFIG_ADBMOUSE. Someone is playing games with
> >the config options. In make menuconfig, the "APPLE DESKTOP MOUSE" is
> >marked as "NEW". Mine doesn't look very new anymore :-)
> 
> It probably changed to CONFIG_ADB_MOUSE.  This may be a side-effect of the
> ADB merger between Linux/m68k and LinuxPPC.

Always run `make oldconfig' after updating your kernel sources, to make sure
you catch the new options.

Greetings,

						Geert

--
Geert Uytterhoeven ----------------- SONY Suprastructure Center Europe (SUPC-E)
Geert.Uytterhoeven@sonycom.com ------------------- Sint-Stevens-Woluwestraat 55
Phone +32-2-7248608 Fax +32-2-7242686 ---------------- B-1130 Brussels, Belgium


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Geert Uytterhoeven]

On Fri, 10 Sep 1999, Martin Costabel wrote:
> Shaw Terwilliger wrote:
> > Martin Costabel wrote:
> > > Kernel 2.3.16 is from cvs@openprojects, with 2 little patches to make it
> > > compile and to recognize my ethernet card. I'll reboot in a minute to
> > > try 2.3.17.
> > 
> > I tried 2.3.16 last week (from cvs from openprojects), and had similar
> > machine check problems on boot.  It _did_ seem to recognize all my
> > hardware (perhaps the BMAC did not work, but I couldn't tell, it wouldn't
> > stay up that long).
> > 
> > If I use the CVS kernel, will I _need_ 2.95, or will EGCS 1.1.2 work?
> 
> I think there is even some explicit warning somewhere that
> gcc-2.95-compiled kernels are bad for your health or something. If you
> ask Linus (or read Documentation/Changes), you are told to use
> gcc-2.7.2.3, no kidding.

Recommended compiler versions in Documentation/Changes apply to ia32 only.

> 2.3.17 is now running, although I had to recompile it once; make
> oldconfig had eaten my CONFIG_ADBMOUSE. Someone is playing games with
> the config options. In make menuconfig, the "APPLE DESKTOP MOUSE" is
> marked as "NEW". Mine doesn't look very new anymore :-)

Woops, didn't read thit that well. So please forgot about my former suggestion
to run `make oldconfig' :-)

Greetings,

						Geert

--
Geert Uytterhoeven ----------------- SONY Suprastructure Center Europe (SUPC-E)
Geert.Uytterhoeven@sonycom.com ------------------- Sint-Stevens-Woluwestraat 55
Phone +32-2-7248608 Fax +32-2-7242686 ---------------- B-1130 Brussels, Belgium


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Paul's devel branch on Lombard?

[Martin Costabel]

Geert Uytterhoeven wrote:
> 

> Woops, didn't read thit that well. So please forgot about my former suggestion
> to run `make oldconfig' :-)

I try to use always 'make oldconfig'. But sometimes this doesn't work
correctly, like in the case of the adbmouse or in 2.3.16 for the Apple
partitions. 
But now the big ADB merger is arriving, and everything is different
again. Curious if 2.3.18 will compile... (1 minute later: No, this is
not yet finished)

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

2.3.18 and CONFIG_ADB_MOUSE

[Martin Costabel]

In the latest vger-2.3.18 and also Paul's linux-pmac-devel, the
ADB_MOUSE cannot be selected any more, neither in 'make oldconfig' nor
in 'make menuconfig'.

While the logic in drivers/char/Config.in:
 
if [ "$CONFIG_ADB" = "y" ; then
    dep_tristate 'Apple Desktop Bus mouse support' CONFIG_ADB_MOUSE
$CONFIG_BUSM
OUSE

looks OK, it doesn't work. This never shows up in 'make menuconfig'.

And in arch/ppc/config.in, the line

   bool 'Support for PowerMac ADB mouse' CONFIG_ADBMOUSE

got trashed, probably by mistake.

Well, I suppose the adb merging is not quite finished, like the irq and
pci parts. I got it to compile, but it doesn't boot.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Joshua M. Thompson]

On Sat, 11 Sep 1999, Martin Costabel wrote:

> And in arch/ppc/config.in, the line
> 
>    bool 'Support for PowerMac ADB mouse' CONFIG_ADBMOUSE
> 
> got trashed, probably by mistake.

Nope, it's superceded by CONFIG_ADB_MOUSE. The ADB mouse driver is
cross-platform now.

-- 
funaho@jurai.org             | Voodoo is a very interesting religion for the
http://www.jurai.org/~funaho | whole family, even those members of it who are
                             | dead.
                             |  -- (Terry Pratchett & Neil Gaiman, Good Omens)


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Paul Mackerras]

Martin Costabel <costabel@wanadoo.fr> wrote:

> if [ "$CONFIG_ADB" = "y" ; then

Simple syntax error, somehow I missed a ] before the ;.

Paul.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[David D. Kilzer]

Martin Costabel wrote:

>In the latest vger-2.3.18 and also Paul's linux-pmac-devel, the
>ADB_MOUSE cannot be selected any more, neither in 'make oldconfig' nor
>in 'make menuconfig'.
>
>While the logic in drivers/char/Config.in:
>
>if [ "$CONFIG_ADB" = "y" ; then
>    dep_tristate 'Apple Desktop Bus mouse support' CONFIG_ADB_MOUSE
>$CONFIG_BUSMOUSE
>
>looks OK, it doesn't work. This never shows up in 'make menuconfig'.

There's a missing ']' before the semi-colon '; then' if you copied that
verbatim.

If that's not the problem, then did you say "y" to ADB bus support (or did
it even ask)?

Dave


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Martin Costabel]

Sorry for this, I was too tired when I wrote it. Sometimes the day job
can get in the way of successful Linux hacking :-)

I had seen the typo in drivers/char/Config.in, but forgot to correct it.
And then it seemed more logical to have the selection of ADB_MOUSE next
to ADB_KEYBOARD, but maybe it isn't. 

Anyway, in 'make menuconfig' the mouse selection is buried rather deep
down, on the  'Character devices' page after things like 'Non-standard
serial port support' and 'Unix98 PTY support'. Let's see... OK, this
seems to be a bug in drivers/char/Config.in. There is an 'endmenu'
missing just before 

mainmenu_option next_comment 
comment 'Mice' 

But, if I put an 'endmenu' there, then 'make xconfig' breaks. And 'make
xconfig' got it right before! (Right = Put 'Mice' and 'Video for Linux'
and so on into the main menu and not in the 'Character devices' menu).
Solution (but this is black magic, I don't understand it and better give
up on this):

Put a line with '#endmenu' instead of 'endmenu' there. Now both 'make
menuconfig' and 'make xconfig' are happy and behave in the same way
(well, almost, in 'make xconfig', the 'watchdog timer' selection is
still on the 'Character devices' page, but the 'watchdog card' selection
is in the main menu). Weird stuff.

Here is the patch:

cvs diff -u drivers/char/Config.in
Index: drivers/char/Config.in
===================================================================
RCS file: /cvs/linux/linux/drivers/char/Config.in,v
retrieving revision 1.80
diff -u -r1.80 Config.in
--- drivers/char/Config.in      1999/09/11 03:39:16     1.80
+++ drivers/char/Config.in      1999/09/11 07:02:10
@@ -61,6 +61,7 @@
   fi
   dep_tristate 'Support for user-space parallel port device drivers'
CONFIG_PPDEV $CONFIG_PARPORT
 fi
+#endmenu
 
 mainmenu_option next_comment
 comment 'Mice'

--
Martin
(Not drunk, it's 10am, but feeling slightly seasick after one hour with
the config.in files)

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Joshua M. Thompson]

On Sat, 11 Sep 1999, Martin Costabel wrote:

> I had seen the typo in drivers/char/Config.in, but forgot to correct it.
> And then it seemed more logical to have the selection of ADB_MOUSE next
> to ADB_KEYBOARD, but maybe it isn't. 

It is more logical, and that's the way I have it listed on the 68K Mac
port. What I believe should happen is that the ADB keyboard driver should
be put in drivers/char where the mouse driver already resides. Then they
can both be listed next to each other in the drivers/char/config.in file,
bracked by a nice clean 'if [ "$CONFIG_ADB" = 'y' ];'

-- 
funaho@jurai.org             | Voodoo is a very interesting religion for the
http://www.jurai.org/~funaho | whole family, even those members of it who are
                             | dead.
                             |  -- (Terry Pratchett & Neil Gaiman, Good Omens)


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Martin Costabel]

"Joshua M. Thompson" wrote:
> 
> On Sat, 11 Sep 1999, Martin Costabel wrote:
> 
> > I had seen the typo in drivers/char/Config.in, but forgot to correct it.
> > And then it seemed more logical to have the selection of ADB_MOUSE next
> > to ADB_KEYBOARD, but maybe it isn't.
> 
> It is more logical, and that's the way I have it listed on the 68K Mac
> port. What I believe should happen is that the ADB keyboard driver should
> be put in drivers/char where the mouse driver already resides. Then they
> can both be listed next to each other in the drivers/char/config.in file,
> bracked by a nice clean 'if [ "$CONFIG_ADB" = 'y' ];'

It's not so clear to me now. If one thinks of menus as in 'make
{x|menu}config', there is a submenu for mice, but nothing for keyboards,
probably because for PCs there is basically only one keyboard (?). When
USB support matures, this will have to change, I guess.

Another small nit to pick: The adbmouse driver is offered as a module,
but it doesn't compile when CONFIG_ADB_MOUSE=m is selected. The reason
is this new
__setup("adb_buttons=", adb_mouse_setup);
line which is a fallout from the implosion of init/main.c.
In <linux/init.h>, __setup is only defined #ifndef MODULE. I don't know
how other drivers handle this problem; in vger-2.3.18, the popular
"atyfb=" kernel argument seems to have disappeared completely, for
instance.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Geert Uytterhoeven]

On Mon, 13 Sep 1999, Martin Costabel wrote:
> Another small nit to pick: The adbmouse driver is offered as a module,
> but it doesn't compile when CONFIG_ADB_MOUSE=m is selected. The reason
> is this new
> __setup("adb_buttons=", adb_mouse_setup);
> line which is a fallout from the implosion of init/main.c.
> In <linux/init.h>, __setup is only defined #ifndef MODULE. I don't know

You have to use the mnoduleparams stuff. Look at de4x5.c for an example.

> how other drivers handle this problem; in vger-2.3.18, the popular
> "atyfb=" kernel argument seems to have disappeared completely, for
> instance.

There never was an `atyfb=' option. Use `video=atyfb:' instead.

Greetings,

						Geert

--
Geert Uytterhoeven ----------------- SONY Suprastructure Center Europe (SUPC-E)
Geert.Uytterhoeven@sonycom.com ------------------- Sint-Stevens-Woluwestraat 55
Phone +32-2-7248608 Fax +32-2-7242686 ---------------- B-1130 Brussels, Belgium


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Martin Costabel]

Geert Uytterhoeven wrote:
> 
> On Mon, 13 Sep 1999, Martin Costabel wrote:
> > Another small nit to pick: The adbmouse driver is offered as a module,
> > but it doesn't compile when CONFIG_ADB_MOUSE=m is selected. The reason
> > is this new
> > __setup("adb_buttons=", adb_mouse_setup);
> > line which is a fallout from the implosion of init/main.c.
> > In <linux/init.h>, __setup is only defined #ifndef MODULE. I don't know
> 
> You have to use the mnoduleparams stuff. Look at de4x5.c for an example.

Don't look at me for this. I'm just a (rather stupid, it seems)
consumer.
 
> > how other drivers handle this problem; in vger-2.3.18, the popular
> > "atyfb=" kernel argument seems to have disappeared completely, for
> > instance.
> 
> There never was an `atyfb=' option. Use `video=atyfb:' instead.

Yes, sorry, and it's parsed in fbmem.c. But I still can't see how atyfb
in 2.3.18 can work as a module. 

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: 2.3.18 and CONFIG_ADB_MOUSE

[Geert Uytterhoeven]

On Mon, 13 Sep 1999, Martin Costabel wrote:
> Geert Uytterhoeven wrote:
> > On Mon, 13 Sep 1999, Martin Costabel wrote:
> > > Another small nit to pick: The adbmouse driver is offered as a module,
> > > but it doesn't compile when CONFIG_ADB_MOUSE=m is selected. The reason
> > > is this new
> > > __setup("adb_buttons=", adb_mouse_setup);
> > > line which is a fallout from the implosion of init/main.c.
> > > In <linux/init.h>, __setup is only defined #ifndef MODULE. I don't know
> > 
> > You have to use the mnoduleparams stuff. Look at de4x5.c for an example.
> 
> Don't look at me for this. I'm just a (rather stupid, it seems)
> consumer.

Well, someone should fix the adbmouse driver...

> > > how other drivers handle this problem; in vger-2.3.18, the popular
> > > "atyfb=" kernel argument seems to have disappeared completely, for
> > > instance.
> > 
> > There never was an `atyfb=' option. Use `video=atyfb:' instead.
> 
> Yes, sorry, and it's parsed in fbmem.c. But I still can't see how atyfb
> in 2.3.18 can work as a module. 

I have to admit I never tried to use it as a module yet. I plan to do that,
though. When loaded as a module, it will use the default video mode.

Greetings,

						Geert

--
Geert Uytterhoeven ----------------- SONY Suprastructure Center Europe (SUPC-E)
Geert.Uytterhoeven@sonycom.com ------------------- Sint Stevens Woluwestraat 55
Phone +32-2-7248620 Fax +32-2-7242686 ---------------- B-1130 Brussels, Belgium


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/