Re: [Cooker] kernel-2.6-test9 config

["Andrey Borzenkov"]

-Original Message-

> 
> "Andrey Borzenkov"  <[EMAIL PROTECTED]> writes:
> 
> > > I'm working on our linuxrc to be able to use cramfs.
> > 
> > I on working on linuxrc based on initramfs. Now when
> > I got Al Viro advice how to free it it should be easy. This allows
> > to avoid *any* filesystem builtin in kernel :)
> 
> his solution would be a small c program that just glob and unlink
> everything ...
> 

sure, but he explained how to access rootfs after it has been
overlaid by normal root mount :) else you have no way to remove
anything

> > hopefully I get prototype in a couple of weeks. IMHO it is better
> > than hack more filesystems.
> 
> 

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

-Original Message-

> 
> Olivier Blin <[EMAIL PROTECTED]> writes:
> 
> > How can harddrake detect the mouse/keyboard type ?
> > It is easy ?
> 
> it's not.
> see:
> - the issues with detecting the wheel of ps/2 mice

wrt to kernel it is irrelevant. we speak about detecting correct
driver (module) for mouse. To this extent we have (on 2.6)

"psmouse" - for any sort of PS/2 (type is autodetected by driver)
"amimouse" - not relevant for x86
"inport" - busmouse (Microsoft) whatever it is :)
"lgibm" - Logitech bus mouse
"logips2pp" - Logitech PS/2++ - no idea what it is
"maplemouse" - irrelevant
"pc110pad" - IBM touchpad (one of)
"rpcmouse" - irrelevant
"sermouse" - requires user-level helper :(
"synaptics" - for synaptics

what I meant is that harddrake could - automatically or manually -
store mouse driver somewhere, probably in /etc/default/mouse,
for later use by mkinitrd or some otehr script, e.g. gpm

same for keyboard where we really have choice between

"atkbd"
"xtkbd"
"98kbd" - anyone knows waht it is?

> - the recent issues in kernel-2.6.0-test9 with sharing mouse &
>   keyboard box for several machine
> 
> 

Re: [Cooker] kernel-2.6-test9 config

["Andrey Borzenkov"]

-Original Message-

> 
> On Mon, 17 Nov 2003 21:22:10 +0300
> Andrey Borzenkov <[EMAIL PROTECTED]> wrote:
> 
> > at least make it depend on CONFIG_CRAMFS_INITRD. it is not that crmafs
> > is available but that it is used for initrd.
> > 
> > but if it is really read-only, our initrd needs some tweaking indeed
> 
> What's this CONFIG_CRAMFS_INITRD option ? I can't find it.
> 

I mean make patch depend on CONFIG_CRAMFS_INITRD.

> 
> I'm working on our linuxrc to be able to use cramfs.

I on working on linuxrc based on initramfs. Now when
I got Al Viro advice how to free it it should be easy. This allows
to avoid *any* filesystem builtin in kernel :)

hopefully I get prototype in a couple of weeks. IMHO it is better
than hack more filesystems.

Re: [Cooker] Re: status of cramfs initrd in 2.6

["Andrey Borzenkov"]

> By the way, what should we do about devfs ?
> It is deprecated, but is there any working alternative yet ?
> udev needs userspace tools IIRC.
> 

leave devfs as is.

udev is not a viable replacement as yet so no real 
infrastructure to replace devfs currently exists. I have not seen
anyhing more than proof of concept.

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Monday 17 November 2003 18:59, Svetoslav Slavtchev wrote:
> > ... cramfs
> >
> > > > > > > Should I make it an option in mkinitrd ?
> > > > > >
> > > > > > I guess yes. But this is compatibility problem. It is not
> > > > > > supported on vanilla kernel so you have to detect patched kernel
> > > > > > somehow. And how are you going to do it? You will get a flood of
> > > > > > "bug reports" from users who compiled their own kernels (vanilla
> > > > > > ones or whatever) and cannot load initrd.
> > > > >
> > > > > just check for /mdk/ and ! /custom/ in uname -r output
> > > >
> > > > that is still just wild guess :(
> > >
> > > but it'll work...
> > > grep /proc/filesystems else
> >
> > not working. you can build initrd for 2.6 under 2.4 or other way
> > round. or even out of miniroot.
>
> what about
>
> grep cramfs /boot/System.map-[KVER] ?

at least make it depend on CONFIG_CRAMFS_INITRD. it is not that crmafs is 
available but that it is used for initrd.

but if it is really read-only, our initrd needs some tweaking indeed

> have you got my last mails with the modified mkbuild.pl & co ?

yes

> have you tried them ?

no :)

Re: [Cooker] kernel-2.6-test9 config

["Andrey Borzenkov"]

... cramfs
> > > > > Should I make it an option in mkinitrd ?
> > > > 
> > > > I guess yes. But this is compatibility problem. It is not
> > > > supported on vanilla kernel so you have to detect patched kernel
> > > > somehow. And how are you going to do it? You will get a flood of
> > > > "bug reports" from users who compiled their own kernels (vanilla
> > > > ones or whatever) and cannot load initrd.
> > >
> > > just check for /mdk/ and ! /custom/ in uname -r output
> > 
> > that is still just wild guess :(
> 
> but it'll work...
> grep /proc/filesystems else
> 

not working. you can build initrd for 2.6 under 2.4 or other way
round. or even out of miniroot.

I'd wish to use initramfs if I knew how to free space taken by it :(

Re: [Cooker] kernel-2.6-test9 config

["Andrey Borzenkov"]

> i've uploaded your work. hotplug is maintained by fredl.

good he has been silent wrt initscirpts as well. I hope he's
not following fpons ...

> btw your changes result in quite some changes (are you sure we haven't
> lost any features ?) :

you never know till you test it.

> --- hotplug--files.old  2003-11-17 12:06:28.0 +0100
> +++ hotplug--files.new  2003-11-17 12:06:28.0 +0100
> @@ -1,37 +1,27 @@
>  /etc/hotplug
> -/etc/hotplug.d
> -/etc/hotplug.d/default
> -/etc/hotplug.d/default/default.hotplug
 /etc/hotplug/blacklist
> -/etc/hotplug/dasd.agent
> -/etc/hotplug/firmware.agent

looks like reverse diff to me. 
[...]
> -/usr/share/doc/hotplug-2003_08_05
> -/usr/share/doc/hotplug-2003_08_05/ChangeLog
> -/usr/share/doc/hotplug-2003_08_05/README
> +/usr/share/doc/hotplug-2002_04_01
> +/usr/share/doc/hotplug-2002_04_01/ChangeLog
> +/usr/share/doc/hotplug-2002_04_01/README

mine is based on post-2003_08_05 CVS :)

> *
> --- hotplug--requires.old   2003-11-17 12:06:28.0 +0100
> +++ hotplug--requires.new   2003-11-17 12:06:28.0 +0100
> @@ -1,5 +1,4 @@
>  /bin/sh  
> -bash  
>  chkconfig  
>  ifplugd  
>  libc.so.6  

I do not know actually; is it direct or reverse diff?

>> > Should I make it an option in mkinitrd ?
>> >
>> 
>> I guess yes. But this is compatibility problem. It is not supported
>> on vanilla kernel so you have to detect patched kernel somehow. And
>> how are you going to do it? You will get a flood of "bug reports"
>> from users who compiled their own kernels (vanilla ones or whatever)
>> and cannot load initrd.
>
> just check for /mdk/ and ! /custom/ in uname -r output

that is still just wild guess :(

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 20:15, Olivier Blin wrote:
> On Sun, 16 Nov 2003 12:53:02 +0300
>
> Andrey Borzenkov <[EMAIL PROTECTED]> wrote:
> > > > they could be loaded unconditionaly after root is mounted
> > > > (there might be some warnings, but nothing that could harm the
> > > > system)
> > >
> > > Do you want to try to load *all* mouse and keyboard drivers at boot
> > > ? That might work, but hey, that's not so smart :)
> >
> > come on it is really just as smart as compiling all of them into
> > kernel.
>
> Yes, I agree, it's the same.
> But can't we use something like hotplug to automatically detect input
> devices and load needed modules ?
>

no. you have to load them and let them figure for themselves :)

> If we use the "load all modules" approach, where should this modules be
> loaded ? In a new init script ?

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 15:45, Svetoslav Slavtchev wrote:
> > On Sunday 16 November 2003 06:20, Svetoslav Slavtchev wrote:
> > > my current patch tarball is uploading
> > > linux-test9s5.tar.bz2
> >
> > I tried 3rdparty but it won't work as is. Dynamically generating Makefile
> > and
> > Kconfig highly confuses build system; BTW it won't even compile if all
> > 3rdparty are modules.
>
> strange,
> i never compiled smth in 3rdparty built in, always as modules,
> and never had problems
>

all of this in build != src. after make vmlinux:

   ld -m elf_i386  -T arch/i386/kernel/vmlinux.lds.s 
arch/i386/kernel/head.o arch/i386/kernel/init_task.o   init/built-in.o 
--start-group  usr/built-in.o  arch/i386/kernel/built-in.o  
arch/i386/mm/built-in.o  arch/i386/mach-default/built-in.o  kernel/built-in.o  
mm/built-in.o  fs/built-in.o  ipc/built-in.o  security/built-in.o  
crypto/built-in.o  lib/lib.a  arch/i386/lib/lib.a  lib/built-in.o  
arch/i386/lib/built-in.o  drivers/built-in.o  sound/built-in.o  
3rdparty/built-in.o  arch/i386/pci/built-in.o  arch/i386/power/built-in.o  
net/built-in.o --end-group  -o .tmp_vmlinux1
ld: cannot open 3rdparty/built-in.o: No such file or directory
make[1]: *** [.tmp_vmlinux1] Error 1
make: *** [vmlinux] Error 2
make: Leaving directory `/home/bor/src/linux-2.6.0-test9'


> > I suggest you use mkbuild.pl to generate static Makefile and Kconfig. How
> > often is it going to change? It could be done as part of making RPM.
>
> i tried to follow the 2.4 way, and i had no problems so far
> except that "make mrproper"  breaks "make clean" cause it deletes
> 3rdparty/Makefile
>
> but it's pretty trivial to build it only once
>

using patch below there is no need to remove files; they should be updated if 
you change subdirs (add or remove).

> > Setting 3rdparty as part of drivers-y forces it to expect
> > 3rdparty/built-in.o
> > but it is not built unless something in 3rdparty is compiled as builtin.
> > I
> >
> > was about to initialize SUBDIRS but then it prevents from overriding it
> > on
> >
> > command line  probably something like
> >
> > ifeq ($(origin SUBDIRS),file)
> >  SUBDIRS+=3rdparty
> > endif
>
> may be adding to the Kconfig's in each 3rdparty tarball "select 3rdparty"
> and
> -
> drivers-y   := drivers/ sound/
> if $(CONFIG_3RDPARTY)
> drivers-y   += 3rdparty/
> endif
>
> would fix it
>

it is not condition. The problem is "-y". It means you are going to build 
subdir contents in kernel. But 3rdparty/Makefile does not build 
3rdparty/built-in.o itself nor has it any dependencies to build it. If 
everything is module it will get only obj-m += ...; and no built-in.o is 
generated. I do not know if it a bug or feature. 

setting SUBDIRS means kbuild just descends and makes modules as expected. I do 
not like this much but I am not sure how to do it differently. check other 
top-level makefiles, all of them has unconditional obj-y.

> > of course neither Makefile nor Kconfig should be removed by mrproper
> > then.
> >
> > BTW your current setup obviously does not work if srctree != buildtree. I
> > attempted to fix it but hopelessly failed :)
> >
> :(
>
> never tried that
>
> what were the errors,
> not finding some of the headers ?
>


{pts/0}% ./makelinux V=1 oldconfig
make: Entering directory `/home/bor/src/linux-2.6.0-test9'
make -C /home/bor/build/linux-2.6.0-test9   \
KBUILD_SRC=/home/bor/src/linux-2.6.0-test9  KBUILD_VERBOSE=1\
KBUILD_CHECK= -f /home/bor/src/linux-2.6.0-test9/Makefile oldconfig
(cd 3rdparty ; perl ./mkbuild.pl)
Can't open perl script "./mkbuild.pl": No such file or directory
make[1]: *** [3rdparty/Makefile] Error 2
make: *** [oldconfig] Error 2
make: Leaving directory `/home/bor/src/linux-2.6.0-test9'


this is for a start :)

attached simple diff for this but that is not what I like very much.

if you are going to do it this way, here is diff against Makefile that 
correctly automatically rebuilds Makefile and Kconfig when you add or change 
subdirs in 3rdparty. and slightly modified mkbuild.pl

it was useful exercise :) still I ask myself if it does make sense to generate 
them dynamically.
--- ../tmp/linux-2.6.0-test9/Makefile	2003-10-27 19:02:59.0 +0300
+++ linux-2.6.0-test9/Makefile	2003-11-16 22:07:49.576939032 +0300
@@ -314,6 +314,18 @@
 scripts/fixdep:
 	$(Q)$(MAKE) $(build)=scripts $@
 
+# Create Makefile/Kconfig for all the 3rdparty stuff
+$(srctree)/3rdparty/Makefile $(srctree)/3rdparty/Kconfig:  FORCE $(srctree)/3rdparty/mkbuild.pl
+	@set -e;\
+	echo '  CHK $@';\
+	mkdir -p $(dir $@); \
+	(cd $(@D) && ./mkbuild.pl $(@F) > [EMAIL PROTECTED]);  \
+	if [ -r $@ ] && cmp -s $@ [EMAIL PROTECTED]; then  \
+		rm -f [EMAIL PROTECTED];   \
+	else\
+		echo '  UPD $@';\
+		mv -f [EMAIL PROTECTED] $@;\
+	fi

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 16:57, Svetoslav Slavtchev wrote:
> > On Sunday 16 November 2003 15:21, Svetoslav Slavtchev wrote:
> > > > you missed my point. modular IDE is unsafe because there is no
> >
> > reference
> >
> > > > counting for IDE modules.
[...]
> wouldn't it be then sufficient just to prevent module unloading

apparently that is not as bad.

for PCI modules (most of) are effectively locked in xxx_init_one() by 
incrementing usage count. As nothing ever decrements it (and ->remove is not 
provided) modules will stay forever if corr. device is available.

For PNP I am not sure not for other non-PCI controller.

given that majority systems now are PCI it hopefully avoids problems in most 
common case.

have you tried build IDE as module?

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 15:21, Svetoslav Slavtchev wrote:
> >
> > you missed my point. modular IDE is unsafe because there is no reference
> > counting for IDE modules. their patch does not change it, it just adds a
> > couple of EXPORT_MODULES. Nor does patch really fixes
> > MOD_{INC,DEC}_USE_COUNT
> > warnings
>
> the warnings could be easily fix> ed, no ?-) 
>

no. you can't just remove MOD_{INC,DEC}_USE_COUNT without providing 
alternative refcount mechanism. Warnings are put there for a reason :)

> > it can't be solved by fixing compilation problems, it needs serious
> > redesign
> > of IDE layer.
>
> that's true, but it's in that state for a long, long time
>
> and how many users are going to  load additional ide drivers once the
> system is up
> and running and unload the ide modules,  once they are loaded ?

PC card IDE? harddrake (or something else) probing for right driver for your 
card?

> without modular ide i don't think there is a way to really reduce the
> kernel size
>

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 15:10, Svetoslav Slavtchev wrote:
> > On Sunday 16 November 2003 05:31, Olivier Blin wrote:
> > > > --
> > > > arent the following pretty important,
> > > > and alwasys needed ?
> >
> > the right question is - are they needed to *boot* kernel?
> >
> > > > > > -CONFIG_PACKET=y
> > > > > > +CONFIG_PACKET=m
> > > > >
> > > > > module
> > > > >
> > > > > >  CONFIG_PACKET_MMAP=y
> > > > > >  CONFIG_NETLINK_DEV=m
> > > > > > -CONFIG_UNIX=y
> > > > > > +CONFIG_UNIX=m
> > > > >
[...]
>
> you are right
> they are needed only for NFS root, which IIRC is disabled in the main 2.4
> kernel,
> but they should be loaded right after boot ( modprobe.preload ? )
>

they are loaded by virtue of MODULE_ALIAS magic :).

{pts/1}% modprobe -c | grep net-pf
alias net-pf-1 unix
alias net-pf-2 ipv4
alias net-pf-3 off
alias net-pf-4 ipx
alias net-pf-5 appletalk
alias net-pf-6 off
alias net-pf-10 off
alias net-pf-17 af_packet
alias net-pf-19 off


(those marked "off" still come from  static modprobe.default because they are 
not in kernel yet).

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 06:20, Svetoslav Slavtchev wrote:
> my current patch tarball is uploading
> linux-test9s5.tar.bz2
>

I tried 3rdparty but it won't work as is. Dynamically generating Makefile and 
Kconfig highly confuses build system; BTW it won't even compile if all 
3rdparty are modules.

I suggest you use mkbuild.pl to generate static Makefile and Kconfig. How 
often is it going to change? It could be done as part of making RPM.

Setting 3rdparty as part of drivers-y forces it to expect 3rdparty/built-in.o 
but it is not built unless something in 3rdparty is compiled as builtin. I 
was about to initialize SUBDIRS but then it prevents from overriding it on 
command line  probably something like

ifeq ($(origin SUBDIRS),file)
 SUBDIRS+=3rdparty
endif

of course neither Makefile nor Kconfig should be removed by mrproper then.

BTW your current setup obviously does not work if srctree != buildtree. I 
attempted to fix it but hopelessly failed :)

-andrey

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 06:20, Svetoslav Slavtchev wrote:
> > On Sat, 15 Nov 2003 23:20:00 +0100 (MET)
> >
> > "Svetoslav Slavtchev" <[EMAIL PROTECTED]> wrote:
> > > > nope, module does not work. Even if it possible to fix all missing
> > > > exported
> > > > symbols, IDE layer does not provide for any sort of module reference
> > > > counting
> > > > (for chipset drivers)  meaning it is simply unsafe. Which implies
> > > > kernel must
> > > > include all IDE drivers :(
> > >
> > > IIRC debian use it for ages
> > > ( and they are far no that experimental as mandrake is)
> > > and the kernel becomes really small
> > >
> > > with Os --  bzImage ~ 1400kb
> > > with Os + modular ide -- ~ 11xx kb
> > >
> > > it should work, we only need to hack mkinitrd
> > > & may be harddrake/ drakx
> > > +++ include the debian patch
> >
> > I wonder how they do it.
> > Is there a patch for modular IDE in their kenrel ?
> > It is officially broken, even in 2.4 .
>
> yes, they do have patches both for 2.4 & for 2.6
>

you missed my point. modular IDE is unsafe because there is no reference 
counting for IDE modules. their patch does not change it, it just adds a 
couple of EXPORT_MODULES. Nor does patch really fixes MOD_{INC,DEC}_USE_COUNT 
warnings

it can't be solved by fixing compilation problems, it needs serious redesign 
of IDE layer.

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Sunday 16 November 2003 00:44, Olivier Blin wrote:
> > > --
> > > -CONFIG_X86_ALIGNMENT_16=y
> > > +CONFIG_X86_GOOD_APIC=y
> >
> > I can't find where it is referenced
>
> in arch/i386/Kconfig
>

I mean - referenced in the *code*. this looks like leftover that is not more 
used.
[...]
>
> > > -CONFIG_PACKET=y
> > > +CONFIG_PACKET=m
> >
> > module
> >
> > > -CONFIG_UNIX=y
> > > +CONFIG_UNIX=m
> >
> > module
>
> I was not sure for those.
> Almost everybody use it, even if the box isn't networked.
>

see another mail. Is it needed to *boot* kernel?
[...]

> > There is no problem loading them out of initrd. Of course it needs
> > harddrake support finally.
> >
> > WDYT?
>

it does it already eevn if it does not autodetect them. We still need some 
extra code to configure correct modules; probably some ldetect enhancements.
[...]

> > > -CONFIG_NFS_FS=y
> >
> > why? OK diskless stations ... that is the only application that may
> > require it. this is rather special case for generic kernel?
>
> i'm not sure, some users may mount /home by NFS
> How many bytes is the nfs code ?
>

{pts/1}% LC_ALL=C ll fs/nfs/*ko
-rw-r--r--1 bor  bor209936 Nov  8 16:27 fs/nfs/nfs.ko
{pts/1}% size fs/nfs/*ko
   textdata bss dec hex filename
 1536303468  60  157158   265e6 fs/nfs/nfs.ko

this is with most debugging options on.

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 05:31, Olivier Blin wrote:
> > --
> > arent the following pretty important,
> > and alwasys needed ?
> >

the right question is - are they needed to *boot* kernel?

> > > > -CONFIG_PACKET=y
> > > > +CONFIG_PACKET=m
> > >
> > > module
> > >
> > > >  CONFIG_PACKET_MMAP=y
> > > >  CONFIG_NETLINK_DEV=m
> > > > -CONFIG_UNIX=y
> > > > +CONFIG_UNIX=m
> > >
> > > module
> >
> > --
> > and what about ipv4
>
> As I said in my other reply, I don't think these one should be build as
> modules.
>

I run all of them as modules and do not have any problems. Unless you boot 
over LAN and need them to access root early. That is what I ask again - just 
how many users do it? Can it go into yet another kernel? :) At least standard 
kernel is targeted at home desktop users - am I wrong? 

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 06:56, Olivier Blin wrote:
> >
> > my current patch tarball is uploading
> > linux-test9s5.tar.bz2
> >
> > nosrc.rpm && may be rpm after successfully recompile
> > under the same kernel :-)
> >

please keep just patches for us poor dial-upers as well :)

> > > by the way, why aren't new module-init-tools updated in main ?
> >
> > because the maintainer is hiding somewhere :(
>
> Well, the packager (Chmouel) is gone, he doesn't work at
> MandrakeSoft anymore.

module-init-tools have been maintained by thierry. I do not know about 
hotplug.

> The real maintener is now Andrey, but I guess he hasn't rights to upload
> in main. Who can volunteer ?
>

once more - I do not have cooker so I cannot test it on cooker so I won't 
upload them to cooker. So I hope someone who can build and test them on 
cooker would do it.

> > > I'll submit a (very small) patch for mkinitrd to add splash in 2.6
> > > initrds.
> >
> > could you add cramfs support ?-)
>
> yes, no problem
> Should I make it an option in mkinitrd ?
>

I guess yes. But this is compatibility problem. It is not supported on vanilla 
kernel so you have to detect patched kernel somehow. And how are you going to 
do it? You will get a flood of "bug reports" from users who compiled their 
own kernels (vanilla ones or whatever) and cannot load initrd.

So please make it off by default. I assume we could turn it on in kernel 
post-install script.

Is it really as good? Does it worth hassles?

> >
> > they could be loaded unconditionaly after root is mounted
> > (there might be some warnings, but nothing that could harm the system)
>
> Do you want to try to load *all* mouse and keyboard drivers at boot ?
> That might work, but hey, that's not so smart :)
>

come on it is really just as smart as compiling all of them into kernel. But 
it will keep *boot* kernel size small making it still possible to fit on a 
floppy. 

Or we should officially use some higher density floppy format. Is floppy over 
1.44 possible?

Re: [Cooker] kernel-2.6-test9 config

[Andrey Borzenkov]

On Sunday 16 November 2003 08:15, Olivier Blin wrote:
>
> > PS.
> > if i only could compile
> > # CONFIG_SCSI_AIC7XXX is not set
> > # CONFIG_SCSI_AIC7XXX_OLD is not set
> > # CONFIG_SCSI_AIC79XX is not set
> >
> > :(
>
> It builds fine here :)

it does not build with src dir != build dir. At lease aic7xxx not sure about 
two others, but I disabled all of them as well :)

Re: [Cooker] Re: kernel 2.6 with bootsplash and fixed IDE

[Andrey Borzenkov]

On Friday 14 November 2003 17:32, Svetoslav Slavtchev wrote:
> > > will upload now nosrc.rpm /patches tarball :-)
> >

remove BS02-kobject-oops-fixes.patch, it is already fixed (differently) in 
-test9. should not be in current -mm either

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Wednesday 12 November 2003 04:44, Olivier Blin wrote:
>
> $ diff -u kernel-config-2.6-{,allmod-}i586smp  | egrep -A 3"^-.*=y"
> -CONFIG_M586=y
> +# CONFIG_M586 is not set
>  # CONFIG_M586TSC is not set
>  # CONFIG_M586MMX is not set

OK that needs tweaking anyway.

> --
> -CONFIG_X86_F00F_BUG=y
>  CONFIG_X86_WP_WORKS_OK=y
>  CONFIG_X86_INVLPG=y
>  CONFIG_X86_BSWAP=y

F00F code is really small, there is no reason to disable it.

> --
> -CONFIG_X86_ALIGNMENT_16=y
> +CONFIG_X86_GOOD_APIC=y

I can't find where it is referenced

>  CONFIG_X86_INTEL_USERCOPY=y
> -# CONFIG_HPET_TIMER is not set

anyone has info about how widely it is used? should it be enabled in generic 
kernel?

> --
> -CONFIG_EDD=y
> -# CONFIG_NOHIGHMEM is not set
> -CONFIG_HIGHMEM4G=y
> +CONFIG_EDD=m
> +CONFIG_NOHIGHMEM=y
> +# CONFIG_HIGHMEM4G is not set

make CONFIG_EDD module. It provides just sysfs interface to EDD data, unless 
we have applications that access sysfs we do not need it in kernel. And those 
applications can always modprobe edd first.

> --
> -CONFIG_HIGHMEM=y
> -CONFIG_HIGHPTE=y
> -# CONFIG_MATH_EMULATION is not set
> +CONFIG_MATH_EMULATION=y

OK that is just stupidity of allmodconfig. Remember it blindly tries to enable 
everything ...

>  CONFIG_MTRR=y
> --
> -CONFIG_ACPI_AC=y
> -CONFIG_ACPI_BATTERY=y
> -CONFIG_ACPI_BUTTON=y

No, please! Everything that can be module should be module. Desktop users has 
no need for most of them and those who need can always load them. It would be 
nice if harddrake could detect needed modules and automatically update 
/etc/sysconfig/acpi.

> +CONFIG_ACPI_AC=m
> +CONFIG_ACPI_BATTERY=m
> +CONFIG_ACPI_BUTTON=m
> --
> -CONFIG_APM=y
> +CONFIG_APM=m

not sure. Probably it should be builtin for legacy reasons - else you get "APM 
missing in Mandrake kernel!". Although technically it can just be loaded out 
of initrd I guess. Given that new systems are likely to support ACPI may be 
we should make it module. And make harddrake detect when APM is needed :)

>  CONFIG_APM_IGNORE_USER_SUSPEND=y
>  CONFIG_APM_DO_ENABLE=y
> --
> -CONFIG_CPU_FREQ_TABLE=y
> +CONFIG_CPU_FREQ_GOV_USERSPACE=m
> +CONFIG_CPU_FREQ_24_API=y
> +CONFIG_CPU_FREQ_TABLE=m

I think those need be modules if possible. Anyone has any experience with it?

> --
> -CONFIG_BINFMT_ELF=y
> +CONFIG_BINFMT_ELF=m
>  CONFIG_BINFMT_AOUT=m
>  CONFIG_BINFMT_MISC=m
> --
> -CONFIG_PARPORT=y
> -CONFIG_PARPORT_PC=y
> -CONFIG_PARPORT_PC_CML1=y
> +CONFIG_PARPORT=m
> +CONFIG_PARPORT_PC=m
> +CONFIG_PARPORT_PC_CML1=m

come on, why should it be builtin?

> --
> -CONFIG_BLK_DEV_FD=y

ditto

> +CONFIG_BLK_DEV_FD=m
> +CONFIG_BLK_DEV_PS2=m
>  CONFIG_BLK_DEV_XD=m
> --
> -CONFIG_PARIDE_PARPORT=y
> +CONFIG_PARIDE_PARPORT=m
>

ditto. I do not have paride device nor I believe do 99% of users.

>  #
> --
> -CONFIG_BLK_DEV_RAM=y
> +CONFIG_BLK_DEV_RAM=m

should it be builtin for initrd to work? If not there is no reason to have it 
builtin.

>  CONFIG_BLK_DEV_RAM_SIZE=4096
>  CONFIG_BLK_DEV_INITRD=y
> --
> -CONFIG_IDE=y
> -CONFIG_BLK_DEV_IDE=y
> +CONFIG_IDE=m
> +CONFIG_BLK_DEV_IDE=m
>
> --
> -CONFIG_BLK_DEV_IDEDISK=y
> +CONFIG_BLK_DEV_HD_IDE=y
> +CONFIG_BLK_DEV_IDEDISK=m
>  CONFIG_IDEDISK_MULTI_MODE=y

nope, module does not work. Even if it possible to fix all missing exported 
symbols, IDE layer does not provide for any sort of module reference counting 
(for chipset drivers)  meaning it is simply unsafe. Which implies kernel must 
include all IDE drivers :(

> --
> -CONFIG_SCSI=y
> +CONFIG_SCSI=m

module, please!

>  CONFIG_SCSI_PROC_FS=y
>
> --
> -CONFIG_BLK_DEV_MD=y
> +CONFIG_BLK_DEV_MD=m

same.

>  CONFIG_MD_LINEAR=m
>  CONFIG_MD_RAID0=m
> --
> -CONFIG_PACKET=y
> +CONFIG_PACKET=m

module

>  CONFIG_PACKET_MMAP=y
>  CONFIG_NETLINK_DEV=m
> -CONFIG_UNIX=y
> +CONFIG_UNIX=m

module

>  CONFIG_NET_KEY=m
>  CONFIG_INET=y
> --
> -CONFIG_ATM=y
> -CONFIG_ATM_CLIP=y
> -# CONFIG_ATM_CLIP_NO_ICMP is not set
> +CONFIG_ATM=m
> +CONFIG_ATM_CLIP=m

come on really, should everyone have ATM?

> --
> -CONFIG_INPUT_MOUSEDEV=y
> +CONFIG_INPUT_MOUSEDEV=m
>  CONFIG_INPUT_MOUSEDEV_PSAUX=y
>  CONFIG_INPUT_MOUSEDEV_SCREEN_X=1024
> --
> -CONFIG_SERIO=y
> -CONFIG_SERIO_I8042=y
> +CONFIG_SERIO=m
> +CONFIG_SERIO_I8042=m
>  CONFIG_SERIO_SERPORT=m
> --
> -CONFIG_KEYBOARD_ATKBD=y
> +CONFIG_KEYBOARD_ATKBD=m
>  CONFIG_KEYBOARD_SUNKBD=m
>  CONFIG_KEYBOARD_XTKBD=m
> --
> -CONFIG_MOUSE_PS2=y
> +CONFIG_MOUSE_PS2=m
>  CONFIG_MOUSE_PS2_SYNAPTICS=y
>  CONFIG_MOUSE_SERIAL=m

that is bad. Apparently to have your mouse (and keyboard) autodetected you 
have 
to build it in. OTOH some drivers are still left as module. Meaning - there 
should be some manual configuration. In which case why not configure others 
as well?

There is no problem loading them out of initrd. Of course it needs harddrake 
support finally.

WDYT?

> --
> -CONFIG_SERIAL_8250=y
> -CONFIG_SERIAL_8250_CONSOLE=y
> +CONFIG_SERIAL_8250=m

it needs to be builtin only to have console support ... not sure actually how 
widely it is used. those n

Re: [Cooker] Kernel 2.6 cutover?

["Andrey Borzenkov"]

-Original Message-

> 
> > 
> > 
> > 
> > > Just wondering, what is the rough development plan for Cooker?
> > > Specifically, at what stage will the kernel 2.6 series be the focus and
> > > the 2.4 go to contribs?
> > 
> > IMNSHO this is post-next release task. Actually i would even delay
> > it for release 11 :))
> > 
> > there are just too many problems to solve
> > 
> 
> IMNSHO it should be in as an alternative, with everything tweaked to 
> work perfectly both under 2.4 & 2.6
> 

the question was when should it become the main kernel. Of course
it should be available, even for marketing reasons.

Re: [Cooker] Re: kernel 2.6 with bootsplash and fixed IDE

["Andrey Borzenkov"]

-Original Message-

> 
> > 
> > 
> > > 
> > > What would you like to have in next release ?
> > > 
> > 
> > dxr3 driver :)
> > 
> 
> will upload now nosrc.rpm /patches tarball :-)

URL, please? :)

Re: [Cooker] Kernel 2.6 cutover?

["Andrey Borzenkov"]

> Just wondering, what is the rough development plan for Cooker?
> Specifically, at what stage will the kernel 2.6 series be the focus and
> the 2.4 go to contribs?

IMNSHO this is post-next release task. Actually i would even delay
it for release 11 :))

there are just too many problems to solve

> Is it a stated goal that the 2.6 kernel (not withstanding
> and "critical"
> show stopper) will be the main kernel for Mandrake 10.0 (due roughly
> next March?)

If my voice counts - no.

> The sooner we get kernel 2.6 into the mainstream Cooker - the more
> testing will take place . . . or is it too soon?

agreed :)

[Cooker] Re: kernel 2.6 with bootsplash and fixed IDE

["Andrey Borzenkov"]

> 
> What would you like to have in next release ?
> 

dxr3 driver :)

Re: [Cooker] no DMA in linux-2.6.0-0.test9.5mdk

["Andrey Borzenkov"]

> A patch has been sent, but Alan waits the IDE code to be stable enough.

he retired from active work for a year now so it is unlikely it
is going to be fixed (by him) any time soon.

> 
> But this problem can be avoided if we keep IDE support
> built-in (CONFIG_IDE=y, CONFIG_BLK_DEV_IDE=y).
> So general IDE module won't load, but we can try to load IDE chipset
> modules.
> 

we are likely to have early initialization problem then. I do not
know if it really possible to change chipset driver after it has
already started (even the most generic one). The same for FB support :(

[...]
> Can we assume that the module size is roughly the same size as if
> it was built-in ?

probably. we get the same exported symbols (that are added to main
kernel) and the same code.

modpost adds some aliases so actually it will be smaller builtin.

> I can provide you the kernel config files from Fedora, made by Arjan
> van de Ven, his kernel is only 1.2 MiB IIRC.
>

I would be really interested to see how Arjan packages kernel source.
 
> I've to work on kernel-2.6-source package too, I don't really know how
> I'll do it, perhaps a main kernel-2.6-source package with additionnals
> kernel-2.6-source- binary packages that conflicts between them.
>

I'm still thinking about it.

Vanilla kernel build requires at least ($src refers to source kernel
directory and $build to uidl kernel directory, They can differe, works
just fine. Third directory $module is external module build directory)

- $build/scripts/* be present. They are the same for all flavours
- $build/vmlinux be present
- it writes into $builld/.tmp_versions during build

modversions adds

- it requires *all* $build/**/module.o files
- it writes in $build/.tmp_versions for new modules
- it expects *all* modules be in $build/.tmp_versions. modpost
is run on that list
- moreover, it also requires all $build/**/*.ko files be present!
- it leaves reference to $module under $build/.tmp_versions which
means, it will attempt to access it when you build different module!

all in all it means that
 - you must be root or $build/.tmp_versions must be all-writable
 - you actually have to ship large part of binary modules; and it
   is possible that due to make rules you even have to ship all
   intermediate *.o files :(

I do not see how it is possible to cleanly compile external module
in 2.6. It is completely screwed up.

Re: [Cooker] no DMA in linux-2.6.0-0.test9.5mdk

["Andrey Borzenkov"]

> 
> > I do not remember having said that :) I said that "modular IDE
> > is known to not work" - but the only proof of that statement I have
> > seen is post-halloween document.
> 
> I've read it on lklm in posts from Andrew Morton IIRC, I'll check it and
> give a more detailed answer at lunch.
> 

ok, I would bei interested. I still like the idea of delayed
initialization and loading IDE out of initrd.

> Regards.
> 
> PS: bootsplash will be in my next 2.6 release :)
>

Could you make your patches available separately - at least, as long,
as it is not in CVS? I use standard kernel but I'd like to test it.

Re: [Cooker] no DMA in linux-2.6.0-0.test9.5mdk

["Andrey Borzenkov"]

> Ok, you're using a VIA82CXXX IDE chipset.
> I've set CONFIG_BLK_DEV_VIA82CXXX=m in test9.5mdk, but, as Andrey
> pointed out, ide chipsets won't even load in current 2.6 state.

I do not remember having said that :) I said that "modular IDE
is known to not work" - but the only proof of that statement I have
seen is post-halloween document.

But I readily admit it is true. It is not an IDE job to actually
load chipset drivers - it is a job of PCI hotplug.

But - we get chicken and egg problem here. You have apparently no way
to dynamically load IDE drivers at the kernel probes for disks. Nor
do you have any way to even STATICALLY load them out of initrd because
initrd is initialized too late.

Ideally we want to skip IDE init until initram has been inited and
load correct driver after that. I appreciate if someone could get a
look.

Olivier, thank you for config comparison, it is very useful, I will
send a detailed comment later.

[Cooker] kdm.log grows indefinitely/XFree86 kbd errors

[Andrey Borzenkov]

I was rather surprised to see /var/log/kdm.log of size 6MB:

{pts/0}% LC_ALL=C ll /var/log/*kdm*
-rw-r-1 root adm   6158990 Nov 11 20:24 /var/log/kdm.log
-rw-r-1 root adm 34143 Apr 17  2003 /var/log/mdkkdm.log

note that I stopped using mdkkdm immediately after seeing it first time :) 
else I expect this log be of the same size.

Apparently there is no logrotate configuration for these files. Why?

The log is full of messages:

XFree86 Version 4.3.0
Release Date: 9 May 2003
X Protocol Version 11, Revision 0, Release 6.6
Build Operating System: Linux 2.4.18-23mdksmp i686 [ELF] 
Build Date: 10 September 2003
Before reporting problems, check http://www.XFree86.Org/
to make sure that you have the latest version.
Module Loader present
Markers: (--) probed, (**) from config file, (==) default setting,
 (++) from command line, (!!) notice, (II) informational,
 (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/XFree86.0.log", Time: Tue Nov 11 20:22:45 2003
(==) Using config file: "/etc/X11/XF86Config-4"
Using vt 7
(II) [GLX]: Initializing GLX extension
The XKEYBOARD keymap compiler (xkbcomp) reports:
> Warning:  Multiple interpretations of "Shift_Lock+AnyOf(Shift+Lock)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "NoSymbol+Exactly(Lock)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "Num_Lock+AnyOf(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "Mode_switch+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "NoSymbol+AnyOf(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_1+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_End+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_2+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Down+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_3+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Next+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_4+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Left+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_6+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Right+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_7+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Home+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_8+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Up+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_9+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Prior+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_5+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Begin+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_F2+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Divide+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_F3+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_Multiply+AnyOfOrNone(all)"
>   Using last definition for duplicate fields
> Warning:  Multiple interpretations of "KP_F4+AnyOfOrNone(all)"
> 

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

-Original Message-

> 
> > On Sunday 09 November 2003 14:07, Svetoslav Slavtchev wrote:
> > [...]
> > >
> > > 2.4 with debuging (CONFIG_KSYMS=yes)
> > >  1337954 Aug 25 17:47 /boot/vmlinuz-2.4.22-1mdk
> > >
> > > 2.4 without KSYMS (~ kernel tmb here)
> > >  1035464 Nov  9 04:04 /boot/vmlinuz-2.4.22-21br.1mdk
> > >
> > > 2.6 without debuging
> > >  1532105 Oct 29 19:37 /boot/vmlinuz-2.6.0-test9-bk2
> > >  1560749 Nov  4 04:42 /boot/vmlinuz-2.6.0-test9-bk8
> > >  1555036 Nov  5 13:11 /boot/vmlinuz-2.6.0-test9-bk9
> > >
> > > isn't this a bit special :(
> > >
> > 
> > That is with standard mdk config? Yes it is bad if kernel does not fit on
> > one 
> > floppy.
> 
> if i didn't missed smth, it's a standard 2.4 mdk config
[...]
> in -mm i think there is config option to compile with -Os not -O2
> other alternatives to reduce the size are:
> - to update/fix the bzip2 compression support patches 
> - to drop some of the schedulers, futexes, ... (CONFIG_EMBEDDED)
> but i don't think the second is a good solution :(
>

I tried allmodconfig which resulted in ~1.2MB bzImage - same as 2.4.22-1mdk in your 
case. You have to add ELF support to it (as
it is also module in this case) which apparently gives you bare
minimum. It is likely that some features can be turned off, because
allmodconfig actually tries to enable everything, either as "m" or
as "y" (except CPU types support).

Note that I am not even sure if it boots. I had at least one report
about allmodconfig+devfs that could not mount IDE root but it may be
due to other problems. Although modular IDE is "officially" not
working (cf. post-halloween).

Doing all as modules has practical problem - we must ensure that
at least drivers for your keyboard available. I plan to add PNP
hotplug support that should load drivers for on-board devices
discovered via PNP BIOS but even then it will load serial chip
driver only (i.e. i8042 in our case) - you still need explcitly
load low level keyboard driver (atkbd or xtkbd). That is week point
of current INPUT architecture :(

This could be handled by initrd, sure - always add some standard
driver unless correct one can be determined from user (drak?) config.
Which adds yet another static configuration that can break :(

Brave hearts can experiment with configuration starting from
allmodconfig to see how much can be removed and how mauch has to be
added for it to boot at all :)

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Sunday 09 November 2003 18:51, Svetoslav Slavtchev wrote:
[...]
> >
> > That is with standard mdk config? Yes it is bad if kernel does not fit on
> > one
> > floppy.
>
> if i didn't missed smth, it's a standard 2.4 mdk config
> + some new ide drivers (part of them probably can be built as modules)
> and agpgart by misstake (agp support, not via-agp, ati-agp, ...)
>
> > There was some discussion on LKML recently about kernel size but I do not
> > remember exact subject. Have you asked there?
>
> i know, but there is no solution (? currently? )
>
> in -mm i think there is config option to compile with -Os not -O2
> other alternatives to reduce the size are:
> - to update/fix the bzip2 compression support patches

this requires bot lilo and grub changes or just some bootstrap code in kernel? 

Re: [Cooker] kernel-2.6.0-0.test5.1mdk-1-1mdk

[Andrey Borzenkov]

On Sunday 09 November 2003 18:31, Svetoslav Slavtchev wrote:
[...]
> > > i'm thinking to rebuild them under 9.2 & cooker
> > > and upload them to the usuak address
> > > (may be update to -pre3 if no reason exist)
[...]
> > > PS.
> > > i'm thinking also to upload them to mandrake-club
> >
> > no. I do not want to start alternative packages that must be in
> > mainstream.
> > The only reason I do not upload them is I do not run cooker so I cannot
> > test
> > not build them on cooker so I won't upload packages that may not function
> > on
> > cooker.
> >
> > These packages belong to main and must be in main. If there are reasons
> > to not
> > having them in mainstream - so be it, anyone can get them and do whatever
> > they like.
>
> no, no
> i'm not talking about alternative packages but update packages for
> mandrake 9.2 through mandrake club
>

Ah, sorry, that's fine.

> IMO there couldn't be a reason for not including them in cooker,
> hotplug may be(?hotplug service?), but module-init-tools must go in
>

no one is forced to turn on this service (it can be off by default as far as I 
am concerned). Using coldpug fixed at least one real problem - usb-storage 
not loaded if you boot with USB stick already plugged.

Else hotplug has quite a bit of clean up since old version available in mdk.

> we probably should submit a bug report &/ or contact the maintainer
>

this is Cc to maintainer of module-init-tools. I do not know who maintains 
hotplug currently.

-andrey

[Cooker] 9.2 funny bootplash problem

[Andrey Borzenkov]

When I boot with vanilla 9.2 kernel and otherwise vanilla 9.2 I get strange 
effect. Using ASUS V7100 (GF2MX) on ASUS CUSL2 (i815) motherboard. Boot is 
completed, I get login prompt - but large "9.2" remains in lower right 
corner. If I do 'echo 0 > /proc/splash' - it mostly disappears (but on this 
vt only, switching to another vt displays it again) except for small remnant 
along bottom of the screen (it looks like mode is switched so some part of 
screen simply is not touched anymore).

Lilo:

boot=/dev/hdb5
map=/boot/map
vga=788
default="linux"
keytable=/boot/ru4.klt
prompt
nowarn
timeout=50
message=/boot/message
menu-scheme=wb:bw:wb:bw
image=/boot/vmlinuz-2.4.22-10mdk
label="2422-10"
root=/dev/hdb5
initrd=/boot/initrd-2.4.22-10mdk.img
append="devfs=mount resume=/dev/hdb6"
read-only

/etc/sysconfig/bootsplash

SPLASH=auto
THEME=Mandrake
LOGO_CONSOLE=theme

anything else not touched upon update install.

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Sunday 09 November 2003 14:07, Svetoslav Slavtchev wrote:
[...]
>
> 2.4 with debuging (CONFIG_KSYMS=yes)
>  1337954 Aug 25 17:47 /boot/vmlinuz-2.4.22-1mdk
>
> 2.4 without KSYMS (~ kernel tmb here)
>  1035464 Nov  9 04:04 /boot/vmlinuz-2.4.22-21br.1mdk
>
> 2.6 without debuging
>  1532105 Oct 29 19:37 /boot/vmlinuz-2.6.0-test9-bk2
>  1560749 Nov  4 04:42 /boot/vmlinuz-2.6.0-test9-bk8
>  1555036 Nov  5 13:11 /boot/vmlinuz-2.6.0-test9-bk9
>
> isn't this a bit special :(
>

That is with standard mdk config? Yes it is bad if kernel does not fit on one 
floppy.

There was some discussion on LKML recently about kernel size but I do not 
remember exact subject. Have you asked there?

-andrey

Re: [Cooker] kernel-2.6.0-0.test5.1mdk-1-1mdk

[Andrey Borzenkov]

On Sunday 09 November 2003 14:14, Svetoslav Slavtchev wrote:
> > {pts/1}% rpm -q module-init-tools
> > module-init-tools-0.9.15-0.pre2.0.2bor
>
> any special reason for not using -pre3 ?
>

any special reason to use -pre3? AFAIK there are no significant changes 
against -pre2 and I have nothing to add as well. Frankly speaking I do not 
have time to rebuild for the sake of rebuild only.

> i'm thinking to rebuild them under 9.2 & cooker
> and upload them to the usuak address
> (may be update to -pre3 if no reason exist)
>
> hint hint
> unless you give me permissions to upload
> them to the supermount-ng.sf.net space :-)
>

do you have SF account?

> best,
>
> svetljo
>
> PS.
> i'm thinking also to upload them to mandrake-club

no. I do not want to start alternative packages that must be in mainstream.
The only reason I do not upload them is I do not run cooker so I cannot test 
not build them on cooker so I won't upload packages that may not function on 
cooker.

These packages belong to main and must be in main. If there are reasons to not 
having them in mainstream - so be it, anyone can get them and do whatever 
they like.

Re: [Cooker] kernel 23mdk panic

[Andrey Borzenkov]

On Wednesday 05 November 2003 14:57, [EMAIL PROTECTED] wrote:
> kernel-desktop can have all the preempt stuff and other things (setpcap?)
> you do not want to have on your server.

actually capabilities are exactly for server for all I can say. It allows you 
to run services with reduced privileges thus reducing the possible intrusion 
impact.

BTW one interesting thing in 2.6 is including of SE Linux. Anyone has tried it 
(as separate distribution possibly) and can comment on impact on system 
configuration? It could finally be used to replace/supplement secure kernel.

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Saturday 01 November 2003 21:01, Svetoslav Slavtchev wrote:
> > On Saturday 01 November 2003 06:03, Svetoslav Slavtchev wrote:
> > > > > all FS's built in ? why that ?
> > > >
> > > > Oops, there is no reason to do that, I've now compiled ext3, jbd,
> > > > ReiserFS, jfs and xfs as modules in -2mdk.
> > > > ext2 is still built-in because our initrd image is ext2.
> > > > By the way, should we switch to ROM FS or Minix FS for initrd in
> > > > order to gain some space ?
> > > >
> > > > my -2mdk rpms are available here :
> > > > http://compil.mandrake.org/~blino/kernel-2.6/
> > >
> > > first we've to hack mkinitrd,
> > > and i'm not sure how much will we gain,
> >
> > Well, adding support for alternate FS to mkinitrd is trivial, it is
> > rather
> >
> > administrative consideration - you need utilities for this available i.e.
> > proper Requires: etc.
> >
> > > besides initramfs is always in,
> > > (but it's pretty experimental stuff yet,
> > > and i haven't heard of successfull initrd's using it)
> > > probably better wait for it
> >
> > initramfs as implemented currently must be done at kernel build time
> > which
> >
> > makes it rather useless. It basically appends compressed cpio archives to
> > kernel. Frankly speaking I do not see any advantage in using it - both
> > initrd
> > and initramfs are initialized at the same time.
> >
> > The only remote advantage would be that initramfs does not require chroot
> > as
> > everything is unpacked into (mini-)root and thus using various programs
> > like
> > hotplug or modprobe is easier. I guess one can add support for initramfs
> > in
> > initrd image if needed then.
>
> OK
> so what options do we have to reduce the kernel/ initrd size ?
>

Why exactly do you want to do it? There is nothing special with 2.6 as far as 
I can tell:

{pts/1}% LC_ALL=C ll /boot/initrd-2.4.22-10mdk.img 
/boot/initrd-2.6.0-test9.img
-rw-r--r--1 root root   357177 Oct 25 18:36 
/boot/initrd-2.4.22-10mdk.img
-rw-r--r--1 root root   139356 Nov  4 21:19 
/boot/initrd-2.6.0-test9.img

now when I turned debugging off ...

> > I'd
> >
> > like to enable PCI coldplug but it apparently requires initscripts tuning
> > (at
> > least ALSA, possibly others, USB is not actually an issue with hotplug).
>
> i don't think enabling PCI cold pug is a good idea
> it might try to load the correct fb driver (which might currently be broken
>
> :(
>
> also it'll load the drivers for any hardware found (do you mentioned alsa
> for that reason?)
> but do we need this, we already have modprobe.[conf, preload, devfs?]
>

exactly to remove any need for static configuration. I still dream about 
loading kernel and picking automatically correct drivers without user 
intervention.

Now really. What we have is

- kernel is expected to provide aliases for device numbers, protocol numbers 
etc. This obviates need to have manual configuration as in modutils

- we have hotplug and device IDs for most subsystems. That allows to match 
driver and load it automatically.

What do we have now? We have exactly the same deviceID-to-module but as static 
static table (ldetect-lst) that is used to create static configuration. But 
all of this information is already available so it looks rather redundant ...

Sure there are open issues. Alternative drivers need to be handled (OSS vs 
ALSA is the most obvious example). Possibly broken drivers need to be 
handled. Some sort of "safe mode" with hotplug disabled need to exist. Sure 
there will remain legacy devices that cannot be hotplugged. 

not that I suggest doing all of this in next release. But as remote goal ...

-andrey

Re: [Cooker] kernel-2.6.0-0.test5.1mdk-1-1mdk

[Andrey Borzenkov]

On Sunday 09 November 2003 05:07, Olivier Blin wrote:
> On Sun, 9 Nov 2003 02:50:46 +0100
>
> Olivier Blin <[EMAIL PROTECTED]> wrote:
> > On Sun, 9 Nov 2003 01:23:35 +0100
> >
> > Oden Eriksson <[EMAIL PROTECTED]> wrote:
> > > Another thing. I noticed ivp6 was loaded per default, why is that?
> >
> > Hum, you're right, I don't know why, perhaps a wrong module
> > dependency.
>
> It's loaded by the network init script (/etc/rc.d/init.d/network) :
>
> -
> # If IPv6 is explicitly configured, make sure it's available.
> if [ "$NETWORKING_IPV6" = "yes" ]; then
> alias=`modprobe -c | awk '/^alias net-pf-10 / { print $3 }'`
> if [ "$alias" != "ipv6" -a ! -f /proc/net/if_inet6 ]; then
> case "$(modprobe -V 2>/dev/null)" in
> modprobe* )
> echo "alias net-pf-10 ipv6" >>
> /etc/modules.conf
> ;;
> module-init-tools* )
> echo "alias net-pf-10 ipv6" >>
> /etc/modprobe.conf
> ;;
> esac
> fi
> fi
> -
>
> $ grep ipv6 /lib/modules/`uname -r`/modules.alias
> alias net-pf-10 ipv6
>
> An alias is made to ipv6 in modules configuration files, so the
> network script loads it.
> I don't know if it's the correct behaviour what do you think ?
>
> I'm cc-ing Andrey because he's the author of modprobe configuration
> files, so he's likely to help us :)

I am not sure what the exact question is, so I try to guess.

/lib/modules/`uname -r`/modules.alias is created by depmod based on alias 
information from modules themselves.

{pts/1}% grep MODULE_ALIAS *
af_inet6.c:MODULE_ALIAS_NETPROTO(PF_INET6);

This is to finally replace static 
driver-to-{major,pf-number,ldisc-number,whatever} mapping from old modutils.

The part you quoted is not relevant because it is executed only as long as 
IPV6 is explicitly enabled. Nor is it needed actually as you have seen. Nor 
is it the correct way to handle it because it breaks if module name is 
changed for whatever reason.

So I guess that some program tries to use IPV6 by default which results in 
module being autoloaded.

The part in /etc/init.d/network was needed in modutils because it defaulted to 
IPV6 off:

{pts/1}% modprobe-24 -c | grep net-pf-10
modprobe: QM_MODULES: Function not implemented

alias net-pf-10 off

Till my last release of module-init-tools I had the same in 
/lib/module-init-tools/modprobe.default. The last version removed this line. 
So I guess for module-init-tools you have to explicitly add "alias net-pf-10 
off" if IPV6 is not configured. Note that it may simply be too late because 
protocol can be already registered by some other before /etc/init.d/network 
ever starts.

So what version of module-init-tools do you run? If you run my last version:

{pts/1}% rpm -q module-init-tools
module-init-tools-0.9.15-0.pre2.0.2bor

loading of ipv6 should be "normal" :) Turn it off in modprobe.conf ...

I do not suggest any patch because we need to establish some policy first. To 
recap:

- kernel is expected to finally provide in-band module aliases that replace 
previous static configuration as found in modutils

- for this reason any program or script that assume that net-pf-N or 
block-major-M or whatever is handled by specific module is broken by 
definition (hmm ... I wonder how they are going to handle dynamic device 
numbers when - if - they ever become available ...)

- so we should start with assumption "everything configured in kernel is 
available" and configuration tools should blacklist or turn off those modules 
user does not need (want). Extreme case is disable kmod.

Unfortunately overriding default values in module-init-tools does not work 
cleanly. Specifically modprobe -c output becomes highly confusing in this 
case. It is not a bug it is design decision :(

-andrey

[Cooker] colorgcc not active after 9.1 -> 9.2 update

[Andrey Borzenkov]

I just noticed that while colorgcc has been correctly updated no alternatives 
for gcc it was setup, nor for g++.

this was update from 9.1 + most security updates; a couple of packages were 
custom it unlikely that hotplug or module-init-tools would have affected 
update-alternatives :)

after manually installling alternatives it works just fine.

While not automatically adding alternative may be intentional in this case

- postinstall apparently does it
- I had them active already so I expect to keep them active after update as 
well

-andrey

Re: [Cooker] Bad performance with supermount and usbstick

[Andrey Borzenkov]

On Wednesday 05 November 2003 12:41, Oliver Lemke wrote:
> On Wed, 2003-11-05 at 09:13, Andrey Borzenkov wrote:
> > Slightly offtopic. Do you have problems with "media change detection" on
> > USB stick? I had report (and verified) that when you unplug stick, change
> > contents on otehr system and plug back again supermount does not detect
> > media change and does not remount it. I.e. new contents or changed
> > contents is not visible.
> >
> > This is because sd skips media change check unless device is removable.
>
> Nope, that is working fine here.

hmm ... I wonder if I did something wrong or it has changed recently. Now my 
USB stick appears as removable device but I definitely remember that under 
2.4 (do not rememeber exact version) I tested it and it was never remounted.

> Other problem is when I connect two 
> different usb sticks one after the other.  Plugging in the first one
> creates /dev/{sda,sda1}. After removing it and plugging in another stick
> /dev/{sdb,sdb1} are created and mount point is /mnt/removable2.
> Doing a rmmod usb-storage before inserting the second stick fixes that.

it is not a bug it is by design :) In 2.4 kernel keeps track of USB UUIDs and 
reserves once allocated host number in case you insert it again so it gets 
the same device. So it is not a bug it is a feature :)

2.6 drops it completely in favor of user-level tools that currently do not 
seem to exist (in a form suitable for deployment in real world that is).

-andrey

Re: [Cooker] Bad performance with supermount and usbstick

[Andrey Borzenkov]

On Wednesday 05 November 2003 01:19, Oliver Lemke wrote:
> So, here are the results from my test.
> Exercise was to copy the kernel rpm from media to hd using dd.
>
> First with CD-ROM (Toshiba FireWire DVD-ROM):
> Direct mount:  2327 KB/sec  6%CPU
> Supermount:1321 KB/sec  9%CPU
>
> 43% performance regression
>
> Now with Lexar USB stick:
> Direct mount: 731 KB/sec   2%CPU
> Supermount:24 KB/sec  42%CPU
>
> 97% performance regression and cpu usage jumps up by 40 percent :-(
>
> Test was done with kernel 2.4.22-23mdk-i686-up-4GB. Are there any
> parameters to tweak supermount? I found nothing in the ReadMe on SF.net.

Short answer - mount with
noatime,nodiratime

this should give you the same level of regression as on CD-ROM. It is as much 
as can be done currently

Long answer.

Supermount attempts to keep subfs mounted rw for as short time as possible to 
ensure consistency. When file is open for reading it does not normally 
require subfs be remounted rw - unless we have to update atime. Currently it 
is done on every read request - subfs is remounted rw before actual read is 
executed and remounted back ro after it has completed. And before remounting 
rw supermount attempts to ensure you can't accidentally remove media and 
tries to lock tray.

This would not actually be as bad - but kernel assumes that if device is 
removable it also can lock tray. USB sticks apparently declare themselves as 
removables (mine does for sure). Which means that driver sends *real* command 
to device. Meaning that every time we have to

 - send LOCK command
 - remount RW
   - do read
 - remount RO (implicitly flushing at least one block with changed inode)
 - send UNLOCK command

note that almost all of this is done synchronously thus eliminating any effect 
of kernel read-ahead :(

well it is rather unfortunate trade-off between security and speed. I am open 
for suggestions but I do not see what can be done easily. Even removing 
useless LOCK/UNLOCK commands would speed things up but I do not know if SCSI 
provides any direct information about this capability (apparently not).

As for CD-ROM it is slower simply due to media change checks. They are 
executed synchronously as well so it will be slower. May be this can be 
relaxed. I again need some feedback.

regards

-andrey

[Cooker] supermount vs konqueror or dev.cdrom.autoclose sysctl

["Andrey Borzenkov"]

flepied as owner of initscripts, warly as release planner and others for
comments :)

may I ask once more that Mandrake include dev.cdrom.autoclose=0 as default
in /etc/sysctl.conf. This is the "least surprise" case for most users.

This should be added on update unless already exists too.

while supermount+FAM does avoid the problem it is too easy to break it. Just
dispaly portmapper and you are done. I myself did the same "mistake" as while
back.

And then you get "My CD is snapping on me" syndrom even with new and
shiny supermount :(

-andrey

PS inspired by browsing a.o.l.m ...

Re: [Cooker] 2.6 and building external modules

["Andrey Borzenkov"]

[...]
> >
> > 3. Use separate build tree approach and ship single kernel sources
> that point
> > to different kernel build trees that are distributed as separate packages.
> > (link established by kheader) That saves us overhead of having sources
> > n x times. It almost certainly requires Makefile fiddling but not very
> large.
> > It may fail for some non-conformant Makefiles (but then anything can fail)
> 
> 
> Since the binary kernel packages ship with vmlinux and binary modules,
> can they not be linked into the kernel-source-$flavour package, so that
> you need kernel-source-base and kernel-source-$flavour (which requires
> kernel-$flavour), thus avoiding the need to include two copies of many
> files? kernel-source would be a virtual provides in kernel-source-$flavour.
> 
> This assumes either modpost can use compressed modules, or we ship
> uncompressed modules in kernel binary packages (or some similar solution).
> 
> Of course, if this is what you meant, just ignore me.
> 
> 

no you are right it is exactly what I meant in next paragraph 4.

This requires possibly non-trivial changes in makefiles (at least non-trivial
with my level of experience with gmake). Besides there is quite nasty issue -
how do we know to use kernel binaries and not in-tree modules? it is new
can of worms.

> 
> >
> > 4. Teach modpost to extract module versions from /boot/vmlinuz and
> /lib/modules.
> > Any takers? :)
> >
> > So far two most attractive approaches are
> >
> > - disable modversions. Frankly speaking, it was meant to allow people
> to reuse
> > modules across different kernel versions - as it stands now I have
> never seen
> > that module from one release could be loaded on another. It seems to
> be more
> > trouble than it is worth.
> 
> You don't maintain binary kernel module packages I think? If it does
> work, it would make life easier ...
> 

just that I understand. You think it makes sense to go without modversions or not?
I personally do not like removing it (it gives extra safety); but if it the
only option so far I would better turn it off and let people test kernel. 

> >
> > - take approach 3. We need to teach users how to use 2.6 kbuild
> anyway, there
> > is quite a bit of jump, so we may teach them how to use separate build
> tree
> > at the same time.
> 
> Option 4 looks the best long-term solution (and useful to others too).
> 

if you manage to convince kbuild maintainers to supply makefile part I promise
to add compressed modules support to modpost. It is not as hard (on the verge
of being trivial :)


-andrey

Re: [Cooker] Re: [CHRPM] urpmi-4.4-42mdk

["Andrey Borzenkov"]

> Guillaume Rousse <[EMAIL PROTECTED]> 4.4-42mdk
> 
> - added bash-completion
> - spec cleanup
> - bziped additional sources

hey, if we start to ship completion functions with packages what about
Zsh? 

[Cooker] 2.6 and building external modules

["Andrey Borzenkov"]

[please Cc me on this thread. thank you]


-Original Message-

> 
> On Tue, 04 Nov 2003 17:21:12 +0300
> "Andrey Borzenkov"  <[EMAIL PROTECTED]> wrote:
> 
> > I fails to see these directions on www.minion.de. Quoting README:
> [ snip ]
> > anyway - it works. It works for may people. Actually so far it worked
> > for everyone :)
> 
> Actually, no, there were some problems with the new kbuild  :

ok, I apologize, I meant "with custom built kernel" :(


> http://www.ussg.iu.edu/hypermail/linux/kernel/0307.1/1162.html
> http://www.mail-archive.com/[EMAIL PROTECTED]/msg114289.html
> 
> I could add all this stuff in the %post of kernel-source, what do
> you think of it ?

no enough see below

> But that would require the correct kernel to be installed before
> kernel-source, because we need the correct
> /boot/config-%name-%version-%release to make oldconfig.
> What to do if both normal an -smp kernels are installed ?
> 

To summarize so far.

On 2.4 module versions have been appended to symbol names themselves
and stored in modversions.h file. It allows using sources without object
modules to built external off-tree modules and using some #ifdef magic
evem prepare single source tree that can be used to automatically build
external modules for currently running kernel.

On 2.6 module versions are stored in object file. As last step kbuild
executes modpost that extracts versions (CRCs) from objects that define
them and stores them in objects that use them. I.e. if module mod1.o defines
mod1_foo that is called by mod2.o, the CRC(mod1_foo) is extracted from mod1.o
and stored in mod2.o *object*. Later on this is used by kernel module loader
to check that module matches kernel for which it was built.

This step as implemented currently in kbuild needs *all* modules and vmlinux
for build tree. It is called as modpopst vmlinux mod1 mod2  - enable verbose
and see :(

Which means that to be able to buid external modules kernel build tree must
contain vmlinux and *all* module objects :(((

Which also means that if we want to build for several kernel flavours we need
vmlinux and modules for *every* kernel flavour :(

So what can be done so far.

1. Disable modversions. I have not tested it as yet but I presume that as
this problem exists only in presence of modversions it should allow you to
build modules. IIRC in this case kernel version is still stored so it should
not allow you to load smp module for up kernel. needs checking.

2. Ship separate kernel-source-up etc for every kernel flavour that contains
vmlinux and kernel objects. Use the same kheader trick to link /usr/src/linux
to currently booted kernel sources. This means quite a bit of space OTOH
most users have just one kernel installed ... apparently this is approach
taken by RH (see quoted thread and recent Arjan reply).

3. Use separate build tree approach and ship single kernel sources that point
to different kernel build trees that are distributed as separate packages.
(link established by kheader) That saves us overhead of having sources
n x times. It almost certainly requires Makefile fiddling but not very large.
It may fail for some non-conformant Makefiles (but then anything can fail)

4. Teach modpost to extract module versions from /boot/vmlinuz and /lib/modules.
Any takers? :)

So far two most attractive approaches are

- disable modversions. Frankly speaking, it was meant to allow people to reuse
modules across different kernel versions - as it stands now I have never seen
that module from one release could be loaded on another. It seems to be more
trouble than it is worth. 

- take approach 3. We need to teach users how to use 2.6 kbuild anyway, there
is quite a bit of jump, so we may teach them how to use separate build tree
at the same time.

-andrey

Re: [Cooker] Bad performance with supermount and usbstick

["Andrey Borzenkov"]

-Original Message-

> 
> So, here are the results from my test.
> Exercise was to copy the kernel rpm from media to hd using dd.
> 
> First with CD-ROM (Toshiba FireWire DVD-ROM):
> Direct mount:  2327 KB/sec  6%CPU
> Supermount:1321 KB/sec  9%CPU
> 
> 43% performance regression
> 
> Now with Lexar USB stick:
> Direct mount: 731 KB/sec   2%CPU
> Supermount:24 KB/sec  42%CPU
> 
> 97% performance regression and cpu usage jumps up by 40 percent :-(
> 

Oh :( I think I know what happens but it will be pretty hard to fix at least
while keeping all functionality.

Slightly offtopic. Do you have problems with "media change detection" on USB
stick? I had report (and verified) that when you unplug stick, change contents
on otehr system and plug back again supermount does not detect media change and
does not remount it. I.e. new contents or changed contents is not visible.

This is because sd skips media change check unless device is removable.

-andrey

Re: [Cooker] test9.4mdk with ext3 as module

["Andrey Borzenkov"]

>> most likely reason for this that you have not used 2.6 patch from
>> www.minion.de (that even provides you 2.6-compatible installer).
>>
>> modversions no more exsts in 2.6
> Sorry, no way out with that step.
> I downloaded 'NVIDIA-Linux-x86-1.0-4496-pkg#.tar.gz' and open it, and got:
> two new files, .manifest and *nvidia-installer. Then I used 
> 'NVIDIA-Linux-x86-1.0-4496-pkg2.run' with ./N.run --extract-only.
> Then I changed the .manifest and *nvidia-installer to the new ones.
> When I did ./.nvidia-installer I got:
> ERROR: Unable to build the NVIDIA kernel module.

I fails to see these directions on www.minion.de. Quoting README:

1) download:

   NVIDIA-Linux-x86-1.0-4496-pkg0.run   (nvidia)
   NVIDIA_kernel-1.0-4496-2.6.diff  (minion.de)

2) install the kernel module (as root):

   # sh NVIDIA-Linux-x86-1.0-4496-pkg0.run --extract-only
   # cd NVIDIA-Linux-x86-1.0-4496-pkg0
   # cd usr/src/nv
   # patch -p1 < ../../../../NVIDIA_kernel-1.0-4496-2.6.diff
   # ln -s Makefile.kbuild Makefile
   # make install

anyway - it works. It works for may people. Actually so far it worked for everyone :)

for this reason and because it is extremly hard to answer off web archive
I leave it as home work for interested parties. With extra
link about how to compile external modules under 2.6:

http://lwn.net/Articles/21823/

-andrey

Oh, and readme does not end on quoted lines of course :)

Re: [Cooker] Bad performance with supermount and usbstick

["Andrey Borzenkov"]

Oops, address was wrong first time.

> 
> 
> [please Cc me]
> 
> > I noticed that copying from usb memory stick to hd is very slow when the
> > stick is mounted by supermount. Transfer rate drops to below
> > 100 kB/sec. When mounting it manually, I get the usual 1 MB/sec. Anybody
> > else seeing this?
> 
> well ... supermount checks for media presence every time it needs access
> device and does it synchronously. So it is possible (although order of
> magnitude looks too much :( this may depend on specific driver, i have to
> look.
> 
> may be this can be relaxed for reading off media (as an option). Another
> idea I had is merging several consequitive media check requests assuming that
> user cannot be as fast in removing media :)
> 
> would you mind enter this in tracker on sourceforge? I guess it already has
> similar issue ... I'd call it Request For Enhancement rather than bug :)
> 
> -andrey
> 

Re: [Cooker] test9.4mdk with ext3 as module

["Andrey Borzenkov"]

> This is on a up to date 9.2, with ext3, and the new kernel installed with no 
> changes.

> When trying to use: *NVIDIA-Linux-x86-1.0-4496-pkg2.run, I got:
> Error: The kernel header file 
> '/lib/modules/2.6.0-0.test9.4mdk/build/include/linux/modversions.h' does not 
> exist.
> The most likely reason for this is that the kernel header files in 
> '/lib/modules/2.6.0-0.test9.4mdk/build/include' have not been included.

most likely reason for this that you have not used 2.6 patch from www.minion.de
(that even provides you 2.6-compatible installer).

modversions no more exsts in 2.6

Re: [Cooker] test9.4mdk with ext3 as module

["Andrey Borzenkov"]

>> After about 4 hours it is still "rock and roll!"  :-)  There are a couple
>> issues with bttv loading and then exiting, but a modprobe after X is loaded
>> solves that.  No nVidia driver will compile, but I expected that as a test
[...]
> You can get patches for nvidia for 2.5 / 2.6, actually. I got one of
> them working on test4...dunno if it'll still work when I get a working
> test9...


nvidia has been working for me since 2.5.75 or like ... that makes me wonder
what does not work using mdk rpm.

[Cooker] Users on "do we need 2.6 in Mandrake"

["Andrey Borzenkov"]

Just hit the following on a.o.l.m:

8.2 was one of the most rock-solid Linux distros ever.  Unless you feel
you need some program or functionality you can't get in 8.2, I'd stay
with it instead of upgrading until the 10.0 release with the 2.6 kernel
   
is released in about 6 months.

oh, well ... :)

Re: [Cooker] test9.4mdk with ext3 as module

["Andrey Borzenkov"]

> After about 4 hours it is still "rock and roll!"  :-)  There are a couple
> issues with bttv loading and then exiting, but a modprobe after X is loaded
> solves that.  No nVidia driver will compile, but I expected that as a test
> kernel. 

What exactly is wrong with nVidia? Have you tried patch from www.minion.de?

[Cooker] Re: New mailing list for server topics

["Andrey Borzenkov"]

Warly wrote:
> 
> http://qa.mandrakesoft.com/twiki/bin/view/Main/MailingListHowTo
>

Is archive available somewhere? I cannot find it on mandrake pages.

Will it appear on marc? I think they need explicit request from list
owners to do it? 

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Saturday 01 November 2003 21:01, Svetoslav Slavtchev wrote:
[...]
> OK
> so what options do we have to reduce the kernel/ initrd size ?
>

optimize compilation?

My initrd consists only of reiserfs. Compare:

{pts/2}% l -s /boot/initrd-2.6.0-test9.img
 992 /boot/initrd-2.6.0-test9.img
{pts/2}% l -s /boot/initrd-2.4.23-pre8.img
 340 /boot/initrd-2.4.23-pre8.img

That is not code bloat - code size increased only marginally:

{pts/2}% size linux-2.6.0-test9/fs/reiserfs/reiserfs.ko
   textdata bss dec hex filename
 21829214006644  226336   37420 
linux-2.6.0-test9/fs/reiserfs/reiserfs.ko
{pts/2}% size linux-2.4.23-pre8/fs/reiserfs/reiserfs.o
   textdata bss dec hex filename
 18028710286612  187927   2de17 
linux-2.4.23-pre8/fs/reiserfs/reiserfs.o

but file size increased dramatically:

{pts/2}% ls -s linux-2.6.0-test9/fs/reiserfs/reiserfs.ko
2612 linux-2.6.0-test9/fs/reiserfs/reiserfs.ko
{pts/2}% ls -s linux-2.4.23-pre8/fs/reiserfs/reiserfs.o
 232 linux-2.4.23-pre8/fs/reiserfs/reiserfs.o

take separate files, extreme example is super.o:

 {pts/2}% size linux-2.6.0-test9/fs/reiserfs/super.o
   textdata bss dec hex filename
  15766 448   4   162183f5a linux-2.6.0-test9/fs/reiserfs/super.o
{pts/2}% size linux-2.4.23-pre8/fs/reiserfs/super.o
   textdata bss dec hex filename
  14520 108   0   146283924 linux-2.4.23-pre8/fs/reiserfs/super.o
{pts/2}% ls -s linux-2.6.0-test9/fs/reiserfs/super.o
 180 linux-2.6.0-test9/fs/reiserfs/super.o
{pts/2}% ls -s linux-2.4.23-pre8/fs/reiserfs/super.o
  24 linux-2.4.23-pre8/fs/reiserfs/super.o

apparently it is symbol table size. My knowledge of GCC internals is not as 
good as it was ten years ago so I do not know why it happens and how to 
change it. 

> bzip compression? (last time i loocked the patches were against 2.5.59 :( )
> compile with Os ?
>

it is not object size :) You do not need to optimize code, you simply has to 
remove all this (redundant? superfluous?) symbol table entries.

Well, I have all debugging options enabled it may have something to do with it 
of course.

-andrey

Re: [Cooker] Congratulations, thanks and coupla comments :)

[Andrey Borzenkov]

On Wednesday 29 October 2003 18:13, Thierry Vignaud wrote:
[...]
> >
> > > the fix is probably to go for using WrappedLabel instead of Label
> > > and feed them with formatAlaTeX() everywhere for better layout.
> >
> > ehh ... could you translate it for us mere mortals? :)
>
> some labels are filled with sized texts and are not marked as
> wrapp-able.
> we should use wrapped labels were needed.
>
> also, some labels look bad with some translations due to wrapping
> issues vs manual return caracters.
>
> so i suggest to separate paragraphs by two lines like in LaTeX and let
> formatAlaTeX() from MDK::Common do the jobs for us.
>
> filling with paragraphs + auto-wrapping should clear most l10n visual
> bugs.

does it require modifying of translation?

[Cooker] syslog-ng as default?

[Andrey Borzenkov]

If somebody still remembers (Oden?) we spoke about transition to syslog-ng. I 
am using it for at least a year now without any apparent ill effects; the 
only open question was klogd Oops decoding and most responses I got from LKML 
indicated that klogd is obsolete; now-a-days we have internal symbol decoding 
in both 2.4 (mdk extra) and 2.6 (built in by default) so it is not a problem 
anymore.

The transition is not as easy, configuration file is not compatible so any 
tool messing with it must be updated. Immediately I can think about msec and 
MNF (it has options for logging configuration); webmin is likely to be 
affected too. Anything else?

so is it worth it? If yes it must be done as soon as possible to leave enough 
time for testing.

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Friday 31 October 2003 13:46, Svetoslav Slavtchev wrote:
[...]
> > >
> > > actually on second look it's a bit later,
> > > the last line is :
> > > "Freeing unused kernel memory:272k freed"
[...]
> > Do you mean that using the same setup test9 stops logging to console and
> > previous kernels not?
>
> i haven't tried for a long time
>
> it's probably a user error :(
> if i stop syslogd serial console comes back,
> so it's either missconfiguration,
> or some problem between your initscripts package mdk9.1 sysklogd
>

actually rc.sysinit resets console log level to 3 (ERR) by default which means 
that only critical errors are logged. Check /etc/sysconfig/init. So there is 
nothing strange you do not see anything after that point. The reason it works 
after syslog is stopped should be that _something_ resets console log level 
to higher value. Quick strace on syslog/klogd do not show anything but I am 
not expert here. Nor do I use syslog - I use syslog-ng.

Just try dmesg -n7 and check if you now get output.

It would be nice to have different levels for each console OTOH it may not 
worth the troubles.

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Saturday 01 November 2003 06:03, Svetoslav Slavtchev wrote:
> > > all FS's built in ? why that ?
> >
> > Oops, there is no reason to do that, I've now compiled ext3, jbd,
> > ReiserFS, jfs and xfs as modules in -2mdk.
> > ext2 is still built-in because our initrd image is ext2.
> > By the way, should we switch to ROM FS or Minix FS for initrd in order
> > to gain some space ?
> >
> > my -2mdk rpms are available here :
> > http://compil.mandrake.org/~blino/kernel-2.6/
>
> first we've to hack mkinitrd,
> and i'm not sure how much will we gain,
>

Well, adding support for alternate FS to mkinitrd is trivial, it is rather 
administrative consideration - you need utilities for this available i.e. 
proper Requires: etc.

> besides initramfs is always in,
> (but it's pretty experimental stuff yet,
> and i haven't heard of successfull initrd's using it)
> probably better wait for it
>

initramfs as implemented currently must be done at kernel build time which 
makes it rather useless. It basically appends compressed cpio archives to 
kernel. Frankly speaking I do not see any advantage in using it - both initrd 
and initramfs are initialized at the same time.

The only remote advantage would be that initramfs does not require chroot as 
everything is unpacked into (mini-)root and thus using various programs like 
hotplug or modprobe is easier. I guess one can add support for initramfs in 
initrd image if needed then.

> me thinks, better get Andrey's module-init-tools
> with gziiped module support & compress the modules
>

But initrd is already built compressed it does not require module-init-tools 
support (the whole image is compressed).

I updated both on usual place (http://supermount-ng.sf.net/mdk-25/). hotplug 
is updated to CVS code, quite a lot of patches are already integrated. I'd 
like to enable PCI coldplug but it apparently requires initscripts tuning (at 
least ALSA, possibly others, USB is not actually an issue with hotplug).

module-init-tools now depends on kernel version - I removed all static aliases 
for which MODULE_ALIAS is defined based on test9. Future versions may add 
more aliases. Ideally modrobe.default should be near empty. This means 
module-init-tools should contain Conflict: kernel-2.6 < 2.6.0-test9 but as I 
am not using kernel RPM I am not sure about proper versioning. Svetoslav, 
would you get a look?

Thierry, I really think both need to be in cooker to enable proper testing.

-andrey

[Cooker] Supermount FAQ

[Andrey Borzenkov]

I compiled a small supermount FAQ, link is 
. 
Home page also has link to DocBook source. Comments, submissions and 
corrections are welcome. This includes DocBook usage and style (it is my 
first document ever, I opted for DocBook because it apparently provides 
generation of both HTML and text, I want to maintain main readme in the same 
format).

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

[Andrey Borzenkov]

On Saturday 01 November 2003 03:34, Olivier Blin wrote:
> I really need to browse all kernel config options, the initial config
> was done by Nanar months ago :)
>

my config for test9 attached. It was initially based on standard mdk config. 
With this I can boot mdk 2.4 and vanilla 2.4 and vanilla 2.6 on the same 
system (I do use my hotplug so for others some things may break).

This is SMP (I need it to test spinlocks) and has all debugging options turned 
on; also it removes all drivers I do not need and makes more of others 
modules.

You can take it as reference. It is known to work :)

-andrey

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

> > 
> > > PS.
> > > is serial console working for everyone ?
> > > i just tried it and when fbcon took over vgacon,
> > > it stopped logging to ttyS0
> > > is this normal behaviour ?
> > >
> > 
> > apparently not. I guess James Simmons may be interested :) BTW have you
> > tried
> 
> actually on second look it's a bit later,
> the last line is :
> "Freeing unused kernel memory:272k freed"
> 
> could it be related to the initrd.c removal ?
>

unlikely. they simply replaced access to initrd image via block device
by direct copying of the initrd memory area into file on rootfs. 

Do you mean that using the same setup test9 stops logging to console and
previous kernels not?
 
> > his patch for framebuffer posted recently?
> > 
> > http://marc.theaimsgroup.com/?t=10668441672&r=1&w=2&n=9
> 
> the last one i didn't, i've tried several prevous version & fbdev bk's
> but no luck
> 

oh well. I guess he can be interested in this too :)

-andrey
[...]
> > seriously it really needs be fixed, it does not make debugging easier.
> 
> i'll probably try to revert the initrd changes
>
> 

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

> 
> > > well vanilla test9 has brocken rlogin/rsh
> > 
> > Do you seriously mean anyone is using them? Or does it apply to ssh as
> > well?
> > 
> 
> well,
> 
> it was reported by 3-4 people on lkml :-)
> 

Ah, if you mean that, it was not rsh. The OOB data delivery was broken
and this is used by rsh to transmit e.g. terminal settings change. This
is likely to affect ssh as well, yes.

> besides Linus is currently accepting only _real_bug_fixes_
> and i have the impression that somehow -bk's are more stable
> then real releases :-)
> 

that is simply due to the fact the real releases are tested by real people
while -bk are tested by developers only. So any bug introduced in -bk is much
less likely to be noticed till next release.

> PS.
> is serial console working for everyone ?
> i just tried it and when fbcon took over vgacon,
> it stopped logging to ttyS0
> is this normal behaviour ?
>

apparently not. I guess James Simmons may be interested :) BTW have you tried
his patch for framebuffer posted recently?

http://marc.theaimsgroup.com/?t=10668441672&r=1&w=2&n=9

seriously it really needs be fixed, it does not make debugging easier.

-andrey
 
> BOOT_IMAGE=260t9serC ro root=2208 devfs=mount elevator=deadline
> console=ttyS0,38400 console=tty0
> 
> -- 
> NEU FэR ALLE - GMX MediaCenter - fЭr Fotos, Musik, Dateien...
> Fotoalbum, File Sharing, MMS, Multimedia-Gruъ, GMX FotoService
> 
> Jetzt kostenlos anmelden unter http://www.gmx.net
> 
> +++ GMX - die erste Adresse fЭr Mail, Message, More! +++
> 
> 

Re: [Cooker] Why doesn't kde require X

["Andrey Borzenkov"]

> I was setting up a chroot environment tonight and I did 
>
> 'urpmi --root /mnt/newroot kmail'
>
> and XFree86 did not get installed.  Is this normal or correct?

you need XFree86-libs to run X11 programs and XFree86 to run them on
local display. You may as well want to run them on remote display only
(consider system without graphic card).

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

> well vanilla test9 has brocken rlogin/rsh

Do you seriously mean anyone is using them? Or does it apply to ssh as well?

Re: [Cooker] Do not renice X by default

["Andrey Borzenkov"]

>> names) or you're not mounting the new usbfs. Put this into your /etc/fstab
>> :
>>
>> none /proc/bus/usb usbfs defaults 0 0

> I tried and still my scanner wouldn't work. The weird thing is that it would 
> appear in the /proc/bus/usb/devices and in usbview.

Do you mean it works under 2.4?

Coldplugging hot-pluggable devices does not work and should not work under
2.4, at least it fails for usb-storage. I think, scanner driver is loaded
statically out of /etc/init.d/usb depending on /etc/sysconfig/usb settings -
could you check if it really works? Chances are driver name has changed
between 2.4 and 2.6 so this loading under fixed name fails, please compare
drivers for 2.4 and 2.6.

Svetljo has fixed hotplug RPM that supports 2.6, if you enable its initscript
it will emulate hotplugging for all devices present at boot and will load
drivers automatically, no manual config is needed.


-andrey

[Cooker] Re[2]: 9.2 - manpages in Russian completely screwed up

["Andrey Borzenkov"]

-Original Message-

> 
> Andrey Borzenkov <[EMAIL PROTECTED]> writes:
> 
> > please, fix. 
> 
> i'll
> 
> 

thank you.

i just realized that koi8-r patch for groff is disabled. any reason?

Re: [Cooker] Congratulations, thanks and coupla comments :)

["Andrey Borzenkov"]

[thierry, would you mind Cc to me directly?]

>> - there are problems in some MCC tools when embedded, they do not
>> fit in parent window (apparently due to different size of translated
>> strings) but no scrollbars are available so some controls (buttons
>> et al) cannot be reached.  More or less the same as always, it
>> happens in every release. I can send screenshots if anyone is
>> interested.

> i would be more interested in patches against share/po/ru.po :-)

sure but as I said tranlsations look mostly right. Russian does have longer
words than English in average so it is generic problem that must be addressed - 
you cannot assume fixed size dialog that suits everybody. Is it possible to compute
parent window size based on child size? This probably needs to be done once
on init for max size as otherweise window starts "flickering".

> the fix is probably to go for using WrappedLabel instead of Label and
> feed them with formatAlaTeX() everywhere for better layout.

ehh ... could you translate it for us mere mortals? :)

-andrey

Re: [Cooker] bootsplash kernel progress bar on non-mdk kernel

["Andrey Borzenkov"]

-Original Message-

> 
> On Wed, 29 Oct 2003, Buchan Milne wrote:
> 
> > -BEGIN PGP SIGNED MESSAGE-
> > Hash: SHA1
> > 
> > Andrey Borzenkov wrote:
> > > Apparently this progress bar is displayed not by kernel but by some
> > external program?
> > 
> > No, it seems just to work by placing "show $num" in /proc/splash where
> > $num is a 16bit integer (see /sbin/splash.sh and the rc_splash function
> > in /etc/rc.d/).
> > 

sorry, but I can't have /proc/splash on a kernel without bootsplash patch. Still
I have progress bar.

Oh never mind, that's definitely not the first priority thing.

just realized it could well be LILO progress bar as it appears before kernel output.
Does it have one? In this case it smells like a bug.

> > > On 2.4.23-pre8 it is too short while on 2.6 it extends over the right
> > side - apparently
> > > it depends on relative kernel version number :)
> > 
> > Does framebuffer work on 2.6?

yes. As has been mentioned on lkml many times - you can't compile several
FB drivers in your kernel. Select one and it will work. It has always been working
here since I started using 2.5. I guess I am useing vesa framebuffer driver, have
to check @home. For other drivers YMMV as usual.

Sure it is PITA for distributions kernels. I hoped new roofs/initramfs would allow
dynamic loading of framebuffer drivers during boot but apparently it is initialized too
late. I vaguelly remember someone mentioned on lkml possibility to load fb driver
after system boots - i.e. out of initrd but I never had time to recompile kernel just
to test it. Besides it is a bit too late, you still have some time window with plain
text output.
 
-andrey

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

> > any offset. Also I had very nasty problem with test8 - no traffic ever
> > flowed over dialup
> > PPP connection. test9 fixed it whatever it was.
> 
> i think it was fixed in test8-bk1 or bk2
> (i had the same problem on a lan conection, but only with http, ssh & ftp
> worked)
> 

good, I was curious it it was just an accident. gone in test9

> > Oh and test8 does not boot
> > with inintrd and
> > devfs :)
> 
> also fixed in bk1 i think,

fixed in test9 anyway.
[...]
> 
> PS.
> talking about test9?
> aren't you scared by the posts on lkml ?
> it might be that there a lot new testers,
> but i never seen that much oopses & fail
> reports with earlier test kernels
>

not much more than usual :) I am afraid 2.6 final will not be in good shape when
it appears (and rumors are it should be RSN).
 
> i would rather wait for a later bk snapshot
> 

if the question is what to release I'd rather used test9 given two mentioned problems.

-andrey

[Cooker] bootsplash kernel progress bar on non-mdk kernel

["Andrey Borzenkov"]

Apparently this progress bar is displayed not by kernel but by some external program?
On 2.4.23-pre8 it is too short while on 2.6 it extends over the right side - apparently
it depends on relative kernel version number :)

seriously, how it is computed? Oh, and anyone knows any attepmt to port bootsplash to
2.6? Even incomplete one - it could be taken for a start

TIA

-andrey

please Cc me I do not receive cooker

Re: [Cooker] kernel-2.6-test8 rpms

["Andrey Borzenkov"]

> I've built rpms of kernel-2.6-test8 (thanks to Andrey who released
> supermount for -test8).

You may just as well jump to test9. I built against test8 only because test9 did not
propagated to locale kernel.org mirror last weekend. patch applies to it even without
any offset. Also I had very nasty problem with test8 - no traffic ever flowed over 
dialup
PPP connection. test9 fixed it whatever it was. Oh and test8 does not boot with 
inintrd and
devfs :)

-andrey


Военные пытаются провести закон об отмене отсрочек от армии
для студентов. Союз Правых Сил убежден, что подобные действия
недопустимы. Нам важно ваше мнение - голосуйте на
http://r.mail.ru/cln2135/mail.ru

[Cooker] Re: Fwd: Re: Supermount / Zip100 bug

["Andrey Borzenkov"]

-Original Message-

> 
> >>>>> "andrey" == Andrey Borzenkov <[EMAIL PROTECTED]> writes:
> 
> Hi
> 
> andrey> As I do not have hardware nor sufficient knowledge about it I
> andrey> hardly can fix it. It appears ide-floppy does not implement
> andrey> media change detection at all - the only thing it does is fake
> andrey> media change on driver ->open. It means supermount never
> andrey> detects any disk change of course.
> 
> Ide-floppy is quite bad, it is easier (for some things) to use
> ide-scsi for zips.  The things that I remembered were that:
> - ide-floppy just locks the drive on open, no way to change that.

is it really the case even with my patches? I added the same logic as in other
drivers for managing drive locked state and so far nobody complained.

> - ide-scsi had other problems that I forgot :(
> 

my concern is mostly 2.6 as in 2.4 you can use ide-scsi; but 2.6 ide-scsi was
broken last time I checked.

> Will take a look.
> 

if you have a patch drop me a note; I'd like to put it (or link to it) on
supermount page, the problem appears to be fairly common.

-andrey

[Cooker] Re: [PATCH][9.2 initscripts] redundant code in /etc/profile.d/10lang.sh

[Andrey Borzenkov]

oops forgot to attach the patch.

On Tuesday 28 October 2003 20:14, Andrey Borzenkov wrote:
> Any reason for duplicating code from mandrake_consmap except because it
> hapens during RH merge :)?
>
> cheers
>
> -andrey
--- /tmp/etc/profile.d/10lang.sh	2003-09-24 18:30:11.0 +0400
+++ /etc/profile.d/10lang.sh	2003-10-28 20:12:03.637746456 +0300
@@ -68,7 +68,6 @@
 [ -n "$_XKB_CHARSET" ] && export _XKB_CHARSET || unset _XKB_CHARSET
 
 [ -f /etc/init.d/mandrake_consmap ] && . /etc/init.d/mandrake_consmap
-
 unset SYSFONTACM SYSFONT
 
 # handling of special cases where localization is done
@@ -82,32 +81,6 @@
 	[ "$CONSOLE_NOT_LOCALIZED" = "yes" ] && LANGUAGE=C
 fi
 
-if [ -n "$CHARSET" ]; then
-	case $CHARSET in
-	8859-1|8859-2|8859-5|8859-15|koi*)
-if [ "$TERM" = "linux" -a "`/sbin/consoletype`" = "vt" ]; then
-   echo -n -e '\033(K' 2>/dev/null > /proc/$$/fd/0
-fi
-;;
-esac
-elif [ -n "$SYSFONTACM" ]; then
-	case $SYSFONTACM in
-	iso01*|iso02*|iso05*|iso15*|koi*|latin2-ucw*)
-		if [ "$TERM" = "linux" -a "`/sbin/consoletype`" = "vt" ]; then
-			echo -n -e '\033(K' 2>/dev/null > /proc/$$/fd/0
-		fi
-		;;
-	esac
-fi
-if [ -n "$LANG" ]; then
-  case $LANG in
-	*.utf8*|*.UTF-8*)
-	if [ "$TERM" = "linux" -a "`/sbin/consoletype`" = "vt" ]; then
-		unicode_start $SYSFONT $SYSFONTACM
-fi
-	;;
-  esac
-fi
 [ -n "$LANGUAGE" ] && export LANGUAGE
 
 fi

[Cooker] 9.2 - manpages in Russian completely screwed up

[Andrey Borzenkov]

reminder - I have ru_RU.KOI8-R locale

Some man pages are in KOI8. Examples are dd, ls, detectloader etc. These pages 
apparently get high bit stripped (due to the fact that nroff sets -Tascii in 
this case) resulting in excellent and useful output like

ioeoaiea
   oOICOAIIA ls  OIAbAIA  xUxIAEO  ODEOIE  xOAE  AEAEIIx  (IA  EAOAIICIx),
   DAOAbEOIAIIUE  x EIIAIAIIE OOOIEA, A UAOAI xUxIAEO ODEOIE xOAE AEAEIIx,
   IAEIANYEEON x EAOAIICAE, DAOAbEOIAIIUE x  EIIAIAIIE  OOOIEA.   aOIE  IA
   OEAUAII  IE  IAIICI  AEAEIA, OI DI OIIIbAIEA AOCOIAIOII IAUIAbAAOON `.'
   (OAEOYEE   EAOAIIC).iDAEN   -dUAOOAxINAOlsIAObEOAOO
   AOCOIAIOU-EAOAIICE   EAOAIICAIE.   aOAOO  IOIAOAOAOOON  OIIOEI  AEAEIU,
   EIOIOUA IA IAbEIAAOON O `.' EIE xOA AEAEIU, AOIE UAAAIA IDAEN -a.

I can manage to view those by using

LESSCHARSET=latin1 man ...

but it may have more glitches that I did not notice as yet.

RPM manpages apparently are in UTF8. These cannot be viewed at all whatever 
LESSCHARSET I try. The only way is to force English that is not that simple 
given current manpage structure:

LANG=C LANGUAGE=C LC_MESSAGES=C man rpm

please, fix. 

-andrey

[Cooker] Fwd: Re: Supermount / Zip100 bug

[Andrey Borzenkov]

As I do not have hardware nor sufficient knowledge about it I hardly can fix 
it. It appears ide-floppy does not implement media change detection at all - 
the only thing it does is fake media change on driver ->open. It means 
supermount never detects any disk change of course.

Anyone could get a look? The right thing is probably add sense code decoding 
to idefloppy_analyze_error and use some innocent command like TEST_UNIT_READY 
in idefloppy_media_change.

what our kernel gurus on cooker say? :)

-andrey

--  Forwarded Message  --

Subject: Re: Supermount / Zip100 bug
Date: Monday 27 October 2003 23:19
From: Andrey Borzenkov <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Cc: Paul Bristow <[EMAIL PROTECTED]>

On Monday 27 October 2003 06:21, you wrote:
> Andrey,
> I downloaded your supermount-ng and successfully
> patched it into kernel 2.4.22. While testing it, I
> found out almost the same problem I used to have when
> I had mandrake in my machine: when I eject a disk, the
> device is not automatically unmounted. Then, when I
> insert a new disk and cd the /mnt/zip dir, the old
> disk tree is still mounted. So, to access another
> disk, I must manually umount and mount it, which would
> make supermount useless.
> Maybe I commited some configuration mistake, or there
> is a bug, but I hope you can help me with this.

Hmm ... now when I got a look it appears that ide-floppy really does not
dynamically check for media presence. The only case when MEDIA_CHANGED flag
is set is on device open and even in this case it is blindly set without ever
consulting real device. So we just assume on every open that media has been
changed.

Paul, is it possible to fix it? I assume it amounts to calling something like
request sense or possibly idefloppy_get_capacity() out of
idefloppy_media_change() except that idefloppy_get_capacity() should then set
IDEFLOPPY_MEDIA_CHANGED if this is the case.

This is the same for both 2.4 and 2.6 trees. Both need fixing to really
support supermount.

Rodrigo, till it is fixed for 2.4 you can use ide-scsi. For 2.6 it was badly
broken last time I checked, do not know exact status now. Hopefully Paul
fixes ide-floppy ...

> My scsi-cdwriter works really fine with supermount.
> I'd say you are doing a fine job.

good. Just use ide-scsi for Zip as well.

-andrey

---

[Cooker] Congratulations, thanks and coupla comments :)

[Andrey Borzenkov]

Hi everybody,

hopefully I am back online now. My congratulations with 9.2 (sorry if it comes 
late) and best wishes. As we all hope - it should be successful.

I did not look into mailbox for a month - so much was surprise when I found 
mail about mandrake club ... thank you, I really appreciate it (I guess I 
have to express my thanks to fred lepied in the first place) ... 

Update to 9.2 went mostly smoothly, some glitches below (Pablo, I Cc to you as 
they mostly are i18n issues - surprise surprise :) I updated from 9.1 with 
most updates.

- initscripts l10n glitch, it becomes routine job :) See separate mail for 
patch

- update again screwed up my system by changing locale to UTF8. Because it was 
already discussed I am forced to use emphasis here - DO NOT CHANGE SYSTEM 
SETTINGS ON UPDATE. That is even more annoying as I explicitly deselected 
"use UTF8" during installer run ...

- printerdrake assumes ISO-8859-1 when reading system files. The effect - it 
creates printer with name "Printer" in Russian and writes it into CUPS file. 
Then it reads it but displays using Western charset. With obvious effect. 
This concerns ALL tools. Any tool should assume that local files have 
encoding determined by current locale. It was discussed with gc already re. 
rpmdrake (.rpmnew merge).

- for some reason kdeutils was not updated. I suspect it is due to new package 
being virtual and some weird dependency resolution problems.

- kdepim was removed but replacement packages have not been installed (with 
effect that address book call from kmail did not work)

- there are problems in some MCC tools when embedded, they do not fit in 
parent window (apparently due to different size of translated strings) but no 
scrollbars are available so some controls (buttons et al) cannot be reached. 
More or less the same as always, it happens in every release. I can send 
screenshots if anyone is interested.

Apart from that funny thing is that KDE does not start non-KDE programs in 
saved session. I had saved session with two konsole and gkrellm - gkrellm 
never appears. Even when I saved session again. I just start it out of 
Autostart now (fortunately new version supports all WM hints that eliminates 
need to use kstart or other dirty tricks).

Of course lack of kernel sources was very bad surprise ... I suspect that most 
people who use downloaded version would miss it actually, to me it makes more 
sense in some stripped down boxed version. Which means I did not even run 
mandrake kernel yet because I have nvidia and cannot compile drivers :) Not a 
big deal as I run 2.6 anyway but still I was curious.

Some sugar.

The overall look is really more polished and professional. MCC looks very 
nice, do I understand right that draksec now takes care of grsecurity 
settings? Some options smell like it. Translations were also mostly good so 
far.

2all - what do you think of translating contributors list? I mean, using 
native names in native locale. I do not know how many non-ISO8859-1 
contributors are there though ...

[Cooker] [PATCH][9.2 initscripts] redundant code in /etc/profile.d/10lang.sh

[Andrey Borzenkov]

Any reason for duplicating code from mandrake_consmap except because it hapens 
during RH merge :)?

cheers

-andrey

[Cooker] [PATCH][9.2 initscripts] fix l10n during shutdown

[Andrey Borzenkov]

Apparently breaking this particular case became favorite game. It has been 
broken and fixed already and may be even more than once.

When you shutdown without logging in on vt1 first console is not switched to 
correct charset. So all nice Russian output is garbled.

Please consider for next initscripts update.

regards

-andrey
--- /etc/rc.d/rc.mandrake_consmap	2003-09-24 18:30:15.0 +0400
+++ /etc/rc.d/rc	2003-10-28 19:58:52.0 +0300
@@ -67,6 +67,11 @@ if [ "$previous" = "N" ]; then
 elif [[ "$newrunlevel" = "0" || "$newrunlevel" = "6" ]]; then
 chvt 1
 exec &> /dev/console
+# Make sure terminal is switched to correct charset.
+# It has to be after chvt of course. For startup case it really
+# does not matter as tty is not changed between rc.sysinit and rc
+# DO NOT REMOVE THIS. Please :)
+[ -f /etc/init.d/mandrake_consmap ] && . /etc/init.d/mandrake_consmap
 fi
 
 # Tell linuxconf what runlevel we are in

[Cooker] New lilo boot image too dark and sad ...

["Andrey Borzenkov"]

I admit it does look more professional than before. But the very first
thing that came in mind as I have seen it first time was funeral.

sorry :)

-andrey

[Cooker] Re: cciss (compaq smart array) devfs support

[Andrey Borzenkov]

On Monday 01 September 2003 17:45, Thierry Vignaud wrote:
> if i remember correcly, current driver works smoothly with cciss
> driver.
> i do not remember what is the exact naming scheme, but it should be
> something like /dev/cciss//{disk,partX)
>

I am interested in exactly 

> so there *should* be no problems.

sure, but people reported that cciss driver in 2.6 does not create any devfs 
entries. Which means it is a Mandrake specific patch.

Could somebody (Thierry?) get a look and show me a sample of canonical devfs 
names for cciss driver? I be damned if I understand how it does it.

[Cooker] cciss (compaq smart array) devfs support

[Andrey Borzenkov]

Anyone knows about it? Reading devfs it appears that just a single drive is 
suppoted (devfsd expects /dev/cciss/{disc,partN}) while driver apparently 
supports more than one (8 if I am not wrong).

But in kernel sources for 2.4.21-0.25mdk there is no trace of cciss driver 
ever creating ANY devfs entry at all.

The question came from Andrew Morton re. devfs support for block devices in 
2.6. For all I can tell the correct structure should be at least

/dev/cciss/hostN/targetM... for controller N; may be full bus and lun are 
overhead, but I do not know this hardware so cannot decide. In any case 
current devfsd patch looks bogus.

-andrey

[Cooker] Re: things to improve now we finally have supermount-ng

[Andrey Borzenkov]

On Saturday 23 August 2003 23:35, [EMAIL PROTECTED] wrote:
[...]
> Use ext2:vfat for floppies and Jaz/ZIP/ide-flopy;
> udf:iso9660 for DVD; iso9660 for CD. This covers most common cases.
>
> [Andrey: I know iso9660 can mount some udf as well, but if it is a real
> cdrom drive, it can only be "packet writing" cds, is it a problem when
> these are mounted with iso9660?]
>

I am not sure I actually understand the question ... I am no expert in CD 
writing after all :)
[...]
> Another thing is that some drives are _very_ slow when trying to autoclose
> the device (see bug 20, there are probably more, since I often saw reports
> about slow cdrom responses). I would prefer that autoclose is
> disabled, or at least that it is made custumizable (sysctl -w
> dev.cdrom.autoclose=0).
>

it is customizable :) the question is what should be default upon install. 
Sometimes I miss autoclose  but you can't have 'em all.

> Andrey, another thing we talked about was sysctl -w
> dev.cdrom.check_media=1. But I currently forgot what it was good for :(
>

mostly to speed up things and to not interfere with CD players. Without this 
setting attempt to mount will try to really mount (incl. media access) that 
may result in skips during playing audio CD. With it attempt to mount does 
not even touch CD because it already knows it is audio and returns EMEDIATYPE 
error.

-andrey

Re: [Cooker] 2.6-tips and docs

["Andrey Borzenkov"]

I meant my RPMs that I uploaded to usual place; I updated them a bit
since then.

please get a look.

-Original Message-

> 
> Andrey Borzenkov <[EMAIL PROTECTED]> writes:
> 
> > > (should module-init-tools not BuildRequire glibc-static-devel ?)
> > 
> > yes it should. But because I build it on 9.1 and 9.1 does not have
> > glibc-static-devel I was a bit selfish and put something that
> > applies to both.
> > 
> > I guess doing urpmi module-init-tools-XXX.src.rpm should pull in
> > that as well, but I may be wrong.
> > 
> > Thierry, please put in correct requires for official package. I
> > guess I can live with rpm -b --nodeps
> 
> already fixed since Jul 24 2003, it seems
> 
> 

Re: [Cooker] LVM2 in cooker & (sort of)rc.sysinit lvm2 fix ?

["Andrey Borzenkov"]

> there was a thread horrible rc.sysinit a while ago,
> but it won't change for 9.2 (and someone mentioned it would break the
> compatibility to redhat, i don't know we need it thought)

IMHO current rc.sysinit is so patched that nothing will be broken
by any change.

Re: resend: Re: [Cooker] APM and time reset after suspend

["Andrey Borzenkov"]

Juan please disable CONFIG_APM_RTC_IS_GMT. Kernel does the right
thing when it is disabled ALWAYS. When it is enabled it just
confuses it.

Others who produce custom kernels may consider it as well :)

> 
> Andrey, as you can expect, your theory was correct: it is indeed the GMT
> setting in kernel:
> 
> standard 2.4.22-0.6mdk kernel (CONFIG_APM_RTC_IS_GMT=y):
>  [EMAIL PROTECTED] jkeller]$ date
>  Mon Aug 18 17:39:56 CEST 2003
> (suspend, wake up)
>  [EMAIL PROTECTED] jkeller]$ date
>  Mon Aug 18 19:40:23 CEST 2003
> 
> modified kernel (# CONFIG_APM_RTC_IS_GMT is not set):
>  [EMAIL PROTECTED] jkeller]$ date
>  Mon Aug 18 17:47:05 CEST 2003
> (suspend, wake up)
>  [EMAIL PROTECTED] jkeller]$ date
>  Mon Aug 18 17:47:40 CEST 2003
> 

background.

when APM_RTC_IS_GMT is disabled, kernel computes the difference
between CMOS and kernel time immediately before going to sleep and
restores kernel time taking difference in account after resuming.
It works irrespectively of whether your system is using GMT or
local time - for GMT diff will be (close to) zero, that's all.

When APM_RTC_IS_GMT is enabled kernel does nothing taking kernel
time be equal to CMOS time. This breaks for the case system is
using local time producing observed time jumps.

I am not sure about hwclock --hctosys (that is done in APM scripts);
I have vague recollection that it did break in this case as well but
I do not remember anymore why I came to this conclusion.

Anyway APM_RTC_IS_GMT is redundant for RTC == GMT, is buggy for
RTC == local so it should be killed.
[...]
> P.S. Andrey, I *think* I'm off your ISP's blacklist now. Knock on wood, this
> will make it to you.
> 

I got this just fine. Recently ISPs seem to go mad about blacklisting
everything possible. Even more amusing is that whether my ISP is
blacklisted or blacklists some address, they recommend contact the
other server's admin in both cases :)

Re: [Cooker] new kernel alsa

["Andrey Borzenkov"]

> Note#2: Please, notice that we merged snd-via686 and snd-via8233 into the
> snd-via82xx module. Also, the snd-usb-midi module users should use the
> snd-usb-audio module which has integrated support for both USB audio and MIDI.

to remind.

We have hotplug. And two modules matching USB audio hardware -
"audio" (OSS) and snd-usb-audio (ALSA). When you boot with external
audio plugged in it appears to work "by virtue" of coldplug not 
being used on Mandrake. If you plug in external audio after boot,
you are at the mercy of depmod - which module is listed first. So
far people hit "audio" :)

The only current way is to blacklist one or other (depending on
whether OSS or ALSA are selected). To do anthing more serious some
internal changes in hotplug are needed + corre. helper.

Just in hope somebody will have idea or even solution before 9.2 is out :)

-andrey

[Cooker] Re: iptables-1.2.8-1mdk: partly broken with 2.6.x kernels

["Andrey Borzenkov"]

> I wanted this documented for when Mandrake includes a 2.6.x in Cooker
> (next to 2.4). The contrib package sounds like a nice solution until
> Cooker includes 2.6. At that time, something which chooses at runtime between
> iptables-24 and iptables-26 would be needed (like the modutils/
> module-init-tools/autoconf/etc). This allows easy switching between the
> kernels.


does iptables compiled under 2.6 work under 2.4?

anyone contacted netfilter guys and asked if this breakage under 2.6
is intentional or a bug?

-andrey

Re: [Cooker] [RFD] supermount behaviour on empty drive

["Andrey Borzenkov"]

> here is poll - should this functionality
>
> - be disabled
> - made optional (mount option) that I do not like
> - left as is.
> ---
> i would say leave as-isish  but have "MountDrake" drop a file called 
> "this is an empty mount point" into the point when is is made so
you have

it has nothing with mountdrake and mount point is NOT empty - it is
mounted. handling it in kernel is nightmare and the last thing I will
do. Besides it does not solve the potential problem because nothing
prevents you from having a media with "this is an empty mount point"
file in it.

thank you for feedback anyway :)

-andrey

[Cooker] Re: Installed 2.6 kernel rpm won't boot.

["Andrey Borzenkov"]

[...]
> I am not sure which version I had compared to the website, I was using
> Cooker version as of Saturday, so I kept it in there.  Regenerated
> initrd and it worked properly.
> 
> >Now I would be _very_ intersted to know why it did not create modprobe.conf. 
> >Have you updated or installed module-init-tools?
> 
> Only what is in cooker.  The version I have on it is dated  Jul 27.  I
> first installed module-init-tools on this box back in February or March.
> 

Ah, OK, then it is clear. I update post to always generate modprobe.conf if empty.

now when thierry is back again? hopefully he can add official package.

Re: [Cooker] [RFD] supermount behaviour on empty drive

["Andrey Borzenkov"]

please Cc me I do not receive cooker

>> > last version makes ls /mnt/cdrom not fail but show empty directory if
>> > drive is empty (or if mount failed in general).
>>
>> IMHO, this is correct behaviour, as it is the same (to the end user) as
>> if they had not mounted by hand the CD on a machine without supermount.
>
> Caution!  Don't wake up an old bug : A blank CD or unformated floppy 
> causes troubles.

yes, that is what danny hinted as well. besides, it is very ugly to
implement (it adds special case and it is unclear *when* this special
case is needed as danny's report shows).

so I thought I ask befire it is too late ...

OK so what this "caution" is about - remove this feature, make
it conditional or what?

Re: [Cooker] 2.6-tips and docs

["Andrey Borzenkov"]

[...]
> 
> It doesn't load 8139too, and maybe cause of that I have completely
> empy routing-table when I do netstat -rn not even localhost seen?
> Usually I see three or four entries in netstat -rn.
> I also see following:
> "error, some other host already uses address 127.0.0.1"
>  loopback  arping:socket:
>  Address family not supported by protocol"
> Modprobe 8139too loads it or at least I see it in lsmod after modprobe,

which means it is not module-init-tools.

I hope someone else will be able to help you I do not have LAN
card (on a system with 2.6 at least) so ...

good luck

-andrey

Re: [Cooker] 2.6-tips and docs

["Andrey Borzenkov"]

-Original Message-

> 
> Mon 18. Aug 2003 14:26, Andrey Borzenkov wrote:
> > please Cc me I do not receive cooker.
> >
> > >8139too for the card rtl-8139c
> > >
> > >> > I have generated
> > >> >
> > >> > modprobe.conf in /etc and I have aliases there but no help.
> > >>
> > >> well you do not tell which modules do not load nor what is in
> > >> modprobe.conf ... hard to help.
> > >
> > > There is alias for that module, but how this mii relate to that 8139too?
> > > mii doesn't have alias and doesn't load automatically?
> >
> > I have no idea what mii is. Would you mind show your modules.conf,
> > modprobe.conf and explain what mii is and why you expect it to load
> > (or be module for that matter).
> >
> I know this not so good network card, but can not easily change it
> cause it is in laptop.
> 
> alias eth0 8139too
> alias usb-controller usb-uhci
> alias ieee1394-controller ohci1394
> alias sound-slot-0 via82cxxx_audio
> install sound-slot-0 /sbin/modprobe --ignore-install sound-slot-0 && { 
> /bin/aumix-minimal -f /etc/.aumixrc -L >/dev/null 2>&1 || :; }
> remove sound-slot-0 { /bin/aumix-minimal -f /etc/.aumixrc -S >/dev/null 2>&1 
> || :; }; /sbin/modprobe -r --ignore-remove sound-slot-0
> 
> Modinfo says about mii that it is:
> description: "MII hardware support library"
> author:  "Jeff Garzik <[EMAIL PROTECTED]>"
> license: "GPL"
> 

OK, I still do not understand your problem.

Does 8139too load or not? If not what is exact error? What happens
when you do modprobe 8139too? What happens if you do modprobe eth0
(of course assuming it is not already loaded)

I am sorry but I checked your posts in this thread in archives and
I cannot find ANY description of your problem except "it did not help". Nor do you 
give ANY description of your problem now.

Re: [Cooker] 2.6-tips and docs

["Andrey Borzenkov"]

please Cc me I do not receive cooker.

>8139too for the card rtl-8139c
>
>> > I have generated
>>
>> > modprobe.conf in /etc and I have aliases there but no help.
>>
>> well you do not tell which modules do not load nor what is in
>> modprobe.conf ... hard to help.
>>
> There is alias for that module, but how this mii relate to that 8139too?
> mii doesn't have alias and doesn't load automatically?

I have no idea what mii is. Would you mind show your modules.conf,
modprobe.conf and explain what mii is and why you expect it to load
(or be module for that matter).

thank you

Re: [Cooker] 2.6-tips and docs

["Andrey Borzenkov"]

> But installing module-init-tools-0.9.13-1bor.src.rpm didn't help.
> Kernel-2.6 still doesn't load modules I need. 

... which (modules) are ... ?

  I have generated
> modprobe.conf in /etc and I have aliases there but no help.

well you do not tell which modules do not load nor what is in
modprobe.conf ... hard to help.

-andrey

Re[2]: [Cooker] [RFD] supermount behaviour on empty drive

["Andrey Borzenkov"]

-Original Message-

> 
> On Mon, 18 Aug 2003, [koi8-r] "Andrey Borzenkov[koi8-r] "  wrote:
> 
> > 
> > last version makes ls /mnt/cdrom not fail but show empty directory if
> > drive is empty (or if mount failed in general).
> Andrey, 
> with 1.2.8 I sometimes get the old "stale NFS handle" error when doing ls 
> when the drive is empty.
> 

and what exactly you do in this case? Does it happen before you
ever attempt to access drive? After disk change? When you are inside
/mnt/cdrom (i.e. cd /mnt/cdrom; ls)?

[Cooker] [RFD] supermount behaviour on empty drive

["Andrey Borzenkov"]

last version makes ls /mnt/cdrom not fail but show empty directory if
drive is empty (or if mount failed in general).

this was done to ensure FAM can open mount point and get notification
when subfs gets mounted.

as danny pointed out it may be confusing for users. but it is the only
possbility I am aware of to make FAM and supermount coexist.

here is poll - should this functionality

- be disabled
- made optional (mount option) that I do not like
- left as is.

and to Laurent - yes, this version fixes FAM + supermount so you can
remove patch that disabled notification of ro media. 

Re: [Cooker] 2.6-tips and docs

[Andrey Borzenkov]

> (should module-init-tools not BuildRequire glibc-static-devel ?)

yes it should. But because I build it on 9.1 and 9.1 does not have 
glibc-static-devel I was a bit selfish and put something that applies to 
both.

I guess doing urpmi module-init-tools-XXX.src.rpm should pull in that as well, 
but I may be wrong.

Thierry, please put in correct requires for official package. I guess I can 
live with rpm -b --nodeps

sorry

-andrey

[Cooker] Re: Installed 2.6 kernel rpm won't boot.

[Andrey Borzenkov]

On Saturday 16 August 2003 12:21, Andrey Borzenkov wrote:
[...]
>
> It is going to be a problem. So what people think if I change spec from
>
> if new install
>   generate modprobe conf
>
> to
>
> if modprobe.conf is empty
>   generate modprobe.conf
>
> ??? It seems that current case more prblems than it helps ...

I uploaded new version that does it. please test. of course, as you have 
non-empty modprobe.conf ...

[Cooker] Re: Installed 2.6 kernel rpm won't boot.

[Andrey Borzenkov]

On Saturday 16 August 2003 05:49, Todd Lyons wrote:
...]
> request_module: failed /sbin/modprobe -- block-major-8. error = -16

block-major-8 == sd

you are missing SCSI in initrd
[...]

> append=" devfs=nomount debug noquiet hdc=ide-scsi acpi=on"

I myself never tried this actually but David runs without devfs, apparently it 
should at least boot :)

> initrd=/boot/initrd-2.6.0-0.test3.1mdk.img
>
> The initrd is present.  I had to take it out of framebuffer because the
> 2.6 kernel just blanks the screen.
>

yes, is appears to be the problem for generic ditribution kernel. So far the 
only advice on lkml was to disable all framebuffer drivers except one you are 
actually using. I have

CONFIG_FB=y
CONFIG_FB_VESA=y
CONFIG_VIDEO_SELECT=y

and run framebuffer just fine.

I guess someone should investigate it or just disable framebuffer entirely 
(putting some comments about this).

[...]
> [i960RM Bridge] [BRIDGE_PCI] megaraid: Intel Corp.|MegaRAID 467
> Enterprise 1500 RAID Controller [INTELLIGENT_I2O] pcnet32 :

is driver for this available in RPM? It may well be missing I do not know. Do 
you actually have megaraid module?
[...]

> probeall scsi_hostadapter megaraid

[...]

> Here is the modprobe.conf:
> [EMAIL PROTECTED] /home/todd]# cat /etc/modprobe.conf | grep -v "^#"
> include /lib/module-init-tools/modprobe.default
> include /lib/module-init-tools/modprobe.compat
>
>
> Does anybody see anything really obvious that I'm overlooking.  I tend
> to think that the lack of lines in /etc/modprobe.conf is probably
> causing my problem.
>

yes. it is empty so mkinitrd can't find neccessary modules. At the very least 
please generate it using

TESTING_MODPROBE_CONF=/etc/modules.conf generate-modprobe,conf >> 
/etc/modprobe.conf

and recreate initrd. If you get current modul-init-tools from 
http://supermount-ng.sf.net/mdk-25/ you can skip TESTING_MODPROBE_CONF - it 
does the right thing.


Now I would be _very_ intersted to know why it did not create modprobe.conf. 
Have you updated or installed module-init-tools?

It is going to be a problem. So what people think if I change spec from

if new install
  generate modprobe conf

to

if modprobe.conf is empty
  generate modprobe.conf

??? It seems that current case more prblems than it helps ...

Re: [Cooker] kernel 2.6 / laptop...almost success!

["Andrey Borzenkov"]

...]
> -o insmod.static -static insmod.o
> /usr/bin/ld: cannot find -lc
> collect2: ld returned 1 exit status
> make: *** [insmod.static] Error 1
^^
> error: Bad exit status from /home/adamw/rpm/tmp/rpm-tmp.166 (%build)
> 
> Which is odd. As far as I understand it, shouldn't ld -lc work so long
> as glibc-devel is installed?

you need glibc-devel-static (or is it glibc-static-devel?)
I'll update buildrequires

thank you for report

-andrey


 Yes, glibc-devel IS installed. :) Any
> ideas? Or could you provide a binary RPM for me to test? 

I do not have cooker so I 'do not want to make binary  RPMs. besides
it is a good test anyway :)

Re: [Cooker] [Bug 4592] [kernel-2.6.0-0.test2.2mdk] No display on Radeon Mobility (Dell C610)

[Andrey Borzenkov]

On Saturday 09 August 2003 08:23, Chmouel Boudjnah wrote:
> Chmouel Boudjnah <[EMAIL PROTECTED]> writes:
> > Andrey Borzenkov <[EMAIL PROTECTED]> writes:
> > > > BTW I need to modprobe 3c59x and sound modules.
> > >
> > > I do not have LAN card so somebody please test this and tell me what is
> > > needed to make it work :)
> >
> > /etc/modules.conf
> > alias eth0 whatmodule
> >

s/modules/modprobe/

that much I know :) I thought some extra steps are needed in 2.6. That should 
be taken care of by module-init-tools install.

sound must work as well, I had single problem - both sound-card-0 and 
sound-slot-0 were defined; apparently modutils do not mind module be loaded 
twice while module-init-tools returns error in this case.

> > ifconfig whatmodule => load_module("whatmodule")
> >
> > you can test even without physical card (it should try to load the
> > module when doing ifconfig interface)
>
> forgot to say that it works for me (2.6t2mm5)