Cooperative Linux

On Sat, Jan 31, 2004 at 03:03:28PM -0500, mylist@... wrote:
> First, thanks a bunch, Dan, for all of the hard work you've done. I'm
> looking forward for working with colinux, like many others who are
> following this project.
> 
> I download and installed the binary files, per the instructions in the
> README. The TAP-Win32 driver seemed to install OK. When I try to run the
> colinux-daemon.exe program, I get "Error initializing the kernel driver"
> and it terminates with a "-1". FWIW, I'm running Windows-XP Pro. Any
> suggestions?

Since more than one person saw this problem, I've compiled a slightly 
modified colinux-daemon.exe and added it to the release.

Now, for the source code part of this message:

It appears that an error is wrongly propagated from kernel space 
to user space when the driver gets initialized by the daemon. The branch  
that gets affected is in co_manager_init() (colinux/user/manager.c). It 
also appears that it doesn't happen on my setup.

In the patched daemon, I changed co_manager_init() so that it would 
always return CO_RC_OK. With this patch, it appears that the daemon 
continues running on Matt's setup without other such failures.

Since this is a developers mailing list, I urge you to look at the 
source and send me a patch :) I mean, help me debug this faster. You do
want a working coLinux, right? ;).

-- 
Dan Aloni
da-x@...

OK, that was different - it didn't work for me. I get a blue-screen when I
run the modified colinux-daemon.exe, due to:

IRQL_NOT_LESS_OR_EQUAL

Larry


> On Sat, Jan 31, 2004 at 03:03:28PM -0500, mylist@... wrote:
>> First, thanks a bunch, Dan, for all of the hard work you've done. I'm
>> looking forward for working with colinux, like many others who are
>> following this project.
>>
>> I download and installed the binary files, per the instructions in the
>> README. The TAP-Win32 driver seemed to install OK. When I try to run the
>> colinux-daemon.exe program, I get "Error initializing the kernel driver"
>> and it terminates with a "-1". FWIW, I'm running Windows-XP Pro. Any
>> suggestions?
>
> Since more than one person saw this problem, I've compiled a slightly
> modified colinux-daemon.exe and added it to the release.
>
> Now, for the source code part of this message:
>
> It appears that an error is wrongly propagated from kernel space
> to user space when the driver gets initialized by the daemon. The branch
> that gets affected is in co_manager_init() (colinux/user/manager.c). It
> also appears that it doesn't happen on my setup.
>
> In the patched daemon, I changed co_manager_init() so that it would
> always return CO_RC_OK. With this patch, it appears that the daemon
> continues running on Matt's setup without other such failures.
>
> Since this is a developers mailing list, I urge you to look at the
> source and send me a patch :) I mean, help me debug this faster. You do
> want a working coLinux, right? ;).
>
> --
> Dan Aloni
> da-x@...
>
>
> -------------------------------------------------------
> The SF.Net email is sponsored by EclipseCon 2004
> Premiere Conference on Open Tools Development and Integration
> See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
> http://www.eclipsecon.org/osdn
> _______________________________________________
> coLinux-devel mailing list
> coLinux-devel@...
> https://lists.sourceforge.net/lists/listinfo/colinux-devel
>

On Sat, Jan 31, 2004 at 04:32:01PM -0500, mylist@... wrote:
> OK, that was different - it didn't work for me. I get a blue-screen when I
> run the modified colinux-daemon.exe, due to:
> 
> IRQL_NOT_LESS_OR_EQUAL

This is bug I'm haunting for awhile, but I haven't managed to reproduce 
it with my computers. I'm not sure what causes this, but it is possible 
that some hardware configurations are responsible. 

If you have some Windows kernel debugging experience or know anyone who 
does and can help, please tell.

-- 
Dan Aloni
da-x@...

Well, its definitely repeatable on my machine! I'm using Windows XP-Pro on
a Dell Inspiron 2650 Notebook, if that helps. I haven't done any Windows
Kernel debugging before, but I'll see what I can come up with.

Larry.


> On Sat, Jan 31, 2004 at 04:32:01PM -0500, mylist@... wrote:
>> OK, that was different - it didn't work for me. I get a blue-screen when
>> I
>> run the modified colinux-daemon.exe, due to:
>>
>> IRQL_NOT_LESS_OR_EQUAL
>
> This is bug I'm haunting for awhile, but I haven't managed to reproduce
> it with my computers. I'm not sure what causes this, but it is possible
> that some hardware configurations are responsible.
>
> If you have some Windows kernel debugging experience or know anyone who
> does and can help, please tell.
>
> --
> Dan Aloni
> da-x@...
>

> OK, that was different - it didn't work for me. I get a blue-screen when I
> run the modified colinux-daemon.exe, due to:
>
> IRQL_NOT_LESS_OR_EQUAL
>
> Larry

I got this as well, and i'm now waiting for the system to dump the whole
memory. would any kind of processing on the memdump be useful to anyone ?

the machine is a Dell Inspiron 5150, 1.6Ghz, P4, 512M ram, and enough HD
to have a couple more full mem dumps...

Let me know...
Jok
---
Nothing is foolproof to a sufficiently talented fool...
  oo
,(..)\
  ~~

On Sat, Jan 31, 2004 at 02:25:45PM -0800, ePAc wrote:
> > OK, that was different - it didn't work for me. I get a blue-screen when I
> > run the modified colinux-daemon.exe, due to:
> >
> > IRQL_NOT_LESS_OR_EQUAL
> >
> > Larry
> 
> I got this as well, and i'm now waiting for the system to dump the whole
> memory. would any kind of processing on the memdump be useful to anyone ?

These memory dumps sound useful. Do you know some reference about how to 
enable them and other info that might be useful?

-- 
Dan Aloni
da-x@...

> > > IRQL_NOT_LESS_OR_EQUAL
> > >
> > > Larry
> >
> > I got this as well, and i'm now waiting for the system to dump the whole
> > memory. would any kind of processing on the memdump be useful to anyone ?
>
> These memory dumps sound useful. Do you know some reference about how to
> enable them and other info that might be useful?

right click on "My Computer"
select "Properties"
select "Advanced" tab
under "Startup And Recovery" there is some option to setup different type
of memory dump (minidump, kernel dump, full memory dump).

i had mine setup as "Full Memory Dump" (512M).

as for analysis, there used to be some tools on the MS web site, i'll
track those down.

DebugView was able to "process" that memdump and give me 3 messages:
1  -  2:06:06 PM  -  Driver Unloaded (*)
2  -  2:07:38 PM  -  Driver Loaded
3  -  2:07:38 PM  -  Allocated cmon0

(*) might be from the first attempt for which Memdump wasn't enabled
during the previous boot or something [/me clueless]...


this was running under XP. Have you done any kind of testing with XP ? or
was it all win2k ?

my other machine (win2K) is loading properly, and the console is coming
up, and i'm having the same set of problem with the e2fsck (i will write
down those numbers next time).

All in all, "This is fantastic !"(TM). You have now access to probably a
couple dozen different config (mine included) which will allow us all to
run linux on our poor windows machine :o)

is there anyone on the list who has access to the newer DDK (2003/XP
version), and that could compile the linux.sys with that OS. It would make
sense that the linux.sys driver need to be compiled for the proper target
OS. is there any success on XP ?

Thanks,
Jok
---
Nothing is foolproof to a sufficiently talented fool...
  oo
,(..)\
  ~~

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

ePAc wrote:

| is there anyone on the list who has access to the newer DDK (2003/XP
| version), and that could compile the linux.sys with that OS. It would make
| sense that the linux.sys driver need to be compiled for the proper target
| OS. is there any success on XP ?

In theory I've got access to it but not enough disk space to install it.
~ I doubt that's the problem though as from the readme the compilation is
done under cygwin and just the final link is done using the DDK.  It's
possible, though, that cygwin is making some assumption about userspace
that's simply not true at the driver level... compiling the source under
visual studio might be worth it.

Tony
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAHDSf5UdHDk9LaRcRArJdAJ4wDZMHEAXyl5wCDR9lkGEELbjAhACePzzO
4GIaHqJEOwHbobC3hW7zpMQ=
=LE8u
-----END PGP SIGNATURE-----

On Sat, Jan 31, 2004 at 10:02:39PM +0000, Tony Hoyle wrote:

> On an XP Pro machine with (most of) the latest patches, even the
> updated colinux-daemon doesn't work:
>=20
> D:\colinux>colinux-daemon
> Cooperative Linux daemon
> Manager is already running
> Manager not initialized (0)
> Removing driver leftover
> Stopping driver service
> Removing driver service
> Installing kernel driver
> Error installing kernel driver: -15

Was this your first time?  I got similar messages earlier when I
was playing around, but did not think to record them.  The catch
was that I was trying to kill the first daemon and left myself
needing a reboot.

Also, you are an Administrator?

> I've installed the TAP driver and this comes up 'not connected', which I
> assume is normal.

Yes, it is.  It'll come up connected when the kernel boots.

--=20
Matt Behrens <matt@...> <http://www.zigg.com/>
Life's too short to recompile what you don't hack. <http://www.debian.org/>

On Sat, Jan 31, 2004 at 05:12:52PM -0500, Matt Behrens wrote:
> On Sat, Jan 31, 2004 at 10:02:39PM +0000, Tony Hoyle wrote:
> 
> > On an XP Pro machine with (most of) the latest patches, even the
> > updated colinux-daemon doesn't work:
> > 
> > D:\colinux>colinux-daemon
> > Cooperative Linux daemon
> > Manager is already running
> > Manager not initialized (0)
> > Removing driver leftover
> > Stopping driver service
> > Removing driver service
> > Installing kernel driver
> > Error installing kernel driver: -15
> 
> Was this your first time?  I got similar messages earlier when I
> was playing around, but did not think to record them.  The catch
> was that I was trying to kill the first daemon and left myself
> needing a reboot.

Matt, killing the daemon forcefully yields unexpected results at this 
time. There is a need to avoid doing that.

> Also, you are an Administrator?

Right, that would have been worth mentioning in the README file I 
supplied.

-- 
Dan Aloni
da-x@...

On Sat, Jan 31, 2004 at 10:02:39PM +0000, Tony Hoyle wrote:

> On an XP Pro machine with (most of) the latest patches, even the
> updated colinux-daemon doesn't work:
> 
> D:\colinux>colinux-daemon
> Cooperative Linux daemon
> Manager is already running
> Manager not initialized (0)
> Removing driver leftover
> Stopping driver service
> Removing driver service
> Installing kernel driver
> Error installing kernel driver: -15

Did the daemon run before the run you sent above? It couldn't have 
reached 'Manager is already running' if it isn't so. It would be very 
interesting to know what it spitted out in the very first run.

> I've installed the TAP driver and this comes up 'not connected', which I
> assume is normal.  The only things running are my normal bootup
> configuration of XFree & sshd, plus sundry grapihics/sounds systray
> things that seem to plague Windows so much.
>
> Running colinux-daemon a second time causes the system to reboot.. I
> guess it's not cleaning up somewhere?

Thanks for your feedback.

I admit there are currently some stability issues concering the removal 
of the driver. I haven't completly addressed them, but I am planning to 
do so.

-- 
Dan Aloni
da-x@...

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Matt Behrens wrote:
|
| Was this your first time?  I got similar messages earlier when I
| was playing around, but did not think to record them.  The catch
| was that I was trying to kill the first daemon and left myself
| needing a reboot.

This was straight after a boot.

| Also, you are an Administrator?

Yes.

I could see if I can run it in a debug/checked WinXP I guess, but if
this driver is as low level as it sounds that might not help.

Tony



-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAHCrd5UdHDk9LaRcRAubNAJ0eyof8/rSqQ+XAvAW29+sIdrAjiACeO87N
YakrkAOvC6RJI847WrHvpEA=
=xs+Y
-----END PGP SIGNATURE-----

On Sat, Jan 31, 2004 at 10:23:25PM +0000, Tony Hoyle wrote:

> I could see if I can run it in a debug/checked WinXP I guess, but if
> this driver is as low level as it sounds that might not help.

I disagree. Running in a debug/checked Windows XP might help to spot
other issue, not only explaining this one.

-- 
Dan Aloni
da-x@...

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dan Aloni wrote:

| Did the daemon run before the run you sent above? It couldn't have
| reached 'Manager is already running' if it isn't so. It would be very
| interesting to know what it spitted out in the very first run.

No, not at all...

It's obviously detecting the daemon when it's not there.

Tony

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAHCyh5UdHDk9LaRcRAkcDAJ9rWj3+jOVdBvMpx26ArvLlHGQn5wCePfI2
siE4VjPQSx9LSMtx1TjWql4=
=im6R
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dan Aloni wrote:


| I disagree. Running in a debug/checked Windows XP might help to spot
| other issue, not only explaining this one.

Not much help, I'm afraid.  The debug build I have is a very minimal
configuration so whatever's conflicting isn't installed there - It
started initialising and vmware blew up...  I'd half expect it not to
work under vmware anyway so that's no surprise.

Driver LoadedAllocated cmon0
cocon: cocon_startup
cocon: cocon_init
cocon: cocon_set_palette
cocon: cocon_set_origin
cocon: cocon_save_screen
cocon: cocon_dummy
cocon: cocon_dummy
cocon: cocon_set_origin
cocon: cocon_switch
cocon: cocon_set_palette
Reported size: 1073741824 bytes
cobd opened (\DosDevices\c:\coLinux\Debian-3.0r0.ext3.1gb)
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Search Route #2: Route not found
Delete Route #3: Route not found
Delete Route #3: Route not found
Delete Route #3: Route not found
MsgErrorLogWrite: COMPONENT = MESSENGER
MsgErrorLogWrite: An unexpected network control block (NCB) was
received. The NCB is the data.

0x000C54CC 912301020C550C0000002A2020202020.#...U....*


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAHDEb5UdHDk9LaRcRAsXfAJ9oswF9LctE4b2MgxWCY6E72unKAgCfSUHZ
kTW41mBZ+TIqT66nSEI/ddQ=
=Dvqv
-----END PGP SIGNATURE-----

On Sat, Jan 31, 2004 at 02:39:36PM -0500, Matt Behrens wrote:

> Result:
> 
> C:\coLinux>colinux-daemon.exe
> Cooperative Linux daemon
> Installing kernel driver
> Error initializing kernel driver
> Removing kernel driver
> Stopping driver service
> Removing driver service
> Daemon failed: -1
>
> C:\coLinux>
> 
> Any hints as to increasing verbosity etc. would be much appreciated.
> All files are as the README asks them to be, and I am a "Computer
> Administrator" in XP Home's user panel (which is significantly
> suckier than 2000's, or I gather XP Pro's.)

Can you please send me the output of DebugView

http://www.sysinternals.com/files/dbgvnt.zip

(Make sure that it captures kernel debug messages)

Thanks,

-- 
Dan Aloni
da-x@...

On Sat, Jan 31, 2004 at 03:06:57PM -0500, Matt Behrens wrote:
> On Sat, Jan 31, 2004 at 09:58:29PM +0200, Dan Aloni wrote:
> 
> > I attached a slightly modified version of the daemon.
> 
> Okay, this one got me booted.  However, now I see:

Oh very good, now you can play with it ;)
 
> /dev/cobd0: The filesystem size (according to the superblock) is 262144 blocks
> The physical size of the device is 30448 blocks
> 262144 sounds much more likely to me, given the gigabyte image I'm
> mounting.

It occured on my setup as well.

I need to see more of these reports to find out whether the 30448 
number that fsck prints there is fixed. I also saw this today. While 
trying to debug it, It sometimes occurs and sometime doesn't, and only 
with that image. 

I am going to look inside the fsck source code to get more clue.
Perhaps it's a cobd bug, related to the way it answers to ioctl()s.

Until it gets fixed, you can hack your way out of it if you edit
that init script from inside coLinux.

-- 
Dan Aloni
da-x@...

On Sat, Jan 31, 2004 at 10:46:56PM +0200, Dan Aloni wrote:

> I need to see more of these reports to find out whether the 30448=20
> number that fsck prints there is fixed. I also saw this today. While=20
> trying to debug it, It sometimes occurs and sometime doesn't, and only=20
> with that image.=20

FWIW, I made a smaller image (128M) by loop-mounting both and moving
the files over; I got the same message on boot, this time 121793
blocks vs. the correct figure of 131072.

Curious -- can I infer that there are disk-style images that do
work?

--=20
Matt Behrens <matt@...> <http://www.zigg.com/>
Life's too short to recompile what you don't hack. <http://www.debian.org/>

Okay, it seems there are worse problems than just misreporting the
size in cobd.

It seems in e2fslibs that the size is actually determined by an
algorithm I can't get my weekend brain around, but it does a lot
of seeking and trying to read a byte to get the "real" size.

Armed with this idea I tried to dd if=3D/dev/cobd0 of=3D/dev/null bs=3D1k
count=3D128k (using my 128M image).  I got a stream of I/O errors
near the end and DebugView is filled with several messages starting
with "block io failed: edbb6000 400 (reason: c0000011)".  Goes up
to edea6c00.

I can't find the above DebugView message anywhere in the 0.5.2
colinux source I pulled today.

So I think I'm stuck for the day... hopefully someone smarter than
me can pick it up from here.

--=20
Matt Behrens <matt@...> <http://www.zigg.com/>
Life's too short to recompile what you don't hack. <http://www.debian.org/>

On Sat, Jan 31, 2004 at 04:55:57PM -0500, Matt Behrens wrote:
> Okay, it seems there are worse problems than just misreporting the
> size in cobd.
> 
> It seems in e2fslibs that the size is actually determined by an
> algorithm I can't get my weekend brain around, but it does a lot
> of seeking and trying to read a byte to get the "real" size.
> 
> Armed with this idea I tried to dd if=/dev/cobd0 of=/dev/null bs=1k
> count=128k (using my 128M image).  I got a stream of I/O errors
> near the end and DebugView is filled with several messages starting
> with "block io failed: edbb6000 400 (reason: c0000011)".  Goes up
> to edea6c00.

c0000011 is STATUS_END_OF_FILE.

It is possible that reading cobd devices from userspace is broken. 
Prehaps in that scenario offsets are passed from Linux to Windows 
wrongly. 

> I can't find the above DebugView message anywhere in the 0.5.2
> colinux source I pulled today.

It's in colinux/os/winnt/kernel/block.c.

-- 
Dan Aloni
da-x@...