Fwd: Re: MBR was destroyed

[seorge]

Please tell me, what exact information should I provide the developers with 
the experiment described below.
I really want to help, but I also don't want to play with my data without 
understanding what exactly is needed from me.
Thanks in advance!

----------  Forwarded Message  ----------
> Ok, and can restore ~/.wine dir to previous state (maybe "rm -rf ~/.wine
> && wineprefixcreate" would help) and reproduce this bug again, please?

I've removed my previous .wine dirrectory, so the only thing I can do is to
reproduce everything happend with me yesturday:
1. try to launch some win32 application - this will create .wine
2. try to setup disks and change labels.
As I undestand after this experiment I need to send something or at the
mailing list from the content of the .wine. Please tell me what it should be.

On Mon, 21 Nov 2005 15:05:28 +0100, seorge  wrote:
> Please tell me, what exact information should I provide the developers  
> with
> the experiment described below.
> I really want to help, but I also don't want to play with my data without
> understanding what exactly is needed from me.
> Thanks in advance!
>
> ----------  Forwarded Message  ----------
> > Ok, and can restore ~/.wine dir to previous state (maybe "rm -rf ~/.wine
> > && wineprefixcreate" would help) and reproduce this bug again, please?
>
> I've removed my previous .wine dirrectory, so the only thing I can do is  
> to
> reproduce everything happend with me yesturday:
> - try to launch some win32 application - this will create .wine
> - try to setup disks and change labels.
>
> As I undestand after this experiment I need to send something or at the
> mailing list from the content of the .wine. Please tell me what it  
> should be.

Saulius wrote:
You may run dd command to retrieve MBR at every moment, calculate it's
checksum and see whether it differs or not.

Excellent idea. makes testing a whole lot easier . Now we all know what  
the result is we just need to find out what actions trigger a write to the  
disk.

I suggest:
add user to disk group, if not already ;)

backup mbr:
dd if=/dev/hda of=/tmp/hda.wine0.mbr bs=512 count=1
mount /mnt/fd  (adjust as necessary to your config)
dd if=/dev/hda of=/mnt/fd/hda.mbr bs=512 count=1
umount /mnt/fd

fiddle with wine, one step at a time , running the following check at each  
step, keeping notes on what is done each time.

dd if=/dev/hda of=/boot/hda.wine1.mbr bs=512 count=1
diff /tmp/*0.mbr /tmp/*1.mbr

If diff returns a blank fiddle a bit harder!

Once you get a hit , post both the files and a note about exactly what  
step you did to provoke a change.

Recover your MBR:
dd of=/dev/hda if=/tmp/hda.wine0.mbr bs=512 count=1

remove user from disk group ( and/or stop using wine !! )

The main thing is to establish exactly what you did, it may not be quite  
where you thought and it will make finding the offending code a lot  
quicker if it is precisely linked to one event.

A new element came out in your last post that you run unconfigured wine  
for the first time by "clicking on a file", this brings a whole lot of  
other things into the picture , so please note _everything_ you do from  
the begining.

I have been unable to reproduce this bug on Gentoo which also has the same  
permissions on /dev/hda .

HTH

[Jonathan Adamczewski]

[e-mail] wrote:
> I have been unable to reproduce this bug on Gentoo which also has the 
> same permissions on /dev/hda .

/dev/hda may be root:disk, but if you've setup your machine based on the 
standard Gentoo install handbook, your regular user account won't be a 
member of the disk group.

j.

On Mon, 21 Nov 2005 21:37:34 +0100, Jonathan Adamczewski  
 wrote:
> [e-mail] wrote:
> > I have been unable to reproduce this bug on Gentoo which also has the  
> > same permissions on /dev/hda .
>
> /dev/hda may be root:disk, but if you've setup your machine based on the  
> standard Gentoo install handbook, your regular user account won't be a  
> member of the disk group.
>
> j.

duh, read the rest of the post :roll:

[Jonathan Adamczewski]

[e-mail] wrote:
> duh, read the rest of the post :roll:

My apologies - your mailer appears to have munged the quoting in that 
post and I misinterpreted.

j.

[Saulius Krasuckas]

* On Mon, 21 Nov 2005 [e-mail] wrote:
> - On Mon, 21 Nov 2005 15:05:28 +0100, seorge  wrote:
> > Please tell me, what exact information should I provide the developers 
> > with the experiment described below.
>
> You may run dd command to retrieve MBR at every moment, calculate it's
> checksum and see whether it differs or not.

BTW, after the Alexandre reply about volume changing I guess this possibly 
should be partition boot sector that is destroyed, not a MBR.

Seorge, can you explain, how did you exactly restored MBR using GRUB, 
please?
> fiddle with wine, one step at a time , running the following check at 
> each step, keeping notes on what is done each time.
>
> dd if=/dev/hda of=/boot/hda.wine1.mbr bs=512 count=1
> diff /tmp/*0.mbr /tmp/*1.mbr

BTW, I think only the text files should be diffed (and -u flag makes 
output very understandable for me):

cat /tmp/hda.wine0.mbr | od -t x1 > /tmp/0.mbr.txt
 cat /boot/hda.wine1.mbr | od -t x1 > /tmp/1.mbr.txt
 diff -u /tmp/0.mbr.txt /tmp/1.mbr.txt

That would show how much different MBRs are.  And if they don't start to 
differ ever, maybe we should refer to appropriate partition (system from 
which cannot boot later)?  Then "hda" string should be replaced with a 
partition name - "hda1" or so.

But comparing hexadecimal dumps should be the only and the final step, as 
it says not only whether sectors are different, but also how much they 
differ.  Which isn't interesting at start of hunting.
> The main thing is to establish exactly what you did, it may not be quite 
> where you thought and it will make finding the offending code a lot 
> quicker if it is precisely linked to one event.

It would be interesting to know, whether latest CVS helps, as Alexandre 
has put some fixes in regarding this issue, I guess:

http://www.winehq.org/pipermail/wine-cvs/2005-November/019377.html

[seorge]

> Seorge, can you explain, how did you exactly restored MBR using GRUB,
> please?

I've booted with grub-floppy, then when the system was up, I executed 
'grub-install /dev/hda'.

On Tue, 22 Nov 2005 12:30:02 +0100, Saulius Krasuckas   
wrote:
> Seorge, can you explain, how did you exactly restored MBR using GRUB,
> please?
> > fiddle with wine, one step at a time , running the following check at
> > each step, keeping notes on what is done each time.
> >
> > dd if=/dev/hda of=/boot/hda.wine1.mbr bs=512 count=1
> > diff /tmp/*0.mbr /tmp/*1.mbr

oops , the first line should of course be :

dd if=/dev/hda of=/tmp/hda.wine1.mbr bs=512 count=1
> BTW, I think only the text files should be diffed (and -u flag makes
> output very understandable for me):

diff will just say the files differ if run on the binaries. That's enough  
here.
> But comparing hexadecimal dumps should be the only and the final step, as
> it says not only whether sectors are different, but also how much they
> differ.  Which isn't interesting at start of hunting.

as you rightly say.
> > The main thing is to establish exactly what you did, it may not be quite
> > where you thought and it will make finding the offending code a lot
> > quicker if it is precisely linked to one event.
> It would be interesting to know, whether latest CVS helps, as Alexandre
> has put some fixes in regarding this issue, I guess:

One step at a time !

I'd like to see the original poster post what he really did. The little  
snippets given so far are little help.

It seems he is using something like kde, clicking on a .exe , and that  
wine was completely unconfigured at this stage.

Again "I used grub boot disk" does not help us know where grub is  
installed.

Maybe someone else who has been able to reproduce this needs to post  
details.