Bug 115827

Summary: Updater does not recognize exiting linux partitions
Product: [openSUSE] SUSE LINUX 10.0 Reporter: Gernot Payer <gpayer>
Component: InstallationAssignee: Hendrik Vogelsang <hvogel>
Status: VERIFIED FIXED QA Contact: Klaus Kämpf <kkaempf>
Severity: Critical    
Priority: P5 - None CC: aj, hvogel, jsrain
Version: RC 1   
Target Milestone: ---   
Hardware: Other   
OS: All   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: y2log
patch for probe_reiserfs

Description Gernot Payer 2005-09-08 11:17:49 UTC 
So far I updated quite a few times from betaN to betaN+1 and the updater always found the root 
partition. However, this didn't happen with RC1.

I'll set this to critical, because this could prevent easy updating from SL93 to SL10.0.

As soon as I get y2log files, I'll attach them.
Comment 1 Gernot Payer 2005-09-08 11:22:34 UTC 
Created attachment 49182 [details]
y2log
Comment 2 Jiri Srain 2005-09-08 13:44:35 UTC 
According to log, all detected partitions have in their target map set: 
 
"detected_fs":`unknown 
 
Thomas, Arvin?
Comment 3 Arvin Schnell 2005-09-08 14:07:07 UTC 
The command blkid to detect the filesystems returns -127 (seg. fault?).

Please run as root the command "BLKID_SKIP_CHECK_MDRAID=1 /sbin/blkid -c
/dev/null" and provide the output (best done in the inst-sys).
Comment 4 Gernot Payer 2005-09-08 14:20:14 UTC 
Doing this gives no output, just a Floating Point Exception.
Comment 5 Arvin Schnell 2005-09-08 14:22:56 UTC 
So it's a blkid problem.
Comment 6 Arvin Schnell 2005-09-08 14:23:42 UTC 
Maybe you can also try without the BLKID_SKIP_CHECK_MDRAID=1.  But it's
unlikely this changes anything.
Comment 7 Hendrik Vogelsang 2005-09-08 14:47:16 UTC 
i need more info. can i access that machine somewhere?
Comment 8 Gernot Payer 2005-09-08 14:54:31 UTC 
Of course. Its name is "testbox". Standard root pwd. If you want physical access, it's on 2nd floor, room 
next to the garbage can.
Comment 9 Hendrik Vogelsang 2005-09-08 17:19:52 UTC 
the machine is offline. cant do anything now......
Comment 10 Gernot Payer 2005-09-09 10:18:18 UTC 
It's online again.
Comment 11 Hendrik Vogelsang 2005-09-09 10:33:40 UTC 
its failing on reiserfs with a blocksize of 512
Comment 12 Hendrik Vogelsang 2005-09-09 11:07:53 UTC 
mason/jeff please have a look at this patch. its just skipping the check if the
blocksize is smaller then 1024. Were not sure what that check is for.
Comment 13 Hendrik Vogelsang 2005-09-09 11:08:26 UTC 
Created attachment 49356 [details]
patch for probe_reiserfs
Comment 14 Hendrik Vogelsang 2005-09-09 11:10:25 UTC 
wtf
Comment 15 Chris L Mason 2005-09-09 12:25:36 UTC 
The check is because reiserfs 3.5.x and higher disks have the super block at offset 64k.    
The older disk format has the super block at 16k and, and 64k into the disk is the   
journal.   
   
If there is a copy of the super block logged at offset 64k on an old format disk, the   
detection code will incorrectly think it is a newer format FS.  So the check compares the   
block number of the super block found with the journal parameters of the FS and makes   
sure the superblock isn't inside its own journal.   
  
We need to keep the check.
Comment 16 Mads Martin Joergensen 2005-09-09 12:27:10 UTC 
But does the patch in Comment #13 hurt?
Comment 17 Chris L Mason 2005-09-09 12:35:28 UTC 
Thinking harder, it doesn't hurt.  The older filesystems don't support a blocksize of 512  
bytes, so you can't possibly find a 512 byte super  in the log of an older FS.  
   
The patch is fine.
Comment 18 Mads Martin Joergensen 2005-09-09 12:36:45 UTC 
It's checked in for RC2