DFU UNDELETE may crash the system

[Karl Rohwedder]

We experienced a system crash together with the corruption of our systemdisk while using the DFU UNDELETE function (DFU V3.1-1).

I reported this to HP and the analysis of HP's VMS engeneering points to DF doing things behind the back of the XQP.

So I would advise you to use this function with extreme care.

regards kalle

[Karl Rohwedder]

I got an additional from the DFU maintainer:

The problem seems to arise, when DFU tries to UNDELETE a deleted, but still opened file (e.g. an installed image).

[David B Sneddon]

Kalle,

What version of VMS?

Dave

[Karl Rohwedder]

David,

it was VMS V7.3-2.

[Volker Halle]

Kalle,

to document the crash footprint, could you provide the CLUE file ?

If possible, just post the first page of the CLUE CRASH output inline and attach the full CLUE file.

Volker.

[Karl Rohwedder]

Volker,

here is the CLUE data:
Crashdump Summary Information:
------------------------------
Crash Time: 8-JUN-2005 12:31:43.07
Bugcheck Type: XQPERR, Error detected by file system XQP
Node: QAK01 (Standalone)
CPU Type: AlphaServer DS10 617 MHz
VMS Version: V7.3-2
Current Process: iTKalle@FTA77
Current Image: DSA0:[CNC_DISK_AXP.][CNCMANAGER]DFU.EXE;1
Failing PC: FFFFFFFF.802133A4 SEARCH_FCB_C+00EE8
Failing PS: 08000000.00000000
Module: F11BXQP (Link Date/Time: 1-OCT-2004 19:58:56.24)
Offset: 000173A4

Boot Time: 24-FEB-2005 19:44:53.00
System Uptime: 103 16:46:50.07
Crash/Primary CPU: 00/00
System/CPU Type: 2208
Saved Processes: 62
Pagesize: 8 KByte (8192 bytes)
Physical Memory: 2048 MByte (262144 PFNs, contiguous memory)
Dumpfile Pagelets: 299869 blocks
Dump Flags: olddump,writecomp,errlogcomp
Dump Type: compressed,selective,shared_mem
EXE$GL_FLAGS: poolpging,init,bugdump
Paging Files: 1 Pagefile and 1 Swapfile installed

regards Kalle

[Petr Spisek]

Hi Karl,
I have the same experience with DFU V3.0 on the same VMS. My colleague tried undelete a lot of files by "macro-script" from PTW32 and system fell down :-( So, DFU is freeware (care with this). For example look to my attachment - compare DFU versus DFO report. DFU returns very different fragmentation report :-) DUMP/HEADER to the most fragmented file says that DFO report is right.
Petr

[Ian Miller.]

Petr,
have you reported the difference in fragmentation reports? I think it may be due
looking at the same thing in two different ways.

[comarow]

DFU is unsupported.
If the file is now being used, you are out of luck. It may be useful, but as you can see dangerous.

[Ton Dorland]

V3.1-1 fixes the issue with reporting the wrong most fragmented file.

It bothers me a bit to see a few crashes lately with the undelete. May some changes in the XQP in the latest VMS version trigger this.

Remember that UNDELETE assumes that a file is really deleted, but when it isn't (that is there are still VMS resources allocated to the file) a UNDELETE may crash the system. The best thing is to first do a UNDELETE/LIST to see waht can be undeleted, and then verify (e.g. with SHOW DEV/FILES) if a file is still open. If it is not, it should be safe to UNDELETE it.

Regards,

Ton Dorland
DFU Author

[comarow]

Remember to install the lastest xf11 and RMS ECOs.