home *** CD-ROM | disk | FTP | other *** search
- OSHTDUMP - OS/2 2.X AND OS/2 WARP MEMORY DUMP PROCESS
- 12/16/94
-
- ================================================================
- OS/2 2.X AND OS/2 WARP MEMORY DUMP PROCESS
- ================================================================
-
- Please Read Entire Document for Full Explanation of Procedures
- ----------------------------------------------------------------
-
-
- A memory dump is performed when a problem is difficult to
- reproduce or other methods of problem determination do not
- identify the problem.
-
-
- THE NEW DUMP PROCEDURE
- ----------------------------------------------------------------
-
- There are new files required for a memory dump. OS/2 Warp
- contains the latest files. The following update does NOT
- apply to OS/2 Warp.
-
-
- OS/2 2.0, 2.1 and 2.11 do not contain the latest dump files. You
- must install the new dump procedure, NEWDUMP.ZIP on the system
- before a doing memory dump. NEWDUMP.ZIP can be downloaded from the
- bulletin board at (407) 443-8000.
-
- 1. Logon with your name and password.
- 2. At the main menu press F for files.
- 3. S to Select a library.
- 4. Type 8 and press Enter.
- 5. Type D to download a file.
- 6. Type NEWDUMP.ZIP and press Enter.
-
- The READMD file explains how to apply newdump to the system.
-
-
- NOTE: OS/2 2.0 requires XR06100 (service pak II) as a
- prerequisite for NEWDUMP.ZIP.
-
-
- You need blank formatted diskettes for a memeory dump. Generaly,
- one 1.44MB diskette is used for each 2MB of RAM. For example, if
- you have a 16MB system, you need 8 formatted diskettes.
-
-
- STARTING THE TRACE
- ----------------------------------------------------------------
-
- NOTE: Do NOT begin TRACE unless an IBM Technical Analyst tells
- you to.
-
-
- For OS/2 prior to OS/2 2.1:
-
- 1. Add the following lines in CONFIG.SYS:
-
- TRACE=ON
- TRACEBUF=63
- TRACE=OFF 4,24,47,50,97,98
-
- 2. Restart the system.
-
-
- The above starts TRACE function and turns off major codes 4, 24,
- 47,50, 97, and 98. The default for TRACE=ON is to trace all
- system events. These major codes are for some housekeeping
- functions that are not normally used by development.
-
-
- For OS/2 2.1:
-
- 1. Add the following lines to the CONFIG.SYS (exactly):
-
- TRACEBUF=63
- TRACE=ON
-
- 2. Restart the system.
-
- 3. When the system restarts, type the following from a command
- prompt:
-
- TRACE ON KERNEL
- TRACE ON DOSCALL1
-
- ----------------------------------------------------------------
-
- Different situations determine the type of memory dump
- required. IBM OS/2 Support can help to determine the correct
- process for your situation.
-
- For example, the OS/2 kernel and device drivers run at Ring 0
- while applications run at Ring 3.
-
- ----------------------------------------------------------------
- OS/2 2.0, 2.1, OR 2.11 WITHOUT FIXPACKS
- ----------------------------------------------------------------
-
- RING 3 TRAPS
- ----------------------------------------------------------------
-
- AUTOMATIC DUMP PROCESS:
-
- 1. Add the following to the CONFIG.SYS:
-
- TRAPDUMP=ON
-
- This causes an automatic DUMP if the failing CS:EIP is in
- Ring 3 code (the CS value ends in either 3, 7, B or F). It
- does not DUMP on Ring 0 code traps. Applications run at Ring
- 3 while the OS/2 Kernel runs at Ring 0.
-
-
- MANUAL DUMP PROCESS :
-
- Press and hold Ctrl+Alt and then press Num Lock twice to start a
- MANUAL DUMP of the system. This starts the dump process in most
- cases, even if the keyboard is locked.
-
-
- ----------------------------------------------------------------
- OS/2 2.11 WITH FIXPACK (36 or later) OR OS/2 WARP
- ----------------------------------------------------------------
-
- AUTOMATIC DUMP PROCESS FOR RING 0 TRAPS - dumps BOTH
- ring 0 and non-ring 0 traps
-
- 1. Add the folling to the CONFIG.SYS:
-
- TRAPDUMP=ON
-
-
-
- RING 0 TRAPS
- ----------------------------------------------------------------
-
- AUTOMATIC DUMP PROCESS:
-
- 1. Add the following to the CONFIG.SYS:
-
- TRAPDUMP=R0
-
- This begins an automatic Dump if the Failing CS:EIP is
- in Ring 0 code (the EIP ends in 0, 4, or 8). It does not
- DUMP Ring 3 traps.
-
-
- MANUAL DUMP PROCESS:
-
- Press and hold Ctrl+Alt and then press Num Lock twice. This
- starts the dump process in most cases, even if the keyboard is
- locked.
-
-
- STARTING THE MEMORY DUMP
- ----------------------------------------------------------------
-
- When the memory dump begins, screen goes blank and you are
- prompted to insert a blank formatted diskette. You are prompted
- for additional diskettes as required, and then asked to insert
- the DUMP diskette again. When the process is completed, restart
- the the system.
-
- Register information is not displayed when an automatic memory
- dump is taken. The register information is in the dump
- information.
-
- Reading the Trap information from the dump diskettes.
- ----------------------------------------------------------------
- To view the dump diskettes for the revision of the kernel and the
- trap information:
-
- 1. Insert the first dump diskette in drive A:
- 2. Type: A: EDUMPDATA.001 and press Enter.
- A message appears stating the editor detected embedded
- null characters as it read the file.
- 3. Press: Continue Loading File. A screen of unreadable
- characters appears.
- 4. Press Edit, Find, Find: revision
- This finds the level of the kernel you are running.
- For example: 6.617 for OS/2 2.11, 8.162 for Warp.
-
- You can also search on the cs:eip to see if the trap screen
- is present.
-
- 5. Place the mouse pointer at the begining of the file.
- 6. Press Edit, Find Find: CS:EIP
- If a trap is present it, displays the registers.
- (Sometimes more than one trap is found in the dump. If the
- registers are all x's, press find again to see if there are
- additional traps.)
-
- **WARNING**
- ----------------------------------------------------------------
- DO NOT edit the file by adding characters or spaces. This can
- cause the memory dump to be unreadable. After viewing
- the dump, always exit without saving the file because it may
- cause the dump to be unreadable. If you have made any changes to
- the file by accident, also exit without saving the file.
-
-
- MAILING THE MEMORY DUMP TO IBM
- ----------------------------------------------------------------
- DO NOT send dump diskettes unless instructed by Level 2 Support.
- Diskettes MUST be clearly labeled with the PMR# or APAR# and your
- name. Make sure the diskettes are numbered.
-
- DUMP diskettes should be sent to:
-
- IBM Corp.
- 1000 NW 51st Street
- Boca Raton, FL 33432
- ATTN: Name of the OS/2 Analyst requesting the dump
- Bldg 001-3
- Internal Zip 1024
-
- ** For IBM Internal Customers ONLY**
-
- Disk images may be created using OS2IMAGE.EXE (OS2TOOLS), no
- other utility is acceptable, for example:LOADDSKF. Please send
- the disk images to the VM ID given to you by the OS/2 Analyst.
- The dump formatter requires disk images as opposed to the
- contents of each diskette. OS/2 Support cannot download dumps
- from IBMLINK, the PCC BBS, COMPUSERVE, etc. OS/2 Sypport may be
- able to arrange an upload to our Service Delivery Machine.
-
-
- DUMPING TO HARD DISK
- ----------------------------------------------------------------
-
- NOTE:
- Memory dump to hard disk must be to a separate, empty partition.
-
- IT WILL ERASE WHATEVER IS ON THE PARTITION.
- -------------------------------------------
-
- You MUST install NEWDUMP on the system if you are running OS/2
- 2.0, 2.1, or 2.11.
-
- 1. Create a partition about 4meg larger than the amount
- of memory installed.
- 2. Format the partition FAT.
- 3. Label the partition SADUMP.
- 4. Edit the CONFIG.SYS as follows:
- trapdump=off,X:
- where X: is the partition letter
- 5. Restart the system.
-
- - To start the memeory dump to hard disk, press and hold
- Ctrl+Alt and press Num Lock key twice. 16megs should
- dump in less than a minute.
-
- A good test would be to run CHKDSK, not CHKDSK/F, on a
- large enough drive. While CHKDSK is still running, begin
- the dump.
-
- Changes to the memory dump to hard disk on 8/1/94:
- ----------------------------------------------------------------
-
- 1. A message tells which partition will be dumped to, and
- how much memory.
- 2. A warning message tells the dump will wipe out the dump
- partition. You have 30 seconds to respond. If you do not
- answer in 30 seconds, the memory dump begins.
- 3. If you choose not to dump to the hard disk by answering N to
- the above prompt, you are asked if you want to dump to
- diskette.
- 4. If an error occurs, and the dump cannot go to the hard
- disk, you are asked if you want to dump to diskettes.
- If the answer is no, or there is no response in 60
- seconds, the system restarts.
-
- Changes to the memory dump to hard disk and diskette on 11/01/94.
- ----------------------------------------------------------------
-
- 1. Improved error checking on dump to hard disk and diskette.
- 2. More descriptive error messages.
- 3. If there is a bad sector encountered when dumping to hard
- disk, a Fatal Dump Error message appears followed by the bad
- sector information to help pinpoint the error.
- There is currently no error recovery in dump to hard disk.
- 4. Support for systems with 64meg or more of memory installed.
- DO NOT USE a memory dump dated before 11/01/94 on systems
- with 64 meg or more.
- 5. Better performance.
- 6. Improved error recovery from bad diskettes.
-
-
- ----------------------------------------------------------------
- IBM disclaims all warranties, whether express or implied,
- including without limitation, warranties of fitness and
- merchantability with respect to the information in this document.
- By furnishing this document, IBM grants no licenses to any
- related patents or copyrights.
-