SLED runs as an application program under IPL-EXEC. IPL-EXEC is part of the initial program loader (IPL), which is loaded and initialized before SLED is loaded.
When loading, a distinction must be made between loading SLED for the first time after a system crash, repeating a SLED run and dumping from the SLED system.
Before loading SLED for the first time the memory areas used by IPL, BOOT and SLED must be saved. This is done partly by the firmware (by copying data areas to save areas in memory or to the service processor) but for the most part by the software by writing data areas to the IPL disk to a save file (SLEDSAVE or BOOTSAVE) before they are used by BOOT, IPL or SLED.
Initialization of SLED varies according to server type. It is described in full in the appropriate manuals for each server type.
For the loading procedure, the load disk is searched for the following files, which must be anchored in the SVL of the disk with SIR:
- $TSOS.SYSPRG.BOOT.DSKnnn.SAVE (BOOTSAVE)
- $TSOS.SYSPRG.IPL.DSKnnn (IPL phase)
- $TSOS.SYSREP.IPL.DSKnnn (corrections for IPL)
- $TSOS.SYSREP.SLED.DSKnnn (corrections for SLED)
- $TSOS.SYSPRG.SLED.DSKnnn.SAVE (SLEDSAVE)
Before SLED is loaded, all the disks required for subsequent processing should already have been mounted and switched online.
Once SLED has been loaded and started, a number of consistency checks are carried out to establish:
whether the version of the loaded SLED matches that of the loading IPL
which system was previously loaded and if this was BS2000 (also under VM2000 or for DUMP from SLED) whether its version matches that of SLED
whether a part of main memory was overwritten without having been saved.
SLED performs these consistency checks regardless of whether the dump runs attended or unattended. Unattended operation means that SLED runs automatically without operator intervention, using the presettings in the SLED parameter file or default settings. Attended operation means that SLED prompts the operator to enter or correct options and SLED is controlled interactively.
Appropriate warnings regarding execution of the consistency checks are output at the console and logged in the SLED output file.
Diagnostic data can be output to the public disks, to private disks or tape.
The appropriate response must be entered to the following message during an attended SLED run:
NSD1003 STANDARD SLED ? REPLY (Y; N; EOT=Y)
Standard SLED
By entering Y or <EOT> in response to message NSD1003, the operator selects standard SLED. This results in the standard SLED behavior described below:
In both attended and unattended operation, a check is made as to the availability of the disks of the home pubset and the paging disks (online scan). Unavailable (offline) disks are logged via message
NSD1400. If data from the missing disks is subsequently required, SLED must be reloaded once these disks have been attached; it is not possible to attach required disks during the SLED run.The more pubsets SLED has to include (the home pubset of the system to be dumped, the load pubset of SLED and possibly another pubset for the parameter file and one for the SLED output file), the longer the online scan takes.
The default values for the MODE and TASK parameters (described below) are initialized as EOT. Regardless of whether or not there has been a preceding system crash, SLED selects the settings MODE=STD (see "Output to a dump file") and TASK=STD (see "Output to a dump file").
Once the SLED initialization phase is complete, it becomes known whether
all disks of the home pubset of the aborted session are online
public disks from different pubsets are online
all paging disks used in the aborted session are online
The operator dialog in standard SLED is continued with message NSD5200 (assignment of a parameter file).
Nonstandard SLED
By entering N in response to message NSD1003, the operator selects nonstandard SLED. The operator requests an extended dialog with SLED for controlling execution and receives the following additional messages:
NSD0900 ONLINE SCAN ? REPLY (Y; N; IPL-CONF=I; GENERAL ONLINE SCAN=X; EOT=Y)
SLED asks whether an online scan is to be performed, i.e. whether the available device configuration is to be checked.
| IPL performs an online scan for each pubset required. |
| IPL is not to perform an online scan |
| Instead of an online scan, the system-specific partition in the file containing the current configuration for system initialization, $TSOS.SYSDAT.IPL-CONF.<ver>, is evaluated. If this partition does not exist in the file or if errors occur during processing, an online scan is initiated again for the pubsets required. After successful processing, SLED knows the home pubset, the paging disks and the SLED load disk if the associated disks were already attached when the system was booted and could therefore be entered in the IPL-CONF file. |
| IPL performs an online scan for all disks. SLED cannot use the general online scan for large systems, since it cannot manage more than 1290 disk device entries. This option should only be selected in special cases |
The response is followed by two further messages which the operator must answer. The responses determine the scope of the SLED file.
NSD3001 SPECIFY NOEDIT MODE. REPLY (STD; NSF; REAL; ALL; EOT=STD; - (BACKTRACK))
For a description of this message, see "Output to a dump file".
NSD3002 SELECT TASKS. REPLY (STD; NONE; ALL;(TSN LIST); EOT=STD/ALL; - (BACKTRACK))
For a description of this message, see "Output to a dump file".
Error conditions for standard and non-standard SLED
Although SLED can run if the SYSRES (system disk) is not available, no system files (TSOSCAT, logging files etc.) and possibly no paging area data can be saved.
If the home pubset is only partially available, some system files may be only partially saved.
Unavailable paging disks may result in incomplete diagnostic information.
If SLED determines that one of the required disks is missing, this is logged via message NSD1400. The operator can attach the missing disks and repeat the SLED run.
Repeat SLED
A repeat SLED refers to the loading and initialization of another SLED run following a first SLED run in order to obtain the dump of the previously aborted session. This may be necessary, for example, if a SLED was inadvertently loaded which was not compatible with the version of the aborted system or the disks required during SLED initialization were not available (online).
Consequently, with a repeat SLED the areas saved in BOOTSAVE and SLEDSAVE must be used again and must not be saved a second time.
Both firmware and software attempt to recognize a repeat SLED and in this case suppress the saving of data areas. This means that no data is lost in the case of a repeat dump. If the newly loaded SLED does not belong to the same operating system version as the first SLED, it may be that the repeat SLED is not recognized and some of the diagnostic data is lost.
SLED dump
If an error occurs during a SLED run (message NSD1002), it may be necessary to take a dump from SLED. This means that SLED is loaded again in order to generate information on the errored SLED run.
Therefore, even though SLED was already loaded, it is necessary in this case to save the memory areas in which SLED was loaded a second time since these areas are required for SLED diagnostics. Both in the firmware and in the software it is necessary to note the following:
In the event of a SLED repetition, the data overwritten by BOOT, IPL and SLED has already been saved and is therefore not saved again.
In the case of a SLED dump, the data overwritten by BOOT, IPL and SLED (SLED data) is now saved again.
It is therefore necessary to take special measures if it is necessary to generate a dump relating to the execution of the dump function itself (SLED dump):
In VM2000 operation, this is achieved at VM start (
/START-VM) by specifying the parameterUNLOCK-SAVEAREA=*YESOn x86 servers at system start with
ipl parameter [d|u]: u(UNLOCK).However, on /390 servers it is necessary to perform certain actions. The selection of the actions depends on the server in question. For a detailed description see the manuals for the various server types.
The actions may include, for example:
Stop CPU
Log register contents
Set the real address stop X'4000'
On multiprocessor systems, set START/STOP mode to TARGET CPU
Start dump function
After the address stop is effective, overwrite real memory location X'1800' with X'00'
Reset address stop
Start CPU
Proceed as for dump function
Reset START/STOP mode after termination of dump function
Function selection
The output medium is selected by responding to the following message:
NSD3000 SPECIFY OUTPUT DEVICE. REPLY (DPUB; DPRIV; TAPE; EOT=DPUB; -(BACKTRACK))
Possible responses:
| output to public disk (default value) |