User manual

Miscellaneous Notes

• Interlocked queue header is genuinely corrupt, for example was overwritten by memory

access to incorrect address. No immediate fix is possible. Reason for memory corruption

need to be investigated and resolved.

• Interlocked queue operation retry loop uses fixed hardwired retry count, such as

900,000, and queue interlock holder takes abnormally long time to release it. Albeit

some limited recourses may be possible, such as tightening SYNCW_ILK window,

normally queue interlock holder is not expected to hold it for that long. Switching from

native interlock mode to portable mode or to portable mode in conjunction with

synchronization window are possible temporary ad hoc emergency recourses, however

the underlying reason for abnormally long interlock holding need to be investigated and

resolved.

• Interlocked queue operation retry loop uses retry count controlled by SYSGEN

parameter LOCKRERTY. In this case additional ad hoc emergency recourse is possible:

increasing the value of LOCKRETRY. However LOCKRETRY should not be increased

excessively high since long spinning time waiting for queue header interlock can lead to

low responsiveness of spinning processor and to CPUSPINWAIT bugcheck being

triggered by other processors. If the latter happens, threshold for CPUSPINWAIT

timeout (SYSGEN parameter SMP_SPINWAIT) may also be increased somewhat as

emergency override measure, such as from 1 second to 2-3 seconds. However the

underlying reason for abnormally long queue holding time should be identified and

resolved as soon as possible.

CPUs do not start

If you execute START / CPU/ALL command but some CPUs won’t start, check if you have

properly increased the value of SYSGEN parameter SPTREQ as described in section “Quick VAX

MP Installation Guide”. If SPTREQ is insufficient to allocate per-CPU data structures and map

them, START /CPU/ALL will fail to start some or all of additional processors, but it won’t display

error message. If you try to start a specific processor with START / CPU <n> command and

SPTREQ is insufficient, the command will display generic error NOSUCHCPU. Unfortunately DCL

START /CPU command ignores actual diagnostic code coming from SYSLOA/VSMP modules and

does not display it properly.

Message “Warning: Unable to lock VAX MP in memory (not critical)” – Linux only

This message may be displayed by VAX MP on Linux only in response to startup script command

DEPOSIT WS_LOCK 1 or commands that set non-zero values of WS_MIN or WS_MAX. When

these values are set, VAX MP tries to lock all of the simulator’s code and data in memory to

reduce paging. If page faults happen during the execution of simulator while VCPU thread that

incurred the fault is holding a lock or OpenVMS level spinlock or needs to process time-critical

operation such as responding to interprocessor interrupt, waiting for the fault to complete may