Questions good, no matter how late they're asked
HP leads Cognos into extra support year

HP liberates some 3000 patches

Customers of all types, self-supporting and third-party supported, as well as those with HP support can now download new HP 3000 patches. Well, not exactly new, but tested long enough to gain the HP support labs' elusive General Release status.

An HP advisory has outlined a raft of patches which HP labeled as Recommended fixes for HP 3000s running MPE/iX 7.5, 7.0 and 6.5. The patches can be downloaded from the HP IT Response Center Web site by any customer running an HP 3000. According to HP support, a Recommended patch

should be applied at your earliest convenience. There is potential for sub-optimal performance, lock-ups or unwanted  shutdowns. This will protect your system from a serious, but not  unrecoverable failure.

More than a dozen such patches were identified as General Release fixes in an HP patch advisory from last week. None bring extra functionality to the HP 3000 customer, unless you count 7.0 patch MPEMXQ3B, which HP described as "Enhancement: Load PDC [Processor Dependent Code] into main memory on A- and N-Class systems." The patch is a fix for a problem which has much more serious consequence — a panic — on HP's Unix systems, according to HP's notes.

The advantage of this patch seems to be giving diagnostics software better performance. HP's details on the "enhancement" patch, from the HP Web page

Intermittent small system freezes and system software clock slips happen on multiprocessor A- and N-Class systems. The clock slips are not gradual, they happen in steps of several seconds each.

The clock slips occur when a CPU that tries to load and run PDC code from NVRAM has to block on a condition where the system bus is too busy. Loading (copying) the PDC code from NVRAM (EEPROM) into normal memory will avoid any blocking for PDC code and will also make its execution significantly faster. 

MPE code using PDC calls (mainly diagnostics) will see increased performance.

This is a solution borrowed from HP-UX where similar issues were seen with A-, L- and N-Classes. Unlike the HP-UX code though, MPE will not ‘panic’ if it cannot relocate the PDC.  MPE will fallback to execute PDC from NVRAM instead (just like without this patch).

HP explains that the patch applies the enhancement by reducing the amount of memory available for the OS and applications by 2MB.

Of such specific repairs and minute enhancements are the General Releases of 2007 made. Other patches now in GR, most to avoid System Aborts (SA):

MPENX08A - SA0 attempting to boost the priority of a completed disk I/O

MPEMXU5C - FSCHECK error header record main_rec count did not match actual count

MPEMXP3C - SA614 when POSIX app writes past limit of fixed-length record file

MPEMXC7C - SA 0 With An Invalid Virtual Address

MPEMX59B - Workaround For OCT Cornercase Involving SCANFMVAT CM Code

MPELXY8B - SA1350 during termination of a process with DEBUG breakpoints set

MPELXJ7C - Data Loss When Using TAR with MPE fixed ASCII Files

MPEMXW2A - SA5414 while running FCSCAN in a loop

MPEMXM7A - DAT fails to open Disc to Disc dump when limit is a multiple of 4096MB

MPELXT1D - Network Spooler: 2 fixes for PJL syntax errors, NMS 9621

INTHDH4A - General Fixes for Internet Services Products (such as INETD and REMSH) on MPE/iX 7.5

INTHD63A - General Fixes for Internet Services Products on MPE/iX 7.0

HP has moved some patches for the 3000 into General Release at a much faster clip, when the potential for data corruption loomed, even for just a few customers. While many sites will leave this list of patches alone, some might be a repair for a system which must continue to carry mission critical data for an indefinite future of homesteading.

Even with the GRs for patches announced in one massive e-mail notice, however, dozens more remain quarantined in beta test status. Genuine advances such as the brand-new SCSI Pass Through driver need to be implemented and tested by HP support customers — the only ones who can use a BT patch.