Opened 5 days ago
Last modified 5 days ago
#440 new defect
Cannot reboot Multics after certain error exits from ISOLTS — at Initial Version
| Reported by: | Eric Swenson | Owned by: | Eric Swenson |
|---|---|---|---|
| Priority: | minor | Milestone: | MR12.9 |
| Component: | Hardcore | Version: | MR12.5 |
| Keywords: | Cc: |
Description
Seen on the DPS8M emulator (2026-11-02): test cpu b under ISOLTS, quit, then shut and boot star in the same emulator session. The controller holding the test processor’s memory (MEM B) had both of its execute-interrupt masks assigned to CPU A’s port.
configure_test_cpu needs a mask on the controller being tested that is assigned to an online processor (the “mask CPU”). If no mask of that controller is assigned to one, but a mask is free(4MW SCU), it calls scr_util$assign mask to assign a free mask to the first online processor.
It then saves the controller’s configuration data (scs$cfg_data_save = scs$cfg_data (iscu)).
During the test, scr_util$isolts_scu_p1 sets mask A to the mask CPU and mask B to the test CPU; at the end isolts_scu_p3 writes the saved configuration back to the controller.
assign_mask$assign_mask loops over the two masks of a 4MW SCU and assigns every free mask it finds: the loop has no exit after the first assignment. On a controller with no masks assigned (the usual case for
a controller with no IOM connected, such as MEM B here), both masks are assigned to the mask CPU, in scs$controller_data (eima _ata) and in scs$cfg_data. The saved copy has both, and isolts_scu_p3 writes both to the controller. The controller keeps that until it is cleared, so the next bootload’s init_scu (via scas_init) finds two masks assigned to one processor and stops.
