mirror of
https://github.com/SDL-Hercules-390/hyperion.git
synced 2026-04-10 22:24:32 +02:00
74 lines
3.3 KiB
Plaintext
74 lines
3.3 KiB
Plaintext
|
|
Please update this file whenever appropriate!
|
|
|
|
Unresolved Issues
|
|
=================
|
|
|
|
*** XXX ***
|
|
Marks the functional subsystem affected
|
|
|
|
*** TAPE DEVICE HANDLER ***
|
|
CE+DE on tape rewind-unload:
|
|
|
|
Technically, they should be presented separately to the guest o/s:
|
|
CE upon completion the actual rewind-unload CCW and then DE once the
|
|
tape actually finishes unloading, but doing so is somewhat problematic
|
|
especially for SCSI tapes. (Might be possible to do for AWS/HET tapes
|
|
but I haven't looked into it yet).
|
|
|
|
Addendum (Ivan 2004/10/02)
|
|
Final DE on Rewind Unload must be accompanied by UC
|
|
(Unit Check) (all D/Ts) and possibly CUE
|
|
(Control Unit End) for 3420 (3803 CU Type, to
|
|
indicate an end of Control Unit Busy condition). It
|
|
is possible the CUE should only occur if a BUSY was
|
|
presented at some point, but I am not 100% sure about
|
|
that (and that's not supported either - as we don't
|
|
support shared CUs). However, VM/370 for example
|
|
insists that CUE be presented.
|
|
On the same line, VM/370 DDR is failing on Rewind
|
|
unload. This could be part of the above problem.
|
|
|
|
*** I/O SUBSYSTEM ***
|
|
Incorrect Channel program check reporting (Ivan 2004/10/02)
|
|
|
|
Because of the current I/O design, where we prefetch all write data
|
|
prior to issuing the I/O, program check is presented unconditionally
|
|
(except for immediate commands which are declared as such by the device
|
|
handlers) - while it should only be presented if the control unit (device
|
|
handler) attempts to consume data from the offending address.
|
|
|
|
*** DEVICE HANDLERS ***
|
|
Some units still returning CE+DE+UC on not-ready conditions (Ivan 2004/10/02)
|
|
|
|
I/O architecture requires non-ready units to return UC only.
|
|
This is done at the device handler lever and has not been implemented
|
|
in all device handlers
|
|
|
|
*** TAPE DEVICE HANDLER ***
|
|
AWS handler fails to read FLEX Generated AWS files (Ivan 2004/10/02)
|
|
|
|
The current AWS handler is not capable of processing multi-segment
|
|
blocks. However, this can be bypassed by forcing the use of the HET
|
|
handler which is backward compatible.
|
|
|
|
*** I/O SUBSYSTEM ***
|
|
I/O Subsystem reset/Device reset not always effective (Ivan 2004/10/02)
|
|
|
|
The I/O subsystem reset function does not actually send a reset signal
|
|
(as described in 17.2.2.2 in POO) to the device handlers.
|
|
If an I/O is in progress, it will not be terminated.
|
|
Subsequently, if status is stacked after the I/O reset, the device may
|
|
be left in an unusable state, with status pending and E bit off
|
|
effectivelly preventing an MSCH to be issued with the E bit on -
|
|
requiring a Power On Reset (hercules stop and start).
|
|
Only ESA/390 and ESAME are affected.
|
|
|
|
*** Hardware panel ***
|
|
Change to Enter key function (Ivan 2004/10/02)
|
|
|
|
Pressing enter while in single step mode used to be the equivalent of
|
|
pressing the start button (which was quite handy for debugging). It now
|
|
displays the last commands typed. Instead, 'start' must be typed every
|
|
time to execute the next instruction.
|