Talk:List of SITS system calls - Revision history

Jnc: /* MFPI call */ typo

2022-10-07T15:20:04Z

‎MFPI call: typo

Larsbrinkhoff: /* MFPI call */ Correct.

2022-10-07T13:10:00Z

‎MFPI call: Correct.

Jnc: MFPI call

2022-10-07T11:58:15Z

MFPI call

New page

==MFPI call==

I have a suspicion about that the MFPI call does. The MFPI instruction ('Move From Previous mode Instruction space') does not work when 'current' and 'previous' modes (in the [[Processor Status Word|PS]]) are both 'user'. (This is the normal way for the PS to be set up when user code is running, to keep it from poking into things.) This is apparently to allow keeping proprietary libraries 'private' - i.e. allow them to be used, but not inspected. (The PDP-10 has a similar 'feature'.)

The problem is that some things (e.g. the nargs() function; also trap handlers which run in user mode, for various kinds of traps) ''require'' the ability to look at user-mode code - and so they don't work when running with 'split I&D' (where one has to use MFPI to read instructions). Maybe the MFPI call simulates the correct operation of MFPI, for those cases?

We actually got started on installing a switch in the CSR -11/45, to disable the disabling of user-user MFPI. I never got around to doing the etch cuts to finish it, though. It would have been labelled, of course, 'Magic/More Magic'! :-) [[User:Jnc|Jnc]] ([[User talk:Jnc|talk]]) 13:57, 7 October 2022 (CEST)

← Older revision		Revision as of 13:10, 7 October 2022
Line 6:		Line 6:

	We actually got started on installing a switch in the CSR -11/45, to disable the disabling of user-user MFPI. I never got around to doing the etch cuts to finish it, though. It would have been labelled, of course, 'Magic/More Magic'! :-) [[User:Jnc\|Jnc]] ([[User talk:Jnc\|talk]]) 13:57, 7 October 2022 (CEST)		We actually got started on installing a switch in the CSR -11/45, to disable the disabling of user-user MFPI. I never got around to doing the etch cuts to finish it, though. It would have been labelled, of course, 'Magic/More Magic'! :-) [[User:Jnc\|Jnc]] ([[User talk:Jnc\|talk]]) 13:57, 7 October 2022 (CEST)
		+
		+	: Thanks. I checked, and that's what it does. [[User:Larsbrinkhoff\|Larsbrinkhoff]] ([[User talk:Larsbrinkhoff\|talk]]) 15:09, 7 October 2022 (CEST)

@@ Line 1: / Line 1: @@
 ==MFPI call==
-I have a suspicion about that the MFPI call does. The MFPI instruction ('Move From Previous mode Instruction space') does not work when 'current' and 'previous' modes (in the [[Processor Status Word|PS]]) are both 'user'. (This is the normal way for the PS to be set up when user code is running, to keep it from poking into things.) This is apparently to allow keeping proprietary libraries 'private' - i.e. allow them to be used, but not inspected. (The PDP-10 has a similar 'feature'.)
+I have a suspicion about what that MFPI call does. The MFPI instruction ('Move From Previous mode Instruction space') does not work when 'current' and 'previous' modes (in the [[Processor Status Word|PS]]) are both 'user'. (This is the normal way for the PS to be set up when user code is running, to keep it from poking into things.) This is apparently to allow keeping proprietary libraries 'private' - i.e. allow them to be used, but not inspected. (The PDP-10 has a similar 'feature'.)
 The problem is that some things (e.g. the nargs() function; also trap handlers which run in user mode, for various kinds of traps) ''require'' the ability to look at user-mode code - and so they don't work when running with 'split I&D' (where one has to use MFPI to read instructions). Maybe the MFPI call simulates the correct operation of MFPI, for those cases?