The Sinclair QL Forum

Posted: **Tue Nov 03, 2015 6:49 pm**

Hi All,

I have a plain QL setup that is, JS ROM, 8049 and TK2.

I use SER2 to transfer data, SER1 as mouse port, and seldom SER1 to printer output. I sometimes experience problems with serial line, after exchanging data with PC, while reseting the BAUD will not help, and I need to reboot the machine to be able to use printer or mouse again.

I was trying to flush any serial data queued, e.g. by COPY SER TO SCR, then brake, but it also did not bring the right syncing with mouse or printer back to normal.

Is there a way to reset SER line along with 8049 chip to its initial state by a program without need of rebooting?

Many thanks in advance.

Tom

Posted: **Tue Nov 03, 2015 7:41 pm**

Not to my knowledge.

You can press <CTRL><ALT>7 , that will reset the 8049. But also resets (and thus, confuses) all comms with the IPC, so it's not really useful because it will also kill your keyboard normally.

If all channels to the serial ports are closed and it still doesn't work, I'm afraid you'll need a full reset.

Tobias

Posted: **Tue Nov 03, 2015 8:04 pm**

Tobias,

Many thanks, I was not aware of "<CTRL><ALT>7" - key combination, I will certainly try. I am just curious, where this trick is documented.

Also, can I somehow achieve <CTRL><ALT>7 by a program?

Many thanks
Tom

Posted: **Wed Nov 04, 2015 12:04 am**

tcat wrote:Tobias,

Many thanks, I was not aware of "<CTRL><ALT>7" - key combination, I will certainly try. I am just curious, where this trick is documented.

Also, can I somehow achieve <CTRL><ALT>7 by a program?

Many thanks
Tom

I scanned an article by Alan Turnbull about implementing a Warm Reset using the CTRL-ALT-7 key combinations. I hope they'll show up below - it should be possible to click on the pages to magnify them, my OCR program couldn't handle the grey boxes, so I just scanned as a graphic.

This key combination was used to invoke and external monitor/debugger during QL development, Some code remained in Sinclair ROMs, the nett effect is usually just to lock up the QL. Here is what Andrew Pennell said about it in QDOS Companion book:

Code: Select all

Level 7 is usually ignored but may be redfined with RAM vector $1C. Level 7  interrupts can normally only be triggered by additional hardware. However, an early debugging tool has been left in the hardware so that pressing CTRL ALT 1, 5 or 7 invokes the equivalent interrupt. A level 7 interrupt will usually cause the machine to lock up as it destroys the communication process between  68008 and 8049 IPC.

warmreset_page1of2.jpg

warmreset_page2of2.jpg

Posted: **Wed Nov 04, 2015 8:30 am**

Hi,

Was there not a bug in the 8049 ALT 7 code, which Hermes cured...

Posted: **Wed Nov 04, 2015 9:52 am**

Derek_Stewart wrote:Hi,

Was there not a bug in the 8049 ALT 7 code, which Hermes cured...

Hermes converts CTL-ALT-7 into CTL-ALT-ESC-7 which is probably just so that it is harder to press by mistake (extra key). I think on a Hermes system that it actually resets the QL as opposed to just locking up on a QL without Hermes.

Posted: **Wed Nov 04, 2015 10:36 am**

Maybe I didn't make myself clear enough:

<CTRL><ALT>7 does re-set the IPC. But as the communication between IPC and "main" QL is in no way secured and breaks as soon as one of the two partners is not compliant with it (which happens if the IPC stops to transmit in the middle of a transfer because it has just been re-set, and we're always in the middle of a transfer in that situation, because after all: you've just pressed a key).......

This key combination is thus entirely useless and normally just locks up the QL.

Tobias

Posted: **Sat Nov 14, 2015 9:02 pm**

Dear All,

I have hand coded the assembly, compiled and run the binary. Pressing <CTL> <ALT> <7>, resets the machine, goes to the initial screen, but does not clear the RAM, but does reset system variables.

I cannot quite make out from the scan, these addresses, so I hope I got them right.

WARM_RESTART_AH EQU $01CA Warm retstart for 'AH' ROM (QDOS v1.02)
WARM_RESTART_JM EQU $01CA Warm retstart for 'JM' ROM (QDOS v1.03)
WARM_RESTART_JS EQU $01CC Warm retstart for 'JS' ROM (QDOS v1.10)
WARM_RESTART_JSU EQU $01CE Warm retstart for 'JSU' ROM (QDOS v1U10)
WARM_RESTART_MG EQU $01CC Warm retstart for 'MG' ROM (QDOS v1.13)

Code follows, please compile with GST MAC as the assembly uses long labels, older GST Assemblers can cope only with short labels up to 8 char in length.

Code: Select all


* Motorola MC68008 program to provide 'warm reset'
* facility by re-defining the <CTRL>, <ALT>, '7' effect.
* Also, vector chich normally cause the QL to 'crash' are
* re-defined to cause a warm system restart instead.
* 
* Assemble, save to Microdrive, reserve memory with
* something like LET a=RESPR(256), load code into memory
* with something like 'LBYTES mdv1_code,a' and activate
* with 'CALL a'.
* 
* Alternatively, run program in Listing 2.
* 
* COPYRIGHT (c) August 1985, Alan Turnbull, B.Sc.  


BASIC_JOB         EQU     $00           SuperBASIC job ID
MT_TRAPV          EQU     $07           TRAP #1 key for TRAP re-direction

WARM_RESTART_AH   EQU     $01CA         Warm retstart for 'AH' ROM (QDOS v1.02)
WARM_RESTART_JM   EQU     $01CA         Warm retstart for 'JM' ROM (QDOS v1.03)
WARM_RESTART_JS   EQU     $01CC         Warm retstart for 'JS' ROM (QDOS v1.10)
WARM_RESTART_JSU  EQU     $01CE         Warm retstart for 'JSU' ROM (QDOS v1U10)
WARM_RESTART_MG   EQU     $01CC         Warm retstart for 'MG' ROM (QDOS v1.13)
 
NEW_DIV_BY_ZERO   EQU     $005E
NEW_CHK_INSTR     EQU     $005E
NEW_TRAPV_INSTR   EQU     $005E
NEW_TRACE_EXCEPT  EQU     $005E
NEW_TRAP_05       EQU     $005E
NEW_TRAP_06       EQU     $005E
NEW_TRAP_07       EQU     $005E
NEW_TRAP_08       EQU     $005E         Address of 'RTE' in all ROMs
NEW_TRAP_09       EQU     $005E
NEW_TRAP_0A       EQU     $005E
NEW_TRAP_0B       EQU     $005E
NEW_TRAP_0C       EQU     $005E
NEW_TRAP_0D       EQU     $005E
NEW_TRAP_0E       EQU     $005E
NEW_TRAP_0F       EQU     $005E

VIDEO_RAM         EQU     $20000        Start of video RAM
SV_RAMT           EQU     $28020        System variable that holds RAM top + 1
SV_STACT          EQU     $28480        End of fixed system variables

                  SECTION NMI7

                  LEA     TRAP_TABLE,A1                    Point to new TRAP vector table
                  LEA     NEW_ADDR_ERROR,A0
                  MOVE.L  A0,ADDRESS_ERROR-TRAP_TABLE(A1)
                  LEA     NEW_ILL_INSTR,A0
                  MOVE.L  A0,ILLEGAL_INSTR-TRAP_TABLE(A1)  Load new vectors
                  LEA     NEW_PRIV_VIOL,A0
                  MOVE.L  A0,PRIVILEGE_VIOL-TRAP_TABLE(A1)
                  LEA     NEW_INT_LEV_7,A0
                  MOVE.L  A0,INT_LEV_7-TRAP_TABLE(A1)
                  MOVEQ   #BASIC_JOB,D1                    Signal SuperBASIC job
                  MOVEQ   #MT_TRAPV,D0                     Re-define trap table
                  TRAP    #1                               Return to SuperBASIC
                  RTS

TRAP_TABLE

ADDRESS_ERROR     DS.L    1
ILLEGAL_INSTR     DS.L    1
                  DC.L    NEW_DIV_BY_ZERO
                  DC.L    NEW_CHK_INSTR
                  DC.L    NEW_TRAPV_INSTR
PRIVILEGE_VIOL    DS.L    1
                  DC.L    NEW_TRACE_EXCEPT
INT_LEV_7         DS.L    1
                  DC.L    NEW_TRAP_05
                  DC.L    NEW_TRAP_06
                  DC.L    NEW_TRAP_07
                  DC.L    NEW_TRAP_08
                  DC.L    NEW_TRAP_09
                  DC.L    NEW_TRAP_0A
                  DC.L    NEW_TRAP_0B
                  DC.L    NEW_TRAP_0C
                  DC.L    NEW_TRAP_0D
                  DC.L    NEW_TRAP_0E
                  DC.L    NEW_TRAP_0F

* [Effectively the same is as the corresponding routine in ROM]

NEW_ADDR_ERROR    ADDQ.L  #8,A7                         Skip over extra words on stack
                  RTE                                   Return from this exception

NEW_ILL_INSTR

NEW_PRIV_VIOL

NEW_INT_LEV_7     MOVE.W  #$2700,SR                     Set 'supervisor' mode, mask-out interrupts 1-6 and clear 'user' flags
                  MOVEA.L SV_RAMT,A5                    Signal value for top of RAM needed by warm restart
                  LEA     VIDEO_RAM,A0
                  MOVE.L  #SV_STACT-VIDEO_RAM-1,D0      Number of bytes to top of fixed system variable 
AGAIN             CLR.B   (A0)+                         Clear a byte (also provides a delay to stop other lev. 7 interrupts)
                  DBF     D0,AGAIN                      Repeat until counter exhausted
                  MOVE.L  #WARM_RESTART_JS,A0           Point to warm restart routine in 'JS' ROM
                  JSR     (A0)                          Do it!

                  END

You may need, to change this line to correspond to your system ROM version.
MOVE.L #WARM_RESTART_JS,A0

It seems that SuperBASIC listing in the article does not have the right codes in DATA statement, at least they do not correspond to the assembly listing.

I have added DATA statements with correct codes, corrected SuperBASIC listing follows below, it is for JS ROM QL.

Code: Select all

100 REMark  SuperBASIC program to implement MC68008 program in Listing 1
110 REMark         COPYRIGHT (c) April 1985, Alan Turnbull, B.Sc.
120 :
130 LET reserved_address=RESPR(160)
140 RESTORE 
150 LET address=reserved_address
160 REPeat read_and_store_data
170   IF EOF THEN EXIT read_and_store_data
180   READ data_item
190   POKE_W address,data_item
200   LET address=address+2
210 END REPeat read_and_store_data
220 CALL reserved_address
230 STOP
240 :
250 DATA 17402, 42, 16890, 114, 9032, 0, 16890, 110
260 DATA 9032, 4, 16890, 102, 9032, 20, 16890, 94
270 DATA 9032, 28, 29184, 28679, 20033, 20085, 0, 0
280 DATA 0, 0, 0, 94, 0, 94, 0, 94
290 DATA 0, 0, 0, 94, 0, 0, 0, 94
300 DATA 0, 94, 0, 94, 0, 94, 0, 94
310 DATA 0, 94, 0, 94, 0, 94, 0, 94
320 DATA 0, 94, 0, 94, 20623, 20083, 18172, 9984
330 DATA 10873, 2, -32736, 16889, 2, 0, 8252, 0
340 DATA -31617, 16920, 20936, -4, 8316, 0, 460, 20112

Cheers
Tom

Posted: **Sat Nov 14, 2015 9:46 pm**

Hi Tom,

Looks like the original code assumed that the System Variables and Screen Base address stay in the same place.

MT.INF (or DO.SMSQ) should been used to find the System Variables address. The Screen Base address is found in the Channel Block.

If Minerva with the second screen enabled, the code will fail.

Posted: **Sat Nov 14, 2015 10:00 pm**

Derek,

I will try to improve the code further, I am just wondering if it would be also possible to somehow restore system variables after restart.

Tom

The Sinclair QL Forum

SERial line reset

SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset

Re: SERial line reset