Hi,

At  0:30 96.5.11 -0400, Nicholas Riley wrote:
>>If it really is the lack of FPU giving you
>>the illegal instruction traps, the kernel itself must be lacking FPE.
>>FPE must be built into the kernel in order to run on an FPU-less
>>machine.
>
>I didn't really understand what Takashi meant:
>
>>This MAY work on the PowerBook 520, 520c, 540, 540c with the FPU emulataion.
>
>I hope (pray?) it will be simple to include the FPE in the PB500 kernel if
>that is all that is needed.

A kernel that I built includes the FPE.
I specified "FPSP" option and "FPU_EMULATE" option in my config file.
Did I forgot to specify any options that are related to the FPE?
I guess that this problem is related to the following message:


At 18:43 96.5.5 +0200, Ignatios Souvatzis wrote:
>Hello,
>
>when checking for things to do for 68060 support integration, I saw
>that in each and every m68k port (checked today after cvs update'ing)
>the following piece of code is found in every locore.s:
>
>_fpfault:
>#ifdef FPCOPROC
>        clrl    sp@-            | stack adjust count
>        moveml  #0xFFFF,sp@-    | save user registers
>        movl    usp,a0          | and save
>        movl    a0,sp@(FR_SP)   |   the user stack pointer
>        clrl    sp@-            | no VA arg
>        movl    _curpcb,a0      | current pcb
>        lea     a0@(PCB_FPCTX),a0 | address of FP savearea
>        fsave   a0@             | save state
>        tstb    a0@             | null state frame?
>        jeq     Lfptnull        | yes, safe
>        clrw    d0              | no, need to tweak BIU
>        movb    a0@(1),d0       | get frame size
>        bset    #3,a0@(0,d0:w)  | set exc_pend bit of BIU
>Lfptnull:
>
>...
>
>This code (or maybe a variation of it) is necessary for the M68881/2
>coprocessor (else we would get the same exception again if the program
>continues to run after servicing the signal); but it is really really
>wrong for 68040 (and 68060) (cf. 68040 UM/AD rev 1, p. 9-41 - 9-42).
>
>I propose something like this:
>
>_fpfault:
>#ifdef FPCOPROC
>        clrl    sp@-            | stack adjust count
>        moveml  #0xFFFF,sp@-    | save user registers
>        movl    usp,a0          | and save
>        movl    a0,sp@(FR_SP)   |   the user stack pointer
>        clrl    sp@-            | no VA arg
>        movl    _curpcb,a0      | current pcb
>        lea     a0@(PCB_FPCTX),a0 | address of FP savearea
>        fsave   a0@             | save state
>|XXX --new code here---
>        test_for_040_or_060     |XXX replace with appropriate code
>        jiftrue Lfptnull        |XXX for your port.
>| on non-060 machines, you can use 
>|       cmpb    #0x41,a0@       | is it the 68040 FPU frame format?
>|       jeq     Lfptnull
>| on Amiga, we would use
>|       movb    _machineid+3,d0
>|       andb    #0x90,d0        | (AMIGA_68060|AMIGA_68040)
>|       jne     Lfptnull        | btw, this has to come before the 
>                                | tstb a0@ below, as the 68060 has a
>                                | different NULL FPU state frame format
>|XXX --- original code continues ---
>        tstb    a0@             | null state frame?
>        jeq     Lfptnull        | yes, safe
>|XXX I'm not sure if the following code is really necessary, I'll
>|XXX check when I get my hands on original Motorola 68881/2 docs:
>        movb    a0@(1),d0       | get frame size
>        and     #0xdf,d0
>        cmp     #0x18,d0        | is it a BUSY frame?
>        jne     Lfptnull
>|XXX --- original code continues ---
>        clrw    d0              | no, need to tweak BIU
>        movb    a0@(1),d0       | get frame size
>        bset    #3,a0@(0,d0:w)  | set exc_pend bit of BIU
>Lfptnull:
>
>Regards,
>        Ignatios Souvatzis


Thanks,

 8[%]8  Takashi HAMADA       Kanazawa Institute of Technology  //
  |#|                        System Design Engineering major   \\
  |#|   =====================================================  //
 (.= )  E-mail: hamada@next.etron.kanazawa-it.ac.jp            \\
 ( =:)          hamada@apricot.etron.kanazawa-it.ac.jp         //