How to use ppcbug to debug my program with mmu?

How to use ppcbug to debug my program with mmu?

[Gong Zhuo]

Hi:
   I am using MVME2600 SBC and I want to debug my program under PPCbug.
According the PPCbug Firmware Package User's Manual chapter 2, a program
with mmu enable should be debugged using PPCbug if some rules are obeyed.
   Those address ranges
         0x0 -- 0x0fffffff,
         0x10000000 -- 0x1fffffff ,
         0xe0000000 -- 0xefffffff,
         0xf0000000 -- 0xffffffff
   have been mapped from physical address to logical address 1 to 1 using
BAT mothed .  Exception vectors 0x100 , 0x700, 0xc00, 0x2000 and spr272,
spr273, spr274, spr275 are not being used.
   I write the code like this under ppcbug at address 0x40000:
          1.     initial all the BAT register
          2.     isync
          3.     msr | 0x30  ==> msr    # open the mmu
          4.     do sth. else
    But it always not come back to PPC1-bug> status after I use command
'T'to singlestep run step 3.
    I also used the mothed like this :
          1.     initial all the BAT register
          2.     isync
          3.     msr | 0x30  ==> srr1    # open the mmu
          4.     logical address of 6. => srr0
          5.     rfi
          6.     do sth. else
    The same error happened when singlestep 5.

   Can this code be debugged under ppcbug? If can , how?


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: How to use ppcbug to debug my program with mmu?

[Gabriel Paubert]

On Fri, 21 Jul 2000, Gong Zhuo wrote:

>
> Hi:
>    I am using MVME2600 SBC and I want to debug my program under PPCbug.
> According the PPCbug Firmware Package User's Manual chapter 2, a program
> with mmu enable should be debugged using PPCbug if some rules are obeyed.
>    Those address ranges
>          0x0 -- 0x0fffffff,
>          0x10000000 -- 0x1fffffff ,

One BAT would be enough for this...

>          0xe0000000 -- 0xefffffff,
>          0xf0000000 -- 0xffffffff
>    have been mapped from physical address to logical address 1 to 1 using
> BAT mothed .  Exception vectors 0x100 , 0x700, 0xc00, 0x2000 and spr272,
> spr273, spr274, spr275 are not being used.
>    I write the code like this under ppcbug at address 0x40000:
>           1.     initial all the BAT register
>           2.     isync
>           3.     msr | 0x30  ==> msr    # open the mmu

Wrong order, isync should be performed after changing the msr, mtmsr is
execution synchronizing but not context synchronizing.

>           4.     do sth. else
>     But it always not come back to PPC1-bug> status after I use command
> 'T'to singlestep run step 3.
>     I also used the mothed like this :
>           1.     initial all the BAT register
>           2.     isync
not necessary (you are still in real mode)

>           3.     msr | 0x30  ==> srr1    # open the mmu
>           4.     logical address of 6. => srr0
>           5.     rfi
>           6.     do sth. else
>     The same error happened when singlestep 5.
>
>    Can this code be debugged under ppcbug? If can , how?

No, don't believe Motorola's docs about this. I tried it and PPCBUG does
not work with MMU enabled, even if the only thing you do is 1:1 mapping as
required in the docs (actually it seems PPCBUG will set the MSR IP bit to
1 and then the interrupt handlers you have installed will never be used).

Been there, done that when debugging prepboot, once you've enabled the MMU
you are on your own. I was using standard paging, you are using BATs but
there are not enough BATs for everybody: PPCBUG would need
- a BAT for RAM (2 if you have more than 256Mb since PPCBUG data is at
top of memory)
- a BAT for code and data in the ROM
- a BAT for I/O at 0x80000000
- a BAT for MMIO at 0xc0000000 in some cases (depending on what PPCBUG
functions you call and whether you have a VGA console)...
- perhaps one BAT for the Raven/Falcon bridge (can't remember) it can be
shared (data only but caching disabled) with the ROM BAT.

Note that there is not much point in enabling the MMU until you enable the
data cache. In your case, you might get it to work with the I/O BAT at
0x80000000 but it basically can only be used as a temporary workaround for
debugging.

	Regards,
	Gabriel.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: How to use ppcbug to debug my program with mmu?

[Gong Zhuo]

----- Original Message -----
From: Gabriel Paubert <paubert@iram.es>
To: Gong Zhuo <gongzhuo@163.net>
Cc: Linux PPC Mailing List <linuxppc-embedded@lists.linuxppc.org>
Sent: Friday, July 21, 2000 5:17 PM
Subject: Re: How to use ppcbug to debug my program with mmu?


>
> On Fri, 21 Jul 2000, Gong Zhuo wrote:
>
> >
> > Hi:
> >    I am using MVME2600 SBC and I want to debug my program under PPCbug.
> > According the PPCbug Firmware Package User's Manual chapter 2, a program
> > with mmu enable should be debugged using PPCbug if some rules are
obeyed.
> >    Those address ranges
> >          0x0 -- 0x0fffffff,
> >          0x10000000 -- 0x1fffffff ,
>
> One BAT would be enough for this...
>
> >          0xe0000000 -- 0xefffffff,
> >          0xf0000000 -- 0xffffffff
> >    have been mapped from physical address to logical address 1 to 1
using
> > BAT mothed .  Exception vectors 0x100 , 0x700, 0xc00, 0x2000 and spr272,
> > spr273, spr274, spr275 are not being used.
> >    I write the code like this under ppcbug at address 0x40000:
> >           1.     initial all the BAT register
> >           2.     isync
> >           3.     msr | 0x30  ==> msr    # open the mmu
>
> Wrong order, isync should be performed after changing the msr, mtmsr is
> execution synchronizing but not context synchronizing.
>
> >           4.     do sth. else
> >     But it always not come back to PPC1-bug> status after I use command
> > 'T'to singlestep run step 3.
> >     I also used the mothed like this :
> >           1.     initial all the BAT register
> >           2.     isync
> not necessary (you are still in real mode)
>
> >           3.     msr | 0x30  ==> srr1    # open the mmu
> >           4.     logical address of 6. => srr0
> >           5.     rfi
> >           6.     do sth. else
> >     The same error happened when singlestep 5.
> >
> >    Can this code be debugged under ppcbug? If can , how?
>
> No, don't believe Motorola's docs about this. I tried it and PPCBUG does
> not work with MMU enabled, even if the only thing you do is 1:1 mapping as
> required in the docs (actually it seems PPCBUG will set the MSR IP bit to
> 1 and then the interrupt handlers you have installed will never be used).
>
> Been there, done that when debugging prepboot, once you've enabled the MMU
> you are on your own. I was using standard paging, you are using BATs but
> there are not enough BATs for everybody: PPCBUG would need
> - a BAT for RAM (2 if you have more than 256Mb since PPCBUG data is at
> top of memory)
> - a BAT for code and data in the ROM
> - a BAT for I/O at 0x80000000
> - a BAT for MMIO at 0xc0000000 in some cases (depending on what PPCBUG
> functions you call and whether you have a VGA console)...
> - perhaps one BAT for the Raven/Falcon bridge (can't remember) it can be
> shared (data only but caching disabled) with the ROM BAT.
>
> Note that there is not much point in enabling the MMU until you enable the
> data cache. In your case, you might get it to work with the I/O BAT at
> 0x80000000 but it basically can only be used as a temporary workaround for
> debugging.
>
Hi:
   I have used your mothed and it works very well . Thank you very much.
Now, I want to use standard paging, but I don't know the really size of each
memory area of PPCbug I shoud map. (0x80000000 --? , 0xc0000000--?,
0xF0000000--?), would you please tell me where can I find the description of
each area?

Regards
Gong Zhuo.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/