problem when read internal memory

problem when read internal memory

[Alan Wang]

Hi, all
  we're porting linux-2.2.13 to our customized board(MPC823e). I modified some head file according
the specification of our board such as memory map. the IMMR is set to 0xff000000. now, the kernel
can be loaded by bootloader(written by ourselves). but inm8xx_cpm_reset() which is called in
setup_arch:
void m8xx_cpm_reset(uint host_page_addr)
{
	volatile immap_t	 *imp;
	volatile cpm8xx_t	*commproc;
	pte_t			*pte;

	imp = (immap_t *)IMAP_ADDR;
	commproc = (cpm8xx_t *)&imp->im_cpm;
 this line generates a machine check exception when write/read to internal memory.

why? any suggestion and comment are welcome. thank you.

A.W.

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

Re: problem when read internal memory

[Daris A Nevil]

Alan,

It sounds as though you have already changed the value of IMAP_ADDR in
the header file for your specific board (found in linux/include/asm).
You must also make sure that your processor's boot code sets the IMMR
register to this same value before booting the Linux kernel.  Your boot
code should also setup an MMU entry for the IMMR's address by writing to
the MD_EPN register (MMU Data Effective Page Number Reg) of the PPC.
For an example of how this is done you can refer to the boot code for
the QS850 (MPC850-based board).  The source is located at
ftp://qslinux.org.  Once you uncompress the tarball you can go to
snmc/qslinux/qs850boot and view the file init_qs850.s.  Hope this helps.

Regards,
Daris Nevil
SNMC


Alan Wang wrote:
>
> Hi, all
>   we're porting linux-2.2.13 to our customized board(MPC823e). I modified some head file according
> the specification of our board such as memory map. the IMMR is set to 0xff000000. now, the kernel
> can be loaded by bootloader(written by ourselves). but inm8xx_cpm_reset() which is called in
> setup_arch:
> void m8xx_cpm_reset(uint host_page_addr)
> {
>         volatile immap_t         *imp;
>         volatile cpm8xx_t       *commproc;
>         pte_t                   *pte;
>
>         imp = (immap_t *)IMAP_ADDR;
>         commproc = (cpm8xx_t *)&imp->im_cpm;
>  this line generates a machine check exception when write/read to internal memory.
>
> why? any suggestion and comment are welcome. thank you.
>
> A.W.
>

--
Daris A Nevil
Simple Network Magic Corporation
401 Kentucky Lane
McKinney, TX 75069
214-793-7757
dnevil@snmc.com
www.snmc.com
quickstack.com
qslinux.org

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

Re: problem when read internal memory

[Wolfgang Denk]

In message <3AD48F2E.2C713B89@snmc.com> Daris A Nevil wrote:
>
> register to this same value before booting the Linux kernel.  Your boot
> code should also setup an MMU entry for the IMMR's address by writing to
> the MD_EPN register (MMU Data Effective Page Number Reg) of the PPC.

What makwes you think this is necessary?

> For an example of how this is done you can refer to the boot code for
> the QS850 (MPC850-based board).  The source is located at
> ftp://qslinux.org.  Once you uncompress the tarball you can go to

We  have  PPCBoot   (see   the   CVS   server   at   SourceForge   or
ftp://ftp.denx.de/pub/ppcboot)  booting  on  many,  many  MPC8xx (and
MPC82xx and IBM 40x) systems, but this has never been needed  to  get
Linux  running  (not with any kernel from the 2.2.13, 2.2.14 or 2.4.x
series).


Wolfgang Denk

--
Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-4596-87  Fax: (+49)-8142-4596-88  Email: wd@denx.de
A student of  probability  soon  realizes  that  by  its  nature  the
billion-to-one  chance  crops  up nine times out of ten, and that the
greatest odds boil down to a double-sided statement: it will  happen,
or it will not.         - Terry Pratchett, _The Dark Side of the Sun_

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

Re: problem when read internal memory

[Alan Wang]

Daris,
  thank you for ur suggestion. yes, i have the loader set the IMMR as 0xff000000 too.
and i don't think the machine check exception is caused by TLB problem. If the tlb
entry mapping internal memory doesn't exist, it's a data tlb missing exception. If the
operation violates the storage protection, i think a tlb error exception should be
generated.
  i found that when access a physical addr that doesn't exist will cause a MC exception.
some other reasons to cause MC are in term of bus error. and more? Thx.

A.W.



-----Original Message-----
From: Daris A Nevil [mailto:dnevil@snmc.com]
Sent: Thursday, April 12, 2001 1:07 AM
To: Alan Wang
Cc: Linuxppc-EmbeddedMaillist
Subject: Re: problem when read internal memory


Alan,

It sounds as though you have already changed the value of IMAP_ADDR in
the header file for your specific board (found in linux/include/asm).
You must also make sure that your processor's boot code sets the IMMR
register to this same value before booting the Linux kernel.  Your boot
code should also setup an MMU entry for the IMMR's address by writing to
the MD_EPN register (MMU Data Effective Page Number Reg) of the PPC.
For an example of how this is done you can refer to the boot code for
the QS850 (MPC850-based board).  The source is located at
ftp://qslinux.org.  Once you uncompress the tarball you can go to
snmc/qslinux/qs850boot and view the file init_qs850.s.  Hope this helps.

Regards,
Daris Nevil
SNMC

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