Skip to contentSkip to conversation

mpc8xx-2.3.10 configuration questions.

DISCUSSION Brendan Simon 4 participants 8 messages Sep 15, 1999 – Sep 17, 1999

lore
Share & tools
lore
#0 lore raw

I have managed to compile a zImage for the BSE target. Great. It took a lot of trial and error to find out which options to leave out. The serial and serial-console support is an obvious one which seems natural to include but causes the build to fail.

It fails to build if I include any of the following options. I can't remember the exact error messages but can post them later if it helps.
* CPU is too slow to handle full bandwidth.
* MS-DOS, VFAT filesystems.
* SMB network filesystems.

Has anyone had similar problems ? Must other options be included for any of these to work ?

I notice that the .config file has both 8xx and 6xx cpus defined even though I've chosen 8xx/860. Similarly, CONFIG_PARIDE_PARPORT=y even though I don't enable it. It looks as though there are some defaults that are set regardless of what the user sets. This worries me a bit, especially having 8xx and 6xx cpus both defined. Is this intentional ?

Thanks,
Brendan Simon.

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

#1 lore raw
Quoted 3 lines · Brendan Simon

It fails to build if I include any of the following options. I can't remember the exact error messages but can post them later if it helps.
* CPU is too slow to handle full bandwidth.

The cpu isn't too slow, so don't select this one.

Quoted 2 lines

* MS-DOS, VFAT filesystems.
* SMB network filesystems.

These are the result of generic VM changes and are not at all related to 8xx. You can't build them on any PPC machine in that release.

Quoted 4 lines

I notice that the .config file has both 8xx and 6xx cpus defined even though I've chosen 8xx/860. Similarly, CONFIG_PARIDE_PARPORT=y even though I don't enable it. It looks as though there are some defaults that are set regardless of what the user sets.

This has all been corrected in later releases. I don't think any of it was 8xx related as I didn't specifically change anything.

Quoted 2 lines

.....This worries me a bit,
especially having 8xx and 6xx cpus both defined. Is this intentional ?

Well, you can either worry or fix it and get on with life. Keep in mind that the 8xx is really a minimal change to the Linux/PPC baseline. If something doesn't work on the 8xx, and it is not specific to the 8xx (like the file system stuff), chances are it doesn't work for any PPC release. Build it for a PMAC or something (you do have some other PPC besides the 8xx, right?) and compare the results. Remember, the 2.3.x CVS is a serious development tree, changing every minute, and can go for many releases without running properly in many aspects. When you play in this sandbox you have to expect things not to work correctly, and contribute to the updates.

People have full multi-user, multiple network, PPP, web server stuff running on 8xx platforms....and I never saw them post a message to any of the PPC lists.....unfortunately no code updates were posted either....

	-- Dan

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

#2 lore raw
Quoted 9 lines · Dan Malek
Quoted 2 lines

.....This worries me a bit,
especially having 8xx and 6xx cpus both defined. Is this intentional ?

Well, you can either worry or fix it and get on with life. Keep in mind that the 8xx is really a minimal change to the Linux/PPC baseline. If something doesn't work on the 8xx, and it is not specific to the 8xx (like the file system stuff), chances are it doesn't work for any PPC release. Build it for a PMAC or something (you do have some other PPC besides the 8xx, right?) and compare the results.

No, unfortunately I do not have access to a powermac. My x86 laptop is just about getting to the end of it's warranty and I am looking at upgrading in the next month or so. Since I have no desire to run DOS/Windows programs AND the powerpc G3/G4 machines kick ass, I will probably upgrade to a PowerBook :)

Quoted 5 lines

Remember, the 2.3.x CVS is a serious development tree, changing every minute, and can go for many releases without running properly in many aspects. When you play in this sandbox you have to expect things not to work correctly, and contribute to the updates.

I do expect things not to work. I do not know the kernel internals very well and I am not confident in changing things willy nilly. Reporting things to the mailing list seems reasonable to me to see if others have seen similar things and have found solutions. I don't mind getting my hands dirty and trying things but I would rather wait for some direction from others first so I don't head down the wrong track and possibly stuff up something that didn't really have a problem.

In a previous email you mentioned that the embedded-2.2.5 release is still the most stable, but you recommended to me that the 2.3.10 kernels are the place to start for new ports. I think I will do the port for the 2.2.5 first and then migrate it to the 2.3.x stuff later. Does this sound reasonable ?

Brendan Simon.

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

#3 lore raw
Quoted 1 line · Brendan Simon

No, unfortunately I do not have access to a powermac.

I strongly recommend that anyone doing 8xx (or any PPC development) use a PPC host of some type. It is sooooo much easier. No cross development hassles, bytes in the proper world order, all of the tools work......

Quoted 1 line

I do expect things not to work.

As I said, if it isn't directly 8xx related, you can probably read about it on the other lists as well.

Quoted 2 lines

.... I think I will do the port for the 2.2.5 first and
then migrate it to the 2.3.x stuff later. Does this sound reasonable ?

I think so. After working with 2.3.18, 2.2.5 just feels "old". In particular there are some configuration changes that work better in the newer releases.

I wouldn't recommend 2.3.10 for anything right now. In fact, I will probably remove that tar image from the server the next time I visit.

	-- Dan

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

#4 lore raw
In reply to Dan Malek
Quoted 2 lines · Dan Malek

When you play in this sandbox you have to expect things not to work correctly, and contribute to the updates.

Accordingly, I offer the following changes, which were necessary to build a kernel for the RPCg CLLF board with an 860T. (I assume this board has the same software model as the RPX Lite, since there is no CLLF option). Both problems relate to the Fast Ethernet Controller code, which didn't compile:

*** arch/ppc/8xx_io/fec.c.dist Wed Sep 15 17:32:26 1999
--- arch/ppc/8xx_io/fec.c Thu Sep 16 17:31:35 1999
***************
*** 820,826 ****

/* Initialize the FEC Ethernet on 860T.
*/
! __initfunc(int fec_enet_init(void))
{
struct net_device *dev;
struct fec_enet_private *fep;
--- 820,826 ----

/* Initialize the FEC Ethernet on 860T.
*/
! int __init fec_enet_init(void)
{
struct net_device *dev;
struct fec_enet_private *fep;

*** include/asm-ppc/rpxlite.h.dist Thu Sep 16 17:30:04 1999
--- include/asm-ppc/rpxlite.h Thu Sep 16 17:31:02 1999
***************
*** 56,61 ****
--- 56,65 ----
#define HIOX_CSR_ENAUDIO ((uint)0x00000200)
#define HIOX_CSR_RSTAUDIO ((uint)0x00000100) /* 0 == reset */

+ /* Interrupt level assignments.
+ */
+ #define FEC_INTERRUPT SIU_LEVEL1 /* FEC interrupt */
+
/* We don't use the 8259.
*/
#define NR_8259_INTS 0

Regards,
Graham

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

#5 lore raw
Quoted 2 lines · Graham Stoney

Accordingly, I offer the following changes, which were necessary to build a kernel for the RPCg CLLF board with an 860T.

The CLLF isn't a really a lite...it's a Classic with less stuff. I am currently working on this now, and if you actually have a CLLF (I don't know how you would) the FEC isn't going to work because of the faulty magnetics on the board (not RPCg fault, the manufacturer acknowledged an error after we found it). The new CLLFs will be available within a few days.

Quoted 1 line

*** arch/ppc/8xx_io/fec.c.dist Wed Sep 15 17:32:26 1999

What is fec.c.dist? It seems like I am always fixing this one. I pretty sure I got it right in the 2.3.18 kernel.

Quoted 4 lines

+ /* Interrupt level assignments.
+ */
+ #define FEC_INTERRUPT SIU_LEVEL1 /* FEC interrupt */
+

The CLLF will be built from the Classic configuration, so it isn't appropriate to put this here. There is also more that needs to be added for the MII interrupts through a Port C bit.

	-- Dan

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

#6 lore raw
Quoted 1 line · Dan Malek

The CLLF isn't a really a lite...it's a Classic with less stuff.

Right; I'm with you now.

Quoted 5 lines

I am currently working on this now, and if you actually have a CLLF (I don't know how you would) the FEC isn't going to work because of the faulty magnetics on the board (not RPCg fault, the manufacturer acknowledged an error after we found it). The new CLLFs will be available within a few days.

No, I don't have one yet; I'm just building the kernel in eager anticipation for when the new CLLF with 860T hits my desk.

Quoted 1 line

What is fec.c.dist? It seems like I am always fixing this one.

It's just what I renamed fec.c from the snapshot to before I hacked it.

Quoted 1 line

I pretty sure I got it right in the 2.3.18 kernel.

I'm using the linux-2.3.18 snapshot kernel from the ftp site, which has it wrong I'm afraid.

Quoted 3 lines

The CLLF will be built from the Classic configuration, so it isn't appropriate to put this here. There is also more that needs to be added for the MII interrupts through a Port C bit.

I've reconfigured/recompiled my kernel for the Classic, but it sounds like it won't Just Work(tm). I'm happy to alpha test any mods you're working on; once the board & doco arrive, I can probably do some kernel hacking as well...

Thanks,
Graham

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

#7 lore raw
Quoted 4 lines · Graham Stoney
Quoted 1 line

I pretty sure I got it right in the 2.3.18 kernel.

I'm using the linux-2.3.18 snapshot kernel from the ftp site, which has it wrong I'm afraid.

Damn it! I forgot to check it in and I don't have the new version here at home with me. I'll fix this and add the Ericsson updates, make sure it works on the Classic and CLLF.

Quoted 2 lines

I've reconfigured/recompiled my kernel for the Classic, but it sounds like it won't Just Work(tm).

The Classic should work, I just never finished all of the FEC when I got stalled on the CLLF. If anyone has a Classic, there is a zero ohm resistor that must be installed to get the MII interrupt to Port C. I'll document this, too.

Quoted 1 line

...I'm happy to alpha test any mods you're working on;

Oh, you will :-).

	-- Dan

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