Marvell 88F5182 SoC support in 2.6.25 vanilla linux kernel !

Sylver · **Posted:** Mon Oct 22, 2007 12:14 pm

Hi,
as some of you may have noticed, the guys at marvell are posting some patches to apply to the vanilla kernel to support the 88F5182 SoC (known as Orion-nas). These patches are available in the "linux arm kernel" mailing list ( http://marc.info/?l=linux-arm-kernel&r=1&b=200710&w=2 ) (I'll upload the files soon). There is even 'native' support for the Kurobox Pro.
We also need to get some other patches for USB/Ethernet/SATA support !
For now I've found the following patches :
- USB Host support : http://sourceforge.net/mailarchive/mess ... stofly.org
- Ethernet support (modification of the existing MV643xx ethernet driver) : 9 posts => 1 2 3 4 5 6 7 8 9
I didn't find the SATA driver yet ...
So I tried to boot this kernel with kurobox pro support and it leads to a boot failing with this only error :

Quote:

Error: unrecognized/unsupported processor variant.

It looks like the "Feroceon" core defined by marvell doesn't fit the CPU defined somewhere (where ?). So I compiled the ARMV926T processor support and It pass this step to failed to the next step => the arm device number is unknown ...
I found that it's because the machine type kurobox pro is defined as n° 1509 (in arch/arm/tools/mach-types) but the number that the kernel get (from u-boot) is 526 ... I corrected that to fit my configuration and now I just have nothing after the kernel loading

I guess that I really need to find why the Feroceon core is not detected correctly !
Feel free to post any of your thought here

Sylver

Edit : the machine type is defined in u-boot (board/mv88fxx81/USP/mv_main.c)

Code:

        /* arch number of Integrator Board */
        gd->bd->bi_arch_number = 526;

Kaiten · **Posted:** Mon Oct 22, 2007 10:43 pm

Hi Sylver, great work your doing. One thing that's got me though.

What's the difference between DB88F5281 & DB88F5182?
I honestly don't know.. Just wondered that's all, as you've stated both in your post & I've seen them in the linked messages.

I've only seen DB88F5281 in relation to LSPro/live in the past.

Sylver · **Posted:** Mon Oct 22, 2007 10:56 pm

Thanks, I'm doing my best for the community, but I'm also learning a lot for my job

5182 is the SoC in the Kurobox Pro/Linkstation Pro.
5281 is another SoC that is almost the same than 5182 but with a hardware floating point unit. It may also have some few other difference but anyway it's not related to our stuff

lb_worm · **Posted:** Tue Oct 23, 2007 1:22 am

This is the SATA interface type code for the HD interface.

Sylver · **Posted:** Tue Oct 23, 2007 10:34 am

lb_worm, I'm afraid not to understand what you're talking about !

lb_worm · **Posted:** Wed Oct 24, 2007 1:23 am

The SoC implementations are emulating either the 5182 or 5281 SCSI devices.

Sylver · **Posted:** Wed Oct 24, 2007 2:51 pm

Hum ok I wasn't aware of that !
I've sent an email to Tzachi Perelstein yesterday and he answered me

I talked about the problems I had with making things run on the Kurobox Pro and told me that the "Error: unrecognized/unsupported processor variant." issue is normal as 5281 SoC are using a new specific ID (that is linked to the feroceon.S definition), but for the 5182 SoC (the one in the kurobox pro), we should use the arm926.S definition (with some Orion specific changes).

lb_worm · **Posted:** Fri Oct 26, 2007 11:31 am

also different processors as you have found out.

Sylver · **Posted:** Thu Nov 08, 2007 10:37 am

from the a-l-k mailing list :

Quote:

Just to let you know that the latest support for the Marvell ARM-based
Orion SOC family aimed for mainline inclusion is now available from the
following Git repository:

git://git.kernel.org/pub/scm/linux/kern ... /orion.git

and can also be browsed from the following gitweb URL:

http://git.kernel.org/?p=linux/kernel/g ... /orion.git
[...]

It looks like I have something to play with

mindbender · **Posted:** Thu Nov 08, 2007 11:26 am

the git stuff looks very promising..especially as it is hosted at kernel.org!

Sylver · **Posted:** Thu Nov 08, 2007 1:23 pm

Well, there are some stuff that are (according to the comments) not to be merged in the vanilla kernel yet (like the SATA driver). Moreover they will update stuff on it to support XOR and CESA engines that can't be merged in the vanilla kernel unless there are some heavy changes in the kernel ...
I'll try the git stuff ASAP, the only thing to do to boot the kernel on a kurobox pro is to change the mach-type file to have the right platform type, and after that, I'll merge the buffalo stuff (MICON V2 ...) in it to have a fully working box

mindbender · **Posted:** Thu Nov 08, 2007 1:27 pm

i ask myself what else needed to be done to boot the kernel on a LS Pro/LS Live?

Sylver · **Posted:** Thu Nov 08, 2007 1:40 pm

Nothing I guess ! Maybe just removing the NAND flash declaration as it's not on the LS Pro/Live ...

goat · **Posted:** Thu Nov 08, 2007 6:49 pm

Sylver wrote:

Well, there are some stuff that are (according to the comments) not to be merged in the vanilla kernel yet (like the SATA driver). Moreover they will update stuff on it to support XOR and CESA engines that can't be merged in the vanilla kernel unless there are some heavy changes in the kernel ...
I'll try the git stuff ASAP, the only thing to do to boot the kernel on a kurobox pro is to change the mach-type file to have the right platform type, and after that, I'll merge the buffalo stuff (MICON V2 ...) in it to have a fully working box

So if I understand that, there isn't going to be inclusion of the SATA driver in the vanilla kernel, but with the SoC support in the kernel, is there a stable SATA driver to be able to patch with? or is that not an option?

bbradley · **Posted:** Thu Nov 08, 2007 7:21 pm

goat wrote:

So if I understand that, there isn't going to be inclusion of the SATA driver in the vanilla kernel, but with the SoC support in the kernel, is there a stable SATA driver to be able to patch with? or is that not an option?

As I understand it the sata_mv driver that is already in mainline should work with minimal changes. The current problem is that sata_mv uses the PCI bus and the stock kernels don't have a proper driver for it (try running lspci). jonli447 mentioned writing a proper PCI driver and then modifying sata_mv to work with the SoC SATA but there's already a PCI/PCIE driver from Tzachi.

The full post about the Orion Git repository is at http://thread.gmane.org/gmane.linux.ports.arm.kernel/35565

I should have some time to do some testing with this on the lspro next week and I'll try and get some more work done on the I2C driver.

Buffalo NAS-Central Forums

Marvell 88F5182 SoC support in 2.6.25 vanilla linux kernel !

Who is online