U-Boot for the Pogoplug 4 Series

[davygravy]

Hi to K Mihelich... a question to you, as I seemed to remember that you are the roller of the upgraded U-Boot for Pogo Series 4...

Do I understand correct that it sees only the top port, and not the back 2 ports?

SATA does an odd thing on reset... ? "ide reset" yields odd results?

Just checking, want to make sure I don't have a bad board...

PS. Attached a file I modified ... base on your uboot.bin. I added the standard Kirkwood UART-boot header to it. It came in handy as I was working w/ current source, and wanted to avoid getting out the JTAG adapter.

[kmihelich]

Yes, it can only use the top USB port because that is the only USB 2.0 port broken out from the SoC. The back ports are USB 3.0 coming from a separate controller connected to the PCIe bus. U-Boot does not have any support for that, and I'm not about to spend time writing code for it.

I can't explain ide reset issues, it has worked fine for me. I have noticed soft-reset issues where the SATA port doesn't get properly re-initialized, but in that case a hard reset clears things up. Again, the PPv4 isn't important enough to warrant the time spent to troubleshoot something like that, since there is no data corruption or anything. Worst case scenario is U-Boot just bails out to a prompt and sits there.

[davygravy]

kk, thanks, so I guess my board is OK after all...

I was some patches for booting from SD/MMC, but its cumbersome at best... maybe a project for this summer...

[kmihelich]

I tried SD, but support was really flaky and hacking in support that might work was very much not worth the guaranteed forum headache from people that were having problems.

[davygravy]

At this point, I have it booting from an SD card and from the top-side USB2.0 port. Once booted in Linux, it recognizes the USB3.0 ports on the aft-side.

My unsolved problem is getting the SATA port to be recognized properly and init properly... I've tried different timings in mv_sata.c, checked register/DRAM setting in kwbimage.cfg... really stuck at this point...

From within u-boot:
$this->bbcode_second_pass_code('', 'SoC: Kirkwood 88F6281_A1
DRAM: 128 MiB
WARNING: Caches not enabled
NAND: 128 MiB
In: serial
Out: serial
Err: serial
Net: egiga0
Hit any key to stop autoboot: 0
Pogov4>
Pogov4> ide reset

Reset IDE: ide_preinit failed
')
From the kernel:
$this->bbcode_second_pass_code('', '[ 21.923957] libata version 3.00 loaded.
[ 21.926720] sata_mv sata_mv.0: version 1.28
[ 21.926898] sata_mv sata_mv.0: slots 32 ports 2
[ 21.948460] scsi0 : sata_mv
[ 21.953758] scsi1 : sata_mv
[ 21.954180] ata1: SATA max UDMA/133 irq 21
[ 21.954212] ata2: SATA max UDMA/133 irq 21
...
[ 22.307810] ata1: SATA link down (SStatus 0 SControl F300)
[ 22.657804] ata2: SATA link down (SStatus 0 SControl F300)
')

@ KMihelich: any chance you could post your sources tarball from when you turned on SATA/IDE access? If I saw that, and diffed it w/ the original PogoV4 uboot source, maybe I'd see something stand out as important, something to try differently. Or maybe a special register setting... I'm not looking for help writing/fixing code, just a peek at what code you used to get SATA turned on in you u-boot. [though if you offered to help, I wouldn't turn it down ]

=================
Afterthought: The u-boot as is would work perfectly on the Pogo Mobile, since it the SATA port on it is not populated. It could still boot w/ the SD card slot.

[kmihelich]

All I did was rebuild the CE U-Boot sources, which already had the ide commands enabled. All they did was disable commands, so I re-enabled them (FAT, USB, something else) and fix the ext2load return value so it can be used correctly in conditional statements. I'd just compare whatever U-Boot version you're using against CE, see where it differs.

[davygravy]

Hmmm... the current stable 2011.12 release is what I'm building it off of... the CE source is based on Marvell's source, which is from about 2008 or so, and is very, very, very dissimilar to what is in u-boot's current tree.

Back in 2009-10 when Kirkwood support was added in to mainline, it was all streamline and made a lot cleaner. There is very little of the old code left intact... so diffing CE vs. mainline would be like diffing Linux 2.4 against 3.3. Literally (no exaggeration...)

anyway, thank you...

[kmihelich]

Like I mentioned, for me it was all about expediently providing a solution that worked with a minimal amount of effort. Re-using CE source let me do that, and then writing out a new board file took only a couple hours on top of that. This really isn't a hugely popular or important platform to spend a whole lot of energy on, it's just another plug computer with a measurable amount of less umph.

[davygravy]

np, thanks for your time & effort, you've already done a lot ... the fact that you got SATA to work means that it is possible... I may drop a line to someone @ Marvell to see if they know what the problem might be.

It is interesting that there are _no_ 6182 Kirkwoods in mainline. No submissions to mainline, either. That means the 6182 code is probably much less tested... so... this could actually be a bug. Tthe Kirkwood custodian may be interested in this, or at least have a better idea than I do in terms of where to start looking.

Again, thanks!

[kmihelich]

There really isn't much different from the standard 6281. It's basically just an underclocked variant. Everything else, if I remember correctly, is exactly the same.