256GB mSATA SSD for iPod 5g/5.5g - possible with current release of Rockbox?

[DMinor]

Thanks Trone for the update and your efforts are greatly appreciated. Very encouraging news.

[DMinor]

OK, I have it booting and able to play music and generally operate, by just disabling the sanity check for whether the ATA registers are writable/readable correctly. It looks like the SATA interface doesn't implement reading back of the four address registers properly, and this is not normally important in usage of parallel ATA devices that aren't cdrom drives, which is why the original firmware works.

It still doesn't work properly, though: writing to the drive in Rockbox mostly doesn't work. It manages to save its settings, playlist control files, etc, but initialising the database hangs when it tries to write the database to disk, and while reading the device over usb works fine, you can't write to it.

So, I'll need to look at it some more, but this is a good start

If you want to try this on your own device, comment out the call to check_registers() in firmware/drivers/ata.c and build; that was enough to get the one I have here to boot. But, don't expect it to work properly.

Trone, we can't wait to hear the ultimate news on the mSATA.   

My dream is to have one affordable 256GB mSATA SSD ipod 5.5g of course Rockbox'ed.

[torne]

The communication with the SSD is completely unreliable no matter what I do, it seems; data are corrupted randomly and writing rarely works, and even the basic identifying information about the SSD such as the model name is corrupted randomly when read :/

I need to do some more experiments, but it's really not obvious what the issue is here.

[crashtack]

:-(

[SpawnHyuuga]

Does anybody have the Super Talent DuraDrive ZT2 installed in an iPod? The reviews on Amazon for the drive itself are rather shaky at best, and I don't want to sink $200 on something that'll fail in 3 months, especially if it's difficult to fix, repair or replace.

I'm highly considering a KingSpec SSD instead, as it's much cheaper.

[Alkor]

torne
Ok ppl still w8 u. So good luck 

[darrysmi]

The communication with the SSD is completely unreliable no matter what I do, it seems; data are corrupted randomly and writing rarely works, and even the basic identifying information about the SSD such as the model name is corrupted randomly when read :/

I need to do some more experiments, but it's really not obvious what the issue is here.

Thanks Torne, progress is progress   Thanks for investigating, hopefully its something simple...

[dreamlayers]

Try different timings in IDE0_PRI_TIMING0. Look in firmware/target/arm/pp/ata-pp5020.c. The register is set in ata_device_init() and below the function there is a comment with some other values.

Rockbox has used 0x10 from before I worked on these files. I never saw iPod 5G OF using that value. When I tried various PIO timings on my 5G 30GB iPod with stock HD, 0x10 gave the best performance, so I left it that way.

In firmware/drivers/ata.c, the first setting in set_features() sets the PIO mode. The OF would match that to the IDE0_PRI_TIMING0 value. You could try setting a slower mode, but that shouldn't be necessary.

[torne]

Aha, thanks for your expertise dreamlayers

I will experiment with that later.

[torne]

Well, that works, it seems. Left it on PIO 4, and changed the timing to the one from pio80mhz[4] (0x3131) and lo and behold check_registers now passes and nothing is corrupted any more.

Do we have any idea what the timing value actually means here? I could change it to 0x3131 for everyone but it seems like we should maybe copy the OF better and use the one that matches the PIO mode we are using?

[DMinor]

Well, that works, it seems. Left it on PIO 4, and changed the timing to the one from pio80mhz[4] (0x3131) and lo and behold check_registers now passes and nothing is corrupted any more.

Do we have any idea what the timing value actually means here? I could change it to 0x3131 for everyone but it seems like we should maybe copy the OF better and use the one that matches the PIO mode we are using?

You are the man Torne, and thanks for the great news. Words can't express my appreciation for what you (and the entire RB team) have been doing for so many RB lovers including myself.

[dreamlayers]

Well, that works, it seems. Left it on PIO 4, and changed the timing to the one from pio80mhz[4] (0x3131) and lo and behold check_registers now passes and nothing is corrupted any more.

Great! It's nice to know you got that working.

Do we have any idea what the timing value actually means here? I could change it to 0x3131 for everyone but it seems like we should maybe copy the OF better and use the one that matches the PIO mode we are using?

This is based on iPod 5G startup code in flash. I think these are timing values for PIO modes 0 through 4, based on how the function using them is written.  The OF first uses value 0 and then after setting PIO mode 4 via the set feature command switches to value 4 (0x3131).  For mwdma and udma, the OF actually checks what the drive supports and sets values based on that, but no detection is performed for PIO.

Support for PIO 3 and 4 can be detected from identify device info and Rockbox does detect them and set 2, 3 or 4 appropriately. This could be used by defining ATA_SET_PIO_TIMING and implementing ata_set_pio_timings with IDE0_PRI_TIMING0 = pio80mhz[mode]. I think it's unlikely that any drives not supporting PIO 4 would be encountered, but this might as well be used since the supporting code is already present. In ata_device_init, IDE0_PRI_TIMING0 = pio80mhz[0] should be used to match the OF. Using the 80 MHz value for PIO0 at a lower CPU frequency is probably fine.

I think this would be a good thing, because of how various adapter boards are popular now and problems like this will come up again. I suspect 0x10 violates the spec but works with most storage devices because the signal path is so short.  The performance decrease is small even with PIO, and now most reads are done via DMA.

[torne]

OK, I'll go ahead and do it like that.

However, while Rockbox appears to work correctly including both reading and writing files, plugging in USB hangs the device immediately. I'm not sure if this is related or not, I'll need to experiment more :/

[DMinor]

OK, I'll go ahead and do it like that.

However, while Rockbox appears to work correctly including both reading and writing files, plugging in USB hangs the device immediately. I'm not sure if this is related or not, I'll need to experiment more :/

By the way, have you tried using an external amp with the ipod? Would a LOD cable hang the device also?

[torne]

Posted a patch for this here: http://gerrit.rockbox.org/r/#/c/469/

There's an ipodvideo test build of this available here if anyone wants to try: http://whitefang.wolfpuppy.org.uk/temp/rockbox-piotimings.zip - if you try this let me know how it works. On the device I have this is not 100% functional yet; it boots and plays audio but test_disk hangs and USB behaves badly.