PTP, USB, and the Samsung Galaxy S3

A couple of years or so ago, I wrote a Picture Transfer Protocol host app for RISC OS and sent it to a few people for beta test, but it failed for (IIRC) all of them in different ways, so I shelved it. Recently I got a Samsung Galaxy S3 phone, which sadly doesn’t appear as a Mass Storage device – it only does PTP and MTP. So, over the last few days, I’ve rewritten a large part of it.

But it doesn’t work reliably on the S3.

I have access to a USB analyser – sadly full speed and low speed only, but I own a USB 1.1 hub, so I can restrict any device to full speed so I can use the analyser, which shows me that, very often, RISC OS attempts to send the same message to the S3 hundreds or even thousands of times, always receiving a NAK from the S3. Needless to say, this is a blocking call on RISC OS, so the machine freezes at that point. Although the mouse cursor and the num lock LED work, not even Alt-Break will get control back; it has to be a reset or a power cycle. The good news is that I’ve never seen it corrupt any disc when this happens.

When I use the same app with a Nikon Cooolpix 3100, it simply works. No problem.

I get the same result with my Iyonix (RO 5.18) or my BBxM (RO 5.19 of 2012 August 23).

Running the S3 direct into the Iyonix or the BBxM has never worked, not even once – yet the Nikon never fails.

USBInfo can see the device, so it’s clear that control transfers work in both directions. The transfer that doesn’t is bulk OUT.

One thing that does appear to be important is to wait for bulk OUT data transfers to complete before requesting a bulk IN transfer that gets the data or status requested by the bulk OUT, so I’ve amended the bulk OUT procedure to wait until the transfer completes or there’s a timeout.

In the hope that someone somewhere is able and willing to try it, and shed some light on why it’s failing, I’ve put the code up on my web site at http://davehigton.me.uk/Progs/PTP.htm (note: this is not linked from the main page).

If I can get it to work reliably, I hope to expand its facilities and release it at my usual price :-)

So please, if you can help, download it and try it. If it works for some devices, that makes it worth working on further.

N8 locks the iyonix, as described for your phone

N8 locks the iyonix, as described for your phone

Not the answer I was hoping for, of course, but thank you for trying, Jess.

I should add that I’ve had exactly the same results on the Raspberry Pi as on the Iyonix and BBxM: the Nikon camera (which is full speed only) is fine; the S3 (which is high speed) locks all apps and needs the RPi to be power cycled.

@Dave

I have tried !PTP on an Iyonix with a Canon EOS40D. It is shown in *usbdevices, but running !PTP opens the ‘filer’ window and then stiffs the machine (finger on blue button required). The window shows the partially corrupted text ‘No images’ in what looks like system font at the top. Could try on BB or RPi, but I guess it’s not likely to work.

Edit: I should add that the Canon does not appear as a mass storage device.

Thanks, Chris.

I should add that the Canon does not appear as a mass storage device.

That’s already clear, thank you anyway. !PTP will only run if it sees a PTP device, otherwise it will quit immediately telling the user it hasn’t found one.

Jess and Chris, can you use the USBInfo app to find the speed of your phone/camera and post the result here, please?

It’s curious that using the device at full speed works for me, whereas using it at high speed doesn’t. There should be no difference in how the devices are opened, read or written.

I’ve been looking at some of the RISC OS source code, and it appears that EtherUSB doesn’t open pipes in the way documented on the Iyonix.com web site. The latter shows:

port% = OPENIN"devices#interrupt;endpoint1:usbd3"

But the EtherUSB source appears to call DeviceFS_CallDevice with parameters including:

Devices#bulk;size2049:$.%s

and it appears that the reason code is &80000007. (I don’t know where to look for what operation that is.) Note the “$.” in the device name in EtherUSB, and note the apparent lack of an endpoint specification.

Am I using an incorrect string to open the endpoints? They appear to open, but the bulk OUT only appears to work at full speed; it always returns NAK at high speed.

I’m in despair here. Help would be appreciated – though I know there are very few people who know enough to help.

Class: 0, Subclass: 0, Protocol: 0
Speed: Hi-Speed (480 Mb/s)
IFs: 1
Endpoints: 3
Max packet: 64 bytes (endpoint 0)

Hope that helps.

Hope that helps.

Yes, thanks, Chris, that’s as I expected.

Class: 0
Subclass: 0
Protocol: 0
Speed: Unknown
IFs: 1
Endpoints: 3
Max packet: 64 bytes (endpoint 0)

On my Galaxy SII I can connect it to my IYONIX in two different ways :-
1..Just plug it in directly which I believe it to not look like a mass storage device and therefore doesnt work
OR
2..If I go to settings and then to USB utilities I get the option to connect via mass storage. It is important to touch the ‘Connect to PC’ button before plugging USB cable into phone. Then after touching above button you are requested to plug USB cable in, do this then follow any prompts that come up. This should conect phone to IYONIX as mass storage.

This second way works fine on my GSII in that I get two icons on icon bar, one for phone and one for external sd card. I can browse all files. I havnt tried reading and writing because I had a bad experience a couple of years ago and my Iyonix trampled all over the operating system and phone would not work after this so had to get fixed.

Good Luck
Greg

@Greg:

Sadly the GS3 doesn’t have a mass storage mode. If you Google, you’ll find lots of other people complaining about Samsung having removed it. It does PTP or Media Transfer (maybe MTP, I haven’t really looked). There are add-ons to enable mass storage mode, and I’ve tried one, but I think it needs another hack applied first, which I haven’t done. Certainly the add-on I’ve tried doesn’t work for me.

Are you running RO of at least version 5.18, or something older?

@Jess:

USBInfo tells you the speed of each device. What does it say for your N8?

Iyonix and Konica in PTP Mode:
!PTP runs if the Konica is on and plugged in.
Computer freezes with files window opened.
I think the camera requires resetting afterwards also.

Using Devices# with the serial port I found once that setting timeout to 0 got things moving.

I set this in line 810 of USBCastle and now I get an error:
Invalid wimp operation in this context (error641) at line 14380

Just some observations for you, Dave.

@Dave

I was going to have another play tonight but got sidetracked. However, I did notice that in the Basic file USBCastle, there is a function reading

DEF FNhex2(x%) = “0” +STR$~x%, 2)

Should that not be

DEF FNhex2(x%) = RIGHT$(“0” + STR$~x%, 2)
I’ll have a play tomorrow.

“Computer freezes with files window opened.”
My description here is not really what is happening, there seems to be various things happening to the log file while this apparent desktop freeze (mouse pointer still moves).
The log file leaves from logging readable text and fills up with 2-8 Mb of unuseable (random?) bytes depending on when you turn things off.
I am wondering if trapping non-asci text going to the log file as an error would stop the looping/desktop freeze in the meantime.

Edit: I have just found my Coolpix 5200 works with !PTP so I can now see what the logfile should look like.

Looking at the log generated on my machine, there are 17 lines of genuine log, finishing with

Write bulk: R3 = 0, R4 = 1

and then the garbage starts. It looks as if there are other files incorporated. Eg part of one of my mail fetch log files, some emails, including the encoded attachments, and parts that look like binary files. Length of this about 2MB.

@Ronald:

I’m delighted to hear that your Coolpix 5200 works. It makes it worthwhile updating the version on my web site, which is now able to delete objects as well as getting them.

What speed is the USB interface of your Coolpix? The USBInfo application will tell you.

@Chris:

I also get these long logs containing all sorts of stuff they shouldn’t. However, the log function is only capable of logging strings. I wonder whether opening a file allocates a lump of space for the file, and this is what we see when the machine is restarted?

@Dave
My Coolpix 5200 shows up in USBinfo as Full speed (12 Mb/s)
Regarding the logfile, yes I think it is just the result of the crash with the OPENIN logfile, rather than garbage being appended. Probably not a good thing to be doing to your filesystem.
I got around this by not using OPENIN and appending log$ with OSCLI.

DEF PROClog(log$)
cmd$ = "echo " log$ " { >> ptplog }"
OSCLIcmd$

and the file closes cleanly, it probably closes after every write.

I have narrowed down the point of the crash to line 1090 of USBCastle.
SYS XOS_GBPB%, 4, inpipe%, buf%, tfrlen% TO,, r2%, r3%, r4%; flags%
If I REM it the program will continue until an ordinary error due to lack of data occurrs.
All of the values going in seem to be equivalent to the Nikon ones, so there is a failure with the Konica protocol somewhere.
I’ve tried varying the Devices# string values, but no improvement.
size0; has a similar effect to REMming the XOS_GBPB… line.
No crash, but no data.

I have narrowed down the point of the crash to line 1090 of USBCastle.
SYS XOS_GBPB%, 4, inpipe%, buf%, tfrlen% TO,, r2%, r3%, r4%; flags%
If I REM it the program will continue until an ordinary error due to lack of data occurrs.

Yes, this is the line that gets the data input from the USB Bulk IN endpoint. No data, no crash – but no progress either :-(

I simply don’t understand why the calls appear to behave differently with full speed and high speed endpoints. It shouldn’t be possible.

@Dave
After you mentioned full speed/high speed, I dug out my Belkin USB 1.1 Hub (F5U021). !PTP is now opening up the contents of the (High Speed) Konica. It doesn’t know what to do with the directories, but I guess that’s a finer point.
Edit: Maybe I didn’t scroll down or something, but it is opening all the directories and images 100% now.
I changed the card to one with more than one image also (-:

The model is Konica Minolta Dimage A2 and I’m expecting my Minolta 7D of the same era will be OK also.
It does open up the possibility of camera control which I think is done by PTP.

Buoyed up by reports of success when devices are limited to full speed, I’ve updated the version on my web site to add the Delete function.

@Ronald: Thank you for trying it with a full speed hub. We see the same thing: when devices are limited to full speed, they work.

I don’t think that can be down to timing in the devices themselves, since we’re talking about entirely different devices that exhibit the same behaviour. There has to be something about the RISC OS USB stack.

@Chris: Thank you for pointing out the error in FNhex2. I can only imagine that was caused by some editing accident. It’s fixed in the version currently on my web site.

We know that RISC OS communicates reliably with bulk endpoints of high speed devices in the form of USB sticks and hard drives. But it looks as if it doesn’t communicate reliably with cameras and phones at high speed.

What’s the difference?

Is it possible that the USB stack reacts differently when called from user and system space?

Just out of interest – is there a version/build-option for the USB stack that is capable of outputting some low-level debug data, perhaps to the serial port? If so, maybe looking at this could help shed some light on the different behaviour?

FWIW – I have had no end of problems with MTP on my XP-SP3 setup. The usual one is a message that the sort of media I’m attempting to copy is not supported, which is complete rubbish as MXPlayer copes with alsorts, not to mention just plain bizarre quirks like WMA is accepted but MP3 rejected (!), seemingly each device behaves differently.
Therefore I have each device set to work in mass storage mode so I plug it in, get a drive letter, and just draggy-droppy the files without the bull (although the Xperia Mini Pro is a bit prone to crashy-crashy if you copy large (>200MiBish) files and it needs to play with the filesystem due to FAT fragmentation). I know this isn’t terribly helpful to why slow works and faster doesn’t; I’m just wondering if protocols such as MTP have a lot of “implementation-defined” parts? Don’t even get me started on the Creative Zen, it is MTP only, but connecting it shows an empty device, it will only work with its own special driver (which causes Explorer to be somewhat less than stable; I know I can point the finger at it because a rollback to before installation was fine, but after installation of the driver, Explorer tends to randomly freeze (task kill, reload, not painful just annoying) – thankfully this is the NAS box so I don’t use Explorer much directly (usually via VNC/share instead), it’d be intolerable on a main machine).
With these experiences, I’d just tend to avoid something that can’t work as a mass storage device…or pull out the memory card and pop it into a regular card reader!