LeapFrog Pollux Platform: DFTPdevice
DFTPdevice is a custom server LeapFrog includes with the Explorer and LeapPad. This is the main interface between LFConnect, and these devices. DFTPdevice seems loosely based on an FTP server, while including customized features for the purposes of updating the firmware, and a general maintenance.
There are currently 2 versions of DFTPdevice, 1.12, mostly used in Surgeon, and 1.8 for firmware. The Explorer and LeapPad version, while containing the same version number, do seem to differ, most notably in that if you disconnect from the Explorer version it reboots, while the LeapPad version goes back to broadcasting its address, waiting for another connection.
|Firmware < 188.8.131.528|
|Firmware >= 184.108.40.2068|
|Firmware < 220.127.116.1130|
|Firmware >= 18.104.22.16830|
The host PC and LeapFrog device communicate via a USBLAN driver. Initial setup is done with Avahi on the device and Linux host, or Bonjour on Windows. To complete the connection on Windows, you will need to add an IP route to the device. When the device starts it opens port 5000, once you make a socket connection, you then need to send the ETH command and wait a second or two. The device then opens up a socket to port 5001, make your second connection to this port, and you can then start communicating with it.
5000 is the port you send commands to.
5001 is the port you read the responses from.
The host PC and LeapFrog device use a USB Mass Storage driver in this version, instead of the USBLAN driver. While the commands and workings of DFTP are fairly similar, the communication connection is a vast departure from Version 1.x. The USB Mass Storage drive shows up as roughly 50mb, but has no file system or partitions. It is essentially just a port to communicate with. Commands and data are written to, and read at two LBAs on the disk, 0x20 and 0x40, with transfer lengths of 0x01 and 0x14 respectively.
Write(10) and read(10) SCSI commands are used exclusively. There is one vendor specific command that is sent, C2 with CDB length of 10, all 0x00, but seems to have no effect on the device. Write(10) and Read(10) are used in pairs for most commands, uploading or download data is the exception read or write is repeated until data sent.
If you write to LBA 0x20, you must read your response from it also. Transfer Length dictates the size of packets you send. Regular commands are 0x01 or 512 bytes, large data is transferred in 0x14 or 10240 bytes. All data must be aligned to this size, padded with 0x00 when necessary. When done a 101 EOF:<padding size> command is sent/received, to determine if the last packet needs to be trimmed.
The success code is sent for most commands to signify everything went okay. Will not be sent for data transfers, except on the 101 EOF packet.
File transfer ACK. This is returned when the device is receiving data from an upload, it also needs to be send when downloading from it. The number seems to be the size of the last amount of data read.
100 ACK <num>
EOF is the marker to signify all data has been sent, and to stop expecting it. When receiving data you need to check for this packet, and when uploading data you need to send it when finished.
Server errors generate a response back that start with 5xx and then the error text.
500 Command not found
504 Unknown Setting
The command list is fairly undocumented, there is a few listed in the sources, but most will be discovered by other means, like using Wireshark. This is the list so far.
All commands must have a trailing 0x00 attached. In python it would look something like.
In OpenLFConnect there is a special command send. It allows you to send any command you like, the 0x00 is appended for you, otherwise its all raw read/write. Obviously STOR, RETR, RUN won't work with it, but the rest you can play around with.
remote>send LIST /path/to
Some commands are Surgeon only.
DFTP Version 1.12
SETS MOUNTPATIENT=0,1, or 2
0 Unmounts both
1 Mounts Bulk and Rootfs
2 Mounts Rootfs
ETH <host ip> 1383
Action: Has to do with establishing a connection.
Must be sent before opening the second port.
<directory path> list
D 00000000 755 root:root
Notes: Seems to not always work as expected.
Action: Initiate file upload.
Next: Send file data till EOF then send 101 EOF.
Device responds with a lot of 100 ACK: <number> (Seems to be last amount of len(data) received.)
Action: Ends STOR data transfer
Notes:Must flush receive buffer first
Action: Downloads file.
Next: Read socket buffer for data and send back 100 ACK: <number> (Seems to be last amount of len(data) received.)
Ends: When 101 EOF is received.
Action: Removes file
Notes:Found listed in source code, does something with firmware.
a number, that doesn't seem to make sense
Action: Sets time.
DATETIME=20120126083014 200 OK
Note: This is the serial number of the device.
Action: Runs shell script
Next: Send data then 101 EOF
Note: Make sure to remove any \r line ends as it will cause an error.
#!/bin/sh /etc/init.d/telnet start
Would start the telnet client.
GETS SCRIPT_RUNNING Returns:
Note: If script is still running returns 1
Reboot after update or in general
Reboot to usb mode
DFTP's main purpose is for Firmware updating by LFConnect. Basically there is a set of special folders created on the device by dftpdevice, which are uploaded to. This triggers the upload to be written to the NAND, instead of to Surgeon's RAM file system. For the Explorer DFTP is used exclusively for this purpose. While the LeapPad uses a program called Fuse-Flasher that can be found in [source]/packages.
The Explorer uses a combination of the file names, and a file system map file on the device itself, to allow dftpdevice to successfully update the firmware.
Firmware File Names
The firmware file names are used by dftpdevice to figure out where to flash the data to on NAND. This will write the data to that location regardless of the fs map. But will cause issues with UBI images if not coordinated with the map. 131072 is the PACKETSIZE returned in the INFO command.
Format: [NAND Address],[Partition Size/131072],[Filename]
1048576,8,FIRST.32.rle 2097152,64,kernel.cbf 10485760,688,erootfs.ubi 100663296,15616
Device FS Map
On the device itself there is a list of files in /var with extension .fs. These numbers are used in the creation of the mtd/ubi devices required for mounting and flashing. To figure out which version your firmware is using, take a look at /etc/init.d/recover-mounts. In that script there is a variable fs, near the beginning, that shows which file to use.
Format: [NAND Address] [Partition Size] [Filename]
0x00000000 0x00100000 payload/emerald-boot.bin 0x00100000 0x00100000 payload/FIRST.32.rle 0x00200000 0x00800000 payload/kernel.cbf 0x00a00000 0x05600000 payload/erootfs.ubi 0x06000000 0x7a000000 payload/bulk.ubi
To change the partitions, a few things need to be accomplished before configuring dftpdevice. The kernel has the partitions hard coded into it, those will need to be changed. And if you move the kernel, Emerald Boot will need to be compiled with the new kernel address and size.
Once those are taken care of, you will need to calculate the file names, and create the new .fs file. From my experience the .fs is most critical for UBI, as once Emerald Boot was compiled with the proper kernel address, it loaded fine. But for UBI there are certain steps required for flashing to NAND, like erase counter management and it's lack of OOB usage, that need to be dealt with. dftpdevice will write the UBI image where ever the file name numbers tell it too, but it will not be usable by the kernel. If you get an error 22 manually mounting it, this is probably the reason. An MTD and UBI device needs to be created for proper flashing, which dftpdevice creates the MTD devices using the .fs file information. For more application of this, look in /etc/init.d/recovery-mounts.