DM365 EVM does not find USB driver

There is a jumper right next to the USB port called J26, this jumper determines the state of the USB ID pin on the port, which is used to determine if the USB port should be used as a host or as a slave. On my board the jumper was set to have the USB port as a slave port by default, however if you are just connecting a flash drive to the EVM you would need to have it setup as a master, therefore you may need to set jumper J26 to pull the ID pin low. After switching J26 and power cycling the EVM I was able to have the kernel recognize a USB flash drive connected to the port through a mini A to standard A USB adapter.

Thanks a lot. I did not find any information about J26 from spectrual digital document.

Before I do anything, I want to ensure not to kill it.

By default, there is a jumper at J26 connecting PIN H and PIN ID.

Should I connect the jumpper to PIN L and PIN ID?

or just take the jumper out?

Best Regards,

Tao

This needs to be set to connect pin L to pin ID, this will pull the ID pin of the USB interface low so that it is configured for host operation, for example see my setup below. Note that the board must be power cycled (or at least the kernel rebooted) before this change will take effect.

yes, it works. I measured the speed by copying a large file.

It is like 4.2 MB/s (bytes). Is it normal or too low?

Anyone has experience with it?

Tao

Unfortunately the PSP datasheet does not currently contain figures for the DM365 USB performance, so it is hard to say if what you are seeing is expected, these figures should be updated in the future. In general you should see higher read values than write values assuming a typical flash USB mass storage device. With a similar device, the DM355, transfers would normally be 6MBps+ assuming an EXT3 filesystem on the USB drive, you may see different performance depending on what file system you are using.

hi,

can you tell how did you determined usb transfer speed.

I believe the measurements here were done by simply timing the transfer of a large file, at least that is how I have measured the bandwidth of mass storage devices like this in the past.

Depending on the make of your USB stick I think ~4MBps is normal.  You can compare the same by measuring it on a windows/Linux PC.  You would probably get around 12-15MBps with a selfpowered USB HDD like IOmega, WD-passport etc.

regards

swami

I was out for travel for a week.[:)]

when I measure the usb speed. A large file like 300Mb is stored in a usb harddrive. I measure the time needed for copy it in the same drive. So the 4.2MB includes both read and write.

Which is around ~9MBps in one direction.  I think that is slightly low but that figure also depends on the USB HDD capabality.  Another way to verify the throughput would be to give this command on the command line

# time dd if=/dev/zero of=/mnt/1gbfile bs=1M count=1024  ------> This would give you the time required for writing a 1gb file to USB HDD media.

# time dd if=/mnt/1gbfile of=/dev/zero bs=1M   -------> This would give you the time required for reading a 1gb file from USB HDD media.

regards

swami

In general you should see higher read values than write values assuming a typical flash USB mass storage device.

Not really on DaVinci hardware. The issue is that the CPPI DMA engine used with USB isn't a very good match for USB, and it hurts *especially* much on reads. The DA8xx chips (OMAP-L13x) evolved that DMA engine in ways that should help. If you want details, see the drivers/usb/musb/cppi C source file; heading "RX woes". (That's the first order problem.)

On other hardware, sure that's correct. Expect read rates of 25 MByte/sec and up; sometimes 35 Mbyte/sec. The bottleneck is usually the USB peripheral.

Dave,

Yes on later platforms linke OMAPL13x the CPPI engine evolved (based on our feedback from DaVinci) and it will evolve further in later platforms.  From performance perspective, we could very well reach ~20MBps on DaVinci (DM644x) similar to the existing platforms.

We have enhanced the DaVinci CPPI DMA implementation (cppi_dma.*) in the LSP1.x releases and had posted a complete patch on the same in the community some time back last year.  We are in the process of breaking down that patch to faciliate acceptance in the community.

I do acknowledge that there is the issue of increased interrupt loading on DaVinci from the perspective of supporting MSC due to a bug in the DMA engine (that got addressed in OMAPL13x) but this definitely does not limit performance to 4MBps.

regards

swami 

just from looking at the results, i would say you are acheiving 48Mbps on the USB port.

you effectively have a 10 bit per byte data stream on the USB, and accounting for overhead, looks like you are running HS