USB host always disconnect after few seconds


ajeandet

Still Fresh
Joined
Jan 14, 2014
Messages
14
Hi,

On my pandora(1GHz edition) I can't get the USB host port running for more than few seconds, for example if I connect a USB key(also tested 2 USB ethernet and one other USB key) I can see it, then if I try to browse on it it disconnect after 5 or 10 seconds. The only way to see it again is to disable and enable the USB host but it work just for few seconds again. I've upadted the firmware to the la st version, I tried with a HUB I got the same results. Is there any known instability on the USB driver?(Don't have this on beagleBone). The dmesg output just says that the device has been disconnected but no error. Is there any power saving mechanism interfering with the USB?

Best regards,

Alexis.
 
Hmm, nope, it should work without any issues. Can you try to not touch it when it's plugged in? Maybe plug it in, reboot and enable USB host carefully.

It could be a bad solder on the port.

I'm not sure if it can be a firmware issue, in that case reflashing should help.
 
The weird thing, is that I initially saw this problem because I wanted to use an USB to ethernet bridge to update the pandora, I never succeed because the dongle always disconnect from USB, at this time the pandora was on a table, so I din't touch it. I tried the opposite, the fun thing is that now I can move a little the USB key it doesn't disconnect anymore even if I try to touch all the buttons or I try to bend a little the pandora, but after 2 minutes without touching it it disconnect again. This time I can see on dmesg errors like:

ehci-omap ehci-omap.0: port 2 reset error -110

and 

hub 2-0:1.0: hub_port_status failed (err = -32)

best regards,

Alexis.
 
The weird thing, is that I initially saw this problem because I wanted to use an USB to ethernet bridge to update the pandora, I never succeed because the dongle always disconnect from USB, at this time the pandora was on a table, so I din't touch it. I tried the opposite, the fun thing is that now I can move a little the USB key it doesn't disconnect anymore even if I try to touch all the buttons or I try to bend a little the pandora, but after 2 minutes without touching it it disconnect again. This time I can see on dmesg errors like:

ehci-omap ehci-omap.0: port 2 reset error -110

and 

hub 2-0:1.0: hub_port_status failed (err = -32)

best regards,

Alexis.
What dongle are you using? I had similar expereiences with a cheap and nasty Apple lookeylikey thingy. After finally getting a pucker Apple one, everything is/was fine....

How does the USB port behave if you plug a USB storage device into it?
 
One more precision it seems very correlated with the USB traffic or the driver use, if I am lucky I can see it connected for 1 minute, but as soon as I transfer data on the key it disconnect. When searching for this problem on google some people are talking about a software issue https://groups.google.com/forum/#!msg/beagleboard/1QZ-FcNKGHs/tN1gxQDk1I4J others about hardware issue (due to tolerances) but fixable by software. I have no idea if this is true, I can try some newer kernels if there is some configs to get at least the screen and on console output.
 
The weird thing, is that I initially saw this problem because I wanted to use an USB to ethernet bridge to update the pandora, I never succeed because the dongle always disconnect from USB, at this time the pandora was on a table, so I din't touch it. I tried the opposite, the fun thing is that now I can move a little the USB key it doesn't disconnect anymore even if I try to touch all the buttons or I try to bend a little the pandora, but after 2 minutes without touching it it disconnect again. This time I can see on dmesg errors like:

ehci-omap ehci-omap.0: port 2 reset error -110

and 

hub 2-0:1.0: hub_port_status failed (err = -32)

best regards,

Alexis.
What dongle are you using? I had similar expereiences with a cheap and nasty Apple lookeylikey thingy. After finally getting a pucker Apple one, everything is/was fine....

How does the USB port behave if you plug a USB storage device into it?
I tried apple one and Asus(the one given with zenbook), both uses Asix chip.

The previous error is given with a 32GB usb key.

With Asix AX88772B USB2ETH chip I get more interesting errors like:

hub 1-2:1.0: port 2 disabled by hub (EMI?), re-enabeling...

-> never reconnected.
 
With suspend that might be normal (haven't tested this myself), but it sure seems like there is something going wrong when a lot of traffic happens...
 
With suspend that might be normal (haven't tested this myself), but it sure seems like there is something going wrong when a lot of traffic happens...
I found this patch which seems interesting and not in the current pandora kernel:

http://www.spinics.net/lists/linux-omap/msg64931.html

I'm going to try it, but it really seems to be a current bug on DM3730, I hope it will work.

I will keep you informed,

Alexis.
 
Finally, I did the naive way, I've cloned the kernel source from pandora git, I just apply this patch. I didn't follow the build instructions(I didn't saw it).

I just did this:

make ARCH=arm omap3_pandora_defconfig

patch -p0 < ../0001-Fix-sprz319-erratum-2.1.patch

make ARCH=arm CROSS_COMPILE=arm-linux-gnu- -j8 uImage

make ARCH=arm CROSS_COMPILE=arm-linux-gnu- INSTALL_MOD_PATH=out modules

make ARCH=arm CROSS_COMPILE=arm-linux-gnu-  INSTALL_MOD_PATH=out modules_install

Then I copied the result on the pandora flash and I overwritten the previous kernel image then reboot the pandora.

then run depmod -ae on the pandora and reboot again.

TADA! Now it works! The USB is working without any trouble now! -> maybe it can also avoid some useless RMA if other people have this trouble.

Here is the kernel I've compiled and the modules.

After this I saw that on your build instructions you apply some patches I would be interesting to check if you finally apply this patch and if yes why it doesn't work.

Best regards,

Alexis.
 
The real question is why no one else is having the issue.
 
Hmm, good question. It only affects 1GHz units and only seems to happen with high traffic. I've only used it for keyboard and joysticks...

Maybe it's something that only happens with one of the last kernels and no one used an USB stick since then, except for loading small files (documents, etc.?)
 
The real question is why no one else is having the issue.
Yes since this issue(DPLL5 frequency drift) is referenced by TI http://www.ti.com/lit/er/sprz319e/sprz319e.pdf , and this seems to be a more or less common issue on DM3730.

EvilDragon what is the number of 1GHz units sold? In my case it append on the pre-installed kernel and the new one, from what I understood this patch isn't integrated in the official kernel at least for the 3.2. What I think is that this patch should be applied on the pandora kernel as this issue a DM3730 issue and must affect some percent of 1GHz units owners, and it doesn't seems to brake anything.

Best regards,

Alexis.
 
Hmm, good question. It only affects 1GHz units and only seems to happen with high traffic. I've only used it for keyboard and joysticks...Maybe it's something that only happens with one of the last kernels and no one used an USB stick since then, except for loading small files (documents, etc.?)
I've just transferred 10GB worth of stuff from 32GB USB stick to one of the SD cards without any issues. I'm running latest SZ 1.60
 
Hmm, good question. It only affects 1GHz units and only seems to happen with high traffic. I've only used it for keyboard and joysticks...Maybe it's something that only happens with one of the last kernels and no one used an USB stick since then, except for loading small files (documents, etc.?)
I've just transferred 10GB worth of stuff from 32GB USB stick to one of the SD cards without any issues. I'm running latest SZ 1.60
On a 1GHz unit? In this case it means that your pandora processor has the normal specs (at least the drift is slower than the actual kernel driver update rate). 
 
I copied Yesterday a 3DO Iso Image and my Pandora freezed while copying but i had overclocked so much i can and thought it was because this.

Ghz Pandora and maximum overlocked for 3DO Emulator Testings.

Will try it this Evening without overclocking and report then.

But a Disconnect was not happening here,only one Freeze.
 
Hi,

Some more information, the DPLL5 issue is present with 26MHz Oscillator as said by Robert Nelson from BeagleBoard community, the Pandora seems to have a 26MHz oscillator(As saw on the PCB). 

My questions are:

   -Will you solve this issue in the official pandora kernel?

   -Do you want me to send back my pandora to check this issue?(I agree to pay for the shipment as a contribution for the community)

If you don't solve this issue in the official kernel my pandora's USB is clearly unusable(for both FS and HS devices), I don't want to spend my time to patch the kernel just for myself.

Once again as said by TI this issue concern just a very small number of DM3730 and OMAP4460 processors, that's why most(all?) of you doesn't have this issue.

best regards,

Alexis.
 
Back
Top