acm/ttyACM0/cdc_acm does not deliver data

acm/ttyACM0/cdc_acm does not deliver data

[AW]

Hi!

I made some USB devices using the CDC-ACM proposal...
They work fine with 2.6.38.8-35.fc15/amd64 and earlier.

But since 2.6.40-4.fc15/amd64 it doesnt work as before...


My device receives a byte and then echoes it...
Sometimes when the 65th byte is sent, a read(2) returns with 0...
When I close the device and open it again and send 1 byte, I receive the echo for the 65th byte...
(btw: after 65 bytes, the roles change: my self-tinkered device sends 65bytes and listens to the echoes...)
Funnily some 65-byte-sequences are always echoed as before, while some are not (I tried it at least 10 times; I think the "bad" sequences had a lot of zeroes...)...

On my headless Atom N270 I have 1 self-tinkered USB-ACM-device and 2 DVB-T USB sticks...
When I boot the 2.6.40 kernel it crashes after some minutes...

Can we change that somehow? *wag tail*
Is it safe to use 2.6.38.8-35 behind a NAT? (I mean: r there big security holes?)

PS: I reported it here https://bugzilla.redhat.com/show_bug.cgi?id=727610 , but fedora/redhat cant help me so much, because i m the only one who has that problem...

Thx.

Bye
Arne

Re: acm/ttyACM0/cdc_acm does not deliver data

[Greg KH]

On Thu, Oct 06, 2011 at 12:19:48PM -0700, AW wrote:
> Hi!
> 
> I made some USB devices using the CDC-ACM proposal...
> They work fine with 2.6.38.8-35.fc15/amd64 and earlier.
> 
> But since 2.6.40-4.fc15/amd64 it doesnt work as before...

Can you try a kernel.org release instead of a Fedora kernel?  Does this
work on 3.0.6?  If not, please post the output of running the usbmon
tool when trying to do this to the linux-usb mailing list and we will
try to help you out there.

greg k-h

Re: acm/ttyACM0/cdc_acm does not deliver data

[Josh Boyer]

On Thu, Oct 6, 2011 at 3:41 PM, Greg KH <greg@kroah.com> wrote:
> On Thu, Oct 06, 2011 at 12:19:48PM -0700, AW wrote:
>> Hi!
>>
>> I made some USB devices using the CDC-ACM proposal...
>> They work fine with 2.6.38.8-35.fc15/amd64 and earlier.
>>
>> But since 2.6.40-4.fc15/amd64 it doesnt work as before...
>
> Can you try a kernel.org release instead of a Fedora kernel?  Does this
> work on 3.0.6?

He's tried 2.6.40.6, which is analogous to 3.0.6.  It still fails
according to the report.

Your suggestion of using a vanilla kernel is fine, but I wouldn't
expect different results.  We have a fairly small set of patches, none
of which seem to jump out as impacting this.

josh

(no subject)

[AW]

> Can you try a kernel.org release instead of a Fedora kernel?
>
never tried that... can i just download a rpm and install it?

> If not, please post the output of running the usbmon [...]

>
I did that:
On the 2.6.38 kernel: each try works fine...
http://www.wgboome.org/cmd:Z-L-0,2.6.38.8-35.fc15.x86_64.usbmon

On the 2.6.40 kernel (=3.0.6?): both tries failed (but the 2nd try with slightly different beginning):
http://www.wgboome.org/cmd:Z-L-0,1st-try,2.6.40.6-0.fc15.x86_64.usbmon

http://www.wgboome.org/cmd:Z-L-0,2nd-try,2.6.40.6-0.fc15.x86_64.usbmon


-arne

Re: acm/ttyACM0/cdc_acm does not deliver data

[AW]

>  Can you try a kernel.org release instead of a Fedora kernel?

>
never tried that... can i just download a rpm and install it?

>  If not, please post the output of running the usbmon [...]
>
I did that:
On the 2.6.38 kernel: each try works fine...
http://www.wgboome.org/cmd:Z-L-0,2.6.38.8-35.fc15.x86_64.usbmon

On the 2.6.40 kernel (=3.0.6?): both tries failed (but the 2nd try with slightly 
different beginning):
http://www.wgboome.org/cmd:Z-L-0,1st-try,2.6.40.6-0.fc15.x86_64.usbmon
http://www.wgboome.org/cmd:Z-L-0,2nd-try,2.6.40.6-0.fc15.x86_64.usbmon


-arne

Re: acm/ttyACM0/cdc_acm does not deliver data

[AW]

I tried a vanilla kernel now (built from the fedora SRPM after some help by Piotr H....)...

The command that works on the vanilla kernel (_and_ on the official fc15 kernel) looks to usbmon on the vanilla kernel like this:
http://www.wgboome.org/cmd:P-C,2.6.40.6-0.local.fc15.x86_64.usbmon


(the command "P C" switches receive/send-roles much faster (i guess: less than 1msec) than the command "Z L 0" (I guess 50msec); during that time interrupts r disabled and the ATtiny doesnt care for USB things... *blush* but why did the "old" USB subsystem didnt complain about it? and what about my professional DVB-T USB sticks?)

On the vanilla kernel the command "Z L 0" doesnt work as before:
http://www.wgboome.org/cmd:Z-L-0,2nd-try,2.6.40.6-0.local.fc15.x86_64.usbmon


-arne

Re: acm/ttyACM0/cdc_acm does not deliver data

[AW]

yesterday i changed some features of my self-tinkered USB device (especially no deep sleep anymore and other measures for faster reaction on interrupts)...

now everything works fine again (with fc15's 2.6.41)... the USB TV sticks work good again, 2... :-)

thx for pointing me to usbmon (it helped me to c that my device is not USBish enough)...

-arne