Current git head oops'es at boot on Overo

Current git head oops'es at boot on Overo

[Steve Sakoman]

Build is stock 62c36fdd962de2db5d62a326490913f41c442a5c, no additional patches.

Without DEBUG_LL, boot silently fails. With DEBUG_LL, one can see that
the failure is due to a NULL pointer dereference in musb.

I suspect that the root cause is related to:

<3>i2c_omap i2c_omap.1: controller timed out
<4>twl4030_usb: probe of twl4030_usb failed with error -22

Adding David Brownell's i2c-omap timeout workaround patch eliminates
this error and the boot then fails on EHCI.

I disabled EHCI in defconfig and was able to get the boot to complete.
 However unlike others, I find that musb is not functional -- and the
trick of disconnecting the USB cable and then reconnecting it does not
work.

Steve

Re: Current git head oops'es at boot on Overo

[David Brownell]

On Monday 06 October 2008, Steve Sakoman wrote:
> Build is stock 62c36fdd962de2db5d62a326490913f41c442a5c, no additional patches.
> 
> Without DEBUG_LL, boot silently fails. With DEBUG_LL, one can see that
> the failure is due to a NULL pointer dereference in musb.
> 
> I suspect that the root cause is related to:
> 
> <3>i2c_omap i2c_omap.1: controller timed out
> <4>twl4030_usb: probe of twl4030_usb failed with error -22
> 
> Adding David Brownell's i2c-omap timeout workaround patch eliminates
> this error

Right.  I'd suggest merging that, except that would provide
zero incentive to actually fix the i2c-omap problem.

And I *have* seen those timeout errors on other requests...


> and the boot then fails on EHCI. 
> 
> I disabled EHCI in defconfig and was able to get the boot to complete.

Ditto.  I suspect one of those no-timeout busy-waits is just
neglecting to complete... :(


>  However unlike others, I find that musb is not functional -- and the
> trick of disconnecting the USB cable and then reconnecting it does not
> work.

I observed the same thing.  More specifically, no IRQs arrive at
the twl4030-usb driver.

The *very same kernel* works fine on Beagle... is there something
different about the USB setup that might explain this?

- Dave

Re: Current git head oops'es at boot on Overo

[Steve Sakoman]

On Mon, Oct 6, 2008 at 12:53 PM, David Brownell <david-b@pacbell.net> wrote:
> On Monday 06 October 2008, Steve Sakoman wrote:

>>  However unlike others, I find that musb is not functional -- and the
>> trick of disconnecting the USB cable and then reconnecting it does not
>> work.
>
> I observed the same thing.  More specifically, no IRQs arrive at
> the twl4030-usb driver.
>
> The *very same kernel* works fine on Beagle... is there something
> different about the USB setup that might explain this?

I took a look at the schematics for that section of both boards and as
far as I can tell the two boards are identical in their omap/twl4030
interconnect for the musb section.  The linux board files are
extremely similar too -- nothing significant as far as I can tell.

I tried the same experiment and used my Overo kernel on Beagle.  I got
exactly the same behavior on Beagle as I did on Overo.  So my results
are different than yours.

How does your kernel config set up musb -- host or OTG?  Mine is set up as host.

Steve

Re: Current git head oops'es at boot on Overo

[Nathan Monson]

I'm set up as host, using a BeagleBoard, and with the current
linux-omap git I can get USB working by unplugging and replugging the
A side of the cable (at the mini-AB connector on the BeagleBoard).

Just disconnecting the cable at the B side is not enough.  It must
have something to do with the ID pin.  So, if you have a soldered ID
pin it wouldn't work either.

Also, my cable is going into a powered 4 port high speed hub.  That
might make a difference.

- Nathan

On Mon, Oct 6, 2008 at 2:04 PM, Steve Sakoman <sakoman@gmail.com> wrote:
> On Mon, Oct 6, 2008 at 12:53 PM, David Brownell <david-b@pacbell.net> wrote:
>> On Monday 06 October 2008, Steve Sakoman wrote:
>
>>>  However unlike others, I find that musb is not functional -- and the
>>> trick of disconnecting the USB cable and then reconnecting it does not
>>> work.
>>
>> I observed the same thing.  More specifically, no IRQs arrive at
>> the twl4030-usb driver.
>>
>> The *very same kernel* works fine on Beagle... is there something
>> different about the USB setup that might explain this?
>
> I took a look at the schematics for that section of both boards and as
> far as I can tell the two boards are identical in their omap/twl4030
> interconnect for the musb section.  The linux board files are
> extremely similar too -- nothing significant as far as I can tell.
>
> I tried the same experiment and used my Overo kernel on Beagle.  I got
> exactly the same behavior on Beagle as I did on Overo.  So my results
> are different than yours.
>
> How does your kernel config set up musb -- host or OTG?  Mine is set up as host.
>
> Steve
> --
> To unsubscribe from this list: send the line "unsubscribe linux-omap" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: Current git head oops'es at boot on Overo

[David Brownell]

On Monday 06 October 2008, Steve Sakoman wrote:
> How does your kernel config set up musb -- host or OTG?  Mine is set up as host.

For now, host-only.

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Current git head oops'es at boot on Overo

[David Brownell]

On Monday 06 October 2008, Nathan Monson wrote:
> I'm set up as host, using a BeagleBoard, and with the current
> linux-omap git I can get USB working by unplugging and replugging the
> A side of the cable (at the mini-AB connector on the BeagleBoard).
> 
> Just disconnecting the cable at the B side is not enough.  It must
> have something to do with the ID pin.  So, if you have a soldered ID
> pin it wouldn't work either.

Right.  That's a bug in the transciever support, note ... on
host side there are TWO relevant cable events:  ID pin grounded,
and D+/D- rising.  It only understands the former for now, and
doesn't check for the state very well unless poked by an IRQ.

(All that can be a PITA to get right.  It helps to have a cable
tester that will let you do things like switch the ID ground,
apply a variable VBUS load, apply D+/D- pullups/pulldowns,
and so on.)


> Also, my cable is going into a powered 4 port high speed hub.  That
> might make a difference.

That's what I was doing too.  It's a common setup ... just a
bit easier to preconfigure networking that way than through
a bridgede configuration.

- Dave

Re: Current git head oops'es at boot on Overo

[Steve Sakoman]

On Mon, Oct 6, 2008 at 2:15 PM, Nathan Monson <nmonson@gmail.com> wrote:
> I'm set up as host, using a BeagleBoard, and with the current
> linux-omap git I can get USB working by unplugging and replugging the
> A side of the cable (at the mini-AB connector on the BeagleBoard).
>
> Just disconnecting the cable at the B side is not enough.  It must
> have something to do with the ID pin.  So, if you have a soldered ID
> pin it wouldn't work either.
>
> Also, my cable is going into a powered 4 port high speed hub.  That
> might make a difference.

Same here -- 4 port high speed hub, disconnect/reconnect cable at the
A (board) side.

No joy on Overo.

Steve