MUSB dual-role on AM335x behaving weirdly

[Bin Liu]

Felipe,

On Thu, May 14, 2015 at 4:04 PM, Bin Liu <binmlist at gmail.com> wrote:
> Felipe,
>
> On Thu, May 14, 2015 at 2:29 PM, Felipe Balbi <balbi at ti.com> wrote:
>> On Thu, May 14, 2015 at 02:29:20PM -0500, Bin Liu wrote:
>>> On Thu, May 14, 2015 at 2:21 PM, Felipe Balbi <balbi at ti.com> wrote:
>>> > On Thu, May 14, 2015 at 02:19:28PM -0500, Bin Liu wrote:
>>> >> Felipe,
>>> >>
>>> >> On Thu, May 14, 2015 at 2:04 PM, Felipe Balbi <balbi at ti.com> wrote:
>>> >> > On Thu, May 14, 2015 at 12:49:07PM -0500, Felipe Balbi wrote:
>>> >> >> On Thu, May 14, 2015 at 12:40:31PM -0500, Felipe Balbi wrote:
>>> >> >> > On Thu, May 14, 2015 at 12:07:00PM -0500, Felipe Balbi wrote:
>>> >> >> > > Hi,
>>> >> >> > >
>>> >> >> > > On Wed, Feb 25, 2015 at 01:11:22PM +0100, Yegor Yefremov wrote:
>>> >> >> > > > On 25.02.2015 12:11, Maxime Ripard wrote:
>>> >> >> > > > > On Tue, Feb 24, 2015 at 11:33:57AM -0600, Felipe Balbi wrote:
>>> >> >> > > > >> Hi,
>>> >> >> > > > >>
>>> >> >> > > > >> On Tue, Feb 24, 2015 at 05:50:50PM +0100, Maxime Ripard wrote:
>>> >> >> > > > >>> Hi Felipe,
>>> >> >> > > > >>>
>>> >> >> > > > >>> On Tue, Feb 24, 2015 at 08:54:01AM -0600, Felipe Balbi wrote:
>>> >> >> > > > >>>> Hi,
>>> >> >> > > > >>>>
>>> >> >> > > > >>>> On Tue, Feb 24, 2015 at 11:39:11AM +0100, Maxime Ripard wrote:
>>> >> >> > > > >>>>> On Thu, Feb 05, 2015 at 02:21:42PM +0100, Maxime Ripard wrote:
>>> >> >> > > > >>>>>> Hi,
>>> >> >> > > > >>>>>>
>>> >> >> > > > >>>>>> On Thu, Jan 22, 2015 at 08:37:45AM +0100, Yegor Yefremov wrote:
>>> >> >> > > > >>>>>>> I have the same experience with 3.15. The switching is working when
>>> >> >> > > > >>>>>>> CONFIG_USB_MUSB_DUAL_ROLE is set and dr_mode = "otg". But since 3.16
>>> >> >> > > > >>
>>> >> >> > > > >> since 3.16 ?
>>> >> >> > > > >
>>> >> >> > > > > That's what Yegor said. I never saw it working with 3.15 either.
>>> >> >> > > >
>>> >> >> > > > I've used 3.15.1 and 3.15.2 with this set of patches:
>>> >> >> > > > https://github.com/visionsystemsgmbh/onrisc_br_bsp/tree/master/board/vscom/kernel-patches/linux-3.15
>>> >> >> > > >
>>> >> >> > > > And it worked so far. The system:
>>> >> >> > > > http://www.visionsystems.de/produkte/baltos-ir-5221.html
>>> >> >> > >
>>> >> >> > > I've had more time to look into this (thanks Yegor for sponsoring a
>>> >> >> > > test/dev platform) what I noticed is that Connect IRQ takes seconds to
>>> >> >> > > fire up.  Below a tiny log snippet after pluging USB OTG adapter cable
>>> >> >> > > that came with IR5521:
>>> >> >> > >
>>> >> >> > > | [ 1227.200514] musb-hdrc musb-hdrc.1.auto: usbintr (100) epintr(0)
>>> >> >> > >
>>> >> >> > > Cable connected. ID is grounded. 0x100 == DRVVBUS IRQ
>>> >> >> > >
>>> >> >> > > | [ 1227.206788] musb-hdrc musb-hdrc.1.auto: VBUS on (a_wait_vrise), devctl 19
>>> >> >> > >
>>> >> >> > > MUSB starts to wait for VBUS to reach Session valid threshold
>>> >> >> > >
>>> >> >> > > | [ 1230.281159] musb-hdrc musb-hdrc.1.auto: usbintr (10) epintr(0)
>>> >> >> > >
>>> >> >> > > 3 seconds later connect interrupt happens. Looking at VBUS charge time
>>> >> >> > > with a scope, it's quite ok. VBUS charges in about 1.77ms. I'll dig
>>> >> >> > > further into this.
>>> >> >> >
>>> >> >> > even more weird. If I disconnect device from OTG adapter, rather than
>>> >> >> > OTG adapter from IR5521, this leave ID pin grounded, which means DRVVBUS
>>> >> >> > is still asserted and VBUS remains above session valid threshold.
>>> >> >> >
>>> >> >> > Even in this case, when I connect the device on the other end of the
>>> >> >> > cable, I still see some 3 seconds delay from the time device is
>>> >> >> > connected, to the time connect IRQ fires up.
>>> >> >>
>>> >> >> seems to be a problem with the USB stick I'm using. Tested two other
>>> >> >> devices and they connect right away.
>>> >> >
>>> >> > ok, fixing DRD on AM335x will take longer than I originally expected,
>>> >> > probably won't be ready for v4.2 :-(
>>> >>
>>> >> Are you able replicate the issue with TI AM335x GP EVM? I am wondering
>>> >> if the is custom board design problem? have you checked the custom
>>> >> board schematics?
>>> >
>>> > don't have either AM335x GP EVM nor schematics for this board. But it's
>>> > certainly not a problem with the board. It's very easy to replicate the
>>> > problem:
>>> >
>>> > Set dr-mode to otg, load g_zero, connect to PC and as quickly as you
>>> > can, remove cable and attach otg cable with a mouse or whatever on the
>>> > other end.
>>> >
>>> > First time, mouse won't enumerate (no IRQs fire) remove and connect
>>> > again. You should see a Babble IRQ.
>>>
>>> And this only happens with 3.16+, not older kernels? I have a GP EVM,
>>> and can try to take a look.
>>
>> yeah, if you can try with v4.1-rc3, that'd be great. I also see that
>> disconnect IRQ takes a while to fire, but VBUS drops rather fast on this
>> board (< 10ms to discharge VBUS).
>
> How low is VBUS after 10ms? I don't have a scope with me right now,
> but I see Diconnect IRQ happened right away if I short VBUS to ground
> right after disconnected from the host. So I suspect is that VBUS
> discharge takes ~20sec on my board, which is the time taken for MUSB
> from b_peripheral to b_idle.
>
> BTY, it seems Disconnect IRQ is expected for MUSB in device mode. I
> never pay attention on this.

I think I found the root cause of the problem: board design issue - I
bet the custom board has too much cap on VBUS line. It should be <
10uF.

I just noticed I have the Jumper 36 on on my EVM, which adds 154.7uF
cap on VBUS causing discharge takes ~20sec. After removed the jumper,
which leaves only 4.7uF cap on VBUS, now it only takes ~0.4sec to
generate Disconnect IRQ. Here is the log.

root@:~# [ 2504.893123] musb-hdrc musb-hdrc.0.auto: usbintr (1) epintr(0)
[ 2504.899198] musb-hdrc musb-hdrc.0.auto: <== DevCtl=99, int_usb=0x1
[ 2504.912751] zero gadget: suspend
[ 2504.916145] zero gadget: zero_suspend
[ 2505.303937] musb-hdrc musb-hdrc.0.auto: usbintr (20) epintr(0)
[ 2505.310072] musb-hdrc musb-hdrc.0.auto: <== DevCtl=88, int_usb=0x20
[ 2505.325355] zero gadget: reset config
[ 2507.303288] musb-hdrc musb-hdrc.0.auto: Poll devctl 80 (b_idle)

Regards,
-Bin.

>
> Regards,
> -Bin.
>
>>
>>> /me just figured out the modem issue, and in a very good mood now ;)
>>
>> hey, congrats :-)
>>
>> --
>> balbi