allwinner/i2c interrupt storm detected

[Daniel Braniss]

> On 5 Jul 2020, at 20:14, Emmanuel Vadot <manu at bidouilliste.com> wrote:
> 
> 
> Hi Daniel, Manuel,
> 
> On Sat, 4 Jul 2020 13:45:50 +0300
> Daniel Braniss <danny at cs.huji.ac.il <mailto:danny at cs.huji.ac.il>> wrote:
> 
>> 
>> 
>>> On 4 Jul 2020, at 13:29, Manuel Stühn <freebsdnewbie at freenet.de> wrote:
>>> 
>>> On Tue, 30 Jun 2020 16:01:41 +0300
>>> Daniel Braniss <danny at cs.huji.ac.il <mailto:danny at cs.huji.ac.il>> wrote:
>>> 
>>>> Hi,
>>>> 
>>>> after a long time I decided to try and upgrade to stable 12.1 r362793 since I saw some changes where done 
>>>> with respect to the DTS and twsi.c, 
>>>> 
>>>> if nothing is connected to the i2c, i2c -s just hangs,
>>>> 
>>>> if something is connected this is what i get on the console after typing ?i2c -s?
>>>> 
>>>> 
>>>> Hardware may not support START/STOP scanning; trinterrupt storm detected on "gic0,s6:"; throttling interrupt source
>>>> ying less-reliable read method.
>>>> interrupt storm detected on "gic0,s6:"; throttling interrupt source
>>>> interrupt storm detected on "gic0,s6:"; throttling interrupt source
>>>> ?
>>>> 
>>>> and
>>>> neo-04> vmstat -i
>>>> interrupt                                             total       rate
>>>> gic0,p13:-ic_timer0                                   16052        164
>>>> gic0,s0: uart2                                          318          3
>>>> gic0,s6: iichb0                                       13034        133
>>>> gic0,s60: aw_mmc0                                      1293         13
>>>> gic0,s82: awg0                                          334          3
>>>> gic0,s120: pmu0                                       49725        509
>>>> cpu0:rendezvous                                          18          0
>>>> cpu1:rendezvous                                          50          1
>>>> cpu2:rendezvous                                          51          1
>>>> cpu3:rendezvous                                          40          0
>>>> cpu0:preempt                                           2691         28
>>>> cpu1:preempt                                           3165         32
>>>> cpu2:preempt                                           2778         28
>>>> cpu3:preempt                                           2986         31
>>>> cpu0:hardclock                                           15          0
>>>> Total                                                 92550        946
>>>> 
>>>> 
>>>> the hardware is an NanoPi Neo
>>>> ---<<BOOT>>---
>>>> KDB: debugger backends: ddb
>>>> KDB: current backend: ddb
>>>> Copyright (c) 1992-2020 The FreeBSD Project.
>>>> Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
>>>>      The Regents of the University of California. All rights reserved.
>>>> FreeBSD is a registered trademark of The FreeBSD Foundation.
>>>> FreeBSD 12.1-STABLE #0 r362793M: Tue Jun 30 11:39:11 IDT 2020
>>>>  danny at nrnd:/home/obj/nrnd/arm/neo/vol/rnd/stable/12/arm.armv7/sys/AWGEN arm
>>>> FreeBSD clang version 10.0.0 (git at github.com <mailto:git at github.com>:llvm/llvm-project.git llvmorg-10.0.0-0-gd32170dbd5b)
>>>> VT: init without driver.
>>>> No PSCI/SMCCC call function found
>>>> CPU: ARM Cortex-A7 r0p5 (ECO: 0x00000000)
>>>> ?
>>>> 
>>> 
>>> I do not have a IRQ-Storm on my NanoPI NEO2, but a "i2s -s" does never return. Commit v356609 broke i2c-support on my hardware (reverting this single commit fixed it, bugreport filed: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=247576 <https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=247576>).
>>> 
>>> Perhaps it is worth a try for you also to revert this commit and test again...
>>> 
>> 
>> before the latest changes it works fine, and if you add my patch to it, i2s -s will not hang:
>> 
>>> -- twsi.c	(revision 346538)
>>> +++ twsi.c	(working copy)
>>> @@ -458,8 +458,15 @@
>>> 		if (sc->msg->len == 1)
>>> 			sc->control_val &= ~TWSI_CONTROL_ACK;
>>> 		TWSI_WRITE(sc, sc->reg_control, sc->control_val | TWSI_CONTROL_START);
>>> -		while (sc->error == 0 && sc->transfer != 0) {
>>> -			pause_sbt("twsi", SBT_1MS * 30, SBT_1MS, 0);
>>> +		{
>>> +			 int count = 10;
>>> +			 while (sc->error == 0 && sc->transfer != 0) {
>>> +				  pause_sbt("twsi", SBT_1MS * 30, SBT_1MS, 0);
>>> +				  if(count-- == 0) {
>>> +					   sc->error = EDEADLK;
>>> +					   break;
>>> +				  }
>>> +			 }
>>> 		}
>>> 
>>> 		debugf(dev, "Done with msg[%d]\n", i);
>> 
>> 
>> cheers,
>> 	danny
>> 
>>> 
>>> BR
>>> Manuel
> 
> Could you test the patch I've just attached to
> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=247576 <https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=247576> please ?
> 
> It doesn't fix everything and I'm still working on doing test on a lot
> of different boards but this is clearly needed.
> 
> -- 
> Emmanuel Vadot <manu at bidouilliste.com <mailto:manu at bidouilliste.com>> <manu at freebsd.org <mailto:manu at freebsd.org>>


short version: it works!

longer version:
first tried with debugging on - worked
next tried with no debugging- worked

the next test will take some time, trying my app.
if I feel up to it, i can try on a neopi neo2 - but that will take longer.

thanks Manu!!!!

	danny