[BCM4718] Broken interrupts

[Michael Zhilin]

Stanislav,

I've added to BCM471x:
makeoptions    MIPS_INTRNG
options        MIPS_INTRNG
to options.mips:
MIPS_INTRNG                   opt_global.h

Then removed all kernel object files and run buildkernel (thank you,
Adrian, for freebsd-wifi-build scripts!). But it's still same behavior.

Here is output of printfs before and after clock_intr:

cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=f2fe11a cmp=deadc0bb cause=8000
after clock_intr cnt=f45c9e6 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=f6225e7 cmp=ffffffff cause=8000
after clock_intr cnt=f780e46 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=f946b1e cmp=ffffffff cause=8000
after clock_intr cnt=faa548a cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=fc6b178 cmp=ffffffff cause=8000
after clock_intr cnt=fdc9ae6 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=ff8f7b8 cmp=ffffffff cause=8000
after clock_intr cnt=100ee112 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=102b8fa8 cmp=ffffffff cause=8000
after clock_intr cnt=1041cc94 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=105e7b26 cmp=ffffffff cause=8000
after clock_intr cnt=1074b6c8 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=10916586 cmp=ffffffff cause=8000
after clock_intr cnt=10a7a24c cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=10c4510a cmp=ffffffff cause=8000
after clock_intr cnt=10da8c62 cmp=ffffffff cause=8000
cpupic0: Servicing interrupt 7
before clock_intr(0x80528100) cnt=10f73c6e cmp=ffffffff cause=8000
after clock_intr cnt=110d7838 cmp=ffffffff cause=8000

As I mentioned before, there are 2 symptoms:
 * cnt != cmp for IRQ5 triggering
 * cause is always 8000 (i.e. IP7 = 1) even after mips_wr_compare.

Thanks!

On Fri, Mar 4, 2016 at 3:33 PM, Stanislav Galabov <sgalabov at gmail.com>
wrote:

> Also, before doing that, you'll have to add:
> MIPS_INTRNG           opt_global.h
> to the end of sys/conf/options.mips
>
> Adrian, we somehow missed that in the commit of
> https://reviews.freebsd.org/D5182 although it is in the last diff I had
> attached... I'll put up a review for this on monday.
>
> Best wishes,
> Stanislav
>
> On Fri, Mar 4, 2016 at 2:17 PM, Stanislav Galabov <sgalabov at gmail.com>
> wrote:
>
>> Hi Michael,
>>
>> If I understand correctly the goal is to use INTRNG for MIPS?
>> If so, I didn't notice 'options MIPS_INTRNG' anywhere in your kernel
>> config files...
>> Please try to add this to your sys/mips/bcm471x/std.bcm471x for example
>> and let us know how this goes.
>>
>> Best wishes,
>> Stanislav
>>
>> On Fri, Mar 4, 2016 at 1:19 PM, Michael Zhilin <mizhka at gmail.com> wrote:
>>
>>> Hi Adrian!
>>>
>>> Yes, KDB works. And yes, (today's finding) something is silly that force
>>> make IRQ5 endless. After processing of previous request, next request is
>>> starting. According to mips docs found in google, IRQ5 must be triggered
>>> only if count == compare. In my case, count increments on 0x20000
>>> roughly,
>>> compare is constant. There is another finding: cause register is always
>>> 0x8000, even after call mips_wr_compare(). I suppose that this is
>>> possible
>>> reason of infinite IRQ5 requests, but still don't know how to fix it...
>>>
>>> Thx,
>>>  Michael
>>>
>>> On Fri, Mar 4, 2016 at 11:56 AM, Adrian Chadd <adrian.chadd at gmail.com>
>>> wrote:
>>>
>>> > hm, does kdb work (ie ,break to debugger at that point), or is the
>>> system
>>> > hosed?
>>> >
>>> > is it something silly like it getting stuck servicing interrupts
>>> > because they're not being correctly ACKed?
>>> >
>>> >
>>> >
>>> > -a
>>> >
>>> >
>>> > On 3 March 2016 at 12:22, Michael Zhilin <mizhka at gmail.com> wrote:
>>> > > Hi,
>>> > >
>>> > > I'm trying to load FreeBSD kernel on Broadcom MIPS router (ASUS
>>> RT-N16).
>>> > > Here is actual dmesg: http://pastebin.com/AtEz6kc9 . But my actual
>>> > problem
>>> > > is broken interrupts. After enabling of interrupts in
>>> > > autoconf.c:configure_final the MipsException code is called, it works
>>> > fine
>>> > > (including INTRNG), but cpu doesn't return to main thread (suprise!).
>>> > >
>>> > > I can see printf from clockintr, i.e. there are regular timer
>>> requests.
>>> > But
>>> > > there is no printf from main thread (autoconf.c) after enabling of
>>> > > interrupts.
>>> > >
>>> > > Also I've tried to simplify MipsException to avoid any issue with
>>> > > saving/restoring registers:
>>> > >     mfc0    k0, MIPS_COP_0_EXC_PC
>>> > >     rfe
>>> > >     jr    k0
>>> > >
>>> > > but still there is no return to main thread.
>>> > >
>>> > > I'm not familiar with JTAG and have no JTAG adapter for debugging.
>>> :( To
>>> > be
>>> > > honest, I've lost hope.
>>> > >
>>> > > What could possibly be wrong? Here is github branch for reference:
>>> > > https://github.com/Cka3o4Huk/freebsd/tree/user/mizhka/bcm471x
>>> > >
>>> > > Thank you in advance,
>>> > >  Michael
>>> > > _______________________________________________
>>> > > freebsd-mips at freebsd.org mailing list
>>> > > https://lists.freebsd.org/mailman/listinfo/freebsd-mips
>>> > > To unsubscribe, send any mail to "
>>> freebsd-mips-unsubscribe at freebsd.org"
>>> >
>>> _______________________________________________
>>> freebsd-mips at freebsd.org mailing list
>>> https://lists.freebsd.org/mailman/listinfo/freebsd-mips
>>> To unsubscribe, send any mail to "freebsd-mips-unsubscribe at freebsd.org"
>>>
>>
>>
>