Message-ID: <51995A78.20207@signal11.us>
Date: Sun, 19 May 2013 19:04:24 -0400
From: Alan Ott <alan@signal11.us>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:14.0) Gecko/20120714 Thunderbird/14.0
MIME-Version: 1.0
To: David Hauweele <david@hauweele.net>
CC: Alexander Smirnov <alex.bluesman.smirnov@gmail.com>,
        Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>,
        linux-zigbee-devel@lists.sourceforge.net, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [Linux-zigbee-devel] [PATCH 2/2] mrf24j40: Keep the interrupt
 line enabled
References: <1368112788-25701-1-git-send-email-david@hauweele.net> <1368112788-25701-2-git-send-email-david@hauweele.net> <5191B5C4.2020609@signal11.us> <CAO+c-UWi5UH-ZN5G0JiJCrCrScaYprKT7jLpFH0wEJecgoHgig@mail.gmail.com>
In-Reply-To: <CAO+c-UWi5UH-ZN5G0JiJCrCrScaYprKT7jLpFH0wEJecgoHgig@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3886
Lines: 106

On 05/16/2013 05:34 PM, David Hauweele wrote:
> I have seen the interrupt line going low forever with heavy traffic.

Hi David,

I've been doing some testing today and I can reproduce your issue if I
ping -f both ways simultaneously (after about 3-4 minutes usually). I
think it has more to do with the tx/rx mutual exclusion that you
described in patch 1/1 than it does with the disable/enable IRQ. With
respect to enable/disable irq, I did some checking of other similar
drivers (non-ieee802154) and noticed that many of them use threaded
interrupts. I think this may be closer to the right way to do it.

Are you running a tickless kernel? What is your preemption model? What's
your hardware platform?

> The at86rf230 driver has two isr, one for edge-triggered and another
> for level-triggered interrupt. The latter uses
> disable_irq_nosync/enable_irq which makes sense for level-triggered
> interrupt. Otherwise the interrupt would be triggered again and again
> until the scheduled work read the status register.
>
> However I don't see the use of disable_irq/enable_irq with an
> edge-triggered interrupt. Perhaps I'm missing something. Though I
> don't see any harm using it anyway.

My understanding based on the datasheet is that the interrupt can be
asserted again as soon as INTSTAT is read (which is done in the
workqueue). I guess even if it happens while the workqueue is running,
it's ok.

I think I had a flawed understanding of schedule_work() before, thinking
that it would not schedule a work_struct it if the work_struct was running.

>  As you said the interrupt should
> be delayed until enable_irq() is called. In particular when
> CONFIG_HARDIRQS_SW_RESEND is set.

I think I agree with you. I'll send you a patch to try with threaded
interrupts to try.

I'm trying to determine exactly what's the cause of the interrupt line
being stuck low.

Alan.

>
> 2013/5/14 Alan Ott <alan@signal11.us>:
>> On 5/9/13 11:19 AM, David Hauweele wrote:
>>> Disabling the interrupt line could miss an IRQ and leave the line into a
>>> low state hence locking the driver.
>>>
>> Have you observed this? My understanding is that the interrupt won't be lost
>> but instead delayed until enable_irq() is called.
>>
>> I got this pattern from the other 802.15.4 drivers. Perhaps my understanding
>> is wrong.
>>
>>
>>
>>> Signed-off-by: David Hauweele <david@hauweele.net>
>>> ---
>>>   drivers/net/ieee802154/mrf24j40.c |    7 +------
>>>   1 file changed, 1 insertion(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/net/ieee802154/mrf24j40.c
>>> b/drivers/net/ieee802154/mrf24j40.c
>>> index 1e3ddf3..6ef32f7 100644
>>> --- a/drivers/net/ieee802154/mrf24j40.c
>>> +++ b/drivers/net/ieee802154/mrf24j40.c
>>> @@ -603,8 +603,6 @@ static irqreturn_t mrf24j40_isr(int irq, void *data)
>>>   {
>>>         struct mrf24j40 *devrec = data;
>>>
>>> -       disable_irq_nosync(irq);
>>> -
>>>         schedule_work(&devrec->irqwork);
>>>
>>>         return IRQ_HANDLED;
>>> @@ -619,7 +617,7 @@ static void mrf24j40_isrwork(struct work_struct *work)
>>>         /* Read the interrupt status */
>>>         ret = read_short_reg(devrec, REG_INTSTAT, &intstat);
>>>         if (ret)
>>> -               goto out;
>>> +               return;
>>>
>>>         /* Check for TX complete */
>>>         if (intstat & 0x1)
>>> @@ -628,9 +626,6 @@ static void mrf24j40_isrwork(struct work_struct *work)
>>>         /* Check for Rx */
>>>         if (intstat & 0x8)
>>>                 schedule_work(&devrec->rxwork);
>>> -
>>> -out:
>>> -       enable_irq(devrec->spi->irq);
>>>   }
>>>
>>>   static void mrf24j40_rxwork(struct work_struct *work)
>>>


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