[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87459424-1083-0716-0481-89d7b7e3f08c@ti.com>
Date: Mon, 15 Oct 2018 15:34:37 +0530
From: Sekhar Nori <nsekhar@...com>
To: Vignesh R <vigneshr@...com>, Mark Brown <broonie@...nel.org>
CC: <linux-spi@...r.kernel.org>, <linux-omap@...r.kernel.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 3/3] spi: omap2-mcspi: Add slave mode support
On Monday 15 October 2018 03:12 PM, Vignesh R wrote:
> Hi Sekhar,
>
> On Monday 15 October 2018 01:53 PM, Sekhar Nori wrote:
>
> [...]
>>>
>>> +static irqreturn_t omap2_mcspi_irq_handler(int irq, void *data)
>>> +{
>>> + struct omap2_mcspi *mcspi = data;
>>> + u32 irqstat;
>>> +
>>> + irqstat = mcspi_read_reg(mcspi->master, OMAP2_MCSPI_IRQSTATUS);
>>> + if (!irqstat)
>>> + return IRQ_NONE;
>>> +
>>> + /* Disable IRQ and wakeup slave xfer task */
>>> + mcspi_write_reg(mcspi->master, OMAP2_MCSPI_IRQENABLE, 0);
>>> + if (irqstat & OMAP2_MCSPI_IRQSTATUS_EOW)
>>> + complete(&mcspi->txdone);
>>> +
>>> + return IRQ_HANDLED;
>>
>> You need to have the:
>>
>> pm_runtime_get_sync();
>>
>> /* access registers */
>>
>> pm_runtime_mark_last_busy();
>> pm_runtime_put_autosuspend();
>>
>> sequence here. I think thats also missing from the dma callbacks.
>> Probably working by chance today.
>>
>
> This is taken care of by the SPI core as part of __spi_pump_messages():
> pm_runtime_get_sync()
> ...
> spi_transfer_one_message
> ...
> omap2_mcspi_transfer_one
> ...
> omap2_mcspi_txrx_dma
>
> So, both in dma callbacks and in IRQ handler, SPI controller is in
> active state.
Ah, okay then. False alarm :)
Regards,
Sekhar
Powered by blists - more mailing lists