Re: [PATCH 11/18] spi: qup: properly detect extra interrupts

On Wed, Jun 14, 2017 at 2:59 PM, Andy Gross [off-list ref] wrote:

On Wed, Jun 14, 2017 at 12:57:25PM +0530, Sricharan R wrote:

quoted

Hi Varada,

On 6/14/2017 11:22 AM, Varadarajan Narayanan wrote:

quoted

It's possible for a SPI transaction to complete and get another
interrupt and have it processed on the same spi_transfer before the
transfer_one can set it to NULL.

This masks unexpected interrupts, so let's set the spi_transfer to
NULL in the interrupt once the transaction is done. So we can
properly detect these bad interrupts and print warning messages.

Signed-off-by: Matthew McClintock <redacted>
Signed-off-by: Varadarajan Narayanan <redacted>
---
 drivers/spi/spi-qup.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)

diff --git a/drivers/spi/spi-qup.c b/drivers/spi/spi-qup.c
index bd53e82..1a2a9d9 100644
--- a/drivers/spi/spi-qup.c
+++ b/drivers/spi/spi-qup.c

@@ -496,13 +496,13 @@ static irqreturn_t spi_qup_qup_irq(int irq, void *dev_id)
    struct spi_qup *controller = dev_id;
    struct spi_transfer *xfer;
    u32 opflags, qup_err, spi_err;
-   unsigned long flags;
    int error = 0;
+   bool done = 0;

-   spin_lock_irqsave(&controller->lock, flags);
+   spin_lock(&controller->lock);
    xfer = controller->xfer;
    controller->xfer = NULL;
-   spin_unlock_irqrestore(&controller->lock, flags);
+   spin_unlock(&controller->lock);

 Why change the locking here ?

quoted

    qup_err = readl_relaxed(controller->base + QUP_ERROR_FLAGS);
    spi_err = readl_relaxed(controller->base + SPI_ERROR_FLAGS);

@@ -556,16 +556,19 @@ static irqreturn_t spi_qup_qup_irq(int irq, void *dev_id)
                    spi_qup_write(controller, xfer);
    }

-   spin_lock_irqsave(&controller->lock, flags);
-   controller->error = error;
-   controller->xfer = xfer;
-   spin_unlock_irqrestore(&controller->lock, flags);
-
    /* re-read opflags as flags may have changed due to actions above */
    opflags = readl_relaxed(controller->base + QUP_OPERATIONAL);

    if ((controller->rx_bytes == xfer->len &&
            (opflags & QUP_OP_MAX_INPUT_DONE_FLAG)) ||  error)
+           done = true;
+
+   spin_lock(&controller->lock);
+   controller->error = error;
+   controller->xfer = done ? NULL : xfer;
+   spin_unlock(&controller->lock);
+
+   if (done)
            complete(&controller->done);

  Its not clear, why the driver is setting the controller->xfer = NULL
  and restoring it inside the irq. This patch seems to fix things on
  top of that.

I think the original intent was to make sure that the irqhandler knew that there
was no outstanding transaction.  This begs the question of why that would ever
be necessary.  I think it would suffice to rework all of that to remove that
behavior and perhaps enable/disable the irq as we need to during transactions.

I've never been a fan of the controller->xfer being set to NULL.

Also, this patch presumably fixes an issue seen on Dakota SoCs,
doesn't add or remote the NULL bits.

-M
--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html