spi/fsl-espi: Don't wait transaction completion forever

Nobuteru Hayashi · broonie · commit aa70e567c4f0 · 2016-03-28T19:41:18.000+01:00
Because the eSPI driver uses wait_for_completion(), any stuck-able
phenomenon at half-way transaction progress made worker task hang up.
This phenomenon is perhaps caused by eSPI device errata which seems not
to be published from vendor site yet.

Anyway, we fix hang task by using fixed 2 seconds timeout
that is our preferred value for eSPI maximum transaction size.

It seems to be better that eSPI driver can detect this stuck and
report error (EMSGSIZE) to the upper device driver because
the upper device driver can decide to retry or recover.

Signed-off-by: Nobuteru Hayashi &lt;hayashi.nbb@ncos.nec.co.jp&gt;
Signed-off-by: Mark Brown &lt;broonie@kernel.org&gt;
diff --git a/drivers/spi/spi-fsl-espi.c b/drivers/spi/spi-fsl-espi.c
@@ -245,7 +245,12 @@ static int fsl_espi_bufs(struct spi_device *spi, struct spi_transfer *t)
 	if (ret)
 		return ret;
 
-	wait_for_completion(&mpc8xxx_spi->done);
+	/* Won't hang up forever, SPI bus sometimes got lost interrupts... */
+	ret = wait_for_completion_timeout(&mpc8xxx_spi->done, 2 * HZ);
+	if (ret == 0)
+		dev_err(mpc8xxx_spi->dev,
+			"Transaction hanging up (left %d bytes)\n",
+			mpc8xxx_spi->count);
 
 	/* disable rx ints */
 	mpc8xxx_spi_write_reg(&reg_base->mask, 0);
