[PATCH] spi/tegra114: Factor runtime PM out into transfer prepare/unprepare

[PATCH] spi/tegra114: Factor runtime PM out into transfer prepare/unprepare

[Mark Brown]

From: Mark Brown <broonie-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>

Currently the tegra114 driver acquires a runtime PM reference for the
duration of each transfer. This may result in the IP being powered down
between transfers which would be at best wasteful. Instead it is better
to do this in the callbacks that are generated before and after starting
a series of transfers, keeping the IP powered throughout.

Signed-off-by: Mark Brown <broonie-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 drivers/spi/spi-tegra114.c | 34 +++++++++++++++++++++++++---------
 1 file changed, 25 insertions(+), 9 deletions(-)

diff --git a/drivers/spi/spi-tegra114.c b/drivers/spi/spi-tegra114.c
index 8c99c00..e7d48a1 100644
--- a/drivers/spi/spi-tegra114.c
+++ b/drivers/spi/spi-tegra114.c
@@ -803,6 +803,20 @@ static int tegra_spi_setup(struct spi_device *spi)
 	return 0;
 }
 
+static int tegra_spi_prepare_transfer(struct spi_master *spi)
+{
+	struct tegra_spi_data *tspi = spi_master_get_devdata(spi);
+	int ret;
+
+	ret = pm_runtime_get_sync(tspi->dev);
+	if (ret < 0) {
+		dev_err(tspi->dev, "runtime PM get failed: %d\n", ret);
+		return ret;
+	}
+
+	return ret;
+}
+
 static int tegra_spi_transfer_one_message(struct spi_master *master,
 			struct spi_message *msg)
 {
@@ -816,14 +830,6 @@ static int tegra_spi_transfer_one_message(struct spi_master *master,
 	msg->status = 0;
 	msg->actual_length = 0;
 
-	ret = pm_runtime_get_sync(tspi->dev);
-	if (ret < 0) {
-		dev_err(tspi->dev, "runtime PM get failed: %d\n", ret);
-		msg->status = ret;
-		spi_finalize_current_message(master);
-		return ret;
-	}
-
 	single_xfer = list_is_singular(&msg->transfers);
 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
 		INIT_COMPLETION(tspi->xfer_completion);
@@ -859,12 +865,20 @@ static int tegra_spi_transfer_one_message(struct spi_master *master,
 	ret = 0;
 exit:
 	tegra_spi_writel(tspi, tspi->def_command1_reg, SPI_COMMAND1);
-	pm_runtime_put(tspi->dev);
 	msg->status = ret;
 	spi_finalize_current_message(master);
 	return ret;
 }
 
+static int tegra_spi_unprepare_transfer(struct spi_master *spi)
+{
+	struct tegra_spi_data *tspi = spi_master_get_devdata(spi);
+
+	pm_runtime_put(tspi->dev);
+
+	return 0;
+}
+
 static irqreturn_t handle_cpu_based_xfer(struct tegra_spi_data *tspi)
 {
 	struct spi_transfer *t = tspi->curr_xfer;
@@ -1050,7 +1064,9 @@ static int tegra_spi_probe(struct platform_device *pdev)
 	/* the spi->mode bits understood by this driver: */
 	master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
 	master->setup = tegra_spi_setup;
+	master->prepare_transfer_hardware = tegra_spi_prepare_transfer;
 	master->transfer_one_message = tegra_spi_transfer_one_message;
+	master->unprepare_transfer_hardware = tegra_spi_unprepare_transfer;
 	master->num_chipselect = MAX_CHIP_SELECT;
 	master->bus_num = -1;
 
-- 
1.8.3.2

Re: [PATCH] spi/tegra114: Factor runtime PM out into transfer prepare/unprepare

[Laxman Dewangan]

On Saturday 27 July 2013 04:53 PM, Mark Brown wrote:
> From: Mark Brown <broonie-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
>
> Currently the tegra114 driver acquires a runtime PM reference for the
> duration of each transfer. This may result in the IP being powered down
> between transfers which would be at best wasteful. Instead it is better
> to do this in the callbacks that are generated before and after starting
> a series of transfers, keeping the IP powered throughout.
>
> Signed-off-by: Mark Brown <broonie-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
> ---

Acked-by: Laxman Dewangan <ldewangan-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>

The reason I kept this in transfer_one() is because prepare/unprepare 
was getting called in atomic context. But after the change,
Author: Bryan Freed <bfreed-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org>
Date:   Wed Mar 13 11:17:40 2013 -0700

     spi: Unlock a spinlock before calling into the controller driver.


It is fine to have prepare as blocking call.

Re: [PATCH] spi/tegra114: Factor runtime PM out into transfer prepare/unprepare

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 346 bytes --]

On Sat, Jul 27, 2013 at 05:49:35PM +0530, Laxman Dewangan wrote:

> The reason I kept this in transfer_one() is because
> prepare/unprepare was getting called in atomic context. But after
> the change,

With something like that the best thing to do is to query why there's a
restriction and/or fix it - it's a fairly obvious thing to want to do.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]