diff mbox series

spi: sprd: add runtime pm for transfer message

Message ID 20201030072444.22122-1-zhang.lyra@gmail.com
State New
Headers show
Series spi: sprd: add runtime pm for transfer message | expand

Commit Message

Chunyan Zhang Oct. 30, 2020, 7:24 a.m. UTC
From: Bangzheng Liu <bangzheng.liu@unisoc.com>

Before transfer one message, spi core would set chipselect, sprd spi
device should be resumed from runtime suspend, otherwise kernel would
crash once access spi registers. The sprd spi device can be suspended
until clearing chipselect which would be executed after transfer.

Fixes: e7d973a31c24 ("spi: sprd: Add SPI driver for Spreadtrum SC9860")
Signed-off-by: Bangzheng Liu <bangzheng.liu@unisoc.com>
Signed-off-by: Chunyan Zhang <chunyan.zhang@unisoc.com>
---
 drivers/spi/spi-sprd.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

Comments

Mark Brown Oct. 30, 2020, 1:42 p.m. UTC | #1
On Fri, Oct 30, 2020 at 03:24:44PM +0800, Chunyan Zhang wrote:
> From: Bangzheng Liu <bangzheng.liu@unisoc.com>

> 

> Before transfer one message, spi core would set chipselect, sprd spi

> device should be resumed from runtime suspend, otherwise kernel would

> crash once access spi registers. The sprd spi device can be suspended

> until clearing chipselect which would be executed after transfer.


The core should be handling runtime PM for normal transfers, if it's not
managing to do that then we should fix the core so it's fixed for all
drivers not just this one.
Chunyan Zhang Nov. 2, 2020, 3:17 a.m. UTC | #2
On Fri, 30 Oct 2020 at 21:42, Mark Brown <broonie@kernel.org> wrote:
>

> On Fri, Oct 30, 2020 at 03:24:44PM +0800, Chunyan Zhang wrote:

> > From: Bangzheng Liu <bangzheng.liu@unisoc.com>

> >

> > Before transfer one message, spi core would set chipselect, sprd spi

> > device should be resumed from runtime suspend, otherwise kernel would

> > crash once access spi registers. The sprd spi device can be suspended

> > until clearing chipselect which would be executed after transfer.

>

> The core should be handling runtime PM for normal transfers, if it's not

> managing to do that then we should fix the core so it's fixed for all

> drivers not just this one.


Sure, I will send a new patch.
diff mbox series

Patch

diff --git a/drivers/spi/spi-sprd.c b/drivers/spi/spi-sprd.c
index 635738f54c73..1733d10eb296 100644
--- a/drivers/spi/spi-sprd.c
+++ b/drivers/spi/spi-sprd.c
@@ -293,15 +293,25 @@  static void sprd_spi_chipselect(struct spi_device *sdev, bool cs)
 	struct spi_controller *sctlr = sdev->controller;
 	struct sprd_spi *ss = spi_controller_get_devdata(sctlr);
 	u32 val;
+	int ret;
 
-	val = readl_relaxed(ss->base + SPRD_SPI_CTL0);
 	/*  The SPI controller will pull down CS pin if cs is 0 */
 	if (!cs) {
-		val &= ~SPRD_SPI_CS0_VALID;
+		ret = pm_runtime_get_sync(ss->dev);
+		if (ret < 0) {
+			pm_runtime_put_noidle(ss->dev);
+			dev_err(ss->dev, "Failed to power device: %d\n", ret);
+			return;
+		}
+		val = readl_relaxed(ss->base + SPRD_SPI_CTL0);
+		val &= ~SPRD_SPI_CS0_VALID; /* set cs0 valid */
 		writel_relaxed(val, ss->base + SPRD_SPI_CTL0);
 	} else {
-		val |= SPRD_SPI_CSN_MASK;
+		val = readl_relaxed(ss->base + SPRD_SPI_CTL0);
+		val |= SPRD_SPI_CSN_MASK; /* set all cs invalid */
 		writel_relaxed(val, ss->base + SPRD_SPI_CTL0);
+		pm_runtime_mark_last_busy(ss->dev);
+		pm_runtime_put_autosuspend(ss->dev);
 	}
 }