Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1425786

[RFC 07/19] HSI: omap_ssi_port: replace wkin_cken with atomic bitmap operations

From Sebastian Reichel <sre@kernel.org>
Newsgroups linux.kernel
Subject [RFC 07/19] HSI: omap_ssi_port: replace wkin_cken with atomic bitmap operations
Date 2016-06-18 18:30 +0200
Message-ID <rLric-6py-101@gated-at.bofh.it> (permalink)
References <rLri9-6py-7@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw


This simplifies the code and avoids holding a spin_lock when
runtime pm calls are made. Once the irq_safe flag is removed
for omap_ssi's runtime pm, pm_runtime_get/put_sync can sleep,
which is a no-go while holding a spin_lock.

Signed-off-by: Sebastian Reichel <sre@kernel.org>
---
 drivers/hsi/controllers/omap_ssi.h      |  6 ++++--
 drivers/hsi/controllers/omap_ssi_port.c | 16 +++-------------
 2 files changed, 7 insertions(+), 15 deletions(-)

diff --git a/drivers/hsi/controllers/omap_ssi.h b/drivers/hsi/controllers/omap_ssi.h
index 88a523902a8f..6cdaad8805f7 100644
--- a/drivers/hsi/controllers/omap_ssi.h
+++ b/drivers/hsi/controllers/omap_ssi.h
@@ -35,6 +35,8 @@
 #define SSI_MAX_GDD_LCH		8
 #define SSI_BYTES_TO_FRAMES(x) ((((x) - 1) >> 2) + 1)
 
+#define SSI_WAKE_EN 0
+
 /**
  * struct omap_ssm_ctx - OMAP synchronous serial module (TX/RX) context
  * @mode: Bit transmission mode
@@ -75,7 +77,7 @@ struct omap_ssm_ctx {
  * @wake_irq: IRQ number for incoming wake line (-1 if none)
  * @wake_gpio: GPIO number for incoming wake line (-1 if none)
  * @pio_tasklet: Bottom half for PIO transfers and events
- * @wkin_cken: Keep track of clock references due to the incoming wake line
+ * @flags: flags to keep track of states
  * @wk_refcount: Reference count for output wake line
  * @sys_mpu_enable: Context for the interrupt enable register for irq 0
  * @sst: Context for the synchronous serial transmitter
@@ -99,7 +101,7 @@ struct omap_ssi_port {
 	struct gpio_desc	*wake_gpio;
 	struct tasklet_struct	pio_tasklet;
 	bool			wktest:1; /* FIXME: HACK to be removed */
-	bool			wkin_cken:1; /* Workaround */
+	unsigned long		flags;
 	unsigned int		wk_refcount;
 	/* OMAP SSI port context */
 	u32			sys_mpu_enable; /* We use only one irq */
diff --git a/drivers/hsi/controllers/omap_ssi_port.c b/drivers/hsi/controllers/omap_ssi_port.c
index 9001b06312c3..0d3452393670 100644
--- a/drivers/hsi/controllers/omap_ssi_port.c
+++ b/drivers/hsi/controllers/omap_ssi_port.c
@@ -751,10 +751,8 @@ static int ssi_release(struct hsi_client *cl)
 		 * Drop the clock reference for the incoming wake line
 		 * if it is still kept high by the other side.
 		 */
-		if (omap_port->wkin_cken) {
+		if (test_and_clear_bit(SSI_WAKE_EN, &omap_port->flags))
 			pm_runtime_put_sync(omap_port->pdev);
-			omap_port->wkin_cken = 0;
-		}
 		pm_runtime_get_sync(omap_port->pdev);
 		/* Stop any SSI TX/RX without a client */
 		ssi_set_port_mode(omap_port, SSI_MODE_SLEEP);
@@ -981,12 +979,8 @@ static irqreturn_t ssi_wake_thread(int irq __maybe_unused, void *ssi_port)
 		 * This workaround will avoid breaking the clock reference
 		 * count when such a situation ocurrs.
 		 */
-		spin_lock(&omap_port->lock);
-		if (!omap_port->wkin_cken) {
-			omap_port->wkin_cken = 1;
+		if (!test_and_set_bit(SSI_WAKE_EN, &omap_port->flags))
 			pm_runtime_get_sync(omap_port->pdev);
-		}
-		spin_unlock(&omap_port->lock);
 		dev_dbg(&ssi->device, "Wake in high\n");
 		if (omap_port->wktest) { /* FIXME: HACK ! To be removed */
 			writel(SSI_WAKE(0),
@@ -1000,12 +994,8 @@ static irqreturn_t ssi_wake_thread(int irq __maybe_unused, void *ssi_port)
 				omap_ssi->sys + SSI_CLEAR_WAKE_REG(port->num));
 		}
 		hsi_event(port, HSI_EVENT_STOP_RX);
-		spin_lock(&omap_port->lock);
-		if (omap_port->wkin_cken) {
+		if (test_and_clear_bit(SSI_WAKE_EN, &omap_port->flags))
 			pm_runtime_put_sync(omap_port->pdev);
-			omap_port->wkin_cken = 0;
-		}
-		spin_unlock(&omap_port->lock);
 	}
 
 	return IRQ_HANDLED;
-- 
2.8.1

Back to linux.kernel | Previous | NextPrevious in thread | Next in thread | Find similar | Unroll thread


Thread

[RFC 00/19] Runtime PM fixes for omap-ssi Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 14/19] HSI: omap_ssi_port: avoid calling runtime_pm_*_sync inside spinlock Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 15/19] HSI: omap_ssi_port: replace pm_runtime_put_sync with non-sync variant Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 02/19] HSI: omap_ssi: do not reset module Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
    Re: [RFC 02/19] HSI: omap_ssi: do not reset module Tony Lindgren <tony@atomide.com> - 2016-06-21 13:50 +0200
  [RFC 12/19] HSI: omap_ssi_port: switch to threaded pio irq Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 13/19] HSI: omap_ssi_port: avoid pm_runtime_get_sync in ssi_start_dma and ssi_start_pio Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 11/19] HSI: omap_ssi_core: remove pm_runtime_get_sync call from tasklet Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 06/19] HSI: omap_ssi: convert cawake irq handler to thread Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 05/19] HSI: ssi_protocol: fix ssip_xmit invocation Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 18/19] HSI: omap_ssi_port: use rpm autosuspend API Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 08/19] HSI: core: switch port event notifier from atomic to blocking Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 07/19] HSI: omap_ssi_port: replace wkin_cken with atomic bitmap operations Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:30 +0200
  [RFC 09/19] HSI: omap_ssi_port: prepare start_tx/stop_tx for blocking pm_runtime calls Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:40 +0200
  [RFC 01/19] HSI: omap_ssi_port: remove useless newline Sebastian Reichel <sre@kernel.org> - 2016-06-18 18:40 +0200
  Re: [RFC 00/19] Runtime PM fixes for omap-ssi Pavel Machek <pavel@ucw.cz> - 2016-06-18 21:00 +0200
    Re: [RFC 00/19] Runtime PM fixes for omap-ssi Sebastian Reichel <sre@kernel.org> - 2016-06-18 22:30 +0200
      Re: [RFC 00/19] Runtime PM fixes for omap-ssi Pavel Machek <pavel@ucw.cz> - 2016-06-19 00:10 +0200

csiph-web