Skip to contentSkip to conversation

mv_xor: add support for clkdev framework

PATCH In review Saeed Bishara 4 participants 9 messages Jun 23, 2009 – Jun 24, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mv_xor: add support for clkdev framework

---
 drivers/dma/mv_xor.c |   16 ++++++++++++++++
 drivers/dma/mv_xor.h |    4 ++++
 2 files changed, 20 insertions(+), 0 deletions(-)
drivers/dma/mv_xor.c +16 -0
index ddab94f..db77903 100644
@@ -1325,6 +1325,14 @@ static int mv_xor_shared_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, msp);
 
+#if defined(CONFIG_HAVE_CLK)
+	msp->clk = clk_get(&pdev->dev, NULL);
+	if (IS_ERR(msp->clk))
+		dev_notice(&pdev->dev, "cannot get clkdev\n");
+	else
+		clk_enable(msp->clk);
+#endif
+
 	/*
 	 * (Re-)program MBUS remapping windows if we are asked to.
 	 */
@@ -1336,6 +1344,14 @@ static int mv_xor_shared_probe(struct platform_device *pdev)
 
 static int mv_xor_shared_remove(struct platform_device *pdev)
 {
+#if defined(CONFIG_HAVE_CLK)
+	struct mv_xor_shared_private *msp = platform_get_drvdata(pdev);
+
+	if (!IS_ERR(msp->clk)) {
+		clk_disable(msp->clk);
+		clk_put(msp->clk);
+	}
+#endif
 	return 0;
 }
 
drivers/dma/mv_xor.h +4 -0
index 06cafe1..08c3815 100644
@@ -22,6 +22,7 @@
 #include <linux/io.h>
 #include <linux/dmaengine.h>
 #include <linux/interrupt.h>
+#include <linux/clk.h>
 
 #define USE_TIMER
 #define MV_XOR_SLOT_SIZE		64
@@ -55,6 +56,9 @@
 struct mv_xor_shared_private {
 	void __iomem	*xor_base;
 	void __iomem	*xor_high_base;
+#if defined(CONFIG_HAVE_CLK)
+	struct clk		*clk;
+#endif
 };
 
 
Signature · 1 line
1.6.0.4
#1 lore raw

PATCH 1/1 mv_xor: add support for power management

---
 drivers/dma/mv_xor.c |   56 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 56 insertions(+), 0 deletions(-)
drivers/dma/mv_xor.c +56 -0
index db77903..ecacd32 100644
@@ -1284,9 +1284,38 @@ mv_xor_conf_mbus_windows(struct mv_xor_shared_private *msp,
 	writel(win_enable, base + WINDOW_BAR_ENABLE(1));
 }
 
+#ifdef CONFIG_PM
+static int mv_xor_suspend(struct platform_device *pdev, pm_message_t state)
+{
+	return 0;
+}
+
+static int mv_xor_resume(struct platform_device *pdev)
+{
+	struct mv_xor_device *device = platform_get_drvdata(pdev);
+	struct dma_chan *chan, *_chan;
+	struct mv_xor_chan *mv_chan;
+
+	list_for_each_entry_safe(chan, _chan, &device->common.channels,
+				device_node) {
+		mv_chan = to_mv_xor_chan(chan);
+		mv_chan_unmask_interrupts(mv_chan);
+
+		mv_set_mode(mv_chan, mv_chan->current_type);
+	}
+	return 0;
+}
+
+#else
+#define mv_xor_suspend NULL
+#define mv_xor_resume NULL
+#endif
+
 static struct platform_driver mv_xor_driver = {
 	.probe		= mv_xor_probe,
 	.remove		= __devexit_p(mv_xor_remove),
+	.suspend	= mv_xor_suspend,
+	.resume		= mv_xor_resume,
 	.driver		= {
 		.owner	= THIS_MODULE,
 		.name	= MV_XOR_NAME,
@@ -1355,9 +1384,36 @@ static int mv_xor_shared_remove(struct platform_device *pdev)
 	return 0;
 }
 
+#ifdef CONFIG_PM
+static int mv_xor_shared_suspend(struct platform_device *pdev,
+				 pm_message_t state)
+{
+	return 0;
+}
+
+static int mv_xor_shared_resume(struct platform_device *pdev)
+{
+	struct mv_xor_platform_shared_data *msd = pdev->dev.platform_data;
+	struct mv_xor_shared_private *msp = platform_get_drvdata(pdev);
+	/*
+	 * (Re-)program MBUS remapping windows if we are asked to.
+	 */
+	if (msd != NULL && msd->dram != NULL)
+		mv_xor_conf_mbus_windows(msp, msd->dram);
+
+	return 0;
+}
+
+#else
+#define mv_xor_shared_suspend NULL
+#define mv_xor_shared_resume NULL
+#endif
+
 static struct platform_driver mv_xor_shared_driver = {
 	.probe		= mv_xor_shared_probe,
 	.remove		= mv_xor_shared_remove,
+	.suspend	= mv_xor_shared_suspend,
+	.resume		= mv_xor_shared_resume,
 	.driver		= {
 		.owner	= THIS_MODULE,
 		.name	= MV_XOR_SHARED_NAME,
Signature · 1 line
1.6.0.4
#2 lore raw
In reply to Saeed Bishara
Quoted 2 lines · Saeed Bishara

---

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge window requires some extra justification and acks because I will not have a chance to get any exposure of this patch in linux-next, in general patches should be available before the merge window opens. Per the schedule the merge window can close any day now, so this may get delayed until 2.6.32.

#3 lore raw
Quoted 7 lines · Dan Williams
Quoted 3 lines · Saeed Bishara

---

Can you expand the changelog to explain why this patch is needed and summarize what it does?

ok, I'll re-send the patch
 Also, releasing patches this late in the merge

Quoted 4 lines

window requires some extra justification and acks because I will not have a chance to get any exposure of this patch in linux-next, in general patches should be available before the merge window opens.  Per the schedule the merge window can close any day now, so this may get delayed until 2.6.32.

actually I was not looking to merge it for next release, I don't mind if the patch merged only in 2.6.32. sorry for not mentioning that.

Quoted 2 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#4 lore raw
In reply to Dan Williams
Quoted 11 lines · Dan Williams
Quoted 2 lines · Saeed Bishara

---

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge window requires some extra justification and acks because I will not have a chance to get any exposure of this patch in linux-next, in general patches should be available before the merge window opens. Per the schedule the merge window can close any day now, so this may get delayed until 2.6.32.

I don't expect for the clock API counter part to this patch to make it in this merge window either, so there is no rush.

Nicolas

#5 lore raw
Quoted 4 lines
Quoted 4 lines

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge

Here is the patch with more explanation:
----------------------------------------<
From c999ed53a913979f1739854cb7c881a9dde980c6 Mon Sep 17 00:00:00 2001
From: Saeed Bishara <[email protected]>
Date: Tue, 12 May 2009 13:29:34 +0300
Subject: [PATCH 1/2 v2] mv_xor: add support for clkdev framework

When the clkdev framework is enabled, the device's clock is gated by default, this patch uses the framework to enable the clock when using the driver loaded and disables it when unloading the driver.

---
 drivers/dma/mv_xor.c |   16 ++++++++++++++++
 drivers/dma/mv_xor.h |    4 ++++
 2 files changed, 20 insertions(+), 0 deletions(-)
drivers/dma/mv_xor.c +16 -0
index ddab94f..db77903 100644
@@ -1325,6 +1325,14 @@ static int mv_xor_shared_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, msp);
 
+#if defined(CONFIG_HAVE_CLK)
+	msp->clk = clk_get(&pdev->dev, NULL);
+	if (IS_ERR(msp->clk))
+		dev_notice(&pdev->dev, "cannot get clkdev\n");
+	else
+		clk_enable(msp->clk);
+#endif
+
 	/*
 	 * (Re-)program MBUS remapping windows if we are asked to.
 	 */
@@ -1336,6 +1344,14 @@ static int mv_xor_shared_probe(struct platform_device *pdev)
 
 static int mv_xor_shared_remove(struct platform_device *pdev)
 {
+#if defined(CONFIG_HAVE_CLK)
+	struct mv_xor_shared_private *msp = platform_get_drvdata(pdev);
+
+	if (!IS_ERR(msp->clk)) {
+		clk_disable(msp->clk);
+		clk_put(msp->clk);
+	}
+#endif
 	return 0;
 }
 
drivers/dma/mv_xor.h +4 -0
index 06cafe1..08c3815 100644
@@ -22,6 +22,7 @@
 #include <linux/io.h>
 #include <linux/dmaengine.h>
 #include <linux/interrupt.h>
+#include <linux/clk.h>
 
 #define USE_TIMER
 #define MV_XOR_SLOT_SIZE		64
@@ -55,6 +56,9 @@
 struct mv_xor_shared_private {
 	void __iomem	*xor_base;
 	void __iomem	*xor_high_base;
+#if defined(CONFIG_HAVE_CLK)
+	struct clk		*clk;
+#endif
 };
 
 
Signature · 1 line
1.6.0.4
#6 lore raw
Quoted 3 lines · Saeed Bishara
Quoted 2 lines
Quoted 2 lines

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge

Here is the patch with more explanation:

Thanks, and thanks for the clarification about the target kernel.

Regards,
Dan

#7 lore raw
In reply to Saeed Bishara
Quoted 7 lines · Saeed Bishara
Quoted 4 lines
Quoted 4 lines

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge

Here is the patch with more explanation:
----------------------------------------<

Quoted 1 line

From c999ed53a913979f1739854cb7c881a9dde980c6 Mon Sep 17 00:00:00 2001

Quoted 8 lines

From: Saeed Bishara <[email protected]>
Date: Tue, 12 May 2009 13:29:34 +0300
Subject: [PATCH 1/2 v2] mv_xor: add support for clkdev framework

When the clkdev framework is enabled, the device's clock is gated by default, this patch uses the framework to enable the clock when using the driver loaded and disables it when unloading the driver.

Can't you be more agressive with clock gating?

Like using clk_get() when loading the driver and clk_put() when unloading, but using clk_enable()/clk_disable() only when a DMA channel is actually requested or active.

Nicolas

#8 lore raw
Quoted 18 lines · Nicolas Pitre
Quoted 7 lines · Saeed Bishara
Quoted 4 lines
Quoted 4 lines

Can you expand the changelog to explain why this patch is needed and summarize what it does? Also, releasing patches this late in the merge

Here is the patch with more explanation:
----------------------------------------<

Quoted 1 line

From c999ed53a913979f1739854cb7c881a9dde980c6 Mon Sep 17 00:00:00 2001

Quoted 8 lines

From: Saeed Bishara <[email protected]>
Date: Tue, 12 May 2009 13:29:34 +0300
Subject: [PATCH 1/2 v2] mv_xor: add support for clkdev framework

When the clkdev framework is enabled, the device's clock is gated by default, this patch uses the framework to enable the clock when using the driver loaded and disables it when unloading the driver.

Quoted 5 lines

Can't you be more agressive with clock gating?

Like using clk_get() when loading the driver and clk_put() when unloading, but using clk_enable()/clk_disable() only when a DMA channel is actually requested or active.

I agree with your suggestion, I'll add it to my todo list :). however, I think that this should be managed by the DMA engine stack, low level drivers should only provide the run time suspend/resume functions, I think that can be done on top of the runtime PM framework, see
http://lwn.net/Articles/336729/

Quoted 2 lines

Nicolas