Thread (38 messages) 38 messages, 5 authors, 2026-04-02

Re: [PATCH v2 08/10] dmaengine: imx-sdma: make use of devm_add_action_or_reset to unregiser the dma_device

From: Frank Li <Frank.li@nxp.com>
Date: 2025-09-12 16:37:54
Also in: dmaengine, imx, lkml

On Fri, Sep 12, 2025 at 05:25:08PM +0200, Marco Felsch wrote:
On 25-09-12, Frank Li wrote:
quoted
On Thu, Sep 11, 2025 at 11:56:49PM +0200, Marco Felsch wrote:
quoted
Make use of the devm_add_action_or_reset() to register a custom devm_
release hook. This is required to turn off the IRQs before calling
dma_async_device_unregister().

Furthermore it removes the last goto error handling within probe() and
trims the remove().

Make use of disable_irq() and let the devm-irq do the job to free the
IRQ, because the only purpose of using devm_free_irq() was to disable
the IRQ before calling dma_async_device_unregister().

Signed-off-by: Marco Felsch <redacted>
---
 drivers/dma/imx-sdma.c | 23 +++++++++++++++--------
 1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
index d39589c20c4b2a26d0239feb86cce8d5a0f5acdd..d6d0d4300f540268a3ab4a6b14af685f7b93275a 100644
--- a/drivers/dma/imx-sdma.c
+++ b/drivers/dma/imx-sdma.c
@@ -2264,6 +2264,14 @@ static struct dma_chan *sdma_xlate(struct of_phandle_args *dma_spec,
 				     ofdma->of_node);
 }

+static void sdma_dma_device_unregister_action(void *data)
+{
+	struct sdma_engine *sdma = data;
+
+	disable_irq(sdma->irq);
May not related this cleanup patch, I am just curious why not mask sdma irq
request.
You mean by setting irq_set_status_flags(irq, IRQ_DISABLE_UNLAZY)
infront? Not sure if this is required since this is just the cleanup
path.
It is not important for this patch.
quoted
quoted
+	dma_async_device_unregister(&sdma->dma_device);
+}
+
 static int sdma_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -2388,10 +2396,16 @@ static int sdma_probe(struct platform_device *pdev)
 		return ret;
 	}

+	ret = devm_add_action_or_reset(dev, sdma_dma_device_unregister_action, sdma);
+	if (ret) {
+		dev_err(dev, "Unable to register release hook\n");
+		return ret;
+	}
why not use dev_err_probe() her?
Please see my last patch.
You can direct use dev_err_probe() here, instead replace all in last one.

Frank
Regards,
  Marco
quoted
quoted
+
 	ret = of_dma_controller_register(np, sdma_xlate, sdma);
 	if (ret) {
 		dev_err(dev, "failed to register controller\n");
-		goto err_register;
+		return ret;
the same here.

Frank
quoted
 	}

 	/*
@@ -2410,11 +2424,6 @@ static int sdma_probe(struct platform_device *pdev)
 	}

 	return 0;
-
-err_register:
-	dma_async_device_unregister(&sdma->dma_device);
-
-	return ret;
 }

 static void sdma_remove(struct platform_device *pdev)
@@ -2423,8 +2432,6 @@ static void sdma_remove(struct platform_device *pdev)
 	int i;

 	of_dma_controller_free(sdma->dev->of_node);
-	devm_free_irq(&pdev->dev, sdma->irq, sdma);
-	dma_async_device_unregister(&sdma->dma_device);
 	/* Kill the tasklet */
 	for (i = 0; i < MAX_DMA_CHANNELS; i++) {
 		struct sdma_channel *sdmac = &sdma->channel[i];

--
2.47.3
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help