Skip to contentSkip to conversation

dmaengine: sprd: Fix runtime PM reference leak in probe

PATCH Applied Ruoyu Wang 5 participants 5 messages Aug 13 – Sep 3

Subsystems (2)
lore
Share & tools
lore

Ruoyu Wang

author
#0 lore raw

PATCH 1/1 dmaengine: sprd: Fix runtime PM reference leak in probe

pm_runtime_get_sync() increments a device's usage counter even when it fails. sprd_dma_probe() currently jumps directly to controller clock cleanup on that error, bypassing both pm_runtime_put_noidle() and pm_runtime_disable(). This can happen if the preceding unchecked pm_runtime_set_active() fails and the following runtime-resume attempt also returns an error.

Enter the existing runtime-PM unwind path instead. This drops the reference without idling the partially initialized device, disables runtime PM, and then releases the controller clocks. The success path and propagated error code are unchanged.

This issue was found by a static analysis checker and confirmed by manual source review.

Fixes 9b3b8171f7f4 ("dmaengine: sprd: Add Spreadtrum DMA driver")
---
 drivers/dma/sprd-dma.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/dma/sprd-dma.c +1 -2
index 087fea3af2e411..19b32a23c882de 100644
@@ -1212,7 +1212,7 @@ static int sprd_dma_probe(struct platform_device *pdev)
 
 	ret = pm_runtime_get_sync(&pdev->dev);
 	if (ret < 0)
-		goto err_rpm;
+		goto err_register;
 
 	ret = dma_async_device_register(&sdev->dma_dev);
 	if (ret < 0) {
@@ -1234,7 +1234,6 @@ static int sprd_dma_probe(struct platform_device *pdev)
 err_register:
 	pm_runtime_put_noidle(&pdev->dev);
 	pm_runtime_disable(&pdev->dev);
-err_rpm:
 	sprd_dma_disable(sdev);
 	return ret;
 }
Signature · 1 line
2.51.0
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Premature IRQ registration and incorrect devres teardown order leading to UAF or bus errors.
- [High] The interrupt handler accesses hardware registers without checking if the device is active. --

--- Patch [1]: [PATCH] dmaengine: sprd: Fix runtime PM reference leak in probe --- commit 274e07c9f2493c369c243df16cab2b58e4834d15
Author: Ruoyu Wang <[email protected]>

dmaengine: sprd: Fix runtime PM reference leak in probe

The commit modifies the error handling path in sprd_dma_probe() to correctly unwind runtime PM state. If pm_runtime_get_sync() fails, it now jumps to err_register to drop the reference without idling the partially initialized device, and disables runtime PM before turning off clocks.

drivers/dma/sprd-dma.c static int sprd_dma_probe(struct platform_device *pdev) · line 1215

Quoted 10 lines
drivers/dma/sprd-dma.c +1 -1
index 087fea3af2e41..19b32a23c882d 100644
@@ -1212,7 +1212,7 @@ static int sprd_dma_probe(struct platform_device *pdev)
 
 	ret = pm_runtime_get_sync(&pdev->dev);
 	if (ret < 0)
-		goto err_rpm;
+		goto err_register;

[Severity: High]
This is a pre-existing issue, but does the devm cleanup order risk a NULL pointer dereference or accessing unmapped memory on probe failure?

Looking at sprd_dma_probe(), the IRQ is requested before the device IO memory is mapped:

drivers/dma/sprd-dma.c:sprd_dma_probe() {
    ...
	if (sdev->irq > 0) {
		ret = devm_request_irq(&pdev->dev, sdev->irq, dma_irq_handle,
				       0, "sprd_dma", (void *)sdev);
    ...
	sdev->glb_base = devm_platform_ioremap_resource(pdev, 0);
    ...
}

If an interrupt is already pending on the line, it could fire immediately upon unmasking. This would cause dma_irq_handle() to read from sdev->glb_base before it is initialized, resulting in a NULL pointer dereference.

Additionally, because devm_request_irq() is called before devm_platform_ioremap_resource(), devres will unmap the memory before freeing the IRQ during probe failure teardown. If an interrupt fires during this window, could the handler access unmapped memory?

[Severity: High]
This is also a pre-existing issue, but does the interrupt handler safely access hardware registers when the device might be runtime suspended?

drivers/dma/sprd-dma.c:dma_irq_handle() {
    ...
	u32 irq_status = readl(sdev->glb_base + SPRD_DMA_GLB_INT_MSK_STS);
    ...
}

If a spurious interrupt occurs while the device is suspended and clocks are disabled, dma_irq_handle() unconditionally reads from the hardware register.

Should the handler verify the device's runtime PM state via pm_runtime_get_if_active() before accessing registers to prevent a synchronous external abort?

[ ... ]

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v1 1/1
#2 lore raw
In reply to Ruoyu Wang
Quoted 12 lines · Ruoyu Wang

pm_runtime_get_sync() increments a device's usage counter even when it fails. sprd_dma_probe() currently jumps directly to controller clock cleanup on that error, bypassing both pm_runtime_put_noidle() and pm_runtime_disable(). This can happen if the preceding unchecked pm_runtime_set_active() fails and the following runtime-resume attempt also returns an error.

Enter the existing runtime-PM unwind path instead. This drops the reference without idling the partially initialized device, disables runtime PM, and then releases the controller clocks. The success path and propagated error code are unchanged.

Quoted 6 lines

This issue was found by a static analysis checker and confirmed by manual source review.

Fixes 9b3b8171f7f4 ("dmaengine: sprd: Add Spreadtrum DMA driver")

---

LGTM. Thanks.

Quoted 24 lines
  drivers/dma/sprd-dma.c | 3 +--
  1 file changed, 1 insertion(+), 2 deletions(-)
drivers/dma/sprd-dma.c +1 -2
index 087fea3af2e411..19b32a23c882de 100644
@@ -1212,7 +1212,7 @@ static int sprd_dma_probe(struct platform_device *pdev)
 
  	ret = pm_runtime_get_sync(&pdev->dev);
  	if (ret < 0)
-		goto err_rpm;
+		goto err_register;
 
  	ret = dma_async_device_register(&sdev->dma_dev);
  	if (ret < 0) {
@@ -1234,7 +1234,6 @@ static int sprd_dma_probe(struct platform_device *pdev)
  err_register:
  	pm_runtime_put_noidle(&pdev->dev);
  	pm_runtime_disable(&pdev->dev);
-err_rpm:
  	sprd_dma_disable(sdev);
  	return ret;
  }
Reviewed-by on v1 1/1
#3 lore raw
In reply to Ruoyu Wang
Quoted 12 lines · Ruoyu Wang

pm_runtime_get_sync() increments a device's usage counter even when it fails. sprd_dma_probe() currently jumps directly to controller clock cleanup on that error, bypassing both pm_runtime_put_noidle() and pm_runtime_disable(). This can happen if the preceding unchecked pm_runtime_set_active() fails and the following runtime-resume attempt also returns an error.

Enter the existing runtime-PM unwind path instead. This drops the reference without idling the partially initialized device, disables runtime PM, and then releases the controller clocks. The success path and propagated error code are unchanged.

Quoted 6 lines

This issue was found by a static analysis checker and confirmed by manual source review.

Fixes 9b3b8171f7f4 ("dmaengine: sprd: Add Spreadtrum DMA driver")

---

Quoted 27 lines
 drivers/dma/sprd-dma.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/dma/sprd-dma.c +1 -3
index 087fea3af2e411..19b32a23c882de 100644
@@ -1212,7 +1212,7 @@ static int sprd_dma_probe(struct platform_device *pdev)
 
 	ret = pm_runtime_get_sync(&pdev->dev);
 	if (ret < 0)
-		goto err_rpm;
+		goto err_register;
 
 	ret = dma_async_device_register(&sdev->dma_dev);
 	if (ret < 0) {
@@ -1234,7 +1234,6 @@ static int sprd_dma_probe(struct platform_device *pdev)
 err_register:
 	pm_runtime_put_noidle(&pdev->dev);
 	pm_runtime_disable(&pdev->dev);
-err_rpm:
 	sprd_dma_disable(sdev);
 	return ret;
 }
--

2.51.0

maintainer Applied
#4 lore raw
In reply to Ruoyu Wang
Quoted 8 lines · Ruoyu Wang

pm_runtime_get_sync() increments a device's usage counter even when it fails. sprd_dma_probe() currently jumps directly to controller clock cleanup on that error, bypassing both pm_runtime_put_noidle() and pm_runtime_disable(). This can happen if the preceding unchecked pm_runtime_set_active() fails and the following runtime-resume attempt also returns an error.

[...]

Applied, thanks!

[1/1] dmaengine: sprd: Fix runtime PM reference leak in probe

      commit: a7df136ec529ee49a789c5029bc37b98b0d4bedd

Best regards,

Signature · 1 line
~Vinod