Skip to contentSkip to conversation

media: qcom: iris: guard IRQ handler with runtime PM check

PATCH Superseded Hungyu Lin 2 participants 2 messages Jun 4

Subsystems (4)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Hungyu Lin

author
#0 lore raw

PATCH v2 1/1 media: qcom: iris: guard IRQ handler with runtime PM check

Use pm_runtime_get_if_active() before accessing hardware registers in the threaded IRQ handler. Skip interrupt processing when the device is not active.

---
v2:
- Use pm_runtime_get_if_active() instead of pm_runtime_get_if_in_use()
- Handle negative runtime PM return values correctly
- Return IRQ_NONE when interrupt processing is skipped

 drivers/media/platform/qcom/iris/iris_hfi_common.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/media/platform/qcom/iris/iris_hfi_common.c +9 -0
index 621c66593d88..1c05f83986fe 100644
@@ -100,10 +100,17 @@ irqreturn_t iris_hfi_isr(int irq, void *data)
 irqreturn_t iris_hfi_isr_handler(int irq, void *data)
 {
 	struct iris_core *core = data;
+	int ret;
 
 	if (!core)
 		return IRQ_NONE;
 
+	ret = pm_runtime_get_if_active(core->dev);
+	if (ret <= 0) {
+		enable_irq(irq);
+		return IRQ_NONE;
+	}
+
 	mutex_lock(&core->lock);
 	pm_runtime_mark_last_busy(core->dev);
 	iris_vpu_clear_interrupt(core);
@@ -111,6 +118,8 @@ irqreturn_t iris_hfi_isr_handler(int irq, void *data)
 
 	core->hfi_response_ops->hfi_response_handler(core);
 
+	pm_runtime_put_autosuspend(core->dev);
+
 	if (!iris_vpu_watchdog(core, core->intr_status))
 		enable_irq(irq);
 
Signature · 1 line
2.34.1
#1 lore raw
Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:

New issues:
- [Critical] Unmasking a level-triggered IRQ without clearing the interrupt source when pm_runtime_get_if_active() returns <= 0 causes an infinite interrupt storm and potential system deadlock.
- [High] core->intr_status is read locklessly after dropping the Runtime PM reference, causing a race that leads to unbalanced enable_irq() and potentially missing watchdog errors.

Pre-existing issues:
- [High] Missing cancel_delayed_work_sync() for sys_error_handler in driver removal path leads to Use-After-Free. --

commit 0c8846847b29684af89fca24c8cacbe7c8cb8ceb
Author: Hungyu Lin <[email protected]>

media: qcom: iris: guard IRQ handler with runtime PM check

This commit adds a runtime PM check to the threaded IRQ handler using pm_runtime_get_if_active(). It skips interrupt processing if the device is not active, preventing hardware access on suspended devices.

Quoted 9 lines
drivers/media/platform/qcom/iris/iris_hfi_common.c +1 -0
index 621c66593d88d..1c05f83986fe2 100644
@@ -100,10 +100,17 @@ irqreturn_t iris_hfi_isr(int irq, void *data)
 irqreturn_t iris_hfi_isr_handler(int irq, void *data)
 {
 	struct iris_core *core = data;
+	int ret;

drivers/media/platform/qcom/iris/iris_hfi_common.c irqreturn_t iris_hfi_isr(int irq, void *data) · line 113

Quoted 10 lines
drivers/media/platform/qcom/iris/iris_hfi_common.c +6 -0
 
 	if (!core)
 		return IRQ_NONE;
 
+	ret = pm_runtime_get_if_active(core->dev);
+	if (ret <= 0) {
+		enable_irq(irq);
+		return IRQ_NONE;
+	}
+

[Severity: Critical]
Is it safe to unmask the IRQ without clearing the hardware source first?

Because the interrupt is level-triggered, if pm_runtime_get_if_active() bails out (for instance, during RPM_RESUMING transitions or if CONFIG_PM is disabled), calling enable_irq() without clearing the interrupt via iris_vpu_clear_interrupt() might trigger an immediate interrupt storm.

Would this cause the CPU to lock up in an infinite loop?

Quoted 3 lines
drivers/media/platform/qcom/iris/iris_hfi_common.c +0 -0
 	mutex_lock(&core->lock);
 	pm_runtime_mark_last_busy(core->dev);
 	iris_vpu_clear_interrupt(core);

[ ... ]

drivers/media/platform/qcom/iris/iris_hfi_common.c irqreturn_t iris_hfi_isr_handler(int irq, void *data) · line 125

Quoted 9 lines
drivers/media/platform/qcom/iris/iris_hfi_common.c +2 -0
@@ -111,6 +118,8 @@ irqreturn_t iris_hfi_isr_handler(int irq, void *data)
 
 	core->hfi_response_ops->hfi_response_handler(core);
 
+	pm_runtime_put_autosuspend(core->dev);
+
 	if (!iris_vpu_watchdog(core, core->intr_status))
 		enable_irq(irq);
 

[Severity: High]
Does dropping the runtime PM reference before checking core->intr_status introduce a race condition?

If the device autosuspends immediately after pm_runtime_put_autosuspend() and is concurrently resumed, the resume path could clear core->intr_status and call enable_irq() before iris_vpu_watchdog() runs here.

If the threaded handler then reads the newly cleared core->intr_status (0), would it execute an unbalanced enable_irq() and potentially miss watchdog errors?

Quoted 2 lines
drivers/media/platform/qcom/iris/iris_probe.c +0 -0

[ ... ]

Quoted 17 lines

static void iris_remove(struct platform_device *pdev)

{
	struct iris_core *core;

	core = platform_get_drvdata(pdev);
	if (!core)
		return;

	iris_core_deinit(core);

	video_unregister_device(core->vdev_dec);
	video_unregister_device(core->vdev_enc);

	v4l2_device_unregister(&core->v4l2_dev);

	mutex_destroy(&core->lock);
}

[Severity: High]
This isn't a bug introduced by this patch, but does iris_remove() need to call cancel_delayed_work_sync() for sys_error_handler before destroying the lock and returning?

Because the core structure is allocated with devm_kzalloc, devres will free it after iris_remove() completes. If there is pending sys_error_handler work, could it execute after the memory is freed or the mutex is destroyed, leading to a use-after-free?

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1