Thread (15 messages) 15 messages, 2 authors, 2021-04-01

Re: [PATCH v3 6/9] media: hantro: add fallback handling for single irq/clk

From: Ezequiel Garcia <hidden>
Date: 2021-03-31 19:58:52
Also in: linux-media, linux-rockchip

Hi Emil,

Thanks for the patch. Looking good!

A few comments below.

On Wed, 2021-03-31 at 18:35 +0100, Emil Velikov wrote:
From: Emil Velikov <redacted>

Currently the driver expects that each irq/clk will have a name
specified.

A valid point was raised by the DT maintainers - when there is a single
interrupt line or clock - the names are not needed.

This is handled:
 - clk - implicitly - ultimately we'll call of_clk_get_hw(..., 0, NULL
   which will get the first clock from the pmc
 - irq - explicitly - platform_get_irq(..., 0)

To gracefully handle potential bugs, add respective WARN_ON() if we're
having more than one irq/clk, yet lacking the respective names.

Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Since this is a new patch, I'm unsure where was this Acked-by?
quoted hunk ↗ jump to hunk
Suggested-by: Ezequiel Garcia <redacted>
Signed-off-by: Emil Velikov <redacted>
---
v3
 - New patch
---
 drivers/staging/media/hantro/hantro_drv.c | 24 +++++++++++++++++++++--
 1 file changed, 22 insertions(+), 2 deletions(-)
diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c
index e5f200e64993..d1294eb9cd07 100644
--- a/drivers/staging/media/hantro/hantro_drv.c
+++ b/drivers/staging/media/hantro/hantro_drv.c
@@ -752,8 +752,16 @@ static int hantro_probe(struct platform_device *pdev)
        if (!vpu->clocks)
                return -ENOMEM;
 
-       for (i = 0; i < vpu->variant->num_clocks; i++)
+       for (i = 0; i < vpu->variant->num_clocks; i++) {
                vpu->clocks[i].id = vpu->variant->clk_names[i];
+
+               /*
+                * Warn and refuse to load if the driver has multiple clocks,
+                * yet they are lacking the respective names.
+                */
+               if (WARN_ON(!vpu->variant->clk_names[i] && 
We already have more WARN_ON than we should in this driver,
I would just remove this condition entirely.
quoted hunk ↗ jump to hunk
+                       return -ENXIO;
+       }
        ret = devm_clk_bulk_get(&pdev->dev, vpu->variant->num_clocks,
                                vpu->clocks);
        if (ret)
@@ -791,7 +799,19 @@ static int hantro_probe(struct platform_device *pdev)
                if (!vpu->variant->irqs[i].handler)
                        continue;
 
-               irq = platform_get_irq_byname(vpu->pdev, irq_name);
+               /*
+                * If the driver has a single IRQ, chances are there will be no
+                * actual name in the DT bindings.
+                */
+               if (!irq_name) {
+                       if (WARN_ON(i))
+                               return -ENXIO;
+
+                       irq_name = "default";
+                       irq = platform_get_irq(vpu->pdev, 0);
+               } else {
+                       irq = platform_get_irq_byname(vpu->pdev, irq_name);
+               }
How about this instead:

irq = platform_get_irq_byname(vpu->pdev, irq_name);
if (irq <= 0)
	irq = platform_get_irq(vpu->pdev, i);

(and leave the irq name in the sama5d4 code).

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