Hello!
On 10/21/21 12:09 PM, Dmitry Osipenko wrote:
[...]
>> If platform_get_irq() returns IRQ0 (considered invalid according to Linus)
>> the driver blithely passes it to usb_add_hcd() that treats IRQ0 as no IRQ
>> at all. Deny IRQ0 right away, returning -EINVAL from the probe() method...
>>
>> Fixes: e84fce0f8837 ("usb: xhci: Add NVIDIA Tegra XUSB controller driver")
>> Signed-off-by: Sergey Shtylyov <s.shtylyov@xxxxxx>
>> ---
>> drivers/usb/host/xhci-tegra.c | 2 ++
>> 1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/usb/host/xhci-tegra.c b/drivers/usb/host/xhci-tegra.c
>> index 1bf494b649bd..7151b1d4f876 100644
>> --- a/drivers/usb/host/xhci-tegra.c
>> +++ b/drivers/usb/host/xhci-tegra.c
>> @@ -1439,6 +1439,8 @@ static int tegra_xusb_probe(struct platform_device *pdev)
>> tegra->xhci_irq = platform_get_irq(pdev, 0);
>> if (tegra->xhci_irq < 0)
>> return tegra->xhci_irq;
>> + if (!tegra->xhci_irq)
>> + return -ENIVAL;
>>
>> tegra->mbox_irq = platform_get_irq(pdev, 1);
>> if (tegra->mbox_irq < 0)
>>
>
> platform_get_irq() never returns zero in accordance to [1], but I see
> that it can return it [2].
Not only that, it also can be returned thru the normal path (from an IRQ descriptor).
I'm not sure whether 0 means an IRQ0 returned from acpi_dev_gpio_irq_get(), looks like yes...
> Should be better to fix [2] and return -EINVAL.
No, we have WARN() before returning IRQ0 -- if we're going to finally declare IRQ0 invalid,
it should be done after this check.
> [1]
> https://elixir.bootlin.com/linux/v5.15-rc6/source/drivers/base/platform.c#L254
>
> [2]
> https://elixir.bootlin.com/linux/v5.15-rc6/source/drivers/base/platform.c#L226
MBR, Sergey
