On Tue, 10 Mar 2020, Peter Chen wrote:
> From: Peter Chen <peter.chen@xxxxxxx>
>
> The code calls pm_runtime_get_sync with irq disabled, it causes below
> warning:
>
> BUG: sleeping function called from invalid context at
> wer/runtime.c:1075
> in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid:
> er/u8:1
...
> Tested-by: Dmitry Osipenko <digetx@xxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> #v5.5
> Fixes: 72dc8df7920f ("usb: chipidea: udc: protect usb interrupt enable")
> Reported-by: Dmitry Osipenko <digetx@xxxxxxxxx>
> Signed-off-by: Peter Chen <peter.chen@xxxxxxx>
> ---
> drivers/usb/chipidea/udc.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/chipidea/udc.c b/drivers/usb/chipidea/udc.c
> index ffaf46f5d062..1fa587ec52fc 100644
> --- a/drivers/usb/chipidea/udc.c
> +++ b/drivers/usb/chipidea/udc.c
> @@ -1539,9 +1539,11 @@ static void ci_hdrc_gadget_connect(struct usb_gadget *_gadget, int is_active)
> if (ci->driver) {
> hw_device_state(ci, ci->ep0out->qh.dma);
> usb_gadget_set_state(_gadget, USB_STATE_POWERED);
> + spin_unlock_irqrestore(&ci->lock, flags);
> usb_udc_vbus_handler(_gadget, true);
> + } else {
> + spin_unlock_irqrestore(&ci->lock, flags);
> }
> - spin_unlock_irqrestore(&ci->lock, flags);
There's something strange about this patch.
Do you really know that interrupts will be enabled following the
spin_unlock_irqrestore()? In other words, do you know that interrupts
were enabled upon entry to this routine?
If they were, then why use spin_lock_irqsave/spin_unlock_irqrestore?
Why not use spin_lock_irq and spin_unlock_irq?
Alan Stern
