[PATCH] serial: 8250_lpss: Switch over to MSI interrupts

[PATCH] serial: 8250_lpss: Switch over to MSI interrupts

[Felipe Balbi]

Some devices support MSI interrupts. Let's at least try to use them in
platforms that provide MSI capability.

Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
---
 drivers/tty/serial/8250/8250_lpss.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/tty/serial/8250/8250_lpss.c b/drivers/tty/serial/8250/8250_lpss.c
index 5f72ef3ea574..f0f7a2777557 100644
--- a/drivers/tty/serial/8250/8250_lpss.c
+++ b/drivers/tty/serial/8250/8250_lpss.c
@@ -293,16 +293,22 @@ static int lpss8250_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (ret)
 		return ret;
 
+	pci_set_master(pdev);
+
 	lpss = devm_kzalloc(&pdev->dev, sizeof(*lpss), GFP_KERNEL);
 	if (!lpss)
 		return -ENOMEM;
 
+	ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);
+	if (ret < 0)
+		return ret;
+
 	lpss->board = (struct lpss8250_board *)id->driver_data;
 
 	memset(&uart, 0, sizeof(struct uart_8250_port));
 
 	uart.port.dev = &pdev->dev;
-	uart.port.irq = pdev->irq;
+	uart.port.irq = pci_irq_vector(pdev, 0);
 	uart.port.private_data = &lpss->data;
 	uart.port.type = PORT_16550A;
 	uart.port.iotype = UPIO_MEM;
@@ -337,6 +343,7 @@ static int lpss8250_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 err_exit:
 	if (lpss->board->exit)
 		lpss->board->exit(lpss);
+	pci_free_irq_vectors(pdev);
 	return ret;
 }
 
@@ -348,6 +355,7 @@ static void lpss8250_remove(struct pci_dev *pdev)
 
 	if (lpss->board->exit)
 		lpss->board->exit(lpss);
+	pci_free_irq_vectors(pdev);
 }
 
 static const struct lpss8250_board byt_board = {
-- 
2.23.0

Re: [PATCH] serial: 8250_lpss: Switch over to MSI interrupts

[Andy Shevchenko]

On Tue, Oct 01, 2019 at 11:16:17AM +0300, Felipe Balbi wrote:
> Some devices support MSI interrupts. Let's at least try to use them in
> platforms that provide MSI capability.

Thanks for the patch!

I think you may clean up qrk_serial_setup() as well.

> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
> ---
>  drivers/tty/serial/8250/8250_lpss.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/tty/serial/8250/8250_lpss.c b/drivers/tty/serial/8250/8250_lpss.c
> index 5f72ef3ea574..f0f7a2777557 100644
> --- a/drivers/tty/serial/8250/8250_lpss.c
> +++ b/drivers/tty/serial/8250/8250_lpss.c
> @@ -293,16 +293,22 @@ static int lpss8250_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>  	if (ret)
>  		return ret;
>  
> +	pci_set_master(pdev);
> +
>  	lpss = devm_kzalloc(&pdev->dev, sizeof(*lpss), GFP_KERNEL);
>  	if (!lpss)
>  		return -ENOMEM;
>  
> +	ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);
> +	if (ret < 0)
> +		return ret;
> +
>  	lpss->board = (struct lpss8250_board *)id->driver_data;
>  
>  	memset(&uart, 0, sizeof(struct uart_8250_port));
>  
>  	uart.port.dev = &pdev->dev;
> -	uart.port.irq = pdev->irq;
> +	uart.port.irq = pci_irq_vector(pdev, 0);
>  	uart.port.private_data = &lpss->data;
>  	uart.port.type = PORT_16550A;
>  	uart.port.iotype = UPIO_MEM;
> @@ -337,6 +343,7 @@ static int lpss8250_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>  err_exit:
>  	if (lpss->board->exit)
>  		lpss->board->exit(lpss);
> +	pci_free_irq_vectors(pdev);
>  	return ret;
>  }
>  
> @@ -348,6 +355,7 @@ static void lpss8250_remove(struct pci_dev *pdev)
>  
>  	if (lpss->board->exit)
>  		lpss->board->exit(lpss);
> +	pci_free_irq_vectors(pdev);
>  }
>  
>  static const struct lpss8250_board byt_board = {
> -- 
> 2.23.0
> 

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] serial: 8250_lpss: Switch over to MSI interrupts

[Felipe Balbi]

[-- Attachment #1: Type: text/plain, Size: 408 bytes --]


Hi,

Andy Shevchenko <andriy.shevchenko@linux.intel.com> writes:

> On Tue, Oct 01, 2019 at 11:16:17AM +0300, Felipe Balbi wrote:
>> Some devices support MSI interrupts. Let's at least try to use them in
>> platforms that provide MSI capability.
>
> Thanks for the patch!
>
> I think you may clean up qrk_serial_setup() as well.

That would break qrk_serial_setup_dma(), though.

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

Re: [PATCH] serial: 8250_lpss: Switch over to MSI interrupts

[Andy Shevchenko]

On Tue, Oct 01, 2019 at 12:35:36PM +0300, Felipe Balbi wrote:
> Andy Shevchenko <andriy.shevchenko@linux.intel.com> writes:
> > On Tue, Oct 01, 2019 at 11:16:17AM +0300, Felipe Balbi wrote:
> >> Some devices support MSI interrupts. Let's at least try to use them in
> >> platforms that provide MSI capability.
> >
> > Thanks for the patch!
> >
> > I think you may clean up qrk_serial_setup() as well.
> 
> That would break qrk_serial_setup_dma(), though.

Hmm... I don't see how.

The DMA setup is called after allocation of IRQ vector and the device in use is
the same, i.e. PCI device.

-- 
With Best Regards,
Andy Shevchenko