2014-11-19 02:36:03

by Ji-Ze Hong (Peter Hong)

Subject: [PATCH] serial: fix io address assign method with Fintek PCI-to-UART Product

The original driver fixed the io address with 0xe000+idx*8, but real io address assigned from BIOS
is dynamically from read PCI configure space 0x24, 0x20, 0x1c.

The Fintek F81504/F81508/F81512 maybe malfunction without this patch and malfunction surely when
more the 1 PCI card.

Signed-off-by: Peter Hung <[email protected]>
---
drivers/tty/serial/8250/8250_pci.c | 63 ++++++++++++++++++++++++++++----------
1 file changed, 47 insertions(+), 16 deletions(-)

diff --git a/drivers/tty/serial/8250/8250_pci.c b/drivers/tty/serial/8250/8250_pci.c
index 0468e15..6dfa227 100644
--- a/drivers/tty/serial/8250/8250_pci.c
+++ b/drivers/tty/serial/8250/8250_pci.c
@@ -1551,28 +1551,51 @@ static int pci_fintek_setup(struct serial_private *priv,
{
struct pci_dev *pdev = priv->dev;
unsigned long base;
- unsigned long iobase;
+ unsigned long iobase = 0;
unsigned long ciobase = 0;
u8 config_base;
+ u32 bar_data[3];

/*
- * We are supposed to be able to read these from the PCI config space,
- * but the values there don't seem to match what we need to use, so
- * just use these hard-coded values for now, as they are correct.
+ * Find each UARTs offset in PCI configuraion space
*/
switch (idx) {
- case 0: iobase = 0xe000; config_base = 0x40; break;
- case 1: iobase = 0xe008; config_base = 0x48; break;
- case 2: iobase = 0xe010; config_base = 0x50; break;
- case 3: iobase = 0xe018; config_base = 0x58; break;
- case 4: iobase = 0xe020; config_base = 0x60; break;
- case 5: iobase = 0xe028; config_base = 0x68; break;
- case 6: iobase = 0xe030; config_base = 0x70; break;
- case 7: iobase = 0xe038; config_base = 0x78; break;
- case 8: iobase = 0xe040; config_base = 0x80; break;
- case 9: iobase = 0xe048; config_base = 0x88; break;
- case 10: iobase = 0xe050; config_base = 0x90; break;
- case 11: iobase = 0xe058; config_base = 0x98; break;
+ case 0:
+ config_base = 0x40;
+ break;
+ case 1:
+ config_base = 0x48;
+ break;
+ case 2:
+ config_base = 0x50;
+ break;
+ case 3:
+ config_base = 0x58;
+ break;
+ case 4:
+ config_base = 0x60;
+ break;
+ case 5:
+ config_base = 0x68;
+ break;
+ case 6:
+ config_base = 0x70;
+ break;
+ case 7:
+ config_base = 0x78;
+ break;
+ case 8:
+ config_base = 0x80;
+ break;
+ case 9:
+ config_base = 0x88;
+ break;
+ case 10:
+ config_base = 0x90;
+ break;
+ case 11:
+ config_base = 0x98;
+ break;
default:
/* Unknown number of ports, get out of here */
return -EINVAL;
@@ -1583,6 +1606,14 @@ static int pci_fintek_setup(struct serial_private *priv,
ciobase = (int)(base + (0x8 * idx));
}

+ /* Get the io address dispatch from the BIOS */
+ pci_read_config_dword(pdev, 0x24, &bar_data[0]);
+ pci_read_config_dword(pdev, 0x20, &bar_data[1]);
+ pci_read_config_dword(pdev, 0x1c, &bar_data[2]);
+
+ /* Calculate Real IO Port */
+ iobase = (bar_data[idx/4] & 0xffffffe0) + (idx % 4) * 8;
+
dev_dbg(&pdev->dev, "%s: idx=%d iobase=0x%lx ciobase=0x%lx config_base=0x%2x\n",
__func__, idx, iobase, ciobase, config_base);

--
1.9.1

2014-11-19 03:46:39

by Greg Kroah-Hartman

Subject: Re: [PATCH] serial: fix io address assign method with Fintek PCI-to-UART Product

On Wed, Nov 19, 2014 at 10:35:53AM +0800, Peter Hung wrote:
> The original driver fixed the io address with 0xe000+idx*8, but real io address assigned from BIOS
> is dynamically from read PCI configure space 0x24, 0x20, 0x1c.
>
> The Fintek F81504/F81508/F81512 maybe malfunction without this patch and malfunction surely when
> more the 1 PCI card.

Please wrap these lines at 72 columns please, like git wants you to.

>
> Signed-off-by: Peter Hung <[email protected]>
> ---
> drivers/tty/serial/8250/8250_pci.c | 63 ++++++++++++++++++++++++++++----------
> 1 file changed, 47 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/tty/serial/8250/8250_pci.c b/drivers/tty/serial/8250/8250_pci.c
> index 0468e15..6dfa227 100644
> --- a/drivers/tty/serial/8250/8250_pci.c
> +++ b/drivers/tty/serial/8250/8250_pci.c
> @@ -1551,28 +1551,51 @@ static int pci_fintek_setup(struct serial_private *priv,
> {
> struct pci_dev *pdev = priv->dev;
> unsigned long base;
> - unsigned long iobase;
> + unsigned long iobase = 0;

Why initialize this to 0? That shouldn't be needed as you set it later
in the function before you use it, right?

Care to try again?

thanks,

greg k-h