Re: [PATCH 4/6] staging: comedi: addi_apci_1564: simplify the PCI bar reading and don't read the unused bars

[Ian Abbott]

On 2014-04-28 11:19, Ian Abbott wrote:
> On 2014-04-27 02:37, Chase Southwood wrote:
> > This driver only uses PCI bar 1 (devpriv->i_IobaseAmcc), and PCI bar 2
> > (dev->iobase) doon't bother reading the unused PCI bars.
> >
> > Signed-off-by: Chase Southwood <chase.southwood@xxxxxxxxx>
> > Cc: Ian Abbott <abbotti@xxxxxxxxx>
> > Cc: H Hartley Sweeten <hsweeten@xxxxxxxxxxxxxxxxxxx>
> > ---
> >
> > Hartley,
> >
> > As far as I can tell from reading the I/O Mapping you sent me, these bar
> > numbers are correct, but it seems a bit odd, so please offer correction
> > if I am interpreting the document incorrectly.
> >
> > Thanks,
> > Chase
> >
> >   drivers/staging/comedi/drivers/addi_apci_1564.c | 11 ++---------
> >   1 file changed, 2 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/staging/comedi/drivers/addi_apci_1564.c
> > b/drivers/staging/comedi/drivers/addi_apci_1564.c
> > index fe42f9d..7e42d47 100644
> > --- a/drivers/staging/comedi/drivers/addi_apci_1564.c
> > +++ b/drivers/staging/comedi/drivers/addi_apci_1564.c
> > @@ -65,15 +65,8 @@ static int apci1564_auto_attach(struct
> > comedi_device *dev,
> >       if (ret)
> >           return ret;
> >
> > -    if (this_board->i_IorangeBase1)
> > -        dev->iobase = pci_resource_start(pcidev, 1);
> > -    else
> > -        dev->iobase = pci_resource_start(pcidev, 0);
> > -
> > -    devpriv->iobase = dev->iobase;
> > -    devpriv->i_IobaseAmcc = pci_resource_start(pcidev, 0);
> > -    devpriv->i_IobaseAddon = pci_resource_start(pcidev, 2);
> > -    devpriv->i_IobaseReserved = pci_resource_start(pcidev, 3);
> > +    dev->iobase = pci_resource_start(pcidev, 2);
> > +    devpriv->i_IobaseAmcc = pci_resource_start(pcidev, 1);
>
> I'm sure the original resources are correct, so since
> this_board->i_IorangeBase1 is non-zero:
>
>      dev->iobase = pci_resource_start(pcidev, 1);
>      devpriv->i_IobaseAmcc = pci_resource_start(pcidev, 0);

Maybe this depends on the revision of the board, though.  The documents 
Chase saw were for APCI1564 revision D, but the PCI BAR assignments 
might have been different for earlier revisions of the board.

All the driver code I've ever seen for this board (including one of 
ADDI-DATA's own Linux drivers, which no longer seems to be available for 
download) has used PCI BARs 0 and 1 with the same functions as the 
comedi driver.

Most "off-the-shelf" PCI interface chips reserve at least PCI BAR 0 for 
their own registers.  The clearest image I've found of this board seems 
to have an Altera ACEX FPGA with built-in PCI interface, so it might not 
have that restriction on BAR 0.

http://pcqt.com/DAQ/images/apci-1564-400.jpg

I think it's best to stick with the existing BAR assignments until 
proved otherwise.

> I suspect the member name 'i_IobaseAmcc' is just a hangover from the
> distant past and could just be changed to 'iobasemain' or something,
> since most of the code seems to use it.


--
-=( Ian Abbott @ MEV Ltd.    E-mail: <abbotti@xxxxxxxxx>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/