Hi Philipp, Thanks for the review! :)
Em Seg, 2009-05-25 às 21:55 +0200, pHilipp Zabel escreveu:
> > +static int pcap_add_subdev(struct spi_device *spi, struct pcap_subdev
> > *subdev)
> > +{
> > + struct platform_device *pdev;
> > +
> > + pdev = platform_device_alloc(subdev->name, subdev->id);
> > + pdev->dev.parent = &spi->dev;
> > + pdev->dev.platform_data = subdev->platform_data;
> > +
> > + return platform_device_add(pdev);
> > +}
>
> This pcap_subdev, pcap_remove/add_subdev business looks a bit like
> you're duplicating the MFD core just because it doesn't handle
> resource-less mfd_cells. I wonder whether it'd be better to fix that
> and then use
> mfd_cell, mfd_add/remove_devices instead?
Due to the lack of comments on this driver, i started grepping LKML for
previous MFD reviews. Now the driver allows multiple instances and no
longer keeps pcap_chip as a global(as per Samuel comments on U300/AB3100
driver), result is that resource-less mfd_cells are no longer the only
issue so I can use the mfd_ API: Now I also need a way to pass a
pcap_chip pointer to the subdevs.
> > + set_irq_type(pdata->irq, IRQ_TYPE_EDGE_RISING);
> > + set_irq_chained_handler(pdata->irq, pcap_irq_handler);
> > + set_irq_wake(pdata->irq, 1);
>
> Looking at the other MFD drivers that register a chained handler,
> providing this IRQ via a resource seems to be the default.
Ok, will do.
I will send a new version of the patch tomorrow, after i finish
converting all the subdevices to the new code.
Thanks!
--
Daniel Ribeiro
