Re: [PATCH v5 02/12] libata: bind the Linux device tree to the ACPIdevice tree

From: Dan Williams
Date: Wed Jun 20 2012 - 23:59:25 EST


On Wed, Jun 20, 2012 at 6:26 PM, Lin Ming <ming.m.lin@xxxxxxxxx> wrote:
> On Wed, 2012-06-20 at 16:50 -0700, Dan Williams wrote:
>> On Sun, Jun 17, 2012 at 11:25 PM, Lin Ming <ming.m.lin@xxxxxxxxx> wrote:
>> > From: Matthew Garrett <mjg@xxxxxxxxxx>
>> >
>> > Associate the ACPI device tree and libata devices.
>> > This patch uses the generic ACPI glue framework to do so.
>> >
>> > Signed-off-by: Matthew Garrett <mjg@xxxxxxxxxx>
>> > Signed-off-by: Holger Macht <holger@xxxxxxxx>
>> > Signed-off-by: Lin Ming <ming.m.lin@xxxxxxxxx>
>> > ---
>> >
>> > Changelog:
>> >
>> > - fixed crash reported at https://lkml.org/lkml/2012/2/24/2 (Aaron Lu)
>> > - rename is_pci_ata to compat_pci_ata (Alan Cox)
>> >
>> >  drivers/acpi/glue.c       |    4 +-
>> >  drivers/ata/libata-acpi.c |  125 +++++++++++++++++++++++++++++++++++++++++++++
>> >  drivers/ata/libata-core.c |    3 ++
>> >  drivers/ata/libata.h      |    4 ++
>> >  4 files changed, 134 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/drivers/acpi/glue.c b/drivers/acpi/glue.c
>> > index 1564e09..18d6812 100644
>> > --- a/drivers/acpi/glue.c
>> > +++ b/drivers/acpi/glue.c
>> [..]
>> > +static int ata_acpi_bind_host(struct device *dev, int host, acpi_handle *handle)
>> > +{
>> > +       struct Scsi_Host *shost = dev_to_shost(dev);
>> > +       struct ata_port *ap = ata_shost_to_port(shost);
>> > +
>> > +       if (ap->flags & ATA_FLAG_ACPI_SATA)
>> > +               return -ENODEV;
>> > +
>> > +       *handle = acpi_get_child(DEVICE_ACPI_HANDLE(dev->parent), ap->port_no);
>> > +
>> > +       if (!*handle)
>> > +               return -ENODEV;
>> > +
>> > +       return 0;
>> > +}
>>
>> I think this patch should be squashed with patch 4, right?  That is if
>> we keep the hard-coded ->parent chasing...
>
> Right. Will merge patch 4.
>
>>
>> > +static int ata_acpi_find_device(struct device *dev, acpi_handle *handle)
>> > +{
>> > +       unsigned int host, channel, id, lun;
>> > +
>> > +       if (sscanf(dev_name(dev), "host%u", &host) == 1) {
>> > +               if (!compat_pci_ata(dev->parent))
>> > +                       return -ENODEV;
>> > +
>> > +               return ata_acpi_bind_host(dev, host, handle);
>> > +       } else if (sscanf(dev_name(dev), "%d:%d:%d:%d",
>> > +                       &host, &channel, &id, &lun) == 4) {
>> > +               if (!compat_pci_ata(dev->parent->parent->parent))
>> > +                       return -ENODEV;
>>
>> ...this looks like it should be using a dev_to_ata_port() helper which
>> like dev_to_shost() skips the need to remember just how many parents
>> until we get back to our ata_port, scsi_host, etc.
>
> Oh, no.
>
> compat_pci_ata checks whether the controller(rather than shost or ata
> port) is pci SATA/IDE controller.
>
> As in patch 4:
> compat_pci_ata(dev->parent->parent->parent->parent)
>
> The "dev" is ata port, and dev->parent->parent->parent->parent points to
> the controller dev.
>
> This looks ugly. How about add two helpers?
> shost_to_controller(dev) and ata_port_to_controller(dev)
>

What about something like the following (untested)? Makes it mostly
independent of the topology, gives some type safety, and drops some
redundant work finding the ata_port.

static int ata_acpi_find_device(struct device *dev, acpi_handle *handle)
{
struct ata_port *ap = dev_to_ata_port(dev);

if (!ap)
return -ENODEV;

if (!compat_pci_ata(ap))
return -ENODEV;

if (scsi_is_host_device(dev))
return ata_acpi_bind_host(ap, handle);
else if (scsi_is_sdev_device(dev)) {
struct scsi_device *sdev = to_scsi_device(dev);

return ata_acpi_bind_device(ap, sdev->channel,
sdev->id, handle);
} else
return -ENODEV;
}
--
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/