[U-Boot] [PATCH 29/29] RFC: x86: Move link to use driver model for SCSI

[Simon Glass]

Hi Bin,

On 12 June 2017 at 21:15, Bin Meng <bmeng.cn at gmail.com> wrote:
> Hi Simon,
>
> On Tue, Jun 6, 2017 at 3:15 AM, Simon Glass <sjg at chromium.org> wrote:
>> As a demonstration of how to use SCSI with driver model, move link over
>> to use this. This patch needs more work, but illustrates the concept.
>>
>> Signed-off-by: Simon Glass <sjg at chromium.org>
>> ---
>>
>>  arch/x86/cpu/ivybridge/sata.c     | 38 +++++++++++++++++++++++++++++++++++++-
>>  configs/chromebook_link_defconfig |  2 ++
>>  2 files changed, 39 insertions(+), 1 deletion(-)
>>
>
> This is great! Thanks a lot.
>
>> diff --git a/arch/x86/cpu/ivybridge/sata.c b/arch/x86/cpu/ivybridge/sata.c
>> index 0f5e190425..5bbe65d442 100644
>> --- a/arch/x86/cpu/ivybridge/sata.c
>> +++ b/arch/x86/cpu/ivybridge/sata.c
>> @@ -6,6 +6,7 @@
>>   */
>>
>>  #include <common.h>
>> +#include <ahci.h>
>>  #include <dm.h>
>>  #include <fdtdec.h>
>>  #include <asm/io.h>
>> @@ -208,6 +209,20 @@ static void bd82x6x_sata_enable(struct udevice *dev)
>>         dm_pci_write_config16(dev, 0x90, map);
>>  }
>>
>> +static int bd82x6x_sata_bind(struct udevice *dev)
>> +{
>> +       struct udevice *scsi_dev;
>> +       int ret;
>> +
>> +       if (gd->flags & GD_FLG_RELOC) {
>> +               ret = ahci_bind_scsi(dev, &scsi_dev);
>> +               if (ret)
>> +                       return ret;
>> +       }
>> +
>> +       return 0;
>> +}
>> +
>>  static int bd82x6x_sata_probe(struct udevice *dev)
>>  {
>>         struct udevice *pch;
>> @@ -219,8 +234,12 @@ static int bd82x6x_sata_probe(struct udevice *dev)
>>
>>         if (!(gd->flags & GD_FLG_RELOC))
>>                 bd82x6x_sata_enable(dev);
>> -       else
>> +       else {
>>                 bd82x6x_sata_init(dev, pch);
>> +               ret = ahci_probe_scsi(dev);
>> +               if (ret)
>> +                       return ret;
>> +       }
>>
>>         return 0;
>>  }
>> @@ -234,5 +253,22 @@ U_BOOT_DRIVER(ahci_ivybridge_drv) = {
>>         .name           = "ahci_ivybridge",
>>         .id             = UCLASS_AHCI,
>>         .of_match       = bd82x6x_ahci_ids,
>> +       .bind           = bd82x6x_sata_bind,
>>         .probe          = bd82x6x_sata_probe,
>>  };
>> +
>> +static struct pci_device_id chromebook_ssd_supported[] = {
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_NM10_AHCI) },
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL,
>> +                    PCI_DEVICE_ID_INTEL_COUGARPOINT_AHCI_MOBILE) },
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL,
>> +                    PCI_DEVICE_ID_INTEL_COUGARPOINT_AHCI_SERIES6) },
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL,
>> +                    PCI_DEVICE_ID_INTEL_PANTHERPOINT_AHCI_MOBILE) },
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_LYNXPOINT_AHCI) },
>> +       { PCI_DEVICE(PCI_VENDOR_ID_INTEL,
>> +                    PCI_DEVICE_ID_INTEL_WILDCATPOINT_AHCI) },
>> +       {},
>> +};
>
> Since  AHCI is pretty generic, instead of providing device list
> filter, can we just use PCI class and sub-class for AHCI, like USB?
>
> For Ivybridge specific stuff, we can call the low level initialization
> function as part of the SoC/platform initialization codes.

Actually this code should be dropped. I was experimenting with how to
enable it generally (as you suggest). But we really need to do what
you say but in a separate file.

The question I have is how to enable a sub-class to use this driver,
since some devices will want to use ahci.c (i.e. SCSI-based) and some
will have their own drivers. I'm not sure that we support finding more
specific PCI drivers first.

Anyway I will respin this series with this patch updated and we can
work it out from there.

Regards,
Simon