[U-Boot] [PATCH 2/8] dm: pci: Only allow serial device to be bound before relocation

[Bin Meng]

Hi Simon,

On Tue, Aug 18, 2015 at 10:11 AM, Simon Glass <sjg at chromium.org> wrote:
> Hi Bin,
>
> On 17 August 2015 at 20:02, Bin Meng <bmeng.cn at gmail.com> wrote:
>> Hi Simon,
>>
>> On Tue, Aug 18, 2015 at 10:00 AM, Simon Glass <sjg at chromium.org> wrote:
>>> Hi Bin,
>>>
>>> On 15 August 2015 at 01:07, Bin Meng <bmeng.cn at gmail.com> wrote:
>>>> To further limit the memory space, we only allow serial device
>>>> to be bound before relocation.
>>>>
>>>> Signed-off-by: Bin Meng <bmeng.cn at gmail.com>
>>>> ---
>>>>
>>>>  drivers/pci/pci-uclass.c | 6 +++++-
>>>>  1 file changed, 5 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/pci/pci-uclass.c b/drivers/pci/pci-uclass.c
>>>> index 4160274..b7dca0f 100644
>>>> --- a/drivers/pci/pci-uclass.c
>>>> +++ b/drivers/pci/pci-uclass.c
>>>> @@ -487,9 +487,13 @@ static int pci_find_and_bind_driver(struct udevice *parent,
>>>>                          * whose driver has the DM_FLAG_PRE_RELOC set, to save
>>>>                          * precious memory space as on some platforms as that
>>>>                          * space is pretty limited (ie: using Cache As RAM).
>>>> +                        *
>>>> +                        * To further limit the memory space, we only allow
>>>> +                        * serial device to be bound.
>>>>                          */
>>>>                         if (!(gd->flags & GD_FLG_RELOC) &&
>>>> -                           !(drv->flags & DM_FLAG_PRE_RELOC))
>>>> +                           !(drv->flags & DM_FLAG_PRE_RELOC) &&
>>>> +                           (drv->id != UCLASS_SERIAL))
>>>>                                 return 0;
>>>>
>>>>                         /*
>>>> --
>>>> 1.8.2.1
>>>>
>>>
>>> This seems to go too far. We may have other drivers that need to do
>>> pre-reloc init. Why not just rely on DM_FLAG_PRE_RELOC?
>>>
>>
>> I was thinking the only usable pci driver before relocation is serial?
>
> I suspect there will be others. We have pre-relocation CPU init that
> might go in a driver one day. In any case this seems a bit hacky so I
> think we should drop this patch unless we need it.

But that is not on PCI bus, right? I can only think of a PCI serial
device so far. But as you said, we can drop this patch and revisit
this in the future.

Regards,
Bin