In article <[email protected]>,
Masanobu SAITOH  <[email protected]> wrote:
>Hi, Christos.
>
>On 2015/08/07 18:53, Christos Zoulas wrote:
>> In article <[email protected]>,
>> Masanobu SAITOH  <[email protected]> wrote:
>>> Hi, all.
>>>
>>> Currently, pci_msi_count() and pci_msix_count() take one
>>> pci_attach_args argument.
>>> These functions may be used in other than attach function. So, it might be
>>> better to use pci_chipset_tag_t and pcitag_t.
>>>
>>> Is the following diff better than current specification?
>> 
>> Ok, but this makes them different than their alloc counterparts. Should
>> we change those too?
>
>Almost all pci related functions don't take pci_attach_args as
>an argument. Some functions are required to take it because
>some elements(e.g. pa_iot and pa_memt) in the structure are required.
>
> Before introducing MSI/MSI-X API, the following functions take
>pci_attach_args as an argument.
>
>pci_find_device()
>pci_mapreg_map()
>pci_mapreg_submap()
>pci_intr_map()
>pci_aprint_devinfo()
>pci_attach_hook()
>
>Some above functions refers pa_iot and pa_memt in them.
>
>And then, the following functions which take pci_attach_args
>as an argument are added:
>
>pci_intx_alloc()
>pci_intr_alloc()
>pci_msi_count()
>pci_msi_alloc()
>pci_msi_alloc_exact()
>pci_msix_count()
>pci_msix_alloc()
>pci_msix_alloc_exact()
>pci_msix_alloc_map()
>
> pci_intr_map() takes pci_attach_args, so it's consistent.
>In reality, it's not required for x86 to take pci_attach_args.
>The MSI/MSI-X related "alloc" function might not required
>to take pci_attach_args, but I think it's ok to keep current
>API for the consistency and possibility of using pci_attach_args
>on other archs.
>
> So, is it ok to change pci_msi[x]_count() only?
>

I don't have a problem with it, I was just questioning the rationale
about passing pci_attach_args to functions...

christos

Reply via email to