Re: [PATCH 1/5, resend] PCI: adjust quirk handler section annotations

[Bjorn Helgaas]

On Wed, May 9, 2012 at 9:02 AM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>>> On 09.05.12 at 17:54, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
>> On Wed, May 9, 2012 at 1:43 AM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>> This is in preparation to adjust modpost to check section mismatches
>>> on most of the .pci_fixup* sections:
>>>
>>> .pci_fixup_final is only used from .init.text, and hence can (along
>>> with the respective handler functions) reside in .init.*.
>>
>> pci_apply_final_quirks() is currently __init, so this probably
>> wouldn't break anything.   But the fact that it's all __init means
>> that pci_fixup_final quirks are only run for devices present at boot,
>> and they don't apply to hot-added devices.  That seems like a bug to
>> me.
>>
>>> Several other .pci_fixup_* sections are needed only during boot and
>>> suspend/resume, and can therefore be moved into .init.* if
>>> !CONFIG_PM.
>>
>> My inclination is that all PCI fixups should work the same for
>> hot-added devices as for those present at boot, which would suggest
>> that we should always use __devinit, not __init.  If I'm missing
>> something, please educate me :)
>
> That's certainly possible - I simply based the patch on what is there
> currently.

Yep, your patch makes perfect sense considering the tree as it is
today.  But if you agree with my sense of what it *should* be, I think
we should leave things as they are, or work on moving everything
towards __devinit (I'd be thrilled if you wanted to work on that :)).
Otherwise we'd just be changing things to __init that we'd have to
change back later.

Bjorn
--
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/