Re: [PATCH 4/6] PCI: acpiphp: check for new devices on enabled host

[Rafael J. Wysocki]

On Friday, June 28, 2013 09:22:32 AM Yinghai Lu wrote:
> On Fri, Jun 28, 2013 at 2:33 AM, Kirill A. Shutemov
> <kirill.shutemov@xxxxxxxxxxxxxxx> wrote:
> > Yinghai Lu wrote:
> >> On Tue, Jun 25, 2013 at 9:22 AM, Mika Westerberg
> >> <mika.westerberg@xxxxxxxxxxxxxxx> wrote:
> >> > From: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
> >> >
> >> > Current acpiphp_check_bridge() implementation is pretty dumb:
> >> >  - it enables the slot if it's not enabled and the slot status is
> >> >    ACPI_STA_ALL;
> >> >  - it disables the slot if it's enabled and slot is not in ACPI_STA_ALL
> >> >    state.
> >> >
> >> > This behavior is not enough to handle Thunderbolt chaining case
> >> > properly. We need to actually rescan for new devices even if a device
> >> > has already in the slot.
> >> >
> >> > The new implementation disables and stops the slot if it's not in
> >> > ACPI_STA_ALL state.
> >> >
> >> > For ACPI_STA_ALL state we first trim devices which don't respond and
> >> > look for the ones after that. We do that even if slot already enabled
> >> > (SLOT_ENABLED).
> >>
> >> that is not right, some time BUS_CHECK is even sent root bus.
> >> in that case, stop all devices in slots and load driver again.
> >>
> >> like you put one card in one slots, but all devices in other slots get stop
> >> and enable again.
> >
> > We don't stop enabled devices, we only stop and remove devices which don't
> > respond.  See patch 3/6.
> >
> > I don't see how it's harmful. Do you?
> 
> then please check with disable_device to put back pci_dev ref,
> also may need to trim corresponding acpi devices.

That's correct.  Thunderbolt may not need that, but ACPI device objects need
to be trimmed too in general.

> so this patch is helping: multiple plug-in and remove?

I think it helps the case when a Thunderbolt device is removed and we only
get a bus check notify for the controller after the fact.

Thanks,
Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
--
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/