On Tue, 3 Jul 2012 11:16:45 -0700
Yinghai Lu <yinghai@xxxxxxxxxx> wrote:
> On Tue, Jul 3, 2012 at 1:33 AM, Taku Izumi <izumi.taku@xxxxxxxxxxxxxx> wrote:
> >
> > Hi Yinghai,
> > Thank you for your comment.
> >
> > On Mon, 2 Jul 2012 13:51:59 -0700
> > Yinghai Lu <yinghai@xxxxxxxxxx> wrote:
> >
> >> On Sun, Jul 1, 2012 at 9:49 PM, Taku Izumi <izumi.taku@xxxxxxxxxxxxxx> wrote:
> >> >
> >> > Currently I/O ports and I/O memory resources are reserved for
> >> > hotplug bridges and we can tune their size by using hpiosize
> >> > and hpmemsize boot option.
> >> >
> >> > This patch extends feature so that we can reserve additional
> >> > bus number resources for hotplug bridges and tune it by
> >> > using "hpbussize" option.
> >> >
> >> > This patch also adds missing document for "hpiosize" and
> >> > "hpmemsize" option.
> >> >
> >> > Signed-off-by: Taku Izumi <izumi.taku@xxxxxxxxxxxxxx>
> >> > ---
> >> > Documentation/kernel-parameters.txt | 10 ++++++++++
> >> > drivers/pci/pci.c | 6 +++++-
> >> > drivers/pci/probe.c | 4 ++++
> >> > include/linux/pci.h | 1 +
> >> > 4 files changed, 20 insertions(+), 1 deletion(-)
> >> >
> >> > Index: linux/drivers/pci/pci.c
> >> > ===================================================================
> >> > --- linux.orig/drivers/pci/pci.c
> >> > +++ linux/drivers/pci/pci.c
> >> > Index: linux/drivers/pci/probe.c
> >> > ===================================================================
> >> > --- linux.orig/drivers/pci/probe.c
> >> > +++ linux/drivers/pci/probe.c
> >> > @@ -865,6 +865,10 @@ int __devinit pci_scan_bridge(struct pci
> >> > /*
> >> > * Set the subordinate bus number to its real value.
> >> > */
> >> > + if (dev->is_hotplug_bridge) {
> >> > + if (max + pci_hotplug_bus_size < child->busn_res.end)
> >> > + max += pci_hotplug_bus_size;
> >> > + }
> >> > pci_bus_update_busn_res_end(child, max);
> >> > pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
> ...
> > I don't think this feature conflicts with your feature.
> > Even if we can reallocate bus number automatically, I'd like to
> > avoid that situation preferably, so I'd like to keep bus number
> > resource for hotplug bridge in advance.
> >
> > Can't you agree with me?
> >
>
> that add may not be safe, before busn_alloc patchset, that
> busn_res.end should be 0xff.
> so you add extra bus range that could overlap other bridge.
>
> after busn_alloc patchset:
>
> /*
> * Set the subordinate bus number to its real value.
> */
> if (dev->is_hotplug_bridge && child->busn_res.end > max &&
> (max - child->busn_res.start) < HOTPLUG_BRIDGE_RESERVE_BUSNR)
> max = min_t(int, child->busn_res.start +
> HOTPLUG_BRIDGE_RESERVE_BUSNR,
> child->busn_res.end);
> shrink_size = (int)child->busn_res.end - max;
> pci_write_config_byte(dev, PCI_SUBORDINATE_BUS, max);
> pci_bus_update_busn_res_end(child, max);
>
> and it will try to shrink less to preserve more space for later usage.
> now
>
> #define HOTPLUG_BRIDGE_RESERVE_BUSNR 8
>
> so I assume that you want that value to be changeable via command line?
Yes. After busn_alloc patchset is merged, "hpbussize" option should work
as you said.
--
Taku Izumi <izumi.taku@xxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
