On Mon, 2025-08-25 at 10:12 -0700, Farhan Ali wrote:
> On s390 systems, which use a machine level hypervisor, PCI devices are
> always accessed through a form of PCI pass-through which fundamentally
> operates on a per PCI function granularity. This is also reflected in the
> s390 PCI hotplug driver which creates hotplug slots for individual PCI
> functions. Its reset_slot() function, which is a wrapper for
> zpci_hot_reset_device(), thus also resets individual functions.
>
> Currently, the kernel's PCI_SLOT() macro assigns the same pci_slot object
> to multifunction devices. This approach worked fine on s390 systems that
> only exposed virtual functions as individual PCI domains to the operating
> system. Since commit 44510d6fa0c0 ("s390/pci: Handling multifunctions")
> s390 supports exposing the topology of multifunction PCI devices by
> grouping them in a shared PCI domain. When attempting to reset a function
> through the hotplug driver, the shared slot assignment causes the wrong
> function to be reset instead of the intended one. It also leaks memory as
> we do create a pci_slot object for the function, but don't correctly free
> it in pci_slot_release().
>
> This patch adds a helper function to allow per function PCI slots for
> functions managed through a hypervisor which exposes individual PCI
> functions while retaining the topology.
>
> Fixes: 44510d6fa0c0 ("s390/pci: Handling multifunctions")
> Co-developed-by: Niklas Schnelle <schnelle@xxxxxxxxxxxxx>
> Signed-off-by: Niklas Schnelle <schnelle@xxxxxxxxxxxxx>
> Signed-off-by: Farhan Ali <alifm@xxxxxxxxxxxxx>
> ---
> drivers/pci/slot.c | 19 ++++++++++++++++---
> 1 file changed, 16 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/pci/slot.c b/drivers/pci/slot.c
> index 50fb3eb595fe..991526af0ffe 100644
> --- a/drivers/pci/slot.c
> +++ b/drivers/pci/slot.c
> @@ -5,6 +5,7 @@
> * Alex Chiang <achiang@xxxxxx>
> */
>
> +#include <linux/hypervisor.h>
> #include <linux/kobject.h>
> #include <linux/slab.h>
> #include <linux/pci.h>
> @@ -73,7 +74,7 @@ static void pci_slot_release(struct kobject *kobj)
>
> down_read(&pci_bus_sem);
> list_for_each_entry(dev, &slot->bus->devices, bus_list)
> - if (PCI_SLOT(dev->devfn) == slot->number)
> + if (dev->slot == slot->number)
> dev->slot = NULL;
> up_read(&pci_bus_sem);
>
> @@ -160,13 +161,25 @@ static int rename_slot(struct pci_slot *slot, const char *name)
> return result;
> }
>
> +static bool pci_dev_matches_slot(struct pci_dev *dev, struct pci_slot *slot)
> +{
> + if (hypervisor_isolated_pci_functions()) {
> + if (dev->devfn == slot->number)
> + return true;
> + } else {
> + if (PCI_SLOT(dev->devfn) == slot->number)
> + return true;
> + }
> + return false;
> +}
> +
> void pci_dev_assign_slot(struct pci_dev *dev)
> {
> struct pci_slot *slot;
>
> mutex_lock(&pci_slot_mutex);
> list_for_each_entry(slot, &dev->bus->slots, list)
> - if (PCI_SLOT(dev->devfn) == slot->number)
> + if (pci_dev_matches_slot(dev, slot))
> dev->slot = slot;
> mutex_unlock(&pci_slot_mutex);
> }
Doing some more digging, I believe this also needs adjustment in
pci_dev_reset_slot_function(). Since commit 10791141a6cf ("PCI:
Simplify pci_dev_reset_slot_function()") that no longer directly looks
at the struct pci_slot linking but instead assumes that slot resets
don't work on multifunction devices. With per PCI function slots the
slot reset should work with pdev->multifunction set. I think adjusting
pci_dev_reset_slot_function() may be easier if instead of using the
hypervisor_isolated_pci_functions() helper we would set up a struct
pci_slot::per_func flag as we had considered as an option.
Thanks,
Niklas
