| Message ID | 5734293A.6040702@sysgo.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Thu, May 12, 2016 at 08:56:58AM +0200, David Engraf wrote: > A virtual PCI controller may not have a bridge device, thus when > __pci_bus_size_bridges is called, it will generate a NULL pointer > exception when accessing bus->self->class. The attached patch adds a > check to verify if a bridge has been set. Without a bridge device, > the function returns. > > Signed-off-by: David Engraf <david.engraf <at> sysgo.com> > > diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c > index 55641a3..ddb3381 100644 > --- a/drivers/pci/setup-bus.c > +++ b/drivers/pci/setup-bus.c > @@ -1245,6 +1245,10 @@ void __pci_bus_size_bridges(struct pci_bus *bus, struct list_head *realloc_head) > if (pci_is_root_bus(bus)) > return; > > + /* bridge device available? */ > + if (!bus->self) > + return; I'd like to see the code that relies on this, i.e., the virtual PCI controller driver. I don't know how we would handle resource assignment if there's no bridge. When we assign resources to a device, we normally look at the upstream bridge to see what resources are available. Here's an example from my laptop: ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-fe]) pci_bus 0000:00: root bus resource [mem 0xbf200000-0xdfffffff window] pci 0000:00:1c.3: PCI bridge to [bus 02] pci 0000:00:1c.3: bridge window [mem 0xd0500000-0xd05fffff] pci 0000:02:00.0: reg 0x10: [mem 0xd0500000-0xd0501fff 64bit] PCI0 is the host bridge, and it forwards transactions in the range [mem 0xbf200000-0xdfffffff] to PCI bus 00. The 00:1c.3 bridge forwards transactions in the range [mem 0xd0500000-0xd05fffff] from bus 00 to bus 02. So if we want to assign resources to the BARs of 02:00.0, we have to take them from the 00:1c.3 window. In your virtual PCI controller situation, it sounds like you could have a device like 02:00.0 that is not on a root bus and does not have an upstream bridge. How would we assign resources to it? > switch (bus->self->class >> 8) { > case PCI_CLASS_BRIDGE_CARDBUS: > /* don't size cardbuses yet. */ -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c index 55641a3..ddb3381 100644 --- a/drivers/pci/setup-bus.c +++ b/drivers/pci/setup-bus.c @@ -1245,6 +1245,10 @@ void __pci_bus_size_bridges(struct pci_bus *bus, struct list_head *realloc_head) if (pci_is_root_bus(bus)) return; + /* bridge device available? */ + if (!bus->self) + return; + switch (bus->self->class >> 8) { case PCI_CLASS_BRIDGE_CARDBUS: /* don't size cardbuses yet. */
A virtual PCI controller may not have a bridge device, thus when __pci_bus_size_bridges is called, it will generate a NULL pointer exception when accessing bus->self->class. The attached patch adds a check to verify if a bridge has been set. Without a bridge device, the function returns. Signed-off-by: David Engraf <david.engraf <at> sysgo.com>