On Mon, Jan 30, 2012 at 12:52 PM, Yinghai Lu <yinghai@xxxxxxxxxx> wrote:
> On Mon, Jan 30, 2012 at 8:57 AM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
>> This adds a list of all PCI host bridges we find and a way to look up
>> the host bridge from a pci_dev.
>> ---
>> drivers/pci/probe.c | 39 ++++++++++++++++++++++++++++++++++-----
>> include/linux/pci.h | 5 +++++
>> 2 files changed, 39 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>> index da0d655..2ffe8a3 100644
>> --- a/drivers/pci/probe.c
>> +++ b/drivers/pci/probe.c
>> @@ -15,6 +15,8 @@
>> #define CARDBUS_LATENCY_TIMER 176 /* secondary latency timer */
>> #define CARDBUS_RESERVE_BUSNR 3
>>
>> +static LIST_HEAD(pci_host_bridges);
>> +
>> /* Ugh. Need to stop exporting this to modules. */
>> LIST_HEAD(pci_root_buses);
>> EXPORT_SYMBOL(pci_root_buses);
>> @@ -42,6 +44,23 @@ int no_pci_devices(void)
>> }
>> EXPORT_SYMBOL(no_pci_devices);
>>
>> +static struct pci_host_bridge *pci_host_bridge(struct pci_dev *dev)
>> +{
>> + struct pci_bus *bus;
>> + struct pci_host_bridge *bridge;
>> +
>> + bus = dev->bus;
>> + while (bus->parent)
>> + bus = bus->parent;
>> +
>> + list_for_each_entry(bridge, &pci_host_bridges, list) {
>> + if (bridge->bus == bus)
>> + return bridge;
>> + }
>> +
>> + return NULL;
>> +}
>> +
>> /*
>> * PCI Bus Class
>> */
>> @@ -1526,20 +1545,23 @@ struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
>> struct pci_ops *ops, void *sysdata, struct list_head *resources)
>> {
>> int error, i;
>> + struct pci_host_bridge *bridge;
>> struct pci_bus *b, *b2;
>> struct device *dev;
>> struct pci_bus_resource *bus_res, *n;
>> struct resource *res;
>>
>> + bridge = kzalloc(sizeof(*bridge), GFP_KERNEL);
>> + if (!bridge)
>> + return NULL;
>> +
>> b = pci_alloc_bus();
>> if (!b)
>> - return NULL;
>> + goto err_bus;
>>
>> dev = kzalloc(sizeof(*dev), GFP_KERNEL);
>> - if (!dev) {
>> - kfree(b);
>> - return NULL;
>> - }
>> + if (!dev)
>> + goto err_dev;
>>
>> b->sysdata = sysdata;
>> b->ops = ops;
>> @@ -1576,6 +1598,8 @@ struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
>>
>> b->number = b->secondary = bus;
>>
>> + bridge->bus = b;
>> +
>> /* Add initial resources to the bus */
>> list_for_each_entry_safe(bus_res, n, resources, list)
>> list_move_tail(&bus_res->list, &b->resources);
>> @@ -1591,6 +1615,7 @@ struct pci_bus *pci_create_root_bus(struct device *parent, int bus,
>> }
>>
>> down_write(&pci_bus_sem);
>> + list_add_tail(&bridge->list, &pci_host_bridges);
>> list_add_tail(&b->node, &pci_root_buses);
>> up_write(&pci_bus_sem);
>
> we already have list for root_bus. So there will be two lists.
Yeah, I'm not sure we need both either. If we do end up adding a
struct pci_host_bridge, then I think it makes sense to keep a list of
*those* (and possibly remove the root_bus list), since that's a
slightly higher-level abstraction than root_bus is.
> wonder if we can overload bus->self point to struct host_bridge.
pci_bus->self points to a struct pci_dev. A host bridge is not a
pci_dev. I don't think it's a good idea to simply reuse "self" to
point to something of a different type, but maybe you have something
different in mind.
> or just do not add host_bridge , and instead only add another field
> like struct list_head windows in pci_sysdata?
The pci_sysdata I see is an x86-specific thing, so I don't see how
that would help.
Bjorn
--
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
