On Sat, Jun 8, 2013 at 1:01 AM, Yinghai Lu <yinghai@xxxxxxxxxx> wrote:
> On Fri, Jun 7, 2013 at 6:51 PM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
>>>
>>> should split pci_release_host_bridge_dev renaming and moving to another patch.
>>
>> I split the rename/move to another patch and added both to my
>> pci/jiang-bus-lock-v3 branch.
>
> good.
>
>>
>>> > @@ -1726,30 +1722,21 @@ struct pci_bus *pci_create_root_bus(struct ..
> ...
>>
>> I think this function would be better if we created the host bridge
>> first, then the bus. But your patch is an improvement, so we can look
>> at doing that later.
>
> Yes, that will more straightforward, create hostbidge before root bus,
> like we have pci bridge before subordinate bus under it.
>
> current logic is
>
> b = pci_alloc_bus();
> if (!b)
> return NULL;
>
> b->sysdata = sysdata;
> b->ops = ops;
> b->number = b->busn_res.start = bus;
> b2 = pci_find_bus(pci_domain_nr(b), bus);
> if (b2) {
> /* If we already got to this bus through a different
> bridge, ignore it */
> dev_dbg(&b2->dev, "bus already known\n");
> goto err_out;
> }
>
> bridge = pci_alloc_host_bridge(b);
>
> We need to find the duplicated hostbridge or racing.
>
> If we need call pci_alloc_host_bridge at first, we will need my
> for_each_host_bridge patchset, and use it to check if there is
> existing hostbridge.
>
>>
>>> looks like we don't need to split device_register to device_del and put_device.
>>>
>>> we can make root bus removal to use device_unregister too, in the patches,
>>
>> If we can use device_register()/device_unregister() directly instead of
>> splitting them into initialize/add/del/put, that would be awesome.
>>
>> Anyway, http://git.kernel.org/cgit/linux/kernel/git/helgaas/pci.git/log/?h=pci/jiang-bus-lock-v3
>> is the current state of what I've got. If you want to iterate on this some
>> more and improve things, that's great. But please start with what's on my
>> branch and post your updates as a complete v4 because I did tweak some
>> changelogs and code formatting, and I don't want to lose that work.
>
> ok, please consider drop
> commit d751727a1792262cb464d8d9608500facb763301
> PCI: Make host bridge/bus creating and destroying logic symmetric
> from pci/jiang-bus-lock-v3, and apply my three patches at
>
> https://patchwork.kernel.org/patch/2562431/
> [1/7] PCI: move back pci_proc_attach_devices calling
> https://patchwork.kernel.org/patch/2562461/
> [2/7] PCI: move resources and bus_list releasing to pci_release_dev
> https://patchwork.kernel.org/patch/2562451/
> [3/7] PCI: Detach driver in pci_stop_device
I dropped the "Make host bridge/bus creating and destroying logic
symmetric" patch for now. I don't think the subsequent patches depend
on that one, so I merged the rest to my -next branch for v3.11, so we
can make at least a tiny bit of progress in this cycle.
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
