Re: [PATCH v3 02/14] PCI: prevent duplicate slot names

[Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>]

Alex Chiang wrote:
> Hi Kenji-san,
> 
> * Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>:
>> Alex Chiang wrote:
>>> * Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>:
>>>> Kenji Kaneshige wrote:
>>>>> Hi Alex-san,
>>>>>
>>>>> Here is one comment, though I have not finished reviewing/testing
>>>>> your patches yet (sorry for the delay).
>>>>>
>>>>> Alex Chiang wrote:
>>>>>
>>>>> (snip.)
>>>>>
>>>>>> diff --git a/drivers/pci/hotplug/pci_hotplug_core.c   
>>>>>> b/drivers/pci/hotplug/pci_hotplug_core.c
>>>>>> index 3e37d63..46802dc 100644
>>>>>> --- a/drivers/pci/hotplug/pci_hotplug_core.c
>>>>>> +++ b/drivers/pci/hotplug/pci_hotplug_core.c
>>>>>> @@ -570,39 +570,32 @@ int pci_hp_register(struct hotplug_slot 
>>>>>> *slot,  struct pci_bus *bus, int slot_nr,
>>>>>>          return -EINVAL;
>>>>>>      }
>>>>>>  -    /* Check if we have already registered a slot with the same 
>>>>>> name. */
>>>>>> -    if (get_slot_from_name(name))
>>>>>> -        return -EEXIST;
>>>>>> -
>>>>>>      /*
>>>>>> -     * No problems if we call this interface from both ACPI_PCI_SLOT
>>>>>> -     * driver and call it here again. If we've already created the
>>>>>> -     * pci_slot, the interface will simply bump the refcount.
>>>>>> +     * Look for existing slot. If we find it, and it was created by a
>>>>>> +     * slot detection driver (ie, doesn't have a ->hotplug()) then we
>>>>>> +     * allow the hotplug driver calling us to rename the slot if 
>>>>>>  desired.
>>>>>> +     *
>>>>>> +     * Otherwise, create the slot and carry on with life.
>>>>>>       */
>>>>>> -    pci_slot = pci_create_slot(bus, slot_nr, name);
>>>>>> -    if (IS_ERR(pci_slot))
>>>>>> -        return PTR_ERR(pci_slot);
>>>>>> -
>>>>>> -    if (pci_slot->hotplug) {
>>>>>> -        dbg("%s: already claimed\n", __func__);
>>>>>> -        pci_destroy_slot(pci_slot);
>>>>>> -        return -EBUSY;
>>>>>> +    pci_slot = pci_get_pci_slot(bus, slot_nr);
>>>>> The pci_get_pci_slot() function refers pci_bus->slots list, so it
>>>>> should be called with pci_bus_sem semaphore held as pci_create_slot()
>>>>> does, or pci_bus_sem semaphore should be held by pci_get_pci_slot()
>>>>> itself.
>>> Yes, I've changed pci_get_pci_slot() to acquire the pci_bus_sem
>>> semaphore.
>>>
>>> Thank you for pointing this out.
>>>
>>> It will be fixed in v4 of this patch series, which I will send
>>> out after I receive the rest of your review comments.
>>>
>> I noticed that changing pci_get_pci_slot() to acquire the pci_bus_sem
>> might be not enough. If slot was created between pci_get_pci_slot() and
>> pci_create_slot() by another thread in the following code, something
>> wrong would happen I think.
>>
>>        pci_slot = pci_get_pci_slot(bus, slot_nr);
>>        if (pci_slot) {
>>                if (pci_slot->hotplug) {
>>                        result = -EBUSY;
>>                        goto err;
>>                }
>>
>>                if (strcmp(kobject_name(&pci_slot->kobj), name))
>>                        if ((result = pci_rename_slot(pci_slot, name)))
>>                                goto err;
>>        } else {
>>                pci_slot = pci_create_slot(bus, slot_nr, name);
>>                if ((result = IS_ERR(pci_slot)))
>>                        goto out;
>>        }
> 
> I'm sorry, I don't think I see the problem you are pointing out.
> 
> If pci_get_pci_slot() finds a pci_slot, we do not modify
> pci_bus->slots any further, so even if a new slot is created, it
> shouldn't affect the pci_slot that we already found.
> 
> I must be missing something, but I don't know what. Would you
> mind explaining what you had in mind with your comment?
> 

Here is one of the example of the problem I imagine.

When pci_get_pci_slot() does NOT find a pci_slot, pci_hp_register()
calls pci_create_slot(). If another hotplug driver creates slot
between pci_get_pci_slot() and pci_create_slot(), hotplug member of
pci_slot returned from pci_create_slot() is not NULL in the following
code.

>>                pci_slot = pci_create_slot(bus, slot_nr, name);
>>                if ((result = IS_ERR(pci_slot)))
>>                        goto out;


In this case, we should return with -EBUSY. But because there is no
check to see if pci->slot is NULL, pci_slot->hotplug will be overwritten
with the new value.

Thanks,
Kenji Kaneshige