Re: [PATCH 2/3] pci: bcm43xx avoid pci_find_device

[Jeff Garzik <jgarzik@pobox.com>]

Jiri Slaby wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Jeff Garzik napsal(a):
>> Jiri Slaby wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> Jeff Garzik napsal(a):
>>>> Jiri Slaby wrote:
>>>>> bcm43xx avoid pci_find_device
>>>>>
>>>>> Change pci_find_device to safer pci_get_device with support for more
>>>>> devices.
>>>>>
>>>>> Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
>>>>>
>>>>> ---
>>>>> commit 1d3b6caf027fe53351c645523587aeac40bc3e47
>>>>> tree ae37c86b633442cdf8a7a19ac287542724081c90
>>>>> parent ab3443d79c94d0ae6a9e020daefa4d29eccff50d
>>>>> author Jiri Slaby <ku@bellona.localdomain> Fri, 26 May 2006 01:49:12
>>>>> +0159
>>>>> committer Jiri Slaby <ku@bellona.localdomain> Fri, 26 May 2006
>>>>> 01:49:12 +0159
>>>>>
>>>>>  drivers/net/wireless/bcm43xx/bcm43xx_main.c |   20
>>>>> ++++++++++++++++----
>>>>>  1 files changed, 16 insertions(+), 4 deletions(-)
>>>>>
>>>>> diff --git a/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>>>>> b/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>>>>> index b488f77..56d2fc6 100644
>>>>> --- a/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>>>>> +++ b/drivers/net/wireless/bcm43xx/bcm43xx_main.c
>>>>> @@ -2131,6 +2131,13 @@ out:
>>>>>      return err;
>>>>>  }
>>>>>  
>>>>> +#ifdef CONFIG_BCM947XX
>>>>> +static struct pci_device_id bcm43xx_ids[] = {
>>>>> +    { PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, 0x4324) },
>>>>> +    { 0 }
>>>>> +};
> Table is here ^^^. You just add an entry, and that's it.
>>>>> +#endif
>>>>> +
>>>>>  static int bcm43xx_initialize_irq(struct bcm43xx_private *bcm)
>>>>>  {
>>>>>      int res;
>>>>> @@ -2141,10 +2148,15 @@ static int bcm43xx_initialize_irq(struct
>>>>>  #ifdef CONFIG_BCM947XX
>>>>>      if (bcm->pci_dev->bus->number == 0) {
>>>>>          struct pci_dev *d = NULL;
>>>>> -        /* FIXME: we will probably need more device IDs here... */
>>>>> -        d = pci_find_device(PCI_VENDOR_ID_BROADCOM, 0x4324, NULL);
>>>>> -        if (d != NULL) {
>>>>> -            bcm->irq = d->irq;
>>>>> +        struct pci_device_id *id = bcm43xx_ids;
>>>>> +        while (id->vendor) {
>>>>> +            d = pci_get_device(id->vendor, id->device, NULL);
>>>>> +            if (d != NULL) {
>>>>> +                bcm->irq = d->irq;
>>>>> +                pci_dev_put(d);
>>>>> +                break;
>>>> You'll want to use pci_match_device() or pci_match_one_device()
>>>> [I forget which one]
>>> Why? Matching is done by pci_get_device() or pci_get_subsys(),
>>> respectively.
>>> [pci_match_device() is for matching dev <-> drv, you meant
>>> pci_match_one_device()]
>> The FIXME says "we will probably need more device IDs here."
> Yup.
>> Thus, if you are touching this area, it would make sense to add the
>> capability to easily add a second (and third, fourth...) PCI ID.  And
>> that means pci_match_one_device() and a pci_device_id table.
> But the while loop do the work: unless id->vendor != NULL, do the matching with
> the current raw (id) of the table, then jump to the next raw (id++).
> 
> pci_get_device returns NULL if the device with id->vendor, id->device wasn't
> found, then we try next raw, otherwise, we break the loop.
> 
> Implementations before and now do the same strangeness -- assume there is only
> one device (?shouldn't matter?, since it is embedded).

The point is that you don't need to loop over the table, 
pci_match_one_device() does that for you.

And this code, like the gt96100_eth code, is testing the existence of 
certain platform devices, to be certain that it can proceed with certain 
platform-specific duties.

Thus we don't care about matching multiple devices -- an unlikely case 
-- but we do care about making the code as small as possible by calling 
a standard PCI match function which searches through a list of PCI IDs.

	Jeff