[dpdk-dev] [PATCH 4/4] igb_uio: bind error if pcie bridge
Scott Branden
scott.branden at broadcom.com
Mon Mar 26 21:13:38 CEST 2018
On 18-03-26 11:20 AM, Ferruh Yigit wrote:
> On 3/26/2018 7:05 PM, Scott Branden wrote:
>> Hi Ferruh,
>>
>>
>> On 18-03-26 10:24 AM, Ferruh Yigit wrote:
>>> On 3/21/2018 6:06 PM, Ajit Khaparde wrote:
>>>> From: Darren Edamura <darren.edamura at broadcom.com>
>>>>
>>>> Probe function should exit immediately if pcie bridge detected
>>>>
>>>> Signed-off-by: Darren Edamura <darren.edamura at broadcom.com>
>>>> Signed-off-by: Rahul Gupta <rahul.gupta at broadcom.com>
>>>> Signed-off-by: Scott Branden <scott.branden at broadcom.com>
>>>> Signed-off-by: Ajit Khaparde <ajit.khaparde at broadcom.com>
>>>> ---
>>>> lib/librte_eal/linuxapp/igb_uio/igb_uio.c | 3 +++
>>>> 1 file changed, 3 insertions(+)
>>>>
>>>> diff --git a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>>>> index 4cae4dd27..3fabbfc4d 100644
>>>> --- a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>>>> +++ b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
>>>> @@ -473,6 +473,9 @@ igbuio_pci_probe(struct pci_dev *dev, const struct pci_device_id *id)
>>>> void *map_addr;
>>>> int err;
>>>>
>>>> + if (pci_is_bridge(dev))
>>>> + return -ENODEV;
>>> What do you think printing a log here?
>> I think it brings little value. ENODEV is already returned?
> User should not provide bridge address at first place, I guess this is a
> protection in case user provides bridge address by mistake.
> In that case no device will be probed and user won't have any idea why.
> I think a log in dmesg saying bridge device is provided may help to the user.
I'll add a dev_warn as we actually encounter this issue on old silicon
revisions due to bridge address equaling a PF.
It's not a user error in such case and just needs to be ignored. So
adding this generic check allows such to occur.
For other use cases like you mentioned it would be a user mistake.
>>>> +
>>>> udev = kzalloc(sizeof(struct rte_uio_pci_dev), GFP_KERNEL);
>>>> if (!udev)
>>>> return -ENOMEM;
>>>>
>> Regards,
>> Scott
>>
More information about the dev
mailing list