Re: [PATCH 6/7] PCI: Make sure VF's driver get attached after PF's

[Yinghai Lu]

On Tue, May 14, 2013 at 1:58 AM, Yan Burman <yanb@xxxxxxxxxxxx> wrote:
>
>
>> -----Original Message-----
>> From: netdev-owner@xxxxxxxxxxxxxxx [mailto:netdev-owner@xxxxxxxxxxxxxxx]
>> On Behalf Of Yinghai Lu
>> Sent: Tuesday, May 14, 2013 05:31
>> To: Bjorn Helgaas
>> Cc: Gu Zheng; linux-pci@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Yinghai
>> Lu; netdev@xxxxxxxxxxxxxxx
>> Subject: [PATCH 6/7] PCI: Make sure VF's driver get attached after PF's
>>
>> Found kernel try to load mlx4 drivers for VFs before PF's is really loaded when
>> the drivers are built-in, and kernel command line include probe_vfs=63,
>> num_vfs=63.
>>
>> It turns that it also happen for hotadd path even drivers are compiled as
>> modules and if they loaded. Esp some VF share the same driver with PF.
>>
>> calling path:
>>       device driver probe
>>               ==> pci_enable_sriov
>>                       ==> virtfn_add
>>                               ==> pci_dev_add
>>                               ==> pci_bus_device_add
>> when pci_bus_device_add is called, the VF's driver will be attached.
>> and at that time PF's driver does not finish yet.
>>
>> Need to move out pci_bus_device_add from virtfn_add and call it later. Fix
>> the problem for two path, 1. hotadd path: use device_schedule_callback.
>> 2. for booting path, use initcall to call that for all VF's.

>> Index: linux-2.6/drivers/net/ethernet/mellanox/mlx4/main.c
>> =================================================================
>> ==
>> --- linux-2.6.orig/drivers/net/ethernet/mellanox/mlx4/main.c
>> +++ linux-2.6/drivers/net/ethernet/mellanox/mlx4/main.c
>> @@ -2308,6 +2308,8 @@ slave_start:
>>       priv->pci_dev_data = pci_dev_data;
>>       pci_set_drvdata(pdev, dev);
>>
>> +     pci_bus_add_device_vfs(pdev);
>> +
>
> This seems wrong, since pci_bus_add_device_vfs() is being called for VF's as well as for PF when SRIOV is not enabled.
> I can see that there are sanity checks in pci_bus_add_vf() that would prevent this code from doing damage,
> but why schedule a callback that will do nothing if we can prevent scheduling altogether?
> It would probably be better to do the following:
>
> +       if (dev->flags & MLX4_FLAG_SRIOV)
> +               pci_bus_add_device_vfs(pdev);
> +

Yes, we can add that check.

BTW, do you have any plan to make mlx4 support sriov_configure via sysfs?

Yinghai
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/