Re: [RFC][PATCH -mm take5 4/7] using symlink for the net_device

[Keiichi KII <k-keiichi@bx.jp.nec.com>]

Hello Satyam,

> In any case, however, the point to extend the critical section here
> to encapsulate all the three parts still stands. We wouldn't want
> ioctl(NETCON_REMOVE_TARGET) on the specified target to
> return without removing the target that the user specified just
> because that target's ethernet interface happens to be currently
> undergoing a name change. The correct behaviour would be to
> sleep on a mutex till the renaming has completed (which will
> then relinquish the mutex) and then (after acquiring the mutex)
> proceed to remove it, IMHO.

You're right. I misunderstood. All the three parts needs to encapsulate.

>> >> +static char *make_netdev_class_name(char *netdev_name)
>> >> +{
>> >> +       char *name;
>> >> +
>> >> +       name = kasprintf(GFP_KERNEL, "net:%s", netdev_name);
>> >
>> > Why the "net:" prefix in the filename?
>>
>> Because I drew upon dev_change_name() method in net/core/dev.c.
>> The device_rename() in the above function makes use of same prefix
>> related to netdev.
> 
> I think you're referring to make_class_name() here? That seems to
> be somewhat bulkier than simply being a wrapper over kasprintf()
> like the make_netdev_class_name() here. I'd definitely recommend
> not obfuscating this simple functionality here.

I understand. The wrapper method such as make_netdev_class_name() isn't
appropriate in this case.

Thanks
-- 
Keiichi KII
NEC Corporation OSS Platform Development Division
E-mail: k-keiichi@bx.jp.nec.com