Re: [PATCH 3/4] myri10ge - Driver core

[Brice Goglin <brice@myri.com>]

Arnd Bergmann wrote:
> Am Friday 19 May 2006 01:56 schrieb Brice Goglin:
>   
>> This place is actually the only one where we don't want to use msleep.
>> This function (myri10ge_send_cmd) might be called from various context
>> (spinlocked or not) and pass orders to the NIC whose processing time
>> depends a lot on the command. Of course, we don't have any place where a
>> long operation is passed from a spinlocked context :) But, we need the
>> tiny udelay granularity for the spinlocked case, and the long loop for
>> operations that are long to process in the NIC.
>>     
>
> I don't see any spinlocks in your code and the function does not
> seem to be called from the interrupt handler or the softirq either.
> Maybe I'm missed something, but where is this ever called in an
> atomic context?
>   

dev_mc_upload() from net/core/dev_mcast.c does

spin_lock_bh(&dev->xmit_lock);
__dev_mc_upload(dev);

which calls dev->set_multicast_list(), which is
myri10ge_set_multicast_list()

which calls myri10ge_change_promisc

which calls myri10ge_send_cmd



> Whenever you have a device associated with the message, it makes
> sense to use the dev_printk family of functions.
>   

Ok, thanks.

> My understanding is that vendor IDs should go to the common file
> because they are likely to be used by multiple drivers whereas
> device IDs only need to be present in the one device driver for
> that particular device.
>   

Make sense. I will change it.

Thanks again,
Brice