netdev
[Top] [All Lists]

Re: NETDEV_CHANGE events when __LINK_STATE_NOCARRIER is modified

To: jamal <hadi@xxxxxxxxxx>
Subject: Re: NETDEV_CHANGE events when __LINK_STATE_NOCARRIER is modified
From: Andrew Morton <andrewm@xxxxxxxxxx>
Date: Sat, 05 May 2001 16:03:41 +1000
Cc: Andi Kleen <ak@xxxxxx>, "David S. Miller" <davem@xxxxxxxxxx>, Janice Girouard <girouard@xxxxxxxxxx>, netdev@xxxxxxxxxxx
References: <20010504201702.A1011@xxxxxxxxxx> <Pine.GSO.4.30.0105042116190.16541-100000@xxxxxxxxxxxxxxxx>
Sender: owner-netdev@xxxxxxxxxxx
jamal wrote:
> 
> Come to think of it: How does cardbus/pcmcia netcard insertion work? That
> sends a netlink message and if i am not mistaken in interupt context.

schedule_task().  The now-standard way of punting asynchronous
events up into process context.

Yeah, we need to sort out the netif_carrier stuff.  Some
userspace apps want async notifications when the ethernet
is unplugged - high availability failover and desktop GUIs
come to mind.   At present, if the driver happens to
implement it you still need to poll the kernel, and the
result you get is munged together with the result of
netif_running().

It'd be better if netif_carrier_on/off were to send up
a select()able message of some form.  rtnetlink would be fine.

<Prev in Thread] Current Thread [Next in Thread>