| To: | shemminger@xxxxxxxx |
|---|---|
| Subject: | Re: [PATCH] network hotplug via class_device/kobject |
| From: | "David S. Miller" <davem@xxxxxxxxxx> |
| Date: | Sun, 15 Jun 2003 00:50:55 -0700 (PDT) |
| Cc: | greg@xxxxxxxxx, netdev@xxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx |
| In-reply-to: | <20030613164119.15209934.shemminger@xxxxxxxx> |
| References: | <20030613164119.15209934.shemminger@xxxxxxxx> |
| Sender: | netdev-bounce@xxxxxxxxxxx |
From: Stephen Hemminger <shemminger@xxxxxxxx>
Date: Fri, 13 Jun 2003 16:41:19 -0700
This patch changes network devices to run hotplug out of the
kobject/class_device infrastructure rather than calling it from the
network core. The code gets simpler and there is only one place for
Greg to fix when he changes the API ;-)
I'll apply this patch, looks fine.
Paranoid about some driver doing something like:
rtnl_lock(); register_netdevice(); unregister_netdevice();
rtnl_unlock() BOOM
These sorts of turds exist at least in two places:
1) drivers/net/wan/comx.c
2) drivers/net/wan/hdlc_fr.c
But it is pretty clear that these two drivers have been
tried by nobody in recent years. They both call into
{un,}register_netdevice without the RTNL semaphore held.
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: [PATCH] convert slip driver to alloc_netdev, David S. Miller |
|---|---|
| Next by Date: | Re: [PATCH] Make xfrm subsystem optional, Andi Kleen |
| Previous by Thread: | [PATCH] network hotplug via class_device/kobject, Stephen Hemminger |
| Next by Thread: | Re: [PATCH] network hotplug via class_device/kobject, Francois Romieu |
| Indexes: | [Date] [Thread] [Top] [All Lists] |