Re: [PATCH 1/8] Notifier chain update

[Alan Stern <stern@rowland.harvard.edu>]

On Wed, 18 Jan 2006, David S. Miller wrote:

> From: Alan Stern <stern@rowland.harvard.edu>
> Date: Wed, 18 Jan 2006 16:57:30 -0500 (EST)
> 
> > On Wed, 18 Jan 2006, Benjamin LaHaise wrote:
> > 
> > > A notifier callee should not be sleeping, if anything it should be putting 
> > > its work onto a workqueue and completing it when it gets scheduled if it 
> > > has to do something that blocks.
> > 
> > Sez who?  If it's not documented in the kernel source, I don't believe 
> > it.
> 
> Many notifiers even get run from software interrupt context,
> making sleeping illegal.
> 
> For example, IPV6 addresses can get added/removed from a device
> in response to packets, and these operations trigger the
> inet6addr_chain notifier in net/ipv6/addrconf.c
> 
> So sleeping in a notifier is indeed illegal.

Correction: sleeping in an atomic notifier (like inet6addr_chain) callout 
is illegal.

But there are plenty of notifier chains that are always invoked in process 
context and where the callout routines may indeed block.

Alan Stern

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