diff mbox

[RFC,V4,02/13] netback: add module unload function.

Message ID 1328256136.13189.34.camel@dagon.hellion.org.uk
State RFC, archived
Delegated to: David Miller
Headers show

Commit Message

Ian Campbell Feb. 3, 2012, 8:02 a.m. UTC 
On Fri, 2012-02-03 at 07:25 +0000, Eric Dumazet wrote:
> Le vendredi 03 février 2012 à 06:38 +0000, Ian Campbell a écrit :
> > On Thu, 2012-02-02 at 22:52 +0000, Paul Gortmaker wrote:
> > > On Thu, Feb 2, 2012 at 3:50 PM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
> > > > On Thu, 2012-02-02 at 20:34 +0000, Eric Dumazet wrote:
> > > 
> > > [...]
> > > 
> > > >
> > > > I don't think it is at all unreasonable to ask for bug fixes but in this
> > > > case Wei's series is removing the code in question (which would also
> > > > undoubtedly fix the bug).
> > > >
> > > > As it happens the fix turns out to be simple but if it were complex I
> > > > would perhaps have disagreed more strongly about spending effort fixing
> > > > code that is removed 2 patches later, although obviously that would have
> > > > depended on the specifics of the fix in that case.
> > > 
> > > Lots of people are relying on git bisect.  If you introduce build failures
> > > or known bugs into any point in history, you take away from the value
> > > in git bisect.  Sure, it happens by accident, but it shouldn't ever be
> > > done knowingly.
> > 
> > Sure. In this case the bug has been there since 2.6.39, it isn't
> > introduced by this series.
> > 
> 
> We are stuck right now with a bug introduced in 2.6.39, (IP redirects),
> and because fix was done in 3.1, we are unable to provide a fix fo
> stable 3.0 kernel.
> 
> Something that takes 15 minutes to fix now, can take several days of
> work later.

Sure.

Here is the patch. I've compile tested it but not run it yet since I'm
supposed to be packing for a trip, I'll be back on Wednesday. It seems
straight forward enough though.

8<--------------------------------

From 6f3d3068f6e049c2d810f9fc667d57667bea77dc Mon Sep 17 00:00:00 2001
From: Ian Campbell <ian.campbell@citrix.com>
Date: Fri, 3 Feb 2012 07:47:23 +0000
Subject: [PATCH] xen: netback: do not bind netback threads to specific CPUs

netback_init does not take proper account of which CPUs is online. However we
don't require a thread per CPU, just a pool of worker threads, of which the
number of CPUs at start of day is as good a number as any.

Therefore do not bind netback threads to particular CPUs.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Wei Lui <wei.lui2@citrix.com>
---
 drivers/net/xen-netback/netback.c |    2 --
 1 files changed, 0 insertions(+), 2 deletions(-)
diff mbox

Patch

diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c
index 59effac..31ad3ee 100644
--- a/drivers/net/xen-netback/netback.c
+++ b/drivers/net/xen-netback/netback.c
@@ -1670,8 +1670,6 @@  static int __init netback_init(void)
 			goto failed_init;
 		}
 
-		kthread_bind(netbk->task, group);
-
 		INIT_LIST_HEAD(&netbk->net_schedule_list);
 
 		spin_lock_init(&netbk->net_schedule_list_lock);