On Mon, 23 Jul 2007 19:06:42 +0200
Andi Kleen <[EMAIL PROTECTED]> wrote:
> On Mon, Jul 23, 2007 at 10:58:22AM +0100, Stephen Hemminger wrote:
> > On 21 Jul 2007 15:26:00 +0200
> > Andi Kleen <[EMAIL PROTECTED]> wrote:
> >
> > > David Miller <[EMAIL PROTECTED]> writes:
> > > >
> > > > Good candid
On Mon, Jul 23, 2007 at 10:58:22AM +0100, Stephen Hemminger wrote:
> On 21 Jul 2007 15:26:00 +0200
> Andi Kleen <[EMAIL PROTECTED]> wrote:
>
> > David Miller <[EMAIL PROTECTED]> writes:
> > >
> > > Good candidates for taking advantage of multi-napi are:
> > >
> > > 1) e1000
> > > 2) ucc_geth
> >
On 21 Jul 2007 15:26:00 +0200
Andi Kleen <[EMAIL PROTECTED]> wrote:
> David Miller <[EMAIL PROTECTED]> writes:
> >
> > Good candidates for taking advantage of multi-napi are:
> >
> > 1) e1000
> > 2) ucc_geth
> > 3) ehea
> > 4) sunvnet
>
> s2io.c
sky2.c could use it because of issues with dual-
From: Rusty Russell <[EMAIL PROTECTED]>
Date: Sun, 22 Jul 2007 17:51:52 +1000
> On Sun, 2007-07-22 at 00:18 -0700, David Miller wrote:
> > --- a/net/core/dev.c
> > +++ b/net/core/dev.c
> > @@ -220,7 +220,8 @@ static RAW_NOTIFIER_HEAD(netdev_chain);
> > * Device drivers call our routines to queue
On Sun, 2007-07-22 at 00:18 -0700, David Miller wrote:
> Rusty, how does it look otherwise?
I like it. For a start, the simplification of the NAPI api was long
overdue, and the damage done by separating the napi_struct is really
minimal. Overall the tg3 driver just looks a little nicer now, and
From: David Miller <[EMAIL PROTECTED]>
Date: Sun, 22 Jul 2007 00:18:18 -0700 (PDT)
> To facilitate discussion I converted the core and tg3, and include
> just that part of the patch below.
And for those for whom it makes it easier to play around with these
changes, here is the updated e1000 conve
From: David Miller <[EMAIL PROTECTED]>
Date: Sat, 21 Jul 2007 20:54:25 -0700 (PDT)
> Ok, I'll put the napi_struct info the driver private and respin
> the patch.
To facilitate discussion I converted the core and tg3, and include
just that part of the patch below.
Consequences:
1) rtnetlink and
From: Rusty Russell <[EMAIL PROTECTED]>
Date: Sun, 22 Jul 2007 12:39:32 +1000
> But if netdev -> driver_private can be optimized, so can driver_private
> -> netdev. We just don't have a wrapper for it. netdev_of() perhaps?
>
> (If it can't be implemented by ptr arith, we have to do some allocat
On Sat, 2007-07-21 at 18:59 -0700, David Miller wrote:
> But note that you'd be adding a pointer deref, the current sequence:
>
> napi_struct --> netdev --> driver_private
>
> involves all pointer arithmetic and no derefs, whereas:
>
> napi_struct --> driver_private --> netdev
>
> will involve
From: Rusty Russell <[EMAIL PROTECTED]>
Date: Sun, 22 Jul 2007 09:10:23 +1000
> In five years' time, the "napi_struct for simple drivers" in net_device
> will look confusing. Since your change touches all NAPI drivers anyway,
> it'd be nice to go straight to "everyone allocates their own NAPI
> s
On Sat, 2007-07-21 at 11:54 -0700, David Miller wrote:
> From: Rusty Russell <[EMAIL PROTECTED]>
> Date: Sat, 21 Jul 2007 18:00:11 +1000
>
> > No, I was just thinking that drivers will put the napi_struct in their
> > driver-specific struct (eg. struct e1000_adapter *adapter =
> > container_of(con
From: Rusty Russell <[EMAIL PROTECTED]>
Date: Sat, 21 Jul 2007 18:00:11 +1000
> No, I was just thinking that drivers will put the napi_struct in their
> driver-specific struct (eg. struct e1000_adapter *adapter =
> container_of(container_of(napi, struct e1000_adapter, napi);).
That works.
> Mu
David Miller <[EMAIL PROTECTED]> writes:
>
> Good candidates for taking advantage of multi-napi are:
>
> 1) e1000
> 2) ucc_geth
> 3) ehea
> 4) sunvnet
s2io.c
-Andi
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo i
On Sat, 2007-07-21 at 00:42 -0700, David Miller wrote:
> From: Rusty Russell <[EMAIL PROTECTED]>
> Date: Sat, 21 Jul 2007 17:14:31 +1000
>
> > This looks good! It might be nice though to go further and remove the
> > internal napi_struct. It's kind of a wart for multi-queue drivers which
> >
From: Rusty Russell <[EMAIL PROTECTED]>
Date: Sat, 21 Jul 2007 17:14:31 +1000
> This looks good! It might be nice though to go further and remove the
> internal napi_struct. It's kind of a wart for multi-queue drivers which
> are going to have their own array (or whatever).
I guess you're
On Fri, 2007-07-20 at 22:31 -0700, David Miller wrote:
> Stephen asked me if I could resurrect the last version of
> his napi_poll patch that I posted a long time ago, I finally
> got to that tonight.
>
> Basically, this disconnects the ->poll() object from the net
> device. This will allow drive
From: Patrick McHardy <[EMAIL PROTECTED]>
Date: Sat, 21 Jul 2007 07:46:39 +0200
> I believe the compat attribute should work fine. One thing thing
> that is not explicitly supported as datatype, but works fine, is
> lists of attributes (using nla_for_each_attr), which looks like a
> good match for
David Miller wrote:
[ huge patch, not looked at it :) ]
> One area for discussion is what to do about rtnetlink, it allows
> setting the dev->napi.weight, but we need to extend it so it can
> do something sane in multi-napi-per-netdev situations.
I believe the compat attribute should work fine.
18 matches
Mail list logo
