[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 03 Mar 2015 05:29:59 -0800
From: Eric Dumazet <eric.dumazet@...il.com>
To: Nicolas Schichan <nschichan@...ebox.fr>
Cc: "David S. Miller" <davem@...emloft.net>,
Tobias Klauser <tklauser@...tanz.ch>,
Felipe Balbi <balbi@...com>,
Wilfried Klaebe <w-lkml@...enslange-mailadresse.de>,
"Eric W. Biederman" <ebiederm@...ssion.com>,
Alexander Duyck <alexander.h.duyck@...hat.com>,
netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] bcm63xx_enet: fix poll callback.
On Tue, 2015-03-03 at 12:45 +0100, Nicolas Schichan wrote:
> In case there was some tx buffer reclaimed and not enough rx packets
> to consume the whole budget, napi_complete would not be called and
> interrupts would be kept disabled, effectively resulting in the
> network core never to call the poll callback again and no rx/tx
> interrupts to be fired either.
>
> Fix that by only accounting the rx work done in the poll callback.
>
> Signed-off-by: Nicolas Schichan <nschichan@...ebox.fr>
> ---
This looks better, thanks.
Acked-by: Eric Dumazet <edumazet@...gle.com>
Note that the way bcm_enet_tx_reclaim() is written, it can livelock on
SMP hosts :
CPU 1,2,3,... keep queuing packets via ndo_start_xmit()
CPU 0 is looping forever in bcm_enet_tx_reclaim() draining queue and
freeing skbs.
To avoid that, I would take priv->tx_lock only once, or add a limit on
the number of skbs that can be drained per round.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists