From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eran Liberty Subject: Re: [PATCH] gainfar.c : skb_over_panic Date: Wed, 23 Jun 2010 18:03:16 +0300 Message-ID: <4C222234.60903@extricom.com> References: <4C1A4E36.5060902@extricom.com> <20100617.122030.112600189.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: galak@kernel.crashing.org, netdev@vger.kernel.org To: David Miller Return-path: Received: from smtp.extricom.com ([192.114.46.18]:56829 "HELO smtp.extricom.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1753335Ab0FWPGo (ORCPT ); Wed, 23 Jun 2010 11:06:44 -0400 In-Reply-To: <20100617.122030.112600189.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: David Miller wrote: > From: Eran Liberty > Date: Thu, 17 Jun 2010 19:32:54 +0300 > > >> I have demonstrated skb_over_panic with linux 2.6.32.15 on a mpc8548 >> based product. >> > > A fix for a similar bug was necessary for the ucc_geth driver, > see below. > > The real problem is that skb->data assignment, the rest of the > SKB state has to be reset, and not doing that is what results in > the skb_over_panic calls. > > >From db176edc89abbf22e6db6853f8581f9475fe8ec1 Mon Sep 17 00:00:00 2001 > From: Sergey Matyukevich > Date: Mon, 14 Jun 2010 06:35:20 +0000 > Subject: [PATCH] ucc_geth: fix for RX skb buffers recycling > > This patch implements a proper modification of RX skb buffers before > recycling. Adjusting only skb->data is not enough because after that > skb->tail and skb->len become incorrect. > > Signed-off-by: Sergey Matyukevich > Signed-off-by: David S. Miller > --- > drivers/net/ucc_geth.c | 2 ++ > 1 files changed, 2 insertions(+), 0 deletions(-) > > diff --git a/drivers/net/ucc_geth.c b/drivers/net/ucc_geth.c > index 4a34833..807470e 100644 > --- a/drivers/net/ucc_geth.c > +++ b/drivers/net/ucc_geth.c > @@ -3215,6 +3215,8 @@ static int ucc_geth_rx(struct ucc_geth_private *ugeth, u8 rxQ, int rx_work_limit > __func__, __LINE__, (u32) skb); > if (skb) { > skb->data = skb->head + NET_SKB_PAD; > + skb->len = 0; > + skb_reset_tail_pointer(skb); > __skb_queue_head(&ugeth->rx_recycle, skb); > } > > When I do go via this code this patch helps. But, I have managed to reach the skb_over_panic without going first via __skb_queue_head() which render this patch useless... So I am investigating this before suggesting any patch. doing something like this: if (unlikely(skb->tail + pkt_len > skb->end)) { pr_err("gfar_clean_rx_ring(): skb_over_panic event avoided\n"); dev_kfree_skb_any(skb); } else { skb_put(skb, pkt_len); dev->stats.rx_bytes += pkt_len; if (in_irq() || irqs_disabled()) printk("Interrupt problem!\n"); gfar_process_frame(dev, skb, amount_pull); } successfully avoids the skb_over_panic(), But I rather find the offending skb creator then continuously defend against its arrival. -- Liberty