[PATCH] 3c589_cs: don't reference skb after passing it to netif

linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed

* [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx
@ 2001-02-27  0:10 Arnaldo Carvalho de Melo
  2001-02-27  1:56 ` Jeff Garzik
  2001-02-27  2:07 ` Jeff Garzik
  0 siblings, 2 replies; 5+ messages in thread
From: Arnaldo Carvalho de Melo @ 2001-02-27  0:10 UTC (permalink / raw)
  To: Alan Cox, dahinds, linux-kernel

now to pcmcia ones

Em Mon, Feb 26, 2001 at 08:33:59PM -0300, Arnaldo Carvalho de Melo escreveu:
Hi,

	I've just read davem's post at netdev about the brokeness of
referencing skbs after passing it to netif_rx, so please consider applying
this patch. Ah, this was just added to the Janitor's TODO list at
http://bazar.conectiva.com.br/~acme/TODO and I'm doing a quick audit in the
net drivers searching for this, maybe some more patches will follow.

- Arnaldo

--- linux-2.4.2/drivers/net/pcmcia/3c589_cs.c	Tue Feb 13 19:15:05 2001
+++ linux-2.4.2.acme/drivers/net/pcmcia/3c589_cs.c	Mon Feb 26 22:44:00 2001
@@ -992,9 +992,9 @@
 			(pkt_len+3)>>2);
 		skb->protocol = eth_type_trans(skb, dev);
 		
+		lp->stats.rx_bytes += skb->len;
 		netif_rx(skb);
 		lp->stats.rx_packets++;
-		lp->stats.rx_bytes += skb->len;
 	    } else {
 		DEBUG(1, "%s: couldn't allocate a sk_buff of"
 		      " size %d.\n", dev->name, pkt_len);

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx
  2001-02-27  0:10 [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx Arnaldo Carvalho de Melo
@ 2001-02-27  1:56 ` Jeff Garzik
  2001-02-27  0:26   ` Arnaldo Carvalho de Melo
  2001-02-27  2:07 ` Jeff Garzik
  1 sibling, 1 reply; 5+ messages in thread
From: Jeff Garzik @ 2001-02-27  1:56 UTC (permalink / raw)
  To: Arnaldo Carvalho de Melo; +Cc: Alan Cox, dahinds, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 919 bytes --]

Arnaldo Carvalho de Melo wrote:
> --- linux-2.4.2/drivers/net/pcmcia/3c589_cs.c   Tue Feb 13 19:15:05 2001
> +++ linux-2.4.2.acme/drivers/net/pcmcia/3c589_cs.c      Mon Feb 26 22:44:00 2001
> @@ -992,9 +992,9 @@
>                         (pkt_len+3)>>2);
>                 skb->protocol = eth_type_trans(skb, dev);
> 
> +               lp->stats.rx_bytes += skb->len;
>                 netif_rx(skb);
>                 lp->stats.rx_packets++;
> -               lp->stats.rx_bytes += skb->len;

I prefer the attached patch instead.  It makes use of the existing local
'pkt_len', and it checks off another item that should probably be on the
janitor's todo list:  Set 'dev->last_rx=jiffies' immediately after
netif_rx.

	Jeff



-- 
Jeff Garzik       | "You see, in this world there's two kinds of
Building 1024     |  people, my friend: Those with loaded guns
MandrakeSoft      |  and those who dig. You dig."  --Blondie

[-- Attachment #2: 3c589-cs.patch --]
[-- Type: text/plain, Size: 684 bytes --]

Index: drivers/net/pcmcia/3c589_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/3c589_cs.c,v
retrieving revision 1.1.1.10.18.1
diff -u -r1.1.1.10.18.1 3c589_cs.c
--- drivers/net/pcmcia/3c589_cs.c	2001/02/25 15:20:31	1.1.1.10.18.1
+++ drivers/net/pcmcia/3c589_cs.c	2001/02/27 01:54:28
@@ -993,8 +993,9 @@
 		skb->protocol = eth_type_trans(skb, dev);
 		
 		netif_rx(skb);
+		dev->last_rx = jiffies;
 		lp->stats.rx_packets++;
-		lp->stats.rx_bytes += skb->len;
+		lp->stats.rx_bytes += pkt_len;
 	    } else {
 		DEBUG(1, "%s: couldn't allocate a sk_buff of"
 		      " size %d.\n", dev->name, pkt_len);

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx
  2001-02-27  1:56 ` Jeff Garzik
@ 2001-02-27  0:26   ` Arnaldo Carvalho de Melo
  0 siblings, 0 replies; 5+ messages in thread
From: Arnaldo Carvalho de Melo @ 2001-02-27  0:26 UTC (permalink / raw)
  To: Jeff Garzik; +Cc: Alan Cox, dahinds, linux-kernel

Em Mon, Feb 26, 2001 at 08:56:06PM -0500, Jeff Garzik escreveu:
> Arnaldo Carvalho de Melo wrote:
> > --- linux-2.4.2/drivers/net/pcmcia/3c589_cs.c   Tue Feb 13 19:15:05 2001
> > +++ linux-2.4.2.acme/drivers/net/pcmcia/3c589_cs.c      Mon Feb 26 22:44:00 2001
> > @@ -992,9 +992,9 @@
> >                         (pkt_len+3)>>2);
> >                 skb->protocol = eth_type_trans(skb, dev);
> > 
> > +               lp->stats.rx_bytes += skb->len;
> >                 netif_rx(skb);
> >                 lp->stats.rx_packets++;
> > -               lp->stats.rx_bytes += skb->len;
> 
> I prefer the attached patch instead.  It makes use of the existing local
> 'pkt_len', and it checks off another item that should probably be on the
> janitor's todo list:  Set 'dev->last_rx=jiffies' immediately after
> netif_rx.

Thanks, I've added your comments and Donald one about grouping the stat
updates, as always the Janitor's TODO list is available at
http://bazar.conectiva.com.br/~acme/TODO, so get your broom and keep on
cleaning 8)

- Arnaldo

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx
  2001-02-27  0:10 [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx Arnaldo Carvalho de Melo
  2001-02-27  1:56 ` Jeff Garzik
@ 2001-02-27  2:07 ` Jeff Garzik
  2001-02-27  0:32   ` Arnaldo Carvalho de Melo
  1 sibling, 1 reply; 5+ messages in thread
From: Jeff Garzik @ 2001-02-27  2:07 UTC (permalink / raw)
  To: Arnaldo Carvalho de Melo; +Cc: Alan Cox, dahinds, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 237 bytes --]

And here are the rest of the ones in pcmcia.
-- 
Jeff Garzik       | "You see, in this world there's two kinds of
Building 1024     |  people, my friend: Those with loaded guns
MandrakeSoft      |  and those who dig. You dig."  --Blondie

[-- Attachment #2: pcmcia-rx.patch --]
[-- Type: text/plain, Size: 4784 bytes --]

Index: drivers/net/pcmcia/3c574_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/3c574_cs.c,v
retrieving revision 1.1.1.11
diff -u -r1.1.1.11 3c574_cs.c
--- drivers/net/pcmcia/3c574_cs.c	2001/02/11 21:28:07	1.1.1.11
+++ drivers/net/pcmcia/3c574_cs.c	2001/02/27 02:05:52
@@ -1166,7 +1166,9 @@
 
 				skb->protocol = eth_type_trans(skb, dev);
 				netif_rx(skb);
+				dev->last_rx = jiffies;
 				lp->stats.rx_packets++;
+				lp->stats.rx_bytes += pkt_len;
 			} else {
 				DEBUG(1, "%s: couldn't allocate a sk_buff of"
 					  " size %d.\n", dev->name, pkt_len);
Index: drivers/net/pcmcia/netwave_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/netwave_cs.c,v
retrieving revision 1.1.1.8
diff -u -r1.1.1.8 netwave_cs.c
--- drivers/net/pcmcia/netwave_cs.c	2001/02/11 21:28:08	1.1.1.8
+++ drivers/net/pcmcia/netwave_cs.c	2001/02/27 02:05:53
@@ -1463,16 +1463,16 @@
 	skb->protocol = eth_type_trans(skb,dev);
 	/* Queue packet for network layer */
 	netif_rx(skb);
-		
+
+	dev->last_rx = jiffies;
+	priv->stats.rx_packets++;
+	priv->stats.rx_bytes += rcvLen;
+
 	/* Got the packet, tell the adapter to skip it */
 	wait_WOC(iobase);
 	writeb(NETWAVE_CMD_SRP, ramBase + NETWAVE_EREG_CB + 0);
 	writeb(NETWAVE_CMD_EOC, ramBase + NETWAVE_EREG_CB + 1);
 	DEBUG(3, "Packet reception ok\n");
-		
-	priv->stats.rx_packets++;
-
-	priv->stats.rx_bytes += skb->len;
     }
     return 0;
 }
Index: drivers/net/pcmcia/nmclan_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/nmclan_cs.c,v
retrieving revision 1.1.1.9
diff -u -r1.1.1.9 nmclan_cs.c
--- drivers/net/pcmcia/nmclan_cs.c	2001/02/11 21:28:08	1.1.1.9
+++ drivers/net/pcmcia/nmclan_cs.c	2001/02/27 02:05:53
@@ -1288,9 +1288,9 @@
 	skb->protocol = eth_type_trans(skb, dev);
 	
 	netif_rx(skb); /* Send the packet to the upper (protocol) layers. */
-
+	dev->last_rx = jiffies;
 	lp->linux_stats.rx_packets++;
-	lp->linux_stats.rx_bytes += skb->len;
+	lp->linux_stats.rx_bytes += pkt_len;
 	outb(0xFF, ioaddr + AM2150_RCV_NEXT); /* skip to next frame */
 	continue;
       } else {
Index: drivers/net/pcmcia/ray_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/ray_cs.c,v
retrieving revision 1.1.1.8
diff -u -r1.1.1.8 ray_cs.c
--- drivers/net/pcmcia/ray_cs.c	2001/02/11 21:28:06	1.1.1.8
+++ drivers/net/pcmcia/ray_cs.c	2001/02/27 02:05:54
@@ -2219,9 +2219,9 @@
 
     skb->protocol = eth_type_trans(skb,dev);
     netif_rx(skb);
-
+    dev->last_rx = jiffies;
     local->stats.rx_packets++;
-    local->stats.rx_bytes += skb->len;
+    local->stats.rx_bytes += total_len;
 
     /* Gather signal strength per address */
 #ifdef WIRELESS_SPY
Index: drivers/net/pcmcia/smc91c92_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/smc91c92_cs.c,v
retrieving revision 1.1.1.12.2.2
diff -u -r1.1.1.12.2.2 smc91c92_cs.c
--- drivers/net/pcmcia/smc91c92_cs.c	2001/02/25 15:20:31	1.1.1.12.2.2
+++ drivers/net/pcmcia/smc91c92_cs.c	2001/02/27 02:05:55
@@ -1617,8 +1617,9 @@
 	
 	skb->dev = dev;
 	netif_rx(skb);
+	dev->last_rx = jiffies;
 	smc->stats.rx_packets++;
-	smc->stats.rx_bytes += skb->len;
+	smc->stats.rx_bytes += packet_length;
 	if (rx_status & RS_MULTICAST)
 	    smc->stats.multicast++;
     } else {
Index: drivers/net/pcmcia/wavelan_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/wavelan_cs.c,v
retrieving revision 1.1.1.16.2.1
diff -u -r1.1.1.16.2.1 wavelan_cs.c
--- drivers/net/pcmcia/wavelan_cs.c	2001/02/23 00:15:34	1.1.1.16.2.1
+++ drivers/net/pcmcia/wavelan_cs.c	2001/02/27 02:05:56
@@ -2733,8 +2733,9 @@
   netif_rx(skb);
 
   /* Keep stats up to date */
+  dev->last_rx = jiffies;
   lp->stats.rx_packets++;
-  lp->stats.rx_bytes += skb->len;
+  lp->stats.rx_bytes += sksize;
 
 #ifdef DEBUG_RX_TRACE
   printk(KERN_DEBUG "%s: <-wv_packet_read()\n", dev->name);
Index: drivers/net/pcmcia/xirc2ps_cs.c
===================================================================
RCS file: /cvsroot/gkernel/linux_2_4/drivers/net/pcmcia/xirc2ps_cs.c,v
retrieving revision 1.1.1.14.2.2
diff -u -r1.1.1.14.2.2 xirc2ps_cs.c
--- drivers/net/pcmcia/xirc2ps_cs.c	2001/02/23 03:37:00	1.1.1.14.2.2
+++ drivers/net/pcmcia/xirc2ps_cs.c	2001/02/27 02:05:57
@@ -1420,6 +1420,7 @@
 		skb->protocol = eth_type_trans(skb, dev);
 		skb->dev = dev;
 		netif_rx(skb);
+		dev->last_rx = jiffies;
 		lp->stats.rx_packets++;
 		lp->stats.rx_bytes += pktlen;
 		if (!(rsr & PhyPkt))

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx
  2001-02-27  2:07 ` Jeff Garzik
@ 2001-02-27  0:32   ` Arnaldo Carvalho de Melo
  0 siblings, 0 replies; 5+ messages in thread
From: Arnaldo Carvalho de Melo @ 2001-02-27  0:32 UTC (permalink / raw)
  To: Jeff Garzik; +Cc: Alan Cox, dahinds, linux-kernel

Em Mon, Feb 26, 2001 at 09:07:49PM -0500, Jeff Garzik escreveu:
> And here are the rest of the ones in pcmcia.

Hey man, thats what I call cooperation 8) I was now on the netwave one, but
had to stop to get another beer, when I came back... Jeff, go get a beer
please, I'll pay you by March, 31, if I meet you in California 8)

- Arnaldo

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2001-02-27  2:12 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-02-27  0:10 [PATCH] 3c589_cs: don't reference skb after passing it to netif_rx Arnaldo Carvalho de Melo
2001-02-27  1:56 ` Jeff Garzik
2001-02-27  0:26   ` Arnaldo Carvalho de Melo
2001-02-27  2:07 ` Jeff Garzik
2001-02-27  0:32   ` Arnaldo Carvalho de Melo

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).