Skip to contentSkip to conversation

PATCH] via-rhine.c: don't reference skb after passing it to netif_rx

PATCH Arnaldo Carvalho de Melo 4 participants 6 messages Feb 27, 2001 Network drivers

lore
Share & tools
lore
#0 lore raw

humm, almost finishing... 8)

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

drivers/net/via-rhine.c +1 -1
@@ -1147,9 +1147,9 @@
 								 np->rx_buf_sz, PCI_DMA_FROMDEVICE);
 			}
 			skb->protocol = eth_type_trans(skb, dev);
+			np->stats.rx_bytes += skb->len;
 			netif_rx(skb);
 			dev->last_rx = jiffies;
-			np->stats.rx_bytes += skb->len;
 			np->stats.rx_packets++;
 		}
 		entry = (++np->cur_rx) % RX_RING_SIZE;
#1 lore raw
Quoted 8 lines · Arnaldo Carvalho de Melo

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

drivers/net/via-rhine.c line 1154

Quoted 9 lines
drivers/net/via-rhine.c +1 -1
@@ -1147,9 +1147,9 @@
 								 np->rx_buf_sz, PCI_DMA_FROMDEVICE);
 			}
 			skb->protocol = eth_type_trans(skb, dev);
+			np->stats.rx_bytes += skb->len;
 			netif_rx(skb);
 			dev->last_rx = jiffies;
-			np->stats.rx_bytes += skb->len;
 			np->stats.rx_packets++;
 		}

Easier fix:
- np->stats.rx_bytes += skb->len;
+ np->stats.rx_bytes += pkt_len;

Grouping the writes to np->stats results in better cache usage.

Donald Becker [email protected]
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993

#2 lore raw
In reply to Arnaldo Carvalho de Melo

Em Mon, Feb 26, 2001 at 08:52:39PM -0500, Donald Becker escreveu:

Quoted 24 lines · Arnaldo Carvalho de Melo
Quoted 9 lines

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

drivers/net/via-rhine.c line 1154

Quoted 9 lines
drivers/net/via-rhine.c +1 -1
@@ -1147,9 +1147,9 @@
 								 np->rx_buf_sz, PCI_DMA_FROMDEVICE);
 			}
 			skb->protocol = eth_type_trans(skb, dev);
+			np->stats.rx_bytes += skb->len;
 			netif_rx(skb);
 			dev->last_rx = jiffies;
-			np->stats.rx_bytes += skb->len;
 			np->stats.rx_packets++;
 		}

Easier fix:

(quoted diff) +1 -1
-			np->stats.rx_bytes += skb->len;
+			np->stats.rx_bytes += pkt_len;
Quoted 1 line

Grouping the writes to np->stats results in better cache usage.

thanks, I'll take that into account for the remaining ones and this should be checked by the driver authors for the ones I've already sent.

- Arnaldo

#3 lore raw
Quoted 2 lines · Arnaldo Carvalho de Melo

thanks, I'll take that into account for the remaining ones and this should be checked by the driver authors for the ones I've already sent.

The pkt_len variant is already in 2.4.1-ac15 and probably before that (changed by Manfred Spraul back in ac4?).

Perhaps you should run through the drivers in -acX instead/also? (too late now I guess :)

/Urban

#4 lore raw

Re: PATCH] via-rhine.c: don't reference skb after passing it tonetif_rx

Quoted 11 lines · Urban Widmark
Quoted 2 lines · Arnaldo Carvalho de Melo

thanks, I'll take that into account for the remaining ones and this should be checked by the driver authors for the ones I've already sent.

The pkt_len variant is already in 2.4.1-ac15 and probably before that (changed by Manfred Spraul back in ac4?).

Perhaps you should run through the drivers in -acX instead/also? (too late now I guess :)

FWIW I sync up all my network driver patches (include most of the last_rx/etc changes) with Alan first, and then ship them to Linus as they are proven stable.

Jeff

Signature · 3 lines
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
#5 lore raw
In reply to Arnaldo Carvalho de Melo

Em Tue, Feb 27, 2001 at 10:07:07AM +0100, Urban Widmark escreveu:

Quoted 10 lines · Arnaldo Carvalho de Melo
Quoted 3 lines

thanks, I'll take that into account for the remaining ones and this should be checked by the driver authors for the ones I've already sent.

The pkt_len variant is already in 2.4.1-ac15 and probably before that (changed by Manfred Spraul back in ac4?).

Perhaps you should run through the drivers in -acX instead/also? (too late now I guess :)

well, I think that some of the patches will apply cleanly to both 2.4.2 stock and 2.4.2-acLATEST, or at least sound as a hint to driver maintainers. I'll wait for the next ac patch and look again.

- Arnaldo