Not sure if NAPI makes sense on transmit because it causes the transmit ring to grow and freeing the transmit skb should be quick. Perhaps other interrupt moderation schemes work better. On receive t
<snip!> <snip!> Okay, well, rx_dropped means "receive no buffers count" in our driver, but doesn't necessarily mean that the packet was completely dropped it just means that the fifo may have had to
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Mon, 10 Jan 2005 15:11:02 -0600
I rebound the card with 2048 RxDescriptors. There appears to be a burst errors right when the port comes online. Down below is an ifconfig ; sleep 60 ; ifconfig. I'll be tesitng Robert's patch shortl
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Tue, 11 Jan 2005 09:11:57 -0600
The code Jesse Brandeburg had me download has the patch in it. -- -- Jeremy M. Guthrie jeremy.guthrie@xxxxxxxxxx Senior Network Engineer Phone: 608-298-1061 Berbee Fax: 608-288-3007 5520 Research Par
About the same load as last time 80 kpps forwarded and droprate of ~200 pps and 2323 fib lookups/sec. Not so bad for PIII but it maybe tuned better. Did the e1000 patch cure the problem that interrup
Yes rtstats is what we estimated. How about RX ring size and the e1000 fix? We give up here? Or you want to try another setting of route hash? It's beyond the scope of this subject. You have to reboo
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Tue, 11 Jan 2005 19:27:13 -0600
ETH3 Interrupts(calc'd from below): 1479968 ETH2 Interrupts: 261543 Packets RX'd on ETH3: 3892720 Packets dropped on RX on ETH3: 10305 This equates to about a 0.26% drop rate. W/ 256 packet RX ring s
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Tue, 11 Jan 2005 19:30:33 -0600
I am up for pushing it if you do not think it is a waste of time. Based on what I am seeing it looks like I just need a faster CPU to do the work. My goal would be to hit zero dropped packets w/ 10-1
Very strange... eth3 is bound to CPU0 which in turn has all packet load... If we were to believe your CPU0 was saturated (due to the drops). We should see no (RX) interrupts on eth3. But there is a l
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Wed, 12 Jan 2005 10:24:59 -0600
I have no idea why it would be doing this. This appears to be where this problem is getting beyond my expertise. I verified NAPI is turned on. I also verified the patch is in place. I am open to sugg
Huh seems you didn't add the patch I sent. Below is diff from my editor to your e1000_main.c --ro -- e1000_main.c.jmg 2005-01-12 20:14:08.324168072 +0100 +++ e1000_main.c 2005-01-12 20:17:24.77730265
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Wed, 12 Jan 2005 14:30:00 -0600
Sorry, here is the latest. BTW, it is VERY sluggish now. Wed Jan 12 14:25:03 CST 2005 eth3 Link encap:Ethernet HWaddr 00:02:B3:D5:7E:30 RX packets:32030222 errors:105026820 dropped:105026820 overruns
Author: "Jeremy M. Guthrie" <jeremy.guthrie@xxxxxxxxxx>
Date: Wed, 12 Jan 2005 14:45:48 -0600
My throughput dropped from 500 mbps to 8mbps. 8( -- -- Jeremy M. Guthrie jeremy.guthrie@xxxxxxxxxx Senior Network Engineer Phone: 608-298-1061 Berbee Fax: 608-288-3007 5520 Research Park Drive NOC: 6
Weird! There are no irq's on eth3 at all so RX softirq is constantly running. Which means it's deferred to ksoftirqd now and running under scheduler context. Do you have anything that competes with k