netdev
[Top] [All Lists]

Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]

To: ak@xxxxxx, hadi@xxxxxxxxxx
Subject: Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]
From: Bob Felderman <feldy@xxxxxxxx>
Date: Fri, 9 Mar 2001 11:12:48 -0800 (PST)
Cc: andrewm@xxxxxxxxxx, davem@xxxxxxxxxx, feldy@xxxxxxxx, kuznet@xxxxxxxxxxxxx, netdev@xxxxxxxxxxx, pp@xxxxxxxxxxxxxx
Sender: owner-netdev@xxxxxxxxxxx
=> Alexey's assesment that the culprit might be he might be loosing some
=> of those frags in the backlog
=> What kind of hardware is this thing running on? 1.5 G seems a lot on
=> a regular PC. very Large MTUs?

9KB MTU 
Supermicro 370DLE motherboard P-III 666MHz
using a pci 64bit 66mhz slot with a Myrinet card.

This board uses the serverworks chipset for PCI and we
have measured 400-500MegaBytes/sec (3.2 - 4.0 Gigabit/sec)
transfer rates across the bus. Our link speed is 2Gigabit/sec.

=> Bob, One of the things that could be done is have the driver stop sending
=> packets after the toplayer gets congested (return is NET_RX_CN_HIGH).
=> linux/include/linux/netdevice.h has the NET_RX_* which are returned
=> by netif_rx()

I'll look at it, but DaveM's zero-copy patches also solved
the packet drop for me. I have not tried yet to put that
patch on my 2.4.2 fixed kernels.


<Prev in Thread] Current Thread [Next in Thread>