Bug 14241
Summary: | r8169 Promiscuous mode on network slows the network down | ||
---|---|---|---|
Product: | Networking | Reporter: | Mudit Wahal (mwahal) |
Component: | IPV4 | Assignee: | Stephen Hemminger (stephen) |
Status: | CLOSED OBSOLETE | ||
Severity: | high | CC: | akpm, alan, mwahal, romieu |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.31 | Subsystem: | |
Regression: | Yes | Bisected commit-id: |
Description
Mudit Wahal
2009-09-26 23:34:38 UTC
Mudit, the comments in http://bugzilla.kernel.org/show_bug.cgi?id=11757 apply : please narrow it down and send a complete dmesg. -- Ueimor (In reply to comment #0) > So, my conclusion is that there is something drastically change in the > ethernet > driver when it runs in promiscuous mode between 2.6.22 and 2.6.28 (or 31) > which > is making it run excruciatingly slow. I marked this as a regression. I'd like to take this bug back. I think something else is going and I need to narrow it down. I tried running same wget script (calls wget about 300 times) without the promiscuous mode and the results are still the same about 30-40mins. If I run a speedtest from a mozilla browser, I get a consistent speed of 5.5Mbps. Also, the responsiveness of 2.6.28 and 2.6.31.1 kernel seems very sluggish over vnc inside my LAN compared to 2.6.22. Then I tried a Linksys 10/100 network card instead of my onboard realtek 8169. I get exact same results - means 300 wgets taking about 40-50mins. So, I don't know if the problem is with networking or some general scheduler issue. I need to narrow down by running some task 300 times which does number crunching in a loop on a 2.6.22 and a 2.6.31.1 kernel. It may be some other issue related to scheduler or something else. I compiled both 2.6.28 and 2.6.31.1 from source using make menuconfig and accepting most of the default options. I want to eliminate networking first. Unfortunately this server is in production network, I will not be able to try it till saturday morning. Thanks Mudit |