ksoftirqd uses 99% CPU triggered by network traffic (maybe RLT-8139 related)
Quoted 10 lines · Will S.
Quoted 5 lines
Yeah, it might be some sort of a bug which is related to VIA Technologies, Inc. VT8366/A/7 [Apollo KT266/A/333]-chipset and RTL-8139? Hard to say at this point. I'll get some other network cards tomorrow from the office and we will see if there is a difference. I hope that it's not the mobo (chipset). =)
I just remembered something. In the kernel config, there's an option for the RTL-8139 driver to use polling I/O instead of memory mapped I/O - and it usually defaults to PIO. My kernel was compiled using the MMIO option. Check and see what you're using.
I'm using MMIO-option, I was just thinking to go with the PIO and see what will happen.
I just found this thread (url below) and some others are having problems also.
http://seclists.org/lists/linux-kernel/2004/Mar/2295.html
Andrew:
Did you find any solution to this one? (I guess not but could I be in help someway to hunt this bug down?)
Quoted 9 lines · Pasi Sjoholm
I just found this thread (url below) and some others are having problems
also.http://seclists.org/lists/linux-kernel/2004/Mar/2295.html
Andrew:
Did you find any solution to this one?
Nope. We still await the intersection of someone who can reproduce it with someone who has the time/ability/inclination to diagnose it.