Hey Liu,
I’ll attempt to answer your questions inline below:
Thanks,
-Don Skidmore
From: lzh [mailto:lhqlzh...@163.com]
Sent: Thursday, December 18, 2014 10:12 PM
To: Skidmore, Donald C; Tom Barbette
Cc: e1000-devel@lists.sourceforge.net
Subject: Re:RE: [E1000-devel] RSS related questions
Hi
tion creating packets you will just have one flow and thus just use one
>queue.
>
>Thanks,
>-Don Skidmore
>
>> -Original Message-
>> From: lzh [mailto:lhqlzh...@163.com]
>> Sent: Thursday, December 18, 2014 12:33 AM
>> To: lzh; e1000-devel@lists.source
12:33 AM
> To: lzh; e1000-devel@lists.sourceforge.net
> Cc: e1000-devel@lists.sourceforge.net
> Subject: Re: [E1000-devel] RSS related questions
>
>
>
>
> There is a another problem. When we set all 20 irqs to both cpu0 and
> cpu1(echo 3 > "/proc/irq//smp
Hi,
Are the packets correctly distributed to different queues? (look at
/proc/interrupts).
RSS which is used to split packets among queues only supports up to 16
queues. Moreover, for the hash function to be effective the packets have to
be different (if you use a pktgen, that might not be the cas
There is a another problem. When we set all 20 irqs to both cpu0 and cpu1(echo
3 > "/proc/irq//smp_affinity".), the irq all go to cpu0.
We use x86 64bit. so the irq balance is done by apic hardware (setting
"msi_set_affinity" in linux/arch/x86/kernel/apic/io_apic.c), why it doesn't
take effe
Hi,
We are using 82599 intel netcard that using 20 rxtx queues. We test it on
cpu E5-2670 which has 20 cpu cores. We bind every queues to each cpu by
"/proc/irq//smp_affinity".
First, we test the throughput and find that the rx irq is not distributed
to all cpu cores. Wd don't know why
