On 12/08/16 2:15 pm, Garrett D'Amore wrote:
That sounds weird. My first thought was to see if the problem manifests as a
receive filter programming problem.
If traffic flows when snoop is running but then ceases then its probably a mac
address programming problem.
Does it emit traffic
That sounds weird. My first thought was to see if the problem manifests as a
receive filter programming problem.
If traffic flows when snoop is running but then ceases then its probably a mac
address programming problem.
Does it emit traffic properly? You can check that by snooping on a
I recall seeing similar behaviour a while ago (much older platform/box) and it
turned out to be an issue with BNX VLAN support for that particular NIC. The
specifics escape me, but it resulted in having to run the NICs as access ports
and let the switch do the VLAN work. This was with vanilla
I know this topic keeps popping up on these lists and I usually reply
with "fit Intel!" but I'm currently stuck with a new Dell box with
embedded Boardcom NICs.
Has anyone had any luck with these recently (I'm using the latest
platform)? They appear to be half working: the OS manages them,
Hi Robert,
Thanks for looking into this issue.
I tried MSI interrupt type on my own but it didn't work, but I will try
your patch again and then report back.
I've studied the nvme driver in Solaris 11.3, it seems they do the same
thing as Linux - MSI-X first, then MSI, finally FIXED, see
Hi Robert,
there are two other KVMs running on the same host, but there comes almost no
load from them. CPU cap ist not set.
If the KVM does nothing, most of the qemu-processes are near 100% SLP rate (as
expected). LAT is very low near 0,0 / 0,1 %.
If i fire up prime things change: SLP is