Re: [E1000-devel] Enabling Loopback TX by default for all VFs

2015-02-13 Thread Yotam Rubin
Just to clarify - enabling the PFVMTXSW register for a VF won't cause all TX traffic to be looped back. Rather, it will enable TX packets whose destination MAC address matches the VF's MAC address to be placed in the RX queue. Is there a performance penalty enabling this? If so, I can implement a n

Re: [E1000-devel] Enabling Loopback TX by default for all VFs

2015-02-13 Thread Alexander Duyck
On 02/13/2015 05:12 AM, Yotam Rubin wrote: > Just to clarify - enabling the PFVMTXSW register for a VF won't cause all > TX traffic to be looped back. > Rather, it will enable TX packets whose destination MAC address matches the > VF's MAC address to be placed in the RX queue. > Is there a performa

Re: [E1000-devel] Enabling Loopback TX by default for all VFs

2015-02-13 Thread Yotam Rubin
Thanks, that's a good point. Some applications such as ours, gains from this behaviour. I was thinking about adding an mbox interface so that individual VF's may be configured for loopback tx. By default, this configuration will be disabled for all VF's. Makes sense? On Fri, Feb 13, 2015 at 3:34 P

Re: [E1000-devel] [bisected regression] e1000e: "Detected Hardware Unit Hang"

2015-02-13 Thread Thomas Jarosch
Hi Aaron, On Thursday, 12. February 2015 23:28:27 Brown, Aaron F wrote: > I do not have any real info. I had been asked to try and reproduce some > unit hangs (maybe for this) recently and did not succeed in producing > them on the parts I have. Reading through the thread I see this is > showing

Re: [E1000-devel] Enabling Loopback TX by default for all VFs

2015-02-13 Thread Alexander Duyck
On 02/13/2015 08:46 AM, Yotam Rubin wrote: > Thanks, that's a good point. > Some applications such as ours, gains from this behaviour. > I was thinking about adding an mbox interface so that individual VF's may > be configured for loopback tx. By default, this configuration will be > disabled for a

Re: [E1000-devel] [PATCH 1/3] ixgbe, ixgbevf: Add new mbox API to enable MC promiscuous mode

2015-02-13 Thread Skidmore, Donald C
> -Original Message- > From: Hiroshi Shimamoto [mailto:h-shimam...@ct.jp.nec.com] > Sent: Thursday, February 12, 2015 8:45 PM > To: Skidmore, Donald C; Kirsher, Jeffrey T > Cc: Alexander Duyck; Bjørn Mork; e1000-devel@lists.sourceforge.net; > net...@vger.kernel.org; Choi, Sy Jong; linux-k

Re: [E1000-devel] [PATCH 1/3] ixgbe, ixgbevf: Add new mbox API to enable MC promiscuous mode

2015-02-13 Thread Vlad Yasevich
On 02/13/2015 12:26 PM, Skidmore, Donald C wrote: > > >> -Original Message- >> From: Hiroshi Shimamoto [mailto:h-shimam...@ct.jp.nec.com] >> Sent: Thursday, February 12, 2015 8:45 PM >> To: Skidmore, Donald C; Kirsher, Jeffrey T >> Cc: Alexander Duyck; Bjørn Mork; e1000-devel@lists.source

Re: [E1000-devel] Enabling Loopback TX by default for all VFs

2015-02-13 Thread Yotam Rubin
Our packets are mostly unicast. I'll elaborate on what we use loopbacking for: 1. Benchmarks - Since we rely on NIC's offloading abilities when measuring our application's throughput, we can't use the loopback as its skews the results. 2. Self-Diagnostics - We'd like to be able to loopback traffic