Hi

Thanks for the reply. Could you not send batches of multicast addresses through 
the mailbox? Alternatively could we make the bits used for the hash 
configurable so that we could fit our multicast groups into a small number of 
hash table entries. There is a comment in the code that says "add handling 
later".

I have an application that needs to join 100s of multicast groups, and I would 
like to use sr-iov. I understand that there is a performance impact but as it 
is (and if I understand correctly) a system using a VF thinks it is a member of 
certain multicast groups but it does not receive that packets; that can't be 
right...

Regards,

Jonathan


-----Original Message-----
From: Fujinaka, Todd [mailto:[email protected]] 
Sent: 11 June 2014 01:35
To: Jonathan Purcell; [email protected]
Subject: RE: igb/igbvf 30 multicast hash limit.

We are limited to 30 because of the size of the mailbox used for communication 
from the PF to VF which is only 16 x 32b registers in size, and the first 32b 
register is reserved for the head.  That leaves 15 registers to split into 2 x 
16b hashes each.

Too many multicast addresses requested by a VF isn't something we can 
accommodate because the PF and every VF receive a copy of ANY 
multicast/broadcast packet received.  As a result, performance with multicast 
and SR-IOV will be very poor with too much multicast address-based traffic.

Todd Fujinaka
Software Application Engineer
Networking Division (ND)
Intel Corporation
[email protected]
(503) 712-4565

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Jonathan Purcell
Sent: Tuesday, June 10, 2014 8:39 AM
To: [email protected]; Linux NICS
Subject: [linux-nics] igb/igbvf 30 multicast hash limit.

Hi

I am trying to understand the impact of the limit of 30 multicast hashes sent 
from igbvf (e1000_update_mc_addr_list_vf() in e1000_vf.c)  to igb 
(igb_set_vf_multicasts() in igb_main.c).

Does this mean that if a system using a VF wants to join more than 30 multicast 
groups whose low order 12 bits are distinct, in fact it can't?

If that is the case is there a way to overcome this?

Thanks in advance

Jonathan



This message is intended only for the stated addressee(s) and may be 
confidential. Access to this email by anyone else is unauthorised. Any opinions 
expressed in this email do not necessarily reflect the opinions of Fidessa. Any 
unauthorised disclosure, use or dissemination, either whole or in part is 
prohibited. If you are not the intended recipient of this message, please 
notify the sender immediately.
Fidessa plc registered in England and Wales no. 3781700. VAT registration no. 
GB688900878. Registered office - Dukes Court, Duke Street, Woking, Surrey, GU21 
5BH, UK Fidessa buy-side ltd registered in England and Wales no. 3656437. VAT 
registration no. GB688900878. Registered office - Dukes Court, Duke Street, 
Woking, Surrey, GU21 5BH, UK Fidessa group plc registered in England and Wales 
no. 3234176. VAT registration no. GB688900878. Registered office - Dukes Court, 
Duke Street, Woking, Surrey, GU21 5BH, UK 
_______________________________________________
Linux-nics mailing list
[email protected]

------------------------------------------------------------------------------
HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions
Find What Matters Most in Your Big Data with HPCC Systems
Open Source. Fast. Scalable. Simple. Ideal for Dirty Data.
Leverages Graph Analysis for Fast Processing & Easy Data Exploration
http://p.sf.net/sfu/hpccsystems
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit 
http://communities.intel.com/community/wired

Reply via email to