netdev - RE: packetloss, on e1000e worse than r8169?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <36D9DB17C6DE9E40B059440DB8D95F52056AD09B@orsmsx418.amr.corp.intel.com>
Date:	Tue, 17 Jun 2008 17:58:00 -0700
From:	"Brandeburg, Jesse" <jesse.brandeburg@...el.com>
To:	"Denys Fedoryshchenko" <denys@...p.net.lb>
Cc:	<netdev@...r.kernel.org>
Subject: RE: packetloss, on e1000e worse than r8169?

Denys Fedoryshchenko wrote:
> More info. Now it is peak time, and flow control disabled. Before it
> was 1% packetloss on small packets and 5-6% on large, now it is 0.3%
> on small, 1-2% on large. The flow-control required on this particular
> setup (probably linksys switches is too much crappy, no idea), and
> without it i experience "small" packetloss"   
> But seems on this motherboard, by some reasons flow-control not
> working, maybe it is related to PBA. 
> 
> I will try to play with PBA now, but not sure i can do that.

yes, on ICH8 playing with PBA will not give you much, ICH8 was meant to
be a "client" part and so sacrificed some FIFO space to save cost/power.

Also, our adapters support variable flow control thresholds (but we have
no good way to tune), it could be that your length of wire to your
switch or the switch itself takes too long to respond to flow control.

This would mean that too many packets get enqueued to the wire/hardware
before the flow control gets reacted to at the switch.

the registers are FCRTL and FCRTH, the numbers are in bytes, the
hardware sends a flow control STOP (txoff) packet when the FIFO fullness
goes past FCRTH, and sends flow control GO (txon) packet when threshold
drops below FCRTL.

Maybe that will help, let me know what else I can help with.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html