netdev - Re: r8169 tx batching(?) causing performance problems

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]

Date:   Wed, 03 Oct 2018 21:19:40 +0100
From:   David Howells <dhowells@...hat.com>
To:     David Miller <davem@...emloft.net>
Cc:     dhowells@...hat.com, nic_swsd@...ltek.com, netdev@...r.kernel.org
Subject: Re: r8169 tx batching(?) causing performance problems

David Miller <davem@...emloft.net> wrote:

> Probably you are seeing some interrupt mitigation.
> 
> It seems there is a difference in how the interrupt mitigation is
> programmed on for 8168 chips vs. others by default.  Most get
> all zeros in the IntrMitigate register, whilst for 8168 chips
> a value of 0x5151 is programmed.

I'm not sure what that means.  I can't seem to find a programmer's manual for
the chip.

> You can play with ethtool to mess with the coalescing settings
> to see if this is part of the problem.

These bits from "ethtool -c enp3s0"?

	rx-usecs: 200
	rx-frames: 4
	rx-usecs-irq: 0
	rx-frames-irq: 0

	tx-usecs: 200
	tx-frames: 4
	tx-usecs-irq: 0
	tx-frames-irq: 0

> I bet this might explain the behavior you see after including
> even Heiner's TXCFG_AUTO_FIFO patch.

Thanks,
David