netdev - RE: [net-next 14/15] i40e: ignore skb->xmit

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:   Mon, 9 Oct 2017 09:07:21 +0000
From:   David Laight <David.Laight@...LAB.COM>
To:     'Jeff Kirsher' <jeffrey.t.kirsher@...el.com>,
        "davem@...emloft.net" <davem@...emloft.net>
CC:     Jacob Keller <jacob.e.keller@...el.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "nhorman@...hat.com" <nhorman@...hat.com>,
        "sassmann@...hat.com" <sassmann@...hat.com>,
        "jogreene@...hat.com" <jogreene@...hat.com>
Subject: RE: [net-next 14/15] i40e: ignore skb->xmit_more when deciding to
 set RS bit

From: Jeff Kirsher
> Sent: 06 October 2017 18:57
> From: Jacob Keller <jacob.e.keller@...el.com>
> 
> Since commit 6a7fded776a7 ("i40e: Fix RS bit update in Tx path and
> disable force WB workaround") we've tried to "optimize" setting the
> RS bit based around skb->xmit_more. This same logic was refactored
> in commit 1dc8b538795f ("i40e: Reorder logic for coalescing RS bits"),
> but ultimately was not functionally changed.

I've tried to understand the above, but without definitions of WB
and RS and the '4-ness' of something it is quite difficult.

I THINK this is all about telling the hardware there is a packet
in the ring to transmit?

I don't understand the 4-ness. Linux requires that the hardware
be notified of a single packet transmit, and that the 'transmit
complete' also be notified in 'a timely manner' for a single packet.

skb->xmit_more ought to be usable to optimise both of these
(assuming it isn't a lie).
The driver would need to ensure a ring full of messages isn't
generated without either wakeup - but that might be 128 frames.

FWIW on the systems I have PCIe writes are relatively cheap
(reads are slow). So different counts would be appropriate
for delaying doorbell writes and requesting completion interrupts.

	David