netdev - Re: [PATCH net-next 3/5] netvsc: optimize receive completions

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] [day] [month] [year] [list]

Message-ID: <20170725140757.5c38ccc0@xeon-e3>
Date:   Tue, 25 Jul 2017 14:07:57 -0700
From:   Stephen Hemminger <stephen@...workplumber.org>
To:     kys@...rosoft.com, haiyangz@...rosoft.com, sthemmin@...rosoft.com
Cc:     devel@...uxdriverproject.org, netdev@...r.kernel.org
Subject: Re: [PATCH net-next 3/5] netvsc: optimize receive completions

On Tue, 25 Jul 2017 13:04:20 -0700
Stephen Hemminger <stephen@...workplumber.org> wrote:

> Optimize how receive completion ring are managed.
>    * Allocate only as many slots as needed for all buffers from host
>    * Allocate before setting up sub channel for better error detection
>    * Don't need to keep copy of initial receive section message
>    * Only needt keep the transaction id, status doesn't matter
>    * Precompute the watermark for when receive flushing is needed
>    * Replace division with conditional test
>    * Replace atomic per-device variable with per-channel check.
>    * Handle corner case where receive completion send
>      fails if ring buffer to host is full.
> 
> Signed-off-by: Stephen Hemminger <sthemmin@...rosoft.com>

I probably pushed the shrinkage too far for now. Hiayang raised the
point that probably need to keep the status value. I dropped storing
it since it was always SUCCESS unless the host sent a bad packet,
or the system was out of skb's.

Will send V2 that keeps the status value for now.