netdev - Re: [PATCH net-next] tls: Add opt-in zerocopy mode of sendfile()

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: <20220503122423.29f48b61@kernel.org>
Date:   Tue, 3 May 2022 12:24:23 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Maxim Mikityanskiy <maximmi@...dia.com>
Cc:     "David S. Miller" <davem@...emloft.net>,
        Daniel Borkmann <daniel@...earbox.net>,
        Paolo Abeni <pabeni@...hat.com>,
        Boris Pismenny <borisp@...dia.com>,
        Tariq Toukan <tariqt@...dia.com>,
        Saeed Mahameed <saeedm@...dia.com>,
        Gal Pressman <gal@...dia.com>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next] tls: Add opt-in zerocopy mode of sendfile()

On Tue, 3 May 2022 21:56:48 +0300 Maxim Mikityanskiy wrote:
> >> Yes, I agree that if the application opted in, it should work properly
> >> regardless of whether the optimization actually did turn on. However,
> >> the indication could be useful, for example, for diagnostic purposes, to
> >> show the user whether zerocopy mode was enabled, if someone is trying to
> >> debug some performance issue. If you insist, though, I can make
> >> setsockopt succeed and getsockopt return 1. What do you think?  
> > 
> > I'd say "whether the optimization is applicable" rather than "whether
> > the optimization is turned on". User can check whether the connection
> > is using SW or HW TLS if they want to make sure it's taken advantage of.
> > 
> > Speaking of which, should we report the state of this knob via socket
> > diag?  
> 
> That sounds like an option, I'll take a look. TLS doesn't expose 
> anything via diag yet, does it? The only option to distinguish SW/HW TLS 
> is ethtool, and there is no per-socket check, right? Cause a HW TLS 
> socket can downgrade to SW after tls_device_down, and ethtool won't show it.

It does - look for tls_get_info()