netdev - Re: [PATCH net-next 3/3] net: stmmac: Introducing support for Page Pool

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: <20190704130635.GA10412@apalos>
Date:   Thu, 4 Jul 2019 16:06:35 +0300
From:   Ilias Apalodimas <ilias.apalodimas@...aro.org>
To:     Jose Abreu <Jose.Abreu@...opsys.com>
Cc:     Jesper Dangaard Brouer <brouer@...hat.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "linux-stm32@...md-mailman.stormreply.com" 
        <linux-stm32@...md-mailman.stormreply.com>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        Joao Pinto <Joao.Pinto@...opsys.com>,
        "David S . Miller" <davem@...emloft.net>,
        Giuseppe Cavallaro <peppe.cavallaro@...com>,
        Alexandre Torgue <alexandre.torgue@...com>,
        Maxime Coquelin <mcoquelin.stm32@...il.com>,
        Maxime Ripard <maxime.ripard@...tlin.com>,
        Chen-Yu Tsai <wens@...e.org>
Subject: Re: [PATCH net-next 3/3] net: stmmac: Introducing support for Page
 Pool

Hi Jose, 

> Thank you all for your review comments !
> 
> From: Ilias Apalodimas <ilias.apalodimas@...aro.org>
> 
> > That's why i was concerned on what will happen on > 1000b frames and what the
> > memory pressure is going to be. 
> > The trade off here is copying vs mapping/unmapping.
> 
> Well, the performance numbers I mentioned are for TSO with default MTU 
> (1500) and using iperf3 with zero-copy. Here follows netperf:
> 

Ok i guess this should be fine. Here's why.
You'll allocate an extra memory from page pool API which equals
the number of descriptors * 1 page.
You also allocate SKB's to copy the data and recycle the page pool buffers.
So page_pool won't add any significant memory pressure since we expect *all*
it's buffers to be recycled. 
The SKBs are allocated anyway in the current driver so bottom line you trade off
some memory (the page_pool buffers) + a memcpy per packet and skip the dma
map/unmap which is the bottleneck in your hardware. 
I think it's fine

Cheers
/Ilias