netdev - Re: [PATCH v3 net-next 07/21] nvme-tcp: Add DDP data-path

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, 3 Feb 2021 02:21:01 -0800
From:   Sagi Grimberg <sagi@...mberg.me>
To:     Christoph Hellwig <hch@....de>
Cc:     Or Gerlitz <gerlitz.or@...il.com>, Keith Busch <kbusch@...nel.org>,
        axboe@...com, Boris Pismenny <borisp@...lanox.com>,
        smalin@...vell.com, yorayz@...dia.com, boris.pismenny@...il.com,
        Ben Ben-Ishay <benishay@...lanox.com>,
        Yoray Zack <yorayz@...lanox.com>,
        linux-nvme@...ts.infradead.org, David Miller <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Linux Netdev List <netdev@...r.kernel.org>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        David Ahern <dsahern@...il.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Or Gerlitz <ogerlitz@...lanox.com>, benishay@...dia.com,
        Saeed Mahameed <saeedm@...dia.com>,
        Or Gerlitz <ogerlitz@...dia.com>
Subject: Re: [PATCH v3 net-next 07/21] nvme-tcp: Add DDP data-path


>>>> Given how much ddp code there is can you split it into a separate file?
>>>
>>> mmm, do we need to check the preferences or get to a consensus among
>>> the maintainers for that one?
>>
>> Not sure if moving it would be better here. Given that the ddp code is
>> working directly on nvme-tcp structs we'll need a new shared header
>> file..
>>
>> Its possible to do, but I'm not sure the end result will be better..
> 
> In the end its your code base.  But I hate having all this offload
> cruft all over the place.

I know, I think that the folks did a solid job consolidating it
given the complexity. But looking at the code again, it is sprinkled
more than I'd like it to be. I think it can be better with a little
more work.

If we can get to a point where we have all the specific logic
moved to dedicated routines and just a few interceptions on
the main flows we should be ok.

> Just saying no to offloads might be an even better position, though.

:)

I've heard complaints about nvme-tcp taking more cpu cycles than
nvme-rdma (well duh..) so I'm not opposed to having mainstream devices
offering offload capabilities to help out with that, not at all.