netdev - Re: [PATCH 4/6 bfp-next] Simplify AF_XDP umem allocation path for Intel drivers.

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: <5061E73B-6980-4055-917D-D35906D98E9A@gmail.com>
Date:   Fri, 28 Jun 2019 14:06:03 -0700
From:   "Jonathan Lemon" <jonathan.lemon@...il.com>
To:     "Jakub Kicinski" <jakub.kicinski@...ronome.com>
Cc:     netdev@...r.kernel.org, bjorn.topel@...el.com,
        magnus.karlsson@...el.com, saeedm@...lanox.com,
        maximmi@...lanox.com, brouer@...hat.com, kernel-team@...com
Subject: Re: [PATCH 4/6 bfp-next] Simplify AF_XDP umem allocation path for
 Intel drivers.

On 28 Jun 2019, at 13:48, Jakub Kicinski wrote:

> On Thu, 27 Jun 2019 19:36:42 -0700, Jonathan Lemon wrote:
>> On 27 Jun 2019, at 15:42, Jakub Kicinski wrote:
>>> Could you be more explicit on the motivation?  I'd call this patch set
>>> "make all drivers use reuse queue" rather than "clean up".
>>
>> The motivation is to have packets which were received on a zero-copy
>> AF_XDP socket, and which returned a TX verdict from the bpf program,
>> queued directly on the TX ring (if they're in the same napi context).
>>
>> When these TX packets are completed, they are placed back onto the
>> reuse queue, as there isn't really any other place to handle them.
>>
>> It also addresses Maxim's concern about having buffers end up sitting
>> on the rq after a ring resize.
>>
>> I was going to send the TX change out as part of this patch, but
>> figured it would be better split unto its own series.
>
> Makes sense, please put that info in the cover letter.  (in other nit
> picks please prefix the patches which the name of the subsystem they
> are touching - net: xsk: ?, and make sure all changes have a commit
> message).
>
> Looking at this yesterday I think I've seen you didn't change the RQ
> sizing -  for the TX side to work the rq size must be RX ring size
> + XDP TX ring size (IIRC that was the worst case number of packets
> which we may need to park during reconfiguration).
>
> Maybe you did do that and I missed it.

Not yet - that's in the TX section of the patches, I'll post the
Intel versions shortly.



>>> Also when you're changing code please make sure you CC the author.
>>
>> Who did I miss?
>
> commit f5bd91388e26557f64ca999e0006038c7a919308
> Author: Jakub Kicinski <jakub.kicinski@...ronome.com>
> Date:   Fri Sep 7 10:18:46 2018 +0200
>
>     net: xsk: add a simple buffer reuse queue
>
>     XSK UMEM is strongly single producer single consumer so reuse of
>     frames is challenging.  Add a simple "stash" of FILL packets to
>     reuse for drivers to optionally make use of.  This is useful
>     when driver has to free (ndo_stop) or resize a ring with an active
>     AF_XDP ZC socket.
>
>     Signed-off-by: Jakub Kicinski <jakub.kicinski@...ronome.com>
>     Tested-by: Andrew Bowers <andrewx.bowers@...el.com>
>     Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@...el.com>
>
>
> Sorry to get upset, I missed a bpf patch recently and had to do
> a revert..

Ack.