netdev - Re: [RFC PATCH net-next 0/3] net: batched receive in GRO 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, 10 Jul 2019 19:39:29 +0200
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Edward Cree <ecree@...arflare.com>,
        Eric Dumazet <eric.dumazet@...il.com>,
        Paolo Abeni <pabeni@...hat.com>,
        David Miller <davem@...emloft.net>
Cc:     netdev <netdev@...r.kernel.org>
Subject: Re: [RFC PATCH net-next 0/3] net: batched receive in GRO path



On 7/10/19 6:47 PM, Edward Cree wrote:
> On 10/07/2019 16:41, Eric Dumazet wrote:
>> On 7/10/19 4:52 PM, Edward Cree wrote:
>>> Hmm, I was caught out by the call to napi_poll() actually being a local
>>>  function pointer, not the static function of the same name.  How did a
>>>  shadow like that ever get allowed?
>>> But in that case I _really_ don't understand napi_busy_loop(); nothing
>>>  in it seems to ever flush GRO, so it's relying on either
>>>  (1) stuff getting flushed because the bucket runs out of space, or
>>>  (2) the next napi poll after busy_poll_stop() doing the flush.
>>> What am I missing, and where exactly in napi_busy_loop() should the
>>>  gro_normal_list() call go?
>> Please look at busy_poll_stop()
> I did look there, but now I've looked again and harder, and I think I get it:
> busy_poll_stop() calls napi->poll(), which (eventually, possibly in the
>  subsequent poll that we schedule if rc == budget) calls napi_complete_done()
>  which does the flush.
> In which case, the same applies to the napi->rx_list, which similarly gets
>  handled in napi_complete_done().  So I don't think napi_busy_loop() needs a
>  gro_normal_list() adding to it(?)

I advise you to try busypoll then, with TCP_RR, with say 50 usec delay in /proc/sys/net/core/busy_read

Holding a small packet in the list up to the point we call busy_poll_stop()
will basically make busypoll non working anymore.

napi_complete_done() has special behavior when busy polling is active.


> 
> As a general rule, I think we need to gro_normal_list() in those places, and
>  only those places, that call napi_gro_flush().  But as I mentioned in the
>  patch 3/3 description, I'm still confused by the (few) drivers that call
>  napi_gro_flush() themselves.
> 
> -Ed
>