netdev - Re: [PATCH 1/2 v2] net: Add function to get SW rxhash

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: <CA+mtBx94yRWRNkfcbkOmFhMcspdyb-Q=jn0=dkAVTZrj-XFCDQ@mail.gmail.com>
Date:	Tue, 24 Sep 2013 15:16:36 -0700
From:	Tom Herbert <therbert@...gle.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	David Miller <davem@...emloft.net>,
	Linux Netdev List <netdev@...r.kernel.org>
Subject: Re: [PATCH 1/2 v2] net: Add function to get SW rxhash

On Tue, Sep 24, 2013 at 2:30 PM, Eric Dumazet <eric.dumazet@...il.com> wrote:
> On Tue, 2013-09-24 at 13:42 -0700, Tom Herbert wrote:
>> Some uses of skb_get_rxhash expect that the function will return
>> a consistent value whether it is called on RX or TX paths. On RX
>> path, we will use the rxhash if provided by the NIC, so this
>> would not normally be the same result computed in TX path would be
>> a software calculation.
>>
>> This patch adds skb_get_sw_rxhash to explicitly request a hash
>> calculated by the stack, disregarding the hash provided by NIC.
>
> As I said, I think this is overhead in network fast path.
>
The overhead is setting one in additional bit in skb_get_rxhash.
Alternatively, we could just take out the symmetric hashing code in
get_rps_code which is two or three conditionals and have tun do its
own hash as you mention.  Since skb_get_rxhash tries to use HW value
anyway, it's pretty non deterministic as to how often we'd set the
symmetric hash, much less be able advantage of the symmetric property.
 If it's a big win for conntrack to process packets on same CPUs for
both directions, we should program the RFS table to accomplish that.

> This can be done without adding a new skb field.
>
> I suspect tun should not use rxhash but a proper hash, as conntrack or
> tcp...
>
>
>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html