linux-kernel - RE: [PATCH net-next,1/3] hv_netvsc: Use per socket hash when available

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] [day] [month] [year] [list]

Date:   Wed, 12 Apr 2017 16:24:36 +0000
From:   Haiyang Zhang <haiyangz@...rosoft.com>
To:     David Miller <davem@...emloft.net>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        KY Srinivasan <kys@...rosoft.com>,
        "olaf@...fle.de" <olaf@...fle.de>,
        "vkuznets@...hat.com" <vkuznets@...hat.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH net-next,1/3] hv_netvsc: Use per socket hash when
 available



> -----Original Message-----
> From: David Miller [mailto:davem@...emloft.net]
> Sent: Tuesday, April 11, 2017 10:13 PM
> To: Haiyang Zhang <haiyangz@...rosoft.com>; Haiyang Zhang
> <haiyangz@...rosoft.com>
> Cc: netdev@...r.kernel.org; KY Srinivasan <kys@...rosoft.com>;
> olaf@...fle.de; vkuznets@...hat.com; linux-kernel@...r.kernel.org
> Subject: Re: [PATCH net-next,1/3] hv_netvsc: Use per socket hash when
> available
> 
> From: Haiyang Zhang <haiyangz@...hange.microsoft.com>
> Date: Sat,  8 Apr 2017 17:53:59 -0700
> 
> > diff --git a/drivers/net/hyperv/netvsc_drv.c
> b/drivers/net/hyperv/netvsc_drv.c
> > index f24c289..0a129cb 100644
> > --- a/drivers/net/hyperv/netvsc_drv.c
> > +++ b/drivers/net/hyperv/netvsc_drv.c
> > @@ -211,9 +211,14 @@ static u16 netvsc_select_queue(struct net_device
> *ndev, struct sk_buff *skb,
> >  	int q_idx = sk_tx_queue_get(sk);
> >
> >  	if (q_idx < 0 || skb->ooo_okay || q_idx >= num_tx_queues) {
> > -		u16 hash = __skb_tx_hash(ndev, skb, VRSS_SEND_TAB_SIZE);
> > +		u16 hash;
> >  		int new_idx;
> >
> > +		if (sk)
> > +			skb_set_hash_from_sk(skb, sk);
> > +
> > +		hash = __skb_tx_hash(ndev, skb, VRSS_SEND_TAB_SIZE);
> 
> Please do not do this.
> 
> TCP performs this operation for you for every pack it emits.
> 
> And also every socket family that uses skb_set_owner_w() either
> directly or indirectly gets this done as well.
> 
> I do not want to see drivers start to get peppered with calls to this
> thing.
> 
> Explain the case which is missing that matters, and we can address
> that instead.
> 
> Thanks.

Thanks for pointing this out. I did some tests, the skb->hash is indeed
set to the sk->sk_txhash by upper layer. I will remove this patch, and
re-submit other patches.

Thanks,
- Haiyang