netdev - Re: [Intel-wired-lan] [net-next PATCH v3 1/3] e1000: track BQL bytes regardless of skb or not

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: <CALx6S364cmWz-6trzyX5pdhk_G_AC4XNoxCga13=_6uoWb5CTg@mail.gmail.com>
Date:   Mon, 12 Sep 2016 21:25:43 -0700
From:   Tom Herbert <tom@...bertland.com>
To:     Alexander Duyck <alexander.duyck@...il.com>
Cc:     John Fastabend <john.fastabend@...il.com>,
        Brenden Blanco <bblanco@...mgrid.com>,
        Alexei Starovoitov <alexei.starovoitov@...il.com>,
        Jeff Kirsher <jeffrey.t.kirsher@...el.com>,
        Jesper Dangaard Brouer <brouer@...hat.com>,
        David Miller <davem@...emloft.net>,
        Cong Wang <xiyou.wangcong@...il.com>,
        intel-wired-lan <intel-wired-lan@...ts.osuosl.org>,
        William Tu <u9012063@...il.com>,
        Netdev <netdev@...r.kernel.org>
Subject: Re: [Intel-wired-lan] [net-next PATCH v3 1/3] e1000: track BQL bytes
 regardless of skb or not

On Mon, Sep 12, 2016 at 8:00 PM, Alexander Duyck
<alexander.duyck@...il.com> wrote:
> On Mon, Sep 12, 2016 at 3:13 PM, John Fastabend
> <john.fastabend@...il.com> wrote:
>> The BQL API does not reference the sk_buff nor does the driver need to
>> reference the sk_buff to calculate the length of a transmitted frame.
>> This patch removes an sk_buff reference from the xmit irq path and
>> also allows packets sent from XDP to use BQL.
>>
>> Signed-off-by: John Fastabend <john.r.fastabend@...el.com>
>> ---
>>  drivers/net/ethernet/intel/e1000/e1000_main.c |    7 ++-----
>>  1 file changed, 2 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/intel/e1000/e1000_main.c b/drivers/net/ethernet/intel/e1000/e1000_main.c
>> index f42129d..62a7f8d 100644
>> --- a/drivers/net/ethernet/intel/e1000/e1000_main.c
>> +++ b/drivers/net/ethernet/intel/e1000/e1000_main.c
>> @@ -3882,11 +3882,8 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter,
>>                         if (cleaned) {
>>                                 total_tx_packets += buffer_info->segs;
>>                                 total_tx_bytes += buffer_info->bytecount;
>> -                               if (buffer_info->skb) {
>> -                                       bytes_compl += buffer_info->skb->len;
>> -                                       pkts_compl++;
>> -                               }
>> -
>> +                               bytes_compl += buffer_info->length;
>> +                               pkts_compl++;
>>                         }
>>                         e1000_unmap_and_free_tx_resource(adapter, buffer_info);
>>                         tx_desc->upper.data = 0;
>
> Actually it might be worth looking into why we have two different
> stats for tracking bytecount and segs.  From what I can tell the
> pkts_compl value is never actually used.  The function doesn't even
> use the value so it is just wasted cycles.  And as far as the bytes go

Transmit flow steering which I posted and is pending on some testing
uses the pkt count BQL to track inflight packets.

> the accounting would be more accurate if you were to use bytecount
> instead of buffer_info->skb->len.  You would just need to update the
> xmit function to use that on the other side so that they match.
>
> - Alex