netdev - Re: [net-next 07/12] ixgbe: add support for modifying UDP RSS flow hash options

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:	Thu, 23 Jun 2011 20:57:51 -0700
From:	Jeff Kirsher <jeffrey.t.kirsher@...el.com>
To:	Ben Hutchings <bhutchings@...arflare.com>
Cc:	davem@...emloft.net, Alexander Duyck <alexander.h.duyck@...el.com>,
	netdev@...r.kernel.org, gospo@...hat.com
Subject: Re: [net-next 07/12] ixgbe: add support for modifying UDP RSS flow
 hash options

On Wed, Jun 22, 2011 at 19:25, Ben Hutchings <bhutchings@...arflare.com> wrote:
> On Wed, 2011-06-22 at 18:44 -0700, Jeff Kirsher wrote:
>> From: Alexander Duyck <alexander.h.duyck@...el.com>
>>
>> This patch adds the ability to add or remove the UDP source and destination
>> ports from the flow hash generated for RSS.  Currently the UDP flow hash is
>> always disabled.  By adding support for enabling the UDP flow hash we are
>> now providing the option of generating an RSS hash based on the UDP source
>> and destination port numbers.
> [...]
>> --- a/drivers/net/ixgbe/ixgbe_ethtool.c
>> +++ b/drivers/net/ixgbe/ixgbe_ethtool.c
> [...]
>> @@ -2681,6 +2726,110 @@ static int ixgbe_del_ethtool_fdir_entry(struct ixgbe_adapter *adapter,
>>       return err;
>>  }
>>
>> +#define UDP_RSS_FLAGS (IXGBE_FLAG2_RSS_FIELD_IPV4_UDP | \
>> +                    IXGBE_FLAG2_RSS_FIELD_IPV6_UDP)
>> +static int ixgbe_set_rss_hash_opt(struct ixgbe_adapter *adapter,
>> +                               struct ethtool_rxnfc *nfc)
>> +{
>> +     u32 flags2 = adapter->flags2;
>> +
>> +     /*
>> +      * RSS does not support anything other than hashing
>> +      * to queues on src and dst IPs and ports
>> +      */
>> +     if (nfc->data & ~(RXH_IP_SRC | RXH_IP_DST |
>> +                       RXH_L4_B_0_1 | RXH_L4_B_2_3))
>> +             return -EINVAL;
>> +
>> +     switch (nfc->flow_type) {
>> +     case TCP_V4_FLOW:
>> +     case TCP_V6_FLOW:
>> +             if (!(nfc->data & RXH_IP_SRC) ||
>> +                 !(nfc->data & RXH_IP_DST) ||
>> +                 !(nfc->data & RXH_L4_B_0_1) ||
>> +                 !(nfc->data & RXH_L4_B_2_3))
>> +                     return -EINVAL;
>> +             break;
>
> This can be written as a simple equality test.
>
>> +     case UDP_V4_FLOW:
>> +             if (!(nfc->data & RXH_IP_SRC) ||
>> +                 !(nfc->data & RXH_IP_DST))
>> +                     return -EINVAL;
>> +             switch (nfc->data & (RXH_L4_B_0_1 | RXH_L4_B_2_3)) {
>
> You could just as well include all flags in the switch.
>
> [...]
>> +     case AH_ESP_V4_FLOW:
>> +     case AH_V4_FLOW:
>> +     case ESP_V4_FLOW:
>> +     case SCTP_V4_FLOW:
>> +     case AH_ESP_V6_FLOW:
>> +     case AH_V6_FLOW:
>> +     case ESP_V6_FLOW:
>> +     case SCTP_V6_FLOW:
>> +             if (!(nfc->data & RXH_IP_SRC) ||
>> +                 !(nfc->data & RXH_IP_DST))
>> +                     return -EINVAL;
>> +               break;
> [...]
>
> This should be written as an equality test, otherwise you will ignore a
> request to include port numbers in the hash for these flow classes.
>
> Ben.
>
>

Alex is on vacation currently, so I will just pull this patch for now
to give Alex time to respond when he returns next week.

-- 
Cheers,
Jeff
--
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