netdev - Re: [PATCH net] be2net: fix initial MAC setting

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 27 Jan 2017 16:21:14 +0100
From:   Ivan Vecera <cera@...a.cz>
To:     Sriharsha Basavapatna <sriharsha.basavapatna@...adcom.com>
Cc:     netdev@...r.kernel.org, Sathya Perla <sathya.perla@...adcom.com>,
        Ajit Khaparde <ajit.khaparde@...adcom.com>,
        Somnath Kotur <somnath.kotur@...adcom.com>
Subject: Re: [PATCH net] be2net: fix initial MAC setting

2017-01-27 13:38 GMT+01:00 Sriharsha Basavapatna
<sriharsha.basavapatna@...adcom.com>:
> Hi Ivan,
>
> This patch is a bit involved. We need some time to review and test
> this to make sure it works with other (non-BE3/VF - Skyhawk, Lancer)
> devices. Also, IIRC we shouldn't see this issue with the latest FW.
>
> Thanks,
> -Harsha

Btw. FW 11.1.215.0 is the latest available... As I described the
problem is reproducible after fresh boot.

1) After _FRESH_ boot we have enp1s0f0 interface that is BE3 PF with
SR-IOV enabled
[root@...02 ~]# uname -r
4.10.0-rc4-00189-ga47b70e
[root@...02 ~]# ethtool -i enp1s0f0 | egrep '(firm|bus)'
firmware-version: 11.1.215.0
bus-info: 0000:01:00.0
[root@...02 ~]# lspci -vv -s 0000:01:00.0 | egrep '(Name:|VFs)'
               Product Name: OCe11102-FM 2P 10GbE Tomcat Enterprise CNA, NIC PF
               Initial VFs: 32, Total VFs: 32, Number of VFs: 0,
Function Dependency Link: 00

2) Now we create 1 VF
[root@...02 ~]# ip link set enp1s0f1 up
be2net 0000:01:00.1 enp1s0f1: Link is Up
[root@...02 ~]# echo 1 > /sys/class/net/enp1s0f1/device/sriov_numvfs
...
be2net 0000:01:00.1: VF0 has FILTMGMT privilege
...
be2net 0000:01:08.0: Emulex OneConnect(be3): VF  port 1
be2net 0000:01:08.0 enp1s8f0: renamed from eth0

3) Now we have 1 privileged VF named enp1s8f0 that is down
[root@...02 ~]# ip link show dev enp1s8f0
12: enp1s8f0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop portid
0200000000000000000000333630384254 state DOWN mode DEFAULT qlen 1000
   link/ether 00:00:c9:9c:5c:60 brd ff:ff:ff:ff:ff:ff

4) Now we set the interface up (this fails)
[root@...02 ~]# ip link set enp1s8f0 up
be2net 0000:01:08.0: opcode 59-1 failed:status 1-76
RTNETLINK answers: Input/output error

5) Delete VF and create 2 new VFs
[root@...02 ~]# echo 0 > /sys/class/net/enp1s0f1/device/sriov_numvfs
...
[root@...02 ~]# echo 2 > /sys/class/net/enp1s0f1/device/sriov_numvfs
...
be2net 0000:01:08.0 enp1s8f0: renamed from eth0
...
be2net 0000:01:08.1 enp1s8f1: renamed from eth0

6) We have 2 privileged VFs named enp1s8f0 and enp1s8f1, both down
[root@...02 ~]# ip link show dev enp1s8f0
13: enp1s8f0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop portid
0200000000000000000000333630384254 state DOWN mode DEFAULT qlen 1000
   link/ether 00:00:c9:9c:5c:60 brd ff:ff:ff:ff:ff:ff
[root@...02 ~]# ip link show dev enp1s8f1
14: enp1s8f1: <BROADCAST,MULTICAST> mtu 1500 qdisc noop portid
0200000000000000000000333630384254 state DOWN mode DEFAULT qlen 1000
   link/ether 00:00:c9:9c:5c:61 brd ff:ff:ff:ff:ff:ff

7) Now set them up
[root@...02 ~]# ip link set enp1s8f0 up
be2net 0000:01:08.0 enp1s8f0: Link is Up
[root@...02 ~]# ip link set enp1s8f1 up
be2net 0000:01:08.1: opcode 59-1 failed:status 1-76
RTNETLINK answers: Input/output error

As you can see the re-used interface enp1s8f0 is OK but not-yet-used
interface enp1s8f1 fails.

8) Try to set enp1s8f1 up again
[root@...02 ~]# ip link set enp1s8f1 up
be2net 0000:01:08.1 enp1s8f1: Link is Up

Now it is OK... This is because be_close() is called when
be_enable_if_filters() fails. be_close() calls be_disable_if_filters()
and there the MAC pre-programmed by PF is deleted by be_dev_mac_del().
So when you try to set the interface up again then the pre-programmed
MAC is away and be_dev_mac_add() succeeds.

Regards,
Ivan