NetXen NX3031 failing test, reporting link down, not working

NetXen NX3031 failing test, reporting link down, not working

[Marc Haber]

Hi,

I have HP DL 120 G6 with two Quad-GE Cards based on NetXen NX3031 (rev
42) [4040:0100]. The boxes are running with a grml daily image with
kernel 2.6.36.1 for the time being, and the NetXens are giving me a
hard time.

This is the dmesg for two of the eight Interfaces (for brevity).

[    1.818833] QLogic/NetXen Network Driver v4.0.74
[    1.818881] netxen_nic 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[    1.818894] netxen_nic 0000:01:00.0: setting latency timer to 64
[    1.819116] netxen_nic 0000:01:00.0: 2MB memory map
[    7.709080] netxen_nic 0000:01:00.0: loading firmware from flash
[   12.866864] netxen_nic 0000:01:00.0: using 64-bit dma mask
[   12.867197] netxen_nic: Quad Gig LP Board S/N TI08BK3987  Chip rev 0x42
[   12.867200] netxen_nic 0000:01:00.0: firmware v4.0.530 [legacy]
[   12.867249] netxen_nic 0000:01:00.0: irq 41 for MSI/MSI-X
[   12.867254] netxen_nic 0000:01:00.0: irq 42 for MSI/MSI-X
[   12.867265] netxen_nic 0000:01:00.0: using msi-x interrupts
[   12.867446] netxen_nic 0000:01:00.0: eth2: GbE port initialized
[   12.867468] netxen_nic 0000:01:00.1: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[   12.867477] netxen_nic 0000:01:00.1: setting latency timer to 64
[   12.867688] netxen_nic 0000:01:00.1: 2MB memory map
[   12.867773] netxen_nic 0000:01:00.1: using 64-bit dma mask
[   12.868097] netxen_nic 0000:01:00.1: firmware v4.0.530 [legacy]
[   12.868120] netxen_nic 0000:01:00.1: irq 43 for MSI/MSI-X
[   12.868126] netxen_nic 0000:01:00.1: irq 44 for MSI/MSI-X
[   12.868136] netxen_nic 0000:01:00.1: using msi-x interrupts
[   12.868288] netxen_nic 0000:01:00.1: eth3: GbE port initialized

They work fine when I have two NetXens connected back-to-back (at
least I can ping the other side), but they do not work when one of the
NetXen Interfaces is connected to an HP ProCurve 2848.

The switch happily says the link is up, while ethtool says "Speed:
1000Mb/s, Duplex: Full, Link detected: no", and ethtool -t offline and
ethtool -t online both say "The test result is FAIL".

Do the netxens need Firmware to work and do not complain when the
firmware is not found? Any ideas how do debug this?

I am not sure whether this is appropriate for netdev@vger. If so,
please tell me and I'll re-send there. I'd appreciate any pointers to
any resources about the NetXen interfaces.

Greetings
Marc

-- 
-----------------------------------------------------------------------------
Marc Haber         | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany  |  lose things."    Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature |  How to make an American Quilt | Fax: *49 3221 2323190

Re: NetXen NX3031 failing test, reporting link down, not working

[Marc Haber]

On Mon, Dec 06, 2010 at 06:17:27PM +0100, Marc Haber wrote:
> I have HP DL 120 G6 with two Quad-GE Cards based on NetXen NX3031 (rev
> 42) [4040:0100]. The boxes are running with a grml daily image with
> kernel 2.6.36.1 for the time being, and the NetXens are giving me a
> hard time.
> 
> They work fine when I have two NetXens connected back-to-back (at
> least I can ping the other side), but they do not work when one of the
> NetXen Interfaces is connected to an HP ProCurve 2848.

Just for the record, the NetXens work fine even against a ProCurve
under Windows Server 2008 R2. This is thus a clear Linux issue.

I have filed bug #24792 in bugzilla. Is there anything more I can do
do pin down the issue (or to help pinning it down)?

Greetings
Marc

-- 
-----------------------------------------------------------------------------
Marc Haber         | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany  |  lose things."    Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature |  How to make an American Quilt | Fax: *49 3221 2323190

Re: NetXen NX3031 failing test, reporting link down, not working

[Marc Haber]

Hi,

On Mon, Dec 06, 2010 at 06:17:27PM +0100, Marc Haber wrote:
> I have HP DL 120 G6 with two Quad-GE Cards based on NetXen NX3031 (rev
> 42) [4040:0100]. The boxes are running with a grml daily image with
> kernel 2.6.36.1 for the time being, and the NetXens are giving me a
> hard time.

it's a pity that nobody cared to even point me into the right
direction. I am not aware of having done anything wrong.

Anyway, the issue could be solved by updating the on-card flashed
firmware from 4.0.530 to 4.0.539. With this firmware, everything seems
to be fine.

Maybe the Linux driver should be adapted to not fail silently if it
doesn't like the on-card firmware but instead log an error message so
that people know they need to update their firmware.

Greetings
Marc

-- 
-----------------------------------------------------------------------------
Marc Haber         | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany  |  lose things."    Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature |  How to make an American Quilt | Fax: *49 3221 2323190

Re: NetXen NX3031 failing test, reporting link down, not working

[Phil Turmel]

Hi Marc,

On 12/25/2010 11:36 AM, Marc Haber wrote:
> On Mon, Dec 06, 2010 at 06:17:27PM +0100, Marc Haber wrote:
>> I have HP DL 120 G6 with two Quad-GE Cards based on NetXen NX3031 (rev
>> 42) [4040:0100]. The boxes are running with a grml daily image with
>> kernel 2.6.36.1 for the time being, and the NetXens are giving me a
>> hard time.
> 
> it's a pity that nobody cared to even point me into the right
> direction. I am not aware of having done anything wrong.

Poor signal-to-noise ratio on lkml itself, so don't take it personally.  The net folk don't listen too closely, but akpm usually tosses these over the wall to 'em.  He seems to have missed this one, so I've added the most important CC.

> Anyway, the issue could be solved by updating the on-card flashed
> firmware from 4.0.530 to 4.0.539. With this firmware, everything seems
> to be fine.
> 
> Maybe the Linux driver should be adapted to not fail silently if it
> doesn't like the on-card firmware but instead log an error message so
> that people know they need to update their firmware.

I'm just a lurker, tester, and occasional bug reporter, so I can't speak to this.

Phil