headers
George Harvey | 7 May 2011 01:13

Trap 0x34 panic on 5.1 on Blade 100

Hi,

I've recently acquired a Blade 100 and, as usual for my Sparc boxes,
installed NetBSD. I run 5.0.1 on an Ultra 5 with no problems so I was
somewhat surprised when 5.1 on the Blade 100 kept crashing and
rebooting. This appears to be related to network activity, xterm and
FTP both cause almost immediate crashes while NFS lasts a bit longer.
After the reboot, dmesg shows messages like this:

trap type 0x34: cpu 0, pc=1316c60 npc=1316c64 pstate=44800006<PRIV,IE>
panic: mem address not aligned
Begin traceback...
End traceback...
Frame pointer is at 0xe0016c51
Call traceback:
13e9ea8(11, 5, 0, 0, e0017658, 100a054, e0016d11) fp = e0016d11
12d1764(104, 0, fffe, 1637889, 12d14d0, 7, e0016dd1) fp = e0016dd1
13f1334(1631bb0, e00178d0, e001775e, 59, 1c05d18, 7, e0016ea1) fp =
e0016ea1 
1008b0c(e00178d0, 34, 1316c60, 44800006, 2e05d70, 2e05d10,
e0017021) fp = e0017021 
13ed8e8(2e57488, c0be000, 6fd762f8, 2f8, 2000, c6fc000, e0017201) fp =
e0017201 
13170a0(c0c1002, c, e0017bbc, 10, 5, 0, e00172c1) fp = e00172c1 
13152dc (2e46180, c0c1002, 0, 0, 5, 0, e00173d1) fp = e00173d1
1175ee0 (2ed4e00, 100a4d0, c0c1002, 0, 0, c7880f0, e00174a1) fp =
e00174a1 
1176998 (c788000, 2d00010, 2, 1, 1ff00400000, 79, e0017561) fp =
e0017561 
1008f54(0, 0, e0017ed0, 110340, 1176914, 1, e0017621) fp = e0017621
(Continue reading)

Permalink | Reply | 10 comments |
headers
Martin Husemann | 7 May 2011 09:18
Picon

Re: Trap 0x34 panic on 5.1 on Blade 100

On Sat, May 07, 2011 at 12:13:02AM +0100, George Harvey wrote:
> Are there any kernel debug options I could set to get a more readable
> traceback?

If this is not a production server, you probably want ddb.onpanic = 1
in /etc/sysctl.conf.

And (default advice): please try a newer netbsd-5 (see releng.netbsd.org
for the daily builds of the NetBSD-5 branch).

Martin
Permalink | Reply | 9 comments |
headers
Matthias Scheler | 7 May 2011 17:12
Picon
Favicon

NetBSD/sparc64 and GbE performance


	Hello,

what performance do people get with GbE cards in NetBSD/sparc64 systems?
I get about 200Mb/s with the wm(4) card in my ULTRA60. If I remember
correctly Solaris 10 achieved twice the throughput with a cas(4) card.
But I seriously doubt that the cas(4) hardware is faster or even as
fast as a wm(4) card.

	Kind regards

--

-- 
Matthias Scheler                                  http://zhadum.org.uk/
Permalink | Reply | 17 comments |
headers
John Nemeth | 7 May 2011 17:45
Picon
Favicon

Re: NetBSD/sparc64 and GbE performance

On Sep 27, 10:48am, Matthias Scheler wrote:
} 
} what performance do people get with GbE cards in NetBSD/sparc64 systems?
} I get about 200Mb/s with the wm(4) card in my ULTRA60. If I remember
} correctly Solaris 10 achieved twice the throughput with a cas(4) card.
} But I seriously doubt that the cas(4) hardware is faster or even as
} fast as a wm(4) card.

     Why?  Have you ever read the specs of cas(4)?  It's a rather
impressive chip.  Unfortunately, NetBSD doesn't take anything like full
advantage of it.  Heck, it doesn't even work on i386.  Now that I have
additional time, hopefully I'll remedy that in the not too distant
future.

}-- End of excerpt from Matthias Scheler
Permalink | Reply | 10 comments |
headers
Matthias Scheler | 7 May 2011 18:36
Picon
Favicon

Re: NetBSD/sparc64 and GbE performance

On Sat, May 07, 2011 at 08:45:34AM -0700, John Nemeth wrote:
> } what performance do people get with GbE cards in NetBSD/sparc64 systems?
> } I get about 200Mb/s with the wm(4) card in my ULTRA60. If I remember
> } correctly Solaris 10 achieved twice the throughput with a cas(4) card.
> } But I seriously doubt that the cas(4) hardware is faster or even as
> } fast as a wm(4) card.
> 
>      Why?  Have you ever read the specs of cas(4)?

No, but I used machines with those cards in the past and I remember
various problems.

> It's a rather impressive chip.

Is it?

Cassini:
http://ecx.images-amazon.com/images/I/51ULkpSVPTL.jpg

Intel PRO 1000 MT:
http://www.buyserverparts.com/item_images/thumb/pro-1000.jpg

I'm sorry but not impressed by the bulk of that thing. And it doesn't
even support things like IP header checksum offload.

	Kind regards

--

-- 
Matthias Scheler                                  http://zhadum.org.uk/
(Continue reading)

Permalink | Reply | 9 comments |
headers
George Harvey | 8 May 2011 00:45

Re: Trap 0x34 panic on 5.1 on Blade 100

On Sat, 7 May 2011 09:18:26 +0200
Martin Husemann <martin <at> duskware.de> wrote:

> On Sat, May 07, 2011 at 12:13:02AM +0100, George Harvey wrote:
> > Are there any kernel debug options I could set to get a more readable
> > traceback?
> 
> If this is not a production server, you probably want ddb.onpanic = 1
> in /etc/sysctl.conf.

Looks like it could be the gem driver, I switched to a 3Com Ethernet
card and it stopped crashing.

George
Permalink | Reply | 8 comments |
headers
John Nemeth | 8 May 2011 02:17
Picon
Favicon

Re: NetBSD/sparc64 and GbE performance

On Sep 27, 12:11pm, Matthias Scheler wrote:
} On Sat, May 07, 2011 at 08:45:34AM -0700, John Nemeth wrote:
} > } what performance do people get with GbE cards in NetBSD/sparc64 systems?
} > } I get about 200Mb/s with the wm(4) card in my ULTRA60. If I remember
} > } correctly Solaris 10 achieved twice the throughput with a cas(4) card.
} > } But I seriously doubt that the cas(4) hardware is faster or even as
} > } fast as a wm(4) card.
} > 
} >      Why?  Have you ever read the specs of cas(4)?
} 
} No, but I used machines with those cards in the past and I remember
} various problems.
} 
} > It's a rather impressive chip.
} 
} Is it?
} 
} Cassini:
} http://ecx.images-amazon.com/images/I/51ULkpSVPTL.jpg
} 
} Intel PRO 1000 MT:
} http://www.buyserverparts.com/item_images/thumb/pro-1000.jpg

     Look at those pictures a bit closer.  The Cassini is a 64-bit card
and the Intel is only a 32-bit card.

} I'm sorry but not impressed by the bulk of that thing. And it doesn't
} even support things like IP header checksum offload.

     I don't have the specsheet in front of me, but considering the
(Continue reading)

Permalink | Reply | 8 comments |
headers
Matthias Scheler | 8 May 2011 02:27
Picon
Favicon

Re: NetBSD/sparc64 and GbE performance

On Sat, May 07, 2011 at 05:17:54PM -0700, John Nemeth wrote:
> } Cassini:
> } http://ecx.images-amazon.com/images/I/51ULkpSVPTL.jpg
> } 
> } Intel PRO 1000 MT:
> } http://www.buyserverparts.com/item_images/thumb/pro-1000.jpg
> 
>      Look at those pictures a bit closer.  The Cassini is a 64-bit card
> and the Intel is only a 32-bit card.

So? The Cassini is still bulky even compared to 64bit Intel card.
And it consumes a lot of power and produces a lot of heat.

> Don't use the simplistic NetBSD driver (derived from OpenBSD) to judge it.

Solaris had a complicated (and highly problematic) driver for it.
And the card performed still far from impressive.

	Kind regards

--

-- 
Matthias Scheler                                  http://zhadum.org.uk/
Permalink | Reply | 7 comments |
headers
Chris Ross | 8 May 2011 02:39

Re: NetBSD/sparc64 and GbE performance


On May 7, 2011, at 11:12 , Matthias Scheler wrote:
> 	Hello,
> 
> what performance do people get with GbE cards in NetBSD/sparc64 systems?
> I get about 200Mb/s with the wm(4) card in my ULTRA60. If I remember
> correctly Solaris 10 achieved twice the throughput with a cas(4) card.
> But I seriously doubt that the cas(4) hardware is faster or even as
> fast as a wm(4) card.

  I have an sk/skc in my E420R, I'd be happy to report what I get.  But, using scp is not a good test because the slow
CPUs are worse on encryption than the network is on transfer.

  What tool(s) were/are you using to measure your 200/MB/sec?

           - Chris
Permalink | Reply | 3 comments |
headers
Dave McGuire | 8 May 2011 02:42

Re: NetBSD/sparc64 and GbE performance

On 5/7/11 8:27 PM, Matthias Scheler wrote:
>>       Look at those pictures a bit closer.  The Cassini is a 64-bit card
>> and the Intel is only a 32-bit card.
>
> So? The Cassini is still bulky even compared to 64bit Intel card.
> And it consumes a lot of power and produces a lot of heat.

   It is physically larger and therefore technologically inferior?

   WTF?

> Solaris had a complicated (and highly problematic) driver for it.
> And the card performed still far from impressive.

   I'm running a few of these in higher-end (V480-range) Sun machines 
running Solaris.  While I've not benchmarked them, I know I get better 
than 500Mbps out of them on those machines.

               -Dave

--

-- 
Dave McGuire
Port Charlotte, FL
Permalink | Reply | 5 comments |

Gmane