From: Brendan Gregg on
G'Day Folks,

Please may I suggest the following to Sun,

could the NIC kstat values be more consistant across interfaces.

....

As some background, lets try ndd for these two Solaris 10 3/05 servers
to check duplex,

# ndd /dev/hme link_mode
1
# ndd /dev/bge link_mode
operation failed: Invalid argument

Ok, I guess we shouldn't be using ndd. That's ok, I prefer kstat anyway.
Lets try kstat,

# kstat ::hme0:link_duplex
module: hme instance: 0
name: hme0 class: net
link_duplex 2
# kstat ::bge0:duplex
module: bge instance: 0
name: bge0 class: net
duplex full

The kstat value is not consistant...

thanks,

Brendan Gregg

[Sydney, Australia]


From: barts on

Noted.

- Bart

From: Chris Lawson on
Brendan Gregg wrote:
> G'Day Folks,
>
> Please may I suggest the following to Sun,
>
> could the NIC kstat values be more consistant across interfaces.
>
> ...
>
> As some background, lets try ndd for these two Solaris 10 3/05 servers
> to check duplex,
>
> # ndd /dev/hme link_mode
> 1
> # ndd /dev/bge link_mode
> operation failed: Invalid argument

try ndd /dev/bge0 link_mode

>
> Ok, I guess we shouldn't be using ndd. That's ok, I prefer kstat anyway.
> Lets try kstat,
>
> # kstat ::hme0:link_duplex
> module: hme instance: 0
> name: hme0 class: net
> link_duplex 2
> # kstat ::bge0:duplex
> module: bge instance: 0
> name: bge0 class: net
> duplex full
>
> The kstat value is not consistant...
>

It may not be consistant with old interfaces, but bge and dmfe give
meaningful information. i.e. full or half.
Seeing as the guy that wrote dmfe and bge isn't at Sun any more new
drivers may bounce back to just giving numeric values for duplexness.

--
Chris

--
Chris
From: barts on
Actually, there's a standard this stuff is supposed to follow now.
We'll just add the new standard kstats to the old drivers; that way
old and new software will work.

- Bart

From: Ben Fitzgerald on
On 8 Apr 2005 20:09:26 -0700, barts(a)smaalders.net <barts(a)smaalders.net> wrote:
> Actually, there's a standard this stuff is supposed to follow now.
> We'll just add the new standard kstats to the old drivers; that way
> old and new software will work.

I'll be pleased if that happens. I have to say that this inconsistency
is amazing within one company. It's the kind of instance you would
expect solaris fans to cite about an open source product to show that
industry can't rely on projects that are not centrally controlled.

ben.

--
Registered Linux user number 339435
 |  Next  |  Last
Pages: 1 2
Prev: Solaris x86 (v10.0)
Next: Ipsec in solaris 9