I have a couple of V280r servers that I have inherited which both have
a load of Ethernet cables connected to a couple of 4-port GigaSwift
Ethernet cards in PCI slots 2 & 3. One server has a SCSI card in PCI
slot 1 with nothing in slot 4 whilst the other has SCSI cards in both
slots 1 & slot 4.

I can see the various ce? interfaces all plumbed via ifconfig -a but
the documentation that I have inherited is extremely poor and I'm
trying to match up the ce? interfaces with specific ports on cards in
specific slots.

I'm working under the assumption that with 2 cards with 4 ports each,
the ce interfaces are grouped into ce0, ce1, ce2 & ce3 on one card and
ce4, ce5, ce6 & ce7 on the other card. Is this a valid assumption to be
making? The only thing is that I don't know which ports are from which
card.

I know I can scan the /etc/path_to_inst file to locate the ports but I
don't have the documentation that came with the V280r to help me
identify/interpret which PCI slot address is associated with the
information in there.

My lowest number ce interface is ce0 whilst the highest is ce7. This is
why I'm assuming 0-3 & 4-7 across the two cards. I'm curious as to why
this numbering convention is in place given that the GigaSwift cards
are in slots 2 & 3. I might have expected ce0-ce3 to be associated with
PCI slot 0. Does this imply that the port number (ce?) is NOT fixed
based on the PCI slot? Why does the numbering start at 0 when the first
card is not in slot 0? If the numbering is dynamic, what would happen,
for example, if a new 4-port GigaSwift card were to be installed in
slot 0, for example? Would the existing numbers all shift?

- CDM

cdmonline@mac.com wrote:
> I can see the various ce? interfaces all plumbed via ifconfig -a but
> the documentation that I have inherited is extremely poor and I'm
> trying to match up the ce? interfaces with specific ports on cards in
> specific slots.
Possibly the cards are labeled with the MAC address. Then you can
reverse map the card with the output from ifconfig -a

Probably the card has four consecutive MAC addresses so the label on de
card will show the first.

If there no MAC address label on the card, try matching the last digits
of the serial number.

If that failes, connect the card with _one_ connection to a network and
try a DHCP client. That will give you a functioning IP address on the
connected interface. Then label it as such.

> I'm working under the assumption that with 2 cards with 4 ports each,
> the ce interfaces are grouped into ce0, ce1, ce2 & ce3 on one card and
> ce4, ce5, ce6 & ce7 on the other card. Is this a valid assumption to be
> making? The only thing is that I don't know which ports are from which
> card.

I'm not sure about Solaris (I guess that's what you're using) but in (at
least RedHat) Linux it's possible to change the device names after
detection.

> I know I can scan the /etc/path_to_inst file to locate the ports but I
> don't have the documentation that came with the V280r to help me
> identify/interpret which PCI slot address is associated with the
> information in there.

> My lowest number ce interface is ce0 whilst the highest is ce7. This is
> why I'm assuming 0-3 & 4-7 across the two cards. I'm curious as to why
> this numbering convention is in place given that the GigaSwift cards
> are in slots 2 & 3. I might have expected ce0-ce3 to be associated with
> PCI slot 0. Does this imply that the port number (ce?) is NOT fixed
> based on the PCI slot?

I guess ce0 will be the first ce device detected.

> Why does the numbering start at 0 when the first
> card is not in slot 0? If the numbering is dynamic, what would happen,
> for example, if a new 4-port GigaSwift card were to be installed in
> slot 0, for example? Would the existing numbers all shift?

Probably.

> - CDM

--
Alex.

In article <1120380211.714339.196350@g14g2000cwa.googlegroups .com>,
<cdmonline@mac.com> wrote:

>PCI slot 0. Does this imply that the port number (ce?) is NOT fixed
>based on the PCI slot? Why does the numbering start at 0 when the first
>card is not in slot 0? If the numbering is dynamic, what would happen,
>for example, if a new 4-port GigaSwift card were to be installed in
>slot 0, for example? Would the existing numbers all shift?

The number are assigend upon installation of the cards and will
never change. You can populate any PCI slot with a new card and
it will get higher ce? numbers. This is one of Solaris major
features.

Ie a new card will not in any way affect the current numbering.
The numbers might change following a new install from scratch.

/wfr
Fredrik

--
Fredrik Lundholm
dol @ ce.chalmers.se

This is a Sun server sunning Solaris 9, the cards are not labelled with
MAC addresses and I don't have the luxury of being able to remove
network cables for testing.

cdmonline@mac.com writes:

> I have a couple of V280r servers that I have inherited which both have
> a load of Ethernet cables connected to a couple of 4-port GigaSwift
> Ethernet cards in PCI slots 2 & 3. One server has a SCSI card in PCI
> slot 1 with nothing in slot 4 whilst the other has SCSI cards in both
> slots 1 & slot 4.
>
> I can see the various ce? interfaces all plumbed via ifconfig -a but
> the documentation that I have inherited is extremely poor and I'm
> trying to match up the ce? interfaces with specific ports on cards in
> specific slots.
>
> I'm working under the assumption that with 2 cards with 4 ports each,
> the ce interfaces are grouped into ce0, ce1, ce2 & ce3 on one card and
> ce4, ce5, ce6 & ce7 on the other card. Is this a valid assumption to be
> making? The only thing is that I don't know which ports are from which
> card.
>
> I know I can scan the /etc/path_to_inst file to locate the ports but I
> don't have the documentation that came with the V280r to help me
> identify/interpret which PCI slot address is associated with the
> information in there.
>

From Sun InfoDoc 21216:

PCI Slot 1 /pci@8,600000/pci@1,* 33/66mhz-32/64bit-3.3V
PCI Slot 2 /pci@8,700000/pci@3,* 33mhz-32/64bit-5V
PCI Slot 3 /pci@8,700000/pci@2,* 33mhz-32/64bit-5V
PCI Slot 4 /pci@8,700000/pci@1,* 33mhz-32/64bit-5V

Cheers,

Paul
--
Paul Robertson
probertson@unixway.com

cdmonline@mac.com wrote:
> I have a couple of V280r servers that I have inherited which both have

Does the 280R have a 'V' in front? I'm not familiar with that.

> I can see the various ce? interfaces all plumbed via ifconfig -a but
> the documentation that I have inherited is extremely poor and I'm
> trying to match up the ce? interfaces with specific ports on cards in
> specific slots.

Then go to sunsolve and try to find the path mapping. It used to be
usually found in the specifics of the motherboard for your model

> I know I can scan the /etc/path_to_inst file to locate the ports but I
> don't have the documentation that came with the V280r to help me
> identify/interpret which PCI slot address is associated with the
> information in there.

If you have access to the system handbook, it should be there in the
documentation for the System Board.

> My lowest number ce interface is ce0 whilst the highest is ce7. This is
> why I'm assuming 0-3 & 4-7 across the two cards. I'm curious as to why
> this numbering convention is in place given that the GigaSwift cards
> are in slots 2 & 3. I might have expected ce0-ce3 to be associated with
> PCI slot 0. Does this imply that the port number (ce?) is NOT fixed
> based on the PCI slot?

Correct. It's not based (directly) on the PCI slot.

> Why does the numbering start at 0 when the first
> card is not in slot 0? If the numbering is dynamic, what would happen,
> for example, if a new 4-port GigaSwift card were to be installed in
> slot 0, for example? Would the existing numbers all shift?

The numbering is dynamically assigned, but static after assignment. It
isbased on the first time that the card is "seen" by the system. This
may depend on installation order or probe order.

Since the numbers are assigned in /etc/path_to_inst, adding or removing
a card should allow the existing ports to retain their identification.

--
Darren Dunham ddunham@taos.com
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >