DEQNA info

[Mouse]

sys/dev/qbus/if_qe.c says, in part, even in -current according to
cvsweb,

/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/

I have a digital copy of EK-DEQNA-UG-001 from bitsavers which appears
to answer this question (though the wording of the first sentence is a
little confusing; I believe there is a "count" missing after the first
"byte"), which I quote here in case anyone wants to pick it up and do
something with it. (I will, but not in any NetBSD version I expect
NetBSD to care about.)

4.3.3.2 Operating Condition Set-Up - When the byte exceeds 128 (177
octal), bits 06:00 of the byte count are used as control parameters,
and are interpreted as follows.

Bit Definition

00 All Multicast. When set, enables the DEQNA to recognize
any multicast address.

01 Promiscuous. When set, enables the DEQNA to recognize any
destination address.

03:02 LED Value. These two bits cause the three LEDs on the
DEQNA to turn off in the following sequence (all the LEDs
are turned on at power-up and when boot/diagnostic code
is loaded from the BD ROM).

00 No effect.

01 Turns off the first LED, indicating the DEQNA
self-test is running in the host.

10 Turns off the second LED, indicating all
internal loopbacks appear to function
correctly.

11 Turns off the third LED, indicating
externalloopback appears to function correctly.

06:04 Sanity Timer Time-Out Value. [...[

It is not clear to me whether this is "of the byte count value in the
descriptor" or "of the actual count of bytes according to the
descriptor", since the buffer descriptor contains the *negated* byte
count. If I had a MicroVAX-II in usable shape, I'd experiment with
this to find out, but, at the moment, I don't - they're still buried
from my move.

/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML ***@rodents-montreal.org
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

[Mark Pizzolato - Info Comm]

sys/dev/qbus/if_qe.c says, in part, even in -current according to cvsweb,
/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/
I have a digital copy of EK-DEQNA-UG-001 from bitsavers which appears to
answer this question (though the wording of the first sentence is a little
confusing; I believe there is a "count" missing after the first "byte"), which I
quote here in case anyone wants to pick it up and do something with it. (I
will, but not in any NetBSD version I expect NetBSD to care about.)
4.3.3.2 Operating Condition Set-Up - When the byte exceeds 128 (177 octal),
bits 06:00 of the byte count are used as control parameters, and are
interpreted as follows.
Bit Definition
00 All Multicast. When set, enables the DEQNA to recognize
any multicast address.
01 Promiscuous. When set, enables the DEQNA to recognize any
destination address.
03:02 LED Value. These two bits cause the three LEDs on the
DEQNA to turn off in the following sequence (all the LEDs
are turned on at power-up and when boot/diagnostic code
is loaded from the BD ROM).
00 No effect.
01 Turns off the first LED, indicating the DEQNA
self-test is running in the host.
10 Turns off the second LED, indicating all
internal loopbacks appear to function
correctly.
11 Turns off the third LED, indicating
externalloopback appears to function correctly.
06:04 Sanity Timer Time-Out Value. [...[
It is not clear to me whether this is "of the byte count value in the descriptor"
or "of the actual count of bytes according to the descriptor", since the buffer
descriptor contains the *negated* byte count. If I had a MicroVAX-II in
usable shape, I'd experiment with this to find out, but, at the moment, I
don't - they're still buried from my move.

There was some ambiguity in this domain when I worked out the simh simulation of the DEQNA/DELQA/DELQA-T devices.

Here are the comments from what I did there:

/*
Under VMS the setup packet that is passed to turn promiscuous
off after it has been on doesn't seem to follow the rules documented
in both the DEQNA and DELQA manuals.
These rules seem to say that setup packets less than 128 should only
modify the address filter set and probably not the All-Multicast and
Promiscuous modes, however, VMS V5-5 and V7.3 seem to send a 127 byte
packet to turn this functionality off. I'm not sure how real hardware
behaves in this case, since the only consequence is extra interrupt
load. To realize and retain the benefits of the newly added BPF
functionality in sim_ether, I've modified the logic implemented here
to disable Promiscuous mode when a "small" setup packet is processed.
I'm deliberately not modifying the All-Multicast mode the same way
since I don't have an observable case of its behavior. These two
different modes come from very different usage situations:
1) Promiscuous mode is usually entered for relatively short periods
of time due to the needs of a specific application program which
is doing some sort of management/monitoring function (i.e. tcpdump)
2) All-Multicast mode is only entered by the OS Kernel Port Driver
when it happens to have clients (usually network stacks or service
programs) which as a group need to listen to more multicast ethernet
addresses than the 12 (or so) which the hardware supports directly.
so, I believe that the All-Multicast mode, is first rarely used, and if
it ever is used, once set, it will probably be set either forever or for
long periods of time, and the additional interrupt processing load to
deal with the distinctly lower multicast traffic set is clearly lower than
that of the promiscuous mode.
*/

- Mark

[Johnny Billquist]

Post by Mouse
sys/dev/qbus/if_qe.c says, in part, even in -current according to
cvsweb,
/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/
I have a digital copy of EK-DEQNA-UG-001 from bitsavers which appears
to answer this question (though the wording of the first sentence is a
little confusing; I believe there is a "count" missing after the first
"byte"), which I quote here in case anyone wants to pick it up and do
something with it. (I will, but not in any NetBSD version I expect
NetBSD to care about.)

[...]

I remember going down this rabbithole many years ago.
I don't remember if it was/is a documentation error, a firmware bug, or
just plain impossible to actually do this in reality, but anyway, Ragge
is well aware of what the documentation says, and as far as I can
recall, we have not been able to actually turn on ALLMULTI.

Johnny

[Anders Magnusson]

Post by Johnny Billquist

Post by Mouse
sys/dev/qbus/if_qe.c says, in part, even in -current according to
cvsweb,
/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/
I have a digital copy of EK-DEQNA-UG-001 from bitsavers which appears
to answer this question (though the wording of the first sentence is a
little confusing; I believe there is a "count" missing after the first
"byte"), which I quote here in case anyone wants to pick it up and do
something with it. (I will, but not in any NetBSD version I expect
NetBSD to care about.)

[...]
I remember going down this rabbithole many years ago.
I don't remember if it was/is a documentation error, a firmware bug,
or just plain impossible to actually do this in reality, but anyway,
Ragge is well aware of what the documentation says, and as far as I
can recall, we have not been able to actually turn on ALLMULTI.

Yetch, this was many years ago, I think I wrote this driver when I lived
in Boston, so I do not remember the specifics.
I had the same docs that are on bitsavers though.

Anyway, the DEQNA is a quite buggy animal, and it may be that I did not
get ALLMULTI to work as the docs said. The DELQA works much better, but
I did not have one of those at hand then.

I probably did not try too much with it, more than 12 addresses could as
well turn on promiscuous mode.

-- Ragge

[Mouse]

Post by Mouse
/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/
I have a digital copy of EK-DEQNA-UG-001 from bitsavers which
appears to answer this question [...]

[...], but anyway, Ragge is well aware of what the documentation
says, [...]

[T]his was many years ago, [...] I do not remember the specifics.
I had the same docs that are on bitsavers though.

Okay. My apologies for doubting!

I may dig out a DEQNA and experiment more myself at some point; I can
make my simulation match what the ROM code and the kernel are
expecting, but I'd like to at least know when I'm differing from the
real thing. (I've actually considered running the simulator and the
real thing in parallel to detect differences, but that is probably
hopeless because of exact timing issues and the like.)

Anyway, the DEQNA is a quite buggy animal, [...]

I don't think I'm going to try to be fully bug-compatible...unless I
run into some code that depends on such a bug. :-)

/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML ***@rodents-montreal.org
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

[Paul Koning]

Post by Johnny Billquist

Post by Mouse
sys/dev/qbus/if_qe.c says, in part, even in -current according to
cvsweb,
/*
* How is the DEQNA turned in ALLMULTI mode???
* Until someone tells me, fall back to PROMISC when more than
* 12 ethernet addresses.
*/
I have a digital copy of EK-DEQNA-UG-001 from bitsavers which appears
to answer this question (though the wording of the first sentence is a
little confusing; I believe there is a "count" missing after the first
"byte"), which I quote here in case anyone wants to pick it up and do
something with it. (I will, but not in any NetBSD version I expect
NetBSD to care about.)

[...]
I remember going down this rabbithole many years ago.
I don't remember if it was/is a documentation error, a firmware bug, or just plain impossible to actually do this in reality, but anyway, Ragge is well aware of what the documentation says, and as far as I can recall, we have not been able to actually turn on ALLMULTI.
Johnny

FWIW, the RSTS device driver for QNA checks for more than 14 multicast addresses being enabled, and if so, sets the all-multicast flag. I haven’t tested it but this suggests that it must have worked for someone. (14 because the address filter has 16 entries, and RSTS mistakenly unconditionally enables broadcast.)

paul