Hi Johnathan:

Hmmm.... That's the problem with not having the hardware on hand to try
out.  Are the NICs built into the motherboard or are they separate
cards?  Any way that you could get the hardware to us to try out?

I don't see the root # prompt there.  Are you running these commands as
root?

After the slay io-pkt-v4-hc, do a pidin and make sure that it isn't
running.  I've occasionally seen cases where the driver faults during
shutdown and this results in the stack doing "funny things" (and you end
up having to "slay -s SIGKILL io-pkt-v4-hc" to make sure that it's gone.

Can you also post the output to "sloginfo" after running the commands.

In terms of the source for the driver, it's already been checked into
our source tree (the last post has the location).  This is the latest
source on the netbsd-4 branch which is where our stack originated from.
The head branch would be for the latest NetBSD incarnation which may
have compatibility problems with out source.

	Robert.



-----Original Message-----
From: Jonathan Mcdougall [mailto:community-noreply@qnx.com] 
Sent: Friday, February 06, 2009 2:07 PM
To: drivers-networking
Subject: Re: MTU on Broadcom BCM5787M

> x86 Driver is attached.  It's got basic sanity testing on it with the 
> card that I have in house, but I don't have a 5787, so you'll be our 
> test case for that card.

We replaced /lib/dll/devnp-bge.so with the one you sent. Restarting the
driver with

  $ slay io-pkt-v4-hc
  $ /sbin/io-pkt-v4 -dbge

slows the machine down to a crawl. After a few seconds, 

  Unable to init devnp-bge.so: No such device or address

is printed and, after a few more seconds, the whole machine hangs and
needs a hard reset.

Trying the same thing with the original bge driver gives the same
message but does not hang the machine.
  
We tried changing /etc/system/enum/devices by removing dev=1693 from the
tigon3 driver and added:

 device(pci, ven=$(PCI_VEND_BROADCOM), dev=1693)  #Broadcom 5787M
Gigabit
    tag(devn)
    append(legacy, ",nonet")
    requires($(IOPKT_CMD),)
    waitfor (/dev/socket)
    mount(-Tio-pkt "-opci=$(index),vid=0x$(ven),did=0x$(dev)"
/lib/dll/devnp-bge.so, )
    use(symbolic=netmgr)

However, that prevented the network card from being recognized after
restarting the computer. sloginfo gives:

  mount: Can't mount / (type io-pkt)
  mount: Possible reason: Unknown error

We get the same thing when running 

  $ mount -Tio-pkt "-opci=$(index),vid=0x$(ven),did=0x$(dev)"
/lib/dll/devnp-bge.so
  
directly.

As for me, I'm the programmer trying to recompile the driver. I'm having
issues (of course, I've got no experience in this), but I'm attaching
the files I modified just in case it might help you. I merged most of
the stuff from the head of netbsd in if_bge.c/h and some associated
files. Although it compiles and links fine, I'm getting some unresolved
names, obviously because I need to recompile and install much more stuff
than I initially thought. Anyways, I'll keep trying.

> On another note, it looks like we've potentially identified the link 
> problem with the tigon3 driver as well, so that may well be another 
> option for you if this doesn't work out.

Can we get an update on this?

--
Jonathan Mcdougall


_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post21641

Hi Johnathan:
	Good news / bad news.  Good news is that we have a box in house
that has those cards in it.   The bad news is that I've reproduced your
problem(well... Not the machine slowing wdown to a crawl, but the "No
such device or address), so the driver doesn't work.  I'm looking into
it now. 

	Robert.

-----Original Message-----
From: Jonathan Mcdougall [mailto:community-noreply@qnx.com] 
Sent: Friday, February 06, 2009 2:07 PM
To: drivers-networking
Subject: Re: MTU on Broadcom BCM5787M

> x86 Driver is attached.  It's got basic sanity testing on it with the 
> card that I have in house, but I don't have a 5787, so you'll be our 
> test case for that card.

We replaced /lib/dll/devnp-bge.so with the one you sent. Restarting the
driver with

  $ slay io-pkt-v4-hc
  $ /sbin/io-pkt-v4 -dbge

slows the machine down to a crawl. After a few seconds, 

  Unable to init devnp-bge.so: No such device or address

is printed and, after a few more seconds, the whole machine hangs and
needs a hard reset.

Trying the same thing with the original bge driver gives the same
message but does not hang the machine.
  
We tried changing /etc/system/enum/devices by removing dev=1693 from the
tigon3 driver and added:

 device(pci, ven=$(PCI_VEND_BROADCOM), dev=1693)  #Broadcom 5787M
Gigabit
    tag(devn)
    append(legacy, ",nonet")
    requires($(IOPKT_CMD),)
    waitfor (/dev/socket)
    mount(-Tio-pkt "-opci=$(index),vid=0x$(ven),did=0x$(dev)"
/lib/dll/devnp-bge.so, )
    use(symbolic=netmgr)

However, that prevented the network card from being recognized after
restarting the computer. sloginfo gives:

  mount: Can't mount / (type io-pkt)
  mount: Possible reason: Unknown error

We get the same thing when running 

  $ mount -Tio-pkt "-opci=$(index),vid=0x$(ven),did=0x$(dev)"
/lib/dll/devnp-bge.so
  
directly.

As for me, I'm the programmer trying to recompile the driver. I'm having
issues (of course, I've got no experience in this), but I'm attaching
the files I modified just in case it might help you. I merged most of
the stuff from the head of netbsd in if_bge.c/h and some associated
files. Although it compiles and links fine, I'm getting some unresolved
names, obviously because I need to recompile and install much more stuff
than I initially thought. Anyways, I'll keep trying.

> On another note, it looks like we've potentially identified the link 
> problem with the tigon3 driver as well, so that may well be another 
> option for you if this doesn't work out.

Can we get an update on this?

--
Jonathan Mcdougall


_______________________________________________
Networking Drivers
http://community.qnx.com/sf/go/post21641