[tulip] Problem with tulip, 2.4.0test kernel

Jonathan Earle jearle@nortelnetworks.com
Mon, 15 Jan 2001 15:54:14 -0500


This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01C07F35.51BC40E0
Content-Type: text/plain;
	charset="iso-8859-1"

Thanks for the tip regarding the options line!  I added "options tulip
options=5,5,5,5" to the modules.conf on the two boxes in question, and now
dmesg will show:

eth1: Using user-specified media 100baseTx-FD.
eth2: Using user-specified media 100baseTx-FD.
eth3: Using user-specified media 100baseTx-FD.
eth4: Using user-specified media 100baseTx-FD.     

>From a cold boot, I now have all four ports operating.  I didn't realize
that all four had to be set individually, although it's fairly obvious now.
Hindsight is always 20/20.

Is it fair to say that the 2.4.0 driver has problems auto-negotiating or is
the problem I saw symptomatic of something else?

Also, I infer that this card does not use MII registers, but emulates them
instead.  What does that ultimately mean?  What does this card use in place
of those registers?

What I'm now observing, is an initial drop in performance after bootup.
Once I was able to ping back and forth betweem the two NICs, I tried an FTP
transfer (20+MB file).  For the first few minutes, the cards will show
transfer rates all over the map - from a few hundred kB/s to perhaps a few
MB/s.  The transfer will generally start at a higher rate, then drop.
Eventually, everything seems to settle down and I then see the appropriate
rates (11.2MB/s for full duplex, 100TX).

This is a lead to another problem I'm having, which I'll post separately.

Many thanks Donald!

Cheers!
Jon

> -----Original Message-----
> From: Donald Becker [mailto:becker@scyld.com]
> Sent: Monday, January 15, 2001 2:30 PM
> To: Earle, Jonathan [KAN:1A31:EXCH]
> Cc: 'Tulip Driver List'
> Subject: Re: [tulip] Problem with tulip, 2.4.0test kernel
> 
> 
> On Mon, 15 Jan 2001, Jonathan Earle wrote:
> 
> > Config is two PIII boxes, RH6.2, kernel 2.4.0-test9 (tulip 
> driver lifted
> > from kernel 2.4.0 - version 0.9.13 (January 2, 2001) but 
> was previously
> > using the proper 2.4.0-test9 driver - version 0.9.10 
> (September 6, 2000))
> > and Znyx ZX346Q 4port cards, with each port connected to 
> the corresponding
> 
> This appears to be a problem in the 2.4.0 driver.
> 
> I've tested the '346 with 2.2 and my drivers.
> 
> > Linux Tulip driver version 0.9.13 (January 2, 2001)
> 
> > root@onc1:~> tulip-diag -aa -f
> > Index #1: Found a Digital DS21143 Tulip adapter at 0xec80.
> >  Port selection is 10mpbs-serial, half-duplex.
> >   The NWay status register is 000000c6.
> >   Internal autonegotiation state is 'Autonegotiation disabled'.
> 
> Hmmm, did you force full dulex?
> 
> >   The Tx process state is 'Waiting for Tx to finish'.
> 
> This Tx state is normal for an incorrect media selection.
> 
> > Index #2: Found a Digital DS21143 Tulip adapter at 0xec00.
> >  Port selection is 100mbps-SYM/PCS 100baseTx scrambler, full-duplex.
> >   The Rx process state is 'Waiting for packets'.
> >   The Tx process state is 'Idle'.
> 
> This interface looks normal.
> 
> > Index #3: Found a Digital DS21143 Tulip adapter at 0xe880.
> >  Port selection is 10mpbs-serial, full-duplex.
> >   The Tx process state is 'Waiting for Tx to finish'.
> 
> Same problem as #1.
> 
> > Now, if I run tulip-diag -m (or -mm), it tells me there are no MII
> > transceivers.  However, if I run mii-diag, it shows:
> > 
> > root@onc1:~> mii-diag eth1
> > Basic registers of MII PHY #32:  0000 784c 0000 0000 0401 
> 0000 0000 0000.
> 
> Valid MII addresses are 0-31.
> Address #32 is used when the driver is internally emulating the MII
> management registers.
> 
> The recent 0.92t has substantially improved the accuracy of the
> emulation, notably the accuracy of the link beat report.
> 
> > Also, how can I force a specific speed?  I know I can pass 
> 'full_duplex=1'
> > to the module to enable full duplex, but what do I pass to 
> force 100TX
> > instead of 10BT?
> 
> See the media table at
>    http://www.scyld.com/network/tulip.html
> 
> options=3,3,5,5  sets 100baseTx, 100baseTx, 100baseTx-FDX, 
> 100baseTx-FDX
> 
> I now recommend setting the speed+duplex based on the table 
> rather than
> using the "full_duplex=" module option.
> 
> <table border=2>
> <tr><td> index	<td>  media</tr>
> <tr><td>  0	<td> Auto-select (default to the 10baseT link)</tr>
> <tr><td>  1	<td> 10base2</tr>
> <tr><td>  2	<td> AUI</tr>
> <tr><td>  3	<td> 100baseTx</tr>
> <tr><td>  4	<td> 10baseT-FD</tr>
> <tr><td>  5	<td> 100baseTx-FD</tr>
> <tr><td>  6	<td> 100baseT4</tr>
> <tr><td>  7	<td> 100baseFx</tr>
> <tr><td>  8	<td> 100baseFx-FD</tr>
> <tr><td>  9	<td> MII 10baseT</tr>
> <tr><td> 10	<td> MII 10baseT-FD</tr>
> <tr><td> 11	<td> MII (autoselect)</tr>
> <tr><td> 12	<td>  Serial 10baseT (no autoselect)</tr>
> <tr><td> 13	<td> MII 100baseTx</tr>
> <tr><td> 14	<td> MII 100baseTx-FD</tr>
> <tr><td> 15	<td> MII 100baseT4</tr>
> <tr><td> 16     <td> MII 100baseFx-HDX (half duplex)</tr>
> <tr><td> 17	<td> MII 100baseFx-FDX (full duplex)</tr>
> <tr><td> 18	<td> MII Home-PNA 1Mbps</tr>
> <tr><td> 0x200 (512 decimal)	<td> Added to other values to 
> set full duplex</tr>
> </table>
> 
> 
> 
> 
> Donald Becker				becker@scyld.com
> Scyld Computing Corporation		http://www.scyld.com
> 410 Severn Ave. Suite 210		Second Generation 
> Beowulf Clusters
> Annapolis MD 21403			410-990-9993
> 
> 
> _______________________________________________
> tulip mailing list
> tulip@scyld.com
> http://www.scyld.com/mailman/listinfo/tulip
> 

------_=_NextPart_001_01C07F35.51BC40E0
Content-Type: text/html;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">



RE: [tulip] Problem with tulip, 2.4.0test kernel



Thanks for the tip regarding the options line!  = I added "options tulip options=3D5,5,5,5" to the modules.conf = on the two boxes in question, and now dmesg will show:

eth1: Using user-specified media 100baseTx-FD.
eth2: Using user-specified media = 100baseTx-FD.
eth3: Using user-specified media = 100baseTx-FD.
eth4: Using user-specified media = 100baseTx-FD.    

From a cold boot, I now have all four ports = operating.  I didn't realize that all four had to be set = individually, although it's fairly obvious now.  Hindsight is = always 20/20.

Is it fair to say that the 2.4.0 driver has problems = auto-negotiating or is the problem I saw symptomatic of something = else?

Also, I infer that this card does not use MII = registers, but emulates them instead.  What does that ultimately = mean?  What does this card use in place of those = registers?

What I'm now observing, is an initial drop in = performance after bootup.  Once I was able to ping back and forth = betweem the two NICs, I tried an FTP transfer (20+MB file).  For = the first few minutes, the cards will show transfer rates all over the = map - from a few hundred kB/s to perhaps a few MB/s.  The transfer = will generally start at a higher rate, then drop.  Eventually, = everything seems to settle down and I then see the appropriate rates = (11.2MB/s for full duplex, 100TX).

This is a lead to another problem I'm having, which = I'll post separately.

Many thanks Donald!

Cheers!
Jon

> -----Original Message-----
> From: Donald Becker [mailto:becker@scyld.com]
> Sent: Monday, January 15, 2001 2:30 PM
> To: Earle, Jonathan [KAN:1A31:EXCH]
> Cc: 'Tulip Driver List'
> Subject: Re: [tulip] Problem with tulip, = 2.4.0test kernel
>
>
> On Mon, 15 Jan 2001, Jonathan Earle = wrote:
>
> > Config is two PIII boxes, RH6.2, kernel = 2.4.0-test9 (tulip
> driver lifted
> > from kernel 2.4.0 - version 0.9.13 = (January 2, 2001) but
> was previously
> > using the proper 2.4.0-test9 driver - = version 0.9.10
> (September 6, 2000))
> > and Znyx ZX346Q 4port cards, with each = port connected to
> the corresponding
>
> This appears to be a problem in the 2.4.0 = driver.
>
> I've tested the '346 with 2.2 and my = drivers.
>
> > Linux Tulip driver version 0.9.13 (January = 2, 2001)
>
> > root@onc1:~> tulip-diag -aa -f
> > Index #1: Found a Digital DS21143 Tulip = adapter at 0xec80.
> >  Port selection is 10mpbs-serial, = half-duplex.
> >   The NWay status register is = 000000c6.
> >   Internal autonegotiation state = is 'Autonegotiation disabled'.
>
> Hmmm, did you force full dulex?
>
> >   The Tx process state is = 'Waiting for Tx to finish'.
>
> This Tx state is normal for an incorrect media = selection.
>
> > Index #2: Found a Digital DS21143 Tulip = adapter at 0xec00.
> >  Port selection is 100mbps-SYM/PCS = 100baseTx scrambler, full-duplex.
> >   The Rx process state is = 'Waiting for packets'.
> >   The Tx process state is = 'Idle'.
>
> This interface looks normal.
>
> > Index #3: Found a Digital DS21143 Tulip = adapter at 0xe880.
> >  Port selection is 10mpbs-serial, = full-duplex.
> >   The Tx process state is = 'Waiting for Tx to finish'.
>
> Same problem as #1.
>
> > Now, if I run tulip-diag -m (or -mm), it = tells me there are no MII
> > transceivers.  However, if I run = mii-diag, it shows:
> >
> > root@onc1:~> mii-diag eth1
> > Basic registers of MII PHY #32:  0000 = 784c 0000 0000 0401
> 0000 0000 0000.
>
> Valid MII addresses are 0-31.
> Address #32 is used when the driver is = internally emulating the MII
> management registers.
>
> The recent 0.92t has substantially improved the = accuracy of the
> emulation, notably the accuracy of the link = beat report.
>
> > Also, how can I force a specific = speed?  I know I can pass
> 'full_duplex=3D1'
> > to the module to enable full duplex, but = what do I pass to
> force 100TX
> > instead of 10BT?
>
> See the media table at
>    http://www.scyld.com/network/tulip.html
>
> options=3D3,3,5,5  sets 100baseTx, = 100baseTx, 100baseTx-FDX,
> 100baseTx-FDX
>
> I now recommend setting the speed+duplex based = on the table
> rather than
> using the "full_duplex=3D" module = option.
>
> <table border=3D2>
> <tr><td> = index        <td>  = media</tr>
> <tr><td>  0   = <td> Auto-select (default to the 10baseT link)</tr>
> <tr><td>  1   = <td> 10base2</tr>
> <tr><td>  2   = <td> AUI</tr>
> <tr><td>  3   = <td> 100baseTx</tr>
> <tr><td>  4   = <td> 10baseT-FD</tr>
> <tr><td>  5   = <td> 100baseTx-FD</tr>
> <tr><td>  6   = <td> 100baseT4</tr>
> <tr><td>  7   = <td> 100baseFx</tr>
> <tr><td>  8   = <td> 100baseFx-FD</tr>
> <tr><td>  9   = <td> MII 10baseT</tr>
> <tr><td> 10   <td> = MII 10baseT-FD</tr>
> <tr><td> 11   <td> = MII (autoselect)</tr>
> <tr><td> 12   = <td>  Serial 10baseT (no autoselect)</tr>
> <tr><td> 13   <td> = MII 100baseTx</tr>
> <tr><td> 14   <td> = MII 100baseTx-FD</tr>
> <tr><td> 15   <td> = MII 100baseT4</tr>
> <tr><td> 16     = <td> MII 100baseFx-HDX (half duplex)</tr>
> <tr><td> 17   <td> = MII 100baseFx-FDX (full duplex)</tr>
> <tr><td> 18   <td> = MII Home-PNA 1Mbps</tr>
> <tr><td> 0x200 (512 decimal)  = <td> Added to other values to
> set full duplex</tr>
> </table>
>
>
>
>
> Donald Becker =         =         =         becker@scyld.com
> Scyld Computing Corporation   =         http://www.scyld.com
> 410 Severn Ave. Suite = 210             = Second Generation
> Beowulf Clusters
> Annapolis MD 21403    =         =         410-990-9993
>
>
> = _______________________________________________
> tulip mailing list
> tulip@scyld.com
> http://www.scyld.com/mailman/listinfo/tulip=
>

------_=_NextPart_001_01C07F35.51BC40E0--