linuxppc_2_4_devel on walnut

linuxppc_2_4_devel on walnut

[Kenneth Johansson]

I tried the latest version on my walnut card and noticed two things.

1. The default configuration don't compile

2. The ethernet driver gets confused. It reports 10mb and half duplex when it
really is 100Mb full duplex. And it reports 100Mb full duplex when only 10Mb
half duplex exist.


--
Kenneth Johansson
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Tom Rini]

On Mon, Oct 29, 2001 at 05:35:39PM +0100, Kenneth Johansson wrote:

> 1. The default configuration don't compile

Can you post a log?

> 2. The ethernet driver gets confused. It reports 10mb and half duplex when it
> really is 100Mb full duplex. And it reports 100Mb full duplex when only 10Mb
> half duplex exist.

Which PHY?

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Armin Kuster]

Kenneth Johansson wrote:
>
> I tried the latest version on my walnut card and noticed two things.
>
> 1. The default configuration don't compile
>
> 2. The ethernet driver gets confused. It reports 10mb and half duplex when it
> really is 100Mb full duplex. And it reports 100Mb full duplex when only 10Mb
> half duplex exist.
>
> --
> Kenneth Johansson
> Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> Viderögatan 3                     Fax: +46 8 404 72 72
> 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se


Kenneth

1) Just pulled this morning (2.4.14-pre3), built  & booted fine

2) Ethernet driver matches what openbios says. I 'll do more tests..

-- armin

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Kenneth Johansson]

This is a litte update.

I did indeed not use the default configuration. The default worked fine.

The problem with the network was a typo in the command line and no problem
with the driver code but something is funny in the ip autoconf code.

So now it sets the ip and stuff over bootp but the nfsroot fails without even
trying to go out on the wire. This was with the code from 2001-10-29.

So I updated to todays version recompiled with the default and now I don't
get any output from the kernel even on the serial console. I did not have
time to read thru every patch since oct 29 today so I ask is this problem
known to exist at this point.


Kenneth Johansson wrote:
>
> I tried the latest version on my walnut card and noticed two things.
>
> 1. The default configuration don't compile
>
> 2. The ethernet driver gets confused. It reports 10mb and half duplex when it
> really is 100Mb full duplex. And it reports 100Mb full duplex when only 10Mb
> half duplex exist.
>
> --
> Kenneth Johansson
> Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> Viderögatan 3                     Fax: +46 8 404 72 72
> 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se
>

--
Kenneth Johansson
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: linuxppc_2_4_devel on walnut

[Wright, David]

If you're trying to boot a Walnut with an NFS root, it can be done,
but things have changed a bit from earlier versions.  We have it
working on 2.4.14-pre3.

1)  You still need a command line with the nfs root directive

2)  Your .config and autoconf.h need to specify both CONFIG_ROOT_NFS
    and CONFIG_IP_PNP_BOOTP.

There may be a way to do it with command line options, but I just
said the heck with it and changed net/ipv4/ipconfig.c to set the
default value of ic_enable to 1.  It used to be default 1, but now
is default 0.

  -- David Wright, Consulting Engineer, InfiniSwitch Corp.



> -----Original Message-----
> From: Kenneth Johansson [mailto:kenneth.johansson@inn.ericsson.se]
> Sent: Wednesday, November 07, 2001 12:39 PM
> To: linuxppc-embedded@lists.linuxppc.org; akuster@mvista.com
> Subject: Re: linuxppc_2_4_devel on walnut
>
>
>
> This is a litte update.
>
> I did indeed not use the default configuration. The default
> worked fine.
>
> The problem with the network was a typo in the command line
> and no problem
> with the driver code but something is funny in the ip autoconf code.
>
> So now it sets the ip and stuff over bootp but the nfsroot
> fails without even
> trying to go out on the wire. This was with the code from 2001-10-29.
>
> So I updated to todays version recompiled with the default
> and now I don't
> get any output from the kernel even on the serial console. I
> did not have
> time to read thru every patch since oct 29 today so I ask is
> this problem
> known to exist at this point.
>
>
> Kenneth Johansson wrote:
> >
> > I tried the latest version on my walnut card and noticed two things.
> >
> > 1. The default configuration don't compile
> >
> > 2. The ethernet driver gets confused. It reports 10mb and
> half duplex when it
> > really is 100Mb full duplex. And it reports 100Mb full
> duplex when only 10Mb
> > half duplex exist.
> >
> > --
> > Kenneth Johansson
> > Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> > Viderögatan 3                     Fax: +46 8 404 72 72
> > 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se
> >
>
> --
> Kenneth Johansson
> Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> Viderögatan 3                     Fax: +46 8 404 72 72
> 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se
>
>
>

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: linuxppc_2_4_devel on walnut

[Martin, Tim]

You can leave the default value of ic_enable at 0 and enter
ip=bootp as part of your boot prompt.

Tim

-----Original Message-----
From: Wright, David [mailto:dwright@infiniswitch.com]
Sent: Wednesday, November 07, 2001 9:47 AM
To: Kenneth Johansson; linuxppc-embedded@lists.linuxppc.org;
akuster@mvista.com
Subject: RE: linuxppc_2_4_devel on walnut



If you're trying to boot a Walnut with an NFS root, it can be done,
but things have changed a bit from earlier versions.  We have it
working on 2.4.14-pre3.

1)  You still need a command line with the nfs root directive

2)  Your .config and autoconf.h need to specify both CONFIG_ROOT_NFS
    and CONFIG_IP_PNP_BOOTP.

There may be a way to do it with command line options, but I just
said the heck with it and changed net/ipv4/ipconfig.c to set the
default value of ic_enable to 1.  It used to be default 1, but now
is default 0.

  -- David Wright, Consulting Engineer, InfiniSwitch Corp.



> -----Original Message-----
> From: Kenneth Johansson [mailto:kenneth.johansson@inn.ericsson.se]
> Sent: Wednesday, November 07, 2001 12:39 PM
> To: linuxppc-embedded@lists.linuxppc.org; akuster@mvista.com
> Subject: Re: linuxppc_2_4_devel on walnut
>
>
>
> This is a litte update.
>
> I did indeed not use the default configuration. The default
> worked fine.
>
> The problem with the network was a typo in the command line
> and no problem
> with the driver code but something is funny in the ip autoconf code.
>
> So now it sets the ip and stuff over bootp but the nfsroot
> fails without even
> trying to go out on the wire. This was with the code from 2001-10-29.
>
> So I updated to todays version recompiled with the default
> and now I don't
> get any output from the kernel even on the serial console. I
> did not have
> time to read thru every patch since oct 29 today so I ask is
> this problem
> known to exist at this point.
>
>
> Kenneth Johansson wrote:
> >
> > I tried the latest version on my walnut card and noticed two things.
> >
> > 1. The default configuration don't compile
> >
> > 2. The ethernet driver gets confused. It reports 10mb and
> half duplex when it
> > really is 100Mb full duplex. And it reports 100Mb full
> duplex when only 10Mb
> > half duplex exist.
> >
> > --
> > Kenneth Johansson
> > Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> > Viderögatan 3                     Fax: +46 8 404 72 72
> > 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se
> >
>
> --
> Kenneth Johansson
> Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> Viderögatan 3                     Fax: +46 8 404 72 72
> 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se
>
>
>


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Kenneth Johansson]

The saga continues.

Well some of the problems I had was of the brown paperbag over head kind of
error but now I can boot and run a version checked out a few hours ago.

But the ethernet reporting of the line is wrong. This is a problem with the
printout only as the function is right. The 100Mb full duplex is reported to
be 10Mb half and vise verse.


I have not run the walnut hardware for some time but I still have a quite
severe problem with ethernet in 10Mb.

What happens is this if I connect to a 3com 10Mb hub that is connected to the
office net (cisco switch) the switch turns off the port after some time
mostly when I run apt-get. Since the walnut card is never directly connected
to the switch this must be some logical error in packets probably the
intergap time is to small that is the only thing I can think of as the switch
only care about the ethernet layer.

This problem could maybee exist if I directly connect to the switch I have
not done that and no I can not test this easy as I don't control the switch
and now only have 1 working port in that room and it is really needed.


--
Kenneth Johansson
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Armin Kuster]

[-- Attachment #1: Type: text/plain, Size: 1588 bytes --]

Kenneth Johansson wrote:
>
> The saga continues.
>
> Well some of the problems I had was of the brown paperbag over head kind of
> error but now I can boot and run a version checked out a few hours ago.
>
> But the ethernet reporting of the line is wrong. This is a problem with the
> printout only as the function is right. The 100Mb full duplex is reported to
> be 10Mb half and vise verse.
>
> I have not run the walnut hardware for some time but I still have a quite
> severe problem with ethernet in 10Mb.
>
> What happens is this if I connect to a 3com 10Mb hub that is connected to the
> office net (cisco switch) the switch turns off the port after some time
> mostly when I run apt-get. Since the walnut card is never directly connected
> to the switch this must be some logical error in packets probably the
> intergap time is to small that is the only thing I can think of as the switch
> only care about the ethernet layer.
>
> This problem could maybee exist if I directly connect to the switch I have
> not done that and no I can not test this easy as I don't control the switch
> and now only have 1 working port in that room and it is really needed.
>
> --
> Kenneth Johansson
> Ericsson Business Innovation AB   Tel: +46 8 404 71 83
> Viderögatan 3                     Fax: +46 8 404 72 72
> 164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

Kenneth;

I finally got arround to looking into this and the problem is the
status.
We use the status to program the emac to the same speed as the PHY.

I have attach a patch the should correct this.


-- armin

[-- Attachment #2: 405_enet_1114.patch.gz --]
[-- Type: application/octet-stream, Size: 2775 bytes --]

Re: linuxppc_2_4_devel on walnut

[Kenneth Johansson]

Armin Kuster wrote:
>
> Kenneth Johansson wrote:
> >
> > But the ethernet reporting of the line is wrong. This is a problem with the
> > printout only as the function is right. The 100Mb full duplex is reported to
> > be 10Mb half and vise verse.
> >
>
> Kenneth;
>
> I finally got arround to looking into this and the problem is the
> status.
> We use the status to program the emac to the same speed as the PHY.
>
> I have attach a patch the should correct this.

I saw that this was also in the devel tree so I updated my copy and yes the
printout is now right.

But "time dd if=/dev/zero of=testing bs=1024k count=100" on a nfs mouted fs
used to be 18s and is now 21s, a 16% slowdown not that I care much but still
quite a large difference.

And I have not tired this on my 10Mb port yet as I still have only one active
but as soon as they work again I test on 10Mb.

--
Kenneth Johansson
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Dan Malek]

Kenneth Johansson wrote:


> But "time dd if=/dev/zero of=testing bs=1024k count=100" on a nfs mouted fs
> used to be 18s and is now 21s, a 16% slowdown not that I care much but still
> quite a large difference.

What versions of kernels/drivers are you comparing?  Do you do this
back-to-back with the same NFS server, or have you introduced more
variables into your testing?  This is why specially prepared network
performance testers are better suited for discussion.

Many versions of the 405GP Ethernet driver didn't use (and didn't have)
cache coherency functions properly.  It was basically lucky, and when
errors occurred you usually didn't see them because the packets were
discarded and retransmitted.  Recent versions of the driver now use
properly mapped and managed cache coherent memory.  When I was making
the driver changes for the new cache attributes, I was thinking we should
be experimenting with cached pages and invalidates/flushes traded off
against uncached pages.  There will certainly be a performance difference.



	-- Dan


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: linuxppc_2_4_devel on walnut

[Kenneth Johansson]

Dan Malek wrote:
>
> Kenneth Johansson wrote:
>
> > But "time dd if=/dev/zero of=testing bs=1024k count=100" on a nfs mouted fs
> > used to be 18s and is now 21s, a 16% slowdown not that I care much but still
> > quite a large difference.
>
> What versions of kernels/drivers are you comparing?  Do you do this
> back-to-back with the same NFS server, or have you introduced more
> variables into your testing?  This is why specially prepared network
> performance testers are better suited for discussion.

This was to the same nfs server on a live network. I did the test only two
times on the first kernel version that was checke out a few hours before the
mail was posted last friday.

This was not done to benchmark anything but a way for me to verify what speed
the networks was using 10/100 Mb was at the time reported wrong by the
driver.

The reason I run it twice was that I thougt 5,5MB/s was not so great but the
second run had the same number. Then I upgraded the tree to a version that
had Armin Kusters changes and just for fun run the test again.

If I test only tcp/ip over the network I now get this..
----
walnut:/tmp# time dd if=/dev/zero  bs=1024k count=100 | nc -q0 <server> 9
100+0 records in
100+0 records out

real    0m14.838s
user    0m0.370s
sys     0m14.380s
---
6,7Mb/s

If I do the same from a intel 400MHz kernel 2.2.20 I get
---
[innkeon@spawn innkeon]$time dd if=/dev/zero  bs=1024k count=100 | nc -q0
<server> 9
100+0 records in
100+0 records out

real    0m12.838s
user    0m0.110s
sys     0m1.890s
---
7,8 MB/s

hmm funny the difference is also 16%

(nc is the netcat program.)

--
Kenneth Johansson
Ericsson Business Innovation AB   Tel: +46 8 404 71 83
Viderögatan 3                     Fax: +46 8 404 72 72
164 80 Stockholm                  kenneth.johansson@inn.ericsson.se

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/