2.5.1-pre11 de2104X tulip driver problem

2.5.1-pre11 de2104X tulip driver problem

[really mason@soo.com]

Just a tulip driver bug report:

i'm one of those dinosaurs using a 10base2 network and really
old DLink-530 21040 and 21041 based ethercards.

This is what a recent kernel version (2.5.1-pre10) working
tulip driver prints out when booting up:

kernel: tulip0: 21041 Media table, default media 0800 (Autosense).
kernel: tulip0:  21041 media #0, 10baseT.
kernel: tulip0:  21041 media #4, 10baseT-FDX.
kernel: tulip0:  21041 media #1, 10base2.
[...]
kernel: eth0: Digital DC21041 Tulip rev 17 at 0xe0800f80, 21041 mode, 00:80:C8:3E:D0:BC, IRQ 12.
[...]
kernel: eth0: No 21041 10baseT link beat, Media switched to AUI.

Card still works on my 10base2 network even tho i haven't got
an AUI port on the ethercard.
======================================================================

This is what the non working 2.5.1-pre11 tulip driver prints out:

kernel: de2104x PCI Ethernet driver v0.5.1 (Nov 20, 2001)
kernel: de0: SROM-listed ports: TP
kernel: eth0: 21041 at 0xe0800f80, 00:80:c8:3e:d0:bc, IRQ 12
[...]
kernel: eth0: set link 10baseT auto, mode 7ffc0040, sia 10c4,ffffef01,ffffffff,ffff0008
kernel:                  set mode 7ffc0000, set sia ef01,ffff,8

====================================================================

regards,
b <mason@soo.com>

Re: 2.5.1-pre11 de2104X tulip driver problem

[Johan Kullstam]

"really mason@soo.com" <lnx-kern@Sophia.soo.com> writes:

> Just a tulip driver bug report:
> 
> i'm one of those dinosaurs using a 10base2 network and really
> old DLink-530 21040 and 21041 based ethercards.
> 
> This is what a recent kernel version (2.5.1-pre10) working
> tulip driver prints out when booting up:
> 
> kernel: tulip0: 21041 Media table, default media 0800 (Autosense).
> kernel: tulip0:  21041 media #0, 10baseT.
> kernel: tulip0:  21041 media #4, 10baseT-FDX.
> kernel: tulip0:  21041 media #1, 10base2.
> [...]
> kernel: eth0: Digital DC21041 Tulip rev 17 at 0xe0800f80, 21041 mode, 00:80:C8:3E:D0:BC, IRQ 12.
> [...]
> kernel: eth0: No 21041 10baseT link beat, Media switched to AUI.
> 
> Card still works on my 10base2 network even tho i haven't got
> an AUI port on the ethercard.
> ======================================================================
> 
> This is what the non working 2.5.1-pre11 tulip driver prints out:
> 
> kernel: de2104x PCI Ethernet driver v0.5.1 (Nov 20, 2001)
> kernel: de0: SROM-listed ports: TP
> kernel: eth0: 21041 at 0xe0800f80, 00:80:c8:3e:d0:bc, IRQ 12
> [...]
> kernel: eth0: set link 10baseT auto, mode 7ffc0040, sia 10c4,ffffef01,ffffffff,ffff0008
> kernel:                  set mode 7ffc0000, set sia ef01,ffff,8
> 
> ====================================================================

i have a DEC DE450 (based on 21041 AA chipset).  i guess, for
2104[01], tulip driver has been broken since 2.4.4 (yes, that's over
six months of brokeness).  yes, jeff garzik knows about it.  i've
emailed the list and sourceforge &c.

rememdies --

0) buy a new network card.

1) use the de4x5 driver instead.

2) use the older tulip driver 0.9.14.  download it from 
   http://sourceforge.net/projects/tulip/

if you go with 2, just download.  have /usr/src/linux be your linux
source (or a symlink to them).  go into tulip-0.9.14/src, make dep,
make.  then copy tulip.o into /lib/modules/2.[45].X/kernel/drivers/net/tulip
which replaces the kernels one.

best of luck.

-- 
J o h a n  K u l l s t a m
[kullstam@mediaone.net]

Re: 2.5.1-pre11 de2104X tulip driver problem

[Jeff Garzik]

Johan Kullstam wrote:
> i have a DEC DE450 (based on 21041 AA chipset).  i guess, for
> 2104[01], tulip driver has been broken since 2.4.4 (yes, that's over
> six months of brokeness).  yes, jeff garzik knows about it.  i've
> emailed the list and sourceforge &c.

This is a bug report for de2104x not tulip.  de2104x was created to fix
the 2104x problems in tulip, which is getting really hairy to maintain.

	Jeff


-- 
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno

Re: 2.5.1-pre11 de2104X tulip driver problem

[Johan Kullstam]

Jeff Garzik <jgarzik@mandrakesoft.com> writes:

> Johan Kullstam wrote:
> > i have a DEC DE450 (based on 21041 AA chipset).  i guess, for
> > 2104[01], tulip driver has been broken since 2.4.4 (yes, that's over
> > six months of brokeness).  yes, jeff garzik knows about it.  i've
> > emailed the list and sourceforge &c.
> 
> This is a bug report for de2104x not tulip.  de2104x was created to fix
> the 2104x problems in tulip, which is getting really hairy to
> maintain.

ah ok.  sorry to crap on you.  i guess the proliferation of tulip
clones is getting pretty crazy.

i got a switch over the weekend.  my DE500 (DEC 21140AB w/o MII)
wouldn't negotiate full-duplex.  i finally just broke down and
scrapped my tulips and put in a 3com905b instead.

-- 
J o h a n  K u l l s t a m
[kullstam@mediaone.net]

Re: 2.5.1-pre11 de2104X tulip driver problem

[Jeff Garzik]

"really mason@soo.com" wrote:
> i'm one of those dinosaurs using a 10base2 network and really
> old DLink-530 21040 and 21041 based ethercards.
[...]
> kernel: de0: SROM-listed ports: TP

Thanks for the report.  If you are using 10base2 that's definitely a
bug.

Can you please change DE_DEF_MSG_ENABLE in de2104x.c as shown below, and
[privately] e-mail me /bin/dmesg and "lspci -vvv" output?

-#define DE_DEF_MSG_ENABLE       (NETIF_MSG_DRV          | \
-                                 NETIF_MSG_PROBE        | \
-                                 NETIF_MSG_LINK         | \
-                                 NETIF_MSG_TIMER        | \
-                                 NETIF_MSG_IFDOWN       | \
-                                 NETIF_MSG_IFUP         | \
-                                 NETIF_MSG_RX_ERR       | \
-                                 NETIF_MSG_TX_ERR)
+#define DE_DEF_MSG_ENABLE       0xffffffff

-- 
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno