e1000-devel Mailing List for Formerly Intel Ethernet Drivers

Those changes did it.

Had to reboot to get the old module to unload, but am running the new
one now.

One minor cosmetic issue showed up at the reboot.  On a soft boot,
the card does not reset, and comes up in whatever config it was in
before the reboot.  The old driver would then reset it to autoneg.
The new driver however left it up and reported that it was running
at 100 full duplex, even though it was still at 10 fd.

On the down side, it remains incompatable with my switch at 100 or
when autoneg is turned on.  (Belkin’s 8 port 10/100 soho switch.)

(It is unfortunate Intel doesn’t provide you the budget to acquire
a known incompatable switch and test it w/ a protocol analyser to
discover why it refuses to work at 100....)

-JimC

Hi Jim.  Sorry, I ran into those same issues the other day.  It seems
compiling against 2.5 is a continual struggle as things break from one
release to the next.  Here's what I did to get things working.  I'll try
to get these changes into the code/makefile and respin the driver.=20

1) Comment out #include "kcompat.h".  That's for back-ward comaptibility
with older kernels, but the driver is written to run unmodified against
the latest 2.5, so we don't need kcompat.h.

2) build mii.ko.  The menuconfig option is something like "Generic Media
Independent Interface device support" under the 10/100 networking
devices menu.  You'll need to load mii.ko before loading e100.ko.

3) In the driver's Makefile, apply this:

--- e100-3.0.0_dev9/src/Makefile        2003-06-22 23:57:20.000000000
-0700
+++ e100-3.0.0_dev9-mod/src/Makefile    2003-07-02 19:44:50.000000000
-0700
@@ -206,6 +206,14 @@ endif
 # depmod version for rpm builds
 DEPVER :=3D $(shell /sbin/depmod -V 2>/dev/null | awk 'BEGIN {FS=3D"."}
NR=3D=3D1 {print $$2}')

+# Check if it is 2.5 or 2.6 kernels
+MATCH=3D$(shell uname -r | awk '{ if (/2\.[5-6]\./) print "1"; else =
print
"0"}')
+ifeq ($(MATCH),1)
+CFLAGS +=3D -DKBUILD_MODNAME=3D$(TARGET:.o=3D)
+ifeq ($(ARCH),i386)
+  CFLAGS +=3D -I$(KSRC)/arch/i386/mach-generic
-I$(KSRC)/include/asm-i386/mach-default
+endif
+endif
=20
########################################################################
###
 # Build rules

Hopefully I didn't foget anything.  Let me know if this works.
Alternatively we could just patch the kernel.

-scott


> -----Original Message-----
> From: James H. Cloos Jr. [mailto:cl...@jh...]=20
> Sent: Wednesday, July 02, 2003 7:16 PM
> To: e10...@li...
> Subject: [E1000-devel] e100-3.0.0_dev9 compile fails
>=20
>=20
> Is 3.0.0_dev9 supposed to work w/ 2.5?  Compiled w/ gcc 3.3?
>=20
> A compile dump is attached below.
>=20
> (
> 2.5.73-jhc1 is bk://linux.bkbits.net/linux-2.5 as of changeset:
>=20
=20
http://linux.bkbits.net:8080/linux-2.5/cset@3efca929Cx6NfOlJIR4pM0ZKOooy
2Q
)

Incidently, irq_vectors.h does exist at:

   include/asm-i386/mach-default/irq_vectors.h

I presume the separation to the mach-foobar dirs under asm-i386 is not
expected by the new module.

The kcompat.h errors are I presume a gcc 3.3 issue?
(It doesn't seem to like < #if ( foo ) > syntax....)

-JimC

Is 3.0.0_dev9 supposed to work w/ 2.5?  Compiled w/ gcc 3.3?

A compile dump is attached below.

(
2.5.73-jhc1 is bk://linux.bkbits.net/linux-2.5 as of changeset:

  http://linux.bkbits.net:8080/linux-2.5/cset@3efca929Cx6NfOlJIR4pM0ZKOooy2Q
)

Incidently, irq_vectors.h does exist at:

   include/asm-i386/mach-default/irq_vectors.h

I presume the separation to the mach-foobar dirs under asm-i386 is
not expected by the new module.

The kcompat.h errors are I presume a gcc 3.3 issue?
(It doesn't seem to like « #if ( foo ) » syntax....)

-JimC