[eepro100] Fwd: Compile-time problems with the newest eepro d river

Estes, Matthew mestes@microstrategy.com
Fri, 20 Oct 2000 11:42:03 -0400


Eric/Donald,
This may not help in this situation.  It is probably something unique with
my setup or my inexperience with building kernels that caused the errors in
the first place.  I could recompile now and it might work fine.  It's not
pretty, I'm just mentioning it in case somebody else runs into the same
thing.  Also, is it kosher to link modules into the kernel and use them as
modules?

Here's the mods to the drivers/net/Makefile:
L_OBJS   += auto_irq.o pci-scan.o
....
ifeq ($(CONFIG_EEXPRESS_PRO100),m)
  M_OBJS += eepro100.o pci-scan.o
  endif

Here's the entry in /lib/modules/[vers]/modules.dep:
/lib/modules/2.2.16-4.mat/net/eepro100.o:
/lib/modules/2.2.16-4.mat/net/pci-scan.o

Here's the output from lsmod:
eepro100               14032   2  (autoclean)
pci-scan                2428   0  (autoclean) [eepro100]

Everything is holding together.

Thanks,
Matt

-----Original Message-----
From: Donald Becker [mailto:becker@scyld.com]
Sent: Friday, October 20, 2000 11:12 AM
To: Estes, Matthew
Cc: Eric Hicks; eepro100@scyld.com
Subject: RE: [eepro100] Fwd: Compile-time problems with the newest eepro
d river


On Fri, 20 Oct 2000, Estes, Matthew wrote:

> Regardless, now I believe I have it compiled into the kernel and load it
as
> a module (through that module dependecy file) and everything works fine...
> except for Timed Out errors, but I'll address those in another thread.

What errors?

The only fix in v1.12 works around what appears to a bug in certain '559
chips that causes the machine to hang when the interface is first started.
This problem does not occur when the driver is compiled with -DUSE_IO_OPS.

Note: I/O operations do not go through the write buffer, and thus have
slightly slower timing.  The '559 only has one documented timing
requirement, and that's a brief delay after triggering a reset.  So this
problem really looks like a hardware bug.

> So, does this fix you described address what I originally experienced as
> well?

The fix is only for linking pci-scan.o into the kernel.
It does nothing if pci-scan.o is loaded as a module.


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


_______________________________________________
eepro100 mailing list
eepro100@scyld.com
http://www.scyld.com/mailman/listinfo/eepro100