[eepro100] 2.4.x: eepro100 does not survive warm boot

Stefan Hoffmeister scyld.vortex@econos.de
Wed, 11 Jul 2001 10:44:42 +0200


: On Sat, 7 Jul 2001 14:18:41 -0400 (EDT), Donald Becker wrote:

>You should be running my driver.  The pci-scan layer includes the PCI/ACPI
>power management support code.  This is especially useful on the Z505.

I wish I were able to do that :-/ First of all, I get these compilation
errors on RedHat 7.1 (gcc 2.96-85):

Although this is addressed in your documentation, the fundamental reason
for the breakage seems to be in 

  kern_compat.h

where you test

  #if ! defined(HAS_NETIF_QUEUE)

while

  include/linux/netdevice.h

of 2.4.6 (pristine) actually does a 

  #define HAVE_NETIF_QUEUE

Changing kern_compat.h to use "HAVE_NETIF_QUEUE" instead of
"HAS_NETIF_QUEUE" gets eepro100.c to compile. Is this change correct? Yes,
there are quite a few compile-time warnings (and resulting insmod errors),
because netdevice.h defines

  netif_stop_queue

and not

  netif_stop_tx_queue

to name just one case. The fact that I cannot find "tbusy" in the 2.4.6
source tree makes me somehow believe that kern_compat.h needs updating to
support 2.4 kernels?

In similar vein, pci-scan.c uses

  apm_register_callback
  apm_unregister_callback

Neither of these symbols appears in my 2.4.6 source tree. Instead, a new
header file "pm.h" is present which offers callback management via
pm_register and pm_unregister; callbacks are called from the APM
implementation.

It seems as if it is not exactly easy to get your eepro100.c driver to
co-operator with a stock 2.4.6 kernel?

Disclaimer: I have no clue about Linux kernel modules, power management,
the differences between 2.2 / 2.4, so I am probably talking rubbish.

>politics won out over the technical side.

Fun.

Thanks very much,
Stefan