Hey gang, not be a pest but...

Another release kernel has come and gone with a severely broken NatSemi
driver.  Attached here is the same patch I sent you guys before (change
listing below).  The first two listings combined are this patch, the third
is the ethtool patch (also attached).

It would be super if we could push these changes into the mainline kernel
ASAP - I know the magic registers thing is causing problems...

> 1) As discussed with Andrew and Jeff: a generic wake-on-lan API for
> ethtool.  A simple struct, 2 ETHTOOL sub-ioctls, and some bitmasks.
> 
> 2) A largish patch against natsemi.c implementing several things.
> * increase the HW timeout to be long enough for an EEPROM reload
> * add several register names
> * add several bit-names for registers
> * remove the wake-on-lan interrupt from the default "on" list
> * call netif_carrier_off if we are not currently opened
> * force an EEPROM reload on a chip reset - some systems have trouble
> loading the EEPROM initially due to the dp83815 starting at voltage < 3.0v
> * ALWAYS set the magic registers
> * decrease maxdma to 256 - makes "Something Wicked" stop happening on
> ServerWorks.
> * Add support for ETHTOOL_GSET and ETHTOOL_SSET
> * Add support for new wake-on-lan ethtools (ETHTOOL_GWOL, ETHTOOL_SWOL)
> * only halt the chip if wake-on-lan is not enabled
> 
> 3) A patch against ethtool-1.2 implementing wake-on-lan support
> * bump version to 1.3
> * compile with -Wall
> * updated man page
> * added cmdline opts for wol and sopass
> * minor restructuring to make it easier

-- 
Tim Hockin
Systems Software Engineer
Sun Microsystems, Cobalt Server Appliances
thockin@sun.com