ACX100 improvements

Hello kernel-discuss,

  Earlier this weak, I fixed suspends/resume handling for acs_cs
driver (well, I just needed to do the same as mem.c does already). I
also spotted and fixed (in rather dirty manner though) buffer overflow
issue when accessing /proc diag files for the driver (could cause
great kernel instability). And as another message to the list show, I
debugged previously reported issue that after resume WEP connection
doesn't work and keys must be reinitialized.

  So, per my testing with h4000, we now have baseline decent acx100
support. The current issue is that full network connectivity recovery
after suspend takes up to 40s. This is apparently due to queues being
not properly shutdown, etc.

  Other issue I see is that driver has poor AP affinity. In case ESSID
is set to "any", it seems to try to use only that last AP it sees in
scan, even if it has very poot signal and is not really connectable,
while there're strong APs before it in the list. I guess, it makes no
sense to try to fix this, instead we should upgrade to 2.6.22 which
has mac80211 (former dscape) WiFi stack, and start to hack on the
corresponding driver version. That would bring us WPA too. One sweet
day, I mean.

  I also reported resume issues with my hx4700, while Bill Reese
hinted it should work, and it does work on other slavemem-based
devices.

  Finally, we should proceed with cleaning up source. We now have 2
source file forks of the original pci.c. So, the aim is to fold cs.c
back into mem.c, and it turn - into pci.c.

     As usual, help with any of these welcome.

-- 
Best regards,
 Paul                          mailto:pmiscml_at_gmail.com