use proper bus_size_t for register offsets and not bus_addr_t; uwe@ miod@ testing

First round of __P removal in sys

remove cold.

move extern int cold to .h

Use 32 pages instead of 4 in mem_alloc. This is needed by cnw, and useful
w/ other drivers as well. The iomem change has changed long ago to keep
using 4 pages; From NetBSD.

Make suspend -> eject card -> resume work.

On suspend, the apm code calls each function in its powerhook list. This list
contains pointers to functions that are to be executed on PWR_SUSPEND and
PWR

Make suspend -> eject card -> resume work.

On suspend, the apm code calls each function in its powerhook list. This list
contains pointers to functions that are to be executed on PWR_SUSPEND and
PWR_RESUME. One of these functions is pcmcia_power() which calls xxx_activate()
for each PC Card that is connected with the action DVACT_ACTIVATE (where xxx
is the name of the driver for the PC Card). On resume, the same thing happens,
except the action is DVACT_ACTIVATE.

Clearly, if a card is gone when we resume, trying to activate it is a bad idea.
This commit adds a pcic_power() function that runs before pcmcia_power(). On
resume, it uses pcic_intr_socket() to detect any events that occurred while we
were suspended. (I had to split pcic_event_thread() into a threaded part and
a non-threaded part, since the thread will not run when we are in interrupt
context, but I need the events to be processed before pcmcia_power() runs.) So,
by the time pcic_power() is through, all events that occurred during suspend
have been processed, and pcmcia_power() will not try to activate cards that are
not there since they have already been completely and properly detached.

Note that event handling for card removals first calls the deactivate functions
on the card then the detach functions. When we suspend, apm does the deactivate
for us. So on resume, if we detect a card has been removed, we skip the
deactivation step and just detach. We use the DVF_ACTIVE bit to determine
whether or not deactivation has already occurred. Deactivating a device that
has already been deactivated causes a panic.

show more ...

Implement PCMCIA event polling. This will either complement interrupt-driven
event notification from the pcic (if an interrupt is available) or allow the
insertion and removal of PCMCIA devices even

Implement PCMCIA event polling. This will either complement interrupt-driven
event notification from the pcic (if an interrupt is available) or allow the
insertion and removal of PCMCIA devices even if there is no IRQ for the pcic.
Each socket is checked for status change every 0.5 seconds using art's timeout
code. Insertion and removal events should no longer ever go unnoticed.

show more ...

Fix a bug introduced after cardbus commit where uart16550a was
being detected as uart16450 as side effect. This also corrects the
posibility of misdetection of vadem controllers.

Initial check-in for support of 32-bit CardBus PC Cards; from NetBSD. On many
machines, this code needs the new PCIBIOS* options enabled in the kernel config
file to work, but your mileage may vary.

Initial check-in for support of 32-bit CardBus PC Cards; from NetBSD. On many
machines, this code needs the new PCIBIOS* options enabled in the kernel config
file to work, but your mileage may vary. Included is a working 3c575 driver for
3Com 10/100 CardBus PC Card NICs (tested only with the 'C' revision). The 3c575
is the pccard version of the PCI EtherLink XL cards, and thus the xl driver has
been split into /sys/dev/ic.

show more ...

Fill in detach support in the i82365 driver; mostly netbsd, as was the
former pcmcia commit.

shrink code, cleanup port printouts, repairs various comments

Add support for Vadem PCMCIA controllers.

provide pcmcia IO bus mapper with list of blocks where it should
preferentially map; idea by me, code by niklas

Added support for Intel 82365SL PCIC controllers and clones from NetBSD.