PCI IO resource behind carbus bridge

I currently support a driver for expanding the PCI bus through a cardbus
card. A customer is using 4 cards that use 0x300 bytes of IO space each
(two windows each card 128 and 256). So the default 4K should be enough IO
to support these devices. However, only two devices can start while the
other two have an error code of 12 (insufficient resources), the driver will
open up more IO then the 4k if needed, but in this case 4k should be enough,
however if I do create an 8K window on the bridge, all four devices will
start.

At first I thought it might be that the IO is allocated on 1k bondary, but
viewing the two devices that start that doesn't seem to be the case Does MS
document there algorithm for assigning IO behind a cardbus bridge? I have
read a couple of white papers that were helpful, but not to the level of
detail I need.


Brian

IO addresses are obsolete, and it was so since around 2000 or even 1999. MS
was always recommending using memory addresses instead, especially for such
large pools.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation

http://www.storagecraft.com

"Brian" <> wrote in message
news:F8790752-F218-4CFE-81B3-...
> I currently support a driver for expanding the PCI bus through a cardbus
> card. A customer is using 4 cards that use 0x300 bytes of IO space each
> (two windows each card 128 and 256). So the default 4K should be enough IO
> to support these devices. However, only two devices can start while the
> other two have an error code of 12 (insufficient resources), the driver will
> open up more IO then the 4k if needed, but in this case 4k should be enough,
> however if I do create an 8K window on the bridge, all four devices will
> start.
>
> At first I thought it might be that the IO is allocated on 1k bondary, but
> viewing the two devices that start that doesn't seem to be the case Does MS
> document there algorithm for assigning IO behind a cardbus bridge? I have
> read a couple of white papers that were helpful, but not to the level of
> detail I need.
>
>
> Brian
>
>

Thanks, but too quote Nicolas Cage in Honeymoon in Las Vegas: "That doesn't
help me at all right now". The OS is not starting devices because of
insufficient resources and I want to know why, so that I can supply the
resources to the devices behind my bridge. For this specific customer I
could just grab enough, but I was hoping to find some information to enable
me to come up with a more elegant solution.

"Maxim S. Shatskih" wrote:

> IO addresses are obsolete, and it was so since around 2000 or even 1999. MS
> was always recommending using memory addresses instead, especially for such
> large pools.
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
>
> http://www.storagecraft.com
>
> "Brian" <> wrote in message
> news:F8790752-F218-4CFE-81B3-...
> > I currently support a driver for expanding the PCI bus through a cardbus
> > card. A customer is using 4 cards that use 0x300 bytes of IO space each
> > (two windows each card 128 and 256). So the default 4K should be enough IO
> > to support these devices. However, only two devices can start while the
> > other two have an error code of 12 (insufficient resources), the driver will
> > open up more IO then the 4k if needed, but in this case 4k should be enough,
> > however if I do create an 8K window on the bridge, all four devices will
> > start.
> >
> > At first I thought it might be that the IO is allocated on 1k bondary, but
> > viewing the two devices that start that doesn't seem to be the case Does MS
> > document there algorithm for assigning IO behind a cardbus bridge? I have
> > read a couple of white papers that were helpful, but not to the level of
> > detail I need.
> >
> >
> > Brian
> >
> >
>
>
>