mapping 4096 bytes into the address space of a process

From the KB:

“Make sure you use mmap_device_memory(), with the PROT_READ |
PROT_WRITE |
PROT_NOCAHCE flags. If you don’t specify PROT_NOCACHE, you’ll
probably get a
CPU cached version of what used to be on the card. Be sure to
check for
error returns from mmap_device_memory() as well-
there may be page alignment issues you have to resolve”.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

How can these issues be resolved if mmap does that mapping ? I have
to map 4096 bytes …

Armin

My understanding would be that mmap_device_memory() will return an error if you
ask it to map non-page aligned addresses/sizes.

When QNX6 finally gets support for >4G memory (c’mon guys, go for it!), isn’t
the page size going to change to either 2 or 4 meg depending on whether or not
big memory is enabled? Could be a lot of error codes what need-a-checkin’ at
some point in the (hopefully) not too distant future.

-Warren “nudge, nudge, know what I mean?” Peece


“Armin Steinhoff” <A-Steinhoff@web_.de> wrote in message
news:3A6E036E.A62A2468@web_.de…
|
| From the KB:
|
| “Make sure you use mmap_device_memory(), with the PROT_READ |
| PROT_WRITE |
| PROT_NOCAHCE flags. If you don’t specify PROT_NOCACHE, you’ll
| probably get a
| CPU cached version of what used to be on the card. Be sure to
| check for
| error returns from mmap_device_memory() as well-
| there may be page alignment issues you have to resolve”.
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| How can these issues be resolved if mmap does that mapping ? I have
| to map 4096 bytes …
|
| Armin

Warren Peece wrote:

My understanding would be that mmap_device_memory() will return an error if you
ask it to map non-page aligned addresses/sizes.

The mapped physical address is page aligned … thanks for your
hint.

Armin


When QNX6 finally gets support for >4G memory (c’mon guys, go for it!), isn’t
the page size going to change to either 2 or 4 meg depending on whether or not
big memory is enabled? Could be a lot of error codes what need-a-checkin’ at
some point in the (hopefully) not too distant future.

-Warren “nudge, nudge, know what I mean?” Peece

“Armin Steinhoff” <A-Steinhoff@web_.de> wrote in message
news:3A6E036E.A62A2468@web_.de…
|
| From the KB:
|
| “Make sure you use mmap_device_memory(), with the PROT_READ |
| PROT_WRITE |
| PROT_NOCAHCE flags. If you don’t specify PROT_NOCACHE, you’ll
| probably get a
| CPU cached version of what used to be on the card. Be sure to
| check for
| error returns from mmap_device_memory() as well-
| there may be page alignment issues you have to resolve”.
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| How can these issues be resolved if mmap does that mapping ? I have
| to map 4096 bytes …
|
| Armin