MmMapIoSpace is it really uncached??
bigus_dingus_at_yahoo.com
Date: 01/24/05
- Next message: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Previous message: MP: "Platform Builder 5.0 installation"
- Next in thread: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Reply: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Messages sorted by: [ date ] [ thread ]
Date: 24 Jan 2005 11:57:21 -0800
I am using the MmMapIoSpace() system call to attempt to access a
physical address in my driver. I am passing in FALSE for the parameter
"CacheType" as I most definitely do _not_ want this access to go
through the processor's cache. However, looking at the bus cycle, I
don't believe that I am getting an un-cached access. Within
OEMAddressTable, this area is mapped to virtual address 0x80000000
which I understand is un-cached.
So, my question is this,
Is it possible to get an un-cached address from MmMapIoSpace() if the
corresponding entry within OEMAddressTable is cached?
-background
I am using a Sharp arm processor. On the Sharp arm bus, sequential bus
accesses to the same chip select space will not toggle the chip select.
This is a problem for some hardware. So, I am creating a barrier
function that will read some other chip select address space in an
effort to toggle that chip select. Now, if my barrier function reads
memory out of cache, then I will not see a bus access. Which is what I
am seeing after I access memory allocated by MmMapIoSpace(some_address,
4, FALSE).
Any insight is greatly appreciated,
Eric
- Next message: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Previous message: MP: "Platform Builder 5.0 installation"
- Next in thread: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Reply: Steve Maillet \(eMVP\): "Re: MmMapIoSpace is it really uncached??"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
|