Mark,
You could ask for as little as a page, but Mm will consume an entire 64K
address range for that one page (that’s an implementation detail, of
course) which is one reason why the 64K number is frequently used -
that’s the smallest contiguous chunk Mm will use.
The cache manager uses 256KB views (on x86, I haven’t verified this is
the same on x64) because *it* has a limited addressable range (x86 -
2048 simultaneous views * 256KB/view = 512MB). Were I coding this
today, I wouldn’t choose less than 64KB. If you are reviewing the
entire file, it might make sense to choose a larger size - like 1MB.
Much would depend upon the target platform and the number of
simultaneous files being accessed plus the limitations on the size of
the address space. You might even want to use a fall-back algorithm
(try a big size and if there’s no address space, fall down to a smaller
size)
So, choose a value that fits for your implementation and you’ll be fine.
If you are happy with 64KB, go for it (I am curious where your address
space is being consumed - mapped view space is not plentiful in kernel
mode, but being unable to do < 256KB seems rather unusual. Are you
mapping a lot of files?)
One trick to extend the address space size is to map them into user
space for a particular process. For example, there’s usually a 2GB
address space between 0-2GB range that people don’t use in the system
process. I also know of products that spawn their own separate system
process precisely to get an empty 2GB address space. The disadvantage
of course is that you must be IN the correct process context to access
those addresses.
This won’t be a big problem in x64 though, because all of these older
limits have been dramatically relaxed.
Regards,
Tony
Tony Mason
Consulting Partner
OSR Open Systems Resources, Inc.
http://www.osr.com
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of Mark Hahn
Sent: Tuesday, July 12, 2005 4:29 AM
To: ntfsd redirect
Subject: Re:[ntfsd] ZwMapViewOfSection on compressed file?
I just figured out that the AllocationSize value from the
FileStandardInformation struct is not the granularity, like I thought,
but
rather the entire allocation for the file. I’m surprised my code worked
as
well as it did. I guess the total allocation would usually (for
non-compressed files) be a multiple of the granularity so my code
worked,
even though it was mapping much bigger views of sections than it needed.
This still leaves the question of what I should use for the allocation
granularity. How do I query that? The docs implied I could pass any
value
to ZwMapViewOfSection and it would adjust for the granularity, but I
found
out right away that I needed to adjust the section offset myself before
calling it.
I just tried 65K and it seems to be working, but it bothers me that I
don’t
really know the correct value. I got the 65K number from several places
on
the web that talked like that was the correct value.
“Mark Hahn” wrote in message news:xxxxx@ntfsd…
> I’m have a problem with a ZwMapViewOfSection call on a particular
file.
> My code has worked with thousands of other calls for other files. I
> noticed that the file is compressed so I’m wondering if there is
something
> I should know about ZwMapViewOfSection with compressed files.
>
> The error returned is STATUS_MAPPED_ALIGNMENT, which says my base
address
> or section offset is not aligned to an allocation size boundary.
>
> For the base address I am passing a null ptr to let the system assign
it.
> I get the AllocationSize value from a FileStandardInformation class
query
> using FltQueryInformationFile call. I then adjust my section offset
to be
> a multiple of the allocation size.
>
> The weird thing is that the AllocationSize I am getting from
> FileStandardInformation, and using for the alignment, is 0x0001c000,
which
> seems like a really strange size. In any case the offset I want is
> 0x00021000 so aligning it amounts to using the section offset value of
> 0x0001c000, which is certainly aligned with 0x0001c000.
>
> Can anyone shed any light on this problem or suggest how I should
proceed?
>
>
—
Questions? First check the IFS FAQ at
https://www.osronline.com/article.cfm?id=17
You are currently subscribed to ntfsd as: xxxxx@osr.com
To unsubscribe send a blank email to xxxxx@lists.osr.com
