Improve company productivity with a Business Account.Sign Up

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 858
  • Last Modified:

VirtualAlloxEx ReadProcessMemory WriteProcessMemory

Process1 reads data to memory and places pointers on a list and remains open

Process 2 needs to access the data from process 1.

Is it possible to save the pointers only of data in a process so the next process can access the data without re-reading the data into a buffer?  I dont want to re-read the data in memory of process1 back into memory again in process2 as that defeats the object, I just want access.

 I had a great post answered yesterday from subn0wa

http://www.experts-exchange.com/Programming/Editors_IDEs/C_CPP_CS/Q_24382024.html

but Im stuck again on whether the above can be done?

If it can could someone provide me with an example?  Thanks
0
TheMoog
Asked:
TheMoog
1 Solution
 
DanRollinsCommented:
Any pointer into that memory would be meaningless/useless to an external process. For instance, if the external process wanted to pass a string from the shared memory to an API function, it would need to copy it to its own address space first -- so that it could pass a process-valid address to the API function.
As to access of the data... for instance, if you need a integer at offset 0x00001234 in the shared memory block, you need to copy it to a variable, anyway.
Shared memory blocks in a DLL can be accessed by all processes that use that same DLL.  I don't know if that buys you anything -- the DLL would need to provide access to the data, which (again) would appear to defeat the purpose -- at some point the data needs to be copied to the local process, doesn't it?
If you describe your needs in more detail (i.e., what will you do with the shared memory block data) then I might be able to provide some additional ideas.
0
 
DanRollinsCommented:
See:
   Creating Named Shared Memory
    http://msdn.microsoft.com/en-us/library/aa366551(VS.85).aspx
...and...
    Using Shared Memory in a Dynamic-Link Library
    http://msdn.microsoft.com/en-us/library/ms686958.aspx
for a technique that uses CreateFileMapping and MapViewOfFile so that multiple processes can access the same data. But, as in all other scenarios, some data eventually needs to be copied to the requesting process.
0
 
MysidiaCommented:
ReadProcessMemory/WriteProcessMemory  are  debugging functions.

You can't just take a pointer from another process and treat it as mapped into your own memory.   You need to setup a shared memory segment.

To do that, you either use native Windows shared memory calls or a third-party framework, tool, or library.     And you attach to the shared memory.

e.g.
SharedMemoryLib:
http://www.inf.pucrs.br/~pinho/shared_memory_library.htm

Microsoft Article on shared memory
http://msdn.microsoft.com/en-us/library/aa366551(VS.85).aspx

the boost library:
http://www.mathematik.uni-ulm.de/help/boost_1_36_0/interprocess/sharedmemorybetweenprocesses.html#interprocess.sharedmemorybetweenprocesses.sharedmemory.windows_shared_memory
http://www.boost.org/doc/libs/1_38_0/doc/html/interprocess/managed_memory_segments.html


http://www.codeproject.com/KB/cpp/Shared_Memory_Heap_Class.aspx

0
Get your problem seen by more experts

Be seen. Boost your question’s priority for more expert views and faster solutions

 
TheMoogAuthor Commented:
Thanks all.  Mysidia, those are great links Ive speedread so I can make a post but will absorb the detail over this week, I see some very credible options,  I didnt know the NAME of what it was I was trying to look for, its always the way I suppose, so its called shared memory, thank you.

What Im attempting to do is read my data to memory on a head node of a computer clustrer and have each instance of the data crunching client app "see the data" and act upon it accordingly.  Im wondering whether shared memory can be seen across servers?  Is this possible, I want the sub-nodes(the other servers) to have access to the shared memory, is that possible?

Also the first link you sent was very simple which is great, it shares arrays and my data is in arrays which is cool.  my arrays end up being pointers in a list and there are a few lists on a list of lists if that makes sense so theres an overall pointer that just accesses the data in memory, is there a way of remembering this structure so that I only call the pointer from shared memory?

As I say Ill absorb the links in more detail, Ive upped teh points as well as I ve asked a couple of side questions, thanks for your support.
0
 
Duncan RoeSoftware DeveloperCommented:
Other servers cannot see your shared memory. You could mmap() a file for writing as well as reading - other servers could see the file if it were NFS shared. I would not let other servers write to it without some agreed protocol between all servers.
0
 
MysidiaCommented:
This is a place where one would normally use a RDBM like Oracle or postgresql for typical application loads,  or sockets, to exchange state info between apps.

How to actually go about this actuallly depends on your needs ands the hardware you will choose; whether you just have a bunch of standalone PCs running normal OSes, or if you actually have specialized clustering hardware  and/or software.

The better performance you need, the hardware the work will be.

Specialized multi-processing  environments hardware/software may actually provide you shared memory facilities, e.g. Kerrighed, OpenSSI, Treadmarks, DIPC.

The feature is called DSM or  "Distributed Shared Memory";  there are many ways you can accomplish memory sharing, from true DSM, to something like a shared file on a server and memcache  for speeding reads of infrequently written data,  but they all involve a fair deal of work.

Implementing a Shared Memory Cluster across multiple independent computers is quite a bit harder than implementing shared memory on one machine.

You'll have to deal with issues like concurrency between your apps; there will be issues if two apps try to write to the same place at once, and you don't have some mechanisms to ensure adequate  read-write consistency,   typical threading issues,  except,  now your lock status has to be persistent across different machines, and surprising issues may show up, e.g. application on another machine crashes, but leaves a record locked...

And ..what happens if your storage server reboots/crashes on you?
Locking issues can be trickier.   With a shared FS, you might go for SCSI-3
reservations, however.



For apps with high performance requirements, NFS (indeed), or a SAN with a clustered filesystem may be your easiest  options.  mmap()'ing a file into memory,  means  the memory addresses will correspond to bytes in the file.


Fastest/robust  would probably be to have specialized distributed filesystems
here ala the Google File system (which is unfortunately proprietary and
not available), and use sockets for most communication...


This all depends on the scale of your app and its requirements.
A  NFS server is great if you don't mind single points of failure.

You might need a more complex architecture though, for a HA cluster.





0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

Join & Write a Comment

Featured Post

The 14th Annual Expert Award Winners

The results are in! Meet the top members of our 2017 Expert Awards. Congratulations to all who qualified!

Tackle projects and never again get stuck behind a technical roadblock.
Join Now