Memory marshalling in Windows CE

CE6 Drivers: What you need to know

Windows CE Base Team Blog — Fri, 10 Nov 2006 07:57:28 GMT

Posted by: Sue Loh One of the biggest concerns people have about the new CE6 release is backward compatibility.

re: Memory marshalling in Windows CE

Pop.Catalin — Fri, 10 Nov 2006 10:28:44 GMT

Great article! Thank you!

re: Memory marshalling in Windows CE

ce_base — Fri, 10 Nov 2006 20:01:14 GMT

You are very welcome. :-)

Sue

Application Compatibility in Windows CE 6.0

Windows CE Base Team Blog — Tue, 14 Nov 2006 21:42:42 GMT

Posted by: Upender Sandadi One of the goals for Windows CE 6.0 design was full backward compatibility

re: Memory marshalling in Windows CE

Asang Dani — Sun, 19 Nov 2006 17:35:43 GMT

Hello,

I have worked with variety of Operating Systems for long time and "buffer duplication" that has been introduced in CE 6 is something that I haven't seen elsewhere. I also feel that CE 6 has done a very good job with "API Registration" which also includes "Signature" identification with "Pointer parameters" access checking being taken over inside trap-handlers. That certainly avoids lot of code duplication and potential mistakes on part of driver developers. No other OS provides such elaborate and general mechanims which can be seen in "RPC" ( Remote Procedure Call ) area, but not in system software. So a great job.

At the same time I feel that CE 6 trap-handlers taking care of "pointer parameters" is going to make some people feel that "no validation" is needed and all pointers are always valid inside drivers. Of course it's not your fault. But doing things "under the hood" may make some DD programmer's forget that "someone" has to do it.

Regards,

Asang..

re: Memory marshalling in Windows CE

ce_base — Mon, 20 Nov 2006 19:24:03 GMT

Hello Asang, thanks for your feedback!

You're right, by automatically access checking pointer parameters, we run the risk that people will mistakenly assume they don't have to access check embedded pointers. On the other hand, if we didn't automatically do those access checks, we'd be completely trusting everyone (including ourselves) to get buffer usage 100% right. This way we at least close off one possible mistake. It's definitely not perfect but I think it's the best choice to make.

One thing that makes it slightly better is that user-mode drivers cannot use embedded pointers at all without first calling CeOpenCallerBuffer. Hopefully that will be enough to get people into the habit of using CeOpenCallerBuffer on embedded pointers, even in kernel-mode drivers.

Sue

re: Memory marshalling in Windows CE

Asang Dani — Tue, 21 Nov 2006 05:42:33 GMT

Hello Sue,

I agree with you completely. Overall, I am really happy to see the manner in which these changes were made in CE 6.0. At the same time, I am afraid MSDN documentation about CE 6 appears to be quite inadequate as of now. Secondly, there need to be one sample for each scenario for most people to make sense - of course I can write some samples myself, but if it comes from your team it's more authentic.

Regards,

Asang..

re: Memory marshalling in Windows CE

ce_base — Tue, 21 Nov 2006 19:12:59 GMT

As to the documentation, I am told that there is a doc update shortly forthcoming. So look forward to lots of improvements in the near future. (You should get one of those notifications that updates are available.)

I am sure our drivers team has good sample drivers for users to look at as examples of how to do the memory marshalling. I would have to ask them for good examples though. That is an excellent suggestion for another post; I hope to have something soon.

Sue

re: Memory marshalling in Windows CE

Jiaoshouwu — Wed, 22 Nov 2006 09:43:35 GMT

Good , good ,... articles!! Thanks Sue!!

From your introduction, "duplication" is one of the approaches to marshall memory as weell as "alias". Obviously, "alias" has higher performance than "duplication" for sparing memory-copy action. Therefore, in WinCE 6, "alias" is the de facto way to implement memory marsharl?

Besides, I cannot refer to the details for some APIs , such as CeAllocAsynchronousBuffer, in MSDN. Could u give me any suggestion about where to look up it? Thanks a lot!!

re: Memory marshalling in Windows CE

Jiao shouwu — Wed, 22 Nov 2006 11:17:47 GMT

'Secure copy' is the specific instance of "duplication" , right?

re: Memory marshalling in Windows CE

ce_base — Wed, 22 Nov 2006 17:58:50 GMT

Thanks!

Actually you might be surprised to learn that we found that aliasing is NOT always faster than duplication. On ARM at least, for small buffers (<8KB or so) duplication is faster than aliasing. ARM has added issues that we have to make the source and dest buffer uncached while aliasing, to avoid cache coherency problems. (Most ARM devices today have a virtually tagged cache, and memory must be uncached if you're going to alias it to two different virtual addresses.) Maybe the boundary is smaller on other CPUs, but I believe that for pretty small buffers we may actually do best by duplicating, not aliasing. So please don't make any assumptions about CeAllocAsynchronousBuffer. It is the OS' job to make the right trade offs for performance and security in these marshalling APIs.

I would say that 'secure copy' is duplication for the purpose of security, rather than for the purpose of marshalling.

I am sorry to hear that some of the APIs did not end up being documented properly in our shipping documentation. I've heard other people say that as well. What I also heard is that there is a documentation update coming soon -- it will show up as a notification that updates are available inside Platform Builder / VS.

Writing services for the WinCE 6.0 Kernel

Windows CE Networking Team WebLog — Tue, 28 Nov 2006 20:03:20 GMT

When rewriting services.exe for the CE 6.0 kernel , we tried to make it so that as few changes as possible

re: Memory marshalling in Windows CE

ce_base — Mon, 04 Dec 2006 18:38:22 GMT

http://www.windowsfordevices.com/articles/AT5831219184.html

MEMORY MARSHALLING IN WINDOWS CE 6.0

Dave Baker's WebLog — Fri, 08 Dec 2006 18:17:05 GMT

Windows Embedded CE 6.0 introduced a new, "third generation" kernel. Unsurprisingly, changes to the kernel

Hardcore pointer marshalling samples for Windows CE 6

Windows CE Networking Team WebLog — Wed, 03 Jan 2007 02:45:16 GMT

This is one of those blogs where I hesitated to post it because it may be going too hard core into the

Books and other resources for learning about Windows CE

Windows CE Base Team Blog — Fri, 12 Jan 2007 09:02:39 GMT

Posted by: Sue Loh I am occasionally asked whether I know any good books or other resources to help learn

Article sur le marshalling mémoire dans Windows CE 6.0 - Memory Marshalling in Windows CE 6.0 great article

Olivier Bloch, Blog embarqué — Thu, 05 Apr 2007 12:55:50 GMT

Dans la série des excellents articles de Sue Loh , voici celui qui présente le marshalling mémoire dans

MARSHAL_FORCE_ALIAS

Windows CE Base Team Blog — Fri, 06 Jun 2008 21:06:39 GMT

Posted by: Sue Loh Hello out there, it's been a long time since I posted anything real, and I feel sorry