Using FltDoCompletionProcessingWhenSafe with IRP_MJ_READ

NTFSD
1

Hi,

I’m looking at swapBuffers Microsoft minifilter sample and it calls FltDoCompletionProcessingWhenSafe in Read Post operation:

https://github.com/Microsoft/Windows-driver-samples/blob/master/filesys/miniFilter/swapBuffers/swapBuffers.c#L1131

However the description of that function says:

https://msdn.microsoft.com/en-us/library/windows/hardware/ff542047(v=vs.85).aspx

“Caution To avoid deadlocks, FltDoCompletionProcessingWhenSafe cannot be called for I/O operations that can be directly completed by a driver in the storage stack, such as the following:
IRP_MJ_READ
IRP_MJ_WRITE
IRP_MJ_FLUSH_BUFFERS”

Is this a bug in the sample code? If yes, what is the proper way to access the buffer in the Read handler?

Thank you!

2

I believe the documentation is misleading here.

The warning doesn’t really have anything to do with
FltDoCompletionProcessingWhenSafe. They just don’t want you doing I/O in the
context of a I/O PostOp callback. FltDoCompletionProcessingWhenSafe calls
your Safe callback immediately if the IRQL is <= APC_LEVEL, which might then
cause you to think it’s “safe” to perform I/O back into the file system.

See the following two posts from Alex:

http://fsfilters.blogspot.com/2010/11/some-thoughts-on-fltdocompletionprocess.html
http://fsfilters.blogspot.com/2010/12/more-thoughts-on-fltdocompletionprocess.html

The sample is pretty safe here because it’s unlikely that the
FltLockUserBuffer would recurse back into the file system for file I/O
(maybe paging file, but that’s always different anyway).

While not directly related, in general if I need to map the user buffer I
prefer to do it in PreOp. That way by the time I get to my PostOp callback I
have everything I need to do my work and it’s a bit tidier to clean up on
error.

-scott
OSR
@OSRDrivers

3

Thanks Scott, much appreciated!