Remove IRP_MJ_ACQUIRE_FOR_SECTION_SYNCHRONIZATION from the list of "g…

[rodwiddowson]

Add explanations about IRP_MJ_ACQUIRE_FOR_SECTION_SYNCHRONIZATION.

Specifically that it has a guaranteed single return value and
the OS versions for which it is supported (the last being a guess and TBC).

This is the stuff I would need before I could add this to a product.

[cgallred]

The new change is correct. What is no longer correct is the requirement that the I/O has to be an IRP-based operation. However, that is only the case for this operation.

You'll never get STATUS_PENDING when calling FltCheckOplockEx for IRP_MJ_ACQUIRE_FOR_SECTION_SYNCHRONIZATION. It only breaks R and RH oplocks and the break does not require acknowledgment. See Types of Opportunistic Locks, third paragraph of the "Read-Handle Opportunistic Locks" (the one after the Note). That specifically talks about what happens if a write operation breaks the oplock, but the behavior is the same for ACQUIRE_FOR_SECTION_SYNCHRONIZATION.

[rodwiddowson]

Thanks @cgallred very much appreciated. I have reprurposed this PR to explain this and also made a guess about supported versions.

[lorihollasch]

@rodwiddowson, thanks for the update. We'll need to remove the last sentence with "TBC" in it; the minimum supported client version is reflected in the metadata. Otherwise, I'll await @cgallred's approval before merging (he's likely super busy prepping for next week's PlugFest!).