[PATCH 2/2] bcrypt: Return STATUS_INVALID_PARAMETER when invalid flag passed to BCryptCreateHash.

[Brendan McGrath]

I just sent a patch to testbot - the patch tests using the 
BCRYPT_HASH_REUSABLE_FLAG and expects a return code of 
STATUS_INVALID_PARAMETER:
https://testbot.winehq.org/JobDetails.pl?Key=48735

This passes on everything but Windows 8 and 10. On Win8 and 10, 
BCRYPT_HASH_REUSABLE_FLAG is implemented - so the return code is 
STATUS_SUCCESS.

My patch to the mailing list is based on a change I made to my local 
build to fix an issue while running .NET Core. I can see in the .NET 
Core code, that it passes BCRYPT_HASH_REUSABLE_FLAG to BCryptCreateHash 
and expects STATUS_INVALID_PARAMETER if the OS is pre-Win8 (I'm using 
Win7). That code can be seen here (it's under the MIT licence):
https://github.com/dotnet/corefx/blob/master/src/Common/src/Internal/Cryptography/HashProviderCng.cs#L40

So until BCRYPT_HASH_REUSABLE_FLAG is implemented - maybe it is better 
to return STATUS_INVALID_PARAMETER?

On 5/3/19 9:07 pm, Hans Leidekker wrote:
> On Tue, 2019-03-05 at 20:33 +1100, Brendan McGrath wrote:
>> OK - so it should return STATUS_NOT_IMPLEMENTED when
>> BCRYPT_HASH_REUSABLE_FLAG is passed and STATUS_INVALID_PARAMETER otherwise?
> Yes, though note that your test is not exhaustive, it checks just one
> flag while your fix rejects all of them. There might be undocumented
> flags, and there's always the possibility that new flags are added.
>
> It's very unlikely that applications depend on the exact error code, so
> it's not worth spending a lot of time getting these checks exactly
> right.
>