Re: ntdll: Introduce new function wine_server_set_reply_user for user provided pointers.

[Alexandre Julliard]

Sebastian Lackner  writes:
> This patch is an alternative strategy for fixing the issue described
> in http://source.winehq.org/patches/data/106864. Please note that the
> patches 2-4 from the previous series are still valid, and can be
> applied independently.
>
> Although Andre_H and ehoover told me that this solution will not be
> accepted, I'll give it a try, because its the cleanest approach I can
> think of.
>
> There is a huge amount of wineserver calls that use
> wine_server_set_reply incorrectly, and manually adding code to all
> these functions seems very complicated. In ntdll we have to use
> virtual_check_buffer_for_write, everywhere else we have to use
> IsBadWritePtr. And even if we have added that, it could still break
> things: When we add the check at the top of each affected function,
> and return some general error codes, it could break apps which expect
> a page fault exception in some other code branch.

There are few places that need this, and even fewer where it's an actual
problem in practice. In fact it wouldn't be a problem at all except for
your writecopy hack, which is in itself a questionable change.  You'll
need more convincing arguments for this series.