Re: [PATCH v2 0/25] MR7226: ntdll: Implement in-process synchronization via the Linux "ntsync" driver.

[Elizabeth Figura (＠zfigura)]

On Thu Feb  6 12:19:19 2025 +0000, Rémi Bernon wrote:
> This effectively leaks the handle? We have the fd cached, why do we care
> about the handle? What about something like that:
>         if (ret || server_get_unix_fd( handle, 0, &fd, &needs_close,
> NULL, NULL )) fd = -1;
>         if (InterlockedCompareExchange( &device, fd, -2 ) != -2 && fd !=
> -1 && needs_close) close( fd );
>         NtClose( handle );

That's not really legal unless you make assumptions about the fd cache. Generally speaking, closing a handle may also close its cached fd.

This *would* work since we don't mark the inproc sync handles cacheable, but there's not actually any reason they can't be cacheable, and making that assumption feels uncomfortable.