Re: Second service: ntsec-patch14

[Corinna Vinschen <corinna at vinschen dot de>]

Corinna Vinschen wrote:
> Chris Faylor wrote:
> > One comment, though: It looks like this could substantially increase the
> > size of cygwin's shared memory area.  I wonder if it would make sense to
> > have a separate area for this which would only be allocated when
> > cygwin=ntsec and when you're not running on Windows 9x.
> 
> Good question. Personally, I don't see the need for this but the real
> problem is, I can't (and won't) examine the code while on vacation ;-)
> 
> Nevertheless I'd like to discuss this next month, ok?

I have taken a look into the sources and the problem is the pinfo
class, isn't it?

The sizeof(pinfo) is grown from 988 to 1324 bytes, that's an increase
of 336 bytes per cygwin process or summarized 43008 bytes for
cygwin_shared (assuming PSIZE=128).

I'm not convinced that it's needful to create an additional shared mem
area if it only could save 42K bytes of memory for some cygwin users.

But I would like to suggest this additional shared mem area, too,
if this area is used to hold all kind of user specific information:

- UID
- primary GID
- name
- SID
- domain
- logon server

and, last but not least

- user's private mount table

Sure I have forgotten something but the advantages should be clear:

- User info would be saved only one time in the dll's workspace instead
  of multiple times.

- The mount info would be multi user aware which isn't yet.

- 9X users would have the space saving as well because they could get
  a shared user memory for only one user while NT or W2K could get this
  for, say, 16 users or alike.

Comments?

Corinna