[PATCH] Cygwin: Allow to set SO_PEERCRED zero (v2)

[Mark Geisert]

The existing code errors as EINVAL any attempt to set a value for
SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
enable the workaround set_no_getpeereid behavior for Python one has
to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
no way to specify a NULL pointer for 'optval'.

This v2 of patch allows the original working (i.e., allow NULL,0 for
optval,optlen to mean turn off SO_PEERCRED) in addition to the new
working described above.  The sense of the 'if' stmt is reversed for
readability.

---
 winsup/cygwin/fhandler_socket_local.cc | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/winsup/cygwin/fhandler_socket_local.cc b/winsup/cygwin/fhandler_socket_local.cc
index c94bf828f..964f3e819 100644
--- a/winsup/cygwin/fhandler_socket_local.cc
+++ b/winsup/cygwin/fhandler_socket_local.cc
@@ -1430,10 +1430,14 @@ fhandler_socket_local::setsockopt (int level, int optname, const void *optval,
      FIXME: In the long run we should find a more generic solution
      which doesn't require a blocking handshake in accept/connect
      to exchange SO_PEERCRED credentials. */
-  if (optval || optlen)
-    set_errno (EINVAL);
-  else
+  /* Temporary: Allow SO_PEERCRED to only be zeroed. Two ways to
+     accomplish this: pass NULL,0 for optval,optlen; or pass the
+     address,length of an '(int) 0' set up by the caller. */
+  if ((!optval && !optlen) ||
+ (optlen == (socklen_t) sizeof (int) && !*(int *) optval))
     ret = af_local_set_no_getpeereid ();
+  else
+    set_errno (EINVAL);
   return ret;
 
  case SO_REUSEADDR:
--
2.29.2

[cygwin-patches mailing list]

On Dec  7 02:29, Mark Geisert wrote:

> The existing code errors as EINVAL any attempt to set a value for
> SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
> enable the workaround set_no_getpeereid behavior for Python one has
> to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
> no way to specify a NULL pointer for 'optval'.
>
> This v2 of patch allows the original working (i.e., allow NULL,0 for
> optval,optlen to mean turn off SO_PEERCRED) in addition to the new
> working described above.  The sense of the 'if' stmt is reversed for
> readability.
>
> ---
>  winsup/cygwin/fhandler_socket_local.cc | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/winsup/cygwin/fhandler_socket_local.cc b/winsup/cygwin/fhandler_socket_local.cc
> index c94bf828f..964f3e819 100644
> --- a/winsup/cygwin/fhandler_socket_local.cc
> +++ b/winsup/cygwin/fhandler_socket_local.cc
> @@ -1430,10 +1430,14 @@ fhandler_socket_local::setsockopt (int level, int optname, const void *optval,
>       FIXME: In the long run we should find a more generic solution
>       which doesn't require a blocking handshake in accept/connect
>       to exchange SO_PEERCRED credentials. */
> -  if (optval || optlen)
> -    set_errno (EINVAL);
> -  else
> +  /* Temporary: Allow SO_PEERCRED to only be zeroed. Two ways to
> +     accomplish this: pass NULL,0 for optval,optlen; or pass the
> +     address,length of an '(int) 0' set up by the caller. */
> +  if ((!optval && !optlen) ||
> + (optlen == (socklen_t) sizeof (int) && !*(int *) optval))
>      ret = af_local_set_no_getpeereid ();
> +  else
> +    set_errno (EINVAL);
>    return ret;
>  
>   case SO_REUSEADDR:
> --
> 2.29.2

Pushed


Thanks,
Corinna

[cygwin-patches mailing list]

On Dec  7 16:30, Corinna Vinschen via Cygwin-patches wrote:

> On Dec  7 02:29, Mark Geisert wrote:
> > The existing code errors as EINVAL any attempt to set a value for
> > SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
> > enable the workaround set_no_getpeereid behavior for Python one has
> > to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
> > no way to specify a NULL pointer for 'optval'.
> >
> > This v2 of patch allows the original working (i.e., allow NULL,0 for
> > optval,optlen to mean turn off SO_PEERCRED) in addition to the new
> > working described above.  The sense of the 'if' stmt is reversed for
> > readability.
> >
> > ---
> >  winsup/cygwin/fhandler_socket_local.cc | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> >
> > diff --git a/winsup/cygwin/fhandler_socket_local.cc b/winsup/cygwin/fhandler_socket_local.cc
> > index c94bf828f..964f3e819 100644
> > --- a/winsup/cygwin/fhandler_socket_local.cc
> > +++ b/winsup/cygwin/fhandler_socket_local.cc
> > @@ -1430,10 +1430,14 @@ fhandler_socket_local::setsockopt (int level, int optname, const void *optval,
> >       FIXME: In the long run we should find a more generic solution
> >       which doesn't require a blocking handshake in accept/connect
> >       to exchange SO_PEERCRED credentials. */
> > -  if (optval || optlen)
> > -    set_errno (EINVAL);
> > -  else
> > +  /* Temporary: Allow SO_PEERCRED to only be zeroed. Two ways to
> > +     accomplish this: pass NULL,0 for optval,optlen; or pass the
> > +     address,length of an '(int) 0' set up by the caller. */
> > +  if ((!optval && !optlen) ||
> > + (optlen == (socklen_t) sizeof (int) && !*(int *) optval))
> >      ret = af_local_set_no_getpeereid ();
> > +  else
> > +    set_errno (EINVAL);
> >    return ret;
> >  
> >   case SO_REUSEADDR:
> > --
> > 2.29.2
>
> Pushed

I created new developer snapshots for testing.


Corinna

[Mark Geisert]

Hi Corinna,

Corinna Vinschen via Cygwin-patches wrote:

> On Dec  7 16:30, Corinna Vinschen via Cygwin-patches wrote:
>> On Dec  7 02:29, Mark Geisert wrote:
>>> The existing code errors as EINVAL any attempt to set a value for
>>> SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
>>> enable the workaround set_no_getpeereid behavior for Python one has
>>> to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
>>> no way to specify a NULL pointer for 'optval'.
>>>
>>> This v2 of patch allows the original working (i.e., allow NULL,0 for
>>> optval,optlen to mean turn off SO_PEERCRED) in addition to the new
>>> working described above.  The sense of the 'if' stmt is reversed for
>>> readability.
>>>
>>> ---

[...]
>>> --
>>> 2.29.2
>>
>> Pushed
>
> I created new developer snapshots for testing.

I didn't phrase my comment somewhere about "future snapshot TBA" as I had
intended.  I just meant some future snapshot, not that I was requesting one for
this patch.  But thank you very much anyway.  I'll sort out with Marco whether the
Python end of the OP's issue patch should go into pythonX.X-test or pythonX.X
itself, separately.  BTW The whole set of Python tests might serve to test Cygwin
in a manner we haven't seen|used yet...
Cheers,

..mark

[cygwin-patches mailing list]

Hi Mark,

On Dec  7 19:25, Mark Geisert wrote:

> Hi Corinna,
>
> Corinna Vinschen via Cygwin-patches wrote:
> > On Dec  7 16:30, Corinna Vinschen via Cygwin-patches wrote:
> > > On Dec  7 02:29, Mark Geisert wrote:
> > > > The existing code errors as EINVAL any attempt to set a value for
> > > > SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
> > > > enable the workaround set_no_getpeereid behavior for Python one has
> > > > to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
> > > > no way to specify a NULL pointer for 'optval'.
> > > >
> > > > This v2 of patch allows the original working (i.e., allow NULL,0 for
> > > > optval,optlen to mean turn off SO_PEERCRED) in addition to the new
> > > > working described above.  The sense of the 'if' stmt is reversed for
> > > > readability.
> > > >
> > > > ---
> [...]
> > > > --
> > > > 2.29.2
> > >
> > > Pushed
> >
> > I created new developer snapshots for testing.
>
> I didn't phrase my comment somewhere about "future snapshot TBA" as I had
> intended.  I just meant some future snapshot, not that I was requesting one
> for this patch.  But thank you very much anyway.

I freely admit I didn't actually read your comment :}

I just created this snapshot because it seemed useful, so all is well :)


Corinna

[cygwin-patches mailing list]

On Dec  8 10:47, Corinna Vinschen via Cygwin-patches wrote:

> Hi Mark,
>
> On Dec  7 19:25, Mark Geisert wrote:
> > Hi Corinna,
> >
> > Corinna Vinschen via Cygwin-patches wrote:
> > > On Dec  7 16:30, Corinna Vinschen via Cygwin-patches wrote:
> > > > On Dec  7 02:29, Mark Geisert wrote:
> > > > > The existing code errors as EINVAL any attempt to set a value for
> > > > > SO_PEERCRED via setsockopt() on an AF_UNIX/AF_LOCAL socket.  But to
> > > > > enable the workaround set_no_getpeereid behavior for Python one has
> > > > > to be able to set SO_PEERCRED to zero.  Ergo, this patch.  Python has
> > > > > no way to specify a NULL pointer for 'optval'.
> > > > >
> > > > > This v2 of patch allows the original working (i.e., allow NULL,0 for
> > > > > optval,optlen to mean turn off SO_PEERCRED) in addition to the new
> > > > > working described above.  The sense of the 'if' stmt is reversed for
> > > > > readability.
> > > > >
> > > > > ---
> > [...]
> > > > > --
> > > > > 2.29.2
> > > >
> > > > Pushed
> > >
> > > I created new developer snapshots for testing.
> >
> > I didn't phrase my comment somewhere about "future snapshot TBA" as I had
> > intended.  I just meant some future snapshot, not that I was requesting one
> > for this patch.  But thank you very much anyway.
>
> I freely admit I didn't actually read your comment :}

...the part talking about a future snapshot, that is...

>
> I just created this snapshot because it seemed useful, so all is well :)
>
>
> Corinna