[openib-general] [PATCH 1/2] ib_cm: cm_destroy_id() cleanup

Sean Hefty mshefty at ichips.intel.com
Tue Jul 18 09:54:08 PDT 2006


Arne Redlich wrote:
> In its current incarnation, cm_destroy_id() will not send a REJ if fed a
> magic number (err = -ENOMEM). This patch replaces this magic number with
> a more generic "reject" parameter.

The intent here is that if the user destroys their cm_id in the middle of 
connection establishment, then the reject code is consumer reject.  The 
cm_destroy_id() interface is generic, taking an error code as input.  Since we 
don't know what state the cm_id is in when the call is made, I'm not sure that 
changing to a reject code is any more generic.  That said, the err parameter is 
currently only used to reject a connection.

See other response for an additional comment.

- Sean




More information about the general mailing list