This patch removes the assert at user-mode because:
1. h_al can be NULL when we got into the al_destroy_cep as a result of a bad
flow within qp_creation
2. h_al is only compared to other values; the code does not attempt to use it
as a pointer
Signed-off by: Alexander Naslednikov (xalex at mellanox.co.il)
Index: D:/windows/MLNX_VPI_trunk/core/al/user/ual_cm_cep.c
===================================================================
--- D:/windows/MLNX_VPI_trunk/core/al/user/ual_cm_cep.c (revision 5850)
+++ D:/windows/MLNX_VPI_trunk/core/al/user/ual_cm_cep.c (revision 5851)
@@ -382,9 +382,13 @@
DWORD bytes_ret;
AL_ENTER( AL_DBG_CM );
+
+#if DBG
+ if ( !h_al) {
+ AL_PRINT(TRACE_LEVEL_ERROR ,AL_DBG_ERROR , ("h_al == NULL, probably because
qp creation was failed.\n") );
+ }
+#endif
- CL_ASSERT( h_al );
-
cl_spinlock_acquire( &gp_cep_mgr->obj.lock );
if( *p_cid < cl_ptr_vector_get_size( &gp_cep_mgr->cep_vector ) )
{
_______________________________________________
ofw mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
