ibat/resolve: retry ibat resolution

Winverbs ND scale out testing showed that IBAT::Resolve() can
return E_PENDING, which requires that the resolution be retried.
A similar issue to this was seen when testing with the librdmacm.
Rather than duplicating retry logic in the winverbs ND provider,
add new functionality to ibat, with retry capability.  To
avoid breaking the ibat.dll interface, extend the API with a
new call ResolvePath() that takes a timeout value.

ResolvePath() automatically retries Resolve() while the result
is E_PENDING, until the request times out.  Modify the winverbs
ND provider to call ResolvePath().  Also update other places
where Resolve() is called in a loop: the librdmacm and wsd.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
git-svn-id: svn://openib.tc.cornell.edu/gen1@2703 ad392aa1-c5ef-ae45-8dd8-e69d62a5ef86

diff --git a/trunk/core/ibat/user/ibat.cpp b/trunk/core/ibat/user/ibat.cpp
index 69d918615789e871b287b5a82bbf978f252a7f0e..773bfd877fe8460c1a6b0b7afff85003f216fad9 100644 (file)
--- a/trunk/core/ibat/user/ibat.cpp
+++ b/trunk/core/ibat/user/ibat.cpp
@@ -358,9 +358,34 @@ Resolve(
     return S_OK;\r
 }\r
 \r
-#endif\r
+#endif // WINVER >= 0x600\r
+\r
+\r
+HRESULT\r
+ResolvePath(\r
+    __in const struct sockaddr* pSrcAddr,\r
+    __in const struct sockaddr* pDestAddr,\r
+    __out IBAT_PATH_BLOB* pPath,\r
+       __in DWORD Timeout)\r
+{\r
+       INT64 to;\r
+       HRESULT hr;\r
+\r
+       to = (Timeout == INFINITE) ? MAXINT64 : (INT64) ((UINT64) Timeout);\r
+       for (;;) {\r
+               hr = Resolve(pSrcAddr, pDestAddr, pPath);\r
+               if( hr != E_PENDING || to <= 0 )\r
+                       break;\r
+\r
+               to -= 10;\r
+               Sleep(10);\r
+       };\r
+\r
+       return hr;\r
 }\r
 \r
+} /* IBAT namespace */\r
+\r
 extern "C"\r
 {\r
 \r
@@ -374,4 +399,14 @@ IbatResolve(
     return IBAT::Resolve( pSrcAddr, pDestAddr, pPath );\r
 }\r
 \r
+HRESULT\r
+IbatResolvePath(\r
+    __in const struct sockaddr* pSrcAddr,\r
+    __in const struct sockaddr* pDestAddr,\r
+    __out IBAT_PATH_BLOB* pPath,\r
+       __in DWORD Timeout)\r
+{\r
+       return IBAT::ResolvePath(pSrcAddr, pDestAddr, pPath, Timeout);\r
+}\r
+\r
 } /* extern "C" */\r

diff --git a/trunk/inc/user/iba/ibat.h b/trunk/inc/user/iba/ibat.h
index c9a174059481380202dbc84a342c80bf705f6fcb..6e84486ca1738d3d9bc0d8a4021622e6c3fb3afd 100644 (file)
--- a/trunk/inc/user/iba/ibat.h
+++ b/trunk/inc/user/iba/ibat.h
@@ -52,6 +52,14 @@ Resolve(
     __out IBAT_PATH_BLOB* pPath\r
     );\r
 \r
+HRESULT\r
+ResolvePath(\r
+    __in const struct sockaddr* pSrcAddr,\r
+    __in const struct sockaddr* pDestAddr,\r
+    __out IBAT_PATH_BLOB* pPath,\r
+       __in DWORD Timeout      /* ms */\r
+    );\r
+\r
 }\r
 #else /* __cplusplus */\r
 \r
@@ -62,6 +70,14 @@ IbatResolve(
     __out IBAT_PATH_BLOB* pPath\r
     );\r
 \r
+HRESULT\r
+IbatResolvePath(\r
+    __in const struct sockaddr* pSrcAddr,\r
+    __in const struct sockaddr* pDestAddr,\r
+    __out IBAT_PATH_BLOB* pPath,\r
+       __in DWORD Timeout      /* ms */\r
+    );\r
+\r
 #endif /* __cplusplus */\r
 \r
-#endif // _IBAT_H_
\ No newline at end of file
+#endif // _IBAT_H_\r

diff --git a/trunk/ulp/librdmacm/src/cma.cpp b/trunk/ulp/librdmacm/src/cma.cpp
index cde309b881321206bbbb73204bd544d8d9989991..40e741136b9a0f2e4be8859024593033c69b21b7 100644 (file)
--- a/trunk/ulp/librdmacm/src/cma.cpp
+++ b/trunk/ulp/librdmacm/src/cma.cpp
@@ -506,25 +506,6 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
        return 0;\r
 }\r
 \r
-static int\r
-ucma_resolve_ibat_path(struct rdma_cm_id *id, int timeout_ms,\r
-                                          IBAT_PATH_BLOB *path)\r
-{\r
-       HRESULT hr;\r
-\r
-       do {\r
-               hr = IBAT::Resolve(&id->route.addr.src_addr, &id->route.addr.dst_addr,\r
-                                                  path);\r
-               if (hr != E_PENDING || timeout_ms <= 0) {\r
-                       break;\r
-               }\r
-               timeout_ms -= 10;\r
-               Sleep(10);\r
-       } while (timeout_ms > 0);\r
-\r
-       return hr;\r
-}\r
-\r
 __declspec(dllexport)\r
 int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)\r
 {\r
@@ -532,7 +513,8 @@ int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms)
        IBAT_PATH_BLOB path;\r
        HRESULT hr;\r
 \r
-       hr = ucma_resolve_ibat_path(id, timeout_ms, &path);\r
+       hr = IBAT::ResolvePath(&id->route.addr.src_addr, &id->route.addr.dst_addr,\r
+                                                  &path, timeout_ms);\r
        if (FAILED(hr)) {\r
                return hr;\r
        }\r

diff --git a/trunk/ulp/netdirect/user/nd_connect.cpp b/trunk/ulp/netdirect/user/nd_connect.cpp
index aa46adac41ce6b8187affee23c5d50b5b9066300..81d5f7bd09bf742bec3380027aaebdce0a705a59 100644 (file)
--- a/trunk/ulp/netdirect/user/nd_connect.cpp
+++ b/trunk/ulp/netdirect/user/nd_connect.cpp
@@ -138,12 +138,13 @@ Connect(INDEndpoint* pEndpoint,
        } else {\r
                addr.Sin6.sin6_port = LocalPort;\r
        }\r
-       hr = m_pWvConnEp->BindAddress(&addr.Sa);\r
+\r
+       hr = IBAT::ResolvePath(&addr.Sa, pAddress, &path, INFINITE);\r
        if (FAILED(hr)) {\r
                goto out;\r
        }\r
 \r
-       hr = IBAT::Resolve(&addr.Sa, pAddress, &path);\r
+       hr = m_pWvConnEp->BindAddress(&addr.Sa);\r
        if (FAILED(hr)) {\r
                goto out;\r
        }\r

diff --git a/trunk/ulp/wsd/user/ibsp_ip.c b/trunk/ulp/wsd/user/ibsp_ip.c
index a0afe89416cd5c287f143a12a118af65c72d8909..b0124bb419a78ba895b97c2d3f1aa5e0a9fe1829 100644 (file)
--- a/trunk/ulp/wsd/user/ibsp_ip.c
+++ b/trunk/ulp/wsd/user/ibsp_ip.c
@@ -270,20 +270,9 @@ query_guid_address(
        HRESULT hr;\r
 \r
        IBSP_ENTER( IBSP_DBG_HW );\r
+       hr = IbatResolvePath(p_src_addr, p_dest_addr, (IBAT_PATH_BLOB*)&path,\r
+               INFINITE);\r
 \r
-       for(;;)\r
-       {\r
-               hr = IbatResolve(\r
-                       p_src_addr,\r
-                       p_dest_addr,\r
-                       (IBAT_PATH_BLOB*)&path\r
-                       );\r
-\r
-               if( hr != E_PENDING )\r
-                       break;\r
-\r
-               Sleep( 100 );\r
-       }\r
        if( hr == S_OK )\r
        {\r
                *port_guid = path.dgid.unicast.interface_id;\r