[PATCH 0/1] MR4386: rpcrt4: Support for explicit handle passing

[Haoyang Chen (＠chenhaoyang)]

If explicit_handle is defined in the *.idl file, c/s are uses explicit handles,
then an explicit handle must be passed in to the server-side  interface.

[PATCH 1/1] rpcrt4: Support for explicit handle passing

[Haoyang Chen]

From: Haoyang Chen

If explicit_handle is defined in the *.idl file, c/s are uses explicit handles,
then an explicit handle must be passed in to the server-side  interface.
---
 dlls/rpcrt4/ndr_stubless.c | 63 +++++++++++++++++++++-----------------
 1 file changed, 35 insertions(+), 28 deletions(-)

diff --git a/dlls/rpcrt4/ndr_stubless.c b/dlls/rpcrt4/ndr_stubless.c
index 2e84d64bba5..b4d7ede2d0a 100644
--- a/dlls/rpcrt4/ndr_stubless.c
+++ b/dlls/rpcrt4/ndr_stubless.c
@@ -1365,6 +1365,32 @@ LONG WINAPI NdrStubCall2(

TRACE("Oi_flags = 0x%02x\n", pProcHeader->Oi_flags);

+    if (pProcHeader->Oi_flags & Oi_OBJECT_PROC)
+        NdrStubInitialize(pRpcMsg, &stubMsg, pStubDesc, pChannel);
+    else
+        NdrServerInitializeNew(pRpcMsg, &stubMsg, pStubDesc);
+
+    /* create the full pointer translation tables, if requested */
+    if (pProcHeader->Oi_flags & Oi_FULL_PTR_USED)
+        stubMsg.FullPtrXlatTables = NdrFullPointerXlatInit(0,XLAT_SERVER);
+
+    /* store the RPC flags away */
+    if (pProcHeader->Oi_flags & Oi_HAS_RPCFLAGS)
+        pRpcMsg->RpcFlags = ((const NDR_PROC_HEADER_RPC *)pProcHeader)->rpc_flags;
+
+    /* use alternate memory allocation routines */
+    if (pProcHeader->Oi_flags & Oi_RPCSS_ALLOC_USED)
+#if 0
+          NdrRpcSsEnableAllocate(&stubMsg);
+#else
+          FIXME("Set RPCSS memory allocation routines\n");
+#endif
+
+    TRACE("allocating memory for stack of size %x\n", stack_size);
+
+    args = calloc(1, stack_size);
+    stubMsg.StackTop = args; /* used by conformance of top-level objects */
+
     /* binding */
     switch (pProcHeader->handle_type)
     {
@@ -1373,8 +1399,15 @@ LONG WINAPI NdrStubCall2(
         switch (*pFormat) /* handle_type */
         {
         case FC_BIND_PRIMITIVE: /* explicit primitive */
-            pFormat += sizeof(NDR_EHD_PRIMITIVE);
-            break;
+            {
+                const NDR_EHD_PRIMITIVE *pDesc = (const NDR_EHD_PRIMITIVE *)pFormat;
+                if (pDesc->flag)
+                    **(handle_t **)ARG_FROM_OFFSET(stubMsg.StackTop, pDesc->offset) = pRpcMsg->Handle ;
+                else
+                    *(handle_t *)ARG_FROM_OFFSET(stubMsg.StackTop, pDesc->offset) = pRpcMsg->Handle;
+                pFormat += sizeof(NDR_EHD_PRIMITIVE);
+                break;
+            }
         case FC_BIND_GENERIC: /* explicit generic */
             pFormat += sizeof(NDR_EHD_GENERIC);
             break;
@@ -1396,32 +1429,6 @@ LONG WINAPI NdrStubCall2(
         RpcRaiseException(RPC_X_BAD_STUB_DATA);
     }

-    if (pProcHeader->Oi_flags & Oi_OBJECT_PROC)
-        NdrStubInitialize(pRpcMsg, &stubMsg, pStubDesc, pChannel);
-    else
-        NdrServerInitializeNew(pRpcMsg, &stubMsg, pStubDesc);
-
-    /* create the full pointer translation tables, if requested */
-    if (pProcHeader->Oi_flags & Oi_FULL_PTR_USED)
-        stubMsg.FullPtrXlatTables = NdrFullPointerXlatInit(0,XLAT_SERVER);
-
-    /* store the RPC flags away */
-    if (pProcHeader->Oi_flags & Oi_HAS_RPCFLAGS)
-        pRpcMsg->RpcFlags = ((const NDR_PROC_HEADER_RPC *)pProcHeader)->rpc_flags;
-
-    /* use alternate memory allocation routines */
-    if (pProcHeader->Oi_flags & Oi_RPCSS_ALLOC_USED)
-#if 0
-          NdrRpcSsEnableAllocate(&stubMsg);
-#else
-          FIXME("Set RPCSS memory allocation routines\n");
-#endif
-
-    TRACE("allocating memory for stack of size %x\n", stack_size);
-
-    args = calloc(1, stack_size);
-    stubMsg.StackTop = args; /* used by conformance of top-level objects */
-
     /* add the implicit This pointer as the first arg to the function if we
      * are calling an object method */
     if (pThis)

[Zebediah Figura (＠zfigura)]

Can you please add a test for this? I *think* it should be possible to extend the existing server.idl with another function.

It turns out that, somewhat surprisingly, FC_BIND_PRIMITIVE really is different from FC_BIND_GENERIC and FC_BIND_CONTEXT. The latter two are written into the proc format string (and so are handled during unmarshalling), whereas FC_BIND_PRIMITIVE isn't. It may be worth adding a comment to that effect.

There's a whitespace error in one of the added lines (space before semicolon).

[Haoyang Chen (＠chenhaoyang)]

On Mon Nov 20 10:35:29 2023 +0000, Zebediah Figura wrote:
> Can you please add a test for this? I *think* it should be possible to
> extend the existing server.idl with another function.
> It turns out that, somewhat surprisingly, FC_BIND_PRIMITIVE really is
> different from FC_BIND_GENERIC and FC_BIND_CONTEXT. The latter two are
> written into the proc format string (and so are handled during
> unmarshalling), whereas FC_BIND_PRIMITIVE isn't. It may be worth adding
> a comment to that effect.
> There's a whitespace error in one of the added lines (space before semicolon).

Hi, There is a space that need to be removed. I wrote a test that may be a little lacking. Take a look at it.[1.diff](/uploads/e9ee2eb7558eaabe6b3c984d38dc18dc/1.diff)

[Zebediah Figura (＠zfigura)]

On Mon Nov 20 10:35:29 2023 +0000, Haoyang Chen wrote:
> Hi, There is a space that need to be removed. I wrote a test that may be
> a little lacking. Take a look at it.[1.diff](/uploads/e9ee2eb7558eaabe6b3c984d38dc18dc/1.diff)

Yes, that looks mostly good, thanks. Note that your EXTRAIDLFLAGS are broken though.

[PATCH 0/1] MR4386: rpcrt4: Support for explicit handle passing - closed

[Haoyang Chen (＠chenhaoyang)]

This merge request was closed by Haoyang Chen.

[Haoyang Chen (＠chenhaoyang)]

I will send v2 patch.