Windows: buffers whose offsets are beyond EOF should be zero filled and locally allocated
authorJeffrey Altman <jaltman@your-file-system.com>
Sat, 27 Mar 2010 20:13:27 +0000 (16:13 -0400)
committerJeffrey Altman <jaltman@openafs.org>
Sun, 28 Mar 2010 21:01:19 +0000 (14:01 -0700)
When a buffer is being allocated for an offset that exceeds the
file length as known to the file server, that buffer should be zero
filled and it does not require server validation.

Previously all buffers were populated with a FetchData call.
This is wasteful of time and server resources when there is a
valid callback registration because the serverLength is known
to be valid.

LICENSE MIT

Change-Id: I13f9a8dbd6387c3ba71638a682e995d7a7d0862f
Reviewed-on: http://gerrit.openafs.org/1664
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

src/WINNT/afsd/cm_dcache.c

index 1f47843..32dcf29 100644 (file)
@@ -457,7 +457,7 @@ long cm_StoreMini(cm_scache_t *scp, cm_user_t *userp, cm_req_t *reqp)
 
 long cm_BufRead(cm_buf_t *bufp, long nbytes, long *bytesReadp, cm_user_t *userp)
 {
-    *bytesReadp = cm_data.buf_blockSize;
+    *bytesReadp = 0;
 
     /* now return a code that means that I/O is done */
     return 0;
@@ -535,6 +535,8 @@ int cm_HaveBuffer(cm_scache_t *scp, cm_buf_t *bufp, int isBufLocked)
         return 1;
     if (bufp->dataVersion <= scp->dataVersion && bufp->dataVersion >= scp->bufDataVersionLow)
         return 1;
+    if (bufp->offset.QuadPart >= scp->serverLength.QuadPart)
+        return 1;
     if (!isBufLocked) {
         code = lock_TryMutex(&bufp->mx);
         if (code == 0) {