When IOR deletes an MPI-IO file, it might not delete it

[roblatham00]

ROMIO allows you to prefix a file with something to specify a file system driver and override ROMIO's automatic detection. We might do this if, for example, PVFS is available, but the kernel module is not installed (pvfs:/path/to/pvfs/testfile), or if we simply wish to avoid the overhead of stat (ufs:/regular/unix/file_system)

When IOR attempts to remove a file with such a prefix, it makes a call to the POSIX routine access(), but does not remove the prefix before doing so. access() does not return F_OK, naturally, and so does not attempt to remove the file. IOR doesn't complain about this, so we ended up thinking we were running tests of "create a new file" when actually we were "reusing existing file."

[roblatham00]

Here's a patch:

From fc6f65e81098075832e53d643e0187946b9e639d Mon Sep 17 00:00:00 2001
From: Rob Latham <[email protected]>
Date: Wed, 19 Feb 2014 10:16:08 -0600
Subject: [PATCH] A more MPI-IO friendly remove file

MPI-IO allows prefxing file names with something-something: (e.g. pvfs2:
or bglockless:). These prefixed file names, however, do not make sense
to posix calls like access().  So, if we must use access(), strip the
prefix first.
---
 src/ior.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/src/ior.c b/src/ior.c
index 163a59c5..6026414e 100644
--- a/src/ior.c
+++ b/src/ior.c
@@ -1254,6 +1254,12 @@ static void PrintRemoveTiming(double start, double finish, int rep)
 static void RemoveFile(char *testFileName, int filePerProc, IOR_param_t * test)
 {
         int tmpRankOffset;
+        char *strippedFileName;
+
+        strippedFileName = strchr(testFileName, ':');
+        if (strippedFileName > testFileName + 1)
+                strippedFileName = strippedFileName + 1;
+
         if (filePerProc) {
                 /* in random tasks, delete own file */
                 if (test->reorderTasksRandom == TRUE) {
@@ -1261,7 +1267,7 @@ static void RemoveFile(char *testFileName, int filePerProc, IOR_param_t * test)
                         rankOffset = 0;
                         GetTestFileName(testFileName, test);
                 }
-                if (access(testFileName, F_OK) == 0) {
+                if (access(strippedFileName, F_OK) == 0) {
                         backend->delete(testFileName, test);
                 }
                 if (test->reorderTasksRandom == TRUE) {
@@ -1269,7 +1275,7 @@ static void RemoveFile(char *testFileName, int filePerProc, IOR_param_t * test)
                         GetTestFileName(testFileName, test);
                 }
         } else {
-                if ((rank == 0) && (access(testFileName, F_OK) == 0)) {
+                if ((rank == 0) && (access(strippedFileName, F_OK) == 0)) {
                         backend->delete(testFileName, test);
                 }
         }
-- 
1.8.3.2

[roblatham00]

I should have added it was poor Florin Isaila [email protected] who encountered this and came up with the fix. I just updated the patch to apply onto master and to match IOR's coding conventions.

[roblatham00]

I found a problem with the patch. See the pull request I'm just about to generate.