Merge pull request #420 from hpc/fix-mdtest-verify

Bugfix: mdtest-verification, posix tolerate short read
hpc · Jun 5, 2022 · 07e2feb · 07e2feb
2 parents c4465be + 5f6bf7b
commit 07e2feb
Show file tree

Hide file tree

Showing 3 changed files with 16 additions and 9 deletions.
diff --git a/src/aiori-POSIX.c b/src/aiori-POSIX.c
@@ -720,21 +720,26 @@ static IOR_offset_t POSIX_Xfer(int access, aiori_fd_t *file, IOR_size_t * buffer
 #ifdef HAVE_GPU_DIRECT
                         }
 #endif
-                        if (rc == 0)
-                                ERRF("read(%d, %p, %lld) returned EOF prematurely",
-                                        fd, (void*)ptr, remaining);
-                        if (rc == -1)
-                                ERRF("read(%d, %p, %lld) failed",
-                                        fd, (void*)ptr, remaining);
+                        if (rc == 0){
+                          WARNF("read(%d, %p, %lld) returned EOF prematurely", fd, (void*)ptr, remaining);
+                          return length - remaining;
+                        }
+
+                        if (rc == -1){
+                          WARNF("read(%d, %p, %lld) failed", fd, (void*)ptr, remaining);
+                          return length - remaining;
+                        }
                 }
                 if (rc < remaining) {
                         WARNF("task %d, partial %s, %lld of %lld bytes at offset %lld\n",
                                 rank,
                                 access == WRITE ? "write()" : "read()",
                                 rc, remaining,
                                 offset + length - remaining);
-                        if (xferRetries > MAX_RETRY || hints->singleXferAttempt)
-                                ERR("too many retries -- aborting");
+                        if (xferRetries > MAX_RETRY || hints->singleXferAttempt){
+                          WARN("too many retries -- aborting");
+                          return length - remaining;
+                        }
                 }
                 assert(rc >= 0);
                 assert(rc <= remaining);

diff --git a/src/mdtest.c b/src/mdtest.c
@@ -733,6 +733,7 @@ void mdtest_read(int random, int dirs, const long dir_iter, char *path) {
             read_buffer[0] = 42;
             if (o.read_bytes != (size_t) o.backend->xfer(READ, aiori_fh, (IOR_size_t *) read_buffer, o.read_bytes, 0, o.backend_options)) {
                 WARNF("unable to read file %s", item);
+                o.verification_error += 1;
                 continue;
             }
             int pretend_rank = (2 * o.nstride + rank) % o.size;
@@ -2578,7 +2579,7 @@ mdtest_results_t * mdtest_run(int argc, char **argv, MPI_Comm world_com, FILE *
         if(rank == 0 && total_errors){
             VERBOSE(0, -1, "\nERROR: verifying the data on read (%lld errors)! Take the performance values with care!\n", total_errors);
         }
-
+        aggregated_results->total_errors += total_errors;
         MPI_Comm_free(&testComm);
     }
 

diff --git a/src/mdtest.h b/src/mdtest.h
@@ -27,6 +27,7 @@ typedef struct
     double time[MDTEST_LAST_NUM]; /* Time */
     double time_before_barrier[MDTEST_TREE_CREATE_NUM]; /* individual time before executing the barrier */
     uint64_t items[MDTEST_LAST_NUM]; /* Number of operations done in this process*/
+    uint64_t total_errors;
 
     /* Statistics when hitting the stonewall */
     double   stonewall_time[MDTEST_LAST_NUM];     /* Max runtime of any process until completion / hit of the stonewall */