fsstress: avoid infinite loops in clonerange_f
authorDarrick J. Wong <darrick.wong@oracle.com>
Fri, 23 Feb 2018 16:22:24 +0000 (08:22 -0800)
committerEryu Guan <eguan@redhat.com>
Sat, 24 Feb 2018 04:04:55 +0000 (12:04 +0800)
Evidently ceph will report a 4M blocksize, which trips clonerange_f's
clumsy attempt to avoid reflinking an extent on top of itself.  The
original code assumed that "pick a random destination up to 1MB past the
end of the file" would suffice, but that clearly won't with a 4M
blocksize.  Instead, we'll change it to 1024*blksize.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Tested-by: Luis Henriques <lhenriques@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
ltp/fsstress.c

index 935f5de2d43cabe261d79bb08af171d338d0a568..e1070997f33e72980a42cf9a74d410eed3e08b70 100644 (file)
@@ -2222,6 +2222,7 @@ clonerange_f(
        off64_t                 lr;
        off64_t                 off1;
        off64_t                 off2;
+       off64_t                 max_off2;
        size_t                  len;
        int                     v1;
        int                     v2;
@@ -2305,9 +2306,10 @@ clonerange_f(
         * If srcfile == destfile, randomly generate destination ranges
         * until we find one that doesn't overlap the source range.
         */
+       max_off2 = MIN(stat2.st_size + (1024ULL * stat2.st_blksize), MAXFSIZE);
        do {
                lr = ((int64_t)random() << 32) + random();
-               off2 = (off64_t)(lr % MIN(stat2.st_size + (1024 * 1024), MAXFSIZE));
+               off2 = (off64_t)(lr % max_off2);
                off2 %= maxfsize;
                off2 &= ~(stat2.st_blksize - 1);
        } while (stat1.st_ino == stat2.st_ino && llabs(off2 - off1) < len);