Commit 833491908a1afd67 introduced the ability to report completion
latency percentiles. It also caused a memory corruption when running
with multiple threads due to out of bound accesses in show_run_stats().
The major index of the io_u_plat two-dimensional array is meant
to be DDIR_ value in {DDIR_READ, DDIR_WRITE} (i.e., {0, 1}). The
code in show_run_stats() incorrectly wrote into the array using
a major index with values {0, 1, 2}. Commit 0a0b49007cbce8d1 fixed
the out of bound accesses by increasing the size of the major
dimension of the io_u_plat array from 2 to 3.
This patch reverts the size change from 0a0b49007cbce8d1 in favor
of avoiding the out-of-bound accesses in show_run_stats().
Signed-off-by: Eric Gouriou <egouriou@xxxxxxxxxx>
---
Jens, Zhu,
Yu-Ju is unlikely to be checking fio email traffic this week,
hence my follow-up. The error was introduced while porting
the patch between different versions of fio. The internal version
was tested appropriately but not the upstream version.
Apologies for the trouble.
Regards - Eric
---
fio.h | 2 +-
stat.c | 5 +++--
2 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/fio.h b/fio.h
index c741162..6c57496 100644
--- a/fio.h
+++ b/fio.h
@@ -217,7 +217,7 @@ struct thread_stat {
unsigned int io_u_complete[FIO_IO_U_MAP_NR];
unsigned int io_u_lat_u[FIO_IO_U_LAT_U_NR];
unsigned int io_u_lat_m[FIO_IO_U_LAT_M_NR];
- unsigned int io_u_plat[3][FIO_IO_U_PLAT_NR];
+ unsigned int io_u_plat[2][FIO_IO_U_PLAT_NR];
unsigned long total_io_u[3];
unsigned long short_io_u[3];
unsigned long total_submit;
diff --git a/stat.c b/stat.c
index ee6ee51..ae3c71a 100644
--- a/stat.c
+++ b/stat.c
@@ -773,11 +773,12 @@ void show_run_stats(void)
for (k = 0; k <= 2; k++) {
- int m;
-
ts->total_io_u[k] += td->ts.total_io_u[k];
ts->short_io_u[k] += td->ts.short_io_u[k];
+ }
+ for (k = 0; k <= DDIR_WRITE; k++) {
+ int m;
for (m = 0; m < FIO_IO_U_PLAT_NR; m++)
ts->io_u_plat[k][m] += td->ts.io_u_plat[k][m];
}
--
1.7.3.1
--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
