diff mbox series

tst_fill_fs: drop safe_macro from fill_flat_vec

Message ID 20230615075307.157688-1-liwang@redhat.com
State Accepted
Headers show
Series tst_fill_fs: drop safe_macro from fill_flat_vec | expand

Commit Message

Li Wang June 15, 2023, 7:53 a.m. UTC 
LTP fs_fill is trying to spawn nCPUS thread to create corresponding
subdirectories and completely fill them. But in the invoke function
tst_fill_fs(, TST_FILL_BLOCKS) -> fill_flat_vec() which has SAFE_MACROs
to break the open behavior once fs is filled with ENOSPC return.

The failure like there aren't enough free inodes in the filesystem for
the tst_fs_fill test program to create it's set of "AOF" data files.

  4108	tst_test.c:1634: TINFO: === Testing on xfs ===
  4109	tst_test.c:1094: TINFO: Formatting /dev/loop0 with xfs opts='' extra opts=''
  4110	fs_fill.c:108: TINFO: Running 98 writer threads
  4111	tst_fill_fs.c:77: TBROK: openat(89</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread84>, 'AOF', 101, 600): ENOSPC (28)
  4112	tst_fill_fs.c:77: TBROK: openat(87</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread83>, 'AOF', 101, 600): ENOSPC (28)
  4113	tst_fill_fs.c:77: TWARN: openat(85</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread82>, 'AOF', 101, 600): ENOSPC (28)
  4114	tst_fill_fs.c:77: TWARN: openat(3</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread81>, 'AOF', 101, 600): ENOSPC (28)

This patch is just to convert the fill_flat_vec() to ignore ENOSPC like
fill_randome() in opening subdirectories.

Reported-by: Shizhao Chen <shichen@redhat.com>
Signed-off-by: Li Wang <liwang@redhat.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Richard Palethorpe <rpalethorpe@suse.com>
---
 lib/tst_fill_fs.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)

Comments

Murphy Zhou June 16, 2023, 12:29 p.m. UTC | #1 
On Thu, Jun 15, 2023 at 3:53 PM Li Wang <liwang@redhat.com> wrote:
>
> LTP fs_fill is trying to spawn nCPUS thread to create corresponding
> subdirectories and completely fill them. But in the invoke function
> tst_fill_fs(, TST_FILL_BLOCKS) -> fill_flat_vec() which has SAFE_MACROs
> to break the open behavior once fs is filled with ENOSPC return.
>
> The failure like there aren't enough free inodes in the filesystem for
> the tst_fs_fill test program to create it's set of "AOF" data files.
>
>   4108  tst_test.c:1634: TINFO: === Testing on xfs ===
>   4109  tst_test.c:1094: TINFO: Formatting /dev/loop0 with xfs opts='' extra opts=''
>   4110  fs_fill.c:108: TINFO: Running 98 writer threads
>   4111  tst_fill_fs.c:77: TBROK: openat(89</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread84>, 'AOF', 101, 600): ENOSPC (28)
>   4112  tst_fill_fs.c:77: TBROK: openat(87</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread83>, 'AOF', 101, 600): ENOSPC (28)
>   4113  tst_fill_fs.c:77: TWARN: openat(85</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread82>, 'AOF', 101, 600): ENOSPC (28)
>   4114  tst_fill_fs.c:77: TWARN: openat(3</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread81>, 'AOF', 101, 600): ENOSPC (28)
>
> This patch is just to convert the fill_flat_vec() to ignore ENOSPC like
> fill_randome() in opening subdirectories.

Looks good to me. Thanks for fixing it!

Murphy
>
> Reported-by: Shizhao Chen <shichen@redhat.com>
> Signed-off-by: Li Wang <liwang@redhat.com>
> Cc: Dave Chinner <dchinner@redhat.com>
> Cc: Richard Palethorpe <rpalethorpe@suse.com>
> ---
>  lib/tst_fill_fs.c | 21 +++++++++++++++++++--
>  1 file changed, 19 insertions(+), 2 deletions(-)
>
> diff --git a/lib/tst_fill_fs.c b/lib/tst_fill_fs.c
> index b9d29755a..243eb279f 100644
> --- a/lib/tst_fill_fs.c
> +++ b/lib/tst_fill_fs.c
> @@ -73,12 +73,29 @@ void fill_random(const char *path, int verbose)
>
>  void fill_flat_vec(const char *path, int verbose)
>  {
> -       int dir = SAFE_OPEN(path, O_PATH | O_DIRECTORY);
> -       int fd = SAFE_OPENAT(dir, "AOF", O_WRONLY | O_CREAT, 0600);
> +       int dir, fd;
>         struct iovec iov[512];
>         int iovcnt = ARRAY_SIZE(iov);
>         int retries = 3;
>
> +       dir = open(path, O_PATH | O_DIRECTORY);
> +       if (dir == -1) {
> +               if (errno == ENOSPC) {
> +                       tst_res(TINFO | TERRNO, "open()");
> +                       return;
> +               }
> +               tst_brk(TBROK | TERRNO, "open(%s, %d) failed", path, O_PATH | O_DIRECTORY);
> +       }
> +
> +       fd = openat(dir, "AOF", O_WRONLY | O_CREAT, 0600);
> +       if (fd == -1) {
> +               if (errno == ENOSPC) {
> +                       tst_res(TINFO | TERRNO, "openat()");
> +                       return;
> +               }
> +               tst_brk(TBROK | TERRNO, "openat(%s, %d, 0600) failed", dir, O_PATH | O_DIRECTORY);
> +       }
> +
>         SAFE_CLOSE(dir);
>
>         for (int i = 0; i < iovcnt; i++) {
> --
> 2.40.1
>
>
> --
> Mailing list info: https://lists.linux.it/listinfo/ltp
Li Wang June 21, 2023, 1:32 a.m. UTC | #2 
On Fri, Jun 16, 2023 at 8:30 PM Murphy Zhou <jencce.kernel@gmail.com> wrote:

> On Thu, Jun 15, 2023 at 3:53 PM Li Wang <liwang@redhat.com> wrote:
> >
> > LTP fs_fill is trying to spawn nCPUS thread to create corresponding
> > subdirectories and completely fill them. But in the invoke function
> > tst_fill_fs(, TST_FILL_BLOCKS) -> fill_flat_vec() which has SAFE_MACROs
> > to break the open behavior once fs is filled with ENOSPC return.
> >
> > The failure like there aren't enough free inodes in the filesystem for
> > the tst_fs_fill test program to create it's set of "AOF" data files.
> >
> >   4108  tst_test.c:1634: TINFO: === Testing on xfs ===
> >   4109  tst_test.c:1094: TINFO: Formatting /dev/loop0 with xfs opts=''
> extra opts=''
> >   4110  fs_fill.c:108: TINFO: Running 98 writer threads
> >   4111  tst_fill_fs.c:77: TBROK:
> openat(89</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread84>,
> 'AOF', 101, 600): ENOSPC (28)
> >   4112  tst_fill_fs.c:77: TBROK:
> openat(87</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread83>,
> 'AOF', 101, 600): ENOSPC (28)
> >   4113  tst_fill_fs.c:77: TWARN:
> openat(85</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread82>,
> 'AOF', 101, 600): ENOSPC (28)
> >   4114  tst_fill_fs.c:77: TWARN:
> openat(3</mnt/testarea/ltp-wGg0XNSHcr/LTP_fs_XLxrLh/mntpoint/thread81>,
> 'AOF', 101, 600): ENOSPC (28)
> >
> > This patch is just to convert the fill_flat_vec() to ignore ENOSPC like
> > fill_randome() in opening subdirectories.
>
> Looks good to me. Thanks for fixing it!
>

Patch applied, thanks for your review.
diff mbox series

Patch

diff --git a/lib/tst_fill_fs.c b/lib/tst_fill_fs.c
index b9d29755a..243eb279f 100644
--- a/lib/tst_fill_fs.c
+++ b/lib/tst_fill_fs.c
@@ -73,12 +73,29 @@  void fill_random(const char *path, int verbose)
 
 void fill_flat_vec(const char *path, int verbose)
 {
-	int dir = SAFE_OPEN(path, O_PATH | O_DIRECTORY);
-	int fd = SAFE_OPENAT(dir, "AOF", O_WRONLY | O_CREAT, 0600);
+	int dir, fd;
 	struct iovec iov[512];
 	int iovcnt = ARRAY_SIZE(iov);
 	int retries = 3;
 
+	dir = open(path, O_PATH | O_DIRECTORY);
+	if (dir == -1) {
+		if (errno == ENOSPC) {
+			tst_res(TINFO | TERRNO, "open()");
+			return;
+		}
+		tst_brk(TBROK | TERRNO, "open(%s, %d) failed", path, O_PATH | O_DIRECTORY);
+	}
+
+	fd = openat(dir, "AOF", O_WRONLY | O_CREAT, 0600);
+	if (fd == -1) {
+		if (errno == ENOSPC) {
+			tst_res(TINFO | TERRNO, "openat()");
+			return;
+		}
+		tst_brk(TBROK | TERRNO, "openat(%s, %d, 0600) failed", dir, O_PATH | O_DIRECTORY);
+	}
+
 	SAFE_CLOSE(dir);
 
 	for (int i = 0; i < iovcnt; i++) {