Patchwork [04/10] xfstets: fsstress add replace file operation

login
register
mail settings
Submitter Dmitri Monakho
Date Feb. 20, 2013, 10:42 a.m.
Message ID <1361356935-29153-5-git-send-email-dmonakhov@openvz.org>
Download mbox | patch
Permalink /patch/222054/
State Not Applicable
Headers show

Comments

Dmitri Monakho - Feb. 20, 2013, 10:42 a.m.
The most common usecase for rename(2) syscall is an atomic replacement
of existing file with newer version. But rename_f() rename some existing
filename to newly generated (non existing) filename. As result the most
important usecase is not covered.
Since rename_f() is already exist in fsstress and it has known behavior,
some tests already depends on that behaviour, let's add another operation
(replace_f) which invoke rename(2) for two existing entries.

OUT_OF_COMMIT_DISCUSSION:
Off course replace_f() break naming convention where fun_name == syscall_f(),
but this is the only way I see to introduce new feature and not break
other tests. May be it is reasonable to call it rename2_f() ?

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 ltp/fsstress.c |   47 +++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 47 insertions(+), 0 deletions(-)
Rich Johnston - March 1, 2013, 3:43 p.m.
On 02/20/2013 04:42 AM, Dmitry Monakhov wrote:
> The most common usecase for rename(2) syscall is an atomic replacement
> of existing file with newer version. But rename_f() rename some existing
> filename to newly generated (non existing) filename. As result the most
> important usecase is not covered.

Good catch.

> Since rename_f() is already exist in fsstress and it has known behavior,
> some tests already depends on that behaviour, let's add another operation
> (replace_f) which invoke rename(2) for two existing entries.

>
> OUT_OF_COMMIT_DISCUSSION:
> Off course replace_f() break naming convention where fun_name == syscall_f(),
> but this is the only way I see to introduce new feature and not break
> other tests. May be it is reasonable to call it rename2_f() ?
>

I think this possible exposes a bug which was not exposed by before when 
running for example test 076 and test 083 on both ext4 and xfs.

Suggest this new function is called rename2_() so that we don't change 
the existing known tests.

Regards
--Rich
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/ltp/fsstress.c b/ltp/fsstress.c
index b4cfb25..85ff72a 100644
--- a/ltp/fsstress.c
+++ b/ltp/fsstress.c
@@ -80,6 +80,7 @@  typedef enum {
 	OP_READ,
 	OP_READLINK,
 	OP_RENAME,
+	OP_REPLACE,
 	OP_RESVSP,
 	OP_RMDIR,
 	OP_SETATTR,
@@ -165,6 +166,7 @@  void	punch_f(int, long);
 void	read_f(int, long);
 void	readlink_f(int, long);
 void	rename_f(int, long);
+void	replace_f(int, long);
 void	resvsp_f(int, long);
 void	rmdir_f(int, long);
 void	setattr_f(int, long);
@@ -202,6 +204,7 @@  opdesc_t	ops[] = {
 	{ OP_READ, "read", read_f, 1, 0 },
 	{ OP_READLINK, "readlink", readlink_f, 1, 0 },
 	{ OP_RENAME, "rename", rename_f, 2, 1 },
+	{ OP_REPLACE, "replace", replace_f, 2, 1 },
 	{ OP_RESVSP, "resvsp", resvsp_f, 1, 1 },
 	{ OP_RMDIR, "rmdir", rmdir_f, 1, 1 },
 	{ OP_SETATTR, "setattr", setattr_f, 0, 1 },
@@ -2680,6 +2683,50 @@  rename_f(int opno, long r)
 }
 
 void
+replace_f(int opno, long r)
+{
+	int		e;
+	pathname_t	src_f, dst_f;
+	fent_t		*src_fep, *dst_fep;
+	flist_t		*src_flp, *dst_flp;
+	int		v;
+	int		v1;
+
+	/* get an existing path for the source of the rename */
+	init_pathname(&src_f);
+	if (!get_fname(FT_ANYm, r, &src_f, &src_flp, &src_fep, &v)) {
+		if (v)
+			printf("%d/%d: replace - no filename\n", procid, opno);
+		free_pathname(&src_f);
+		return;
+	}
+	/* get an existing path for the destination of the rename */
+	init_pathname(&dst_f);
+	if (!get_fname(1 << (src_flp - flist), rand(), &dst_f, &dst_flp, &dst_fep, &v1)) {
+		if (v1)
+			printf("%d/%d: replace - no filename\n", procid, opno);
+		free_pathname(&dst_f);
+		return;
+	}
+	
+	v |= v1;
+	e = rename_path(&src_f, &dst_f) < 0 ? errno : 0;
+	check_cwd();
+	if (e == 0 && src_fep->id != dst_fep->id ) {
+		del_from_flist(src_flp - flist, src_fep - src_flp->fents);
+	}
+	if (v) {
+		printf("%d/%d: replace %s with %s %d\n", procid, opno, dst_f.path,
+			src_f.path, e);
+		if (e == 0 && src_fep->id != dst_fep->id)
+			printf("%d/%d: replace del entry: id=%d,parent=%d\n",
+				procid, opno, src_fep->id, src_fep->parent);
+	}
+	free_pathname(&src_f);
+	free_pathname(&dst_f);
+}
+
+void
 resvsp_f(int opno, long r)
 {
 	int		e;