fs: fsstress: make test more POSIX compliant

PATCH Superseded [email protected] 3 participants 5 messages Mar 16, 2011 1 message not in the archive

lore

Share & tools

lore

A message that is not in the archive (sent off-list or not yet received): [PATCH 1/2]

[email protected]

author

2011-03-16 15:44 UTC #1 lore raw

PATCH v4 1/2 fs: fsstress: make test more POSIX compliant

From: Caspar Zhang <[email protected]>

in commit bacc849720ec4efda5a0a8a9ea6a0e93a1415541, malloc.h was removed and stdlib.h got used instead. However, it's not enough. The function memalign() is also obsolete (via man memalign) and we should use posix_memalign instead. Also if we keep memalign() function here, the program would probably hit segfault once it enters dwrite_t and dread_t.

v2: use correct &buf as argument
v3: remove incorrect disabling warning, will try to fix warning and compiling without -DNO_XFS in another patch
v4: code style

Signed-off-by Caspar Zhang <[email protected]>

---
 testcases/kernel/fs/fsstress/fsstress.c |   24 +++++++++++++++++++-----
 1 files changed, 19 insertions(+), 5 deletions(-)

testcases/kernel/fs/fsstress/fsstress.c +19 -5

index e3b48ea..d6521f4 100644
@@ -1746,7 +1746,7 @@ void
 dread_f(int opno, long r)
 {
 	__int64_t	align;
-	char		*buf;
+	char		*buf = NULL;
 	struct dioattr	diob;
 	int		e;
 	pathname_t	f;

@@ -1826,7 +1826,14 @@ dread_f(int opno, long r)
 		len = align;
 	else if (len > diob.d_maxiosz)
 		len = diob.d_maxiosz;
-	buf = memalign(diob.d_mem, len);
+	if (posix_memalign((void **)&buf, diob.d_mem, len) != 0) {
+		perror("posix_memalign");
+		exit(1);
+	}
+	if (buf == NULL) {
+		fprintf(stderr, "posix_memalign: buf is NULL\n");
+		exit(1);
+	}
 	e = read(fd, buf, len) < 0 ? errno : 0;
 	free(buf);
 	if (v)

@@ -1840,7 +1847,7 @@ void
 dwrite_f(int opno, long r)
 {
 	__int64_t	align;
-	char		*buf;
+	char		*buf = NULL;
 	struct dioattr	diob;
 	int		e;
 	pathname_t	f;

@@ -1909,7 +1916,14 @@ dwrite_f(int opno, long r)
 		len = align;
 	else if (len > diob.d_maxiosz)
 		len = diob.d_maxiosz;
-	buf = memalign(diob.d_mem, len);
+	if (posix_memalign((void **)&buf, diob.d_mem, len) != 0) {
+		perror("posix_memalign");
+		exit(1);
+	}
+	if (buf == NULL) {
+		fprintf(stderr, "posix_memalign: buf is NULL\n");
+		exit(1);
+	}
 	off %= maxfsize;
 	lseek64(fd, off, SEEK_SET);
 	memset(buf, nameseq & 0xff, len);

@@ -2650,4 +2664,4 @@ write_f(int opno, long r)
 			procid, opno, f.path, (long long)off, (long int)len, e);
 	free_pathname(&f);
 	close(fd);
-}
\ No newline at end of file
+}

Signature · 12 lines

1.7.4.1


------------------------------------------------------------------------------
Colocation vs. Managed Hosting
A question and answer guide to determining the best fit
for your organization - today and in the future.
http://p.sf.net/sfu/internap-sfd2d
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Cyril Hrubis

2011-03-16 16:06 UTC #2 lore raw

Hi!

Quoted 9 lines

dread_f(int opno, long r)
{

testcases/kernel/fs/fsstress/fsstress.c +1 -1

 	__int64_t	align;
-	char		*buf;
+	char		*buf = NULL;
 	struct dioattr	diob;
 	int		e;
 	pathname_t	f;

testcases/kernel/fs/fsstress/fsstress.c dread_f(int opno, long r) · line 1830

Quoted 6 lines

testcases/kernel/fs/fsstress/fsstress.c +2 -1

@@ -1826,7 +1826,14 @@ dread_f(int opno, long r)
 		len = align;
 	else if (len > diob.d_maxiosz)
 		len = diob.d_maxiosz;
-	buf = memalign(diob.d_mem, len);
+	if (posix_memalign((void **)&buf, diob.d_mem, len) != 0) {
+		perror("posix_memalign");

Sorry for not spotting this earlier, but I've doublechecked the manual page and it says:

...
posix_memalign() returns zero on success, or one of the error values listed in the next section on failure. Note that errno is not set.
...

So you can't use perror() to print the error message. It would probably print something like:

posix_memalign: Succeded

The correct solution is to save the return value from posix_memalign() and use strerror().

Quoted 8 lines

testcases/kernel/fs/fsstress/fsstress.c +6 -0

+		exit(1);
+	}
+	if (buf == NULL) {
+		fprintf(stderr, "posix_memalign: buf is NULL\n");
+		exit(1);
+	}
 	e = read(fd, buf, len) < 0 ? errno : 0;
 	free(buf);

Signature · 12 lines

Cyril Hrubis
[email protected]

------------------------------------------------------------------------------
Colocation vs. Managed Hosting
A question and answer guide to determining the best fit
for your organization - today and in the future.
http://p.sf.net/sfu/internap-sfd2d
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

[email protected]

author

2011-03-16 15:45 UTC #3 lore raw

PATCH v2 2/2 fs: fsstress: fix compile when -DNO_XFS disabled

In reply to a message not in the archive

From: Caspar Zhang <[email protected]>

When -DNO_XFS disabled in Makefile, the compilation couldn't pass and many errors and warnings outputed. This patch fixes this issue.

v2: modify CPPFLAGS to LDLIBS

Signed-off-by Caspar Zhang <[email protected]>

---
 testcases/kernel/fs/fsstress/Makefile   |    3 +++
 testcases/kernel/fs/fsstress/fsstress.c |   16 ++++++++--------
 testcases/kernel/fs/fsstress/global.h   |    4 ++--
 3 files changed, 13 insertions(+), 10 deletions(-)

testcases/kernel/fs/fsstress/Makefile +3 -0

index 0a53891..38acce0 100644
@@ -27,6 +27,9 @@ include $(top_srcdir)/include/mk/env_pre.mk
 CPPFLAGS			+= -DNO_XFS -I$(abs_srcdir) \
 				   -D_LARGEFILE64_SOURCE -D_GNU_SOURCE
 
+# if removed -DNO_XFS, you should unmask the following line
+#LDLIBS				+= -lattr
+
 # XXX (garrcoop): not -Wuninitialized clean.
 CPPFLAGS			+= -Wno-error

testcases/kernel/fs/fsstress/fsstress.c +8 -8

index d6521f4..089edfb 100644
@@ -1436,7 +1436,7 @@ allocsp_f(int opno, long r)
 	e = ioctl(fd, XFS_IOC_ALLOCSP64, &fl) < 0 ? errno : 0;
 	if (v)
 		printf("%d/%d: ioctl(XFS_IOC_ALLOCSP64) %s %lld 0 %d\n",
-			procid, opno, f.path, off, e);
+			procid, opno, f.path, (long long)off, e);
 	free_pathname(&f);
 	close(fd);
 }

@@ -1571,7 +1571,7 @@ bulkstat_f(int opno, long r)
 	free(t);
 	if (verbose)
 		printf("%d/%d: bulkstat nent %d total %lld\n",
-			procid, opno, nent, total);
+			procid, opno, nent, (long long)total);
 	close(fd);
 }

@@ -1704,7 +1704,7 @@ creat_f(int opno, long r)
 				geom.rtextsize * geom.blocksize * extsize;
 			if (ioctl(fd, XFS_IOC_FSSETXATTR, &a) < 0)
 				e1 = errno;
-			esz = a.fsx_estsize;
+			esz = a.fsx_extsize;
 
 		}
 #endif

@@ -2016,7 +2016,7 @@ freesp_f(int opno, long r)
 	e = ioctl(fd, XFS_IOC_FREESP64, &fl) < 0 ? errno : 0;
 	if (v)
 		printf("%d/%d: ioctl(XFS_IOC_FREESP64) %s %lld 0 %d\n",
-			procid, opno, f.path, off, e);
+			procid, opno, f.path, (long long)off, e);
 	free_pathname(&f);
 	close(fd);
 }

@@ -2388,8 +2388,8 @@ resvsp_f(int opno, long r)
 	fl.l_len = (off64_t)(random() % (1024 * 1024));
 	e = ioctl(fd, XFS_IOC_RESVSP64, &fl) < 0 ? errno : 0;
 	if (v)
-		printf("%d/%d: ioctl(XFS_IOC_RESVSP64) %s %lld %lld %d\n",
-			procid, opno, f.path, off, fl.l_len, e);
+		printf("%d/%d: ioctl(XFS_IOC_RESVSP64) %s %lld %ld %d\n",
+			procid, opno, f.path, (long long)off, fl.l_len, e);
 	free_pathname(&f);
 	close(fd);
 }

@@ -2605,8 +2605,8 @@ unresvsp_f(int opno, long r)
 	fl.l_len = (off64_t)(random() % (1 << 20));
 	e = ioctl(fd, XFS_IOC_UNRESVSP64, &fl) < 0 ? errno : 0;
 	if (v)
-		printf("%d/%d: ioctl(XFS_IOC_UNRESVSP64) %s %lld %lld %d\n",
-			procid, opno, f.path, off, fl.l_len, e);
+		printf("%d/%d: ioctl(XFS_IOC_UNRESVSP64) %s %lld %ld %d\n",
+			procid, opno, f.path, (long long)off, fl.l_len, e);
 	free_pathname(&f);
 	close(fd);
 }

testcases/kernel/fs/fsstress/global.h +2 -2

index f788395..9644666 100644
@@ -38,8 +38,8 @@
 #if defined(NO_XFS)
 # include "xfscompat.h"
 #else
-# include <libxfs.h>
-# include <attributes.h>
+# include <xfs/libxfs.h>
+# include <attr/attributes.h>
 #endif
 
 /* libc includes */

Signature · 12 lines

1.7.4.1


------------------------------------------------------------------------------
Colocation vs. Managed Hosting
A question and answer guide to determining the best fit
for your organization - today and in the future.
http://p.sf.net/sfu/internap-sfd2d
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Garrett Cooper

2011-03-16 17:26 UTC #4 lore raw

Re: [LTP] [PATCH v2 2/2] fs: fsstress: fix compile when -DNO_XFS disabled

Quoted 6 lines · On Wed, Mar 16, 2011 at 8:45 AM, <[email protected]> wrote:

From: Caspar Zhang <[email protected]>
When -DNO_XFS disabled in Makefile, the compilation couldn't pass and many errors and warnings outputed. This patch fixes this issue.
v2: modify CPPFLAGS to LDLIBS

Once upon a time I assume this worked for someone when compiling
on legacy systems. Please add any and all necessary autoconf tests for these headers.
Thanks,
-Garrett

------------------------------------------------------------------------------
Colocation vs. Managed Hosting
A question and answer guide to determining the best fit for your organization - today and in the future.
http://p.sf.net/sfu/internap-sfd2d
_______________________________________________
Ltp-list mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Cyril Hrubis

2011-03-16 17:34 UTC #5 lore raw

Re: [LTP] [PATCH v2 2/2] fs: fsstress: fix compile when -DNO_XFS disabled

Hi!

Quoted 3 lines

Once upon a time I assume this worked for someone when compiling
on legacy systems. Please add any and all necessary autoconf tests for these headers.

I've commited the patch as it is a couple of minutes ago as you may say it's improvement to the previous state. However that it doesn't compile here.

Signature · 12 lines

Cyril Hrubis
[email protected]

------------------------------------------------------------------------------
Colocation vs. Managed Hosting
A question and answer guide to determining the best fit
for your organization - today and in the future.
http://p.sf.net/sfu/internap-sfd2d
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list