[PATCH] src/t_stripealign.c: Use block size reported by fstatfs()

[PATCH] src/t_stripealign.c: Use block size reported by fstatfs()

[Chandan Rajendra]

On ppc64le, block size reported by fstat(2) is 64k (the page size)
i.e. the "preferred I/O size". However src/t_stripealign.c requires the
actual block size of the filesystem. Hence this commit now makes use of
the block size reported by fstatfs(2) syscall.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
---
 src/t_stripealign.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/t_stripealign.c b/src/t_stripealign.c
index c61e7632..5cdadaae 100644
--- a/src/t_stripealign.c
+++ b/src/t_stripealign.c
@@ -7,7 +7,7 @@
 #include <unistd.h>
 #include <stdlib.h>
 #include <sys/types.h>
-#include <sys/stat.h>
+#include <sys/vfs.h>
 #include <fcntl.h>
 #include <stdio.h>
 #include <sys/ioctl.h>
@@ -31,7 +31,7 @@
 
 int main(int argc, char ** argv)
 {
-	struct stat		sb;
+	struct statfs		sb;
 	struct fiemap		*fie;
 	struct fiemap_extent	*fe;
 	int			fd;
@@ -54,7 +54,7 @@ int main(int argc, char ** argv)
                 return 1;
         }
 
-	ret = fstat(fd, &sb);
+	ret = fstatfs(fd, &sb);
 	if (ret) {
 		perror(filename);
 		close(fd);
@@ -101,7 +101,7 @@ int main(int argc, char ** argv)
 		return 1;
 	}
 
-	block = fie->fm_extents[0].fe_physical / sb.st_blksize;
+	block = fie->fm_extents[0].fe_physical / sb.f_bsize;
 check:
 	if (block % sunit) {
 		printf("%s: Start block %llu not multiple of sunit %u\n",
-- 
2.19.1

Re: [PATCH] src/t_stripealign.c: Use block size reported by fstatfs()

[Eryu Guan]

On Wed, Nov 21, 2018 at 09:56:46PM +0530, Chandan Rajendra wrote:
> On ppc64le, block size reported by fstat(2) is 64k (the page size)
> i.e. the "preferred I/O size". However src/t_stripealign.c requires the
> actual block size of the filesystem. Hence this commit now makes use of
> the block size reported by fstatfs(2) syscall.
> 
> Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>

Looks like that only XFS returns the preferred iosize
(xfs_preferred_iosize()) on stat(2), other filesystems will return block
size (I checked the code of ext4, f2fs, ocfs2).

On the other hand, statfs(2) returns the actual block size on all
filesystems I've checked, including XFS. So this patch looks fine to me.

Thanks,
Eryu