Skip to contentSkip to conversation

do_sys_truncate: call do_truncate with ATTR_MTIME|ATTR_CTIME

RFC PATCH In review Vladimir V. Saveliev 2 participants 2 messages Jun 30, 2006 VFS

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 do_sys_truncate: call do_truncate with ATTR_MTIME|ATTR_CTIME

Hello

do_sys_ftruncate calls do_truncate with time_attrs set to ATTR_MTIME| ATTR_CTIME. Is there a reason for do_sys_truncate to not set time_attrs to the same value or it is a bug?

From: Vladimir Saveliev <[email protected]>

When do_truncate is called from do_sys_truncate it is not given ATTR_MTIME|ATTR_CTIME as in case when it is called from do_sys_ftruncate. The patch fixes this unfairness.

fs/open.c renamed from fs/open.c~do_truncate-time_attrs +1 -1
diff -puN fs/open.c~do_truncate-time_attrs fs/open.c
@@ -270,7 +270,7 @@ static long do_sys_truncate(const char _
 	error = locks_verify_truncate(inode, NULL, length);
 	if (!error) {
 		DQUOT_INIT(inode);
-		error = do_truncate(nd.dentry, length, 0, NULL);
+		error = do_truncate(nd.dentry, length, ATTR_MTIME|ATTR_CTIME, NULL);
 	}
 	put_write_access(inode);
 

_

#1 lore raw
Quoted 5 lines · Vladimir V. Saveliev

Hello

do_sys_ftruncate calls do_truncate with time_attrs set to ATTR_MTIME| ATTR_CTIME. Is there a reason for do_sys_truncate to not set time_attrs to the same value or it is a bug?

It would be a bug.

In the case of truncate(), the filesystem has to decide if the file size will actually change. If it does, then it has to update mtime/ctime. In the case of ftruncate(), it would appear that the filesystem always has to update mtime/ctime on success, so it is appropriate to call do_truncate with ATTR_MTIME/ATTR_CTIME.

Cheers,
Trond

Quoted 5 lines

email message attachment (do_truncate-time_attrs.patch)

Quoted 4 lines

-------- Forwarded Message --------
Subject: No Subject
Date: Fri, 30 Jun 2006 17:33:29 +0400