[patch?] truncate and timestamps

From: Andries.Brouwer@cwi.nl
Date: Thu May 22 2003 - 14:09:43 EST


Investigating why some SuSE init script no longer works, I find:
The shell command
> file
does not update the time stamp of file in case it existed and had size 0.
The corresponding system call is
    open(file, O_WRONLY | O_TRUNC);

Time stamp here is st_mtime, and the behaviour is not unreasonable:
after all, the contents do not change.

And indeed, in do_truncate() we have
    newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
that is, ATTR_MTIME is not mentioned.

Older Linux systems did update st_mtime, old FreeBSD systems did not,
SunOS 5.7 does, SunOS 5.8 does not. There is no uniform Unix behaviour.

Remains the question what POSIX says.
For opening with O_TRUNC, and for the ftruncate system call
POSIX has always said and still says: st_ctime and st_mtime
are updated.
For the truncate call POSIX.1-2001 says: st_ctime and st_mtime
are updated if the size changed.

If we want to follow this, then do_truncate() needs an additional
parameter indicating what kind of call we have.
A little bit ugly. A possible patch follows.

Andries

------------------
diff -u --recursive --new-file -X /linux/dontdiff a/fs/exec.c b/fs/exec.c
--- a/fs/exec.c Thu May 22 13:17:02 2003
+++ b/fs/exec.c Thu May 22 20:14:02 2003
@@ -1352,7 +1352,7 @@
                 goto close_fail;
         if (!file->f_op->write)
                 goto close_fail;
- if (do_truncate(file->f_dentry, 0) != 0)
+ if (do_truncate(file->f_dentry, 0, 1) != 0)
                 goto close_fail;
 
         retval = binfmt->core_dump(signr, regs, file);
diff -u --recursive --new-file -X /linux/dontdiff a/fs/namei.c b/fs/namei.c
--- a/fs/namei.c Thu May 22 13:17:02 2003
+++ b/fs/namei.c Thu May 22 20:13:41 2003
@@ -1178,7 +1178,7 @@
                 if (!error) {
                         DQUOT_INIT(inode);
                         
- error = do_truncate(dentry, 0);
+ error = do_truncate(dentry, 0, 1);
                 }
                 put_write_access(inode);
                 if (error)
diff -u --recursive --new-file -X /linux/dontdiff a/fs/open.c b/fs/open.c
--- a/fs/open.c Thu May 22 13:17:02 2003
+++ b/fs/open.c Thu May 22 20:15:25 2003
@@ -75,7 +75,7 @@
         return error;
 }
 
-int do_truncate(struct dentry *dentry, loff_t length)
+int do_truncate(struct dentry *dentry, loff_t length, int update_timestamp)
 {
         int err;
         struct iattr newattrs;
@@ -85,7 +85,9 @@
                 return -EINVAL;
 
         newattrs.ia_size = length;
- newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
+ newattrs.ia_valid = ATTR_SIZE;
+ if (update_timestamp || length != dentry->d_inode->i_size)
+ newattrs.ia_valid |= ATTR_CTIME | ATTR_MTIME;
         down(&dentry->d_inode->i_sem);
         err = notify_change(dentry, &newattrs);
         up(&dentry->d_inode->i_sem);
@@ -142,7 +144,7 @@
         error = locks_verify_truncate(inode, NULL, length);
         if (!error) {
                 DQUOT_INIT(inode);
- error = do_truncate(nd.dentry, length);
+ error = do_truncate(nd.dentry, length, 0);
         }
         put_write_access(inode);
 
@@ -194,7 +196,7 @@
 
         error = locks_verify_truncate(inode, file, length);
         if (!error)
- error = do_truncate(dentry, length);
+ error = do_truncate(dentry, length, 1);
 out_putf:
         fput(file);
 out:
diff -u --recursive --new-file -X /linux/dontdiff a/include/linux/fs.h b/include/linux/fs.h
--- a/include/linux/fs.h Thu May 22 13:17:05 2003
+++ b/include/linux/fs.h Thu May 22 20:14:39 2003
@@ -1019,7 +1019,7 @@
 
 asmlinkage long sys_open(const char __user *, int, int);
 asmlinkage long sys_close(unsigned int); /* yes, it's really unsigned */
-extern int do_truncate(struct dentry *, loff_t start);
+extern int do_truncate(struct dentry *, loff_t start, int update_timestamp);
 
 extern struct file *filp_open(const char *, int, int);
 extern struct file * dentry_open(struct dentry *, struct vfsmount *, int);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Fri May 23 2003 - 22:00:51 EST