> It seems like you need to fix jffs2 to implement the proper semantics in its ->setattr.
Yes I have thought of this solution. However, when I tried to
track this problem down, I found that ftruncate() had similar
problem and it was fixed by commit 6e656be89999 ("ftruncate
does not always update m/ctime"):
diff --git a/fs/open.c b/fs/open.c
index 5fb16e5267dc..303f06d2a7b9 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -322,7 +322,7 @@ static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
error = locks_verify_truncate(inode, file, length);
if (!error)
- error = do_truncate(dentry, length, 0, file);
+ error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
out_putf:
fput(file);
out:
In my opinion, there are two advantages if we fix it in
vfs_truncate():
1. All filesystems can reuse the scheme without adapting
Separately, just like what we did for ftruncate().
2. In the case when old_size = new_size, we can avoid
calling do_truncate() and return without doing anything.
Hope that you can consider my suggestion, thanks.
