Re: [PATCH 09/10] VFS: Remove read-only checks from dentry_permission

[Miklos Szeredi]

On Mon, 06 Sep 2010, NeilBrown wrote:
> It is not sufficient to depend on the the "filesystem is readonly"
> tests in dentry_permission as it does not check if the vfsmnt is
> readonly.
> All call sites already call mnt_want_write or __mnt_is_readonly which
> includes the test on MS_RDONLY.

Last time I checked I found some holes (in nfsd IIRC).  That was a
long time ago and things may have changed.

That check could be replaced with a

		if (IS_RDONLY(inode) &&
		    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
			BUG();

which would catch these cases but only if the superblock was marked
r/o.  Otherwise it's basically impossible to make sure the callers of
the VFS play by the rules.  That was one reason I advocated a
path_... interface for the VFS instead of the current dentry based
one, but Al didn't like it.

Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/