login
Search
Hi Andrew,
For autofs maps that recursively reference other map entries in their
mount point path we need to ensure that these mounts are performed
prior to the current mount request being done. We've used the access
system call, made just prior to performing the mount, to make this
happen. In the case of bind mounts, however, the lookup flags in
the path walk don't trigger a mount. To do this we need to also
trigger a mount when we see the LOOKUP_ACCESS flag.
Signed-off-by: Ian Kent <raven@themaw.net>
Ian
---
diff -up linux-2.6.25-mm1/fs/autofs4/root.c.lookup-access-intent-trigger linux-2.6.25-mm1/fs/autofs4/root.c
--- linux-2.6.25-mm1/fs/autofs4/root.c.lookup-access-intent-trigger 2008-04-28 11:33:20.000000000 +0800
+++ linux-2.6.25-mm1/fs/autofs4/root.c 2008-04-28 11:36:06.000000000 +0800
@@ -19,6 +19,8 @@
#include <linux/time.h>
#include "autofs_i.h"
+#define LOOKUP_TRIGGER_FLAGS (LOOKUP_CONTINUE|LOOKUP_DIRECTORY|LOOKUP_ACCESS)
+
static int autofs4_dir_symlink(struct inode *,struct dentry *,const char *);
static int autofs4_dir_unlink(struct inode *,struct dentry *);
static int autofs4_dir_rmdir(struct inode *,struct dentry *);
@@ -291,7 +293,7 @@ static int try_to_fill_dentry(struct den
return status;
}
/* Trigger mount for path component or follow link */
- } else if (flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY) ||
+ } else if (flags & LOOKUP_TRIGGER_FLAGS ||
current->link_count) {
DPRINTK("waiting for mount name=%.*s",
dentry->d_name.len, dentry->d_name.name);
@@ -335,7 +337,7 @@ static void *autofs4_follow_link(struct
nd->flags);
/* If it's our master or we shouldn't trigger a mount we're done */
- lookup_type = nd->flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY);
+ lookup_type = nd->flags & LOOKUP_TRIGGER_FLAGS;
if (oz_mode || !lookup_type)
goto done;
--
That's too fscking bad, since ->lookup() is about to lose access to nameidata *and* flags for anything but the last step. IOW, we'll need cleaner solution... --
->lookup() and friends are going to stop getting struct nameidata *. ->follow_link() obviously will keep it; ->d_revalidate() should lose the sucker, but we might have to split it in two methods (which is probably what'll be needed in your case). At the same time struct open_intent is going to die, BTW, but that won't matter for autofs, AFAICT. --
Right. I don't see that yet in 2.6.25-mm1, are there any preliminary patches I Yep, autofs doesn't care about this, it only cares about being able to ask userspace to try a mount when needed. Actually, using the flags to decide when to call back the daemon wouldn't be needed at all except for mount storms caused by aggressive GUI file system scanning and color ls and the like. Ian --