[PATCH] fix use after free in smbfs.

[Dave Jones]

>From code inspection it seems that after freeing 'req', in smb_request()
we continue to dereference it a dozen or so times.

That whole area of code looks suspect.
		...
        smb_lock_server(server);
        if (!(req->rq_flags & SMB_REQ_RECEIVED)) {
            list_del_init(&req->rq_queue);
            smb_rput(req);
        }

smb_rput() also does a list_del_init, but only if its safe
to do so (ie, the refcount has dropped to 0).
To my not-smbfs-savvy eyes, it would seem that we could
potentially nuke the ->rq_queue list, and return from
smb_rput() if the request was still in use. What the rest
of the code does with such a buggered-up request, I've no
idea, but it probably isn't pretty.

Perhaps smb_rput should be taking a pointer to a request that can
be null'd on success ?

Signed-off-by: Dave Jones <davej@xxxxxxxxxx>

--- linux-2.6.12/fs/smbfs/request.c~	2005-07-07 14:41:11.000000000 -0400
+++ linux-2.6.12/fs/smbfs/request.c	2005-07-07 14:41:22.000000000 -0400
@@ -348,6 +348,7 @@ int smb_add_request(struct smb_request *
 			smb_rput(req);
 		}
 		smb_unlock_server(server);
+		return -EINTR;
 	}
 
 	if (!timeleft) {



Looking further, we do exactly the same thing in smb_request_recv()

        smb_rput(req);
        wake_up_interruptible(&req->rq_wait);
    }

ditto in smbiod.c..

What am I missing here?  smb_rput() -> smb_free_request() does
a kmem_cache_free(req_cachep, req); making further use of that
cache item invalid.

		Dave

-
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/