Re: running out of tags in 9P (was Re: [git pull] vfs part 2)

[Al Viro]

On Thu, Jul 02, 2015 at 03:19:57PM +0300, Andrey Ryabinin wrote:

> Added:
> +    if (total > count)
> +           *(char *)0 = 0
> 
> and never hit this condition.
> 

OK, so it's definitely a mismatched response.

>         req->tc->tag = tag-1;
> +       if (WARN_ON(req->status != REQ_STATUS_IDLE))
> +               pr_err("req->status: %d\n", req->status);
>         req->status = REQ_STATUS_ALLOC;
> 
>         return req;

> [  150.259076] 9pnet: req->status: 4

IOW, REQ_STATUS_RCVD.  Hmm...  Stray tag seen by req_done() after we'd already
freed the tag in question?  That, or it really would have to had wrapped
around...  Note that req_done() does *not* check anything about the req -
not even that p9_tag_lookup() hasn't returned NULL, so a server sending you
any response tagged with number well above anything you'd ever sent will
reliably oops you.

Frankly, the whole thing needs fuzzing from the server side - start throwing
crap at the client and see how badly does it get fucked...  Folks, it's
a network protocol, with userland servers, no less.  You *can't* assume
them competent and non-malicious...

How much traffic does it take to reproduce that fun, BTW?  IOW, is attempting
to log the sequence of tag {allocation,freeing}/tag of packet being {sent,
received} something completely suicidal, or is it more or less feasible?

> I didn't get this. c->reqs[row] is always non-NULL as it should be, so this warning
> will trigger all the time.

????
                        row = (tag / P9_ROW_MAXTAG);
                        c->reqs[row] = kcalloc(P9_ROW_MAXTAG,
                                        sizeof(struct p9_req_t), GFP_ATOMIC);

and you are seeing c->reqs[row] != NULL *BEFORE* that kcalloc()?  All the time,
no less?  Just to make sure we are on the same page - the delta against
mainline I would like tested is this:

diff --git a/net/9p/client.c b/net/9p/client.c
index 6f4c4c8..fa88c9a 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -248,6 +248,9 @@ p9_tag_alloc(struct p9_client *c, u16 tag, unsigned int max_size)
 		/* check again since original check was outside of lock */
 		while (tag >= c->max_tag) {
 			row = (tag / P9_ROW_MAXTAG);
+
+			WARN_ON_ONCE(c->reqs[row]);	// are we about to leak?
+
 			c->reqs[row] = kcalloc(P9_ROW_MAXTAG,
 					sizeof(struct p9_req_t), GFP_ATOMIC);
 
@@ -286,6 +289,8 @@ p9_tag_alloc(struct p9_client *c, u16 tag, unsigned int max_size)
 	p9pdu_reset(req->rc);
 
 	req->tc->tag = tag-1;
+	if (req->status != REQ_STATUS_IDLE)
+		pr_err("using tag %d with odd status (%d)", tag, req->status);
 	req->status = REQ_STATUS_ALLOC;
 
 	return req;
@@ -425,6 +430,8 @@ void p9_client_cb(struct p9_client *c, struct p9_req_t *req, int status)
 	 * the other thread wakes up will indeed be seen by the waiting side.
 	 */
 	smp_wmb();
+	if (req->status == REQ_STATUS_IDLE)
+		pr_err("late delivery, tag %d already freed", req->tc->tag);
 	req->status = status;
 
 	wake_up(req->wq);
@@ -693,6 +700,10 @@ static struct p9_req_t *p9_client_prepare_req(struct p9_client *c,
 		tag = p9_idpool_get(c->tagpool);
 		if (tag < 0)
 			return ERR_PTR(-ENOMEM);
+		if (WARN_ON_ONCE(tag != (u16)tag)) {	// wrapped around?
+			p9_idpool_put(tag, c->tagpool);
+			return ERR_PTR(-ENOMEM);
+		}
 	}
 
 	req = p9_tag_alloc(c, tag, req_size);
@@ -1647,7 +1658,10 @@ p9_client_write(struct p9_fid *fid, u64 offset, struct iov_iter *from, int *err)
 		if (*err) {
 			trace_9p_protocol_dump(clnt, req->rc);
 			p9_free_req(clnt, req);
+			break;
 		}
+		if (rsize < count)
+			pr_err("mismatched reply [tag = %d]\n", req->tc->tag);
 
 		p9_debug(P9_DEBUG_9P, "<<< RWRITE count %d\n", count);
 
--
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/