d2b0825bc927e494105d7cf014cb9544218881fe6bd22d13ab0b4965.patch

From 039761fb8bef114e2d1b272957980b501c0328a3 Mon Sep 17 00:00:00 2001
From: Ikjoon Jang <ikjn@chromium.org>
Date: Thu, 8 Jul 2021 16:43:01 +0800
Subject: [PATCH] BACKPORT: FROMLIST: xhci: fix unmatched num_trbs_free

When unlinked urbs are queued to the cancelled td list, many tds
might be located after hw dequeue pointer and just marked as no-op
but not reclaimed to num_trbs_free. This bias can leads to unnecessary
ring expansions and leaks in atomic pool.

To prevent this bias, this patch counts free TRBs every time xhci moves
dequeue pointer. This patch utilizes existing
update_ring_for_set_deq_completion() function, renamed it to move_deq().

When it walks through to the new dequeue pointer, it also counts
free TRBs manually. This patch adds a fast path for the most cases
where the new dequeue pointer is still in the current segment.

Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
(am from https://patchwork.kernel.org/patch/12364751/)
(also found at https://lore.kernel.org/r/20210708164256.1.Ib344a977b52486ec81b60f9820338f1b43655f8d@changeid)

BUG=b:190087110
TEST=r8152 network test with suspend loop

Change-Id: Ib344a977b52486ec81b60f9820338f1b43655f8d
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/3013506
Tested-by: Ikjoon Jang <ikjn@chromium.org>
Reviewed-by: Tzung-Bi Shih <tzungbi@chromium.org>
Commit-Queue: Ikjoon Jang <ikjn@chromium.org>
(cherry picked from commit 3277997e948e78d58d7ca70182004e5b55345ae1)
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/3147790
Commit-Queue: Hsin-Yi Wang <hsinyi@chromium.org>
Commit-Queue: Tzung-Bi Shih <tzungbi@chromium.org>
Tested-by: Hsin-Yi Wang <hsinyi@chromium.org>
Auto-Submit: Hsin-Yi Wang <hsinyi@chromium.org>
---
 drivers/usb/host/xhci-ring.c | 106 +++++++++++++++++------------------
 1 file changed, 52 insertions(+), 54 deletions(-)

diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -152,6 +152,54 @@ static void next_trb(struct xhci_hcd *xhci,
 	}
 }
 
+/* Forward dequeue pointer to the specific position,
+ * walk through the ring and reclaim free trb slots to num_trbs_free
+ */
+static int move_deq(struct xhci_hcd *xhci, struct xhci_ring *ep_ring,
+		    struct xhci_segment *new_seg, union xhci_trb *new_deq)
+{
+	unsigned int steps;
+	union xhci_trb *deq;
+	struct xhci_segment *seg = ep_ring->deq_seg;
+
+	/* direct paths */
+	if (ep_ring->dequeue == new_deq) {
+		return 0;
+	} else if ((ep_ring->deq_seg == new_seg) &&
+	    (ep_ring->dequeue <= new_deq)) {
+		steps = new_deq - ep_ring->dequeue;
+		deq = new_deq;
+		goto found;
+	}
+
+	/* fast walk to the next segment */
+	seg = seg->next;
+	steps = (TRBS_PER_SEGMENT - 1) -
+		(ep_ring->dequeue - ep_ring->deq_seg->trbs);
+	deq = &seg->trbs[0];
+
+	while (deq != new_deq) {
+		if (trb_is_link(deq)) {
+			seg = seg->next;
+			deq = seg->trbs;
+		} else {
+			steps++;
+			deq++;
+		}
+		if (deq == ep_ring->dequeue) {
+			xhci_warn(xhci, "Unable to find new dequeue pointer\n");
+			return -ENOENT;
+		}
+	}
+
+found:
+	ep_ring->deq_seg = seg;
+	ep_ring->dequeue = deq;
+	ep_ring->num_trbs_free += steps;
+
+	return 0;
+}
+
 /*
  * See Cycle bit rules. SW is the consumer for the event ring only.
  */
@@ -1267,52 +1315,6 @@ void xhci_stop_endpoint_command_watchdog(struct timer_list *t)
 			"xHCI host controller is dead.");
 }
 
-static void update_ring_for_set_deq_completion(struct xhci_hcd *xhci,
-		struct xhci_virt_device *dev,
-		struct xhci_ring *ep_ring,
-		unsigned int ep_index)
-{
-	union xhci_trb *dequeue_temp;
-	int num_trbs_free_temp;
-	bool revert = false;
-
-	num_trbs_free_temp = ep_ring->num_trbs_free;
-	dequeue_temp = ep_ring->dequeue;
-
-	/* If we get two back-to-back stalls, and the first stalled transfer
-	 * ends just before a link TRB, the dequeue pointer will be left on
-	 * the link TRB by the code in the while loop.  So we have to update
-	 * the dequeue pointer one segment further, or we'll jump off
-	 * the segment into la-la-land.
-	 */
-	if (trb_is_link(ep_ring->dequeue)) {
-		ep_ring->deq_seg = ep_ring->deq_seg->next;
-		ep_ring->dequeue = ep_ring->deq_seg->trbs;
-	}
-
-	while (ep_ring->dequeue != dev->eps[ep_index].queued_deq_ptr) {
-		/* We have more usable TRBs */
-		ep_ring->num_trbs_free++;
-		ep_ring->dequeue++;
-		if (trb_is_link(ep_ring->dequeue)) {
-			if (ep_ring->dequeue ==
-					dev->eps[ep_index].queued_deq_ptr)
-				break;
-			ep_ring->deq_seg = ep_ring->deq_seg->next;
-			ep_ring->dequeue = ep_ring->deq_seg->trbs;
-		}
-		if (ep_ring->dequeue == dequeue_temp) {
-			revert = true;
-			break;
-		}
-	}
-
-	if (revert) {
-		xhci_dbg(xhci, "Unable to find new dequeue pointer\n");
-		ep_ring->num_trbs_free = num_trbs_free_temp;
-	}
-}
-
 /*
  * When we get a completion for a Set Transfer Ring Dequeue Pointer command,
  * we need to clear the set deq pending flag in the endpoint ring state, so that
@@ -1399,8 +1401,8 @@ static void xhci_handle_cmd_set_deq(struct xhci_hcd *xhci, int slot_id,
 			/* Update the ring's dequeue segment and dequeue pointer
 			 * to reflect the new position.
 			 */
-			update_ring_for_set_deq_completion(xhci, ep->vdev,
-				ep_ring, ep_index);
+			move_deq(xhci, ep_ring, ep->queued_deq_seg,
+				 ep->queued_deq_ptr);
 		} else {
 			xhci_warn(xhci, "Mismatch between completed Set TR Deq Ptr command & xHCI internal state.\n");
 			xhci_warn(xhci, "ep deq seg = %p, deq ptr = %p\n",
@@ -2240,9 +2242,7 @@ static int finish_td(struct xhci_hcd *xhci, struct xhci_virt_ep *ep,
 	}
 
 	/* Update ring dequeue pointer */
-	ep_ring->dequeue = td->last_trb;
-	ep_ring->deq_seg = td->last_trb_seg;
-	ep_ring->num_trbs_free += td->num_trbs - 1;
+	move_deq(xhci, ep_ring, td->last_trb_seg, td->last_trb);
 	inc_deq(xhci, ep_ring);
 
 	return xhci_td_cleanup(xhci, td, ep_ring, td->status);
@@ -2462,9 +2462,7 @@ static int skip_isoc_td(struct xhci_hcd *xhci, struct xhci_td *td,
 	frame->actual_length = 0;
 
 	/* Update ring dequeue pointer */
-	ep->ring->dequeue = td->last_trb;
-	ep->ring->deq_seg = td->last_trb_seg;
-	ep->ring->num_trbs_free += td->num_trbs - 1;
+	move_deq(xhci, ep->ring, td->last_trb_seg, td->last_trb);
 	inc_deq(xhci, ep->ring);
 
 	return xhci_td_cleanup(xhci, td, ep->ring, status);
-- 
2.33.0.685.g46640cef36-goog