Skip to contentSkip to conversation

mei: hbm: drop BUG() from the hbm handler

PATCH New Tomas Winkler 1 participant 3 messages Nov 6, 2018

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/3 mei: hbm: drop BUG() from the hbm handler

Drop BUG() from the hbm handler in order not to crash the whole kernel on faulty firmware implementation. Instead of it, just return an error resulting into link reset.
There is no any known issue of faulty firmware in this matter, the change is just to ease the development.

---
 drivers/misc/mei/hbm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/misc/mei/hbm.c +2 -2
index e56f3e72d57a..fe67381b34c3 100644
@@ -1271,8 +1271,8 @@ int mei_hbm_dispatch(struct mei_device *dev, struct mei_msg_hdr *hdr)
 		break;
 
 	default:
-		BUG();
-		break;
+		WARN(1, "hbm: wrong command %d\n", mei_msg->hbm_cmd);
+		return -EPROTO;
 
 	}
 	return 0;
Signature · 1 line
2.14.5
#1 lore raw

PATCH 2/3 mei: flush queues by the host client only

From: Alexander Usyskin <[email protected]>

During queues flush, the me client in most cases is already unlinked hence the me client id is unavailable. The host client structure pointer is enough for identification.
The function mei_cl_cmp_id() is dropped as it has no more usage.

---
 drivers/misc/mei/client.c | 21 ++-------------------
 1 file changed, 2 insertions(+), 19 deletions(-)
drivers/misc/mei/client.c +2 -19
index ebdcf0b450e2..ee58561ceb08 100644
@@ -317,23 +317,6 @@ void mei_me_cl_rm_all(struct mei_device *dev)
 	up_write(&dev->me_clients_rwsem);
 }
 
-/**
- * mei_cl_cmp_id - tells if the clients are the same
- *
- * @cl1: host client 1
- * @cl2: host client 2
- *
- * Return: true  - if the clients has same host and me ids
- *         false - otherwise
- */
-static inline bool mei_cl_cmp_id(const struct mei_cl *cl1,
-				const struct mei_cl *cl2)
-{
-	return cl1 && cl2 &&
-		(cl1->host_client_id == cl2->host_client_id) &&
-		(mei_cl_me_id(cl1) == mei_cl_me_id(cl2));
-}
-
 /**
  * mei_io_cb_free - free mei_cb_private related memory
  *
@@ -418,7 +401,7 @@ static void mei_io_list_flush_cl(struct list_head *head,
 	struct mei_cl_cb *cb, *next;
 
 	list_for_each_entry_safe(cb, next, head, list) {
-		if (mei_cl_cmp_id(cl, cb->cl))
+		if (cl == cb->cl)
 			list_del_init(&cb->list);
 	}
 }
@@ -435,7 +418,7 @@ static void mei_io_tx_list_free_cl(struct list_head *head,
 	struct mei_cl_cb *cb, *next;
 
 	list_for_each_entry_safe(cb, next, head, list) {
-		if (mei_cl_cmp_id(cl, cb->cl))
+		if (cl == cb->cl)
 			mei_tx_cb_dequeue(cb);
 	}
 }
Signature · 1 line
2.14.5
#2 lore raw

PATCH 3/3 mei: expedite ioctl return on the notify set operation error

In reply to Tomas Winkler

From: Alexander Usyskin <[email protected]>

The notify set operation ioctl will wait till timeout is expired even in case when the FW returned an error.
Check the status field of the client object in wait_event_timeout() to determine if the caller can return earlier.

---
 drivers/misc/mei/client.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/misc/mei/client.c +3 -1
index ee58561ceb08..1fe9426ce48b 100644
@@ -1357,7 +1357,9 @@ int mei_cl_notify_request(struct mei_cl *cl,
 
 	mutex_unlock(&dev->device_lock);
 	wait_event_timeout(cl->wait,
-			   cl->notify_en == request || !mei_cl_is_connected(cl),
+			   cl->notify_en == request ||
+			   cl->status ||
+			   !mei_cl_is_connected(cl),
 			   mei_secs_to_jiffies(MEI_CL_CONNECT_TIMEOUT));
 	mutex_lock(&dev->device_lock);
 
Signature · 1 line
2.14.5