[PATCH 1/1] Bluetooth: make the balance of judgement condition to fix a false report

[PATCH 1/1] Bluetooth: make the balance of judgement condition to fix a false report

[Zumeng Chen]

This patch is to balance the condition scope between hci_get_cmd_complete and
hci_event_packet about orig_skb as follows: 

        if (req_complete_skb || event == HCI_EV_CMD_STATUS ||
            event == HCI_EV_CMD_COMPLETE)
                orig_skb = skb_clone(skb, GFP_KERNEL);

And hci_get_cmd_complete will bt_dev_err out when HCI_EV_CMD_STATUS, so a lot
of asymmetric conditions are triggered. Since both of them are the entry into
hci_get_cmd_complete, we'd better get STATUS into judge the false out there.

Signed-off-by: Zumeng Chen <zumeng.chen@gmail.com>
---

Hi expert,

This issue existed whether or not T_DBG had been changed into bt_dev_err, which
just shows the issue explicitly. I noticed actually that opcode doesn't match
ev->opcode either at the same time. And there might be some logic issue about
HCI_EV_CMD_COMPLETE between protocol and drivers. I'm not familar with the whole
bluetooth protocol, and not gonna to dig more, so all yours guys~

Cheers,
Zumeng

 net/bluetooth/hci_event.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
index 235b5aa..d848663 100644
--- a/net/bluetooth/hci_event.c
+++ b/net/bluetooth/hci_event.c
@@ -5217,7 +5217,8 @@ static bool hci_get_cmd_complete(struct hci_dev *hdev, u16 opcode,
 		return true;
 	}
 
-	if (hdr->evt != HCI_EV_CMD_COMPLETE) {
+	if (!((hdr->evt == HCI_EV_CMD_COMPLETE) ||
+		(hdr->evt == HCI_EV_CMD_STATUS))) {
 		bt_dev_err(hdev, "last event is not cmd complete (0x%2.2x)",
 			   hdr->evt);
 		return false;
-- 
2.7.4

Re: [PATCH 1/1] Bluetooth: make the balance of judgement condition to fix a false report

[Marcel Holtmann]

Hi Zumeng,

> This patch is to balance the condition scope between hci_get_cmd_complete and
> hci_event_packet about orig_skb as follows: 
> 
>        if (req_complete_skb || event == HCI_EV_CMD_STATUS ||
>            event == HCI_EV_CMD_COMPLETE)
>                orig_skb = skb_clone(skb, GFP_KERNEL);
> 
> And hci_get_cmd_complete will bt_dev_err out when HCI_EV_CMD_STATUS, so a lot
> of asymmetric conditions are triggered. Since both of them are the entry into
> hci_get_cmd_complete, we'd better get STATUS into judge the false out there.
> 
> Signed-off-by: Zumeng Chen <zumeng.chen@gmail.com>
> ---
> 
> Hi expert,
> 
> This issue existed whether or not T_DBG had been changed into bt_dev_err, which
> just shows the issue explicitly. I noticed actually that opcode doesn't match
> ev->opcode either at the same time. And there might be some logic issue about
> HCI_EV_CMD_COMPLETE between protocol and drivers. I'm not familar with the whole
> bluetooth protocol, and not gonna to dig more, so all yours guys~
> 
> Cheers,
> Zumeng
> 
> net/bluetooth/hci_event.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
> index 235b5aa..d848663 100644
> --- a/net/bluetooth/hci_event.c
> +++ b/net/bluetooth/hci_event.c
> @@ -5217,7 +5217,8 @@ static bool hci_get_cmd_complete(struct hci_dev *hdev, u16 opcode,
> 		return true;
> 	}
> 
> -	if (hdr->evt != HCI_EV_CMD_COMPLETE) {
> +	if (!((hdr->evt == HCI_EV_CMD_COMPLETE) ||
> +		(hdr->evt == HCI_EV_CMD_STATUS))) {

this indentation is messed up. Also some braces are not needed.

	if (!(hdr->evt == HCI_EV_CMD_COMPLETE ||
	      hdr->evt == HCI_EV_CMD_STATUS)) {

> 		bt_dev_err(hdev, "last event is not cmd complete (0x%2.2x)",
> 			   hdr->evt);
> 		return false;

Regards

Marcel

[PATCH 1/1 v2] Bluetooth: make the balance of judgement condition to fix a false report

[Zumeng Chen]

This patch is to balance the condition scope between hci_get_cmd_complete and
hci_event_packet about orig_skb as follows:

        if (req_complete_skb || event == HCI_EV_CMD_STATUS ||
            event == HCI_EV_CMD_COMPLETE)
                orig_skb = skb_clone(skb, GFP_KERNEL);

And hci_get_cmd_complete will bt_dev_err out when HCI_EV_CMD_STATUS, so a lot
of asymmetric conditions are triggered. Since both of them are the entry into
hci_get_cmd_complete, we'd better get STATUS into judge the false out there.

Signed-off-by: Zumeng Chen <zumeng.chen@gmail.com>
---

v2: remove redundant braces and adjust the indentation.

Cheers,
Zumeng

 net/bluetooth/hci_event.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
index 235b5aa..1d2a8fe 100644
--- a/net/bluetooth/hci_event.c
+++ b/net/bluetooth/hci_event.c
@@ -5217,7 +5217,8 @@ static bool hci_get_cmd_complete(struct hci_dev *hdev, u16 opcode,
 		return true;
 	}
 
-	if (hdr->evt != HCI_EV_CMD_COMPLETE) {
+	if (!(hdr->evt == HCI_EV_CMD_COMPLETE ||
+	      hdr->evt == HCI_EV_CMD_STATUS)) {
 		bt_dev_err(hdev, "last event is not cmd complete (0x%2.2x)",
 			   hdr->evt);
 		return false;
-- 
2.7.4

Re: [PATCH 1/1 v2] Bluetooth: make the balance of judgement condition to fix a false report

[Luiz Augusto von Dentz]

Hi Marcel,
On Thu, Nov 15, 2018 at 3:37 AM Zumeng Chen <zumeng.chen@gmail.com> wrote:
>
> This patch is to balance the condition scope between hci_get_cmd_complete and
> hci_event_packet about orig_skb as follows:
>
>         if (req_complete_skb || event == HCI_EV_CMD_STATUS ||
>             event == HCI_EV_CMD_COMPLETE)
>                 orig_skb = skb_clone(skb, GFP_KERNEL);
>
> And hci_get_cmd_complete will bt_dev_err out when HCI_EV_CMD_STATUS, so a lot
> of asymmetric conditions are triggered. Since both of them are the entry into
> hci_get_cmd_complete, we'd better get STATUS into judge the false out there.
>
> Signed-off-by: Zumeng Chen <zumeng.chen@gmail.com>
> ---
>
> v2: remove redundant braces and adjust the indentation.
>
> Cheers,
> Zumeng
>
>  net/bluetooth/hci_event.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
> index 235b5aa..1d2a8fe 100644
> --- a/net/bluetooth/hci_event.c
> +++ b/net/bluetooth/hci_event.c
> @@ -5217,7 +5217,8 @@ static bool hci_get_cmd_complete(struct hci_dev *hdev, u16 opcode,
>                 return true;
>         }
>
> -       if (hdr->evt != HCI_EV_CMD_COMPLETE) {
> +       if (!(hdr->evt == HCI_EV_CMD_COMPLETE ||
> +             hdr->evt == HCI_EV_CMD_STATUS)) {
>                 bt_dev_err(hdev, "last event is not cmd complete (0x%2.2x)",
>                            hdr->evt);
>                 return false;
> --
> 2.7.4

It appears we need this also for enabling vendor_diag with intel controllers:

[399314.236288] hci_cmd_status_evt:3138: hci0 opcode 0xfc43
[399314.236291] Bluetooth: hci0: last event is not cmd complete (0x0f)
[399314.236359] Bluetooth: hci0: Changing Intel diagnostic mode failed (-16)


-- 
Luiz Augusto von Dentz