Hi Johan,
> The initialization function used by hci_open_dev (hci_init_req) sends
> many different HCI commands. The __hci_request function should only
> return when all of these commands have completed (or a timeout occurs).
> Several of these commands cause hci_req_complete to be called which
> causes __hci_request to return prematurely.
>
> This patch fixes the issue by adding a new hdev->req_last_cmd variable
> which is set during the initialization procedure. The hci_req_complete
> function will no longer mark the request as complete until the command
> matching hdev->req_last_cmd completes.
>
> Signed-off-by: Johan Hedberg <johan.hedberg@xxxxxxxxx>
> ---
> v4: Use __u16 instead of int for req_last_cmd
>
> include/net/bluetooth/hci_core.h | 3 ++-
> net/bluetooth/hci_core.c | 10 +++++++---
> net/bluetooth/hci_event.c | 33 +++++++++++++++++++++++----------
> 3 files changed, 32 insertions(+), 14 deletions(-)
>
> diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
> index 3786ee8..a29feb0 100644
> --- a/include/net/bluetooth/hci_core.h
> +++ b/include/net/bluetooth/hci_core.h
> @@ -129,6 +129,7 @@ struct hci_dev {
> wait_queue_head_t req_wait_q;
> __u32 req_status;
> __u32 req_result;
> + __u16 req_last_cmd;
>
> struct inquiry_cache inq_cache;
> struct hci_conn_hash conn_hash;
> @@ -693,6 +694,6 @@ struct hci_sec_filter {
> #define hci_req_lock(d) mutex_lock(&d->req_lock)
> #define hci_req_unlock(d) mutex_unlock(&d->req_lock)
>
> -void hci_req_complete(struct hci_dev *hdev, int result);
> +void hci_req_complete(struct hci_dev *hdev, __u16 cmd, int result);
>
> #endif /* __HCI_CORE_H */
> diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> index 1a4ec97..787c8df 100644
> --- a/net/bluetooth/hci_core.c
> +++ b/net/bluetooth/hci_core.c
> @@ -91,9 +91,12 @@ static void hci_notify(struct hci_dev *hdev, int event)
>
> /* ---- HCI requests ---- */
>
> -void hci_req_complete(struct hci_dev *hdev, int result)
> +void hci_req_complete(struct hci_dev *hdev, __u16 cmd, int result)
> {
> - BT_DBG("%s result 0x%2.2x", hdev->name, result);
> + BT_DBG("%s command 0x%04x result 0x%2.2x", hdev->name, cmd, result);
> +
> + if (hdev->req_last_cmd && cmd != hdev->req_last_cmd)
> + return;
>
> if (hdev->req_status == HCI_REQ_PEND) {
> hdev->req_result = result;
> @@ -149,7 +152,7 @@ static int __hci_request(struct hci_dev *hdev, void (*req)(struct hci_dev *hdev,
> break;
> }
>
> - hdev->req_status = hdev->req_result = 0;
> + hdev->req_last_cmd = hdev->req_status = hdev->req_result = 0;
>
> BT_DBG("%s end: err %d", hdev->name, err);
>
> @@ -252,6 +255,7 @@ static void hci_init_req(struct hci_dev *hdev, unsigned long opt)
> /* Connection accept timeout ~20 secs */
> param = cpu_to_le16(0x7d00);
> hci_send_cmd(hdev, HCI_OP_WRITE_CA_TIMEOUT, 2, ¶m);
> + hdev->req_last_cmd = HCI_OP_WRITE_CA_TIMEOUT;
just for style consistency add an empty line before this command.
And actually hci_init_req is not the only function where you would need
to add hdev->req_last_cmd entries. There are hci_reset_req and some
others. Without that, the rest of your patch makes hci_req_complete a
non functional operation.
Regards
Marcel
--
To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
