Re: hci_ldisc: suspicious/buggy code

From: Pavel Machek
Date: Mon Oct 17 2016 - 17:54:31 EST


On Mon 2016-10-17 23:32:06, Pavel Machek wrote:
> Hi!
>
> hci_uart_set_flow_control() contains some rather suspicious
> code. AFAICT:
>
> set = 0.
>
> Then we do set &= (magic constant). But as set is 0, it stays zero. It
> just does not make sense. [Nor does it make sense to do |= on known
> value of zero just after else. But that's probably just some artifact
> of editing.]
>
> (And if you edit the code anyway, temporary variable to store
> TIOCM_DTR | TIOCM_RTS | TIOCM_OUT1 | TIOCM_OUT2 | TIOCM_LOOP would
> make sense.)

I think this might be the fix, but it breaks h4p protocol on n900...

Pavel

diff --git a/drivers/bluetooth/hci_ldisc.c b/drivers/bluetooth/hci_ldisc.c
index 0506806..b549c47 100644
--- a/drivers/bluetooth/hci_ldisc.c
+++ b/drivers/bluetooth/hci_ldisc.c
@@ -271,8 +271,8 @@ void hci_uart_set_rts(struct hci_uart *hu, bool enable)
status = tty->driver->ops->tiocmget(tty);
BT_DBG("Current tiocm 0x%x", status);

- set &= ~(TIOCM_OUT2 | TIOCM_RTS);
- clear = ~set;
+ clear |= (TIOCM_OUT2 | TIOCM_RTS);
+ set = ~clear;
set &= TIOCM_DTR | TIOCM_RTS | TIOCM_OUT1 |
TIOCM_OUT2 | TIOCM_LOOP;
clear &= TIOCM_DTR | TIOCM_RTS | TIOCM_OUT1 |

> /* Flow control or un-flow control the device */
> void hci_uart_set_flow_control(struct hci_uart *hu, bool enable)
> {
> ...
> unsigned int set = 0;
> unsigned int clear = 0;
>
> if (enable) {
> ...
> set &= ~(TIOCM_OUT2 | TIOCM_RTS);
> clear = ~set;
> set &= TIOCM_DTR | TIOCM_RTS | TIOCM_OUT1 |
> TIOCM_OUT2 | TIOCM_LOOP;
> clear &= TIOCM_DTR | TIOCM_RTS | TIOCM_OUT1 |
> TIOCM_OUT2 | TIOCM_LOOP;
> status = tty->driver->ops->tiocmset(tty, set, clear);
> BT_DBG("Clearing RTS: %s", status ? "failed" : "success");
> } else {
> ...
> set |= (TIOCM_OUT2 | TIOCM_RTS);
>



--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Attachment: signature.asc
Description: Digital signature