Re: [PATCH] synclink_gt fix transmit race and timeout

[Paul Fulghum <paulkf@microgate.com>]

Andrew Morton wrote:
> I'm referring to
> 
>   Where did the "7" come from?

Oops, I did not scroll down far enough.

+	/*
+	 * use worst case speed of 1200bps to calculate transmit timeout
+	 * based on data in buffers (tbuf_bytes) and FIFO (128 bytes)
+	 */
+	if (info->params.mode == MGSL_MODE_HDLC) {
+		int timeout  = (tbuf_bytes(info) * 7) + 1000;
+		mod_timer(&info->tx_timer, jiffies + msecs_to_jiffies(timeout));
+	}

7 is roughly the number of milliseconds to send a byte at 1200bps.

The problem with externally provided data clocks is that you
don't necessarily know the data rate before hand, so a somewhat
arbitrary worst case assumption is used.

> and
> 
>   I have a suspicion that tx_stop() should use del_timer_sync(), not
>   del_timer().  What happens if the timer handler is concurrently
>   running?

Everything is synchronized with info->lock spin_lock,
so nothing critical runs concurrently. tx_stop() is sometimes
called in interrupt context so it can't call del_timer_sync().
If the timer has already fired but has not
run yet it does nothing more than call tx_stop() itself and wake
any transmit waiters so there are no ill effects.

-- 
Paul Fulghum
MicroGate Systems, Ltd.
=Customer Driven, by Design=
(800)444-1982
(512)345-7791 (Direct)
(512)343-9046 (Fax)
Central Time Zone (GMT -5h)
www.microgate.com