From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <25f8de8836cc6c592e70cd662b61d0a2@hamnavoe.com>
To: 9fans@9fans.net
From: Richard Miller <9fans@hamnavoe.com>
Date: Thu, 14 Feb 2013 19:24:06 +0000
In-Reply-To: <54c0ba490a70bb58de6700d2c4e7bd96@brasstown.quanstro.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] usb serial bug
Topicbox-Message-UUID: 178ceb12-ead8-11e9-9d60-3106f5b1d025

I said:

>> The current Plan 9 usb architecture perpetuates the confusion by
>> referring to them both with one name epN.1, but you still have to open
>> them both independently.

Erik replied:
> in that case, shouldn't these three blocks be reverted?

Erik is right, I was talking through my hat.  It's OK to open bulk
endpoints read/write, and the kernel will do the right thing.  The
actual problem, which neither of us had spotted although it was
staring us in the face, is this:

			if((ep->dir == Ein || ep->dir == Eboth) && epin == -1)
				epin = ep->id;
			if((ep->dir == Ein || ep->dir == Eboth) && epout == -1)
				epout = ep->id;

Notice the two occurrences of Ein?  The second one obviously should
be Eout.  It was a typo (mine, I blush to admit).

My usb serial adapter uses the same ep number for input and output,
so my testing didn't reveal this error.  I think the same typo will
account for the double-free bug which Jeff (and Lucio on 4 Feb) reported.

Erik, Jeff, Lucio - please try changing the offending Ein to Eout in
/sys/src/cmd/usb/serial/serial.c:721 and see if your troubles are resolved.