Re: [patch] Attempting to bring arcmsr(4) up to date

[Sevan Janiyan <venture37%geeklan.co.uk@localhost>]

Hello,
Forgot to post the updated diff[1].

On 11/05/17 05:57, Sevan Janiyan wrote:
> As a work around to get the driver "working", by reducing the tested
> values from 3 down to 2 in arc_msgbuf()
> 
> if ((rlen > 3) && (rdone == 3)) {
> 	rlen = *(u_int16_t *)rwbuf;
> 	rlen = sizeof(struct arc_fw_bufhdr) + rlen + 1;
> 	}
> 
> if ((rlen > 2) && (rdone == 2)) { ... }

Ended up resolving this issue by not reusing the rlen variable halfway
through the process but using a new variable instead, rlen2 and using
that to do the calculation of how much IO we've done so far.

if ((rlen > 3) && (rdone == 3)) {
	rlen2 = *(u_int16_t *)rwbuf;
	rlen2 = sizeof(struct arc_fw_bufhdr) + rlen2 + 1;
	if ((rdone + rwlen) > rlen2) {
		DNPRINTF(ARC_D_DB, "%s:  rwbuf too big\n", device_xname(sc->sc_dev));
		error = EIO;
		goto out;
	}
}

This results in the driver binding successfully and is happy to run
without crashing.[1]
Moving on from there, it turns out the sensor framework stuff and bioctl
support is completely broken. I've not made any progress with that yet
(haven't tried).


Sevan
[1] http://www.netbsd.org/~sevan/patch-arcmsr-mk4.txt
[2] http://dmesgd.nycbug.org/index.cgi?do=view&id=3391