Firewire drives not recognized when I load sbp

[Sean Bruno]

Jim:

thanks for the detailed information.  Let me digest and review this
tonight and see if I can determine what the badness here is.

Sean


On Sat, 2011-08-06 at 08:55 -0700, Jim Pryor wrote:
> I also posted this to the Peripheral Hardware section of the forums.
> Apologies if it's inappropriate to also mail here. This mailing list
> seems the right place to discuss my issue, but it looks like it's been
> very quiet for the past year.
> 
> 
> What I'll call Computer Bad: Mac mini (Intel Core 2), was running
> FreeBSD 8.2-release, tried upgrading to 8-STABLE, but these issues
> persist. Has a single Firewire 400 port.
> 
> Device 1: a dual Firewire enclosure, has both a Firewire 400 port and
> two Firewire 800 ports. Contains two drives.
> 
> Device 2: a pocket-sized Firewire enclosure, has one Firewire 400 and
> one Firewire 800 port.
> 
> Device 3 is this: <http://tech.acomdata.com/minipal.asp>. It attaches to
> the mini via Firewire and USB, and provides a hub for each, and
> additionally has a drive on the Firewire bus. We can ignore the USB
> aspects of this.
> 
> I'll compare against Computer Good: which is a Sony Vaio, whose details
> probably not important, except it's also running FreeBSD 8-STABLE. After
> I do a `kldload sbp` on this machine, and attach either of Device 1 or
> 2, or both of them in a chain, they all are immediately recognized and
> show up as /dev/da0, da1, and da2.
> 
> However that's just for testing. I need to get these devices recognized
> on Computer Bad.
> 
> And on Computer Bad, **when Device 3 is in the mix**, and the other
> devices are chained off it, I can do so--- although it takes some
> voodoo. I cannot have the sbp module loaded at boot time; if I do so the
> boot will hang until I disconnect or power down all attached Firewire
> devices. However, if I wait until boot has completed, then do a `kldload
> sbp`, once or twice the devices were then recognized. Most often,
> though, I have to do a dance of unplugging the devices, replugging them,
> power-cycling them, fwcontrol -r'ing, and so on. If there's a system
> that works reliably, I haven't been able to find it. Nor have I found
> any way to make this work solely from the commandline. `camcontrol
> rescan all` hangs until I disconnect or power down the devices.
> `fwcontrol -r` resets the bus, but the devices don't show up. And so on.
> However, with some persistence at the cable/power-cycling dance, I have
> always been able to get the drives recognized when they're chained off
> of Device 3. And once the drives get recognized, they give me no further
> trouble during that session. Even if I then power Device 3 down. I've
> transferred 1 TB of data to Device 1, I think without any trouble.
> 
> I'd live with this if I could. However, I can't keep Device 3 in this
> setup; it's required elsewhere.
> 
> And **without Device 3**, I can't find any way to get Devices 1 and 2,
> either together or solo, to be recognized on Computer Bad. I've tried
> connecting to either their Firewire 400 or their Firewire 800 ports
> (though on the other side, Computer Bad is only attempting Firewire
> 400). I've tried doing the `kldload sbp` before connecting them, or
> after, or whatever, and no luck. I've tried `sysctl -w
> hw.firewire.sbp.auto_login=0` and `sysctl -w
> hw.firewire.nocyclemaster=1` at various points, but this doesn't seem to
> change anything.
> 
> 
> These problems showed up when Computer Bad had a fresh 8.2-release
> installation. I tried upgrading to 8-STABLE, which is what Computer Good
> had, but no joy there.
> 
> I've been searching and searching but haven't found much, even ideas to
> try. Any suggestions would be much appreciated.
> 
> Here are some details.
> 
> 1. Here's uname -a, pciconf -lcv, kldstat, and sysctl hw.firewire on Computer Good:
>    <http://dpaste.com/hold/588061/>
> 2. Here's the stretch of dmesg where the devices are connected and
>    successfully recognized on Computer Good:
>    <http://dpaste.com/hold/588063/>
> 3. And here's what I then see on Computer Good (fwcontrol, fwcontrol -t,
>    and camcontrol devlist -v): <http://dpaste.com/hold/588066/>
> 
> 4. Here's uname -a, pciconf -lcv, kldstat, and sysctl hw.firewire on
>    Computer Bad: <http://dpaste.com/hold/588062/>. Notice the
>    hw.firewire.sbp.auto_login is at 0, that's because following some
>    diagnostic advice I found while searching I disabled that in
>    /boot/loader.conf. It doesn't seem to affect my results at all.
> 5. Here's the stretch of dmesg where the devices are connected but fail
>    to be recognized on Computer Bad: <http://dpaste.com/hold/588076/>
> 6. Here are fwcontrol, fwcontrol -t, and camcontrol devlist -v on
>    Computer Bad, before kldloading sbp and attaching the devices:
>    <http://dpaste.com/hold/588068/>
> 7. Here are the results of the same, after kldloading sbp and attaching
>    the devices: <http://dpaste.com/hold/588069/>. At this point, doing
>     `sysctl -w hw.firewire.sbp.auto_login=1; fwcontrol -r` seems to produce
>     no change. Allowing sbp to load with auto_login=1 from the beginning
>     produces no change either.
> 
> At some points in my experimenting, I was able to get one or two
> additional entries in the `fwcontrol` output on Computer Bad, but I
> can't reproduce that now. And I didn't notice any additional progress
> when that occurred. 
>