Re: Kernel-Crash when working with ubt0

On Mon, Aug 26, 2019 at 6:13 PM Lizbeth Mutterhunt, Ph.D. <
mutterhunt_gesta_at_yahoo.com> wrote:

> On Monday, 26 August 2019 23:35:25 CEST maksim yevmenkin wrote:
>
> > > On Aug 26, 2019, at 9:14 AM, Warner Losh <imp_at_bsdimp.com> wrote:
>
> > >
>
> > > Is it from read_connection_list? If so I have a 'patch' that I'm using
> but
>
> > > haven't committed because it's just too gross: drop the lock before the
>
> > > copyout and pick it up again after...
>
> >
>
> > Hmm... interesting....
>
> >
>
> > I only took a brief look at it. I suppose I can ensure user space
> address is
>
> > wired and then copyout() can be called with mutex held
>
> >
>
> > It’s odd that I’ve never seen this before. I’m traveling at the moment
> will
>
> > take a closer look this later this week
>
> >
>
> > Thanks
>
> > Max
>
>
>
> Sorry for late reply, suffering on a "VPN-Jetlag".
>
>
>
> Here's the promised vmcore.o(last) <http://pastebin.com/SB0HWHyz> on
> pastepin. It's not a "mini"-dump at all but a huge skirt up to the heals
> :-) as it has 360M of text and the editor loads about ten minutes on
> grandmother's laptop with her 70 years after 36 hours of being online. The
> options are with DEBUG=-g and VERBOSE set in the config.
>
>
>
> With the ubt1, a 2070X stack by Asus it's probably the kernel driver I'd
> have to parse into the kernel as here at linux a quarrel about it is for
> years systematically delaunching "bills driver"; you get the patch as a
> .zip file on Arch, not as a .deb or .tar.gz or whatever and it must be
> compiled by hand; after that you'd to reboot and install the A2DP-fix and
> than you can pray (where ever you want to and think it might help you) with
> the initial pulseaudio-connect at boottime (sometimes it just doesn't
> recognize the adapter) - so as a LiLo-substitute I prefer EFI-Clover not
> the GRUB-stuff that doesn't work at all for BSD for many, many times) and I
> really don't know how many hours I spent with "sudo bluetoothctl" and on
> kernel builts for this 'Ubitek Magic Box'.
>
>
>
> Hope someone can think it over, I know it's old hardware but she'd (the
> laptop) to to spend 50 mankind-years in bills slavery when I got her
> finally - probably one of the SALVAGED - messages at fsck -y in single mode
> :-)
>
>
>
> So take care of it, it's not the way I normally act to sent guys boxing
> about and for the box but much more than beeing dead can't happen, can it?
> - This happened to my cat "Rambo" who just flew away from me and went
> straight away under a car - buhuhuhu.... so far for the revenge thoughts,
>
>
>
> 'svn up' is good at this, subversion is much better than git --diff, which
> really doens't work at all here.
>
>
>
Here's the patches that I'm running with. I also have to restart bthidd
after manually connecting my keyboard for reasons i've not had time to dig
into. These patches aren't quite right, since dropping the lock to copy the
data out that the lock seems to be protecting can't be 'right', it just
satisfies the invariant that you can't hold a non-sleep-able lock around
copyout.

Warner