Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 29 Aug 2013 09:10:49 -0700
From:      Adrian Chadd <adrian@freebsd.org>
To:        Ian Lepore <ian@freebsd.org>
Cc:        "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org>, Ronald Klop <ronald-freebsd8@klop.yi.org>
Subject:   Re: Reminder: Removal of WITHOUT_ARM_EABI
Message-ID:  <CAJ-VmonVdMVD3=O-COJ8yJq-ysHzc3OBuiNE-tJcznnta8HFTw@mail.gmail.com>
In-Reply-To: <1377787319.1111.277.camel@revolution.hippie.lan>
References:  <20130820091527.42127170@bender.Home> <1377271598.1111.78.camel@revolution.hippie.lan> <op.w2k0j5i08527sy@212-182-167-131.ip.telfort.nl> <CAJ-VmokTrMDJtw0OwjMamW-T=ZAx_6%2BhLxq=mxUbfLLA84ecPA@mail.gmail.com> <1377787319.1111.277.camel@revolution.hippie.lan>

next in thread | previous in thread | raw e-mail | index | archive | help
On 29 August 2013 07:41, Ian Lepore <ian@freebsd.org> wrote:

> On Thu, 2013-08-29 at 07:23 -0700, Adrian Chadd wrote:
> > Hm, which commit broke this?
> >
> > Did someone commit something to the random number framework that is
> hanging
> > the kernel early because there's not enough entropy?
> >
>
> As mentioned in the text below (which of course will get trimmed now by
> sensible mail clients because you top-posted in a bottom-posted thread),
> it has been broken since at least April.
>

.. still? Then how exactly are people commiiting to -arm? :)


> This problem is somewhat resistant to simple debugging -- the kernel
> isn't hung overall, but some userland processes become hung while
> waiting for something in the kernel.  It's not a deadlock that witness
> can detect, it seems to be something more like waiting for IO to
> complete and it never does.  Running newfs on a /dev/md# will hang for
> sure (but you can ^C out of it).  While it's in this state, you cannot
> launch top or ps or any of the usual tools for figuring out what the
> system is doing -- the tools also hang, and they do NOT respond to a ^C
> when hung (but will exit with a kill -9 from another shell).
>

Well, a 'show alllocks' and I think 'show allproc' or something like that
should show the locks+state and the PIDs w/ their sleep channel. Hopefully
they're "just" asleep waiting for some IO to occur.

If it's been broken since April - does someone have a known working
revision on any arm platform from April that is easy to just go kernel
commit revision bisecting with?



-adrian



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmonVdMVD3=O-COJ8yJq-ysHzc3OBuiNE-tJcznnta8HFTw>