Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 2 Apr 2004 22:20:11 +0800
From:      Xin LI <delphij@frontfree.net>
To:        Claudiu Dragalina-Paraipan <clau@reversedhell.net>
Cc:        freebsd-current@FreeBSD.org
Subject:   Re: P4 HT with SMP kernel, FreeBSD 5.2.1-p4 problem
Message-ID:  <20040402142011.GA3491@frontfree.net>
In-Reply-To: <406C6AC0.8030506@reversedhell.net>
References:  <406C6AC0.8030506@reversedhell.net>

next in thread | previous in thread | raw e-mail | index | archive | help

--TB36FDmn/VVEgNH/
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

This is a "me too" post, and I think current@ is more appropriate
so I redirected this to current@ instead of hackers@.

I have a computer with the following hardware configuration:
Intel D865GBF, Pentium 4 2.8-E, 512MB DDR *2, 40GB PATA

And I have same problem you have with CURRENT code when having HT
enabled. Additionally, I was unable to produce a useful backtrace,
as the core dumps generated shows addresses like 0xc078dbf6 and
no symbols associated with it.

A hang could be triggered by the following process:
	o Start with boot -vs
	o fsck -fy
	o Go to multiuser mode
	o Run cvsup, and press Ctrl+Alt+Del
You will likely to get the box hanged there after vulru stopped.

A second panic, which is more strange, appears to be possibly triggered
by the following process:
	o Right after the system hangs, reset it
	o boot directly to multiuser mode, and login as root
	o killall sleep - this will start bgfsck immediately
	o then, press Ctrl+Alt+Del
Strangely, this often causes a double panic and the panic message
seemed to be messy, looks like that the two logical CPUs are intending
to output panic message in the same time and the message appears to
be a mixture of them.

Another case is to panic somewhere at ffs_balloc with kernel trap 12.

Unfortunatelly, the core dumps produced by these panic appears to have
damaged call stacks, as only dumpadd could be resolved and all other
addresses are resolved to "??" so I don't think they are useful here.
Because the box is intending to be serving as a production mail server
it will not be acceptable to reboot it remotely and try a untested new
kernel, but if someone has some idea I can help do some tests when I
am at work, on that computer.

Personally, I guess there might exist a problem which is lock/mutex
related. Being lack of time I have disabled HTT support as a work around,
and of course, any help would be appreciated, for example, to let it
possible to produce a useful crashdump :-)

Not sure if this is BIOS related (I don't believe so, though), I have
tried both version P10 and P14 on the motherboard.

More information available up on request. Thanks in advance!

On Thu, Apr 01, 2004 at 10:17:20PM +0300, Claudiu Dragalina-Paraipan wrote:
> Delivered-To: freebsd-hackers@freebsd.org
> Date: Thu, 01 Apr 2004 22:17:20 +0300
> From: Claudiu Dragalina-Paraipan <clau@reversedhell.net>
> User-Agent: Mozilla Thunderbird 0.5 (X11/20040301)
> To: freebsd-hackers@freebsd.org
> Subject: P4 HT with SMP kernel, FreeBSD 5.2.1-p4 problem
> X-BeenThere: freebsd-hackers@freebsd.org
>=20
> Hello,
>=20
> I have a Pentium 4 2.8E (with HT), and an Intel 865GBF motherboard.
> After I made a bios upgrade I have this problem with it: every time I do=
=20
> a reboot or a shutdown the kernel hangs. Sometimes it gives a specific=20
> error message, sometimes the last lines of text are garbage, and rarely=
=20
> it just resets before it should.
> Seems that it happens before the file system is unmounted, because every=
=20
> next boot I have problems with it. Until now I didn't lose any important=
=20
> file, but I assume that this can happen giving this situation.
> Also, this does not happen if I compile a kernel without SMP support, or=
=20
> if I disable HT from bios (with SMP or not).
> Running a kernel with INVARIANTS and WITNESS seems to fix this problem to=
o.
>=20
> I have attached a dmesg result after a boot -v.
> Please let me know what other informations you need for solving this=20
> situation.


Cheers,
--=20
Xin LI <delphij frontfree net>	http://www.delphij.net/
See complete headers for GPG key and other information.


--TB36FDmn/VVEgNH/
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (FreeBSD)

iD8DBQFAbXabOfuToMruuMARAiHNAJ49q+v/vL37OjGyMswXHyfpBJLKHgCfcx+p
bcR9wnKqGAjebcN0ivqSk28=
=7/LP
-----END PGP SIGNATURE-----

--TB36FDmn/VVEgNH/--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040402142011.GA3491>