Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 18 Mar 2018 20:50:49 -0700
From:      Mark Millard <marklmi26-fbsd@yahoo.com>
To:        Steve Wills <swills@FreeBSD.org>
Cc:        Nathan Whitehorn <nwhitehorn@freebsd.org>, freebsd-ppc@freebsd.org
Subject:   Re: fatal kernel trap
Message-ID:  <401F7AE1-3B2B-43C8-A643-316596556D62@yahoo.com>
In-Reply-To: <7708841b-a124-7cf1-8a17-8d55fa837d59@FreeBSD.org>
References:  <d2e22843-7fb2-c2c9-1dfc-70d019ca24e7@FreeBSD.org> <CAHSQbTAUwQHu3SQMVC098=s7EBHEgzf6U1w8kmxU9b4fTf3x1A@mail.gmail.com> <f02d76b2-0c09-290b-9f6d-65eff8fd36e8@FreeBSD.org> <a3ec563a-1b76-1525-08d8-d0a5a427fb20@freebsd.org> <7708841b-a124-7cf1-8a17-8d55fa837d59@FreeBSD.org>

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


On 2018-Mar-18, at 8:16 PM, Steve Wills <swills at FreeBSD.org> wrote:

> Sure. The system has 8GB of RAM and 2 CPUs. The root fs is on UFS and =
it also has a disk with ZFS. The panic doesn't happen until very late in =
boot, when it's mounting disks. I updated the whole system to r328835 =
back in Feb. Then I switched the compiler over gcc 6.3.0 and rebuilt =
kernel/world again. Not sure what other info might be helpful, let me =
know if there's some particular that might help.
>=20
> Steve
>=20
> On 03/17/2018 19:57, Nathan Whitehorn wrote:
>> Could you provide any other detail on your system? Amount of RAM, =
etc.?
>> -Nathan
>> On 03/17/18 08:39, Steve Wills wrote:
>>> Finished bisecting, r329611 boots fine, r329612 does not.
>>>=20
>>> Steve
>>>=20
>>> On 03/16/2018 10:43, Justin Hibbits wrote:
>>>>=20
>>>> On Mar 16, 2018 09:29, "Steve Wills" <swills@freebsd.org =
<mailto:swills@freebsd.org>> wrote:
>>>>=20
>>>>     My PowerMac G5 runs r328835 fine, but upgrading to r330240 =
results in:
>>>>=20
>>>>     fatal kernel trap:
>>>>=20
>>>>        exception           =3D 0x300 (data storage interrupt)
>>>>        virtual address     =3D 0xc186bff8
>>>>        dsisr               =3D 0x40000000
>>>>        srr0                =3D 0x8ef510 (0x8ef510)
>>>>        srr1                =3D 0x9000000000009032
>>>>        lr                  =3D 0x7be3e4 (0x7be3e4)
>>>>        curthread           =3D 0x11b26560
>>>>               pid =3D 38, comm =3D kldload

Given that it reports "kldload": Do you have a way to do a simple
boot that would avoid loading any .ko files?

In my clang-based experiments I ran into .ko files getting a
different type of linkage (R_PPC64_JMP_SLOT) for the kldload to
resolve (that it did not handle, leading to panics). I ended up
adding:

device 	filemon
device 	geom_label

to the kernel so that my normal activity could happen without
kldload's of .ko files.

[This goes back to 2017-Dec-24 or so. I've not been very active
on FreeBSD most of the time after that.]

I'm not claiming you have R_PPC64_JMP_SLOT use. I'm just
illustrating avoiding kldload activity as a possible test.

>>>>     [ thread pid 38 tid 100087 ]
>>>>     Stopped at      strchr+0x70:    lbzu   r10, 0x1(r4)
>>>>     db>
>>>>=20
>>>>     Note this was transcribed by hand and may have typos. Also note =
my
>>>>     kernel was compiled with gcc 6.3.0, although the previous =
kernel was
>>>>     as well and it works fine.
>>>>=20
>>>>     Any suggestions would be appreciated.
>>>>=20
>>>>     Thanks,
>>>>     Steve
>>>>=20
>>>>=20
>>>> Hi Steve,
>>>>=20
>>>> I see the same thing with any attempt to upgrade the kernel on my =
G5 (PowerMac11,2) from a November 6 build to more recent. I've seen it =
since the January timeframe, and haven't yet figured it out.
>>>>=20
>>>> - Justin



=3D=3D=3D
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?401F7AE1-3B2B-43C8-A643-316596556D62>