2002-04-09 16:19:54

by Randy.Dunlap

Subject: 2.5.7 and runaway modprobe loop?

Hi,

If I build/boot 2.5.7 with 64 GB support (with or without
high_pte), I get:

Freeing unused kernel memory: 448k freed
INIT: version 2.78 booting
kmod: runaway modprobe loop assumed and stopped
kmod: runaway modprobe loop assumed and stopped
kmod: runaway modprobe loop assumed and stopped
kmod: runaway modprobe loop assumed and stopped
kmod: runaway modprobe loop assumed and stopped

If I build/boot it with 4 GB support, it boots fine.

Fixes, suggestion?

This is a 16 GB 8-proc machine (IBM SMP).

Thanks,
--
~Randy

2002-04-10 06:52:07

by Keith Owens

Subject: Re: 2.5.7 and runaway modprobe loop?

On Tue, 9 Apr 2002 09:17:08 -0700 (PDT),
"Randy.Dunlap" <[email protected]> wrote:
>If I build/boot 2.5.7 with 64 GB support (with or without
>high_pte), I get:
>
>Freeing unused kernel memory: 448k freed
>INIT: version 2.78 booting
>kmod: runaway modprobe loop assumed and stopped
>kmod: runaway modprobe loop assumed and stopped
>kmod: runaway modprobe loop assumed and stopped
>kmod: runaway modprobe loop assumed and stopped
>kmod: runaway modprobe loop assumed and stopped
>
>If I build/boot it with 4 GB support, it boots fine.
>
>Fixes, suggestion?

If /var/log/ksymoops exists and is mounted rw when that message is
issued, look at 20020409.log to see what modprobe is asking for.
Otherwise add a printk to kernel/kmod::request_module() to print
module_name. In either case, work out why it is going recursive.

2002-04-10 18:03:51

by Randy.Dunlap

Subject: Re: 2.5.7 and runaway modprobe loop?

On Wed, 10 Apr 2002, Keith Owens wrote:

| On Tue, 9 Apr 2002 09:17:08 -0700 (PDT),
| "Randy.Dunlap" <[email protected]> wrote:
| >If I build/boot 2.5.7 with 64 GB support (with or without
| >high_pte), I get:
| >
| >Freeing unused kernel memory: 448k freed
| >INIT: version 2.78 booting
| >kmod: runaway modprobe loop assumed and stopped
| >kmod: runaway modprobe loop assumed and stopped
| >kmod: runaway modprobe loop assumed and stopped
| >kmod: runaway modprobe loop assumed and stopped
| >kmod: runaway modprobe loop assumed and stopped
| >
| >If I build/boot it with 4 GB support, it boots fine.
| >
| >Fixes, suggestion?
|
| If /var/log/ksymoops exists and is mounted rw when that message is
| issued, look at 20020409.log to see what modprobe is asking for.
| Otherwise add a printk to kernel/kmod::request_module() to print
| module_name. In either case, work out why it is going recursive.

Thanks for replying, Keith.

I added module_name to the runaway message (OK ?) and its
answer is binfmt-0000.

I also moved from 2.5.7 to 2.5.8-pre2 and don't have this
problem.

--
~Randy

2002-04-10 23:28:03

by Keith Owens

Subject: Re: 2.5.7 and runaway modprobe loop?

On Wed, 10 Apr 2002 11:00:49 -0700 (PDT),
"Randy.Dunlap" <[email protected]> wrote:
>| On Tue, 9 Apr 2002 09:17:08 -0700 (PDT),
>| "Randy.Dunlap" <[email protected]> wrote:
>| >If I build/boot 2.5.7 with 64 GB support (with or without
>| >high_pte), I get:
>| >
>| >Freeing unused kernel memory: 448k freed
>| >INIT: version 2.78 booting
>| >kmod: runaway modprobe loop assumed and stopped
>| >
>| >If I build/boot it with 4 GB support, it boots fine.
>
>I added module_name to the runaway message (OK ?) and its
>answer is binfmt-0000.
>
>I also moved from 2.5.7 to 2.5.8-pre2 and don't have this
>problem.

Interesting. The binfmt-0000 implies that search_binary_handler() is
reading garbage for the executable. It tries to load a handler for
binfmt-0000 which tries to execute modprobe which hits the same bug.
At a guess, the executable binary is not being mapped correctly with
64GB support.