SEO

#1 (**permalink**) 01-18-2008, 05:13 AM

The 2.4.27 kernel prepared two unpleasant surprises when
upgrading my faithful Sun SS20 (two Hypersparcs) from 2.2.21:

1. The first attempt with everything needed in the kernel,
no modules, stopped at boot with SILO complaining of
a too big uncompressed image (gunzipped it: 2.8 Mbytes).

Is there a hard limit on the Sparc / SILO image size?
AFAIK, there was still well over 220 Mbytes of free RAM
at boot time.

2. The SMP kernel hangs occasionally. It seems that the hang
is associated with heavy disk write activity, like at the
end of a mke2fs when everything is being written to disk.
The process hangs so that it cannot be killed, even with
a kill -15. The rest of the system stays available, so
it is possible to log on from another terminal and bring
the thing cleanly down. Spinlock or uniterruptible sleep?
A known issue with 2.4.27-SMP?

Uniprocessor compile of the very same kernel works (but
slower, of course, sigh!).

Any light / pointers?

TIA

Tauno Voipio
tauno voipio (at) iki fi

#2 (**permalink**) 01-18-2008, 05:21 AM

Tauno Voipio wrote:
> The 2.4.27 kernel prepared two unpleasant surprises when
> upgrading my faithful Sun SS20 (two Hypersparcs) from 2.2.21:
>
> 1. The first attempt with everything needed in the kernel,
> no modules, stopped at boot with SILO complaining of
> a too big uncompressed image (gunzipped it: 2.8 Mbytes).
>
> Is there a hard limit on the Sparc / SILO image size?
> AFAIK, there was still well over 220 Mbytes of free RAM
> at boot time.
>
> 2. The SMP kernel hangs occasionally. It seems that the hang
> is associated with heavy disk write activity, like at the
> end of a mke2fs when everything is being written to disk.
> The process hangs so that it cannot be killed, even with
> a kill -15. The rest of the system stays available, so
> it is possible to log on from another terminal and bring
> the thing cleanly down. Spinlock or uniterruptible sleep?
> A known issue with 2.4.27-SMP?
>
> Uniprocessor compile of the very same kernel works (but
> slower, of course, sigh!).
>
>
> Any light / pointers?
>

1. There is a size limitation with SILO. It is itself linked
to 0x280000 (2.5 Mbytes), which puts a roof on the
code loaded (oh, why!).

2. There is a bug in SMP kernel. It is patched in the
Debian release of 2.4.27 sources. The work as should:
12 hours of crashme did not succeed to crash it.

Tauno Voipio
tauno voipio (at) iki fi

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

SEO

Sparcstation 20 problems with 2.4.27 kernel