/proc/ppc_htab

/proc/ppc_htab

[Geert Uytterhoeven]

A few days ago I saw `mmu_context_overflow' in my kernel messages.
The machine locked up on `cat /proc/ppc_htab'. Fortunately it autorebooted
after 180 seconds and came back to life again.

Since I don't have direct access to my PPC box now (I'm recovering from knee
surgery) and rely on it for receiving email, I'm not very eager to do some
more experiments with /proc/ppc_htab.

Note that I have no idea what mmu_context_overflow means nor what ppc_htab
really does. Are these two related?

For what it's worth, I'm running the bk tree at the 2.3.99-pre3 level.


Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- Linux/{m68k~Amiga,PPC~CHRP} -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: /proc/ppc_htab

[Dan Malek]

Geert Uytterhoeven wrote:

> For what it's worth, I'm running the bk tree at the 2.3.99-pre3 level.

Wow.  That's more than I can do, even with two good knees :-).


	-- Dan

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: /proc/ppc_htab and 2.3 oddities

[Ani Joshi]

On Mon, 17 Apr 2000, Geert Uytterhoeven wrote:

>
> A few days ago I saw `mmu_context_overflow' in my kernel messages.
> The machine locked up on `cat /proc/ppc_htab'. Fortunately it autorebooted
> after 180 seconds and came back to life again.

I also saw this a few weeks ago, and was meaning to fix it but I was
unable to reproduce it.

>
> Since I don't have direct access to my PPC box now (I'm recovering from knee
> surgery) and rely on it for receiving email, I'm not very eager to do some
> more experiments with /proc/ppc_htab.

Ouch, hope you feel better soon.

>
> Note that I have no idea what mmu_context_overflow means nor what ppc_htab
> really does. Are these two related?

/proc/ppc_htab is the proc entry for the hash table.  it shows some
current stats on the hash table and allows some manipulation.

an mmu_context_overflow occurs when the counter, well...  overflows.

I'm sure you don't feel like crashing a remote box again, but if you or
anyone else could cat /proc/ppc_htab and tell if panics and perhaps post
the OOPS.

>
> For what it's worth, I'm running the bk tree at the 2.3.99-pre3 level.
>

Speaking of 2.3.99, i think the shared mem or something is terrible
broken.  I am currently in the process of writing an XFree86 driver for
Savage chipset cards, (for those unfamiliar with xfree 4.0, you just
build the driver as a module, ie just an object and place it in a set path
for the server to load), and each time I would build a new driver to test
and copy/replace over the old one and then run the server, it would not
replace the old driver correctly.

Just to prove i wasn't going insane, I used an old build, found out it
worked.  Then made a new build with a ton of ErrorF's (xfree printf's) all
over the place, and replaced the driver.  To my amazement, it was still
using the old build, even though the new one was sitting right over the
old one.  This was all with 2.3.99pre3 with all the /var/shm done.  When I
switched back to 2.2.4 kernel, it worked fine as normal (copy'ing over DID
work as expected).

Has anyone else seen this?

>
> Gr{oetje,eeting}s,
>
> 						Geert


Ani


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: /proc/ppc_htab

[Iain Sandoe]

>> A few days ago I saw `mmu_context_overflow' in my kernel messages.
>> The machine locked up on `cat /proc/ppc_htab'. Fortunately it autorebooted
>> after 180 seconds and came back to life again.
>
> I also saw this a few weeks ago, and was meaning to fix it but I was
> unable to reproduce it.

I have not seen the mmu_context_overflow - but the /proc/ppc_htab has been
causing a lock-up since at least 2.3.51 (sorry it slipped my mind).

I'm using Paul's rsync (latest I've tried is 2.3.99-pre3).

> I'm sure you don't feel like crashing a remote box again, but if you or
> anyone else could cat /proc/ppc_htab and tell if panics and perhaps post
> the OOPS.

Unfortunately there was no panic, OOPS or anything - and - unlike Geert's my
system never came back - just locked up completely - requiring a three
finger salute to get anywhere.  G3/minitower (beige).

Not as much help as I would like - apart from the fact that it was
completely repeatable - I'm out of circulation for 5 weeks now - paid work
;-) but if there's anything I can supply by tomorrow night that will throw
more light on it - let me know.

Iain.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/