Re: Kernel2.4.20 & PowerbookG3-Lombard awakening hiccup
Hello,
Sorry I've been delayed with answering.
At 11:50 Uhr +0100 20.01.2003, Benjamin Herrenschmidt wrote:
On Mon, 2003-01-20 at 05:04, Christian Jaeger wrote:
Hello
3 weeks ago I've compiled a stock kernel.org 2.4.20 kernel and
replaced the previously used 2.4.18 one. I took the 2.4.18 .config
and added a second wirelesslan driver (and maybe a few other
modules). About 5 weeks ago I upgraded my RAM to 256+512MB (from
which only 2*256 were recognized)
Did you try CONFIG_HIGHMEM ? What do dmesg says about BAT allocation ?
(very first few lines of dmesg).
Oh, I don't have the 512MB RAM anymore. Based of the claims of the
technician that the lombard doesn't support more than 512MB and the
fact that there wasn't a firmware upgrade to be found I believed that
it's simply not possible - is this wrong? (Maybe I should get the
512MB dimm again to check? It would probably cost me money though.)
I didn't (and don't) have CONFIG_HIGHMEM set. I didn't know that this
would do any effect below 960MB. (And MacOS9 had the problem as well,
so being a hardware problem sounded plausible.)
> Since about 3 weeks I've been
experiencing frequent power cutoffs after waking up the pb from sleep
- after opening the lid and hitting a key, it would switch on the
hard disk and the display, then about 1-2 seconds later it would just
switch completely off. Pressing the power-on button would boot it,
but the function key setting has each time been resetted (which means
that the PRAM has been deleted, right? BTW I'm still looking for a
version of fnset that really works on this lombard, unlike some
claims from others, looks like there are multiple versions of
lombards, I have a 333Mhz CPU). Independantly from wakeup, the
poweroff did only happen once, and this was when MacOS9 was running
natively - well I think there is a loose connection somewhere, I've
had sudden poweroffs quite a few times already, maybe once per month
or two. But the poweroffs right after wakeup looked like a new
phenomena and it happened in about 30% of all wakeups.
I didn't quite understand your statement.
Ok I try to summarize.
Phenomenon 1, which exists since I've bought this powerbook
(second-hand) 1 and 1/2 years ago: about every (1-)2 months it just
switches off (poweroff, or worse, PRAM erase IIRC) during normal
usage without any apparent reason, the accu with enough power,
sometimes when connected to AC, but most of the times when sitting in
train or bus (mechanical movements?). This also happened when
running MacOS9. At some point in time I moved the battery from the
left to the right bay (and removed the cdrom). This seemed to cure it
a bit but not completely. Once (beginning of october 2002) I've also
had an incident where it smelled like something had been burning
inside, (I think immediately) after having closed the lid and put it
into my bag. It was switched off thereafter, but (removing the
batteries and) hitting poweron was enough to make it work again. It
took a few days until I didn't smell it anymore. It didn't have a
burning incident again since (but the sporadic non-wakeup-related
poweroffs seem to persist as happened right before the eyes of the
mac shop technician while in MacOS9, just when he put the keyboard
back into place that he opened to take a look beneath).
Phenomenon 2, which happened between about 30 Dec 2002 (upgrade to
2.4.20) and 16 Jan 2003 (downgrade to 2.4.18): poweroff (or worse,
always PRAM erase) immediately after about 30% of all wakeups from
sleep. The 512MB DIMM I did have built in between about 15 Dec 2002
and 16 Jan 2003 *might* have to do something with it but it really
looks like the kernel version makes the major difference.
Is this specific to waking up
in linux or does this happen in MacOS as well ?
Hmm. Did I test wakeup in MacOS? I'm not sure anymore.
Oh, you RAM isn't recongnized by MacOS either ?
Yep, MacOS9 saw 512MB in total (256 + 1/2*512) just like linux.
Have I already said?: the interesting thing was that the day I bought
the RAM I told the seller to install it immediately to see if it
works. He moved the 256 Dimm from the lower to the upper slot
(&removed old 64MB dimm), and put the new 512MB dimm into the lower
slot. Upon booting MacOS9 we saw that only 512MB total were
recognized, so he took out the 512 dimm again and replaced it with a
second 256 dimm. Upon boot only 384 MB total RAM was available. So he
shrugged and said he would better put the 512MB dimm in so I have at
least 512MB useable RAM in total. Meanwhile a colleague of him
arrived and said that the firmware needs to be upgraded to
make >512MB possible. I then didn't find an upgrade for the lombard
and thus a few weeks later returned to the shop, where the technician
then said that no upgrade exists and so I changed the 512 to another
256 dimm which then (surprisingly) correctly worked so I now still
have 512MB of recognized RAM (2*256).
BTW we always tested in MacOS9, and most of the time in linux
(2.4.18), never in MacOSX.
Looks like some BAD dimm
indeed. Now, regarding the poweroffs, that's strange. Those are
typically triggered by the PMU timing out, so that can eventually be a
kernel crash at a bad moment, but still, that is weird.
So my conclusion: I *think* that there must be some software problem
with wakeup with 2.4.20. (BTW if I remember correctly it was with
2.4.20 that I've seen the wakeup cycle take really long (~20 seconds)
until X is back and the beep is played. This only happened 2 times.)
The beep is some random crap due to power management of the sound chip.
(Isn't it played by the pmud? With ~potato it didn't beep.)
20 seconds does indeed sounds too long though. Is your kernel compiled
with CONFIG_XMON ?
No.
Do you see it entering the xmon debugger sometimes on
wakeup ?
No
Also make sure you have apm_emu enabled
Yes it is.
and /dev/apm_bios
created so that XFree is properly power managed.
It's a link to /dev/misc/apm_bios. (I'm using devfs(d).)
> Of course it might also be a *combination* of flaky hardware and this
kernel. If that's the case, I'll have to take precautions (buy
> another laptop soon enough).
It's definitely largely caused by the kernel I've been using. I have
not had a single wakeup poweroff since downgrading to 2.4.18.
Christian.
Reply to: